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
191 -Wsign-promo -Wsynth}
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-follow-jumps -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
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}
405 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
406 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
407 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
408 -mstack-align -mdata-align -mconst-align @gol
409 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
410 -melf -maout -melinux -mlinux -sim -sim2 @gol
411 -mmul-bug-workaround -mno-mul-bug-workaround}
413 @emph{Darwin Options}
414 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
415 -arch_only -bind_at_load -bundle -bundle_loader @gol
416 -client_name -compatibility_version -current_version @gol
418 -dependency-file -dylib_file -dylinker_install_name @gol
419 -dynamic -dynamiclib -exported_symbols_list @gol
420 -filelist -flat_namespace -force_cpusubtype_ALL @gol
421 -force_flat_namespace -headerpad_max_install_names @gol
422 -image_base -init -install_name -keep_private_externs @gol
423 -multi_module -multiply_defined -multiply_defined_unused @gol
424 -noall_load -no_dead_strip_inits_and_terms @gol
425 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
426 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
427 -private_bundle -read_only_relocs -sectalign @gol
428 -sectobjectsymbols -whyload -seg1addr @gol
429 -sectcreate -sectobjectsymbols -sectorder @gol
430 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
431 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
432 -segprot -segs_read_only_addr -segs_read_write_addr @gol
433 -single_module -static -sub_library -sub_umbrella @gol
434 -twolevel_namespace -umbrella -undefined @gol
435 -unexported_symbols_list -weak_reference_mismatches @gol
436 -whatsloaded -F -gused -gfull -mone-byte-bool}
438 @emph{DEC Alpha Options}
439 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
440 -mieee -mieee-with-inexact -mieee-conformant @gol
441 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
442 -mtrap-precision=@var{mode} -mbuild-constants @gol
443 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
444 -mbwx -mmax -mfix -mcix @gol
445 -mfloat-vax -mfloat-ieee @gol
446 -mexplicit-relocs -msmall-data -mlarge-data @gol
447 -msmall-text -mlarge-text @gol
448 -mmemory-latency=@var{time}}
450 @emph{DEC Alpha/VMS Options}
451 @gccoptlist{-mvms-return-codes}
454 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
455 -mhard-float -msoft-float @gol
456 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
457 -mdouble -mno-double @gol
458 -mmedia -mno-media -mmuladd -mno-muladd @gol
459 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
460 -mlinked-fp -mlong-calls -malign-labels @gol
461 -mlibrary-pic -macc-4 -macc-8 @gol
462 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
463 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
464 -mvliw-branch -mno-vliw-branch @gol
465 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
466 -mno-nested-cond-exec -mtomcat-stats @gol
470 @emph{H8/300 Options}
471 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
474 @gccoptlist{-march=@var{architecture-type} @gol
475 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
476 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
477 -mfixed-range=@var{register-range} @gol
478 -mjump-in-delay -mlinker-opt -mlong-calls @gol
479 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
480 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
481 -mno-jump-in-delay -mno-long-load-store @gol
482 -mno-portable-runtime -mno-soft-float @gol
483 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
484 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
485 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
486 -munix=@var{unix-std} -nolibdld -static -threads}
488 @emph{i386 and x86-64 Options}
489 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
490 -mfpmath=@var{unit} @gol
491 -masm=@var{dialect} -mno-fancy-math-387 @gol
492 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
493 -mno-wide-multiply -mrtd -malign-double @gol
494 -mpreferred-stack-boundary=@var{num} @gol
495 -mmmx -msse -msse2 -msse3 -m3dnow @gol
496 -mthreads -mno-align-stringops -minline-all-stringops @gol
497 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
498 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
499 -mno-red-zone -mno-tls-direct-seg-refs @gol
500 -mcmodel=@var{code-model} @gol
504 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
505 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
506 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
507 -minline-float-divide-max-throughput @gol
508 -minline-int-divide-min-latency @gol
509 -minline-int-divide-max-throughput @gol
510 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
511 -mno-dwarf2-asm -mearly-stop-bits @gol
512 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
513 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64}
515 @emph{M32R/D Options}
516 @gccoptlist{-m32r2 -m32rx -m32r @gol
518 -malign-loops -mno-align-loops @gol
519 -missue-rate=@var{number} @gol
520 -mbranch-cost=@var{number} @gol
521 -mmodel=@var{code-size-model-type} @gol
522 -msdata=@var{sdata-type} @gol
523 -mno-flush-func -mflush-func=@var{name} @gol
524 -mno-flush-trap -mflush-trap=@var{number} @gol
527 @emph{M680x0 Options}
528 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
529 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
530 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
531 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
532 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
534 @emph{M68hc1x Options}
535 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
536 -mauto-incdec -minmax -mlong-calls -mshort @gol
537 -msoft-reg-count=@var{count}}
540 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
541 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
542 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
543 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
544 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
547 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
548 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
549 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
550 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
551 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
552 -mpaired-single -mips3d @gol
553 -mint64 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
554 -G@var{num} -membedded-data -mno-embedded-data @gol
555 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
556 -msplit-addresses -mno-split-addresses @gol
557 -mexplicit-relocs -mno-explicit-relocs @gol
558 -mcheck-zero-division -mno-check-zero-division @gol
559 -mdivide-traps -mdivide-breaks @gol
560 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
561 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
562 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
563 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
564 -mfix-sb1 -mno-fix-sb1 @gol
565 -mflush-func=@var{func} -mno-flush-func @gol
566 -mbranch-likely -mno-branch-likely @gol
567 -mfp-exceptions -mno-fp-exceptions @gol
568 -mvr4130-align -mno-vr4130-align}
571 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
572 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
573 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
574 -mno-base-addresses -msingle-exit -mno-single-exit}
576 @emph{MN10300 Options}
577 @gccoptlist{-mmult-bug -mno-mult-bug @gol
578 -mam33 -mno-am33 @gol
579 -mam33-2 -mno-am33-2 @gol
583 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
584 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
585 -mregparam -mnoregparam -msb -mnosb @gol
586 -mbitfield -mnobitfield -mhimem -mnohimem}
588 @emph{PDP-11 Options}
589 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
590 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
591 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
592 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
593 -mbranch-expensive -mbranch-cheap @gol
594 -msplit -mno-split -munix-asm -mdec-asm}
596 @emph{PowerPC Options}
597 See RS/6000 and PowerPC Options.
599 @emph{RS/6000 and PowerPC Options}
600 @gccoptlist{-mcpu=@var{cpu-type} @gol
601 -mtune=@var{cpu-type} @gol
602 -mpower -mno-power -mpower2 -mno-power2 @gol
603 -mpowerpc -mpowerpc64 -mno-powerpc @gol
604 -maltivec -mno-altivec @gol
605 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
606 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
607 -mnew-mnemonics -mold-mnemonics @gol
608 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
609 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
610 -malign-power -malign-natural @gol
611 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
612 -mstring -mno-string -mupdate -mno-update @gol
613 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
614 -mstrict-align -mno-strict-align -mrelocatable @gol
615 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
616 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
617 -mdynamic-no-pic @gol
618 -mprioritize-restricted-insns=@var{priority} @gol
619 -msched-costly-dep=@var{dependence_type} @gol
620 -minsert-sched-nops=@var{scheme} @gol
621 -mcall-sysv -mcall-netbsd @gol
622 -maix-struct-return -msvr4-struct-return @gol
623 -mabi=altivec -mabi=no-altivec @gol
624 -mabi=spe -mabi=no-spe @gol
625 -misel=yes -misel=no @gol
626 -mspe=yes -mspe=no @gol
627 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
628 -mprototype -mno-prototype @gol
629 -msim -mmvme -mads -myellowknife -memb -msdata @gol
630 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
632 @emph{S/390 and zSeries Options}
633 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
634 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
635 -mpacked-stack -mno-packed-stack @gol
636 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
637 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
638 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
639 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
642 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
643 -m4-nofpu -m4-single-only -m4-single -m4 @gol
644 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
645 -m5-64media -m5-64media-nofpu @gol
646 -m5-32media -m5-32media-nofpu @gol
647 -m5-compact -m5-compact-nofpu @gol
648 -mb -ml -mdalign -mrelax @gol
649 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
650 -mieee -misize -mpadstruct -mspace @gol
651 -mprefergot -musermode}
654 @gccoptlist{-mcpu=@var{cpu-type} @gol
655 -mtune=@var{cpu-type} @gol
656 -mcmodel=@var{code-model} @gol
657 -m32 -m64 -mapp-regs -mno-app-regs @gol
658 -mfaster-structs -mno-faster-structs @gol
659 -mfpu -mno-fpu -mhard-float -msoft-float @gol
660 -mhard-quad-float -msoft-quad-float @gol
661 -mimpure-text -mno-impure-text -mlittle-endian @gol
662 -mstack-bias -mno-stack-bias @gol
663 -munaligned-doubles -mno-unaligned-doubles @gol
664 -mv8plus -mno-v8plus -mvis -mno-vis
667 @emph{System V Options}
668 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
670 @emph{TMS320C3x/C4x Options}
671 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
672 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
673 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
674 -mparallel-insns -mparallel-mpy -mpreserve-float}
677 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
678 -mprolog-function -mno-prolog-function -mspace @gol
679 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
680 -mapp-regs -mno-app-regs @gol
681 -mdisable-callt -mno-disable-callt @gol
687 @gccoptlist{-mg -mgnu -munix}
689 @emph{x86-64 Options}
690 See i386 and x86-64 Options.
692 @emph{Xstormy16 Options}
695 @emph{Xtensa Options}
696 @gccoptlist{-mconst16 -mno-const16 @gol
697 -mfused-madd -mno-fused-madd @gol
698 -mtext-section-literals -mno-text-section-literals @gol
699 -mtarget-align -mno-target-align @gol
700 -mlongcalls -mno-longcalls}
702 @emph{zSeries Options}
703 See S/390 and zSeries Options.
705 @item Code Generation Options
706 @xref{Code Gen Options,,Options for Code Generation Conventions}.
707 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
708 -ffixed-@var{reg} -fexceptions @gol
709 -fnon-call-exceptions -funwind-tables @gol
710 -fasynchronous-unwind-tables @gol
711 -finhibit-size-directive -finstrument-functions @gol
712 -fno-common -fno-ident @gol
713 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
714 -freg-struct-return -fshared-data -fshort-enums @gol
715 -fshort-double -fshort-wchar @gol
716 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
717 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
718 -fargument-alias -fargument-noalias @gol
719 -fargument-noalias-global -fleading-underscore @gol
720 -ftls-model=@var{model} @gol
721 -ftrapv -fwrapv -fbounds-check @gol
726 * Overall Options:: Controlling the kind of output:
727 an executable, object files, assembler files,
728 or preprocessed source.
729 * C Dialect Options:: Controlling the variant of C language compiled.
730 * C++ Dialect Options:: Variations on C++.
731 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
733 * Language Independent Options:: Controlling how diagnostics should be
735 * Warning Options:: How picky should the compiler be?
736 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
737 * Optimize Options:: How much optimization?
738 * Preprocessor Options:: Controlling header files and macro definitions.
739 Also, getting dependency information for Make.
740 * Assembler Options:: Passing options to the assembler.
741 * Link Options:: Specifying libraries and so on.
742 * Directory Options:: Where to find header files and libraries.
743 Where to find the compiler executable files.
744 * Spec Files:: How to pass switches to sub-processes.
745 * Target Options:: Running a cross-compiler, or an old version of GCC.
748 @node Overall Options
749 @section Options Controlling the Kind of Output
751 Compilation can involve up to four stages: preprocessing, compilation
752 proper, assembly and linking, always in that order. GCC is capable of
753 preprocessing and compiling several files either into several
754 assembler input files, or into one assembler input file; then each
755 assembler input file produces an object file, and linking combines all
756 the object files (those newly compiled, and those specified as input)
757 into an executable file.
759 @cindex file name suffix
760 For any given input file, the file name suffix determines what kind of
765 C source code which must be preprocessed.
768 C source code which should not be preprocessed.
771 C++ source code which should not be preprocessed.
774 Objective-C source code. Note that you must link with the @file{libobjc}
775 library to make an Objective-C program work.
778 Objective-C source code which should not be preprocessed.
782 Objective-C++ source code. Note that you must link with the @file{libobjc}
783 library to make an Objective-C++ program work. Note that @samp{.M} refers
784 to a literal capital M@.
787 Objective-C++ source code which should not be preprocessed.
790 C, C++, Objective-C or Objective-C++ header file to be turned into a
795 @itemx @var{file}.cxx
796 @itemx @var{file}.cpp
797 @itemx @var{file}.CPP
798 @itemx @var{file}.c++
800 C++ source code which must be preprocessed. Note that in @samp{.cxx},
801 the last two letters must both be literally @samp{x}. Likewise,
802 @samp{.C} refers to a literal capital C@.
806 C++ header file to be turned into a precompiled header.
809 @itemx @var{file}.for
810 @itemx @var{file}.FOR
811 Fortran source code which should not be preprocessed.
814 @itemx @var{file}.fpp
815 @itemx @var{file}.FPP
816 Fortran source code which must be preprocessed (with the traditional
820 Fortran source code which must be preprocessed with a RATFOR
821 preprocessor (not included with GCC)@.
824 @itemx @var{file}.f95
825 Fortran 90/95 source code which should not be preprocessed.
827 @c FIXME: Descriptions of Java file types.
834 Ada source code file which contains a library unit declaration (a
835 declaration of a package, subprogram, or generic, or a generic
836 instantiation), or a library unit renaming declaration (a package,
837 generic, or subprogram renaming declaration). Such files are also
840 @itemx @var{file}.adb
841 Ada source code file containing a library unit body (a subprogram or
842 package body). Such files are also called @dfn{bodies}.
844 @c GCC also knows about some suffixes for languages not yet included:
853 Assembler code which must be preprocessed.
856 An object file to be fed straight into linking.
857 Any file name with no recognized suffix is treated this way.
861 You can specify the input language explicitly with the @option{-x} option:
864 @item -x @var{language}
865 Specify explicitly the @var{language} for the following input files
866 (rather than letting the compiler choose a default based on the file
867 name suffix). This option applies to all following input files until
868 the next @option{-x} option. Possible values for @var{language} are:
870 c c-header c-cpp-output
871 c++ c++-header c++-cpp-output
872 objective-c objective-c-header objective-c-cpp-output
873 objective-c++ objective-c++-header objective-c++-cpp-output
874 assembler assembler-with-cpp
876 f77 f77-cpp-input ratfor
883 Turn off any specification of a language, so that subsequent files are
884 handled according to their file name suffixes (as they are if @option{-x}
885 has not been used at all).
887 @item -pass-exit-codes
888 @opindex pass-exit-codes
889 Normally the @command{gcc} program will exit with the code of 1 if any
890 phase of the compiler returns a non-success return code. If you specify
891 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
892 numerically highest error produced by any phase that returned an error
896 If you only want some of the stages of compilation, you can use
897 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
898 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
899 @command{gcc} is to stop. Note that some combinations (for example,
900 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
905 Compile or assemble the source files, but do not link. The linking
906 stage simply is not done. The ultimate output is in the form of an
907 object file for each source file.
909 By default, the object file name for a source file is made by replacing
910 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
912 Unrecognized input files, not requiring compilation or assembly, are
917 Stop after the stage of compilation proper; do not assemble. The output
918 is in the form of an assembler code file for each non-assembler input
921 By default, the assembler file name for a source file is made by
922 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
924 Input files that don't require compilation are ignored.
928 Stop after the preprocessing stage; do not run the compiler proper. The
929 output is in the form of preprocessed source code, which is sent to the
932 Input files which don't require preprocessing are ignored.
934 @cindex output file option
937 Place output in file @var{file}. This applies regardless to whatever
938 sort of output is being produced, whether it be an executable file,
939 an object file, an assembler file or preprocessed C code.
941 If @option{-o} is not specified, the default is to put an executable
942 file in @file{a.out}, the object file for
943 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
944 assembler file in @file{@var{source}.s}, a precompiled header file in
945 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
950 Print (on standard error output) the commands executed to run the stages
951 of compilation. Also print the version number of the compiler driver
952 program and of the preprocessor and the compiler proper.
956 Like @option{-v} except the commands are not executed and all command
957 arguments are quoted. This is useful for shell scripts to capture the
958 driver-generated command lines.
962 Use pipes rather than temporary files for communication between the
963 various stages of compilation. This fails to work on some systems where
964 the assembler is unable to read from a pipe; but the GNU assembler has
969 If you are compiling multiple source files, this option tells the driver
970 to pass all the source files to the compiler at once (for those
971 languages for which the compiler can handle this). This will allow
972 intermodule analysis (IMA) to be performed by the compiler. Currently the only
973 language for which this is supported is C@. If you pass source files for
974 multiple languages to the driver, using this option, the driver will invoke
975 the compiler(s) that support IMA once each, passing each compiler all the
976 source files appropriate for it. For those languages that do not support
977 IMA this option will be ignored, and the compiler will be invoked once for
978 each source file in that language. If you use this option in conjunction
979 with @option{-save-temps}, the compiler will generate multiple
981 (one for each source file), but only one (combined) @file{.o} or
986 Print (on the standard output) a description of the command line options
987 understood by @command{gcc}. If the @option{-v} option is also specified
988 then @option{--help} will also be passed on to the various processes
989 invoked by @command{gcc}, so that they can display the command line options
990 they accept. If the @option{-Wextra} option is also specified then command
991 line options which have no documentation associated with them will also
996 Print (on the standard output) a description of target specific command
997 line options for each tool.
1001 Display the version number and copyrights of the invoked GCC@.
1005 @section Compiling C++ Programs
1007 @cindex suffixes for C++ source
1008 @cindex C++ source file suffixes
1009 C++ source files conventionally use one of the suffixes @samp{.C},
1010 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1011 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1012 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1013 files with these names and compiles them as C++ programs even if you
1014 call the compiler the same way as for compiling C programs (usually
1015 with the name @command{gcc}).
1019 However, C++ programs often require class libraries as well as a
1020 compiler that understands the C++ language---and under some
1021 circumstances, you might want to compile programs or header files from
1022 standard input, or otherwise without a suffix that flags them as C++
1023 programs. You might also like to precompile a C header file with a
1024 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
1025 program that calls GCC with the default language set to C++, and
1026 automatically specifies linking against the C++ library. On many
1027 systems, @command{g++} is also installed with the name @command{c++}.
1029 @cindex invoking @command{g++}
1030 When you compile C++ programs, you may specify many of the same
1031 command-line options that you use for compiling programs in any
1032 language; or command-line options meaningful for C and related
1033 languages; or options that are meaningful only for C++ programs.
1034 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1035 explanations of options for languages related to C@.
1036 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1037 explanations of options that are meaningful only for C++ programs.
1039 @node C Dialect Options
1040 @section Options Controlling C Dialect
1041 @cindex dialect options
1042 @cindex language dialect options
1043 @cindex options, dialect
1045 The following options control the dialect of C (or languages derived
1046 from C, such as C++, Objective-C and Objective-C++) that the compiler
1050 @cindex ANSI support
1054 In C mode, support all ISO C90 programs. In C++ mode,
1055 remove GNU extensions that conflict with ISO C++.
1057 This turns off certain features of GCC that are incompatible with ISO
1058 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1059 such as the @code{asm} and @code{typeof} keywords, and
1060 predefined macros such as @code{unix} and @code{vax} that identify the
1061 type of system you are using. It also enables the undesirable and
1062 rarely used ISO trigraph feature. For the C compiler,
1063 it disables recognition of C++ style @samp{//} comments as well as
1064 the @code{inline} keyword.
1066 The alternate keywords @code{__asm__}, @code{__extension__},
1067 @code{__inline__} and @code{__typeof__} continue to work despite
1068 @option{-ansi}. You would not want to use them in an ISO C program, of
1069 course, but it is useful to put them in header files that might be included
1070 in compilations done with @option{-ansi}. Alternate predefined macros
1071 such as @code{__unix__} and @code{__vax__} are also available, with or
1072 without @option{-ansi}.
1074 The @option{-ansi} option does not cause non-ISO programs to be
1075 rejected gratuitously. For that, @option{-pedantic} is required in
1076 addition to @option{-ansi}. @xref{Warning Options}.
1078 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1079 option is used. Some header files may notice this macro and refrain
1080 from declaring certain functions or defining certain macros that the
1081 ISO standard doesn't call for; this is to avoid interfering with any
1082 programs that might use these names for other things.
1084 Functions which would normally be built in but do not have semantics
1085 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1086 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1087 built-in functions provided by GCC}, for details of the functions
1092 Determine the language standard. This option is currently only
1093 supported when compiling C or C++. A value for this option must be
1094 provided; possible values are
1099 ISO C90 (same as @option{-ansi}).
1101 @item iso9899:199409
1102 ISO C90 as modified in amendment 1.
1108 ISO C99. Note that this standard is not yet fully supported; see
1109 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1110 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1113 Default, ISO C90 plus GNU extensions (including some C99 features).
1117 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1118 this will become the default. The name @samp{gnu9x} is deprecated.
1121 The 1998 ISO C++ standard plus amendments.
1124 The same as @option{-std=c++98} plus GNU extensions. This is the
1125 default for C++ code.
1128 Even when this option is not specified, you can still use some of the
1129 features of newer standards in so far as they do not conflict with
1130 previous C standards. For example, you may use @code{__restrict__} even
1131 when @option{-std=c99} is not specified.
1133 The @option{-std} options specifying some version of ISO C have the same
1134 effects as @option{-ansi}, except that features that were not in ISO C90
1135 but are in the specified version (for example, @samp{//} comments and
1136 the @code{inline} keyword in ISO C99) are not disabled.
1138 @xref{Standards,,Language Standards Supported by GCC}, for details of
1139 these standard versions.
1141 @item -aux-info @var{filename}
1143 Output to the given filename prototyped declarations for all functions
1144 declared and/or defined in a translation unit, including those in header
1145 files. This option is silently ignored in any language other than C@.
1147 Besides declarations, the file indicates, in comments, the origin of
1148 each declaration (source file and line), whether the declaration was
1149 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1150 @samp{O} for old, respectively, in the first character after the line
1151 number and the colon), and whether it came from a declaration or a
1152 definition (@samp{C} or @samp{F}, respectively, in the following
1153 character). In the case of function definitions, a K&R-style list of
1154 arguments followed by their declarations is also provided, inside
1155 comments, after the declaration.
1159 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1160 keyword, so that code can use these words as identifiers. You can use
1161 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1162 instead. @option{-ansi} implies @option{-fno-asm}.
1164 In C++, this switch only affects the @code{typeof} keyword, since
1165 @code{asm} and @code{inline} are standard keywords. You may want to
1166 use the @option{-fno-gnu-keywords} flag instead, which has the same
1167 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1168 switch only affects the @code{asm} and @code{typeof} keywords, since
1169 @code{inline} is a standard keyword in ISO C99.
1172 @itemx -fno-builtin-@var{function}
1173 @opindex fno-builtin
1174 @cindex built-in functions
1175 Don't recognize built-in functions that do not begin with
1176 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1177 functions provided by GCC}, for details of the functions affected,
1178 including those which are not built-in functions when @option{-ansi} or
1179 @option{-std} options for strict ISO C conformance are used because they
1180 do not have an ISO standard meaning.
1182 GCC normally generates special code to handle certain built-in functions
1183 more efficiently; for instance, calls to @code{alloca} may become single
1184 instructions that adjust the stack directly, and calls to @code{memcpy}
1185 may become inline copy loops. The resulting code is often both smaller
1186 and faster, but since the function calls no longer appear as such, you
1187 cannot set a breakpoint on those calls, nor can you change the behavior
1188 of the functions by linking with a different library. In addition,
1189 when a function is recognized as a built-in function, GCC may use
1190 information about that function to warn about problems with calls to
1191 that function, or to generate more efficient code, even if the
1192 resulting code still contains calls to that function. For example,
1193 warnings are given with @option{-Wformat} for bad calls to
1194 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1195 known not to modify global memory.
1197 With the @option{-fno-builtin-@var{function}} option
1198 only the built-in function @var{function} is
1199 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1200 function is named this is not built-in in this version of GCC, this
1201 option is ignored. There is no corresponding
1202 @option{-fbuiltin-@var{function}} option; if you wish to enable
1203 built-in functions selectively when using @option{-fno-builtin} or
1204 @option{-ffreestanding}, you may define macros such as:
1207 #define abs(n) __builtin_abs ((n))
1208 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1213 @cindex hosted environment
1215 Assert that compilation takes place in a hosted environment. This implies
1216 @option{-fbuiltin}. A hosted environment is one in which the
1217 entire standard library is available, and in which @code{main} has a return
1218 type of @code{int}. Examples are nearly everything except a kernel.
1219 This is equivalent to @option{-fno-freestanding}.
1221 @item -ffreestanding
1222 @opindex ffreestanding
1223 @cindex hosted environment
1225 Assert that compilation takes place in a freestanding environment. This
1226 implies @option{-fno-builtin}. A freestanding environment
1227 is one in which the standard library may not exist, and program startup may
1228 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1229 This is equivalent to @option{-fno-hosted}.
1231 @xref{Standards,,Language Standards Supported by GCC}, for details of
1232 freestanding and hosted environments.
1234 @item -fms-extensions
1235 @opindex fms-extensions
1236 Accept some non-standard constructs used in Microsoft header files.
1238 Some cases of unnamed fields in structures and unions are only
1239 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1240 fields within structs/unions}, for details.
1244 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1245 options for strict ISO C conformance) implies @option{-trigraphs}.
1247 @item -no-integrated-cpp
1248 @opindex no-integrated-cpp
1249 Performs a compilation in two passes: preprocessing and compiling. This
1250 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1251 @option{-B} option. The user supplied compilation step can then add in
1252 an additional preprocessing step after normal preprocessing but before
1253 compiling. The default is to use the integrated cpp (internal cpp)
1255 The semantics of this option will change if "cc1", "cc1plus", and
1256 "cc1obj" are merged.
1258 @cindex traditional C language
1259 @cindex C language, traditional
1261 @itemx -traditional-cpp
1262 @opindex traditional-cpp
1263 @opindex traditional
1264 Formerly, these options caused GCC to attempt to emulate a pre-standard
1265 C compiler. They are now only supported with the @option{-E} switch.
1266 The preprocessor continues to support a pre-standard mode. See the GNU
1267 CPP manual for details.
1269 @item -fcond-mismatch
1270 @opindex fcond-mismatch
1271 Allow conditional expressions with mismatched types in the second and
1272 third arguments. The value of such an expression is void. This option
1273 is not supported for C++.
1275 @item -funsigned-char
1276 @opindex funsigned-char
1277 Let the type @code{char} be unsigned, like @code{unsigned char}.
1279 Each kind of machine has a default for what @code{char} should
1280 be. It is either like @code{unsigned char} by default or like
1281 @code{signed char} by default.
1283 Ideally, a portable program should always use @code{signed char} or
1284 @code{unsigned char} when it depends on the signedness of an object.
1285 But many programs have been written to use plain @code{char} and
1286 expect it to be signed, or expect it to be unsigned, depending on the
1287 machines they were written for. This option, and its inverse, let you
1288 make such a program work with the opposite default.
1290 The type @code{char} is always a distinct type from each of
1291 @code{signed char} or @code{unsigned char}, even though its behavior
1292 is always just like one of those two.
1295 @opindex fsigned-char
1296 Let the type @code{char} be signed, like @code{signed char}.
1298 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1299 the negative form of @option{-funsigned-char}. Likewise, the option
1300 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1302 @item -fsigned-bitfields
1303 @itemx -funsigned-bitfields
1304 @itemx -fno-signed-bitfields
1305 @itemx -fno-unsigned-bitfields
1306 @opindex fsigned-bitfields
1307 @opindex funsigned-bitfields
1308 @opindex fno-signed-bitfields
1309 @opindex fno-unsigned-bitfields
1310 These options control whether a bit-field is signed or unsigned, when the
1311 declaration does not use either @code{signed} or @code{unsigned}. By
1312 default, such a bit-field is signed, because this is consistent: the
1313 basic integer types such as @code{int} are signed types.
1316 @node C++ Dialect Options
1317 @section Options Controlling C++ Dialect
1319 @cindex compiler options, C++
1320 @cindex C++ options, command line
1321 @cindex options, C++
1322 This section describes the command-line options that are only meaningful
1323 for C++ programs; but you can also use most of the GNU compiler options
1324 regardless of what language your program is in. For example, you
1325 might compile a file @code{firstClass.C} like this:
1328 g++ -g -frepo -O -c firstClass.C
1332 In this example, only @option{-frepo} is an option meant
1333 only for C++ programs; you can use the other options with any
1334 language supported by GCC@.
1336 Here is a list of options that are @emph{only} for compiling C++ programs:
1340 @item -fabi-version=@var{n}
1341 @opindex fabi-version
1342 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1343 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1344 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1345 the version that conforms most closely to the C++ ABI specification.
1346 Therefore, the ABI obtained using version 0 will change as ABI bugs
1349 The default is version 2.
1351 @item -fno-access-control
1352 @opindex fno-access-control
1353 Turn off all access checking. This switch is mainly useful for working
1354 around bugs in the access control code.
1358 Check that the pointer returned by @code{operator new} is non-null
1359 before attempting to modify the storage allocated. This check is
1360 normally unnecessary because the C++ standard specifies that
1361 @code{operator new} will only return @code{0} if it is declared
1362 @samp{throw()}, in which case the compiler will always check the
1363 return value even without this option. In all other cases, when
1364 @code{operator new} has a non-empty exception specification, memory
1365 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1366 @samp{new (nothrow)}.
1368 @item -fconserve-space
1369 @opindex fconserve-space
1370 Put uninitialized or runtime-initialized global variables into the
1371 common segment, as C does. This saves space in the executable at the
1372 cost of not diagnosing duplicate definitions. If you compile with this
1373 flag and your program mysteriously crashes after @code{main()} has
1374 completed, you may have an object that is being destroyed twice because
1375 two definitions were merged.
1377 This option is no longer useful on most targets, now that support has
1378 been added for putting variables into BSS without making them common.
1380 @item -fno-const-strings
1381 @opindex fno-const-strings
1382 Give string constants type @code{char *} instead of type @code{const
1383 char *}. By default, G++ uses type @code{const char *} as required by
1384 the standard. Even if you use @option{-fno-const-strings}, you cannot
1385 actually modify the value of a string constant.
1387 This option might be removed in a future release of G++. For maximum
1388 portability, you should structure your code so that it works with
1389 string constants that have type @code{const char *}.
1391 @item -fno-elide-constructors
1392 @opindex fno-elide-constructors
1393 The C++ standard allows an implementation to omit creating a temporary
1394 which is only used to initialize another object of the same type.
1395 Specifying this option disables that optimization, and forces G++ to
1396 call the copy constructor in all cases.
1398 @item -fno-enforce-eh-specs
1399 @opindex fno-enforce-eh-specs
1400 Don't check for violation of exception specifications at runtime. This
1401 option violates the C++ standard, but may be useful for reducing code
1402 size in production builds, much like defining @samp{NDEBUG}. The compiler
1403 will still optimize based on the exception specifications.
1406 @itemx -fno-for-scope
1408 @opindex fno-for-scope
1409 If @option{-ffor-scope} is specified, the scope of variables declared in
1410 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1411 as specified by the C++ standard.
1412 If @option{-fno-for-scope} is specified, the scope of variables declared in
1413 a @i{for-init-statement} extends to the end of the enclosing scope,
1414 as was the case in old versions of G++, and other (traditional)
1415 implementations of C++.
1417 The default if neither flag is given to follow the standard,
1418 but to allow and give a warning for old-style code that would
1419 otherwise be invalid, or have different behavior.
1421 @item -fno-gnu-keywords
1422 @opindex fno-gnu-keywords
1423 Do not recognize @code{typeof} as a keyword, so that code can use this
1424 word as an identifier. You can use the keyword @code{__typeof__} instead.
1425 @option{-ansi} implies @option{-fno-gnu-keywords}.
1427 @item -fno-implicit-templates
1428 @opindex fno-implicit-templates
1429 Never emit code for non-inline templates which are instantiated
1430 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1431 @xref{Template Instantiation}, for more information.
1433 @item -fno-implicit-inline-templates
1434 @opindex fno-implicit-inline-templates
1435 Don't emit code for implicit instantiations of inline templates, either.
1436 The default is to handle inlines differently so that compiles with and
1437 without optimization will need the same set of explicit instantiations.
1439 @item -fno-implement-inlines
1440 @opindex fno-implement-inlines
1441 To save space, do not emit out-of-line copies of inline functions
1442 controlled by @samp{#pragma implementation}. This will cause linker
1443 errors if these functions are not inlined everywhere they are called.
1445 @item -fms-extensions
1446 @opindex fms-extensions
1447 Disable pedantic warnings about constructs used in MFC, such as implicit
1448 int and getting a pointer to member function via non-standard syntax.
1450 @item -fno-nonansi-builtins
1451 @opindex fno-nonansi-builtins
1452 Disable built-in declarations of functions that are not mandated by
1453 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1454 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1456 @item -fno-operator-names
1457 @opindex fno-operator-names
1458 Do not treat the operator name keywords @code{and}, @code{bitand},
1459 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1460 synonyms as keywords.
1462 @item -fno-optional-diags
1463 @opindex fno-optional-diags
1464 Disable diagnostics that the standard says a compiler does not need to
1465 issue. Currently, the only such diagnostic issued by G++ is the one for
1466 a name having multiple meanings within a class.
1469 @opindex fpermissive
1470 Downgrade some diagnostics about nonconformant code from errors to
1471 warnings. Thus, using @option{-fpermissive} will allow some
1472 nonconforming code to compile.
1476 Enable automatic template instantiation at link time. This option also
1477 implies @option{-fno-implicit-templates}. @xref{Template
1478 Instantiation}, for more information.
1482 Disable generation of information about every class with virtual
1483 functions for use by the C++ runtime type identification features
1484 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1485 of the language, you can save some space by using this flag. Note that
1486 exception handling uses the same information, but it will generate it as
1491 Emit statistics about front-end processing at the end of the compilation.
1492 This information is generally only useful to the G++ development team.
1494 @item -ftemplate-depth-@var{n}
1495 @opindex ftemplate-depth
1496 Set the maximum instantiation depth for template classes to @var{n}.
1497 A limit on the template instantiation depth is needed to detect
1498 endless recursions during template class instantiation. ANSI/ISO C++
1499 conforming programs must not rely on a maximum depth greater than 17.
1501 @item -fno-threadsafe-statics
1502 @opindex fno-threadsafe-statics
1503 Do not emit the extra code to use the routines specified in the C++
1504 ABI for thread-safe initialization of local statics. You can use this
1505 option to reduce code size slightly in code that doesn't need to be
1508 @item -fuse-cxa-atexit
1509 @opindex fuse-cxa-atexit
1510 Register destructors for objects with static storage duration with the
1511 @code{__cxa_atexit} function rather than the @code{atexit} function.
1512 This option is required for fully standards-compliant handling of static
1513 destructors, but will only work if your C library supports
1514 @code{__cxa_atexit}.
1516 @item -fvisibility-inlines-hidden
1517 @opindex fvisibility-inlines-hidden
1518 Causes all inlined methods to be marked with
1519 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1520 appear in the export table of a DSO and do not require a PLT indirection
1521 when used within the DSO@. Enabling this option can have a dramatic effect
1522 on load and link times of a DSO as it massively reduces the size of the
1523 dynamic export table when the library makes heavy use of templates. While
1524 it can cause bloating through duplication of code within each DSO where
1525 it is used, often the wastage is less than the considerable space occupied
1526 by a long symbol name in the export table which is typical when using
1527 templates and namespaces. For even more savings, combine with the
1528 @option{-fvisibility=hidden} switch.
1532 Do not use weak symbol support, even if it is provided by the linker.
1533 By default, G++ will use weak symbols if they are available. This
1534 option exists only for testing, and should not be used by end-users;
1535 it will result in inferior code and has no benefits. This option may
1536 be removed in a future release of G++.
1540 Do not search for header files in the standard directories specific to
1541 C++, but do still search the other standard directories. (This option
1542 is used when building the C++ library.)
1545 In addition, these optimization, warning, and code generation options
1546 have meanings only for C++ programs:
1549 @item -fno-default-inline
1550 @opindex fno-default-inline
1551 Do not assume @samp{inline} for functions defined inside a class scope.
1552 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1553 functions will have linkage like inline functions; they just won't be
1556 @item -Wabi @r{(C++ only)}
1558 Warn when G++ generates code that is probably not compatible with the
1559 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1560 all such cases, there are probably some cases that are not warned about,
1561 even though G++ is generating incompatible code. There may also be
1562 cases where warnings are emitted even though the code that is generated
1565 You should rewrite your code to avoid these warnings if you are
1566 concerned about the fact that code generated by G++ may not be binary
1567 compatible with code generated by other compilers.
1569 The known incompatibilities at this point include:
1574 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1575 pack data into the same byte as a base class. For example:
1578 struct A @{ virtual void f(); int f1 : 1; @};
1579 struct B : public A @{ int f2 : 1; @};
1583 In this case, G++ will place @code{B::f2} into the same byte
1584 as@code{A::f1}; other compilers will not. You can avoid this problem
1585 by explicitly padding @code{A} so that its size is a multiple of the
1586 byte size on your platform; that will cause G++ and other compilers to
1587 layout @code{B} identically.
1590 Incorrect handling of tail-padding for virtual bases. G++ does not use
1591 tail padding when laying out virtual bases. For example:
1594 struct A @{ virtual void f(); char c1; @};
1595 struct B @{ B(); char c2; @};
1596 struct C : public A, public virtual B @{@};
1600 In this case, G++ will not place @code{B} into the tail-padding for
1601 @code{A}; other compilers will. You can avoid this problem by
1602 explicitly padding @code{A} so that its size is a multiple of its
1603 alignment (ignoring virtual base classes); that will cause G++ and other
1604 compilers to layout @code{C} identically.
1607 Incorrect handling of bit-fields with declared widths greater than that
1608 of their underlying types, when the bit-fields appear in a union. For
1612 union U @{ int i : 4096; @};
1616 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1617 union too small by the number of bits in an @code{int}.
1620 Empty classes can be placed at incorrect offsets. For example:
1630 struct C : public B, public A @{@};
1634 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1635 it should be placed at offset zero. G++ mistakenly believes that the
1636 @code{A} data member of @code{B} is already at offset zero.
1639 Names of template functions whose types involve @code{typename} or
1640 template template parameters can be mangled incorrectly.
1643 template <typename Q>
1644 void f(typename Q::X) @{@}
1646 template <template <typename> class Q>
1647 void f(typename Q<int>::X) @{@}
1651 Instantiations of these templates may be mangled incorrectly.
1655 @item -Wctor-dtor-privacy @r{(C++ only)}
1656 @opindex Wctor-dtor-privacy
1657 Warn when a class seems unusable because all the constructors or
1658 destructors in that class are private, and it has neither friends nor
1659 public static member functions.
1661 @item -Wnon-virtual-dtor @r{(C++ only)}
1662 @opindex Wnon-virtual-dtor
1663 Warn when a class appears to be polymorphic, thereby requiring a virtual
1664 destructor, yet it declares a non-virtual one.
1665 This warning is enabled by @option{-Wall}.
1667 @item -Wreorder @r{(C++ only)}
1669 @cindex reordering, warning
1670 @cindex warning for reordering of member initializers
1671 Warn when the order of member initializers given in the code does not
1672 match the order in which they must be executed. For instance:
1678 A(): j (0), i (1) @{ @}
1682 The compiler will rearrange the member initializers for @samp{i}
1683 and @samp{j} to match the declaration order of the members, emitting
1684 a warning to that effect. This warning is enabled by @option{-Wall}.
1687 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1690 @item -Weffc++ @r{(C++ only)}
1692 Warn about violations of the following style guidelines from Scott Meyers'
1693 @cite{Effective C++} book:
1697 Item 11: Define a copy constructor and an assignment operator for classes
1698 with dynamically allocated memory.
1701 Item 12: Prefer initialization to assignment in constructors.
1704 Item 14: Make destructors virtual in base classes.
1707 Item 15: Have @code{operator=} return a reference to @code{*this}.
1710 Item 23: Don't try to return a reference when you must return an object.
1714 Also warn about violations of the following style guidelines from
1715 Scott Meyers' @cite{More Effective C++} book:
1719 Item 6: Distinguish between prefix and postfix forms of increment and
1720 decrement operators.
1723 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1727 When selecting this option, be aware that the standard library
1728 headers do not obey all of these guidelines; use @samp{grep -v}
1729 to filter out those warnings.
1731 @item -Wno-deprecated @r{(C++ only)}
1732 @opindex Wno-deprecated
1733 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1735 @item -Wno-non-template-friend @r{(C++ only)}
1736 @opindex Wno-non-template-friend
1737 Disable warnings when non-templatized friend functions are declared
1738 within a template. Since the advent of explicit template specification
1739 support in G++, if the name of the friend is an unqualified-id (i.e.,
1740 @samp{friend foo(int)}), the C++ language specification demands that the
1741 friend declare or define an ordinary, nontemplate function. (Section
1742 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1743 could be interpreted as a particular specialization of a templatized
1744 function. Because this non-conforming behavior is no longer the default
1745 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1746 check existing code for potential trouble spots and is on by default.
1747 This new compiler behavior can be turned off with
1748 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1749 but disables the helpful warning.
1751 @item -Wold-style-cast @r{(C++ only)}
1752 @opindex Wold-style-cast
1753 Warn if an old-style (C-style) cast to a non-void type is used within
1754 a C++ program. The new-style casts (@samp{static_cast},
1755 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1756 unintended effects and much easier to search for.
1758 @item -Woverloaded-virtual @r{(C++ only)}
1759 @opindex Woverloaded-virtual
1760 @cindex overloaded virtual fn, warning
1761 @cindex warning for overloaded virtual fn
1762 Warn when a function declaration hides virtual functions from a
1763 base class. For example, in:
1770 struct B: public A @{
1775 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1783 will fail to compile.
1785 @item -Wno-pmf-conversions @r{(C++ only)}
1786 @opindex Wno-pmf-conversions
1787 Disable the diagnostic for converting a bound pointer to member function
1790 @item -Wsign-promo @r{(C++ only)}
1791 @opindex Wsign-promo
1792 Warn when overload resolution chooses a promotion from unsigned or
1793 enumerated type to a signed type, over a conversion to an unsigned type of
1794 the same size. Previous versions of G++ would try to preserve
1795 unsignedness, but the standard mandates the current behavior.
1797 @item -Wsynth @r{(C++ only)}
1799 @cindex warning for synthesized methods
1800 @cindex synthesized methods, warning
1801 Warn when G++'s synthesis behavior does not match that of cfront. For
1807 A& operator = (int);
1817 In this example, G++ will synthesize a default @samp{A& operator =
1818 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1821 @node Objective-C and Objective-C++ Dialect Options
1822 @section Options Controlling Objective-C and Objective-C++ Dialects
1824 @cindex compiler options, Objective-C and Objective-C++
1825 @cindex Objective-C and Objective-C++ options, command line
1826 @cindex options, Objective-C and Objective-C++
1827 (NOTE: This manual does not describe the Objective-C and Objective-C++
1828 languages themselves. See @xref{Standards,,Language Standards
1829 Supported by GCC}, for references.)
1831 This section describes the command-line options that are only meaningful
1832 for Objective-C and Objective-C++ programs, but you can also use most of
1833 the language-independent GNU compiler options.
1834 For example, you might compile a file @code{some_class.m} like this:
1837 gcc -g -fgnu-runtime -O -c some_class.m
1841 In this example, @option{-fgnu-runtime} is an option meant only for
1842 Objective-C and Objective-C++ programs; you can use the other options with
1843 any language supported by GCC@.
1845 Note that since Objective-C is an extension of the C language, Objective-C
1846 compilations may also use options specific to the C front-end (e.g.,
1847 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1848 C++-specific options (e.g., @option{-Wabi}).
1850 Here is a list of options that are @emph{only} for compiling Objective-C
1851 and Objective-C++ programs:
1854 @item -fconstant-string-class=@var{class-name}
1855 @opindex fconstant-string-class
1856 Use @var{class-name} as the name of the class to instantiate for each
1857 literal string specified with the syntax @code{@@"@dots{}"}. The default
1858 class name is @code{NXConstantString} if the GNU runtime is being used, and
1859 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1860 @option{-fconstant-cfstrings} option, if also present, will override the
1861 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1862 to be laid out as constant CoreFoundation strings.
1865 @opindex fgnu-runtime
1866 Generate object code compatible with the standard GNU Objective-C
1867 runtime. This is the default for most types of systems.
1869 @item -fnext-runtime
1870 @opindex fnext-runtime
1871 Generate output compatible with the NeXT runtime. This is the default
1872 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1873 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1876 @item -fno-nil-receivers
1877 @opindex fno-nil-receivers
1878 Assume that all Objective-C message dispatches (e.g.,
1879 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1880 is not @code{nil}. This allows for more efficient entry points in the runtime
1881 to be used. Currently, this option is only available in conjunction with
1882 the NeXT runtime on Mac OS X 10.3 and later.
1884 @item -fobjc-exceptions
1885 @opindex fobjc-exceptions
1886 Enable syntactic support for structured exception handling in Objective-C,
1887 similar to what is offered by C++ and Java. Currently, this option is only
1888 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1896 @@catch (AnObjCClass *exc) @{
1903 @@catch (AnotherClass *exc) @{
1906 @@catch (id allOthers) @{
1916 The @code{@@throw} statement may appear anywhere in an Objective-C or
1917 Objective-C++ program; when used inside of a @code{@@catch} block, the
1918 @code{@@throw} may appear without an argument (as shown above), in which case
1919 the object caught by the @code{@@catch} will be rethrown.
1921 Note that only (pointers to) Objective-C objects may be thrown and
1922 caught using this scheme. When an object is thrown, it will be caught
1923 by the nearest @code{@@catch} clause capable of handling objects of that type,
1924 analogously to how @code{catch} blocks work in C++ and Java. A
1925 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1926 any and all Objective-C exceptions not caught by previous @code{@@catch}
1929 The @code{@@finally} clause, if present, will be executed upon exit from the
1930 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1931 regardless of whether any exceptions are thrown, caught or rethrown
1932 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1933 of the @code{finally} clause in Java.
1935 There are several caveats to using the new exception mechanism:
1939 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
1940 idioms provided by the @code{NSException} class, the new
1941 exceptions can only be used on Mac OS X 10.3 (Panther) and later
1942 systems, due to additional functionality needed in the (NeXT) Objective-C
1946 As mentioned above, the new exceptions do not support handling
1947 types other than Objective-C objects. Furthermore, when used from
1948 Objective-C++, the Objective-C exception model does not interoperate with C++
1949 exceptions at this time. This means you cannot @code{@@throw} an exception
1950 from Objective-C and @code{catch} it in C++, or vice versa
1951 (i.e., @code{throw @dots{} @@catch}).
1954 The @option{-fobjc-exceptions} switch also enables the use of synchronization
1955 blocks for thread-safe execution:
1958 @@synchronized (ObjCClass *guard) @{
1963 Upon entering the @code{@@synchronized} block, a thread of execution shall
1964 first check whether a lock has been placed on the corresponding @code{guard}
1965 object by another thread. If it has, the current thread shall wait until
1966 the other thread relinquishes its lock. Once @code{guard} becomes available,
1967 the current thread will place its own lock on it, execute the code contained in
1968 the @code{@@synchronized} block, and finally relinquish the lock (thereby
1969 making @code{guard} available to other threads).
1971 Unlike Java, Objective-C does not allow for entire methods to be marked
1972 @code{@@synchronized}. Note that throwing exceptions out of
1973 @code{@@synchronized} blocks is allowed, and will cause the guarding object
1974 to be unlocked properly.
1976 @item -freplace-objc-classes
1977 @opindex freplace-objc-classes
1978 Emit a special marker instructing @command{ld(1)} not to statically link in
1979 the resulting object file, and allow @command{dyld(1)} to load it in at
1980 run time instead. This is used in conjunction with the Fix-and-Continue
1981 debugging mode, where the object file in question may be recompiled and
1982 dynamically reloaded in the course of program execution, without the need
1983 to restart the program itself. Currently, Fix-and-Continue functionality
1984 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
1989 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
1990 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
1991 compile time) with static class references that get initialized at load time,
1992 which improves run-time performance. Specifying the @option{-fzero-link} flag
1993 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
1994 to be retained. This is useful in Zero-Link debugging mode, since it allows
1995 for individual class implementations to be modified during program execution.
1999 Dump interface declarations for all classes seen in the source file to a
2000 file named @file{@var{sourcename}.decl}.
2003 @opindex Wno-protocol
2004 If a class is declared to implement a protocol, a warning is issued for
2005 every method in the protocol that is not implemented by the class. The
2006 default behavior is to issue a warning for every method not explicitly
2007 implemented in the class, even if a method implementation is inherited
2008 from the superclass. If you use the @option{-Wno-protocol} option, then
2009 methods inherited from the superclass are considered to be implemented,
2010 and no warning is issued for them.
2014 Warn if multiple methods of different types for the same selector are
2015 found during compilation. The check is performed on the list of methods
2016 in the final stage of compilation. Additionally, a check is performed
2017 for each selector appearing in a @code{@@selector(@dots{})}
2018 expression, and a corresponding method for that selector has been found
2019 during compilation. Because these checks scan the method table only at
2020 the end of compilation, these warnings are not produced if the final
2021 stage of compilation is not reached, for example because an error is
2022 found during compilation, or because the @option{-fsyntax-only} option is
2025 @item -Wundeclared-selector
2026 @opindex Wundeclared-selector
2027 Warn if a @code{@@selector(@dots{})} expression referring to an
2028 undeclared selector is found. A selector is considered undeclared if no
2029 method with that name has been declared before the
2030 @code{@@selector(@dots{})} expression, either explicitly in an
2031 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2032 an @code{@@implementation} section. This option always performs its
2033 checks as soon as a @code{@@selector(@dots{})} expression is found,
2034 while @option{-Wselector} only performs its checks in the final stage of
2035 compilation. This also enforces the coding style convention
2036 that methods and selectors must be declared before being used.
2038 @item -print-objc-runtime-info
2039 @opindex print-objc-runtime-info
2040 Generate C header describing the largest structure that is passed by
2045 @node Language Independent Options
2046 @section Options to Control Diagnostic Messages Formatting
2047 @cindex options to control diagnostics formatting
2048 @cindex diagnostic messages
2049 @cindex message formatting
2051 Traditionally, diagnostic messages have been formatted irrespective of
2052 the output device's aspect (e.g.@: its width, @dots{}). The options described
2053 below can be used to control the diagnostic messages formatting
2054 algorithm, e.g.@: how many characters per line, how often source location
2055 information should be reported. Right now, only the C++ front end can
2056 honor these options. However it is expected, in the near future, that
2057 the remaining front ends would be able to digest them correctly.
2060 @item -fmessage-length=@var{n}
2061 @opindex fmessage-length
2062 Try to format error messages so that they fit on lines of about @var{n}
2063 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2064 the front ends supported by GCC@. If @var{n} is zero, then no
2065 line-wrapping will be done; each error message will appear on a single
2068 @opindex fdiagnostics-show-location
2069 @item -fdiagnostics-show-location=once
2070 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2071 reporter to emit @emph{once} source location information; that is, in
2072 case the message is too long to fit on a single physical line and has to
2073 be wrapped, the source location won't be emitted (as prefix) again,
2074 over and over, in subsequent continuation lines. This is the default
2077 @item -fdiagnostics-show-location=every-line
2078 Only meaningful in line-wrapping mode. Instructs the diagnostic
2079 messages reporter to emit the same source location information (as
2080 prefix) for physical lines that result from the process of breaking
2081 a message which is too long to fit on a single line.
2085 @node Warning Options
2086 @section Options to Request or Suppress Warnings
2087 @cindex options to control warnings
2088 @cindex warning messages
2089 @cindex messages, warning
2090 @cindex suppressing warnings
2092 Warnings are diagnostic messages that report constructions which
2093 are not inherently erroneous but which are risky or suggest there
2094 may have been an error.
2096 You can request many specific warnings with options beginning @samp{-W},
2097 for example @option{-Wimplicit} to request warnings on implicit
2098 declarations. Each of these specific warning options also has a
2099 negative form beginning @samp{-Wno-} to turn off warnings;
2100 for example, @option{-Wno-implicit}. This manual lists only one of the
2101 two forms, whichever is not the default.
2103 The following options control the amount and kinds of warnings produced
2104 by GCC; for further, language-specific options also refer to
2105 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2109 @cindex syntax checking
2111 @opindex fsyntax-only
2112 Check the code for syntax errors, but don't do anything beyond that.
2116 Issue all the warnings demanded by strict ISO C and ISO C++;
2117 reject all programs that use forbidden extensions, and some other
2118 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2119 version of the ISO C standard specified by any @option{-std} option used.
2121 Valid ISO C and ISO C++ programs should compile properly with or without
2122 this option (though a rare few will require @option{-ansi} or a
2123 @option{-std} option specifying the required version of ISO C)@. However,
2124 without this option, certain GNU extensions and traditional C and C++
2125 features are supported as well. With this option, they are rejected.
2127 @option{-pedantic} does not cause warning messages for use of the
2128 alternate keywords whose names begin and end with @samp{__}. Pedantic
2129 warnings are also disabled in the expression that follows
2130 @code{__extension__}. However, only system header files should use
2131 these escape routes; application programs should avoid them.
2132 @xref{Alternate Keywords}.
2134 Some users try to use @option{-pedantic} to check programs for strict ISO
2135 C conformance. They soon find that it does not do quite what they want:
2136 it finds some non-ISO practices, but not all---only those for which
2137 ISO C @emph{requires} a diagnostic, and some others for which
2138 diagnostics have been added.
2140 A feature to report any failure to conform to ISO C might be useful in
2141 some instances, but would require considerable additional work and would
2142 be quite different from @option{-pedantic}. We don't have plans to
2143 support such a feature in the near future.
2145 Where the standard specified with @option{-std} represents a GNU
2146 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2147 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2148 extended dialect is based. Warnings from @option{-pedantic} are given
2149 where they are required by the base standard. (It would not make sense
2150 for such warnings to be given only for features not in the specified GNU
2151 C dialect, since by definition the GNU dialects of C include all
2152 features the compiler supports with the given option, and there would be
2153 nothing to warn about.)
2155 @item -pedantic-errors
2156 @opindex pedantic-errors
2157 Like @option{-pedantic}, except that errors are produced rather than
2162 Inhibit all warning messages.
2166 Inhibit warning messages about the use of @samp{#import}.
2168 @item -Wchar-subscripts
2169 @opindex Wchar-subscripts
2170 Warn if an array subscript has type @code{char}. This is a common cause
2171 of error, as programmers often forget that this type is signed on some
2173 This warning is enabled by @option{-Wall}.
2177 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2178 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2179 This warning is enabled by @option{-Wall}.
2181 @item -Wfatal-errors
2182 @opindex Wfatal-errors
2183 This option causes the compiler to abort compilation on the first error
2184 occurred rather than trying to keep going and printing further error
2189 @opindex ffreestanding
2190 @opindex fno-builtin
2191 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2192 the arguments supplied have types appropriate to the format string
2193 specified, and that the conversions specified in the format string make
2194 sense. This includes standard functions, and others specified by format
2195 attributes (@pxref{Function Attributes}), in the @code{printf},
2196 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2197 not in the C standard) families (or other target-specific families).
2198 Which functions are checked without format attributes having been
2199 specified depends on the standard version selected, and such checks of
2200 functions without the attribute specified are disabled by
2201 @option{-ffreestanding} or @option{-fno-builtin}.
2203 The formats are checked against the format features supported by GNU
2204 libc version 2.2. These include all ISO C90 and C99 features, as well
2205 as features from the Single Unix Specification and some BSD and GNU
2206 extensions. Other library implementations may not support all these
2207 features; GCC does not support warning about features that go beyond a
2208 particular library's limitations. However, if @option{-pedantic} is used
2209 with @option{-Wformat}, warnings will be given about format features not
2210 in the selected standard version (but not for @code{strfmon} formats,
2211 since those are not in any version of the C standard). @xref{C Dialect
2212 Options,,Options Controlling C Dialect}.
2214 Since @option{-Wformat} also checks for null format arguments for
2215 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2217 @option{-Wformat} is included in @option{-Wall}. For more control over some
2218 aspects of format checking, the options @option{-Wformat-y2k},
2219 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2220 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2221 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2224 @opindex Wformat-y2k
2225 If @option{-Wformat} is specified, also warn about @code{strftime}
2226 formats which may yield only a two-digit year.
2228 @item -Wno-format-extra-args
2229 @opindex Wno-format-extra-args
2230 If @option{-Wformat} is specified, do not warn about excess arguments to a
2231 @code{printf} or @code{scanf} format function. The C standard specifies
2232 that such arguments are ignored.
2234 Where the unused arguments lie between used arguments that are
2235 specified with @samp{$} operand number specifications, normally
2236 warnings are still given, since the implementation could not know what
2237 type to pass to @code{va_arg} to skip the unused arguments. However,
2238 in the case of @code{scanf} formats, this option will suppress the
2239 warning if the unused arguments are all pointers, since the Single
2240 Unix Specification says that such unused arguments are allowed.
2242 @item -Wno-format-zero-length
2243 @opindex Wno-format-zero-length
2244 If @option{-Wformat} is specified, do not warn about zero-length formats.
2245 The C standard specifies that zero-length formats are allowed.
2247 @item -Wformat-nonliteral
2248 @opindex Wformat-nonliteral
2249 If @option{-Wformat} is specified, also warn if the format string is not a
2250 string literal and so cannot be checked, unless the format function
2251 takes its format arguments as a @code{va_list}.
2253 @item -Wformat-security
2254 @opindex Wformat-security
2255 If @option{-Wformat} is specified, also warn about uses of format
2256 functions that represent possible security problems. At present, this
2257 warns about calls to @code{printf} and @code{scanf} functions where the
2258 format string is not a string literal and there are no format arguments,
2259 as in @code{printf (foo);}. This may be a security hole if the format
2260 string came from untrusted input and contains @samp{%n}. (This is
2261 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2262 in future warnings may be added to @option{-Wformat-security} that are not
2263 included in @option{-Wformat-nonliteral}.)
2267 Enable @option{-Wformat} plus format checks not included in
2268 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2269 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2273 Warn about passing a null pointer for arguments marked as
2274 requiring a non-null value by the @code{nonnull} function attribute.
2276 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2277 can be disabled with the @option{-Wno-nonnull} option.
2279 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2281 Warn about uninitialized variables which are initialized with themselves.
2282 Note this option can only be used with the @option{-Wuninitialized} option,
2283 which in turn only works with @option{-O1} and above.
2285 For example, GCC will warn about @code{i} being uninitialized in the
2286 following snippet only when @option{-Winit-self} has been specified:
2297 @item -Wimplicit-int
2298 @opindex Wimplicit-int
2299 Warn when a declaration does not specify a type.
2300 This warning is enabled by @option{-Wall}.
2302 @item -Wimplicit-function-declaration
2303 @itemx -Werror-implicit-function-declaration
2304 @opindex Wimplicit-function-declaration
2305 @opindex Werror-implicit-function-declaration
2306 Give a warning (or error) whenever a function is used before being
2307 declared. The form @option{-Wno-error-implicit-function-declaration}
2309 This warning is enabled by @option{-Wall} (as a warning, not an error).
2313 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2314 This warning is enabled by @option{-Wall}.
2318 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2319 function with external linkage, returning int, taking either zero
2320 arguments, two, or three arguments of appropriate types.
2321 This warning is enabled by @option{-Wall}.
2323 @item -Wmissing-braces
2324 @opindex Wmissing-braces
2325 Warn if an aggregate or union initializer is not fully bracketed. In
2326 the following example, the initializer for @samp{a} is not fully
2327 bracketed, but that for @samp{b} is fully bracketed.
2330 int a[2][2] = @{ 0, 1, 2, 3 @};
2331 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2334 This warning is enabled by @option{-Wall}.
2336 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2337 @opindex Wmissing-include-dirs
2338 Warn if a user-supplied include directory does not exist.
2341 @opindex Wparentheses
2342 Warn if parentheses are omitted in certain contexts, such
2343 as when there is an assignment in a context where a truth value
2344 is expected, or when operators are nested whose precedence people
2345 often get confused about. Only the warning for an assignment used as
2346 a truth value is supported when compiling C++; the other warnings are
2347 only supported when compiling C@.
2349 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2350 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2351 interpretation from that of ordinary mathematical notation.
2353 Also warn about constructions where there may be confusion to which
2354 @code{if} statement an @code{else} branch belongs. Here is an example of
2369 In C, every @code{else} branch belongs to the innermost possible @code{if}
2370 statement, which in this example is @code{if (b)}. This is often not
2371 what the programmer expected, as illustrated in the above example by
2372 indentation the programmer chose. When there is the potential for this
2373 confusion, GCC will issue a warning when this flag is specified.
2374 To eliminate the warning, add explicit braces around the innermost
2375 @code{if} statement so there is no way the @code{else} could belong to
2376 the enclosing @code{if}. The resulting code would look like this:
2392 This warning is enabled by @option{-Wall}.
2394 @item -Wsequence-point
2395 @opindex Wsequence-point
2396 Warn about code that may have undefined semantics because of violations
2397 of sequence point rules in the C standard.
2399 The C standard defines the order in which expressions in a C program are
2400 evaluated in terms of @dfn{sequence points}, which represent a partial
2401 ordering between the execution of parts of the program: those executed
2402 before the sequence point, and those executed after it. These occur
2403 after the evaluation of a full expression (one which is not part of a
2404 larger expression), after the evaluation of the first operand of a
2405 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2406 function is called (but after the evaluation of its arguments and the
2407 expression denoting the called function), and in certain other places.
2408 Other than as expressed by the sequence point rules, the order of
2409 evaluation of subexpressions of an expression is not specified. All
2410 these rules describe only a partial order rather than a total order,
2411 since, for example, if two functions are called within one expression
2412 with no sequence point between them, the order in which the functions
2413 are called is not specified. However, the standards committee have
2414 ruled that function calls do not overlap.
2416 It is not specified when between sequence points modifications to the
2417 values of objects take effect. Programs whose behavior depends on this
2418 have undefined behavior; the C standard specifies that ``Between the
2419 previous and next sequence point an object shall have its stored value
2420 modified at most once by the evaluation of an expression. Furthermore,
2421 the prior value shall be read only to determine the value to be
2422 stored.''. If a program breaks these rules, the results on any
2423 particular implementation are entirely unpredictable.
2425 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2426 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2427 diagnosed by this option, and it may give an occasional false positive
2428 result, but in general it has been found fairly effective at detecting
2429 this sort of problem in programs.
2431 The present implementation of this option only works for C programs. A
2432 future implementation may also work for C++ programs.
2434 The C standard is worded confusingly, therefore there is some debate
2435 over the precise meaning of the sequence point rules in subtle cases.
2436 Links to discussions of the problem, including proposed formal
2437 definitions, may be found on the GCC readings page, at
2438 @w{@uref{http://gcc.gnu.org/readings.html}}.
2440 This warning is enabled by @option{-Wall}.
2443 @opindex Wreturn-type
2444 Warn whenever a function is defined with a return-type that defaults to
2445 @code{int}. Also warn about any @code{return} statement with no
2446 return-value in a function whose return-type is not @code{void}.
2448 For C, also warn if the return type of a function has a type qualifier
2449 such as @code{const}. Such a type qualifier has no effect, since the
2450 value returned by a function is not an lvalue. ISO C prohibits
2451 qualified @code{void} return types on function definitions, so such
2452 return types always receive a warning even without this option.
2454 For C++, a function without return type always produces a diagnostic
2455 message, even when @option{-Wno-return-type} is specified. The only
2456 exceptions are @samp{main} and functions defined in system headers.
2458 This warning is enabled by @option{-Wall}.
2462 Warn whenever a @code{switch} statement has an index of enumerated type
2463 and lacks a @code{case} for one or more of the named codes of that
2464 enumeration. (The presence of a @code{default} label prevents this
2465 warning.) @code{case} labels outside the enumeration range also
2466 provoke warnings when this option is used.
2467 This warning is enabled by @option{-Wall}.
2469 @item -Wswitch-default
2470 @opindex Wswitch-switch
2471 Warn whenever a @code{switch} statement does not have a @code{default}
2475 @opindex Wswitch-enum
2476 Warn whenever a @code{switch} statement has an index of enumerated type
2477 and lacks a @code{case} for one or more of the named codes of that
2478 enumeration. @code{case} labels outside the enumeration range also
2479 provoke warnings when this option is used.
2483 Warn if any trigraphs are encountered that might change the meaning of
2484 the program (trigraphs within comments are not warned about).
2485 This warning is enabled by @option{-Wall}.
2487 @item -Wunused-function
2488 @opindex Wunused-function
2489 Warn whenever a static function is declared but not defined or a
2490 non\-inline static function is unused.
2491 This warning is enabled by @option{-Wall}.
2493 @item -Wunused-label
2494 @opindex Wunused-label
2495 Warn whenever a label is declared but not used.
2496 This warning is enabled by @option{-Wall}.
2498 To suppress this warning use the @samp{unused} attribute
2499 (@pxref{Variable Attributes}).
2501 @item -Wunused-parameter
2502 @opindex Wunused-parameter
2503 Warn whenever a function parameter is unused aside from its declaration.
2505 To suppress this warning use the @samp{unused} attribute
2506 (@pxref{Variable Attributes}).
2508 @item -Wunused-variable
2509 @opindex Wunused-variable
2510 Warn whenever a local variable or non-constant static variable is unused
2511 aside from its declaration
2512 This warning is enabled by @option{-Wall}.
2514 To suppress this warning use the @samp{unused} attribute
2515 (@pxref{Variable Attributes}).
2517 @item -Wunused-value
2518 @opindex Wunused-value
2519 Warn whenever a statement computes a result that is explicitly not used.
2520 This warning is enabled by @option{-Wall}.
2522 To suppress this warning cast the expression to @samp{void}.
2526 All the above @option{-Wunused} options combined.
2528 In order to get a warning about an unused function parameter, you must
2529 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2530 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2532 @item -Wuninitialized
2533 @opindex Wuninitialized
2534 Warn if an automatic variable is used without first being initialized or
2535 if a variable may be clobbered by a @code{setjmp} call.
2537 These warnings are possible only in optimizing compilation,
2538 because they require data flow information that is computed only
2539 when optimizing. If you don't specify @option{-O}, you simply won't
2542 If you want to warn about code which uses the uninitialized value of the
2543 variable in its own initializer, use the @option{-Winit-self} option.
2545 These warnings occur only for variables that are candidates for
2546 register allocation. Therefore, they do not occur for a variable that
2547 is declared @code{volatile}, or whose address is taken, or whose size
2548 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
2549 structures, unions or arrays, even when they are in registers.
2551 Note that there may be no warning about a variable that is used only
2552 to compute a value that itself is never used, because such
2553 computations may be deleted by data flow analysis before the warnings
2556 These warnings are made optional because GCC is not smart
2557 enough to see all the reasons why the code might be correct
2558 despite appearing to have an error. Here is one example of how
2579 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2580 always initialized, but GCC doesn't know this. Here is
2581 another common case:
2586 if (change_y) save_y = y, y = new_y;
2588 if (change_y) y = save_y;
2593 This has no bug because @code{save_y} is used only if it is set.
2595 @cindex @code{longjmp} warnings
2596 This option also warns when a non-volatile automatic variable might be
2597 changed by a call to @code{longjmp}. These warnings as well are possible
2598 only in optimizing compilation.
2600 The compiler sees only the calls to @code{setjmp}. It cannot know
2601 where @code{longjmp} will be called; in fact, a signal handler could
2602 call it at any point in the code. As a result, you may get a warning
2603 even when there is in fact no problem because @code{longjmp} cannot
2604 in fact be called at the place which would cause a problem.
2606 Some spurious warnings can be avoided if you declare all the functions
2607 you use that never return as @code{noreturn}. @xref{Function
2610 This warning is enabled by @option{-Wall}.
2612 @item -Wunknown-pragmas
2613 @opindex Wunknown-pragmas
2614 @cindex warning for unknown pragmas
2615 @cindex unknown pragmas, warning
2616 @cindex pragmas, warning of unknown
2617 Warn when a #pragma directive is encountered which is not understood by
2618 GCC@. If this command line option is used, warnings will even be issued
2619 for unknown pragmas in system header files. This is not the case if
2620 the warnings were only enabled by the @option{-Wall} command line option.
2622 @item -Wstrict-aliasing
2623 @opindex Wstrict-aliasing
2624 This option is only active when @option{-fstrict-aliasing} is active.
2625 It warns about code which might break the strict aliasing rules that the
2626 compiler is using for optimization. The warning does not catch all
2627 cases, but does attempt to catch the more common pitfalls. It is
2628 included in @option{-Wall}.
2630 @item -Wstrict-aliasing=2
2631 @opindex Wstrict-aliasing=2
2632 This option is only active when @option{-fstrict-aliasing} is active.
2633 It warns about all code which might break the strict aliasing rules that the
2634 compiler is using for optimization. This warning catches all cases, but
2635 it will also give a warning for some ambiguous cases that are safe.
2639 All of the above @samp{-W} options combined. This enables all the
2640 warnings about constructions that some users consider questionable, and
2641 that are easy to avoid (or modify to prevent the warning), even in
2642 conjunction with macros. This also enables some language-specific
2643 warnings described in @ref{C++ Dialect Options} and
2644 @ref{Objective-C and Objective-C++ Dialect Options}.
2647 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2648 Some of them warn about constructions that users generally do not
2649 consider questionable, but which occasionally you might wish to check
2650 for; others warn about constructions that are necessary or hard to avoid
2651 in some cases, and there is no simple way to modify the code to suppress
2658 (This option used to be called @option{-W}. The older name is still
2659 supported, but the newer name is more descriptive.) Print extra warning
2660 messages for these events:
2664 A function can return either with or without a value. (Falling
2665 off the end of the function body is considered returning without
2666 a value.) For example, this function would evoke such a
2680 An expression-statement or the left-hand side of a comma expression
2681 contains no side effects.
2682 To suppress the warning, cast the unused expression to void.
2683 For example, an expression such as @samp{x[i,j]} will cause a warning,
2684 but @samp{x[(void)i,j]} will not.
2687 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2690 Storage-class specifiers like @code{static} are not the first things in
2691 a declaration. According to the C Standard, this usage is obsolescent.
2694 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2698 A comparison between signed and unsigned values could produce an
2699 incorrect result when the signed value is converted to unsigned.
2700 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2703 An aggregate has an initializer which does not initialize all members.
2704 This warning can be independently controlled by
2705 @option{-Wmissing-field-initializers}.
2708 A function parameter is declared without a type specifier in K&R-style
2716 An empty body occurs in an @samp{if} or @samp{else} statement.
2719 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2720 @samp{>}, or @samp{>=}.
2723 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2726 Any of several floating-point events that often indicate errors, such as
2727 overflow, underflow, loss of precision, etc.
2729 @item @r{(C++ only)}
2730 An enumerator and a non-enumerator both appear in a conditional expression.
2732 @item @r{(C++ only)}
2733 A non-static reference or non-static @samp{const} member appears in a
2734 class without constructors.
2736 @item @r{(C++ only)}
2737 Ambiguous virtual bases.
2739 @item @r{(C++ only)}
2740 Subscripting an array which has been declared @samp{register}.
2742 @item @r{(C++ only)}
2743 Taking the address of a variable which has been declared @samp{register}.
2745 @item @r{(C++ only)}
2746 A base class is not initialized in a derived class' copy constructor.
2749 @item -Wno-div-by-zero
2750 @opindex Wno-div-by-zero
2751 @opindex Wdiv-by-zero
2752 Do not warn about compile-time integer division by zero. Floating point
2753 division by zero is not warned about, as it can be a legitimate way of
2754 obtaining infinities and NaNs.
2756 @item -Wsystem-headers
2757 @opindex Wsystem-headers
2758 @cindex warnings from system headers
2759 @cindex system headers, warnings from
2760 Print warning messages for constructs found in system header files.
2761 Warnings from system headers are normally suppressed, on the assumption
2762 that they usually do not indicate real problems and would only make the
2763 compiler output harder to read. Using this command line option tells
2764 GCC to emit warnings from system headers as if they occurred in user
2765 code. However, note that using @option{-Wall} in conjunction with this
2766 option will @emph{not} warn about unknown pragmas in system
2767 headers---for that, @option{-Wunknown-pragmas} must also be used.
2770 @opindex Wfloat-equal
2771 Warn if floating point values are used in equality comparisons.
2773 The idea behind this is that sometimes it is convenient (for the
2774 programmer) to consider floating-point values as approximations to
2775 infinitely precise real numbers. If you are doing this, then you need
2776 to compute (by analyzing the code, or in some other way) the maximum or
2777 likely maximum error that the computation introduces, and allow for it
2778 when performing comparisons (and when producing output, but that's a
2779 different problem). In particular, instead of testing for equality, you
2780 would check to see whether the two values have ranges that overlap; and
2781 this is done with the relational operators, so equality comparisons are
2784 @item -Wtraditional @r{(C only)}
2785 @opindex Wtraditional
2786 Warn about certain constructs that behave differently in traditional and
2787 ISO C@. Also warn about ISO C constructs that have no traditional C
2788 equivalent, and/or problematic constructs which should be avoided.
2792 Macro parameters that appear within string literals in the macro body.
2793 In traditional C macro replacement takes place within string literals,
2794 but does not in ISO C@.
2797 In traditional C, some preprocessor directives did not exist.
2798 Traditional preprocessors would only consider a line to be a directive
2799 if the @samp{#} appeared in column 1 on the line. Therefore
2800 @option{-Wtraditional} warns about directives that traditional C
2801 understands but would ignore because the @samp{#} does not appear as the
2802 first character on the line. It also suggests you hide directives like
2803 @samp{#pragma} not understood by traditional C by indenting them. Some
2804 traditional implementations would not recognize @samp{#elif}, so it
2805 suggests avoiding it altogether.
2808 A function-like macro that appears without arguments.
2811 The unary plus operator.
2814 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2815 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2816 constants.) Note, these suffixes appear in macros defined in the system
2817 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2818 Use of these macros in user code might normally lead to spurious
2819 warnings, however GCC's integrated preprocessor has enough context to
2820 avoid warning in these cases.
2823 A function declared external in one block and then used after the end of
2827 A @code{switch} statement has an operand of type @code{long}.
2830 A non-@code{static} function declaration follows a @code{static} one.
2831 This construct is not accepted by some traditional C compilers.
2834 The ISO type of an integer constant has a different width or
2835 signedness from its traditional type. This warning is only issued if
2836 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2837 typically represent bit patterns, are not warned about.
2840 Usage of ISO string concatenation is detected.
2843 Initialization of automatic aggregates.
2846 Identifier conflicts with labels. Traditional C lacks a separate
2847 namespace for labels.
2850 Initialization of unions. If the initializer is zero, the warning is
2851 omitted. This is done under the assumption that the zero initializer in
2852 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2853 initializer warnings and relies on default initialization to zero in the
2857 Conversions by prototypes between fixed/floating point values and vice
2858 versa. The absence of these prototypes when compiling with traditional
2859 C would cause serious problems. This is a subset of the possible
2860 conversion warnings, for the full set use @option{-Wconversion}.
2863 Use of ISO C style function definitions. This warning intentionally is
2864 @emph{not} issued for prototype declarations or variadic functions
2865 because these ISO C features will appear in your code when using
2866 libiberty's traditional C compatibility macros, @code{PARAMS} and
2867 @code{VPARAMS}. This warning is also bypassed for nested functions
2868 because that feature is already a GCC extension and thus not relevant to
2869 traditional C compatibility.
2872 @item -Wdeclaration-after-statement @r{(C only)}
2873 @opindex Wdeclaration-after-statement
2874 Warn when a declaration is found after a statement in a block. This
2875 construct, known from C++, was introduced with ISO C99 and is by default
2876 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2877 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2881 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2883 @item -Wno-endif-labels
2884 @opindex Wno-endif-labels
2885 @opindex Wendif-labels
2886 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2890 Warn whenever a local variable shadows another local variable, parameter or
2891 global variable or whenever a built-in function is shadowed.
2893 @item -Wlarger-than-@var{len}
2894 @opindex Wlarger-than
2895 Warn whenever an object of larger than @var{len} bytes is defined.
2897 @item -Wpointer-arith
2898 @opindex Wpointer-arith
2899 Warn about anything that depends on the ``size of'' a function type or
2900 of @code{void}. GNU C assigns these types a size of 1, for
2901 convenience in calculations with @code{void *} pointers and pointers
2904 @item -Wbad-function-cast @r{(C only)}
2905 @opindex Wbad-function-cast
2906 Warn whenever a function call is cast to a non-matching type.
2907 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2911 Warn whenever a pointer is cast so as to remove a type qualifier from
2912 the target type. For example, warn if a @code{const char *} is cast
2913 to an ordinary @code{char *}.
2916 @opindex Wcast-align
2917 Warn whenever a pointer is cast such that the required alignment of the
2918 target is increased. For example, warn if a @code{char *} is cast to
2919 an @code{int *} on machines where integers can only be accessed at
2920 two- or four-byte boundaries.
2922 @item -Wwrite-strings
2923 @opindex Wwrite-strings
2924 When compiling C, give string constants the type @code{const
2925 char[@var{length}]} so that
2926 copying the address of one into a non-@code{const} @code{char *}
2927 pointer will get a warning; when compiling C++, warn about the
2928 deprecated conversion from string constants to @code{char *}.
2929 These warnings will help you find at
2930 compile time code that can try to write into a string constant, but
2931 only if you have been very careful about using @code{const} in
2932 declarations and prototypes. Otherwise, it will just be a nuisance;
2933 this is why we did not make @option{-Wall} request these warnings.
2936 @opindex Wconversion
2937 Warn if a prototype causes a type conversion that is different from what
2938 would happen to the same argument in the absence of a prototype. This
2939 includes conversions of fixed point to floating and vice versa, and
2940 conversions changing the width or signedness of a fixed point argument
2941 except when the same as the default promotion.
2943 Also, warn if a negative integer constant expression is implicitly
2944 converted to an unsigned type. For example, warn about the assignment
2945 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2946 casts like @code{(unsigned) -1}.
2948 @item -Wsign-compare
2949 @opindex Wsign-compare
2950 @cindex warning for comparison of signed and unsigned values
2951 @cindex comparison of signed and unsigned values, warning
2952 @cindex signed and unsigned values, comparison warning
2953 Warn when a comparison between signed and unsigned values could produce
2954 an incorrect result when the signed value is converted to unsigned.
2955 This warning is also enabled by @option{-Wextra}; to get the other warnings
2956 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2958 @item -Waggregate-return
2959 @opindex Waggregate-return
2960 Warn if any functions that return structures or unions are defined or
2961 called. (In languages where you can return an array, this also elicits
2964 @item -Wstrict-prototypes @r{(C only)}
2965 @opindex Wstrict-prototypes
2966 Warn if a function is declared or defined without specifying the
2967 argument types. (An old-style function definition is permitted without
2968 a warning if preceded by a declaration which specifies the argument
2971 @item -Wold-style-definition @r{(C only)}
2972 @opindex Wold-style-definition
2973 Warn if an old-style function definition is used. A warning is given
2974 even if there is a previous prototype.
2976 @item -Wmissing-prototypes @r{(C only)}
2977 @opindex Wmissing-prototypes
2978 Warn if a global function is defined without a previous prototype
2979 declaration. This warning is issued even if the definition itself
2980 provides a prototype. The aim is to detect global functions that fail
2981 to be declared in header files.
2983 @item -Wmissing-declarations @r{(C only)}
2984 @opindex Wmissing-declarations
2985 Warn if a global function is defined without a previous declaration.
2986 Do so even if the definition itself provides a prototype.
2987 Use this option to detect global functions that are not declared in
2990 @item -Wmissing-field-initializers
2991 @opindex Wmissing-field-initializers
2994 Warn if a structure's initializer has some fields missing. For
2995 example, the following code would cause such a warning, because
2996 @code{x.h} is implicitly zero:
2999 struct s @{ int f, g, h; @};
3000 struct s x = @{ 3, 4 @};
3003 This option does not warn about designated initializers, so the following
3004 modification would not trigger a warning:
3007 struct s @{ int f, g, h; @};
3008 struct s x = @{ .f = 3, .g = 4 @};
3011 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3012 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3014 @item -Wmissing-noreturn
3015 @opindex Wmissing-noreturn
3016 Warn about functions which might be candidates for attribute @code{noreturn}.
3017 Note these are only possible candidates, not absolute ones. Care should
3018 be taken to manually verify functions actually do not ever return before
3019 adding the @code{noreturn} attribute, otherwise subtle code generation
3020 bugs could be introduced. You will not get a warning for @code{main} in
3021 hosted C environments.
3023 @item -Wmissing-format-attribute
3024 @opindex Wmissing-format-attribute
3026 If @option{-Wformat} is enabled, also warn about functions which might be
3027 candidates for @code{format} attributes. Note these are only possible
3028 candidates, not absolute ones. GCC will guess that @code{format}
3029 attributes might be appropriate for any function that calls a function
3030 like @code{vprintf} or @code{vscanf}, but this might not always be the
3031 case, and some functions for which @code{format} attributes are
3032 appropriate may not be detected. This option has no effect unless
3033 @option{-Wformat} is enabled (possibly by @option{-Wall}).
3035 @item -Wno-multichar
3036 @opindex Wno-multichar
3038 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3039 Usually they indicate a typo in the user's code, as they have
3040 implementation-defined values, and should not be used in portable code.
3042 @item -Wnormalized=<none|id|nfc|nfkc>
3043 @opindex Wnormalized
3046 @cindex character set, input normalization
3047 In ISO C and ISO C++, two identifiers are different if they are
3048 different sequences of characters. However, sometimes when characters
3049 outside the basic ASCII character set are used, you can have two
3050 different character sequences that look the same. To avoid confusion,
3051 the ISO 10646 standard sets out some @dfn{normalization rules} which
3052 when applied ensure that two sequences that look the same are turned into
3053 the same sequence. GCC can warn you if you are using identifiers which
3054 have not been normalized; this option controls that warning.
3056 There are four levels of warning that GCC supports. The default is
3057 @option{-Wnormalized=nfc}, which warns about any identifier which is
3058 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3059 recommended form for most uses.
3061 Unfortunately, there are some characters which ISO C and ISO C++ allow
3062 in identifiers that when turned into NFC aren't allowable as
3063 identifiers. That is, there's no way to use these symbols in portable
3064 ISO C or C++ and have all your identifiers in NFC.
3065 @option{-Wnormalized=id} suppresses the warning for these characters.
3066 It is hoped that future versions of the standards involved will correct
3067 this, which is why this option is not the default.
3069 You can switch the warning off for all characters by writing
3070 @option{-Wnormalized=none}. You would only want to do this if you
3071 were using some other normalization scheme (like ``D''), because
3072 otherwise you can easily create bugs that are literally impossible to see.
3074 Some characters in ISO 10646 have distinct meanings but look identical
3075 in some fonts or display methodologies, especially once formatting has
3076 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3077 LETTER N'', will display just like a regular @code{n} which has been
3078 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3079 normalisation scheme to convert all these into a standard form as
3080 well, and GCC will warn if your code is not in NFKC if you use
3081 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3082 about every identifier that contains the letter O because it might be
3083 confused with the digit 0, and so is not the default, but may be
3084 useful as a local coding convention if the programming environment is
3085 unable to be fixed to display these characters distinctly.
3087 @item -Wno-deprecated-declarations
3088 @opindex Wno-deprecated-declarations
3089 Do not warn about uses of functions, variables, and types marked as
3090 deprecated by using the @code{deprecated} attribute.
3091 (@pxref{Function Attributes}, @pxref{Variable Attributes},
3092 @pxref{Type Attributes}.)
3096 Warn if a structure is given the packed attribute, but the packed
3097 attribute has no effect on the layout or size of the structure.
3098 Such structures may be mis-aligned for little benefit. For
3099 instance, in this code, the variable @code{f.x} in @code{struct bar}
3100 will be misaligned even though @code{struct bar} does not itself
3101 have the packed attribute:
3108 @} __attribute__((packed));
3118 Warn if padding is included in a structure, either to align an element
3119 of the structure or to align the whole structure. Sometimes when this
3120 happens it is possible to rearrange the fields of the structure to
3121 reduce the padding and so make the structure smaller.
3123 @item -Wredundant-decls
3124 @opindex Wredundant-decls
3125 Warn if anything is declared more than once in the same scope, even in
3126 cases where multiple declaration is valid and changes nothing.
3128 @item -Wnested-externs @r{(C only)}
3129 @opindex Wnested-externs
3130 Warn if an @code{extern} declaration is encountered within a function.
3132 @item -Wunreachable-code
3133 @opindex Wunreachable-code
3134 Warn if the compiler detects that code will never be executed.
3136 This option is intended to warn when the compiler detects that at
3137 least a whole line of source code will never be executed, because
3138 some condition is never satisfied or because it is after a
3139 procedure that never returns.
3141 It is possible for this option to produce a warning even though there
3142 are circumstances under which part of the affected line can be executed,
3143 so care should be taken when removing apparently-unreachable code.
3145 For instance, when a function is inlined, a warning may mean that the
3146 line is unreachable in only one inlined copy of the function.
3148 This option is not made part of @option{-Wall} because in a debugging
3149 version of a program there is often substantial code which checks
3150 correct functioning of the program and is, hopefully, unreachable
3151 because the program does work. Another common use of unreachable
3152 code is to provide behavior which is selectable at compile-time.
3156 Warn if a function can not be inlined and it was declared as inline.
3157 Even with this option, the compiler will not warn about failures to
3158 inline functions declared in system headers.
3160 The compiler uses a variety of heuristics to determine whether or not
3161 to inline a function. For example, the compiler takes into account
3162 the size of the function being inlined and the the amount of inlining
3163 that has already been done in the current function. Therefore,
3164 seemingly insignificant changes in the source program can cause the
3165 warnings produced by @option{-Winline} to appear or disappear.
3167 @item -Wno-invalid-offsetof @r{(C++ only)}
3168 @opindex Wno-invalid-offsetof
3169 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3170 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3171 to a non-POD type is undefined. In existing C++ implementations,
3172 however, @samp{offsetof} typically gives meaningful results even when
3173 applied to certain kinds of non-POD types. (Such as a simple
3174 @samp{struct} that fails to be a POD type only by virtue of having a
3175 constructor.) This flag is for users who are aware that they are
3176 writing nonportable code and who have deliberately chosen to ignore the
3179 The restrictions on @samp{offsetof} may be relaxed in a future version
3180 of the C++ standard.
3183 @opindex Winvalid-pch
3184 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3185 the search path but can't be used.
3189 @opindex Wno-long-long
3190 Warn if @samp{long long} type is used. This is default. To inhibit
3191 the warning messages, use @option{-Wno-long-long}. Flags
3192 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3193 only when @option{-pedantic} flag is used.
3195 @item -Wvariadic-macros
3196 @opindex Wvariadic-macros
3197 @opindex Wno-variadic-macros
3198 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3199 alternate syntax when in pedantic ISO C99 mode. This is default.
3200 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3202 @item -Wdisabled-optimization
3203 @opindex Wdisabled-optimization
3204 Warn if a requested optimization pass is disabled. This warning does
3205 not generally indicate that there is anything wrong with your code; it
3206 merely indicates that GCC's optimizers were unable to handle the code
3207 effectively. Often, the problem is that your code is too big or too
3208 complex; GCC will refuse to optimize programs when the optimization
3209 itself is likely to take inordinate amounts of time.
3211 @item -Wno-pointer-sign
3212 @opindex Wno-pointer-sign
3213 Don't warn for pointer argument passing or assignment with different signedness.
3214 Only useful in the negative form since this warning is enabled by default.
3215 This option is only supported for C and Objective-C@.
3219 Make all warnings into errors.
3222 @node Debugging Options
3223 @section Options for Debugging Your Program or GCC
3224 @cindex options, debugging
3225 @cindex debugging information options
3227 GCC has various special options that are used for debugging
3228 either your program or GCC:
3233 Produce debugging information in the operating system's native format
3234 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3237 On most systems that use stabs format, @option{-g} enables use of extra
3238 debugging information that only GDB can use; this extra information
3239 makes debugging work better in GDB but will probably make other debuggers
3241 refuse to read the program. If you want to control for certain whether
3242 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3243 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3245 GCC allows you to use @option{-g} with
3246 @option{-O}. The shortcuts taken by optimized code may occasionally
3247 produce surprising results: some variables you declared may not exist
3248 at all; flow of control may briefly move where you did not expect it;
3249 some statements may not be executed because they compute constant
3250 results or their values were already at hand; some statements may
3251 execute in different places because they were moved out of loops.
3253 Nevertheless it proves possible to debug optimized output. This makes
3254 it reasonable to use the optimizer for programs that might have bugs.
3256 The following options are useful when GCC is generated with the
3257 capability for more than one debugging format.
3261 Produce debugging information for use by GDB@. This means to use the
3262 most expressive format available (DWARF 2, stabs, or the native format
3263 if neither of those are supported), including GDB extensions if at all
3268 Produce debugging information in stabs format (if that is supported),
3269 without GDB extensions. This is the format used by DBX on most BSD
3270 systems. On MIPS, Alpha and System V Release 4 systems this option
3271 produces stabs debugging output which is not understood by DBX or SDB@.
3272 On System V Release 4 systems this option requires the GNU assembler.
3274 @item -feliminate-unused-debug-symbols
3275 @opindex feliminate-unused-debug-symbols
3276 Produce debugging information in stabs format (if that is supported),
3277 for only symbols that are actually used.
3281 Produce debugging information in stabs format (if that is supported),
3282 using GNU extensions understood only by the GNU debugger (GDB)@. The
3283 use of these extensions is likely to make other debuggers crash or
3284 refuse to read the program.
3288 Produce debugging information in COFF format (if that is supported).
3289 This is the format used by SDB on most System V systems prior to
3294 Produce debugging information in XCOFF format (if that is supported).
3295 This is the format used by the DBX debugger on IBM RS/6000 systems.
3299 Produce debugging information in XCOFF format (if that is supported),
3300 using GNU extensions understood only by the GNU debugger (GDB)@. The
3301 use of these extensions is likely to make other debuggers crash or
3302 refuse to read the program, and may cause assemblers other than the GNU
3303 assembler (GAS) to fail with an error.
3307 Produce debugging information in DWARF version 2 format (if that is
3308 supported). This is the format used by DBX on IRIX 6. With this
3309 option, GCC uses features of DWARF version 3 when they are useful;
3310 version 3 is upward compatible with version 2, but may still cause
3311 problems for older debuggers.
3315 Produce debugging information in VMS debug format (if that is
3316 supported). This is the format used by DEBUG on VMS systems.
3319 @itemx -ggdb@var{level}
3320 @itemx -gstabs@var{level}
3321 @itemx -gcoff@var{level}
3322 @itemx -gxcoff@var{level}
3323 @itemx -gvms@var{level}
3324 Request debugging information and also use @var{level} to specify how
3325 much information. The default level is 2.
3327 Level 1 produces minimal information, enough for making backtraces in
3328 parts of the program that you don't plan to debug. This includes
3329 descriptions of functions and external variables, but no information
3330 about local variables and no line numbers.
3332 Level 3 includes extra information, such as all the macro definitions
3333 present in the program. Some debuggers support macro expansion when
3334 you use @option{-g3}.
3336 @option{-gdwarf-2} does not accept a concatenated debug level, because
3337 GCC used to support an option @option{-gdwarf} that meant to generate
3338 debug information in version 1 of the DWARF format (which is very
3339 different from version 2), and it would have been too confusing. That
3340 debug format is long obsolete, but the option cannot be changed now.
3341 Instead use an additional @option{-g@var{level}} option to change the
3342 debug level for DWARF2.
3344 @item -feliminate-dwarf2-dups
3345 @opindex feliminate-dwarf2-dups
3346 Compress DWARF2 debugging information by eliminating duplicated
3347 information about each symbol. This option only makes sense when
3348 generating DWARF2 debugging information with @option{-gdwarf-2}.
3350 @cindex @command{prof}
3353 Generate extra code to write profile information suitable for the
3354 analysis program @command{prof}. You must use this option when compiling
3355 the source files you want data about, and you must also use it when
3358 @cindex @command{gprof}
3361 Generate extra code to write profile information suitable for the
3362 analysis program @command{gprof}. You must use this option when compiling
3363 the source files you want data about, and you must also use it when
3368 Makes the compiler print out each function name as it is compiled, and
3369 print some statistics about each pass when it finishes.
3372 @opindex ftime-report
3373 Makes the compiler print some statistics about the time consumed by each
3374 pass when it finishes.
3377 @opindex fmem-report
3378 Makes the compiler print some statistics about permanent memory
3379 allocation when it finishes.
3381 @item -fprofile-arcs
3382 @opindex fprofile-arcs
3383 Add code so that program flow @dfn{arcs} are instrumented. During
3384 execution the program records how many times each branch and call is
3385 executed and how many times it is taken or returns. When the compiled
3386 program exits it saves this data to a file called
3387 @file{@var{auxname}.gcda} for each source file. The data may be used for
3388 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3389 test coverage analysis (@option{-ftest-coverage}). Each object file's
3390 @var{auxname} is generated from the name of the output file, if
3391 explicitly specified and it is not the final executable, otherwise it is
3392 the basename of the source file. In both cases any suffix is removed
3393 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3394 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3396 @cindex @command{gcov}
3400 This option is used to compile and link code instrumented for coverage
3401 analysis. The option is a synonym for @option{-fprofile-arcs}
3402 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3403 linking). See the documentation for those options for more details.
3408 Compile the source files with @option{-fprofile-arcs} plus optimization
3409 and code generation options. For test coverage analysis, use the
3410 additional @option{-ftest-coverage} option. You do not need to profile
3411 every source file in a program.
3414 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3415 (the latter implies the former).
3418 Run the program on a representative workload to generate the arc profile
3419 information. This may be repeated any number of times. You can run
3420 concurrent instances of your program, and provided that the file system
3421 supports locking, the data files will be correctly updated. Also
3422 @code{fork} calls are detected and correctly handled (double counting
3426 For profile-directed optimizations, compile the source files again with
3427 the same optimization and code generation options plus
3428 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3429 Control Optimization}).
3432 For test coverage analysis, use @command{gcov} to produce human readable
3433 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3434 @command{gcov} documentation for further information.
3438 With @option{-fprofile-arcs}, for each function of your program GCC
3439 creates a program flow graph, then finds a spanning tree for the graph.
3440 Only arcs that are not on the spanning tree have to be instrumented: the
3441 compiler adds code to count the number of times that these arcs are
3442 executed. When an arc is the only exit or only entrance to a block, the
3443 instrumentation code can be added to the block; otherwise, a new basic
3444 block must be created to hold the instrumentation code.
3446 @item -ftree-based-profiling
3447 @opindex ftree-based-profiling
3448 This option is used in addition to @option{-fprofile-arcs} or
3449 @option{-fbranch-probabilities} to control whether those optimizations
3450 are performed on a tree-based or rtl-based internal representation.
3451 If you use this option when compiling with @option{-fprofile-arcs},
3452 you must also use it when compiling later with @option{-fbranch-probabilities}.
3453 Currently the tree-based optimization is in an early stage of
3454 development, and this option is recommended only for those people
3455 working on improving it.
3458 @item -ftest-coverage
3459 @opindex ftest-coverage
3460 Produce a notes file that the @command{gcov} code-coverage utility
3461 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3462 show program coverage. Each source file's note file is called
3463 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3464 above for a description of @var{auxname} and instructions on how to
3465 generate test coverage data. Coverage data will match the source files
3466 more closely, if you do not optimize.
3468 @item -d@var{letters}
3469 @item -fdump-rtl-@var{pass}
3471 Says to make debugging dumps during compilation at times specified by
3472 @var{letters}. This is used for debugging the RTL-based passes of the
3473 compiler. The file names for most of the dumps are made by appending a
3474 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3475 from the name of the output file, if explicitly specified and it is not
3476 an executable, otherwise it is the basename of the source file.
3478 Most debug dumps can be enabled either passing a letter to the @option{-d}
3479 option, or with a long @option{-fdump-rtl} switch; here are the possible
3480 letters for use in @var{letters} and @var{pass}, and their meanings:
3485 Annotate the assembler output with miscellaneous debugging information.
3488 @itemx -fdump-rtl-bp
3490 @opindex fdump-rtl-bp
3491 Dump after computing branch probabilities, to @file{@var{file}.09.bp}.
3494 @itemx -fdump-rtl-bbro
3496 @opindex fdump-rtl-bbro
3497 Dump after block reordering, to @file{@var{file}.30.bbro}.
3500 @itemx -fdump-rtl-combine
3502 @opindex fdump-rtl-combine
3503 Dump after instruction combination, to the file @file{@var{file}.17.combine}.
3506 @itemx -fdump-rtl-ce1
3507 @itemx -fdump-rtl-ce2
3509 @opindex fdump-rtl-ce1
3510 @opindex fdump-rtl-ce2
3511 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3512 first if conversion, to the file @file{@var{file}.11.ce1}. @option{-dC}
3513 and @option{-fdump-rtl-ce2} enable dumping after the second if
3514 conversion, to the file @file{@var{file}.18.ce2}.
3517 @itemx -fdump-rtl-btl
3518 @itemx -fdump-rtl-dbr
3520 @opindex fdump-rtl-btl
3521 @opindex fdump-rtl-dbr
3522 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3523 target load optimization, to to @file{@var{file}.31.btl}. @option{-dd}
3524 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3525 scheduling, to @file{@var{file}.36.dbr}.
3529 Dump all macro definitions, at the end of preprocessing, in addition to
3533 @itemx -fdump-rtl-ce3
3535 @opindex fdump-rtl-ce3
3536 Dump after the third if conversion, to @file{@var{file}.28.ce3}.
3539 @itemx -fdump-rtl-cfg
3540 @itemx -fdump-rtl-life
3542 @opindex fdump-rtl-cfg
3543 @opindex fdump-rtl-life
3544 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3545 and data flow analysis, to @file{@var{file}.08.cfg}. @option{-df}
3546 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3547 to @file{@var{file}.16.life}.
3550 @itemx -fdump-rtl-greg
3552 @opindex fdump-rtl-greg
3553 Dump after global register allocation, to @file{@var{file}.23.greg}.
3556 @itemx -fdump-rtl-gcse
3557 @itemx -fdump-rtl-bypass
3559 @opindex fdump-rtl-gcse
3560 @opindex fdump-rtl-bypass
3561 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3562 @file{@var{file}.05.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3563 enable dumping after jump bypassing and control flow optimizations, to
3564 @file{@var{file}.07.bypass}.
3567 @itemx -fdump-rtl-eh
3569 @opindex fdump-rtl-eh
3570 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3573 @itemx -fdump-rtl-sibling
3575 @opindex fdump-rtl-sibling
3576 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
3579 @itemx -fdump-rtl-jump
3581 @opindex fdump-rtl-jump
3582 Dump after the first jump optimization, to @file{@var{file}.03.jump}.
3585 @itemx -fdump-rtl-stack
3587 @opindex fdump-rtl-stack
3588 Dump after conversion from registers to stack, to @file{@var{file}.33.stack}.
3591 @itemx -fdump-rtl-lreg
3593 @opindex fdump-rtl-lreg
3594 Dump after local register allocation, to @file{@var{file}.22.lreg}.
3597 @itemx -fdump-rtl-loop
3598 @itemx -fdump-rtl-loop2
3600 @opindex fdump-rtl-loop
3601 @opindex fdump-rtl-loop2
3602 @option{-dL} and @option{-fdump-rtl-loop} enable dumping after the first
3603 loop optimization pass, to @file{@var{file}.06.loop}. @option{-dL} and
3604 @option{-fdump-rtl-loop2} enable dumping after the second pass, to
3605 @file{@var{file}.13.loop2}.
3608 @itemx -fdump-rtl-sms
3610 @opindex fdump-rtl-sms
3611 Dump after modulo scheduling, to @file{@var{file}.20.sms}.
3614 @itemx -fdump-rtl-mach
3616 @opindex fdump-rtl-mach
3617 Dump after performing the machine dependent reorganization pass, to
3618 @file{@var{file}.35.mach}.
3621 @itemx -fdump-rtl-rnreg
3623 @opindex fdump-rtl-rnreg
3624 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3627 @itemx -fdump-rtl-regmove
3629 @opindex fdump-rtl-regmove
3630 Dump after the register move pass, to @file{@var{file}.19.regmove}.
3633 @itemx -fdump-rtl-postreload
3635 @opindex fdump-rtl-postreload
3636 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3639 @itemx -fdump-rtl-expand
3641 @opindex fdump-rtl-expand
3642 Dump after RTL generation, to @file{@var{file}.00.expand}.
3645 @itemx -fdump-rtl-sched2
3647 @opindex fdump-rtl-sched2
3648 Dump after the second scheduling pass, to @file{@var{file}.32.sched2}.
3651 @itemx -fdump-rtl-cse
3653 @opindex fdump-rtl-cse
3654 Dump after CSE (including the jump optimization that sometimes follows
3655 CSE), to @file{@var{file}.04.cse}.
3658 @itemx -fdump-rtl-sched
3660 @opindex fdump-rtl-sched
3661 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3664 @itemx -fdump-rtl-cse2
3666 @opindex fdump-rtl-cse2
3667 Dump after the second CSE pass (including the jump optimization that
3668 sometimes follows CSE), to @file{@var{file}.15.cse2}.
3671 @itemx -fdump-rtl-tracer
3673 @opindex fdump-rtl-tracer
3674 Dump after running tracer, to @file{@var{file}.12.tracer}.
3677 @itemx -fdump-rtl-vpt
3678 @itemx -fdump-rtl-vartrack
3680 @opindex fdump-rtl-vpt
3681 @opindex fdump-rtl-vartrack
3682 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3683 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3684 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3685 to @file{@var{file}.34.vartrack}.
3688 @itemx -fdump-rtl-flow2
3690 @opindex fdump-rtl-flow2
3691 Dump after the second flow pass, to @file{@var{file}.26.flow2}.
3694 @itemx -fdump-rtl-peephole2
3696 @opindex fdump-rtl-peephole2
3697 Dump after the peephole pass, to @file{@var{file}.27.peephole2}.
3700 @itemx -fdump-rtl-web
3702 @opindex fdump-rtl-web
3703 Dump after live range splitting, to @file{@var{file}.14.web}.
3706 @itemx -fdump-rtl-all
3708 @opindex fdump-rtl-all
3709 Produce all the dumps listed above.
3713 Produce a core dump whenever an error occurs.
3717 Print statistics on memory usage, at the end of the run, to
3722 Annotate the assembler output with a comment indicating which
3723 pattern and alternative was used. The length of each instruction is
3728 Dump the RTL in the assembler output as a comment before each instruction.
3729 Also turns on @option{-dp} annotation.
3733 For each of the other indicated dump files (either with @option{-d} or
3734 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3735 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3739 Just generate RTL for a function instead of compiling it. Usually used
3740 with @samp{r} (@option{-fdump-rtl-expand}).
3744 Dump debugging information during parsing, to standard error.
3747 @item -fdump-unnumbered
3748 @opindex fdump-unnumbered
3749 When doing debugging dumps (see @option{-d} option above), suppress instruction
3750 numbers and line number note output. This makes it more feasible to
3751 use diff on debugging dumps for compiler invocations with different
3752 options, in particular with and without @option{-g}.
3754 @item -fdump-translation-unit @r{(C and C++ only)}
3755 @itemx -fdump-translation-unit-@var{options} @r{(C and C++ only)}
3756 @opindex fdump-translation-unit
3757 Dump a representation of the tree structure for the entire translation
3758 unit to a file. The file name is made by appending @file{.tu} to the
3759 source file name. If the @samp{-@var{options}} form is used, @var{options}
3760 controls the details of the dump as described for the
3761 @option{-fdump-tree} options.
3763 @item -fdump-class-hierarchy @r{(C++ only)}
3764 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3765 @opindex fdump-class-hierarchy
3766 Dump a representation of each class's hierarchy and virtual function
3767 table layout to a file. The file name is made by appending @file{.class}
3768 to the source file name. If the @samp{-@var{options}} form is used,
3769 @var{options} controls the details of the dump as described for the
3770 @option{-fdump-tree} options.
3772 @item -fdump-ipa-@var{switch}
3774 Control the dumping at various stages of inter-procedural analysis
3775 language tree to a file. The file name is generated by appending a switch
3776 specific suffix to the source file name. The following dumps are possible:
3780 Enables all inter-procedural analysis dumps; currently the only produced
3781 dump is the @samp{cgraph} dump.
3784 Dumps information about call-graph optimization, unused function removal,
3785 and inlining decisions.
3788 @item -fdump-tree-@var{switch} @r{(C and C++ only)}
3789 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C and C++ only)}
3791 Control the dumping at various stages of processing the intermediate
3792 language tree to a file. The file name is generated by appending a switch
3793 specific suffix to the source file name. If the @samp{-@var{options}}
3794 form is used, @var{options} is a list of @samp{-} separated options that
3795 control the details of the dump. Not all options are applicable to all
3796 dumps, those which are not meaningful will be ignored. The following
3797 options are available
3801 Print the address of each node. Usually this is not meaningful as it
3802 changes according to the environment and source file. Its primary use
3803 is for tying up a dump file with a debug environment.
3805 Inhibit dumping of members of a scope or body of a function merely
3806 because that scope has been reached. Only dump such items when they
3807 are directly reachable by some other path. When dumping pretty-printed
3808 trees, this option inhibits dumping the bodies of control structures.
3810 Print a raw representation of the tree. By default, trees are
3811 pretty-printed into a C-like representation.
3813 Enable more detailed dumps (not honored by every dump option).
3815 Enable dumping various statistics about the pass (not honored by every dump
3818 Enable showing basic block boundaries (disabled in raw dumps).
3820 Enable showing virtual operands for every statement.
3822 Enable showing line numbers for statements.
3824 Enable showing the unique ID (@code{DECL_UID}) for each variable.
3826 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
3829 The following tree dumps are possible:
3833 Dump before any tree based optimization, to @file{@var{file}.original}.
3836 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3839 Dump after function inlining, to @file{@var{file}.inlined}.
3842 @opindex fdump-tree-gimple
3843 Dump each function before and after the gimplification pass to a file. The
3844 file name is made by appending @file{.gimple} to the source file name.
3847 @opindex fdump-tree-cfg
3848 Dump the control flow graph of each function to a file. The file name is
3849 made by appending @file{.cfg} to the source file name.
3852 @opindex fdump-tree-vcg
3853 Dump the control flow graph of each function to a file in VCG format. The
3854 file name is made by appending @file{.vcg} to the source file name. Note
3855 that if the file contains more than one function, the generated file cannot
3856 be used directly by VCG@. You will need to cut and paste each function's
3857 graph into its own separate file first.
3860 @opindex fdump-tree-ch
3861 Dump each function after copying loop headers. The file name is made by
3862 appending @file{.ch} to the source file name.
3865 @opindex fdump-tree-ssa
3866 Dump SSA related information to a file. The file name is made by appending
3867 @file{.ssa} to the source file name.
3870 @opindex fdump-tree-salias
3871 Dump structure aliasing variable information to a file. This file name
3872 is made by appending @file{.salias} to the source file name.
3875 @opindex fdump-tree-alias
3876 Dump aliasing information for each function. The file name is made by
3877 appending @file{.alias} to the source file name.
3880 @opindex fdump-tree-ccp
3881 Dump each function after CCP@. The file name is made by appending
3882 @file{.ccp} to the source file name.
3885 @opindex fdump-tree-pre
3886 Dump trees after partial redundancy elimination. The file name is made
3887 by appending @file{.pre} to the source file name.
3890 @opindex fdump-tree-fre
3891 Dump trees after full redundancy elimination. The file name is made
3892 by appending @file{.fre} to the source file name.
3895 @opindex fdump-tree-dce
3896 Dump each function after dead code elimination. The file name is made by
3897 appending @file{.dce} to the source file name.
3900 @opindex fdump-tree-mudflap
3901 Dump each function after adding mudflap instrumentation. The file name is
3902 made by appending @file{.mudflap} to the source file name.
3905 @opindex fdump-tree-sra
3906 Dump each function after performing scalar replacement of aggregates. The
3907 file name is made by appending @file{.sra} to the source file name.
3910 @opindex fdump-tree-sink
3911 Dump each function after performing code sinking. The file name is made
3912 by appending @file{.sink} to the source file name.
3915 @opindex fdump-tree-dom
3916 Dump each function after applying dominator tree optimizations. The file
3917 name is made by appending @file{.dom} to the source file name.
3920 @opindex fdump-tree-dse
3921 Dump each function after applying dead store elimination. The file
3922 name is made by appending @file{.dse} to the source file name.
3925 @opindex fdump-tree-phiopt
3926 Dump each function after optimizing PHI nodes into straightline code. The file
3927 name is made by appending @file{.phiopt} to the source file name.
3930 @opindex fdump-tree-forwprop
3931 Dump each function after forward propagating single use variables. The file
3932 name is made by appending @file{.forwprop} to the source file name.
3935 @opindex fdump-tree-copyrename
3936 Dump each function after applying the copy rename optimization. The file
3937 name is made by appending @file{.copyrename} to the source file name.
3940 @opindex fdump-tree-nrv
3941 Dump each function after applying the named return value optimization on
3942 generic trees. The file name is made by appending @file{.nrv} to the source
3946 @opindex fdump-tree-vect
3947 Dump each function after applying vectorization of loops. The file name is
3948 made by appending @file{.vect} to the source file name.
3951 @opindex fdump-tree-all
3952 Enable all the available tree dumps with the flags provided in this option.
3955 @item -ftree-vectorizer-verbose=@var{n}
3956 @opindex ftree-vectorizer-verbose
3957 This option controls the amount of debugging output the vectorizer prints.
3958 This information is written to standard error, unless @option{-fdump-tree-all}
3959 or @option{-fdump-tree-vect} is specified, in which case it is output to the
3960 usual dump listing file, @file{.vect}.
3962 @item -frandom-seed=@var{string}
3963 @opindex frandom-string
3964 This option provides a seed that GCC uses when it would otherwise use
3965 random numbers. It is used to generate certain symbol names
3966 that have to be different in every compiled file. It is also used to
3967 place unique stamps in coverage data files and the object files that
3968 produce them. You can use the @option{-frandom-seed} option to produce
3969 reproducibly identical object files.
3971 The @var{string} should be different for every file you compile.
3973 @item -fsched-verbose=@var{n}
3974 @opindex fsched-verbose
3975 On targets that use instruction scheduling, this option controls the
3976 amount of debugging output the scheduler prints. This information is
3977 written to standard error, unless @option{-dS} or @option{-dR} is
3978 specified, in which case it is output to the usual dump
3979 listing file, @file{.sched} or @file{.sched2} respectively. However
3980 for @var{n} greater than nine, the output is always printed to standard
3983 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
3984 same information as @option{-dRS}. For @var{n} greater than one, it
3985 also output basic block probabilities, detailed ready list information
3986 and unit/insn info. For @var{n} greater than two, it includes RTL
3987 at abort point, control-flow and regions info. And for @var{n} over
3988 four, @option{-fsched-verbose} also includes dependence info.
3992 Store the usual ``temporary'' intermediate files permanently; place them
3993 in the current directory and name them based on the source file. Thus,
3994 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
3995 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
3996 preprocessed @file{foo.i} output file even though the compiler now
3997 normally uses an integrated preprocessor.
3999 When used in combination with the @option{-x} command line option,
4000 @option{-save-temps} is sensible enough to avoid over writing an
4001 input source file with the same extension as an intermediate file.
4002 The corresponding intermediate file may be obtained by renaming the
4003 source file before using @option{-save-temps}.
4007 Report the CPU time taken by each subprocess in the compilation
4008 sequence. For C source files, this is the compiler proper and assembler
4009 (plus the linker if linking is done). The output looks like this:
4016 The first number on each line is the ``user time'', that is time spent
4017 executing the program itself. The second number is ``system time'',
4018 time spent executing operating system routines on behalf of the program.
4019 Both numbers are in seconds.
4021 @item -fvar-tracking
4022 @opindex fvar-tracking
4023 Run variable tracking pass. It computes where variables are stored at each
4024 position in code. Better debugging information is then generated
4025 (if the debugging information format supports this information).
4027 It is enabled by default when compiling with optimization (@option{-Os},
4028 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4029 the debug info format supports it.
4031 @item -print-file-name=@var{library}
4032 @opindex print-file-name
4033 Print the full absolute name of the library file @var{library} that
4034 would be used when linking---and don't do anything else. With this
4035 option, GCC does not compile or link anything; it just prints the
4038 @item -print-multi-directory
4039 @opindex print-multi-directory
4040 Print the directory name corresponding to the multilib selected by any
4041 other switches present in the command line. This directory is supposed
4042 to exist in @env{GCC_EXEC_PREFIX}.
4044 @item -print-multi-lib
4045 @opindex print-multi-lib
4046 Print the mapping from multilib directory names to compiler switches
4047 that enable them. The directory name is separated from the switches by
4048 @samp{;}, and each switch starts with an @samp{@@} instead of the
4049 @samp{-}, without spaces between multiple switches. This is supposed to
4050 ease shell-processing.
4052 @item -print-prog-name=@var{program}
4053 @opindex print-prog-name
4054 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4056 @item -print-libgcc-file-name
4057 @opindex print-libgcc-file-name
4058 Same as @option{-print-file-name=libgcc.a}.
4060 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4061 but you do want to link with @file{libgcc.a}. You can do
4064 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4067 @item -print-search-dirs
4068 @opindex print-search-dirs
4069 Print the name of the configured installation directory and a list of
4070 program and library directories @command{gcc} will search---and don't do anything else.
4072 This is useful when @command{gcc} prints the error message
4073 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4074 To resolve this you either need to put @file{cpp0} and the other compiler
4075 components where @command{gcc} expects to find them, or you can set the environment
4076 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4077 Don't forget the trailing @samp{/}.
4078 @xref{Environment Variables}.
4081 @opindex dumpmachine
4082 Print the compiler's target machine (for example,
4083 @samp{i686-pc-linux-gnu})---and don't do anything else.
4086 @opindex dumpversion
4087 Print the compiler version (for example, @samp{3.0})---and don't do
4092 Print the compiler's built-in specs---and don't do anything else. (This
4093 is used when GCC itself is being built.) @xref{Spec Files}.
4095 @item -feliminate-unused-debug-types
4096 @opindex feliminate-unused-debug-types
4097 Normally, when producing DWARF2 output, GCC will emit debugging
4098 information for all types declared in a compilation
4099 unit, regardless of whether or not they are actually used
4100 in that compilation unit. Sometimes this is useful, such as
4101 if, in the debugger, you want to cast a value to a type that is
4102 not actually used in your program (but is declared). More often,
4103 however, this results in a significant amount of wasted space.
4104 With this option, GCC will avoid producing debug symbol output
4105 for types that are nowhere used in the source file being compiled.
4108 @node Optimize Options
4109 @section Options That Control Optimization
4110 @cindex optimize options
4111 @cindex options, optimization
4113 These options control various sorts of optimizations.
4115 Without any optimization option, the compiler's goal is to reduce the
4116 cost of compilation and to make debugging produce the expected
4117 results. Statements are independent: if you stop the program with a
4118 breakpoint between statements, you can then assign a new value to any
4119 variable or change the program counter to any other statement in the
4120 function and get exactly the results you would expect from the source
4123 Turning on optimization flags makes the compiler attempt to improve
4124 the performance and/or code size at the expense of compilation time
4125 and possibly the ability to debug the program.
4127 The compiler performs optimization based on the knowledge it has of
4128 the program. Optimization levels @option{-O2} and above, in
4129 particular, enable @emph{unit-at-a-time} mode, which allows the
4130 compiler to consider information gained from later functions in
4131 the file when compiling a function. Compiling multiple files at
4132 once to a single output file in @emph{unit-at-a-time} mode allows
4133 the compiler to use information gained from all of the files when
4134 compiling each of them.
4136 Not all optimizations are controlled directly by a flag. Only
4137 optimizations that have a flag are listed.
4144 Optimize. Optimizing compilation takes somewhat more time, and a lot
4145 more memory for a large function.
4147 With @option{-O}, the compiler tries to reduce code size and execution
4148 time, without performing any optimizations that take a great deal of
4151 @option{-O} turns on the following optimization flags:
4152 @gccoptlist{-fdefer-pop @gol
4153 -fdelayed-branch @gol
4154 -fguess-branch-probability @gol
4155 -fcprop-registers @gol
4156 -floop-optimize @gol
4157 -fif-conversion @gol
4158 -fif-conversion2 @gol
4164 -ftree-live_range_split @gol
4166 -ftree-copyrename @gol
4171 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4172 where doing so does not interfere with debugging.
4176 Optimize even more. GCC performs nearly all supported optimizations
4177 that do not involve a space-speed tradeoff. The compiler does not
4178 perform loop unrolling or function inlining when you specify @option{-O2}.
4179 As compared to @option{-O}, this option increases both compilation time
4180 and the performance of the generated code.
4182 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4183 also turns on the following optimization flags:
4184 @gccoptlist{-fthread-jumps @gol
4186 -foptimize-sibling-calls @gol
4187 -fcse-follow-jumps -fcse-skip-blocks @gol
4188 -fgcse -fgcse-lm @gol
4189 -fexpensive-optimizations @gol
4190 -fstrength-reduce @gol
4191 -frerun-cse-after-loop -frerun-loop-opt @gol
4195 -fschedule-insns -fschedule-insns2 @gol
4196 -fsched-interblock -fsched-spec @gol
4198 -fstrict-aliasing @gol
4199 -fdelete-null-pointer-checks @gol
4200 -freorder-blocks -freorder-functions @gol
4201 -funit-at-a-time @gol
4202 -falign-functions -falign-jumps @gol
4203 -falign-loops -falign-labels @gol
4206 Please note the warning under @option{-fgcse} about
4207 invoking @option{-O2} on programs that use computed gotos.
4211 Optimize yet more. @option{-O3} turns on all optimizations specified by
4212 @option{-O2} and also turns on the @option{-finline-functions},
4213 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4217 Do not optimize. This is the default.
4221 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4222 do not typically increase code size. It also performs further
4223 optimizations designed to reduce code size.
4225 @option{-Os} disables the following optimization flags:
4226 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4227 -falign-labels -freorder-blocks -freorder-blocks-and-partition -fprefetch-loop-arrays}
4229 If you use multiple @option{-O} options, with or without level numbers,
4230 the last such option is the one that is effective.
4233 Options of the form @option{-f@var{flag}} specify machine-independent
4234 flags. Most flags have both positive and negative forms; the negative
4235 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4236 below, only one of the forms is listed---the one you typically will
4237 use. You can figure out the other form by either removing @samp{no-}
4240 The following options control specific optimizations. They are either
4241 activated by @option{-O} options or are related to ones that are. You
4242 can use the following flags in the rare cases when ``fine-tuning'' of
4243 optimizations to be performed is desired.
4246 @item -fno-default-inline
4247 @opindex fno-default-inline
4248 Do not make member functions inline by default merely because they are
4249 defined inside the class scope (C++ only). Otherwise, when you specify
4250 @w{@option{-O}}, member functions defined inside class scope are compiled
4251 inline by default; i.e., you don't need to add @samp{inline} in front of
4252 the member function name.
4254 @item -fno-defer-pop
4255 @opindex fno-defer-pop
4256 Always pop the arguments to each function call as soon as that function
4257 returns. For machines which must pop arguments after a function call,
4258 the compiler normally lets arguments accumulate on the stack for several
4259 function calls and pops them all at once.
4261 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4265 Force memory operands to be copied into registers before doing
4266 arithmetic on them. This produces better code by making all memory
4267 references potential common subexpressions. When they are not common
4268 subexpressions, instruction combination should eliminate the separate
4271 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4274 @opindex fforce-addr
4275 Force memory address constants to be copied into registers before
4276 doing arithmetic on them. This may produce better code just as
4277 @option{-fforce-mem} may.
4279 @item -fomit-frame-pointer
4280 @opindex fomit-frame-pointer
4281 Don't keep the frame pointer in a register for functions that
4282 don't need one. This avoids the instructions to save, set up and
4283 restore frame pointers; it also makes an extra register available
4284 in many functions. @strong{It also makes debugging impossible on
4287 On some machines, such as the VAX, this flag has no effect, because
4288 the standard calling sequence automatically handles the frame pointer
4289 and nothing is saved by pretending it doesn't exist. The
4290 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4291 whether a target machine supports this flag. @xref{Registers,,Register
4292 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4294 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4296 @item -foptimize-sibling-calls
4297 @opindex foptimize-sibling-calls
4298 Optimize sibling and tail recursive calls.
4300 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4304 Don't pay attention to the @code{inline} keyword. Normally this option
4305 is used to keep the compiler from expanding any functions inline.
4306 Note that if you are not optimizing, no functions can be expanded inline.
4308 @item -finline-functions
4309 @opindex finline-functions
4310 Integrate all simple functions into their callers. The compiler
4311 heuristically decides which functions are simple enough to be worth
4312 integrating in this way.
4314 If all calls to a given function are integrated, and the function is
4315 declared @code{static}, then the function is normally not output as
4316 assembler code in its own right.
4318 Enabled at level @option{-O3}.
4320 @item -finline-limit=@var{n}
4321 @opindex finline-limit
4322 By default, GCC limits the size of functions that can be inlined. This flag
4323 allows the control of this limit for functions that are explicitly marked as
4324 inline (i.e., marked with the inline keyword or defined within the class
4325 definition in c++). @var{n} is the size of functions that can be inlined in
4326 number of pseudo instructions (not counting parameter handling). The default
4327 value of @var{n} is 600.
4328 Increasing this value can result in more inlined code at
4329 the cost of compilation time and memory consumption. Decreasing usually makes
4330 the compilation faster and less code will be inlined (which presumably
4331 means slower programs). This option is particularly useful for programs that
4332 use inlining heavily such as those based on recursive templates with C++.
4334 Inlining is actually controlled by a number of parameters, which may be
4335 specified individually by using @option{--param @var{name}=@var{value}}.
4336 The @option{-finline-limit=@var{n}} option sets some of these parameters
4340 @item max-inline-insns-single
4341 is set to @var{n}/2.
4342 @item max-inline-insns-auto
4343 is set to @var{n}/2.
4344 @item min-inline-insns
4345 is set to 130 or @var{n}/4, whichever is smaller.
4346 @item max-inline-insns-rtl
4350 See below for a documentation of the individual
4351 parameters controlling inlining.
4353 @emph{Note:} pseudo instruction represents, in this particular context, an
4354 abstract measurement of function's size. In no way, it represents a count
4355 of assembly instructions and as such its exact meaning might change from one
4356 release to an another.
4358 @item -fkeep-inline-functions
4359 @opindex fkeep-inline-functions
4360 In C, emit @code{static} functions that are declared @code{inline}
4361 into the object file, even if the function has been inlined into all
4362 of its callers. This switch does not affect functions using the
4363 @code{extern inline} extension in GNU C@. In C++, emit any and all
4364 inline functions into the object file.
4366 @item -fkeep-static-consts
4367 @opindex fkeep-static-consts
4368 Emit variables declared @code{static const} when optimization isn't turned
4369 on, even if the variables aren't referenced.
4371 GCC enables this option by default. If you want to force the compiler to
4372 check if the variable was referenced, regardless of whether or not
4373 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4375 @item -fmerge-constants
4376 Attempt to merge identical constants (string constants and floating point
4377 constants) across compilation units.
4379 This option is the default for optimized compilation if the assembler and
4380 linker support it. Use @option{-fno-merge-constants} to inhibit this
4383 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4385 @item -fmerge-all-constants
4386 Attempt to merge identical constants and identical variables.
4388 This option implies @option{-fmerge-constants}. In addition to
4389 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4390 arrays or initialized constant variables with integral or floating point
4391 types. Languages like C or C++ require each non-automatic variable to
4392 have distinct location, so using this option will result in non-conforming
4395 @item -fmodulo-sched
4396 @opindex fmodulo-sched
4397 Perform swing modulo scheduling immediately before the first scheduling
4398 pass. This pass looks at innermost loops and reorders their
4399 instructions by overlapping different iterations.
4401 @item -fno-branch-count-reg
4402 @opindex fno-branch-count-reg
4403 Do not use ``decrement and branch'' instructions on a count register,
4404 but instead generate a sequence of instructions that decrement a
4405 register, compare it against zero, then branch based upon the result.
4406 This option is only meaningful on architectures that support such
4407 instructions, which include x86, PowerPC, IA-64 and S/390.
4409 The default is @option{-fbranch-count-reg}, enabled when
4410 @option{-fstrength-reduce} is enabled.
4412 @item -fno-function-cse
4413 @opindex fno-function-cse
4414 Do not put function addresses in registers; make each instruction that
4415 calls a constant function contain the function's address explicitly.
4417 This option results in less efficient code, but some strange hacks
4418 that alter the assembler output may be confused by the optimizations
4419 performed when this option is not used.
4421 The default is @option{-ffunction-cse}
4423 @item -fno-zero-initialized-in-bss
4424 @opindex fno-zero-initialized-in-bss
4425 If the target supports a BSS section, GCC by default puts variables that
4426 are initialized to zero into BSS@. This can save space in the resulting
4429 This option turns off this behavior because some programs explicitly
4430 rely on variables going to the data section. E.g., so that the
4431 resulting executable can find the beginning of that section and/or make
4432 assumptions based on that.
4434 The default is @option{-fzero-initialized-in-bss}.
4436 @item -fbounds-check
4437 @opindex fbounds-check
4438 For front-ends that support it, generate additional code to check that
4439 indices used to access arrays are within the declared range. This is
4440 currently only supported by the Java and Fortran front-ends, where
4441 this option defaults to true and false respectively.
4443 @item -fmudflap -fmudflapth -fmudflapir
4447 @cindex bounds checking
4449 For front-ends that support it (C and C++), instrument all risky
4450 pointer/array dereferencing operations, some standard library
4451 string/heap functions, and some other associated constructs with
4452 range/validity tests. Modules so instrumented should be immune to
4453 buffer overflows, invalid heap use, and some other classes of C/C++
4454 programming errors. The instrumentation relies on a separate runtime
4455 library (@file{libmudflap}), which will be linked into a program if
4456 @option{-fmudflap} is given at link time. Run-time behavior of the
4457 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4458 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4461 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4462 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4463 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4464 instrumentation should ignore pointer reads. This produces less
4465 instrumentation (and therefore faster execution) and still provides
4466 some protection against outright memory corrupting writes, but allows
4467 erroneously read data to propagate within a program.
4469 @item -fstrength-reduce
4470 @opindex fstrength-reduce
4471 Perform the optimizations of loop strength reduction and
4472 elimination of iteration variables.
4474 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4476 @item -fthread-jumps
4477 @opindex fthread-jumps
4478 Perform optimizations where we check to see if a jump branches to a
4479 location where another comparison subsumed by the first is found. If
4480 so, the first branch is redirected to either the destination of the
4481 second branch or a point immediately following it, depending on whether
4482 the condition is known to be true or false.
4484 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4486 @item -fcse-follow-jumps
4487 @opindex fcse-follow-jumps
4488 In common subexpression elimination, scan through jump instructions
4489 when the target of the jump is not reached by any other path. For
4490 example, when CSE encounters an @code{if} statement with an
4491 @code{else} clause, CSE will follow the jump when the condition
4494 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4496 @item -fcse-skip-blocks
4497 @opindex fcse-skip-blocks
4498 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4499 follow jumps which conditionally skip over blocks. When CSE
4500 encounters a simple @code{if} statement with no else clause,
4501 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4502 body of the @code{if}.
4504 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4506 @item -frerun-cse-after-loop
4507 @opindex frerun-cse-after-loop
4508 Re-run common subexpression elimination after loop optimizations has been
4511 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4513 @item -frerun-loop-opt
4514 @opindex frerun-loop-opt
4515 Run the loop optimizer twice.
4517 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4521 Perform a global common subexpression elimination pass.
4522 This pass also performs global constant and copy propagation.
4524 @emph{Note:} When compiling a program using computed gotos, a GCC
4525 extension, you may get better runtime performance if you disable
4526 the global common subexpression elimination pass by adding
4527 @option{-fno-gcse} to the command line.
4529 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4533 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4534 attempt to move loads which are only killed by stores into themselves. This
4535 allows a loop containing a load/store sequence to be changed to a load outside
4536 the loop, and a copy/store within the loop.
4538 Enabled by default when gcse is enabled.
4542 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4543 global common subexpression elimination. This pass will attempt to move
4544 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4545 loops containing a load/store sequence can be changed to a load before
4546 the loop and a store after the loop.
4548 Not enabled at any optimization level.
4552 When @option{-fgcse-las} is enabled, the global common subexpression
4553 elimination pass eliminates redundant loads that come after stores to the
4554 same memory location (both partial and full redundancies).
4556 Not enabled at any optimization level.
4558 @item -fgcse-after-reload
4559 @opindex fgcse-after-reload
4560 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4561 pass is performed after reload. The purpose of this pass is to cleanup
4564 @item -floop-optimize
4565 @opindex floop-optimize
4566 Perform loop optimizations: move constant expressions out of loops, simplify
4567 exit test conditions and optionally do strength-reduction as well.
4569 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4571 @item -floop-optimize2
4572 @opindex floop-optimize2
4573 Perform loop optimizations using the new loop optimizer. The optimizations
4574 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4577 @item -fcrossjumping
4578 @opindex crossjumping
4579 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4580 resulting code may or may not perform better than without cross-jumping.
4582 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4584 @item -fif-conversion
4585 @opindex if-conversion
4586 Attempt to transform conditional jumps into branch-less equivalents. This
4587 include use of conditional moves, min, max, set flags and abs instructions, and
4588 some tricks doable by standard arithmetics. The use of conditional execution
4589 on chips where it is available is controlled by @code{if-conversion2}.
4591 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4593 @item -fif-conversion2
4594 @opindex if-conversion2
4595 Use conditional execution (where available) to transform conditional jumps into
4596 branch-less equivalents.
4598 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4600 @item -fdelete-null-pointer-checks
4601 @opindex fdelete-null-pointer-checks
4602 Use global dataflow analysis to identify and eliminate useless checks
4603 for null pointers. The compiler assumes that dereferencing a null
4604 pointer would have halted the program. If a pointer is checked after
4605 it has already been dereferenced, it cannot be null.
4607 In some environments, this assumption is not true, and programs can
4608 safely dereference null pointers. Use
4609 @option{-fno-delete-null-pointer-checks} to disable this optimization
4610 for programs which depend on that behavior.
4612 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4614 @item -fexpensive-optimizations
4615 @opindex fexpensive-optimizations
4616 Perform a number of minor optimizations that are relatively expensive.
4618 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4620 @item -foptimize-register-move
4622 @opindex foptimize-register-move
4624 Attempt to reassign register numbers in move instructions and as
4625 operands of other simple instructions in order to maximize the amount of
4626 register tying. This is especially helpful on machines with two-operand
4629 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4632 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4634 @item -fdelayed-branch
4635 @opindex fdelayed-branch
4636 If supported for the target machine, attempt to reorder instructions
4637 to exploit instruction slots available after delayed branch
4640 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4642 @item -fschedule-insns
4643 @opindex fschedule-insns
4644 If supported for the target machine, attempt to reorder instructions to
4645 eliminate execution stalls due to required data being unavailable. This
4646 helps machines that have slow floating point or memory load instructions
4647 by allowing other instructions to be issued until the result of the load
4648 or floating point instruction is required.
4650 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4652 @item -fschedule-insns2
4653 @opindex fschedule-insns2
4654 Similar to @option{-fschedule-insns}, but requests an additional pass of
4655 instruction scheduling after register allocation has been done. This is
4656 especially useful on machines with a relatively small number of
4657 registers and where memory load instructions take more than one cycle.
4659 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4661 @item -fno-sched-interblock
4662 @opindex fno-sched-interblock
4663 Don't schedule instructions across basic blocks. This is normally
4664 enabled by default when scheduling before register allocation, i.e.@:
4665 with @option{-fschedule-insns} or at @option{-O2} or higher.
4667 @item -fno-sched-spec
4668 @opindex fno-sched-spec
4669 Don't allow speculative motion of non-load instructions. This is normally
4670 enabled by default when scheduling before register allocation, i.e.@:
4671 with @option{-fschedule-insns} or at @option{-O2} or higher.
4673 @item -fsched-spec-load
4674 @opindex fsched-spec-load
4675 Allow speculative motion of some load instructions. This only makes
4676 sense when scheduling before register allocation, i.e.@: with
4677 @option{-fschedule-insns} or at @option{-O2} or higher.
4679 @item -fsched-spec-load-dangerous
4680 @opindex fsched-spec-load-dangerous
4681 Allow speculative motion of more load instructions. This only makes
4682 sense when scheduling before register allocation, i.e.@: with
4683 @option{-fschedule-insns} or at @option{-O2} or higher.
4685 @item -fsched-stalled-insns=@var{n}
4686 @opindex fsched-stalled-insns
4687 Define how many insns (if any) can be moved prematurely from the queue
4688 of stalled insns into the ready list, during the second scheduling pass.
4690 @item -fsched-stalled-insns-dep=@var{n}
4691 @opindex fsched-stalled-insns-dep
4692 Define how many insn groups (cycles) will be examined for a dependency
4693 on a stalled insn that is candidate for premature removal from the queue
4694 of stalled insns. Has an effect only during the second scheduling pass,
4695 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4697 @item -fsched2-use-superblocks
4698 @opindex fsched2-use-superblocks
4699 When scheduling after register allocation, do use superblock scheduling
4700 algorithm. Superblock scheduling allows motion across basic block boundaries
4701 resulting on faster schedules. This option is experimental, as not all machine
4702 descriptions used by GCC model the CPU closely enough to avoid unreliable
4703 results from the algorithm.
4705 This only makes sense when scheduling after register allocation, i.e.@: with
4706 @option{-fschedule-insns2} or at @option{-O2} or higher.
4708 @item -fsched2-use-traces
4709 @opindex fsched2-use-traces
4710 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4711 allocation and additionally perform code duplication in order to increase the
4712 size of superblocks using tracer pass. See @option{-ftracer} for details on
4715 This mode should produce faster but significantly longer programs. Also
4716 without @option{-fbranch-probabilities} the traces constructed may not
4717 match the reality and hurt the performance. This only makes
4718 sense when scheduling after register allocation, i.e.@: with
4719 @option{-fschedule-insns2} or at @option{-O2} or higher.
4721 @item -freschedule-modulo-scheduled-loops
4722 @opindex fscheduling-in-modulo-scheduled-loops
4723 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4724 we may want to prevent the later scheduling passes from changing its schedule, we use this
4725 option to control that.
4727 @item -fcaller-saves
4728 @opindex fcaller-saves
4729 Enable values to be allocated in registers that will be clobbered by
4730 function calls, by emitting extra instructions to save and restore the
4731 registers around such calls. Such allocation is done only when it
4732 seems to result in better code than would otherwise be produced.
4734 This option is always enabled by default on certain machines, usually
4735 those which have no call-preserved registers to use instead.
4737 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4740 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4741 enabled by default at @option{-O2} and @option{-O3}.
4744 Perform Full Redundancy Elimination (FRE) on trees. The difference
4745 between FRE and PRE is that FRE only considers expressions
4746 that are computed on all paths leading to the redundant computation.
4747 This analysis faster than PRE, though it exposes fewer redundancies.
4748 This flag is enabled by default at @option{-O} and higher.
4751 Perform structural alias analysis on trees. This flag
4752 is enabled by default at @option{-O} and higher.
4755 Perform forward store motion on trees. This flag is
4756 enabled by default at @option{-O} and higher.
4759 Perform sparse conditional constant propagation (CCP) on trees. This flag
4760 is enabled by default at @option{-O} and higher.
4763 Perform dead code elimination (DCE) on trees. This flag is enabled by
4764 default at @option{-O} and higher.
4766 @item -ftree-dominator-opts
4767 Perform dead code elimination (DCE) on trees. This flag is enabled by
4768 default at @option{-O} and higher.
4771 Perform loop header copying on trees. This is beneficial since it increases
4772 effectiveness of code motion optimizations. It also saves one jump. This flag
4773 is enabled by default at @option{-O} and higher. It is not enabled
4774 for @option{-Os}, since it usually increases code size.
4776 @item -ftree-loop-optimize
4777 Perform loop optimizations on trees. This flag is enabled by default
4778 at @option{-O} and higher.
4780 @item -ftree-loop-linear
4781 Perform linear loop transformations on tree. This flag can improve cache
4782 performance and allow further loop optimizations to take place.
4785 Perform loop invariant motion on trees. This pass moves only invartiants that
4786 would be hard to handle on rtl level (function calls, operations that expand to
4787 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
4788 operands of conditions that are invariant out of the loop, so that we can use
4789 just trivial invariantness analysis in loop unswitching. The pass also includes
4793 Create a canonical counter for number of iterations in the loop for that
4794 determining number of iterations requires complicated analysis. Later
4795 optimizations then may determine the number easily. Useful especially
4796 in connection with unrolling.
4799 Perform induction variable optimizations (strength reduction, induction
4800 variable merging and induction variable elimination) on trees.
4803 Perform scalar replacement of aggregates. This pass replaces structure
4804 references with scalars to prevent committing structures to memory too
4805 early. This flag is enabled by default at @option{-O} and higher.
4807 @item -ftree-copyrename
4808 Perform copy renaming on trees. This pass attempts to rename compiler
4809 temporaries to other variables at copy locations, usually resulting in
4810 variable names which more closely resemble the original variables. This flag
4811 is enabled by default at @option{-O} and higher.
4814 Perform temporary expression replacement during the SSA->normal phase. Single
4815 use/single def temporaries are replaced at their use location with their
4816 defining expression. This results in non-GIMPLE code, but gives the expanders
4817 much more complex trees to work on resulting in better RTL generation. This is
4818 enabled by default at @option{-O} and higher.
4821 Perform live range splitting during the SSA->normal phase. Distinct live
4822 ranges of a variable are split into unique variables, allowing for better
4823 optimization later. This is enabled by default at @option{-O} and higher.
4825 @item -ftree-vectorize
4826 Perform loop vectorization on trees.
4830 Perform tail duplication to enlarge superblock size. This transformation
4831 simplifies the control flow of the function allowing other optimizations to do
4834 @item -funroll-loops
4835 @opindex funroll-loops
4836 Unroll loops whose number of iterations can be determined at compile
4837 time or upon entry to the loop. @option{-funroll-loops} implies both
4838 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4839 option makes code larger, and may or may not make it run faster.
4841 @item -funroll-all-loops
4842 @opindex funroll-all-loops
4843 Unroll all loops, even if their number of iterations is uncertain when
4844 the loop is entered. This usually makes programs run more slowly.
4845 @option{-funroll-all-loops} implies the same options as
4846 @option{-funroll-loops},
4848 @item -fsplit-ivs-in-unroller
4849 @opindex -fsplit-ivs-in-unroller
4850 Enables expressing of values of induction variables in later iterations
4851 of the unrolled loop using the value in the first iteration. This breaks
4852 long dependency chains, thus improving efficiency of the scheduling passes
4853 (for best results, @option{-fweb} should be used as well).
4855 Combination of @option{-fweb} and CSE is often sufficient to obtain the
4856 same effect. However in cases the loop body is more complicated than
4857 a single basic block, this is not reliable. It also does not work at all
4858 on some of the architectures due to restrictions in the CSE pass.
4860 This optimization is enabled by default.
4862 @item -fvariable-expansion-in-unroller
4863 @opindex -fvariable-expansion-in-unroller
4864 With this option, the compiler will create multiple copies of some
4865 local variables when unrolling a loop which can result in superior code.
4867 @item -fprefetch-loop-arrays
4868 @opindex fprefetch-loop-arrays
4869 If supported by the target machine, generate instructions to prefetch
4870 memory to improve the performance of loops that access large arrays.
4872 These options may generate better or worse code; results are highly
4873 dependent on the structure of loops within the source code.
4876 @itemx -fno-peephole2
4877 @opindex fno-peephole
4878 @opindex fno-peephole2
4879 Disable any machine-specific peephole optimizations. The difference
4880 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
4881 are implemented in the compiler; some targets use one, some use the
4882 other, a few use both.
4884 @option{-fpeephole} is enabled by default.
4885 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4887 @item -fno-guess-branch-probability
4888 @opindex fno-guess-branch-probability
4889 Do not guess branch probabilities using heuristics.
4891 GCC will use heuristics to guess branch probabilities if they are
4892 not provided by profiling feedback (@option{-fprofile-arcs}). These
4893 heuristics are based on the control flow graph. If some branch probabilities
4894 are specified by @samp{__builtin_expect}, then the heuristics will be
4895 used to guess branch probabilities for the rest of the control flow graph,
4896 taking the @samp{__builtin_expect} info into account. The interactions
4897 between the heuristics and @samp{__builtin_expect} can be complex, and in
4898 some cases, it may be useful to disable the heuristics so that the effects
4899 of @samp{__builtin_expect} are easier to understand.
4901 The default is @option{-fguess-branch-probability} at levels
4902 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4904 @item -freorder-blocks
4905 @opindex freorder-blocks
4906 Reorder basic blocks in the compiled function in order to reduce number of
4907 taken branches and improve code locality.
4909 Enabled at levels @option{-O2}, @option{-O3}.
4911 @item -freorder-blocks-and-partition
4912 @opindex freorder-blocks-and-partition
4913 In addition to reordering basic blocks in the compiled function, in order
4914 to reduce number of taken branches, partitions hot and cold basic blocks
4915 into separate sections of the assembly and .o files, to improve
4916 paging and cache locality performance.
4918 This optimization is automatically turned off in the presence of
4919 exception handling, for linkonce sections, for functions with a user-defined
4920 section attribute and on any architecture that does not support named
4923 @item -freorder-functions
4924 @opindex freorder-functions
4925 Reorder functions in the object file in order to
4926 improve code locality. This is implemented by using special
4927 subsections @code{.text.hot} for most frequently executed functions and
4928 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
4929 the linker so object file format must support named sections and linker must
4930 place them in a reasonable way.
4932 Also profile feedback must be available in to make this option effective. See
4933 @option{-fprofile-arcs} for details.
4935 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4937 @item -fstrict-aliasing
4938 @opindex fstrict-aliasing
4939 Allows the compiler to assume the strictest aliasing rules applicable to
4940 the language being compiled. For C (and C++), this activates
4941 optimizations based on the type of expressions. In particular, an
4942 object of one type is assumed never to reside at the same address as an
4943 object of a different type, unless the types are almost the same. For
4944 example, an @code{unsigned int} can alias an @code{int}, but not a
4945 @code{void*} or a @code{double}. A character type may alias any other
4948 Pay special attention to code like this:
4961 The practice of reading from a different union member than the one most
4962 recently written to (called ``type-punning'') is common. Even with
4963 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
4964 is accessed through the union type. So, the code above will work as
4965 expected. However, this code might not:
4976 Every language that wishes to perform language-specific alias analysis
4977 should define a function that computes, given an @code{tree}
4978 node, an alias set for the node. Nodes in different alias sets are not
4979 allowed to alias. For an example, see the C front-end function
4980 @code{c_get_alias_set}.
4982 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4984 @item -falign-functions
4985 @itemx -falign-functions=@var{n}
4986 @opindex falign-functions
4987 Align the start of functions to the next power-of-two greater than
4988 @var{n}, skipping up to @var{n} bytes. For instance,
4989 @option{-falign-functions=32} aligns functions to the next 32-byte
4990 boundary, but @option{-falign-functions=24} would align to the next
4991 32-byte boundary only if this can be done by skipping 23 bytes or less.
4993 @option{-fno-align-functions} and @option{-falign-functions=1} are
4994 equivalent and mean that functions will not be aligned.
4996 Some assemblers only support this flag when @var{n} is a power of two;
4997 in that case, it is rounded up.
4999 If @var{n} is not specified or is zero, use a machine-dependent default.
5001 Enabled at levels @option{-O2}, @option{-O3}.
5003 @item -falign-labels
5004 @itemx -falign-labels=@var{n}
5005 @opindex falign-labels
5006 Align all branch targets to a power-of-two boundary, skipping up to
5007 @var{n} bytes like @option{-falign-functions}. This option can easily
5008 make code slower, because it must insert dummy operations for when the
5009 branch target is reached in the usual flow of the code.
5011 @option{-fno-align-labels} and @option{-falign-labels=1} are
5012 equivalent and mean that labels will not be aligned.
5014 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5015 are greater than this value, then their values are used instead.
5017 If @var{n} is not specified or is zero, use a machine-dependent default
5018 which is very likely to be @samp{1}, meaning no alignment.
5020 Enabled at levels @option{-O2}, @option{-O3}.
5023 @itemx -falign-loops=@var{n}
5024 @opindex falign-loops
5025 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5026 like @option{-falign-functions}. The hope is that the loop will be
5027 executed many times, which will make up for any execution of the dummy
5030 @option{-fno-align-loops} and @option{-falign-loops=1} are
5031 equivalent and mean that loops will not be aligned.
5033 If @var{n} is not specified or is zero, use a machine-dependent default.
5035 Enabled at levels @option{-O2}, @option{-O3}.
5038 @itemx -falign-jumps=@var{n}
5039 @opindex falign-jumps
5040 Align branch targets to a power-of-two boundary, for branch targets
5041 where the targets can only be reached by jumping, skipping up to @var{n}
5042 bytes like @option{-falign-functions}. In this case, no dummy operations
5045 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5046 equivalent and mean that loops will not be aligned.
5048 If @var{n} is not specified or is zero, use a machine-dependent default.
5050 Enabled at levels @option{-O2}, @option{-O3}.
5052 @item -funit-at-a-time
5053 @opindex funit-at-a-time
5054 Parse the whole compilation unit before starting to produce code.
5055 This allows some extra optimizations to take place but consumes
5056 more memory (in general). There are some compatibility issues
5057 with @emph{unit-at-at-time} mode:
5060 enabling @emph{unit-at-a-time} mode may change the order
5061 in which functions, variables, and top-level @code{asm} statements
5062 are emitted, and will likely break code relying on some particular
5063 ordering. The majority of such top-level @code{asm} statements,
5064 though, can be replaced by @code{section} attributes.
5067 @emph{unit-at-a-time} mode removes unreferenced static variables
5068 and functions are removed. This may result in undefined references
5069 when an @code{asm} statement refers directly to variables or functions
5070 that are otherwise unused. In that case either the variable/function
5071 shall be listed as an operand of the @code{asm} statement operand or,
5072 in the case of top-level @code{asm} statements the attribute @code{used}
5073 shall be used on the declaration.
5076 Static functions now can use non-standard passing conventions that
5077 may break @code{asm} statements calling functions directly. Again,
5078 attribute @code{used} will prevent this behavior.
5081 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5082 but this scheme may not be supported by future releases of GCC@.
5084 Enabled at levels @option{-O2}, @option{-O3}.
5088 Constructs webs as commonly used for register allocation purposes and assign
5089 each web individual pseudo register. This allows the register allocation pass
5090 to operate on pseudos directly, but also strengthens several other optimization
5091 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5092 however, make debugging impossible, since variables will no longer stay in a
5095 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os},
5096 on targets where the default format for debugging information supports
5099 @item -fno-cprop-registers
5100 @opindex fno-cprop-registers
5101 After register allocation and post-register allocation instruction splitting,
5102 we perform a copy-propagation pass to try to reduce scheduling dependencies
5103 and occasionally eliminate the copy.
5105 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5107 @item -fprofile-generate
5108 @opindex fprofile-generate
5110 Enable options usually used for instrumenting application to produce
5111 profile useful for later recompilation with profile feedback based
5112 optimization. You must use @option{-fprofile-generate} both when
5113 compiling and when linking your program.
5115 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5118 @opindex fprofile-use
5119 Enable profile feedback directed optimizations, and optimizations
5120 generally profitable only with profile feedback available.
5122 The following options are enabled: @code{-fbranch-probabilities},
5123 @code{-fvpt}, @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}.
5127 The following options control compiler behavior regarding floating
5128 point arithmetic. These options trade off between speed and
5129 correctness. All must be specifically enabled.
5133 @opindex ffloat-store
5134 Do not store floating point variables in registers, and inhibit other
5135 options that might change whether a floating point value is taken from a
5138 @cindex floating point precision
5139 This option prevents undesirable excess precision on machines such as
5140 the 68000 where the floating registers (of the 68881) keep more
5141 precision than a @code{double} is supposed to have. Similarly for the
5142 x86 architecture. For most programs, the excess precision does only
5143 good, but a few programs rely on the precise definition of IEEE floating
5144 point. Use @option{-ffloat-store} for such programs, after modifying
5145 them to store all pertinent intermediate computations into variables.
5149 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5150 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5151 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5152 and @option{fcx-limited-range}.
5154 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5156 This option should never be turned on by any @option{-O} option since
5157 it can result in incorrect output for programs which depend on
5158 an exact implementation of IEEE or ISO rules/specifications for
5161 @item -fno-math-errno
5162 @opindex fno-math-errno
5163 Do not set ERRNO after calling math functions that are executed
5164 with a single instruction, e.g., sqrt. A program that relies on
5165 IEEE exceptions for math error handling may want to use this flag
5166 for speed while maintaining IEEE arithmetic compatibility.
5168 This option should never be turned on by any @option{-O} option since
5169 it can result in incorrect output for programs which depend on
5170 an exact implementation of IEEE or ISO rules/specifications for
5173 The default is @option{-fmath-errno}.
5175 @item -funsafe-math-optimizations
5176 @opindex funsafe-math-optimizations
5177 Allow optimizations for floating-point arithmetic that (a) assume
5178 that arguments and results are valid and (b) may violate IEEE or
5179 ANSI standards. When used at link-time, it may include libraries
5180 or startup files that change the default FPU control word or other
5181 similar optimizations.
5183 This option should never be turned on by any @option{-O} option since
5184 it can result in incorrect output for programs which depend on
5185 an exact implementation of IEEE or ISO rules/specifications for
5188 The default is @option{-fno-unsafe-math-optimizations}.
5190 @item -ffinite-math-only
5191 @opindex ffinite-math-only
5192 Allow optimizations for floating-point arithmetic that assume
5193 that arguments and results are not NaNs or +-Infs.
5195 This option should never be turned on by any @option{-O} option since
5196 it can result in incorrect output for programs which depend on
5197 an exact implementation of IEEE or ISO rules/specifications.
5199 The default is @option{-fno-finite-math-only}.
5201 @item -fno-trapping-math
5202 @opindex fno-trapping-math
5203 Compile code assuming that floating-point operations cannot generate
5204 user-visible traps. These traps include division by zero, overflow,
5205 underflow, inexact result and invalid operation. This option implies
5206 @option{-fno-signaling-nans}. Setting this option may allow faster
5207 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5209 This option should never be turned on by any @option{-O} option since
5210 it can result in incorrect output for programs which depend on
5211 an exact implementation of IEEE or ISO rules/specifications for
5214 The default is @option{-ftrapping-math}.
5216 @item -frounding-math
5217 @opindex frounding-math
5218 Disable transformations and optimizations that assume default floating
5219 point rounding behavior. This is round-to-zero for all floating point
5220 to integer conversions, and round-to-nearest for all other arithmetic
5221 truncations. This option should be specified for programs that change
5222 the FP rounding mode dynamically, or that may be executed with a
5223 non-default rounding mode. This option disables constant folding of
5224 floating point expressions at compile-time (which may be affected by
5225 rounding mode) and arithmetic transformations that are unsafe in the
5226 presence of sign-dependent rounding modes.
5228 The default is @option{-fno-rounding-math}.
5230 This option is experimental and does not currently guarantee to
5231 disable all GCC optimizations that are affected by rounding mode.
5232 Future versions of GCC may provide finer control of this setting
5233 using C99's @code{FENV_ACCESS} pragma. This command line option
5234 will be used to specify the default state for @code{FENV_ACCESS}.
5236 @item -fsignaling-nans
5237 @opindex fsignaling-nans
5238 Compile code assuming that IEEE signaling NaNs may generate user-visible
5239 traps during floating-point operations. Setting this option disables
5240 optimizations that may change the number of exceptions visible with
5241 signaling NaNs. This option implies @option{-ftrapping-math}.
5243 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5246 The default is @option{-fno-signaling-nans}.
5248 This option is experimental and does not currently guarantee to
5249 disable all GCC optimizations that affect signaling NaN behavior.
5251 @item -fsingle-precision-constant
5252 @opindex fsingle-precision-constant
5253 Treat floating point constant as single precision constant instead of
5254 implicitly converting it to double precision constant.
5256 @item -fcx-limited-range
5257 @itemx -fno-cx-limited-range
5258 @opindex fcx-limited-range
5259 @opindex fno-cx-limited-range
5260 When enabled, this option states that a range reduction step is not
5261 needed when performing complex division. The default is
5262 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5264 This option controls the default setting of the ISO C99
5265 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5270 The following options control optimizations that may improve
5271 performance, but are not enabled by any @option{-O} options. This
5272 section includes experimental options that may produce broken code.
5275 @item -fbranch-probabilities
5276 @opindex fbranch-probabilities
5277 After running a program compiled with @option{-fprofile-arcs}
5278 (@pxref{Debugging Options,, Options for Debugging Your Program or
5279 @command{gcc}}), you can compile it a second time using
5280 @option{-fbranch-probabilities}, to improve optimizations based on
5281 the number of times each branch was taken. When the program
5282 compiled with @option{-fprofile-arcs} exits it saves arc execution
5283 counts to a file called @file{@var{sourcename}.gcda} for each source
5284 file The information in this data file is very dependent on the
5285 structure of the generated code, so you must use the same source code
5286 and the same optimization options for both compilations.
5288 With @option{-fbranch-probabilities}, GCC puts a
5289 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5290 These can be used to improve optimization. Currently, they are only
5291 used in one place: in @file{reorg.c}, instead of guessing which path a
5292 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5293 exactly determine which path is taken more often.
5295 @item -fprofile-values
5296 @opindex fprofile-values
5297 If combined with @option{-fprofile-arcs}, it adds code so that some
5298 data about values of expressions in the program is gathered.
5300 With @option{-fbranch-probabilities}, it reads back the data gathered
5301 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5302 notes to instructions for their later usage in optimizations.
5304 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5308 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5309 a code to gather information about values of expressions.
5311 With @option{-fbranch-probabilities}, it reads back the data gathered
5312 and actually performs the optimizations based on them.
5313 Currently the optimizations include specialization of division operation
5314 using the knowledge about the value of the denominator.
5316 @item -fspeculative-prefetching
5317 @opindex fspeculative-prefetching
5318 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5319 a code to gather information about addresses of memory references in the
5322 With @option{-fbranch-probabilities}, it reads back the data gathered
5323 and issues prefetch instructions according to them. In addition to the opportunities
5324 noticed by @option{-fprefetch-loop-arrays}, it also notices more complicated
5325 memory access patterns---for example accesses to the data stored in linked
5326 list whose elements are usually allocated sequentially.
5328 In order to prevent issuing double prefetches, usage of
5329 @option{-fspeculative-prefetching} implies @option{-fno-prefetch-loop-arrays}.
5331 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5333 @item -frename-registers
5334 @opindex frename-registers
5335 Attempt to avoid false dependencies in scheduled code by making use
5336 of registers left over after register allocation. This optimization
5337 will most benefit processors with lots of registers. Depending on the
5338 debug information format adopted by the target, however, it can
5339 make debugging impossible, since variables will no longer stay in
5340 a ``home register''.
5342 Not enabled by default at any level because it has known bugs.
5346 Perform tail duplication to enlarge superblock size. This transformation
5347 simplifies the control flow of the function allowing other optimizations to do
5350 Enabled with @option{-fprofile-use}.
5352 @item -funroll-loops
5353 @opindex funroll-loops
5354 Unroll loops whose number of iterations can be determined at compile time or
5355 upon entry to the loop. @option{-funroll-loops} implies
5356 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
5357 (i.e.@: complete removal of loops with small constant number of iterations).
5358 This option makes code larger, and may or may not make it run faster.
5360 Enabled with @option{-fprofile-use}.
5362 @item -funroll-all-loops
5363 @opindex funroll-all-loops
5364 Unroll all loops, even if their number of iterations is uncertain when
5365 the loop is entered. This usually makes programs run more slowly.
5366 @option{-funroll-all-loops} implies the same options as
5367 @option{-funroll-loops}.
5370 @opindex fpeel-loops
5371 Peels the loops for that there is enough information that they do not
5372 roll much (from profile feedback). It also turns on complete loop peeling
5373 (i.e.@: complete removal of loops with small constant number of iterations).
5375 Enabled with @option{-fprofile-use}.
5377 @item -fmove-loop-invariants
5378 @opindex fmove-loop-invariants
5379 Enables the loop invariant motion pass in the new loop optimizer. Enabled
5380 at level @option{-O1}
5382 @item -funswitch-loops
5383 @opindex funswitch-loops
5384 Move branches with loop invariant conditions out of the loop, with duplicates
5385 of the loop on both branches (modified according to result of the condition).
5387 @item -fprefetch-loop-arrays
5388 @opindex fprefetch-loop-arrays
5389 If supported by the target machine, generate instructions to prefetch
5390 memory to improve the performance of loops that access large arrays.
5392 Disabled at level @option{-Os}.
5394 @item -ffunction-sections
5395 @itemx -fdata-sections
5396 @opindex ffunction-sections
5397 @opindex fdata-sections
5398 Place each function or data item into its own section in the output
5399 file if the target supports arbitrary sections. The name of the
5400 function or the name of the data item determines the section's name
5403 Use these options on systems where the linker can perform optimizations
5404 to improve locality of reference in the instruction space. Most systems
5405 using the ELF object format and SPARC processors running Solaris 2 have
5406 linkers with such optimizations. AIX may have these optimizations in
5409 Only use these options when there are significant benefits from doing
5410 so. When you specify these options, the assembler and linker will
5411 create larger object and executable files and will also be slower.
5412 You will not be able to use @code{gprof} on all systems if you
5413 specify this option and you may have problems with debugging if
5414 you specify both this option and @option{-g}.
5416 @item -fbranch-target-load-optimize
5417 @opindex fbranch-target-load-optimize
5418 Perform branch target register load optimization before prologue / epilogue
5420 The use of target registers can typically be exposed only during reload,
5421 thus hoisting loads out of loops and doing inter-block scheduling needs
5422 a separate optimization pass.
5424 @item -fbranch-target-load-optimize2
5425 @opindex fbranch-target-load-optimize2
5426 Perform branch target register load optimization after prologue / epilogue
5429 @item -fbtr-bb-exclusive
5430 @opindex fbtr-bb-exclusive
5431 When performing branch target register load optimization, don't reuse
5432 branch target registers in within any basic block.
5434 @item --param @var{name}=@var{value}
5436 In some places, GCC uses various constants to control the amount of
5437 optimization that is done. For example, GCC will not inline functions
5438 that contain more that a certain number of instructions. You can
5439 control some of these constants on the command-line using the
5440 @option{--param} option.
5442 The names of specific parameters, and the meaning of the values, are
5443 tied to the internals of the compiler, and are subject to change
5444 without notice in future releases.
5446 In each case, the @var{value} is an integer. The allowable choices for
5447 @var{name} are given in the following table:
5450 @item sra-max-structure-size
5451 The maximum structure size, in bytes, at which the scalar replacement
5452 of aggregates (SRA) optimization will perform block copies. The
5453 default value, 0, implies that GCC will select the most appropriate
5456 @item sra-field-structure-ratio
5457 The threshold ratio (as a percentage) between instantiated fields and
5458 the complete structure size. We say that if the ratio of the number
5459 of bytes in instantiated fields to the number of bytes in the complete
5460 structure exceeds this parameter, then block copies are not used. The
5463 @item max-crossjump-edges
5464 The maximum number of incoming edges to consider for crossjumping.
5465 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5466 the number of edges incoming to each block. Increasing values mean
5467 more aggressive optimization, making the compile time increase with
5468 probably small improvement in executable size.
5470 @item min-crossjump-insns
5471 The minimum number of instructions which must be matched at the end
5472 of two blocks before crossjumping will be performed on them. This
5473 value is ignored in the case where all instructions in the block being
5474 crossjumped from are matched. The default value is 5.
5476 @item max-goto-duplication-insns
5477 The maximum number of instructions to duplicate to a block that jumps
5478 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5479 passes, GCC factors computed gotos early in the compilation process,
5480 and unfactors them as late as possible. Only computed jumps at the
5481 end of a basic blocks with no more than max-goto-duplication-insns are
5482 unfactored. The default value is 8.
5484 @item max-delay-slot-insn-search
5485 The maximum number of instructions to consider when looking for an
5486 instruction to fill a delay slot. If more than this arbitrary number of
5487 instructions is searched, the time savings from filling the delay slot
5488 will be minimal so stop searching. Increasing values mean more
5489 aggressive optimization, making the compile time increase with probably
5490 small improvement in executable run time.
5492 @item max-delay-slot-live-search
5493 When trying to fill delay slots, the maximum number of instructions to
5494 consider when searching for a block with valid live register
5495 information. Increasing this arbitrarily chosen value means more
5496 aggressive optimization, increasing the compile time. This parameter
5497 should be removed when the delay slot code is rewritten to maintain the
5500 @item max-gcse-memory
5501 The approximate maximum amount of memory that will be allocated in
5502 order to perform the global common subexpression elimination
5503 optimization. If more memory than specified is required, the
5504 optimization will not be done.
5506 @item max-gcse-passes
5507 The maximum number of passes of GCSE to run. The default is 1.
5509 @item max-pending-list-length
5510 The maximum number of pending dependencies scheduling will allow
5511 before flushing the current state and starting over. Large functions
5512 with few branches or calls can create excessively large lists which
5513 needlessly consume memory and resources.
5515 @item max-inline-insns-single
5516 Several parameters control the tree inliner used in gcc.
5517 This number sets the maximum number of instructions (counted in GCC's
5518 internal representation) in a single function that the tree inliner
5519 will consider for inlining. This only affects functions declared
5520 inline and methods implemented in a class declaration (C++).
5521 The default value is 500.
5523 @item max-inline-insns-auto
5524 When you use @option{-finline-functions} (included in @option{-O3}),
5525 a lot of functions that would otherwise not be considered for inlining
5526 by the compiler will be investigated. To those functions, a different
5527 (more restrictive) limit compared to functions declared inline can
5529 The default value is 120.
5531 @item large-function-insns
5532 The limit specifying really large functions. For functions larger than this
5533 limit after inlining inlining is constrained by
5534 @option{--param large-function-growth}. This parameter is useful primarily
5535 to avoid extreme compilation time caused by non-linear algorithms used by the
5537 This parameter is ignored when @option{-funit-at-a-time} is not used.
5538 The default value is 3000.
5540 @item large-function-growth
5541 Specifies maximal growth of large function caused by inlining in percents.
5542 This parameter is ignored when @option{-funit-at-a-time} is not used.
5543 The default value is 100 which limits large function growth to 2.0 times
5546 @item inline-unit-growth
5547 Specifies maximal overall growth of the compilation unit caused by inlining.
5548 This parameter is ignored when @option{-funit-at-a-time} is not used.
5549 The default value is 50 which limits unit growth to 1.5 times the original
5552 @item max-inline-insns-recursive
5553 @itemx max-inline-insns-recursive-auto
5554 Specifies maximum number of instructions out-of-line copy of self recursive inline
5555 function can grow into by performing recursive inlining.
5557 For functions declared inline @option{--param max-inline-insns-recursive} is
5558 taken into acount. For function not declared inline, recursive inlining
5559 happens only when @option{-finline-functions} (included in @option{-O3}) is
5560 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5561 default value is 500.
5563 @item max-inline-recursive-depth
5564 @itemx max-inline-recursive-depth-auto
5565 Specifies maximum recursion depth used by the recursive inlining.
5567 For functions declared inline @option{--param max-inline-recursive-depth} is
5568 taken into acount. For function not declared inline, recursive inlining
5569 happens only when @option{-finline-functions} (included in @option{-O3}) is
5570 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5571 default value is 500.
5573 @item max-unrolled-insns
5574 The maximum number of instructions that a loop should have if that loop
5575 is unrolled, and if the loop is unrolled, it determines how many times
5576 the loop code is unrolled.
5578 @item max-average-unrolled-insns
5579 The maximum number of instructions biased by probabilities of their execution
5580 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5581 it determines how many times the loop code is unrolled.
5583 @item max-unroll-times
5584 The maximum number of unrollings of a single loop.
5586 @item max-peeled-insns
5587 The maximum number of instructions that a loop should have if that loop
5588 is peeled, and if the loop is peeled, it determines how many times
5589 the loop code is peeled.
5591 @item max-peel-times
5592 The maximum number of peelings of a single loop.
5594 @item max-completely-peeled-insns
5595 The maximum number of insns of a completely peeled loop.
5597 @item max-completely-peel-times
5598 The maximum number of iterations of a loop to be suitable for complete peeling.
5600 @item max-unswitch-insns
5601 The maximum number of insns of an unswitched loop.
5603 @item max-unswitch-level
5604 The maximum number of branches unswitched in a single loop.
5607 The minimum cost of an expensive expression in the loop invariant motion.
5609 @item iv-consider-all-candidates-bound
5610 Bound on number of candidates for induction variables below that
5611 all candidates are considered for each use in induction variable
5612 optimizations. Only the most relevant candidates are considered
5613 if there are more candidates, to avoid quadratic time complexity.
5615 @item iv-max-considered-uses
5616 The induction variable optimizations give up on loops that contain more
5617 induction variable uses.
5619 @item iv-always-prune-cand-set-bound
5620 If number of candidates in the set is smaller than this value,
5621 we always try to remove unnecessary ivs from the set during its
5622 optimization when a new iv is added to the set.
5624 @item max-iterations-to-track
5626 The maximum number of iterations of a loop the brute force algorithm
5627 for analysis of # of iterations of the loop tries to evaluate.
5629 @item hot-bb-count-fraction
5630 Select fraction of the maximal count of repetitions of basic block in program
5631 given basic block needs to have to be considered hot.
5633 @item hot-bb-frequency-fraction
5634 Select fraction of the maximal frequency of executions of basic block in
5635 function given basic block needs to have to be considered hot
5637 @item tracer-dynamic-coverage
5638 @itemx tracer-dynamic-coverage-feedback
5640 This value is used to limit superblock formation once the given percentage of
5641 executed instructions is covered. This limits unnecessary code size
5644 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5645 feedback is available. The real profiles (as opposed to statically estimated
5646 ones) are much less balanced allowing the threshold to be larger value.
5648 @item tracer-max-code-growth
5649 Stop tail duplication once code growth has reached given percentage. This is
5650 rather hokey argument, as most of the duplicates will be eliminated later in
5651 cross jumping, so it may be set to much higher values than is the desired code
5654 @item tracer-min-branch-ratio
5656 Stop reverse growth when the reverse probability of best edge is less than this
5657 threshold (in percent).
5659 @item tracer-min-branch-ratio
5660 @itemx tracer-min-branch-ratio-feedback
5662 Stop forward growth if the best edge do have probability lower than this
5665 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
5666 compilation for profile feedback and one for compilation without. The value
5667 for compilation with profile feedback needs to be more conservative (higher) in
5668 order to make tracer effective.
5670 @item max-cse-path-length
5672 Maximum number of basic blocks on path that cse considers. The default is 10.
5674 @item global-var-threshold
5676 Counts the number of function calls (@var{n}) and the number of
5677 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
5678 single artificial variable will be created to represent all the
5679 call-clobbered variables at function call sites. This artificial
5680 variable will then be made to alias every call-clobbered variable.
5681 (done as @code{int * size_t} on the host machine; beware overflow).
5683 @item max-aliased-vops
5685 Maximum number of virtual operands allowed to represent aliases
5686 before triggering the alias grouping heuristic. Alias grouping
5687 reduces compile times and memory consumption needed for aliasing at
5688 the expense of precision loss in alias information.
5690 @item ggc-min-expand
5692 GCC uses a garbage collector to manage its own memory allocation. This
5693 parameter specifies the minimum percentage by which the garbage
5694 collector's heap should be allowed to expand between collections.
5695 Tuning this may improve compilation speed; it has no effect on code
5698 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
5699 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
5700 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
5701 GCC is not able to calculate RAM on a particular platform, the lower
5702 bound of 30% is used. Setting this parameter and
5703 @option{ggc-min-heapsize} to zero causes a full collection to occur at
5704 every opportunity. This is extremely slow, but can be useful for
5707 @item ggc-min-heapsize
5709 Minimum size of the garbage collector's heap before it begins bothering
5710 to collect garbage. The first collection occurs after the heap expands
5711 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
5712 tuning this may improve compilation speed, and has no effect on code
5715 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
5716 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
5717 with a lower bound of 4096 (four megabytes) and an upper bound of
5718 131072 (128 megabytes). If GCC is not able to calculate RAM on a
5719 particular platform, the lower bound is used. Setting this parameter
5720 very large effectively disables garbage collection. Setting this
5721 parameter and @option{ggc-min-expand} to zero causes a full collection
5722 to occur at every opportunity.
5724 @item max-reload-search-insns
5725 The maximum number of instruction reload should look backward for equivalent
5726 register. Increasing values mean more aggressive optimization, making the
5727 compile time increase with probably slightly better performance. The default
5730 @item max-cselib-memory-location
5731 The maximum number of memory locations cselib should take into acount.
5732 Increasing values mean more aggressive optimization, making the compile time
5733 increase with probably slightly better performance. The default value is 500.
5735 @item reorder-blocks-duplicate
5736 @itemx reorder-blocks-duplicate-feedback
5738 Used by basic block reordering pass to decide whether to use unconditional
5739 branch or duplicate the code on its destination. Code is duplicated when its
5740 estimated size is smaller than this value multiplied by the estimated size of
5741 unconditional jump in the hot spots of the program.
5743 The @option{reorder-block-duplicate-feedback} is used only when profile
5744 feedback is available and may be set to higher values than
5745 @option{reorder-block-duplicate} since information about the hot spots is more
5748 @item max-sched-region-blocks
5749 The maximum number of blocks in a region to be considered for
5750 interblock scheduling. The default value is 10.
5752 @item max-sched-region-insns
5753 The maximum number of insns in a region to be considered for
5754 interblock scheduling. The default value is 100.
5756 @item max-last-value-rtl
5758 The maximum size measured as number of RTLs that can be recorded in an expression
5759 in combiner for a pseudo register as last known value of that register. The default
5762 @item integer-share-limit
5763 Small integer constants can use a shared data structure, reducing the
5764 compiler's memory usage and increasing its speed. This sets the maximum
5765 value of a shared integer constant's. The default value is 256.
5770 @node Preprocessor Options
5771 @section Options Controlling the Preprocessor
5772 @cindex preprocessor options
5773 @cindex options, preprocessor
5775 These options control the C preprocessor, which is run on each C source
5776 file before actual compilation.
5778 If you use the @option{-E} option, nothing is done except preprocessing.
5779 Some of these options make sense only together with @option{-E} because
5780 they cause the preprocessor output to be unsuitable for actual
5785 You can use @option{-Wp,@var{option}} to bypass the compiler driver
5786 and pass @var{option} directly through to the preprocessor. If
5787 @var{option} contains commas, it is split into multiple options at the
5788 commas. However, many options are modified, translated or interpreted
5789 by the compiler driver before being passed to the preprocessor, and
5790 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
5791 interface is undocumented and subject to change, so whenever possible
5792 you should avoid using @option{-Wp} and let the driver handle the
5795 @item -Xpreprocessor @var{option}
5796 @opindex preprocessor
5797 Pass @var{option} as an option to the preprocessor. You can use this to
5798 supply system-specific preprocessor options which GCC does not know how to
5801 If you want to pass an option that takes an argument, you must use
5802 @option{-Xpreprocessor} twice, once for the option and once for the argument.
5805 @include cppopts.texi
5807 @node Assembler Options
5808 @section Passing Options to the Assembler
5810 @c prevent bad page break with this line
5811 You can pass options to the assembler.
5814 @item -Wa,@var{option}
5816 Pass @var{option} as an option to the assembler. If @var{option}
5817 contains commas, it is split into multiple options at the commas.
5819 @item -Xassembler @var{option}
5821 Pass @var{option} as an option to the assembler. You can use this to
5822 supply system-specific assembler options which GCC does not know how to
5825 If you want to pass an option that takes an argument, you must use
5826 @option{-Xassembler} twice, once for the option and once for the argument.
5831 @section Options for Linking
5832 @cindex link options
5833 @cindex options, linking
5835 These options come into play when the compiler links object files into
5836 an executable output file. They are meaningless if the compiler is
5837 not doing a link step.
5841 @item @var{object-file-name}
5842 A file name that does not end in a special recognized suffix is
5843 considered to name an object file or library. (Object files are
5844 distinguished from libraries by the linker according to the file
5845 contents.) If linking is done, these object files are used as input
5854 If any of these options is used, then the linker is not run, and
5855 object file names should not be used as arguments. @xref{Overall
5859 @item -l@var{library}
5860 @itemx -l @var{library}
5862 Search the library named @var{library} when linking. (The second
5863 alternative with the library as a separate argument is only for
5864 POSIX compliance and is not recommended.)
5866 It makes a difference where in the command you write this option; the
5867 linker searches and processes libraries and object files in the order they
5868 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
5869 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
5870 to functions in @samp{z}, those functions may not be loaded.
5872 The linker searches a standard list of directories for the library,
5873 which is actually a file named @file{lib@var{library}.a}. The linker
5874 then uses this file as if it had been specified precisely by name.
5876 The directories searched include several standard system directories
5877 plus any that you specify with @option{-L}.
5879 Normally the files found this way are library files---archive files
5880 whose members are object files. The linker handles an archive file by
5881 scanning through it for members which define symbols that have so far
5882 been referenced but not defined. But if the file that is found is an
5883 ordinary object file, it is linked in the usual fashion. The only
5884 difference between using an @option{-l} option and specifying a file name
5885 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
5886 and searches several directories.
5890 You need this special case of the @option{-l} option in order to
5891 link an Objective-C or Objective-C++ program.
5894 @opindex nostartfiles
5895 Do not use the standard system startup files when linking.
5896 The standard system libraries are used normally, unless @option{-nostdlib}
5897 or @option{-nodefaultlibs} is used.
5899 @item -nodefaultlibs
5900 @opindex nodefaultlibs
5901 Do not use the standard system libraries when linking.
5902 Only the libraries you specify will be passed to the linker.
5903 The standard startup files are used normally, unless @option{-nostartfiles}
5904 is used. The compiler may generate calls to @code{memcmp},
5905 @code{memset}, @code{memcpy} and @code{memmove}.
5906 These entries are usually resolved by entries in
5907 libc. These entry points should be supplied through some other
5908 mechanism when this option is specified.
5912 Do not use the standard system startup files or libraries when linking.
5913 No startup files and only the libraries you specify will be passed to
5914 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
5915 @code{memcpy} and @code{memmove}.
5916 These entries are usually resolved by entries in
5917 libc. These entry points should be supplied through some other
5918 mechanism when this option is specified.
5920 @cindex @option{-lgcc}, use with @option{-nostdlib}
5921 @cindex @option{-nostdlib} and unresolved references
5922 @cindex unresolved references and @option{-nostdlib}
5923 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
5924 @cindex @option{-nodefaultlibs} and unresolved references
5925 @cindex unresolved references and @option{-nodefaultlibs}
5926 One of the standard libraries bypassed by @option{-nostdlib} and
5927 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
5928 that GCC uses to overcome shortcomings of particular machines, or special
5929 needs for some languages.
5930 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
5931 Collection (GCC) Internals},
5932 for more discussion of @file{libgcc.a}.)
5933 In most cases, you need @file{libgcc.a} even when you want to avoid
5934 other standard libraries. In other words, when you specify @option{-nostdlib}
5935 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
5936 This ensures that you have no unresolved references to internal GCC
5937 library subroutines. (For example, @samp{__main}, used to ensure C++
5938 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
5939 GNU Compiler Collection (GCC) Internals}.)
5943 Produce a position independent executable on targets which support it.
5944 For predictable results, you must also specify the same set of options
5945 that were used to generate code (@option{-fpie}, @option{-fPIE},
5946 or model suboptions) when you specify this option.
5950 Remove all symbol table and relocation information from the executable.
5954 On systems that support dynamic linking, this prevents linking with the shared
5955 libraries. On other systems, this option has no effect.
5959 Produce a shared object which can then be linked with other objects to
5960 form an executable. Not all systems support this option. For predictable
5961 results, you must also specify the same set of options that were used to
5962 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
5963 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
5964 needs to build supplementary stub code for constructors to work. On
5965 multi-libbed systems, @samp{gcc -shared} must select the correct support
5966 libraries to link against. Failing to supply the correct flags may lead
5967 to subtle defects. Supplying them in cases where they are not necessary
5970 @item -shared-libgcc
5971 @itemx -static-libgcc
5972 @opindex shared-libgcc
5973 @opindex static-libgcc
5974 On systems that provide @file{libgcc} as a shared library, these options
5975 force the use of either the shared or static version respectively.
5976 If no shared version of @file{libgcc} was built when the compiler was
5977 configured, these options have no effect.
5979 There are several situations in which an application should use the
5980 shared @file{libgcc} instead of the static version. The most common
5981 of these is when the application wishes to throw and catch exceptions
5982 across different shared libraries. In that case, each of the libraries
5983 as well as the application itself should use the shared @file{libgcc}.
5985 Therefore, the G++ and GCJ drivers automatically add
5986 @option{-shared-libgcc} whenever you build a shared library or a main
5987 executable, because C++ and Java programs typically use exceptions, so
5988 this is the right thing to do.
5990 If, instead, you use the GCC driver to create shared libraries, you may
5991 find that they will not always be linked with the shared @file{libgcc}.
5992 If GCC finds, at its configuration time, that you have a non-GNU linker
5993 or a GNU linker that does not support option @option{--eh-frame-hdr},
5994 it will link the shared version of @file{libgcc} into shared libraries
5995 by default. Otherwise, it will take advantage of the linker and optimize
5996 away the linking with the shared version of @file{libgcc}, linking with
5997 the static version of libgcc by default. This allows exceptions to
5998 propagate through such shared libraries, without incurring relocation
5999 costs at library load time.
6001 However, if a library or main executable is supposed to throw or catch
6002 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6003 for the languages used in the program, or using the option
6004 @option{-shared-libgcc}, such that it is linked with the shared
6009 Bind references to global symbols when building a shared object. Warn
6010 about any unresolved references (unless overridden by the link editor
6011 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6014 @item -Xlinker @var{option}
6016 Pass @var{option} as an option to the linker. You can use this to
6017 supply system-specific linker options which GCC does not know how to
6020 If you want to pass an option that takes an argument, you must use
6021 @option{-Xlinker} twice, once for the option and once for the argument.
6022 For example, to pass @option{-assert definitions}, you must write
6023 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6024 @option{-Xlinker "-assert definitions"}, because this passes the entire
6025 string as a single argument, which is not what the linker expects.
6027 @item -Wl,@var{option}
6029 Pass @var{option} as an option to the linker. If @var{option} contains
6030 commas, it is split into multiple options at the commas.
6032 @item -u @var{symbol}
6034 Pretend the symbol @var{symbol} is undefined, to force linking of
6035 library modules to define it. You can use @option{-u} multiple times with
6036 different symbols to force loading of additional library modules.
6039 @node Directory Options
6040 @section Options for Directory Search
6041 @cindex directory options
6042 @cindex options, directory search
6045 These options specify directories to search for header files, for
6046 libraries and for parts of the compiler:
6051 Add the directory @var{dir} to the head of the list of directories to be
6052 searched for header files. This can be used to override a system header
6053 file, substituting your own version, since these directories are
6054 searched before the system header file directories. However, you should
6055 not use this option to add directories that contain vendor-supplied
6056 system header files (use @option{-isystem} for that). If you use more than
6057 one @option{-I} option, the directories are scanned in left-to-right
6058 order; the standard system directories come after.
6060 If a standard system include directory, or a directory specified with
6061 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6062 option will be ignored. The directory will still be searched but as a
6063 system directory at its normal position in the system include chain.
6064 This is to ensure that GCC's procedure to fix buggy system headers and
6065 the ordering for the include_next directive are not inadvertently changed.
6066 If you really need to change the search order for system directories,
6067 use the @option{-nostdinc} and/or @option{-isystem} options.
6069 @item -iquote@var{dir}
6071 Add the directory @var{dir} to the head of the list of directories to
6072 be searched for header files only for the case of @samp{#include
6073 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6074 otherwise just like @option{-I}.
6078 Add directory @var{dir} to the list of directories to be searched
6081 @item -B@var{prefix}
6083 This option specifies where to find the executables, libraries,
6084 include files, and data files of the compiler itself.
6086 The compiler driver program runs one or more of the subprograms
6087 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6088 @var{prefix} as a prefix for each program it tries to run, both with and
6089 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6091 For each subprogram to be run, the compiler driver first tries the
6092 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6093 was not specified, the driver tries two standard prefixes, which are
6094 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6095 those results in a file name that is found, the unmodified program
6096 name is searched for using the directories specified in your
6097 @env{PATH} environment variable.
6099 The compiler will check to see if the path provided by the @option{-B}
6100 refers to a directory, and if necessary it will add a directory
6101 separator character at the end of the path.
6103 @option{-B} prefixes that effectively specify directory names also apply
6104 to libraries in the linker, because the compiler translates these
6105 options into @option{-L} options for the linker. They also apply to
6106 includes files in the preprocessor, because the compiler translates these
6107 options into @option{-isystem} options for the preprocessor. In this case,
6108 the compiler appends @samp{include} to the prefix.
6110 The run-time support file @file{libgcc.a} can also be searched for using
6111 the @option{-B} prefix, if needed. If it is not found there, the two
6112 standard prefixes above are tried, and that is all. The file is left
6113 out of the link if it is not found by those means.
6115 Another way to specify a prefix much like the @option{-B} prefix is to use
6116 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6119 As a special kludge, if the path provided by @option{-B} is
6120 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6121 9, then it will be replaced by @file{[dir/]include}. This is to help
6122 with boot-strapping the compiler.
6124 @item -specs=@var{file}
6126 Process @var{file} after the compiler reads in the standard @file{specs}
6127 file, in order to override the defaults that the @file{gcc} driver
6128 program uses when determining what switches to pass to @file{cc1},
6129 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6130 @option{-specs=@var{file}} can be specified on the command line, and they
6131 are processed in order, from left to right.
6135 This option has been deprecated. Please use @option{-iquote} instead for
6136 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6137 Any directories you specify with @option{-I} options before the @option{-I-}
6138 option are searched only for the case of @samp{#include "@var{file}"};
6139 they are not searched for @samp{#include <@var{file}>}.
6141 If additional directories are specified with @option{-I} options after
6142 the @option{-I-}, these directories are searched for all @samp{#include}
6143 directives. (Ordinarily @emph{all} @option{-I} directories are used
6146 In addition, the @option{-I-} option inhibits the use of the current
6147 directory (where the current input file came from) as the first search
6148 directory for @samp{#include "@var{file}"}. There is no way to
6149 override this effect of @option{-I-}. With @option{-I.} you can specify
6150 searching the directory which was current when the compiler was
6151 invoked. That is not exactly the same as what the preprocessor does
6152 by default, but it is often satisfactory.
6154 @option{-I-} does not inhibit the use of the standard system directories
6155 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6162 @section Specifying subprocesses and the switches to pass to them
6165 @command{gcc} is a driver program. It performs its job by invoking a
6166 sequence of other programs to do the work of compiling, assembling and
6167 linking. GCC interprets its command-line parameters and uses these to
6168 deduce which programs it should invoke, and which command-line options
6169 it ought to place on their command lines. This behavior is controlled
6170 by @dfn{spec strings}. In most cases there is one spec string for each
6171 program that GCC can invoke, but a few programs have multiple spec
6172 strings to control their behavior. The spec strings built into GCC can
6173 be overridden by using the @option{-specs=} command-line switch to specify
6176 @dfn{Spec files} are plaintext files that are used to construct spec
6177 strings. They consist of a sequence of directives separated by blank
6178 lines. The type of directive is determined by the first non-whitespace
6179 character on the line and it can be one of the following:
6182 @item %@var{command}
6183 Issues a @var{command} to the spec file processor. The commands that can
6187 @item %include <@var{file}>
6189 Search for @var{file} and insert its text at the current point in the
6192 @item %include_noerr <@var{file}>
6193 @cindex %include_noerr
6194 Just like @samp{%include}, but do not generate an error message if the include
6195 file cannot be found.
6197 @item %rename @var{old_name} @var{new_name}
6199 Rename the spec string @var{old_name} to @var{new_name}.
6203 @item *[@var{spec_name}]:
6204 This tells the compiler to create, override or delete the named spec
6205 string. All lines after this directive up to the next directive or
6206 blank line are considered to be the text for the spec string. If this
6207 results in an empty string then the spec will be deleted. (Or, if the
6208 spec did not exist, then nothing will happened.) Otherwise, if the spec
6209 does not currently exist a new spec will be created. If the spec does
6210 exist then its contents will be overridden by the text of this
6211 directive, unless the first character of that text is the @samp{+}
6212 character, in which case the text will be appended to the spec.
6214 @item [@var{suffix}]:
6215 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6216 and up to the next directive or blank line are considered to make up the
6217 spec string for the indicated suffix. When the compiler encounters an
6218 input file with the named suffix, it will processes the spec string in
6219 order to work out how to compile that file. For example:
6226 This says that any input file whose name ends in @samp{.ZZ} should be
6227 passed to the program @samp{z-compile}, which should be invoked with the
6228 command-line switch @option{-input} and with the result of performing the
6229 @samp{%i} substitution. (See below.)
6231 As an alternative to providing a spec string, the text that follows a
6232 suffix directive can be one of the following:
6235 @item @@@var{language}
6236 This says that the suffix is an alias for a known @var{language}. This is
6237 similar to using the @option{-x} command-line switch to GCC to specify a
6238 language explicitly. For example:
6245 Says that .ZZ files are, in fact, C++ source files.
6248 This causes an error messages saying:
6251 @var{name} compiler not installed on this system.
6255 GCC already has an extensive list of suffixes built into it.
6256 This directive will add an entry to the end of the list of suffixes, but
6257 since the list is searched from the end backwards, it is effectively
6258 possible to override earlier entries using this technique.
6262 GCC has the following spec strings built into it. Spec files can
6263 override these strings or create their own. Note that individual
6264 targets can also add their own spec strings to this list.
6267 asm Options to pass to the assembler
6268 asm_final Options to pass to the assembler post-processor
6269 cpp Options to pass to the C preprocessor
6270 cc1 Options to pass to the C compiler
6271 cc1plus Options to pass to the C++ compiler
6272 endfile Object files to include at the end of the link
6273 link Options to pass to the linker
6274 lib Libraries to include on the command line to the linker
6275 libgcc Decides which GCC support library to pass to the linker
6276 linker Sets the name of the linker
6277 predefines Defines to be passed to the C preprocessor
6278 signed_char Defines to pass to CPP to say whether @code{char} is signed
6280 startfile Object files to include at the start of the link
6283 Here is a small example of a spec file:
6289 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6292 This example renames the spec called @samp{lib} to @samp{old_lib} and
6293 then overrides the previous definition of @samp{lib} with a new one.
6294 The new definition adds in some extra command-line options before
6295 including the text of the old definition.
6297 @dfn{Spec strings} are a list of command-line options to be passed to their
6298 corresponding program. In addition, the spec strings can contain
6299 @samp{%}-prefixed sequences to substitute variable text or to
6300 conditionally insert text into the command line. Using these constructs
6301 it is possible to generate quite complex command lines.
6303 Here is a table of all defined @samp{%}-sequences for spec
6304 strings. Note that spaces are not generated automatically around the
6305 results of expanding these sequences. Therefore you can concatenate them
6306 together or combine them with constant text in a single argument.
6310 Substitute one @samp{%} into the program name or argument.
6313 Substitute the name of the input file being processed.
6316 Substitute the basename of the input file being processed.
6317 This is the substring up to (and not including) the last period
6318 and not including the directory.
6321 This is the same as @samp{%b}, but include the file suffix (text after
6325 Marks the argument containing or following the @samp{%d} as a
6326 temporary file name, so that that file will be deleted if GCC exits
6327 successfully. Unlike @samp{%g}, this contributes no text to the
6330 @item %g@var{suffix}
6331 Substitute a file name that has suffix @var{suffix} and is chosen
6332 once per compilation, and mark the argument in the same way as
6333 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6334 name is now chosen in a way that is hard to predict even when previously
6335 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6336 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6337 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6338 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6339 was simply substituted with a file name chosen once per compilation,
6340 without regard to any appended suffix (which was therefore treated
6341 just like ordinary text), making such attacks more likely to succeed.
6343 @item %u@var{suffix}
6344 Like @samp{%g}, but generates a new temporary file name even if
6345 @samp{%u@var{suffix}} was already seen.
6347 @item %U@var{suffix}
6348 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6349 new one if there is no such last file name. In the absence of any
6350 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6351 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6352 would involve the generation of two distinct file names, one
6353 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6354 simply substituted with a file name chosen for the previous @samp{%u},
6355 without regard to any appended suffix.
6357 @item %j@var{suffix}
6358 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6359 writable, and if save-temps is off; otherwise, substitute the name
6360 of a temporary file, just like @samp{%u}. This temporary file is not
6361 meant for communication between processes, but rather as a junk
6364 @item %|@var{suffix}
6365 @itemx %m@var{suffix}
6366 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6367 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6368 all. These are the two most common ways to instruct a program that it
6369 should read from standard input or write to standard output. If you
6370 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6371 construct: see for example @file{f/lang-specs.h}.
6373 @item %.@var{SUFFIX}
6374 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6375 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6376 terminated by the next space or %.
6379 Marks the argument containing or following the @samp{%w} as the
6380 designated output file of this compilation. This puts the argument
6381 into the sequence of arguments that @samp{%o} will substitute later.
6384 Substitutes the names of all the output files, with spaces
6385 automatically placed around them. You should write spaces
6386 around the @samp{%o} as well or the results are undefined.
6387 @samp{%o} is for use in the specs for running the linker.
6388 Input files whose names have no recognized suffix are not compiled
6389 at all, but they are included among the output files, so they will
6393 Substitutes the suffix for object files. Note that this is
6394 handled specially when it immediately follows @samp{%g, %u, or %U},
6395 because of the need for those to form complete file names. The
6396 handling is such that @samp{%O} is treated exactly as if it had already
6397 been substituted, except that @samp{%g, %u, and %U} do not currently
6398 support additional @var{suffix} characters following @samp{%O} as they would
6399 following, for example, @samp{.o}.
6402 Substitutes the standard macro predefinitions for the
6403 current target machine. Use this when running @code{cpp}.
6406 Like @samp{%p}, but puts @samp{__} before and after the name of each
6407 predefined macro, except for macros that start with @samp{__} or with
6408 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6412 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6413 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
6414 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6418 Current argument is the name of a library or startup file of some sort.
6419 Search for that file in a standard list of directories and substitute
6420 the full name found.
6423 Print @var{str} as an error message. @var{str} is terminated by a newline.
6424 Use this when inconsistent options are detected.
6427 Substitute the contents of spec string @var{name} at this point.
6430 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6432 @item %x@{@var{option}@}
6433 Accumulate an option for @samp{%X}.
6436 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6440 Output the accumulated assembler options specified by @option{-Wa}.
6443 Output the accumulated preprocessor options specified by @option{-Wp}.
6446 Process the @code{asm} spec. This is used to compute the
6447 switches to be passed to the assembler.
6450 Process the @code{asm_final} spec. This is a spec string for
6451 passing switches to an assembler post-processor, if such a program is
6455 Process the @code{link} spec. This is the spec for computing the
6456 command line passed to the linker. Typically it will make use of the
6457 @samp{%L %G %S %D and %E} sequences.
6460 Dump out a @option{-L} option for each directory that GCC believes might
6461 contain startup files. If the target supports multilibs then the
6462 current multilib directory will be prepended to each of these paths.
6465 Process the @code{lib} spec. This is a spec string for deciding which
6466 libraries should be included on the command line to the linker.
6469 Process the @code{libgcc} spec. This is a spec string for deciding
6470 which GCC support library should be included on the command line to the linker.
6473 Process the @code{startfile} spec. This is a spec for deciding which
6474 object files should be the first ones passed to the linker. Typically
6475 this might be a file named @file{crt0.o}.
6478 Process the @code{endfile} spec. This is a spec string that specifies
6479 the last object files that will be passed to the linker.
6482 Process the @code{cpp} spec. This is used to construct the arguments
6483 to be passed to the C preprocessor.
6486 Process the @code{cc1} spec. This is used to construct the options to be
6487 passed to the actual C compiler (@samp{cc1}).
6490 Process the @code{cc1plus} spec. This is used to construct the options to be
6491 passed to the actual C++ compiler (@samp{cc1plus}).
6494 Substitute the variable part of a matched option. See below.
6495 Note that each comma in the substituted string is replaced by
6499 Remove all occurrences of @code{-S} from the command line. Note---this
6500 command is position dependent. @samp{%} commands in the spec string
6501 before this one will see @code{-S}, @samp{%} commands in the spec string
6502 after this one will not.
6504 @item %:@var{function}(@var{args})
6505 Call the named function @var{function}, passing it @var{args}.
6506 @var{args} is first processed as a nested spec string, then split
6507 into an argument vector in the usual fashion. The function returns
6508 a string which is processed as if it had appeared literally as part
6509 of the current spec.
6511 The following built-in spec functions are provided:
6514 @item @code{if-exists}
6515 The @code{if-exists} spec function takes one argument, an absolute
6516 pathname to a file. If the file exists, @code{if-exists} returns the
6517 pathname. Here is a small example of its usage:
6521 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6524 @item @code{if-exists-else}
6525 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6526 spec function, except that it takes two arguments. The first argument is
6527 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6528 returns the pathname. If it does not exist, it returns the second argument.
6529 This way, @code{if-exists-else} can be used to select one file or another,
6530 based on the existence of the first. Here is a small example of its usage:
6534 crt0%O%s %:if-exists(crti%O%s) \
6535 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6538 @item @code{replace-outfile}
6539 The @code{replace-outfile} spec function takes two arguments. It looks for the
6540 first argument in the outfiles array and replaces it with the second argument. Here
6541 is a small example of its usage:
6544 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
6550 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6551 If that switch was not specified, this substitutes nothing. Note that
6552 the leading dash is omitted when specifying this option, and it is
6553 automatically inserted if the substitution is performed. Thus the spec
6554 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6555 and would output the command line option @option{-foo}.
6557 @item %W@{@code{S}@}
6558 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6561 @item %@{@code{S}*@}
6562 Substitutes all the switches specified to GCC whose names start
6563 with @code{-S}, but which also take an argument. This is used for
6564 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6565 GCC considers @option{-o foo} as being
6566 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6567 text, including the space. Thus two arguments would be generated.
6569 @item %@{@code{S}*&@code{T}*@}
6570 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6571 (the order of @code{S} and @code{T} in the spec is not significant).
6572 There can be any number of ampersand-separated variables; for each the
6573 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6575 @item %@{@code{S}:@code{X}@}
6576 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6578 @item %@{!@code{S}:@code{X}@}
6579 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6581 @item %@{@code{S}*:@code{X}@}
6582 Substitutes @code{X} if one or more switches whose names start with
6583 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6584 once, no matter how many such switches appeared. However, if @code{%*}
6585 appears somewhere in @code{X}, then @code{X} will be substituted once
6586 for each matching switch, with the @code{%*} replaced by the part of
6587 that switch that matched the @code{*}.
6589 @item %@{.@code{S}:@code{X}@}
6590 Substitutes @code{X}, if processing a file with suffix @code{S}.
6592 @item %@{!.@code{S}:@code{X}@}
6593 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6595 @item %@{@code{S}|@code{P}:@code{X}@}
6596 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6597 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
6598 although they have a stronger binding than the @samp{|}. If @code{%*}
6599 appears in @code{X}, all of the alternatives must be starred, and only
6600 the first matching alternative is substituted.
6602 For example, a spec string like this:
6605 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
6608 will output the following command-line options from the following input
6609 command-line options:
6614 -d fred.c -foo -baz -boggle
6615 -d jim.d -bar -baz -boggle
6618 @item %@{S:X; T:Y; :D@}
6620 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
6621 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
6622 be as many clauses as you need. This may be combined with @code{.},
6623 @code{!}, @code{|}, and @code{*} as needed.
6628 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
6629 construct may contain other nested @samp{%} constructs or spaces, or
6630 even newlines. They are processed as usual, as described above.
6631 Trailing white space in @code{X} is ignored. White space may also
6632 appear anywhere on the left side of the colon in these constructs,
6633 except between @code{.} or @code{*} and the corresponding word.
6635 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
6636 handled specifically in these constructs. If another value of
6637 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
6638 @option{-W} switch is found later in the command line, the earlier
6639 switch value is ignored, except with @{@code{S}*@} where @code{S} is
6640 just one letter, which passes all matching options.
6642 The character @samp{|} at the beginning of the predicate text is used to
6643 indicate that a command should be piped to the following command, but
6644 only if @option{-pipe} is specified.
6646 It is built into GCC which switches take arguments and which do not.
6647 (You might think it would be useful to generalize this to allow each
6648 compiler's spec to say which switches take arguments. But this cannot
6649 be done in a consistent fashion. GCC cannot even decide which input
6650 files have been specified without knowing which switches take arguments,
6651 and it must know which input files to compile in order to tell which
6654 GCC also knows implicitly that arguments starting in @option{-l} are to be
6655 treated as compiler output files, and passed to the linker in their
6656 proper position among the other output files.
6658 @c man begin OPTIONS
6660 @node Target Options
6661 @section Specifying Target Machine and Compiler Version
6662 @cindex target options
6663 @cindex cross compiling
6664 @cindex specifying machine version
6665 @cindex specifying compiler version and target machine
6666 @cindex compiler version, specifying
6667 @cindex target machine, specifying
6669 The usual way to run GCC is to run the executable called @file{gcc}, or
6670 @file{<machine>-gcc} when cross-compiling, or
6671 @file{<machine>-gcc-<version>} to run a version other than the one that
6672 was installed last. Sometimes this is inconvenient, so GCC provides
6673 options that will switch to another cross-compiler or version.
6676 @item -b @var{machine}
6678 The argument @var{machine} specifies the target machine for compilation.
6680 The value to use for @var{machine} is the same as was specified as the
6681 machine type when configuring GCC as a cross-compiler. For
6682 example, if a cross-compiler was configured with @samp{configure
6683 i386v}, meaning to compile for an 80386 running System V, then you
6684 would specify @option{-b i386v} to run that cross compiler.
6686 @item -V @var{version}
6688 The argument @var{version} specifies which version of GCC to run.
6689 This is useful when multiple versions are installed. For example,
6690 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
6693 The @option{-V} and @option{-b} options work by running the
6694 @file{<machine>-gcc-<version>} executable, so there's no real reason to
6695 use them if you can just run that directly.
6697 @node Submodel Options
6698 @section Hardware Models and Configurations
6699 @cindex submodel options
6700 @cindex specifying hardware config
6701 @cindex hardware models and configurations, specifying
6702 @cindex machine dependent options
6704 Earlier we discussed the standard option @option{-b} which chooses among
6705 different installed compilers for completely different target
6706 machines, such as VAX vs.@: 68000 vs.@: 80386.
6708 In addition, each of these target machine types can have its own
6709 special options, starting with @samp{-m}, to choose among various
6710 hardware models or configurations---for example, 68010 vs 68020,
6711 floating coprocessor or none. A single installed version of the
6712 compiler can compile for any model or configuration, according to the
6715 Some configurations of the compiler also support additional special
6716 options, usually for compatibility with other compilers on the same
6719 These options are defined by the macro @code{TARGET_SWITCHES} in the
6720 machine description. The default for the options is also defined by
6721 that macro, which enables you to change the defaults.
6723 @c This list is ordered alphanumerically by subsection name.
6724 @c It should be the same order and spelling as these options are listed
6725 @c in Machine Dependent Options
6733 * DEC Alpha Options::
6734 * DEC Alpha/VMS Options::
6738 * i386 and x86-64 Options::
6750 * RS/6000 and PowerPC Options::
6751 * S/390 and zSeries Options::
6754 * System V Options::
6755 * TMS320C3x/C4x Options::
6759 * Xstormy16 Options::
6765 @subsection ARC Options
6768 These options are defined for ARC implementations:
6773 Compile code for little endian mode. This is the default.
6777 Compile code for big endian mode.
6780 @opindex mmangle-cpu
6781 Prepend the name of the cpu to all public symbol names.
6782 In multiple-processor systems, there are many ARC variants with different
6783 instruction and register set characteristics. This flag prevents code
6784 compiled for one cpu to be linked with code compiled for another.
6785 No facility exists for handling variants that are ``almost identical''.
6786 This is an all or nothing option.
6788 @item -mcpu=@var{cpu}
6790 Compile code for ARC variant @var{cpu}.
6791 Which variants are supported depend on the configuration.
6792 All variants support @option{-mcpu=base}, this is the default.
6794 @item -mtext=@var{text-section}
6795 @itemx -mdata=@var{data-section}
6796 @itemx -mrodata=@var{readonly-data-section}
6800 Put functions, data, and readonly data in @var{text-section},
6801 @var{data-section}, and @var{readonly-data-section} respectively
6802 by default. This can be overridden with the @code{section} attribute.
6803 @xref{Variable Attributes}.
6808 @subsection ARM Options
6811 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6815 @item -mabi=@var{name}
6817 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
6818 @samp{atpcs}, @samp{aapcs} and @samp{iwmmxt}.
6821 @opindex mapcs-frame
6822 Generate a stack frame that is compliant with the ARM Procedure Call
6823 Standard for all functions, even if this is not strictly necessary for
6824 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6825 with this option will cause the stack frames not to be generated for
6826 leaf functions. The default is @option{-mno-apcs-frame}.
6830 This is a synonym for @option{-mapcs-frame}.
6833 @c not currently implemented
6834 @item -mapcs-stack-check
6835 @opindex mapcs-stack-check
6836 Generate code to check the amount of stack space available upon entry to
6837 every function (that actually uses some stack space). If there is
6838 insufficient space available then either the function
6839 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
6840 called, depending upon the amount of stack space required. The run time
6841 system is required to provide these functions. The default is
6842 @option{-mno-apcs-stack-check}, since this produces smaller code.
6844 @c not currently implemented
6846 @opindex mapcs-float
6847 Pass floating point arguments using the float point registers. This is
6848 one of the variants of the APCS@. This option is recommended if the
6849 target hardware has a floating point unit or if a lot of floating point
6850 arithmetic is going to be performed by the code. The default is
6851 @option{-mno-apcs-float}, since integer only code is slightly increased in
6852 size if @option{-mapcs-float} is used.
6854 @c not currently implemented
6855 @item -mapcs-reentrant
6856 @opindex mapcs-reentrant
6857 Generate reentrant, position independent code. The default is
6858 @option{-mno-apcs-reentrant}.
6861 @item -mthumb-interwork
6862 @opindex mthumb-interwork
6863 Generate code which supports calling between the ARM and Thumb
6864 instruction sets. Without this option the two instruction sets cannot
6865 be reliably used inside one program. The default is
6866 @option{-mno-thumb-interwork}, since slightly larger code is generated
6867 when @option{-mthumb-interwork} is specified.
6869 @item -mno-sched-prolog
6870 @opindex mno-sched-prolog
6871 Prevent the reordering of instructions in the function prolog, or the
6872 merging of those instruction with the instructions in the function's
6873 body. This means that all functions will start with a recognizable set
6874 of instructions (or in fact one of a choice from a small set of
6875 different function prologues), and this information can be used to
6876 locate the start if functions inside an executable piece of code. The
6877 default is @option{-msched-prolog}.
6880 @opindex mhard-float
6881 Generate output containing floating point instructions. This is the
6885 @opindex msoft-float
6886 Generate output containing library calls for floating point.
6887 @strong{Warning:} the requisite libraries are not available for all ARM
6888 targets. Normally the facilities of the machine's usual C compiler are
6889 used, but this cannot be done directly in cross-compilation. You must make
6890 your own arrangements to provide suitable library functions for
6893 @option{-msoft-float} changes the calling convention in the output file;
6894 therefore, it is only useful if you compile @emph{all} of a program with
6895 this option. In particular, you need to compile @file{libgcc.a}, the
6896 library that comes with GCC, with @option{-msoft-float} in order for
6899 @item -mfloat-abi=@var{name}
6901 Specifies which ABI to use for floating point values. Permissible values
6902 are: @samp{soft}, @samp{softfp} and @samp{hard}.
6904 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
6905 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
6906 of floating point instructions, but still uses the soft-float calling
6909 @item -mlittle-endian
6910 @opindex mlittle-endian
6911 Generate code for a processor running in little-endian mode. This is
6912 the default for all standard configurations.
6915 @opindex mbig-endian
6916 Generate code for a processor running in big-endian mode; the default is
6917 to compile code for a little-endian processor.
6919 @item -mwords-little-endian
6920 @opindex mwords-little-endian
6921 This option only applies when generating code for big-endian processors.
6922 Generate code for a little-endian word order but a big-endian byte
6923 order. That is, a byte order of the form @samp{32107654}. Note: this
6924 option should only be used if you require compatibility with code for
6925 big-endian ARM processors generated by versions of the compiler prior to
6928 @item -mcpu=@var{name}
6930 This specifies the name of the target ARM processor. GCC uses this name
6931 to determine what kind of instructions it can emit when generating
6932 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
6933 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
6934 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
6935 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
6936 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
6937 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
6938 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
6939 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
6940 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
6941 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
6942 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
6943 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
6944 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
6945 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
6948 @itemx -mtune=@var{name}
6950 This option is very similar to the @option{-mcpu=} option, except that
6951 instead of specifying the actual target processor type, and hence
6952 restricting which instructions can be used, it specifies that GCC should
6953 tune the performance of the code as if the target were of the type
6954 specified in this option, but still choosing the instructions that it
6955 will generate based on the cpu specified by a @option{-mcpu=} option.
6956 For some ARM implementations better performance can be obtained by using
6959 @item -march=@var{name}
6961 This specifies the name of the target ARM architecture. GCC uses this
6962 name to determine what kind of instructions it can emit when generating
6963 assembly code. This option can be used in conjunction with or instead
6964 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
6965 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
6966 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
6967 @samp{iwmmxt}, @samp{ep9312}.
6969 @item -mfpu=@var{name}
6970 @itemx -mfpe=@var{number}
6971 @itemx -mfp=@var{number}
6975 This specifies what floating point hardware (or hardware emulation) is
6976 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
6977 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
6978 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
6979 with older versions of GCC@.
6981 If @option{-msoft-float} is specified this specifies the format of
6982 floating point values.
6984 @item -mstructure-size-boundary=@var{n}
6985 @opindex mstructure-size-boundary
6986 The size of all structures and unions will be rounded up to a multiple
6987 of the number of bits set by this option. Permissible values are 8, 32
6988 and 64. The default value varies for different toolchains. For the COFF
6989 targeted toolchain the default value is 8. A value of 64 is only allowed
6990 if the underlying ABI supports it.
6992 Specifying the larger number can produce faster, more efficient code, but
6993 can also increase the size of the program. Different values are potentially
6994 incompatible. Code compiled with one value cannot necessarily expect to
6995 work with code or libraries compiled with another value, if they exchange
6996 information using structures or unions.
6998 @item -mabort-on-noreturn
6999 @opindex mabort-on-noreturn
7000 Generate a call to the function @code{abort} at the end of a
7001 @code{noreturn} function. It will be executed if the function tries to
7005 @itemx -mno-long-calls
7006 @opindex mlong-calls
7007 @opindex mno-long-calls
7008 Tells the compiler to perform function calls by first loading the
7009 address of the function into a register and then performing a subroutine
7010 call on this register. This switch is needed if the target function
7011 will lie outside of the 64 megabyte addressing range of the offset based
7012 version of subroutine call instruction.
7014 Even if this switch is enabled, not all function calls will be turned
7015 into long calls. The heuristic is that static functions, functions
7016 which have the @samp{short-call} attribute, functions that are inside
7017 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7018 definitions have already been compiled within the current compilation
7019 unit, will not be turned into long calls. The exception to this rule is
7020 that weak function definitions, functions with the @samp{long-call}
7021 attribute or the @samp{section} attribute, and functions that are within
7022 the scope of a @samp{#pragma long_calls} directive, will always be
7023 turned into long calls.
7025 This feature is not enabled by default. Specifying
7026 @option{-mno-long-calls} will restore the default behavior, as will
7027 placing the function calls within the scope of a @samp{#pragma
7028 long_calls_off} directive. Note these switches have no effect on how
7029 the compiler generates code to handle function calls via function
7032 @item -mnop-fun-dllimport
7033 @opindex mnop-fun-dllimport
7034 Disable support for the @code{dllimport} attribute.
7036 @item -msingle-pic-base
7037 @opindex msingle-pic-base
7038 Treat the register used for PIC addressing as read-only, rather than
7039 loading it in the prologue for each function. The run-time system is
7040 responsible for initializing this register with an appropriate value
7041 before execution begins.
7043 @item -mpic-register=@var{reg}
7044 @opindex mpic-register
7045 Specify the register to be used for PIC addressing. The default is R10
7046 unless stack-checking is enabled, when R9 is used.
7048 @item -mcirrus-fix-invalid-insns
7049 @opindex mcirrus-fix-invalid-insns
7050 @opindex mno-cirrus-fix-invalid-insns
7051 Insert NOPs into the instruction stream to in order to work around
7052 problems with invalid Maverick instruction combinations. This option
7053 is only valid if the @option{-mcpu=ep9312} option has been used to
7054 enable generation of instructions for the Cirrus Maverick floating
7055 point co-processor. This option is not enabled by default, since the
7056 problem is only present in older Maverick implementations. The default
7057 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7060 @item -mpoke-function-name
7061 @opindex mpoke-function-name
7062 Write the name of each function into the text section, directly
7063 preceding the function prologue. The generated code is similar to this:
7067 .ascii "arm_poke_function_name", 0
7070 .word 0xff000000 + (t1 - t0)
7071 arm_poke_function_name
7073 stmfd sp!, @{fp, ip, lr, pc@}
7077 When performing a stack backtrace, code can inspect the value of
7078 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7079 location @code{pc - 12} and the top 8 bits are set, then we know that
7080 there is a function name embedded immediately preceding this location
7081 and has length @code{((pc[-3]) & 0xff000000)}.
7085 Generate code for the 16-bit Thumb instruction set. The default is to
7086 use the 32-bit ARM instruction set.
7089 @opindex mtpcs-frame
7090 Generate a stack frame that is compliant with the Thumb Procedure Call
7091 Standard for all non-leaf functions. (A leaf function is one that does
7092 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7094 @item -mtpcs-leaf-frame
7095 @opindex mtpcs-leaf-frame
7096 Generate a stack frame that is compliant with the Thumb Procedure Call
7097 Standard for all leaf functions. (A leaf function is one that does
7098 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7100 @item -mcallee-super-interworking
7101 @opindex mcallee-super-interworking
7102 Gives all externally visible functions in the file being compiled an ARM
7103 instruction set header which switches to Thumb mode before executing the
7104 rest of the function. This allows these functions to be called from
7105 non-interworking code.
7107 @item -mcaller-super-interworking
7108 @opindex mcaller-super-interworking
7109 Allows calls via function pointers (including virtual functions) to
7110 execute correctly regardless of whether the target code has been
7111 compiled for interworking or not. There is a small overhead in the cost
7112 of executing a function pointer if this option is enabled.
7117 @subsection AVR Options
7120 These options are defined for AVR implementations:
7123 @item -mmcu=@var{mcu}
7125 Specify ATMEL AVR instruction set or MCU type.
7127 Instruction set avr1 is for the minimal AVR core, not supported by the C
7128 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7129 attiny11, attiny12, attiny15, attiny28).
7131 Instruction set avr2 (default) is for the classic AVR core with up to
7132 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7133 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7134 at90c8534, at90s8535).
7136 Instruction set avr3 is for the classic AVR core with up to 128K program
7137 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7139 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7140 memory space (MCU types: atmega8, atmega83, atmega85).
7142 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7143 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7144 atmega64, atmega128, at43usb355, at94k).
7148 Output instruction sizes to the asm file.
7150 @item -minit-stack=@var{N}
7151 @opindex minit-stack
7152 Specify the initial stack address, which may be a symbol or numeric value,
7153 @samp{__stack} is the default.
7155 @item -mno-interrupts
7156 @opindex mno-interrupts
7157 Generated code is not compatible with hardware interrupts.
7158 Code size will be smaller.
7160 @item -mcall-prologues
7161 @opindex mcall-prologues
7162 Functions prologues/epilogues expanded as call to appropriate
7163 subroutines. Code size will be smaller.
7165 @item -mno-tablejump
7166 @opindex mno-tablejump
7167 Do not generate tablejump insns which sometimes increase code size.
7170 @opindex mtiny-stack
7171 Change only the low 8 bits of the stack pointer.
7175 Assume int to be 8 bit integer. This affects the sizes of all types: A
7176 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7177 and long long will be 4 bytes. Please note that this option does not
7178 comply to the C standards, but it will provide you with smaller code
7183 @subsection CRIS Options
7184 @cindex CRIS Options
7186 These options are defined specifically for the CRIS ports.
7189 @item -march=@var{architecture-type}
7190 @itemx -mcpu=@var{architecture-type}
7193 Generate code for the specified architecture. The choices for
7194 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7195 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7196 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7199 @item -mtune=@var{architecture-type}
7201 Tune to @var{architecture-type} everything applicable about the generated
7202 code, except for the ABI and the set of available instructions. The
7203 choices for @var{architecture-type} are the same as for
7204 @option{-march=@var{architecture-type}}.
7206 @item -mmax-stack-frame=@var{n}
7207 @opindex mmax-stack-frame
7208 Warn when the stack frame of a function exceeds @var{n} bytes.
7210 @item -melinux-stacksize=@var{n}
7211 @opindex melinux-stacksize
7212 Only available with the @samp{cris-axis-aout} target. Arranges for
7213 indications in the program to the kernel loader that the stack of the
7214 program should be set to @var{n} bytes.
7220 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7221 @option{-march=v3} and @option{-march=v8} respectively.
7223 @item -mmul-bug-workaround
7224 @itemx -mno-mul-bug-workaround
7225 @opindex mmul-bug-workaround
7226 @opindex mno-mul-bug-workaround
7227 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7228 models where it applies. This option is active by default.
7232 Enable CRIS-specific verbose debug-related information in the assembly
7233 code. This option also has the effect to turn off the @samp{#NO_APP}
7234 formatted-code indicator to the assembler at the beginning of the
7239 Do not use condition-code results from previous instruction; always emit
7240 compare and test instructions before use of condition codes.
7242 @item -mno-side-effects
7243 @opindex mno-side-effects
7244 Do not emit instructions with side-effects in addressing modes other than
7248 @itemx -mno-stack-align
7250 @itemx -mno-data-align
7251 @itemx -mconst-align
7252 @itemx -mno-const-align
7253 @opindex mstack-align
7254 @opindex mno-stack-align
7255 @opindex mdata-align
7256 @opindex mno-data-align
7257 @opindex mconst-align
7258 @opindex mno-const-align
7259 These options (no-options) arranges (eliminate arrangements) for the
7260 stack-frame, individual data and constants to be aligned for the maximum
7261 single data access size for the chosen CPU model. The default is to
7262 arrange for 32-bit alignment. ABI details such as structure layout are
7263 not affected by these options.
7271 Similar to the stack- data- and const-align options above, these options
7272 arrange for stack-frame, writable data and constants to all be 32-bit,
7273 16-bit or 8-bit aligned. The default is 32-bit alignment.
7275 @item -mno-prologue-epilogue
7276 @itemx -mprologue-epilogue
7277 @opindex mno-prologue-epilogue
7278 @opindex mprologue-epilogue
7279 With @option{-mno-prologue-epilogue}, the normal function prologue and
7280 epilogue that sets up the stack-frame are omitted and no return
7281 instructions or return sequences are generated in the code. Use this
7282 option only together with visual inspection of the compiled code: no
7283 warnings or errors are generated when call-saved registers must be saved,
7284 or storage for local variable needs to be allocated.
7290 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7291 instruction sequences that load addresses for functions from the PLT part
7292 of the GOT rather than (traditional on other architectures) calls to the
7293 PLT@. The default is @option{-mgotplt}.
7297 Legacy no-op option only recognized with the cris-axis-aout target.
7301 Legacy no-op option only recognized with the cris-axis-elf and
7302 cris-axis-linux-gnu targets.
7306 Only recognized with the cris-axis-aout target, where it selects a
7307 GNU/linux-like multilib, include files and instruction set for
7312 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7316 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7317 to link with input-output functions from a simulator library. Code,
7318 initialized data and zero-initialized data are allocated consecutively.
7322 Like @option{-sim}, but pass linker options to locate initialized data at
7323 0x40000000 and zero-initialized data at 0x80000000.
7326 @node Darwin Options
7327 @subsection Darwin Options
7328 @cindex Darwin options
7330 These options are defined for all architectures running the Darwin operating
7333 FSF GCC on Darwin does not create ``fat'' object files; it will create
7334 an object file for the single architecture that it was built to
7335 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7336 @option{-arch} options are used; it does so by running the compiler or
7337 linker multiple times and joining the results together with
7340 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7341 @samp{i686}) is determined by the flags that specify the ISA
7342 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7343 @option{-force_cpusubtype_ALL} option can be used to override this.
7345 The Darwin tools vary in their behaviour when presented with an ISA
7346 mismatch. The assembler, @file{as}, will only permit instructions to
7347 be used that are valid for the subtype of the file it is generating,
7348 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7349 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
7350 and print an error if asked to create a shared library with a less
7351 restrictive subtype than its input files (for instance, trying to put
7352 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
7353 for executables, @file{ld}, will quietly give the executable the most
7354 restrictive subtype of any of its input files.
7359 Add the framework directory @var{dir} to the head of the list of
7360 directories to be searched for header files. These directories are
7361 interleaved with those specified by @option{-I} options and are
7362 scanned in a left-to-right order.
7364 A framework directory is a directory with frameworks in it. A
7365 framework is a directory with a @samp{"Headers"} and/or
7366 @samp{"PrivateHeaders"} directory contained directly in it that ends
7367 in @samp{".framework"}. The name of a framework is the name of this
7368 directory excluding the @samp{".framework"}. Headers associated with
7369 the framework are found in one of those two directories, with
7370 @samp{"Headers"} being searched first. A subframework is a framework
7371 directory that is in a framework's @samp{"Frameworks"} directory.
7372 Includes of subframework headers can only appear in a header of a
7373 framework that contains the subframework, or in a sibling subframework
7374 header. Two subframeworks are siblings if they occur in the same
7375 framework. A subframework should not have the same name as a
7376 framework, a warning will be issued if this is violated. Currently a
7377 subframework cannot have subframeworks, in the future, the mechanism
7378 may be extended to support this. The standard frameworks can be found
7379 in @samp{"/System/Library/Frameworks"} and
7380 @samp{"/Library/Frameworks"}. An example include looks like
7381 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
7382 the name of the framework and header.h is found in the
7383 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
7387 Emit debugging information for symbols that are used. For STABS
7388 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
7389 This is by default ON@.
7393 Emit debugging information for all symbols and types.
7395 @item -mone-byte-bool
7396 @opindex -mone-byte-bool
7397 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7398 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7399 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7400 option has no effect on x86.
7402 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
7403 to generate code that is not binary compatible with code generated
7404 without that switch. Using this switch may require recompiling all
7405 other modules in a program, including system libraries. Use this
7406 switch to conform to a non-default data model.
7408 @item -mfix-and-continue
7409 @itemx -ffix-and-continue
7410 @itemx -findirect-data
7411 @opindex mfix-and-continue
7412 @opindex ffix-and-continue
7413 @opindex findirect-data
7414 Generate code suitable for fast turn around development. Needed to
7415 enable gdb to dynamically load @code{.o} files into already running
7416 programs. @option{-findirect-data} and @option{-ffix-and-continue}
7417 are provided for backwards compatibility.
7421 Loads all members of static archive libraries.
7422 See man ld(1) for more information.
7424 @item -arch_errors_fatal
7425 @opindex arch_errors_fatal
7426 Cause the errors having to do with files that have the wrong architecture
7430 @opindex bind_at_load
7431 Causes the output file to be marked such that the dynamic linker will
7432 bind all undefined references when the file is loaded or launched.
7436 Produce a Mach-o bundle format file.
7437 See man ld(1) for more information.
7439 @item -bundle_loader @var{executable}
7440 @opindex bundle_loader
7441 This option specifies the @var{executable} that will be loading the build
7442 output file being linked. See man ld(1) for more information.
7445 @opindex -dynamiclib
7446 When passed this option, GCC will produce a dynamic library instead of
7447 an executable when linking, using the Darwin @file{libtool} command.
7449 @item -force_cpusubtype_ALL
7450 @opindex -force_cpusubtype_ALL
7451 This causes GCC's output file to have the @var{ALL} subtype, instead of
7452 one controlled by the @option{-mcpu} or @option{-march} option.
7454 @item -allowable_client @var{client_name}
7456 @itemx -compatibility_version
7457 @itemx -current_version
7459 @itemx -dependency-file
7461 @itemx -dylinker_install_name
7463 @itemx -exported_symbols_list
7465 @itemx -flat_namespace
7466 @itemx -force_flat_namespace
7467 @itemx -headerpad_max_install_names
7470 @itemx -install_name
7471 @itemx -keep_private_externs
7472 @itemx -multi_module
7473 @itemx -multiply_defined
7474 @itemx -multiply_defined_unused
7476 @itemx -no_dead_strip_inits_and_terms
7477 @itemx -nofixprebinding
7480 @itemx -noseglinkedit
7481 @itemx -pagezero_size
7483 @itemx -prebind_all_twolevel_modules
7484 @itemx -private_bundle
7485 @itemx -read_only_relocs
7487 @itemx -sectobjectsymbols
7491 @itemx -sectobjectsymbols
7494 @itemx -segs_read_only_addr
7495 @itemx -segs_read_write_addr
7496 @itemx -seg_addr_table
7497 @itemx -seg_addr_table_filename
7500 @itemx -segs_read_only_addr
7501 @itemx -segs_read_write_addr
7502 @itemx -single_module
7505 @itemx -sub_umbrella
7506 @itemx -twolevel_namespace
7509 @itemx -unexported_symbols_list
7510 @itemx -weak_reference_mismatches
7513 @opindex allowable_client
7514 @opindex client_name
7515 @opindex compatibility_version
7516 @opindex current_version
7518 @opindex dependency-file
7520 @opindex dylinker_install_name
7522 @opindex exported_symbols_list
7524 @opindex flat_namespace
7525 @opindex force_flat_namespace
7526 @opindex headerpad_max_install_names
7529 @opindex install_name
7530 @opindex keep_private_externs
7531 @opindex multi_module
7532 @opindex multiply_defined
7533 @opindex multiply_defined_unused
7535 @opindex no_dead_strip_inits_and_terms
7536 @opindex nofixprebinding
7537 @opindex nomultidefs
7539 @opindex noseglinkedit
7540 @opindex pagezero_size
7542 @opindex prebind_all_twolevel_modules
7543 @opindex private_bundle
7544 @opindex read_only_relocs
7546 @opindex sectobjectsymbols
7550 @opindex sectobjectsymbols
7553 @opindex segs_read_only_addr
7554 @opindex segs_read_write_addr
7555 @opindex seg_addr_table
7556 @opindex seg_addr_table_filename
7557 @opindex seglinkedit
7559 @opindex segs_read_only_addr
7560 @opindex segs_read_write_addr
7561 @opindex single_module
7563 @opindex sub_library
7564 @opindex sub_umbrella
7565 @opindex twolevel_namespace
7568 @opindex unexported_symbols_list
7569 @opindex weak_reference_mismatches
7570 @opindex whatsloaded
7572 These options are passed to the Darwin linker. The Darwin linker man page
7573 describes them in detail.
7576 @node DEC Alpha Options
7577 @subsection DEC Alpha Options
7579 These @samp{-m} options are defined for the DEC Alpha implementations:
7582 @item -mno-soft-float
7584 @opindex mno-soft-float
7585 @opindex msoft-float
7586 Use (do not use) the hardware floating-point instructions for
7587 floating-point operations. When @option{-msoft-float} is specified,
7588 functions in @file{libgcc.a} will be used to perform floating-point
7589 operations. Unless they are replaced by routines that emulate the
7590 floating-point operations, or compiled in such a way as to call such
7591 emulations routines, these routines will issue floating-point
7592 operations. If you are compiling for an Alpha without floating-point
7593 operations, you must ensure that the library is built so as not to call
7596 Note that Alpha implementations without floating-point operations are
7597 required to have floating-point registers.
7602 @opindex mno-fp-regs
7603 Generate code that uses (does not use) the floating-point register set.
7604 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
7605 register set is not used, floating point operands are passed in integer
7606 registers as if they were integers and floating-point results are passed
7607 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
7608 so any function with a floating-point argument or return value called by code
7609 compiled with @option{-mno-fp-regs} must also be compiled with that
7612 A typical use of this option is building a kernel that does not use,
7613 and hence need not save and restore, any floating-point registers.
7617 The Alpha architecture implements floating-point hardware optimized for
7618 maximum performance. It is mostly compliant with the IEEE floating
7619 point standard. However, for full compliance, software assistance is
7620 required. This option generates code fully IEEE compliant code
7621 @emph{except} that the @var{inexact-flag} is not maintained (see below).
7622 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
7623 defined during compilation. The resulting code is less efficient but is
7624 able to correctly support denormalized numbers and exceptional IEEE
7625 values such as not-a-number and plus/minus infinity. Other Alpha
7626 compilers call this option @option{-ieee_with_no_inexact}.
7628 @item -mieee-with-inexact
7629 @opindex mieee-with-inexact
7630 This is like @option{-mieee} except the generated code also maintains
7631 the IEEE @var{inexact-flag}. Turning on this option causes the
7632 generated code to implement fully-compliant IEEE math. In addition to
7633 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
7634 macro. On some Alpha implementations the resulting code may execute
7635 significantly slower than the code generated by default. Since there is
7636 very little code that depends on the @var{inexact-flag}, you should
7637 normally not specify this option. Other Alpha compilers call this
7638 option @option{-ieee_with_inexact}.
7640 @item -mfp-trap-mode=@var{trap-mode}
7641 @opindex mfp-trap-mode
7642 This option controls what floating-point related traps are enabled.
7643 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
7644 The trap mode can be set to one of four values:
7648 This is the default (normal) setting. The only traps that are enabled
7649 are the ones that cannot be disabled in software (e.g., division by zero
7653 In addition to the traps enabled by @samp{n}, underflow traps are enabled
7657 Like @samp{su}, but the instructions are marked to be safe for software
7658 completion (see Alpha architecture manual for details).
7661 Like @samp{su}, but inexact traps are enabled as well.
7664 @item -mfp-rounding-mode=@var{rounding-mode}
7665 @opindex mfp-rounding-mode
7666 Selects the IEEE rounding mode. Other Alpha compilers call this option
7667 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
7672 Normal IEEE rounding mode. Floating point numbers are rounded towards
7673 the nearest machine number or towards the even machine number in case
7677 Round towards minus infinity.
7680 Chopped rounding mode. Floating point numbers are rounded towards zero.
7683 Dynamic rounding mode. A field in the floating point control register
7684 (@var{fpcr}, see Alpha architecture reference manual) controls the
7685 rounding mode in effect. The C library initializes this register for
7686 rounding towards plus infinity. Thus, unless your program modifies the
7687 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
7690 @item -mtrap-precision=@var{trap-precision}
7691 @opindex mtrap-precision
7692 In the Alpha architecture, floating point traps are imprecise. This
7693 means without software assistance it is impossible to recover from a
7694 floating trap and program execution normally needs to be terminated.
7695 GCC can generate code that can assist operating system trap handlers
7696 in determining the exact location that caused a floating point trap.
7697 Depending on the requirements of an application, different levels of
7698 precisions can be selected:
7702 Program precision. This option is the default and means a trap handler
7703 can only identify which program caused a floating point exception.
7706 Function precision. The trap handler can determine the function that
7707 caused a floating point exception.
7710 Instruction precision. The trap handler can determine the exact
7711 instruction that caused a floating point exception.
7714 Other Alpha compilers provide the equivalent options called
7715 @option{-scope_safe} and @option{-resumption_safe}.
7717 @item -mieee-conformant
7718 @opindex mieee-conformant
7719 This option marks the generated code as IEEE conformant. You must not
7720 use this option unless you also specify @option{-mtrap-precision=i} and either
7721 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
7722 is to emit the line @samp{.eflag 48} in the function prologue of the
7723 generated assembly file. Under DEC Unix, this has the effect that
7724 IEEE-conformant math library routines will be linked in.
7726 @item -mbuild-constants
7727 @opindex mbuild-constants
7728 Normally GCC examines a 32- or 64-bit integer constant to
7729 see if it can construct it from smaller constants in two or three
7730 instructions. If it cannot, it will output the constant as a literal and
7731 generate code to load it from the data segment at runtime.
7733 Use this option to require GCC to construct @emph{all} integer constants
7734 using code, even if it takes more instructions (the maximum is six).
7736 You would typically use this option to build a shared library dynamic
7737 loader. Itself a shared library, it must relocate itself in memory
7738 before it can find the variables and constants in its own data segment.
7744 Select whether to generate code to be assembled by the vendor-supplied
7745 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
7763 Indicate whether GCC should generate code to use the optional BWX,
7764 CIX, FIX and MAX instruction sets. The default is to use the instruction
7765 sets supported by the CPU type specified via @option{-mcpu=} option or that
7766 of the CPU on which GCC was built if none was specified.
7771 @opindex mfloat-ieee
7772 Generate code that uses (does not use) VAX F and G floating point
7773 arithmetic instead of IEEE single and double precision.
7775 @item -mexplicit-relocs
7776 @itemx -mno-explicit-relocs
7777 @opindex mexplicit-relocs
7778 @opindex mno-explicit-relocs
7779 Older Alpha assemblers provided no way to generate symbol relocations
7780 except via assembler macros. Use of these macros does not allow
7781 optimal instruction scheduling. GNU binutils as of version 2.12
7782 supports a new syntax that allows the compiler to explicitly mark
7783 which relocations should apply to which instructions. This option
7784 is mostly useful for debugging, as GCC detects the capabilities of
7785 the assembler when it is built and sets the default accordingly.
7789 @opindex msmall-data
7790 @opindex mlarge-data
7791 When @option{-mexplicit-relocs} is in effect, static data is
7792 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
7793 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
7794 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
7795 16-bit relocations off of the @code{$gp} register. This limits the
7796 size of the small data area to 64KB, but allows the variables to be
7797 directly accessed via a single instruction.
7799 The default is @option{-mlarge-data}. With this option the data area
7800 is limited to just below 2GB@. Programs that require more than 2GB of
7801 data must use @code{malloc} or @code{mmap} to allocate the data in the
7802 heap instead of in the program's data segment.
7804 When generating code for shared libraries, @option{-fpic} implies
7805 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
7809 @opindex msmall-text
7810 @opindex mlarge-text
7811 When @option{-msmall-text} is used, the compiler assumes that the
7812 code of the entire program (or shared library) fits in 4MB, and is
7813 thus reachable with a branch instruction. When @option{-msmall-data}
7814 is used, the compiler can assume that all local symbols share the
7815 same @code{$gp} value, and thus reduce the number of instructions
7816 required for a function call from 4 to 1.
7818 The default is @option{-mlarge-text}.
7820 @item -mcpu=@var{cpu_type}
7822 Set the instruction set and instruction scheduling parameters for
7823 machine type @var{cpu_type}. You can specify either the @samp{EV}
7824 style name or the corresponding chip number. GCC supports scheduling
7825 parameters for the EV4, EV5 and EV6 family of processors and will
7826 choose the default values for the instruction set from the processor
7827 you specify. If you do not specify a processor type, GCC will default
7828 to the processor on which the compiler was built.
7830 Supported values for @var{cpu_type} are
7836 Schedules as an EV4 and has no instruction set extensions.
7840 Schedules as an EV5 and has no instruction set extensions.
7844 Schedules as an EV5 and supports the BWX extension.
7849 Schedules as an EV5 and supports the BWX and MAX extensions.
7853 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
7857 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
7860 @item -mtune=@var{cpu_type}
7862 Set only the instruction scheduling parameters for machine type
7863 @var{cpu_type}. The instruction set is not changed.
7865 @item -mmemory-latency=@var{time}
7866 @opindex mmemory-latency
7867 Sets the latency the scheduler should assume for typical memory
7868 references as seen by the application. This number is highly
7869 dependent on the memory access patterns used by the application
7870 and the size of the external cache on the machine.
7872 Valid options for @var{time} are
7876 A decimal number representing clock cycles.
7882 The compiler contains estimates of the number of clock cycles for
7883 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
7884 (also called Dcache, Scache, and Bcache), as well as to main memory.
7885 Note that L3 is only valid for EV5.
7890 @node DEC Alpha/VMS Options
7891 @subsection DEC Alpha/VMS Options
7893 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
7896 @item -mvms-return-codes
7897 @opindex mvms-return-codes
7898 Return VMS condition codes from main. The default is to return POSIX
7899 style condition (e.g.@ error) codes.
7903 @subsection FRV Options
7910 Only use the first 32 general purpose registers.
7915 Use all 64 general purpose registers.
7920 Use only the first 32 floating point registers.
7925 Use all 64 floating point registers
7928 @opindex mhard-float
7930 Use hardware instructions for floating point operations.
7933 @opindex msoft-float
7935 Use library routines for floating point operations.
7940 Dynamically allocate condition code registers.
7945 Do not try to dynamically allocate condition code registers, only
7946 use @code{icc0} and @code{fcc0}.
7951 Change ABI to use double word insns.
7956 Do not use double word instructions.
7961 Use floating point double instructions.
7966 Do not use floating point double instructions.
7971 Use media instructions.
7976 Do not use media instructions.
7981 Use multiply and add/subtract instructions.
7986 Do not use multiply and add/subtract instructions.
7991 Select the FDPIC ABI, that uses function descriptors to represent
7992 pointers to functions. Without any PIC/PIE-related options, it
7993 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
7994 assumes GOT entries and small data are within a 12-bit range from the
7995 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
7996 are computed with 32 bits.
7999 @opindex minline-plt
8001 Enable inlining of PLT entries in function calls to functions that are
8002 not known to bind locally. It has no effect without @option{-mfdpic}.
8003 It's enabled by default if optimizing for speed and compiling for
8004 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8005 optimization option such as @option{-O3} or above is present in the
8011 Assume a large TLS segment when generating thread-local code.
8016 Do not assume a large TLS segment when generating thread-local code.
8021 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8022 that is known to be in read-only sections. It's enabled by default,
8023 except for @option{-fpic} or @option{-fpie}: even though it may help
8024 make the global offset table smaller, it trades 1 instruction for 4.
8025 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8026 one of which may be shared by multiple symbols, and it avoids the need
8027 for a GOT entry for the referenced symbol, so it's more likely to be a
8028 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8030 @item -multilib-library-pic
8031 @opindex multilib-library-pic
8033 Link with the (library, not FD) pic libraries. It's implied by
8034 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8035 @option{-fpic} without @option{-mfdpic}. You should never have to use
8041 Follow the EABI requirement of always creating a frame pointer whenever
8042 a stack frame is allocated. This option is enabled by default and can
8043 be disabled with @option{-mno-linked-fp}.
8046 @opindex mlong-calls
8048 Use indirect addressing to call functions outside the current
8049 compilation unit. This allows the functions to be placed anywhere
8050 within the 32-bit address space.
8052 @item -malign-labels
8053 @opindex malign-labels
8055 Try to align labels to an 8-byte boundary by inserting nops into the
8056 previous packet. This option only has an effect when VLIW packing
8057 is enabled. It doesn't create new packets; it merely adds nops to
8061 @opindex mlibrary-pic
8063 Generate position-independent EABI code.
8068 Use only the first four media accumulator registers.
8073 Use all eight media accumulator registers.
8078 Pack VLIW instructions.
8083 Do not pack VLIW instructions.
8088 Do not mark ABI switches in e_flags.
8093 Enable the use of conditional-move instructions (default).
8095 This switch is mainly for debugging the compiler and will likely be removed
8096 in a future version.
8098 @item -mno-cond-move
8099 @opindex mno-cond-move
8101 Disable the use of conditional-move instructions.
8103 This switch is mainly for debugging the compiler and will likely be removed
8104 in a future version.
8109 Enable the use of conditional set instructions (default).
8111 This switch is mainly for debugging the compiler and will likely be removed
8112 in a future version.
8117 Disable the use of conditional set instructions.
8119 This switch is mainly for debugging the compiler and will likely be removed
8120 in a future version.
8125 Enable the use of conditional execution (default).
8127 This switch is mainly for debugging the compiler and will likely be removed
8128 in a future version.
8130 @item -mno-cond-exec
8131 @opindex mno-cond-exec
8133 Disable the use of conditional execution.
8135 This switch is mainly for debugging the compiler and will likely be removed
8136 in a future version.
8139 @opindex mvliw-branch
8141 Run a pass to pack branches into VLIW instructions (default).
8143 This switch is mainly for debugging the compiler and will likely be removed
8144 in a future version.
8146 @item -mno-vliw-branch
8147 @opindex mno-vliw-branch
8149 Do not run a pass to pack branches into VLIW instructions.
8151 This switch is mainly for debugging the compiler and will likely be removed
8152 in a future version.
8154 @item -mmulti-cond-exec
8155 @opindex mmulti-cond-exec
8157 Enable optimization of @code{&&} and @code{||} in conditional execution
8160 This switch is mainly for debugging the compiler and will likely be removed
8161 in a future version.
8163 @item -mno-multi-cond-exec
8164 @opindex mno-multi-cond-exec
8166 Disable optimization of @code{&&} and @code{||} in conditional execution.
8168 This switch is mainly for debugging the compiler and will likely be removed
8169 in a future version.
8171 @item -mnested-cond-exec
8172 @opindex mnested-cond-exec
8174 Enable nested conditional execution optimizations (default).
8176 This switch is mainly for debugging the compiler and will likely be removed
8177 in a future version.
8179 @item -mno-nested-cond-exec
8180 @opindex mno-nested-cond-exec
8182 Disable nested conditional execution optimizations.
8184 This switch is mainly for debugging the compiler and will likely be removed
8185 in a future version.
8187 @item -mtomcat-stats
8188 @opindex mtomcat-stats
8190 Cause gas to print out tomcat statistics.
8192 @item -mcpu=@var{cpu}
8195 Select the processor type for which to generate code. Possible values are
8196 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8197 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8201 @node H8/300 Options
8202 @subsection H8/300 Options
8204 These @samp{-m} options are defined for the H8/300 implementations:
8209 Shorten some address references at link time, when possible; uses the
8210 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8211 ld, Using ld}, for a fuller description.
8215 Generate code for the H8/300H@.
8219 Generate code for the H8S@.
8223 Generate code for the H8S and H8/300H in the normal mode. This switch
8224 must be used either with @option{-mh} or @option{-ms}.
8228 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8232 Make @code{int} data 32 bits by default.
8236 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8237 The default for the H8/300H and H8S is to align longs and floats on 4
8239 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8240 This option has no effect on the H8/300.
8244 @subsection HPPA Options
8245 @cindex HPPA Options
8247 These @samp{-m} options are defined for the HPPA family of computers:
8250 @item -march=@var{architecture-type}
8252 Generate code for the specified architecture. The choices for
8253 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8254 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8255 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8256 architecture option for your machine. Code compiled for lower numbered
8257 architectures will run on higher numbered architectures, but not the
8261 @itemx -mpa-risc-1-1
8262 @itemx -mpa-risc-2-0
8263 @opindex mpa-risc-1-0
8264 @opindex mpa-risc-1-1
8265 @opindex mpa-risc-2-0
8266 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8269 @opindex mbig-switch
8270 Generate code suitable for big switch tables. Use this option only if
8271 the assembler/linker complain about out of range branches within a switch
8274 @item -mjump-in-delay
8275 @opindex mjump-in-delay
8276 Fill delay slots of function calls with unconditional jump instructions
8277 by modifying the return pointer for the function call to be the target
8278 of the conditional jump.
8280 @item -mdisable-fpregs
8281 @opindex mdisable-fpregs
8282 Prevent floating point registers from being used in any manner. This is
8283 necessary for compiling kernels which perform lazy context switching of
8284 floating point registers. If you use this option and attempt to perform
8285 floating point operations, the compiler will abort.
8287 @item -mdisable-indexing
8288 @opindex mdisable-indexing
8289 Prevent the compiler from using indexing address modes. This avoids some
8290 rather obscure problems when compiling MIG generated code under MACH@.
8292 @item -mno-space-regs
8293 @opindex mno-space-regs
8294 Generate code that assumes the target has no space registers. This allows
8295 GCC to generate faster indirect calls and use unscaled index address modes.
8297 Such code is suitable for level 0 PA systems and kernels.
8299 @item -mfast-indirect-calls
8300 @opindex mfast-indirect-calls
8301 Generate code that assumes calls never cross space boundaries. This
8302 allows GCC to emit code which performs faster indirect calls.
8304 This option will not work in the presence of shared libraries or nested
8307 @item -mfixed-range=@var{register-range}
8308 @opindex mfixed-range
8309 Generate code treating the given register range as fixed registers.
8310 A fixed register is one that the register allocator can not use. This is
8311 useful when compiling kernel code. A register range is specified as
8312 two registers separated by a dash. Multiple register ranges can be
8313 specified separated by a comma.
8315 @item -mlong-load-store
8316 @opindex mlong-load-store
8317 Generate 3-instruction load and store sequences as sometimes required by
8318 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8321 @item -mportable-runtime
8322 @opindex mportable-runtime
8323 Use the portable calling conventions proposed by HP for ELF systems.
8327 Enable the use of assembler directives only GAS understands.
8329 @item -mschedule=@var{cpu-type}
8331 Schedule code according to the constraints for the machine type
8332 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8333 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8334 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8335 proper scheduling option for your machine. The default scheduling is
8339 @opindex mlinker-opt
8340 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8341 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8342 linkers in which they give bogus error messages when linking some programs.
8345 @opindex msoft-float
8346 Generate output containing library calls for floating point.
8347 @strong{Warning:} the requisite libraries are not available for all HPPA
8348 targets. Normally the facilities of the machine's usual C compiler are
8349 used, but this cannot be done directly in cross-compilation. You must make
8350 your own arrangements to provide suitable library functions for
8351 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8352 does provide software floating point support.
8354 @option{-msoft-float} changes the calling convention in the output file;
8355 therefore, it is only useful if you compile @emph{all} of a program with
8356 this option. In particular, you need to compile @file{libgcc.a}, the
8357 library that comes with GCC, with @option{-msoft-float} in order for
8362 Generate the predefine, @code{_SIO}, for server IO@. The default is
8363 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8364 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
8365 options are available under HP-UX and HI-UX@.
8369 Use GNU ld specific options. This passes @option{-shared} to ld when
8370 building a shared library. It is the default when GCC is configured,
8371 explicitly or implicitly, with the GNU linker. This option does not
8372 have any affect on which ld is called, it only changes what parameters
8373 are passed to that ld. The ld that is called is determined by the
8374 @option{--with-ld} configure option, GCC's program search path, and
8375 finally by the user's @env{PATH}. The linker used by GCC can be printed
8376 using @samp{which `gcc -print-prog-name=ld`}.
8380 Use HP ld specific options. This passes @option{-b} to ld when building
8381 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8382 links. It is the default when GCC is configured, explicitly or
8383 implicitly, with the HP linker. This option does not have any affect on
8384 which ld is called, it only changes what parameters are passed to that
8385 ld. The ld that is called is determined by the @option{--with-ld}
8386 configure option, GCC's program search path, and finally by the user's
8387 @env{PATH}. The linker used by GCC can be printed using @samp{which
8388 `gcc -print-prog-name=ld`}.
8391 @opindex mno-long-calls
8392 Generate code that uses long call sequences. This ensures that a call
8393 is always able to reach linker generated stubs. The default is to generate
8394 long calls only when the distance from the call site to the beginning
8395 of the function or translation unit, as the case may be, exceeds a
8396 predefined limit set by the branch type being used. The limits for
8397 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8398 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8401 Distances are measured from the beginning of functions when using the
8402 @option{-ffunction-sections} option, or when using the @option{-mgas}
8403 and @option{-mno-portable-runtime} options together under HP-UX with
8406 It is normally not desirable to use this option as it will degrade
8407 performance. However, it may be useful in large applications,
8408 particularly when partial linking is used to build the application.
8410 The types of long calls used depends on the capabilities of the
8411 assembler and linker, and the type of code being generated. The
8412 impact on systems that support long absolute calls, and long pic
8413 symbol-difference or pc-relative calls should be relatively small.
8414 However, an indirect call is used on 32-bit ELF systems in pic code
8415 and it is quite long.
8417 @item -munix=@var{unix-std}
8419 Generate compiler predefines and select a startfile for the specified
8420 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
8421 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
8422 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
8423 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
8424 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
8427 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
8428 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
8429 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
8430 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
8431 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
8432 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
8434 It is @emph{important} to note that this option changes the interfaces
8435 for various library routines. It also affects the operational behavior
8436 of the C library. Thus, @emph{extreme} care is needed in using this
8439 Library code that is intended to operate with more than one UNIX
8440 standard must test, set and restore the variable @var{__xpg4_extended_mask}
8441 as appropriate. Most GNU software doesn't provide this capability.
8445 Suppress the generation of link options to search libdld.sl when the
8446 @option{-static} option is specified on HP-UX 10 and later.
8450 The HP-UX implementation of setlocale in libc has a dependency on
8451 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8452 when the @option{-static} option is specified, special link options
8453 are needed to resolve this dependency.
8455 On HP-UX 10 and later, the GCC driver adds the necessary options to
8456 link with libdld.sl when the @option{-static} option is specified.
8457 This causes the resulting binary to be dynamic. On the 64-bit port,
8458 the linkers generate dynamic binaries by default in any case. The
8459 @option{-nolibdld} option can be used to prevent the GCC driver from
8460 adding these link options.
8464 Add support for multithreading with the @dfn{dce thread} library
8465 under HP-UX@. This option sets flags for both the preprocessor and
8469 @node i386 and x86-64 Options
8470 @subsection Intel 386 and AMD x86-64 Options
8471 @cindex i386 Options
8472 @cindex x86-64 Options
8473 @cindex Intel 386 Options
8474 @cindex AMD x86-64 Options
8476 These @samp{-m} options are defined for the i386 and x86-64 family of
8480 @item -mtune=@var{cpu-type}
8482 Tune to @var{cpu-type} everything applicable about the generated code, except
8483 for the ABI and the set of available instructions. The choices for
8487 Original Intel's i386 CPU@.
8489 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
8491 Intel Pentium CPU with no MMX support.
8493 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
8494 @item i686, pentiumpro
8495 Intel PentiumPro CPU@.
8497 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
8498 @item pentium3, pentium3m
8499 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
8502 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
8503 support. Used by Centrino notebooks.
8504 @item pentium4, pentium4m
8505 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
8507 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
8510 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
8511 SSE2 and SSE3 instruction set support.
8513 AMD K6 CPU with MMX instruction set support.
8515 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
8516 @item athlon, athlon-tbird
8517 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
8519 @item athlon-4, athlon-xp, athlon-mp
8520 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
8521 instruction set support.
8522 @item k8, opteron, athlon64, athlon-fx
8523 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
8524 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
8526 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
8529 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
8530 instruction set support.
8532 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
8533 implemented for this chip.)
8535 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
8536 implemented for this chip.)
8539 While picking a specific @var{cpu-type} will schedule things appropriately
8540 for that particular chip, the compiler will not generate any code that
8541 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8544 @item -march=@var{cpu-type}
8546 Generate instructions for the machine type @var{cpu-type}. The choices
8547 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8548 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8550 @item -mcpu=@var{cpu-type}
8552 A deprecated synonym for @option{-mtune}.
8561 @opindex mpentiumpro
8562 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8563 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8564 These synonyms are deprecated.
8566 @item -mfpmath=@var{unit}
8568 Generate floating point arithmetics for selected unit @var{unit}. The choices
8573 Use the standard 387 floating point coprocessor present majority of chips and
8574 emulated otherwise. Code compiled with this option will run almost everywhere.
8575 The temporary results are computed in 80bit precision instead of precision
8576 specified by the type resulting in slightly different results compared to most
8577 of other chips. See @option{-ffloat-store} for more detailed description.
8579 This is the default choice for i386 compiler.
8582 Use scalar floating point instructions present in the SSE instruction set.
8583 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8584 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8585 instruction set supports only single precision arithmetics, thus the double and
8586 extended precision arithmetics is still done using 387. Later version, present
8587 only in Pentium4 and the future AMD x86-64 chips supports double precision
8590 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
8591 or @option{-msse2} switches to enable SSE extensions and make this option
8592 effective. For the x86-64 compiler, these extensions are enabled by default.
8594 The resulting code should be considerably faster in the majority of cases and avoid
8595 the numerical instability problems of 387 code, but may break some existing
8596 code that expects temporaries to be 80bit.
8598 This is the default choice for the x86-64 compiler.
8601 Attempt to utilize both instruction sets at once. This effectively double the
8602 amount of available registers and on chips with separate execution units for
8603 387 and SSE the execution resources too. Use this option with care, as it is
8604 still experimental, because the GCC register allocator does not model separate
8605 functional units well resulting in instable performance.
8608 @item -masm=@var{dialect}
8609 @opindex masm=@var{dialect}
8610 Output asm instructions using selected @var{dialect}. Supported choices are
8611 @samp{intel} or @samp{att} (the default one).
8616 @opindex mno-ieee-fp
8617 Control whether or not the compiler uses IEEE floating point
8618 comparisons. These handle correctly the case where the result of a
8619 comparison is unordered.
8622 @opindex msoft-float
8623 Generate output containing library calls for floating point.
8624 @strong{Warning:} the requisite libraries are not part of GCC@.
8625 Normally the facilities of the machine's usual C compiler are used, but
8626 this can't be done directly in cross-compilation. You must make your
8627 own arrangements to provide suitable library functions for
8630 On machines where a function returns floating point results in the 80387
8631 register stack, some floating point opcodes may be emitted even if
8632 @option{-msoft-float} is used.
8634 @item -mno-fp-ret-in-387
8635 @opindex mno-fp-ret-in-387
8636 Do not use the FPU registers for return values of functions.
8638 The usual calling convention has functions return values of types
8639 @code{float} and @code{double} in an FPU register, even if there
8640 is no FPU@. The idea is that the operating system should emulate
8643 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8644 in ordinary CPU registers instead.
8646 @item -mno-fancy-math-387
8647 @opindex mno-fancy-math-387
8648 Some 387 emulators do not support the @code{sin}, @code{cos} and
8649 @code{sqrt} instructions for the 387. Specify this option to avoid
8650 generating those instructions. This option is the default on FreeBSD,
8651 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8652 indicates that the target cpu will always have an FPU and so the
8653 instruction will not need emulation. As of revision 2.6.1, these
8654 instructions are not generated unless you also use the
8655 @option{-funsafe-math-optimizations} switch.
8657 @item -malign-double
8658 @itemx -mno-align-double
8659 @opindex malign-double
8660 @opindex mno-align-double
8661 Control whether GCC aligns @code{double}, @code{long double}, and
8662 @code{long long} variables on a two word boundary or a one word
8663 boundary. Aligning @code{double} variables on a two word boundary will
8664 produce code that runs somewhat faster on a @samp{Pentium} at the
8665 expense of more memory.
8667 @strong{Warning:} if you use the @option{-malign-double} switch,
8668 structures containing the above types will be aligned differently than
8669 the published application binary interface specifications for the 386
8670 and will not be binary compatible with structures in code compiled
8671 without that switch.
8673 @item -m96bit-long-double
8674 @itemx -m128bit-long-double
8675 @opindex m96bit-long-double
8676 @opindex m128bit-long-double
8677 These switches control the size of @code{long double} type. The i386
8678 application binary interface specifies the size to be 96 bits,
8679 so @option{-m96bit-long-double} is the default in 32 bit mode.
8681 Modern architectures (Pentium and newer) would prefer @code{long double}
8682 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8683 conforming to the ABI, this would not be possible. So specifying a
8684 @option{-m128bit-long-double} will align @code{long double}
8685 to a 16 byte boundary by padding the @code{long double} with an additional
8688 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8689 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8691 Notice that neither of these options enable any extra precision over the x87
8692 standard of 80 bits for a @code{long double}.
8694 @strong{Warning:} if you override the default value for your target ABI, the
8695 structures and arrays containing @code{long double} variables will change
8696 their size as well as function calling convention for function taking
8697 @code{long double} will be modified. Hence they will not be binary
8698 compatible with arrays or structures in code compiled without that switch.
8702 @itemx -mno-svr3-shlib
8703 @opindex msvr3-shlib
8704 @opindex mno-svr3-shlib
8705 Control whether GCC places uninitialized local variables into the
8706 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8707 into @code{bss}. These options are meaningful only on System V Release 3.
8711 Use a different function-calling convention, in which functions that
8712 take a fixed number of arguments return with the @code{ret} @var{num}
8713 instruction, which pops their arguments while returning. This saves one
8714 instruction in the caller since there is no need to pop the arguments
8717 You can specify that an individual function is called with this calling
8718 sequence with the function attribute @samp{stdcall}. You can also
8719 override the @option{-mrtd} option by using the function attribute
8720 @samp{cdecl}. @xref{Function Attributes}.
8722 @strong{Warning:} this calling convention is incompatible with the one
8723 normally used on Unix, so you cannot use it if you need to call
8724 libraries compiled with the Unix compiler.
8726 Also, you must provide function prototypes for all functions that
8727 take variable numbers of arguments (including @code{printf});
8728 otherwise incorrect code will be generated for calls to those
8731 In addition, seriously incorrect code will result if you call a
8732 function with too many arguments. (Normally, extra arguments are
8733 harmlessly ignored.)
8735 @item -mregparm=@var{num}
8737 Control how many registers are used to pass integer arguments. By
8738 default, no registers are used to pass arguments, and at most 3
8739 registers can be used. You can control this behavior for a specific
8740 function by using the function attribute @samp{regparm}.
8741 @xref{Function Attributes}.
8743 @strong{Warning:} if you use this switch, and
8744 @var{num} is nonzero, then you must build all modules with the same
8745 value, including any libraries. This includes the system libraries and
8748 @item -mpreferred-stack-boundary=@var{num}
8749 @opindex mpreferred-stack-boundary
8750 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8751 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8752 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8753 size (@option{-Os}), in which case the default is the minimum correct
8754 alignment (4 bytes for x86, and 8 bytes for x86-64).
8756 On Pentium and PentiumPro, @code{double} and @code{long double} values
8757 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8758 suffer significant run time performance penalties. On Pentium III, the
8759 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8760 penalties if it is not 16 byte aligned.
8762 To ensure proper alignment of this values on the stack, the stack boundary
8763 must be as aligned as that required by any value stored on the stack.
8764 Further, every function must be generated such that it keeps the stack
8765 aligned. Thus calling a function compiled with a higher preferred
8766 stack boundary from a function compiled with a lower preferred stack
8767 boundary will most likely misalign the stack. It is recommended that
8768 libraries that use callbacks always use the default setting.
8770 This extra alignment does consume extra stack space, and generally
8771 increases code size. Code that is sensitive to stack space usage, such
8772 as embedded systems and operating system kernels, may want to reduce the
8773 preferred alignment to @option{-mpreferred-stack-boundary=2}.
8791 These switches enable or disable the use of built-in functions that allow
8792 direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow extensions of the
8795 @xref{X86 Built-in Functions}, for details of the functions enabled
8796 and disabled by these switches.
8798 To have SSE/SSE2 instructions generated automatically from floating-point
8799 code, see @option{-mfpmath=sse}.
8802 @itemx -mno-push-args
8804 @opindex mno-push-args
8805 Use PUSH operations to store outgoing parameters. This method is shorter
8806 and usually equally fast as method using SUB/MOV operations and is enabled
8807 by default. In some cases disabling it may improve performance because of
8808 improved scheduling and reduced dependencies.
8810 @item -maccumulate-outgoing-args
8811 @opindex maccumulate-outgoing-args
8812 If enabled, the maximum amount of space required for outgoing arguments will be
8813 computed in the function prologue. This is faster on most modern CPUs
8814 because of reduced dependencies, improved scheduling and reduced stack usage
8815 when preferred stack boundary is not equal to 2. The drawback is a notable
8816 increase in code size. This switch implies @option{-mno-push-args}.
8820 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8821 on thread-safe exception handling must compile and link all code with the
8822 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8823 @option{-D_MT}; when linking, it links in a special thread helper library
8824 @option{-lmingwthrd} which cleans up per thread exception handling data.
8826 @item -mno-align-stringops
8827 @opindex mno-align-stringops
8828 Do not align destination of inlined string operations. This switch reduces
8829 code size and improves performance in case the destination is already aligned,
8830 but GCC doesn't know about it.
8832 @item -minline-all-stringops
8833 @opindex minline-all-stringops
8834 By default GCC inlines string operations only when destination is known to be
8835 aligned at least to 4 byte boundary. This enables more inlining, increase code
8836 size, but may improve performance of code that depends on fast memcpy, strlen
8837 and memset for short lengths.
8839 @item -momit-leaf-frame-pointer
8840 @opindex momit-leaf-frame-pointer
8841 Don't keep the frame pointer in a register for leaf functions. This
8842 avoids the instructions to save, set up and restore frame pointers and
8843 makes an extra register available in leaf functions. The option
8844 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8845 which might make debugging harder.
8847 @item -mtls-direct-seg-refs
8848 @itemx -mno-tls-direct-seg-refs
8849 @opindex mtls-direct-seg-refs
8850 Controls whether TLS variables may be accessed with offsets from the
8851 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
8852 or whether the thread base pointer must be added. Whether or not this
8853 is legal depends on the operating system, and whether it maps the
8854 segment to cover the entire TLS area.
8856 For systems that use GNU libc, the default is on.
8859 These @samp{-m} switches are supported in addition to the above
8860 on AMD x86-64 processors in 64-bit environments.
8867 Generate code for a 32-bit or 64-bit environment.
8868 The 32-bit environment sets int, long and pointer to 32 bits and
8869 generates code that runs on any i386 system.
8870 The 64-bit environment sets int to 32 bits and long and pointer
8871 to 64 bits and generates code for AMD's x86-64 architecture.
8874 @opindex no-red-zone
8875 Do not use a so called red zone for x86-64 code. The red zone is mandated
8876 by the x86-64 ABI, it is a 128-byte area beyond the location of the
8877 stack pointer that will not be modified by signal or interrupt handlers
8878 and therefore can be used for temporary data without adjusting the stack
8879 pointer. The flag @option{-mno-red-zone} disables this red zone.
8881 @item -mcmodel=small
8882 @opindex mcmodel=small
8883 Generate code for the small code model: the program and its symbols must
8884 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
8885 Programs can be statically or dynamically linked. This is the default
8888 @item -mcmodel=kernel
8889 @opindex mcmodel=kernel
8890 Generate code for the kernel code model. The kernel runs in the
8891 negative 2 GB of the address space.
8892 This model has to be used for Linux kernel code.
8894 @item -mcmodel=medium
8895 @opindex mcmodel=medium
8896 Generate code for the medium model: The program is linked in the lower 2
8897 GB of the address space but symbols can be located anywhere in the
8898 address space. Programs can be statically or dynamically linked, but
8899 building of shared libraries are not supported with the medium model.
8901 @item -mcmodel=large
8902 @opindex mcmodel=large
8903 Generate code for the large model: This model makes no assumptions
8904 about addresses and sizes of sections. Currently GCC does not implement
8909 @subsection IA-64 Options
8910 @cindex IA-64 Options
8912 These are the @samp{-m} options defined for the Intel IA-64 architecture.
8916 @opindex mbig-endian
8917 Generate code for a big endian target. This is the default for HP-UX@.
8919 @item -mlittle-endian
8920 @opindex mlittle-endian
8921 Generate code for a little endian target. This is the default for AIX5
8928 Generate (or don't) code for the GNU assembler. This is the default.
8929 @c Also, this is the default if the configure option @option{--with-gnu-as}
8936 Generate (or don't) code for the GNU linker. This is the default.
8937 @c Also, this is the default if the configure option @option{--with-gnu-ld}
8942 Generate code that does not use a global pointer register. The result
8943 is not position independent code, and violates the IA-64 ABI@.
8945 @item -mvolatile-asm-stop
8946 @itemx -mno-volatile-asm-stop
8947 @opindex mvolatile-asm-stop
8948 @opindex mno-volatile-asm-stop
8949 Generate (or don't) a stop bit immediately before and after volatile asm
8952 @item -mregister-names
8953 @itemx -mno-register-names
8954 @opindex mregister-names
8955 @opindex mno-register-names
8956 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
8957 the stacked registers. This may make assembler output more readable.
8963 Disable (or enable) optimizations that use the small data section. This may
8964 be useful for working around optimizer bugs.
8967 @opindex mconstant-gp
8968 Generate code that uses a single constant global pointer value. This is
8969 useful when compiling kernel code.
8973 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
8974 This is useful when compiling firmware code.
8976 @item -minline-float-divide-min-latency
8977 @opindex minline-float-divide-min-latency
8978 Generate code for inline divides of floating point values
8979 using the minimum latency algorithm.
8981 @item -minline-float-divide-max-throughput
8982 @opindex minline-float-divide-max-throughput
8983 Generate code for inline divides of floating point values
8984 using the maximum throughput algorithm.
8986 @item -minline-int-divide-min-latency
8987 @opindex minline-int-divide-min-latency
8988 Generate code for inline divides of integer values
8989 using the minimum latency algorithm.
8991 @item -minline-int-divide-max-throughput
8992 @opindex minline-int-divide-max-throughput
8993 Generate code for inline divides of integer values
8994 using the maximum throughput algorithm.
8996 @item -minline-sqrt-min-latency
8997 @opindex minline-sqrt-min-latency
8998 Generate code for inline square roots
8999 using the minimum latency algorithm.
9001 @item -minline-sqrt-max-throughput
9002 @opindex minline-sqrt-max-throughput
9003 Generate code for inline square roots
9004 using the maximum throughput algorithm.
9006 @item -mno-dwarf2-asm
9008 @opindex mno-dwarf2-asm
9009 @opindex mdwarf2-asm
9010 Don't (or do) generate assembler code for the DWARF2 line number debugging
9011 info. This may be useful when not using the GNU assembler.
9013 @item -mearly-stop-bits
9014 @itemx -mno-early-stop-bits
9015 @opindex mearly-stop-bits
9016 @opindex mno-early-stop-bits
9017 Allow stop bits to be placed earlier than immediately preceding the
9018 instruction that triggered the stop bit. This can improve instruction
9019 scheduling, but does not always do so.
9021 @item -mfixed-range=@var{register-range}
9022 @opindex mfixed-range
9023 Generate code treating the given register range as fixed registers.
9024 A fixed register is one that the register allocator can not use. This is
9025 useful when compiling kernel code. A register range is specified as
9026 two registers separated by a dash. Multiple register ranges can be
9027 specified separated by a comma.
9029 @item -mtls-size=@var{tls-size}
9031 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9034 @item -mtune-arch=@var{cpu-type}
9036 Tune the instruction scheduling for a particular CPU, Valid values are
9037 itanium, itanium1, merced, itanium2, and mckinley.
9043 Add support for multithreading using the POSIX threads library. This
9044 option sets flags for both the preprocessor and linker. It does
9045 not affect the thread safety of object code produced by the compiler or
9046 that of libraries supplied with it. These are HP-UX specific flags.
9052 Generate code for a 32-bit or 64-bit environment.
9053 The 32-bit environment sets int, long and pointer to 32 bits.
9054 The 64-bit environment sets int to 32 bits and long and pointer
9055 to 64 bits. These are HP-UX specific flags.
9059 @node M32R/D Options
9060 @subsection M32R/D Options
9061 @cindex M32R/D options
9063 These @option{-m} options are defined for Renesas M32R/D architectures:
9068 Generate code for the M32R/2@.
9072 Generate code for the M32R/X@.
9076 Generate code for the M32R@. This is the default.
9079 @opindex mmodel=small
9080 Assume all objects live in the lower 16MB of memory (so that their addresses
9081 can be loaded with the @code{ld24} instruction), and assume all subroutines
9082 are reachable with the @code{bl} instruction.
9083 This is the default.
9085 The addressability of a particular object can be set with the
9086 @code{model} attribute.
9088 @item -mmodel=medium
9089 @opindex mmodel=medium
9090 Assume objects may be anywhere in the 32-bit address space (the compiler
9091 will generate @code{seth/add3} instructions to load their addresses), and
9092 assume all subroutines are reachable with the @code{bl} instruction.
9095 @opindex mmodel=large
9096 Assume objects may be anywhere in the 32-bit address space (the compiler
9097 will generate @code{seth/add3} instructions to load their addresses), and
9098 assume subroutines may not be reachable with the @code{bl} instruction
9099 (the compiler will generate the much slower @code{seth/add3/jl}
9100 instruction sequence).
9103 @opindex msdata=none
9104 Disable use of the small data area. Variables will be put into
9105 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
9106 @code{section} attribute has been specified).
9107 This is the default.
9109 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
9110 Objects may be explicitly put in the small data area with the
9111 @code{section} attribute using one of these sections.
9114 @opindex msdata=sdata
9115 Put small global and static data in the small data area, but do not
9116 generate special code to reference them.
9120 Put small global and static data in the small data area, and generate
9121 special instructions to reference them.
9125 @cindex smaller data references
9126 Put global and static objects less than or equal to @var{num} bytes
9127 into the small data or bss sections instead of the normal data or bss
9128 sections. The default value of @var{num} is 8.
9129 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
9130 for this option to have any effect.
9132 All modules should be compiled with the same @option{-G @var{num}} value.
9133 Compiling with different values of @var{num} may or may not work; if it
9134 doesn't the linker will give an error message---incorrect code will not be
9139 Makes the M32R specific code in the compiler display some statistics
9140 that might help in debugging programs.
9143 @opindex malign-loops
9144 Align all loops to a 32-byte boundary.
9146 @item -mno-align-loops
9147 @opindex mno-align-loops
9148 Do not enforce a 32-byte alignment for loops. This is the default.
9150 @item -missue-rate=@var{number}
9151 @opindex missue-rate=@var{number}
9152 Issue @var{number} instructions per cycle. @var{number} can only be 1
9155 @item -mbranch-cost=@var{number}
9156 @opindex mbranch-cost=@var{number}
9157 @var{number} can only be 1 or 2. If it is 1 then branches will be
9158 preferred over conditional code, if it is 2, then the opposite will
9161 @item -mflush-trap=@var{number}
9162 @opindex mflush-trap=@var{number}
9163 Specifies the trap number to use to flush the cache. The default is
9164 12. Valid numbers are between 0 and 15 inclusive.
9166 @item -mno-flush-trap
9167 @opindex mno-flush-trap
9168 Specifies that the cache cannot be flushed by using a trap.
9170 @item -mflush-func=@var{name}
9171 @opindex mflush-func=@var{name}
9172 Specifies the name of the operating system function to call to flush
9173 the cache. The default is @emph{_flush_cache}, but a function call
9174 will only be used if a trap is not available.
9176 @item -mno-flush-func
9177 @opindex mno-flush-func
9178 Indicates that there is no OS function for flushing the cache.
9182 @node M680x0 Options
9183 @subsection M680x0 Options
9184 @cindex M680x0 options
9186 These are the @samp{-m} options defined for the 68000 series. The default
9187 values for these options depends on which style of 68000 was selected when
9188 the compiler was configured; the defaults for the most common choices are
9196 Generate output for a 68000. This is the default
9197 when the compiler is configured for 68000-based systems.
9199 Use this option for microcontrollers with a 68000 or EC000 core,
9200 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
9206 Generate output for a 68020. This is the default
9207 when the compiler is configured for 68020-based systems.
9211 Generate output containing 68881 instructions for floating point.
9212 This is the default for most 68020 systems unless @option{--nfp} was
9213 specified when the compiler was configured.
9217 Generate output for a 68030. This is the default when the compiler is
9218 configured for 68030-based systems.
9222 Generate output for a 68040. This is the default when the compiler is
9223 configured for 68040-based systems.
9225 This option inhibits the use of 68881/68882 instructions that have to be
9226 emulated by software on the 68040. Use this option if your 68040 does not
9227 have code to emulate those instructions.
9231 Generate output for a 68060. This is the default when the compiler is
9232 configured for 68060-based systems.
9234 This option inhibits the use of 68020 and 68881/68882 instructions that
9235 have to be emulated by software on the 68060. Use this option if your 68060
9236 does not have code to emulate those instructions.
9240 Generate output for a CPU32. This is the default
9241 when the compiler is configured for CPU32-based systems.
9243 Use this option for microcontrollers with a
9244 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
9245 68336, 68340, 68341, 68349 and 68360.
9249 Generate output for a 520X ``coldfire'' family cpu. This is the default
9250 when the compiler is configured for 520X-based systems.
9252 Use this option for microcontroller with a 5200 core, including
9253 the MCF5202, MCF5203, MCF5204 and MCF5202.
9258 Generate output for a 68040, without using any of the new instructions.
9259 This results in code which can run relatively efficiently on either a
9260 68020/68881 or a 68030 or a 68040. The generated code does use the
9261 68881 instructions that are emulated on the 68040.
9265 Generate output for a 68060, without using any of the new instructions.
9266 This results in code which can run relatively efficiently on either a
9267 68020/68881 or a 68030 or a 68040. The generated code does use the
9268 68881 instructions that are emulated on the 68060.
9271 @opindex msoft-float
9272 Generate output containing library calls for floating point.
9273 @strong{Warning:} the requisite libraries are not available for all m68k
9274 targets. Normally the facilities of the machine's usual C compiler are
9275 used, but this can't be done directly in cross-compilation. You must
9276 make your own arrangements to provide suitable library functions for
9277 cross-compilation. The embedded targets @samp{m68k-*-aout} and
9278 @samp{m68k-*-coff} do provide software floating point support.
9282 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9283 Additionally, parameters passed on the stack are also aligned to a
9284 16-bit boundary even on targets whose API mandates promotion to 32-bit.
9287 @opindex mnobitfield
9288 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
9289 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
9293 Do use the bit-field instructions. The @option{-m68020} option implies
9294 @option{-mbitfield}. This is the default if you use a configuration
9295 designed for a 68020.
9299 Use a different function-calling convention, in which functions
9300 that take a fixed number of arguments return with the @code{rtd}
9301 instruction, which pops their arguments while returning. This
9302 saves one instruction in the caller since there is no need to pop
9303 the arguments there.
9305 This calling convention is incompatible with the one normally
9306 used on Unix, so you cannot use it if you need to call libraries
9307 compiled with the Unix compiler.
9309 Also, you must provide function prototypes for all functions that
9310 take variable numbers of arguments (including @code{printf});
9311 otherwise incorrect code will be generated for calls to those
9314 In addition, seriously incorrect code will result if you call a
9315 function with too many arguments. (Normally, extra arguments are
9316 harmlessly ignored.)
9318 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
9319 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
9322 @itemx -mno-align-int
9324 @opindex mno-align-int
9325 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
9326 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
9327 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
9328 Aligning variables on 32-bit boundaries produces code that runs somewhat
9329 faster on processors with 32-bit busses at the expense of more memory.
9331 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
9332 align structures containing the above types differently than
9333 most published application binary interface specifications for the m68k.
9337 Use the pc-relative addressing mode of the 68000 directly, instead of
9338 using a global offset table. At present, this option implies @option{-fpic},
9339 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
9340 not presently supported with @option{-mpcrel}, though this could be supported for
9341 68020 and higher processors.
9343 @item -mno-strict-align
9344 @itemx -mstrict-align
9345 @opindex mno-strict-align
9346 @opindex mstrict-align
9347 Do not (do) assume that unaligned memory references will be handled by
9351 Generate code that allows the data segment to be located in a different
9352 area of memory from the text segment. This allows for execute in place in
9353 an environment without virtual memory management. This option implies
9357 Generate code that assumes that the data segment follows the text segment.
9358 This is the default.
9360 @item -mid-shared-library
9361 Generate code that supports shared libraries via the library ID method.
9362 This allows for execute in place and shared libraries in an environment
9363 without virtual memory management. This option implies @option{-fPIC}.
9365 @item -mno-id-shared-library
9366 Generate code that doesn't assume ID based shared libraries are being used.
9367 This is the default.
9369 @item -mshared-library-id=n
9370 Specified the identification number of the ID based shared library being
9371 compiled. Specifying a value of 0 will generate more compact code, specifying
9372 other values will force the allocation of that number to the current
9373 library but is no more space or time efficient than omitting this option.
9377 @node M68hc1x Options
9378 @subsection M68hc1x Options
9379 @cindex M68hc1x options
9381 These are the @samp{-m} options defined for the 68hc11 and 68hc12
9382 microcontrollers. The default values for these options depends on
9383 which style of microcontroller was selected when the compiler was configured;
9384 the defaults for the most common choices are given below.
9391 Generate output for a 68HC11. This is the default
9392 when the compiler is configured for 68HC11-based systems.
9398 Generate output for a 68HC12. This is the default
9399 when the compiler is configured for 68HC12-based systems.
9405 Generate output for a 68HCS12.
9408 @opindex mauto-incdec
9409 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
9416 Enable the use of 68HC12 min and max instructions.
9419 @itemx -mno-long-calls
9420 @opindex mlong-calls
9421 @opindex mno-long-calls
9422 Treat all calls as being far away (near). If calls are assumed to be
9423 far away, the compiler will use the @code{call} instruction to
9424 call a function and the @code{rtc} instruction for returning.
9428 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9430 @item -msoft-reg-count=@var{count}
9431 @opindex msoft-reg-count
9432 Specify the number of pseudo-soft registers which are used for the
9433 code generation. The maximum number is 32. Using more pseudo-soft
9434 register may or may not result in better code depending on the program.
9435 The default is 4 for 68HC11 and 2 for 68HC12.
9440 @subsection MCore Options
9441 @cindex MCore options
9443 These are the @samp{-m} options defined for the Motorola M*Core
9451 @opindex mno-hardlit
9452 Inline constants into the code stream if it can be done in two
9453 instructions or less.
9459 Use the divide instruction. (Enabled by default).
9461 @item -mrelax-immediate
9462 @itemx -mno-relax-immediate
9463 @opindex mrelax-immediate
9464 @opindex mno-relax-immediate
9465 Allow arbitrary sized immediates in bit operations.
9467 @item -mwide-bitfields
9468 @itemx -mno-wide-bitfields
9469 @opindex mwide-bitfields
9470 @opindex mno-wide-bitfields
9471 Always treat bit-fields as int-sized.
9473 @item -m4byte-functions
9474 @itemx -mno-4byte-functions
9475 @opindex m4byte-functions
9476 @opindex mno-4byte-functions
9477 Force all functions to be aligned to a four byte boundary.
9479 @item -mcallgraph-data
9480 @itemx -mno-callgraph-data
9481 @opindex mcallgraph-data
9482 @opindex mno-callgraph-data
9483 Emit callgraph information.
9486 @itemx -mno-slow-bytes
9487 @opindex mslow-bytes
9488 @opindex mno-slow-bytes
9489 Prefer word access when reading byte quantities.
9491 @item -mlittle-endian
9493 @opindex mlittle-endian
9494 @opindex mbig-endian
9495 Generate code for a little endian target.
9501 Generate code for the 210 processor.
9505 @subsection MIPS Options
9506 @cindex MIPS options
9512 Generate big-endian code.
9516 Generate little-endian code. This is the default for @samp{mips*el-*-*}
9519 @item -march=@var{arch}
9521 Generate code that will run on @var{arch}, which can be the name of a
9522 generic MIPS ISA, or the name of a particular processor.
9524 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
9525 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
9526 The processor names are:
9527 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
9529 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
9530 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
9534 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
9535 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
9536 The special value @samp{from-abi} selects the
9537 most compatible architecture for the selected ABI (that is,
9538 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
9540 In processor names, a final @samp{000} can be abbreviated as @samp{k}
9541 (for example, @samp{-march=r2k}). Prefixes are optional, and
9542 @samp{vr} may be written @samp{r}.
9544 GCC defines two macros based on the value of this option. The first
9545 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
9546 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
9547 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
9548 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
9549 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
9551 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
9552 above. In other words, it will have the full prefix and will not
9553 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
9554 the macro names the resolved architecture (either @samp{"mips1"} or
9555 @samp{"mips3"}). It names the default architecture when no
9556 @option{-march} option is given.
9558 @item -mtune=@var{arch}
9560 Optimize for @var{arch}. Among other things, this option controls
9561 the way instructions are scheduled, and the perceived cost of arithmetic
9562 operations. The list of @var{arch} values is the same as for
9565 When this option is not used, GCC will optimize for the processor
9566 specified by @option{-march}. By using @option{-march} and
9567 @option{-mtune} together, it is possible to generate code that will
9568 run on a family of processors, but optimize the code for one
9569 particular member of that family.
9571 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
9572 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
9573 @samp{-march} ones described above.
9577 Equivalent to @samp{-march=mips1}.
9581 Equivalent to @samp{-march=mips2}.
9585 Equivalent to @samp{-march=mips3}.
9589 Equivalent to @samp{-march=mips4}.
9593 Equivalent to @samp{-march=mips32}.
9597 Equivalent to @samp{-march=mips32r2}.
9601 Equivalent to @samp{-march=mips64}.
9607 Use (do not use) the MIPS16 ISA@.
9619 Generate code for the given ABI@.
9621 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
9622 generates 64-bit code when you select a 64-bit architecture, but you
9623 can use @option{-mgp32} to get 32-bit code instead.
9625 For information about the O64 ABI, see
9626 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
9629 @itemx -mno-abicalls
9631 @opindex mno-abicalls
9632 Generate (do not generate) SVR4-style position-independent code.
9633 @option{-mabicalls} is the default for SVR4-based systems.
9639 Lift (do not lift) the usual restrictions on the size of the global
9642 GCC normally uses a single instruction to load values from the GOT@.
9643 While this is relatively efficient, it will only work if the GOT
9644 is smaller than about 64k. Anything larger will cause the linker
9645 to report an error such as:
9647 @cindex relocation truncated to fit (MIPS)
9649 relocation truncated to fit: R_MIPS_GOT16 foobar
9652 If this happens, you should recompile your code with @option{-mxgot}.
9653 It should then work with very large GOTs, although it will also be
9654 less efficient, since it will take three instructions to fetch the
9655 value of a global symbol.
9657 Note that some linkers can create multiple GOTs. If you have such a
9658 linker, you should only need to use @option{-mxgot} when a single object
9659 file accesses more than 64k's worth of GOT entries. Very few do.
9661 These options have no effect unless GCC is generating position
9666 Assume that general-purpose registers are 32 bits wide.
9670 Assume that general-purpose registers are 64 bits wide.
9674 Assume that floating-point registers are 32 bits wide.
9678 Assume that floating-point registers are 64 bits wide.
9681 @opindex mhard-float
9682 Use floating-point coprocessor instructions.
9685 @opindex msoft-float
9686 Do not use floating-point coprocessor instructions. Implement
9687 floating-point calculations using library calls instead.
9689 @item -msingle-float
9690 @opindex msingle-float
9691 Assume that the floating-point coprocessor only supports single-precision
9694 @itemx -mdouble-float
9695 @opindex mdouble-float
9696 Assume that the floating-point coprocessor supports double-precision
9697 operations. This is the default.
9699 @itemx -mpaired-single
9700 @itemx -mno-paired-single
9701 @opindex mpaired-single
9702 @opindex mno-paired-single
9703 Use (do not use) paired-single floating-point instructions.
9704 @xref{MIPS Paired-Single Support}. This option can only be used
9705 when generating 64-bit code and requires hardware floating-point
9706 support to be enabled.
9712 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
9713 The option @option{-mips3d} implies @option{-mpaired-single}.
9717 Force @code{int} and @code{long} types to be 64 bits wide. See
9718 @option{-mlong32} for an explanation of the default and the way
9719 that the pointer size is determined.
9721 This option has been deprecated and will be removed in a future release.
9725 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
9726 an explanation of the default and the way that the pointer size is
9731 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
9733 The default size of @code{int}s, @code{long}s and pointers depends on
9734 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
9735 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
9736 32-bit @code{long}s. Pointers are the same size as @code{long}s,
9737 or the same size as integer registers, whichever is smaller.
9743 Assume (do not assume) that all symbols have 32-bit values, regardless
9744 of the selected ABI@. This option is useful in combination with
9745 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
9746 to generate shorter and faster references to symbolic addresses.
9750 @cindex smaller data references (MIPS)
9751 @cindex gp-relative references (MIPS)
9752 Put global and static items less than or equal to @var{num} bytes into
9753 the small data or bss section instead of the normal data or bss section.
9754 This allows the data to be accessed using a single instruction.
9756 All modules should be compiled with the same @option{-G @var{num}}
9759 @item -membedded-data
9760 @itemx -mno-embedded-data
9761 @opindex membedded-data
9762 @opindex mno-embedded-data
9763 Allocate variables to the read-only data section first if possible, then
9764 next in the small data section if possible, otherwise in data. This gives
9765 slightly slower code than the default, but reduces the amount of RAM required
9766 when executing, and thus may be preferred for some embedded systems.
9768 @item -muninit-const-in-rodata
9769 @itemx -mno-uninit-const-in-rodata
9770 @opindex muninit-const-in-rodata
9771 @opindex mno-uninit-const-in-rodata
9772 Put uninitialized @code{const} variables in the read-only data section.
9773 This option is only meaningful in conjunction with @option{-membedded-data}.
9775 @item -msplit-addresses
9776 @itemx -mno-split-addresses
9777 @opindex msplit-addresses
9778 @opindex mno-split-addresses
9779 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
9780 relocation operators. This option has been superceded by
9781 @option{-mexplicit-relocs} but is retained for backwards compatibility.
9783 @item -mexplicit-relocs
9784 @itemx -mno-explicit-relocs
9785 @opindex mexplicit-relocs
9786 @opindex mno-explicit-relocs
9787 Use (do not use) assembler relocation operators when dealing with symbolic
9788 addresses. The alternative, selected by @option{-mno-explicit-relocs},
9789 is to use assembler macros instead.
9791 @option{-mexplicit-relocs} is the default if GCC was configured
9792 to use an assembler that supports relocation operators.
9794 @item -mcheck-zero-division
9795 @itemx -mno-check-zero-division
9796 @opindex mcheck-zero-division
9797 @opindex mno-check-zero-division
9798 Trap (do not trap) on integer division by zero. The default is
9799 @option{-mcheck-zero-division}.
9801 @item -mdivide-traps
9802 @itemx -mdivide-breaks
9803 @opindex mdivide-traps
9804 @opindex mdivide-breaks
9805 MIPS systems check for division by zero by generating either a
9806 conditional trap or a break instruction. Using traps results in
9807 smaller code, but is only supported on MIPS II and later. Also, some
9808 versions of the Linux kernel have a bug that prevents trap from
9809 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
9810 allow conditional traps on architectures that support them and
9811 @option{-mdivide-breaks} to force the use of breaks.
9813 The default is usually @option{-mdivide-traps}, but this can be
9814 overridden at configure time using @option{--with-divide=breaks}.
9815 Divide-by-zero checks can be completely disabled using
9816 @option{-mno-check-zero-division}.
9822 Force (do not force) the use of @code{memcpy()} for non-trivial block
9823 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
9824 most constant-sized copies.
9827 @itemx -mno-long-calls
9828 @opindex mlong-calls
9829 @opindex mno-long-calls
9830 Disable (do not disable) use of the @code{jal} instruction. Calling
9831 functions using @code{jal} is more efficient but requires the caller
9832 and callee to be in the same 256 megabyte segment.
9834 This option has no effect on abicalls code. The default is
9835 @option{-mno-long-calls}.
9841 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
9842 instructions, as provided by the R4650 ISA@.
9845 @itemx -mno-fused-madd
9846 @opindex mfused-madd
9847 @opindex mno-fused-madd
9848 Enable (disable) use of the floating point multiply-accumulate
9849 instructions, when they are available. The default is
9850 @option{-mfused-madd}.
9852 When multiply-accumulate instructions are used, the intermediate
9853 product is calculated to infinite precision and is not subject to
9854 the FCSR Flush to Zero bit. This may be undesirable in some
9859 Tell the MIPS assembler to not run its preprocessor over user
9860 assembler files (with a @samp{.s} suffix) when assembling them.
9863 @itemx -mno-fix-r4000
9865 @opindex mno-fix-r4000
9866 Work around certain R4000 CPU errata:
9869 A double-word or a variable shift may give an incorrect result if executed
9870 immediately after starting an integer division.
9872 A double-word or a variable shift may give an incorrect result if executed
9873 while an integer multiplication is in progress.
9875 An integer division may give an incorrect result if started in a delay slot
9876 of a taken branch or a jump.
9880 @itemx -mno-fix-r4400
9882 @opindex mno-fix-r4400
9883 Work around certain R4400 CPU errata:
9886 A double-word or a variable shift may give an incorrect result if executed
9887 immediately after starting an integer division.
9891 @itemx -mno-fix-vr4120
9892 @opindex mfix-vr4120
9893 Work around certain VR4120 errata:
9896 @code{dmultu} does not always produce the correct result.
9898 @code{div} and @code{ddiv} do not always produce the correct result if one
9899 of the operands is negative.
9901 The workarounds for the division errata rely on special functions in
9902 @file{libgcc.a}. At present, these functions are only provided by
9903 the @code{mips64vr*-elf} configurations.
9905 Other VR4120 errata require a nop to be inserted between certain pairs of
9906 instructions. These errata are handled by the assembler, not by GCC itself.
9909 @opindex mfix-vr4130
9910 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
9911 workarounds are implemented by the assembler rather than by GCC,
9912 although GCC will avoid using @code{mflo} and @code{mfhi} if the
9913 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
9914 instructions are available instead.
9919 Work around certain SB-1 CPU core errata.
9920 (This flag currently works around the SB-1 revision 2
9921 ``F1'' and ``F2'' floating point errata.)
9923 @item -mflush-func=@var{func}
9924 @itemx -mno-flush-func
9925 @opindex mflush-func
9926 Specifies the function to call to flush the I and D caches, or to not
9927 call any such function. If called, the function must take the same
9928 arguments as the common @code{_flush_func()}, that is, the address of the
9929 memory range for which the cache is being flushed, the size of the
9930 memory range, and the number 3 (to flush both caches). The default
9931 depends on the target GCC was configured for, but commonly is either
9932 @samp{_flush_func} or @samp{__cpu_flush}.
9934 @item -mbranch-likely
9935 @itemx -mno-branch-likely
9936 @opindex mbranch-likely
9937 @opindex mno-branch-likely
9938 Enable or disable use of Branch Likely instructions, regardless of the
9939 default for the selected architecture. By default, Branch Likely
9940 instructions may be generated if they are supported by the selected
9941 architecture. An exception is for the MIPS32 and MIPS64 architectures
9942 and processors which implement those architectures; for those, Branch
9943 Likely instructions will not be generated by default because the MIPS32
9944 and MIPS64 architectures specifically deprecate their use.
9946 @item -mfp-exceptions
9947 @itemx -mno-fp-exceptions
9948 @opindex mfp-exceptions
9949 Specifies whether FP exceptions are enabled. This affects how we schedule
9950 FP instructions for some processors. The default is that FP exceptions are
9953 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
9954 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
9957 @item -mvr4130-align
9958 @itemx -mno-vr4130-align
9959 @opindex mvr4130-align
9960 The VR4130 pipeline is two-way superscalar, but can only issue two
9961 instructions together if the first one is 8-byte aligned. When this
9962 option is enabled, GCC will align pairs of instructions that it
9963 thinks should execute in parallel.
9965 This option only has an effect when optimizing for the VR4130.
9966 It normally makes code faster, but at the expense of making it bigger.
9967 It is enabled by default at optimization level @option{-O3}.
9971 @subsection MMIX Options
9972 @cindex MMIX Options
9974 These options are defined for the MMIX:
9978 @itemx -mno-libfuncs
9980 @opindex mno-libfuncs
9981 Specify that intrinsic library functions are being compiled, passing all
9982 values in registers, no matter the size.
9987 @opindex mno-epsilon
9988 Generate floating-point comparison instructions that compare with respect
9989 to the @code{rE} epsilon register.
9991 @item -mabi=mmixware
9993 @opindex mabi-mmixware
9995 Generate code that passes function parameters and return values that (in
9996 the called function) are seen as registers @code{$0} and up, as opposed to
9997 the GNU ABI which uses global registers @code{$231} and up.
10000 @itemx -mno-zero-extend
10001 @opindex mzero-extend
10002 @opindex mno-zero-extend
10003 When reading data from memory in sizes shorter than 64 bits, use (do not
10004 use) zero-extending load instructions by default, rather than
10005 sign-extending ones.
10008 @itemx -mno-knuthdiv
10010 @opindex mno-knuthdiv
10011 Make the result of a division yielding a remainder have the same sign as
10012 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10013 remainder follows the sign of the dividend. Both methods are
10014 arithmetically valid, the latter being almost exclusively used.
10016 @item -mtoplevel-symbols
10017 @itemx -mno-toplevel-symbols
10018 @opindex mtoplevel-symbols
10019 @opindex mno-toplevel-symbols
10020 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10021 code can be used with the @code{PREFIX} assembly directive.
10025 Generate an executable in the ELF format, rather than the default
10026 @samp{mmo} format used by the @command{mmix} simulator.
10028 @item -mbranch-predict
10029 @itemx -mno-branch-predict
10030 @opindex mbranch-predict
10031 @opindex mno-branch-predict
10032 Use (do not use) the probable-branch instructions, when static branch
10033 prediction indicates a probable branch.
10035 @item -mbase-addresses
10036 @itemx -mno-base-addresses
10037 @opindex mbase-addresses
10038 @opindex mno-base-addresses
10039 Generate (do not generate) code that uses @emph{base addresses}. Using a
10040 base address automatically generates a request (handled by the assembler
10041 and the linker) for a constant to be set up in a global register. The
10042 register is used for one or more base address requests within the range 0
10043 to 255 from the value held in the register. The generally leads to short
10044 and fast code, but the number of different data items that can be
10045 addressed is limited. This means that a program that uses lots of static
10046 data may require @option{-mno-base-addresses}.
10048 @item -msingle-exit
10049 @itemx -mno-single-exit
10050 @opindex msingle-exit
10051 @opindex mno-single-exit
10052 Force (do not force) generated code to have a single exit point in each
10056 @node MN10300 Options
10057 @subsection MN10300 Options
10058 @cindex MN10300 options
10060 These @option{-m} options are defined for Matsushita MN10300 architectures:
10065 Generate code to avoid bugs in the multiply instructions for the MN10300
10066 processors. This is the default.
10068 @item -mno-mult-bug
10069 @opindex mno-mult-bug
10070 Do not generate code to avoid bugs in the multiply instructions for the
10071 MN10300 processors.
10075 Generate code which uses features specific to the AM33 processor.
10079 Do not generate code which uses features specific to the AM33 processor. This
10084 Do not link in the C run-time initialization object file.
10088 Indicate to the linker that it should perform a relaxation optimization pass
10089 to shorten branches, calls and absolute memory addresses. This option only
10090 has an effect when used on the command line for the final link step.
10092 This option makes symbolic debugging impossible.
10095 @node NS32K Options
10096 @subsection NS32K Options
10097 @cindex NS32K options
10099 These are the @samp{-m} options defined for the 32000 series. The default
10100 values for these options depends on which style of 32000 was selected when
10101 the compiler was configured; the defaults for the most common choices are
10109 Generate output for a 32032. This is the default
10110 when the compiler is configured for 32032 and 32016 based systems.
10116 Generate output for a 32332. This is the default
10117 when the compiler is configured for 32332-based systems.
10123 Generate output for a 32532. This is the default
10124 when the compiler is configured for 32532-based systems.
10128 Generate output containing 32081 instructions for floating point.
10129 This is the default for all systems.
10133 Generate output containing 32381 instructions for floating point. This
10134 also implies @option{-m32081}. The 32381 is only compatible with the 32332
10135 and 32532 cpus. This is the default for the pc532-netbsd configuration.
10138 @opindex mmulti-add
10139 Try and generate multiply-add floating point instructions @code{polyF}
10140 and @code{dotF}. This option is only available if the @option{-m32381}
10141 option is in effect. Using these instructions requires changes to
10142 register allocation which generally has a negative impact on
10143 performance. This option should only be enabled when compiling code
10144 particularly likely to make heavy use of multiply-add instructions.
10146 @item -mnomulti-add
10147 @opindex mnomulti-add
10148 Do not try and generate multiply-add floating point instructions
10149 @code{polyF} and @code{dotF}. This is the default on all platforms.
10152 @opindex msoft-float
10153 Generate output containing library calls for floating point.
10154 @strong{Warning:} the requisite libraries may not be available.
10156 @item -mieee-compare
10157 @itemx -mno-ieee-compare
10158 @opindex mieee-compare
10159 @opindex mno-ieee-compare
10160 Control whether or not the compiler uses IEEE floating point
10161 comparisons. These handle correctly the case where the result of a
10162 comparison is unordered.
10163 @strong{Warning:} the requisite kernel support may not be available.
10166 @opindex mnobitfield
10167 Do not use the bit-field instructions. On some machines it is faster to
10168 use shifting and masking operations. This is the default for the pc532.
10172 Do use the bit-field instructions. This is the default for all platforms
10177 Use a different function-calling convention, in which functions
10178 that take a fixed number of arguments return pop their
10179 arguments on return with the @code{ret} instruction.
10181 This calling convention is incompatible with the one normally
10182 used on Unix, so you cannot use it if you need to call libraries
10183 compiled with the Unix compiler.
10185 Also, you must provide function prototypes for all functions that
10186 take variable numbers of arguments (including @code{printf});
10187 otherwise incorrect code will be generated for calls to those
10190 In addition, seriously incorrect code will result if you call a
10191 function with too many arguments. (Normally, extra arguments are
10192 harmlessly ignored.)
10194 This option takes its name from the 680x0 @code{rtd} instruction.
10199 Use a different function-calling convention where the first two arguments
10200 are passed in registers.
10202 This calling convention is incompatible with the one normally
10203 used on Unix, so you cannot use it if you need to call libraries
10204 compiled with the Unix compiler.
10207 @opindex mnoregparam
10208 Do not pass any arguments in registers. This is the default for all
10213 It is OK to use the sb as an index register which is always loaded with
10214 zero. This is the default for the pc532-netbsd target.
10218 The sb register is not available for use or has not been initialized to
10219 zero by the run time system. This is the default for all targets except
10220 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
10221 @option{-fpic} is set.
10225 Many ns32000 series addressing modes use displacements of up to 512MB@.
10226 If an address is above 512MB then displacements from zero can not be used.
10227 This option causes code to be generated which can be loaded above 512MB@.
10228 This may be useful for operating systems or ROM code.
10232 Assume code will be loaded in the first 512MB of virtual address space.
10233 This is the default for all platforms.
10237 @node PDP-11 Options
10238 @subsection PDP-11 Options
10239 @cindex PDP-11 Options
10241 These options are defined for the PDP-11:
10246 Use hardware FPP floating point. This is the default. (FIS floating
10247 point on the PDP-11/40 is not supported.)
10250 @opindex msoft-float
10251 Do not use hardware floating point.
10255 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10259 Return floating-point results in memory. This is the default.
10263 Generate code for a PDP-11/40.
10267 Generate code for a PDP-11/45. This is the default.
10271 Generate code for a PDP-11/10.
10273 @item -mbcopy-builtin
10274 @opindex bcopy-builtin
10275 Use inline @code{movmemhi} patterns for copying memory. This is the
10280 Do not use inline @code{movmemhi} patterns for copying memory.
10286 Use 16-bit @code{int}. This is the default.
10292 Use 32-bit @code{int}.
10295 @itemx -mno-float32
10297 @opindex mno-float32
10298 Use 64-bit @code{float}. This is the default.
10301 @itemx -mno-float64
10303 @opindex mno-float64
10304 Use 32-bit @code{float}.
10308 Use @code{abshi2} pattern. This is the default.
10312 Do not use @code{abshi2} pattern.
10314 @item -mbranch-expensive
10315 @opindex mbranch-expensive
10316 Pretend that branches are expensive. This is for experimenting with
10317 code generation only.
10319 @item -mbranch-cheap
10320 @opindex mbranch-cheap
10321 Do not pretend that branches are expensive. This is the default.
10325 Generate code for a system with split I&D@.
10329 Generate code for a system without split I&D@. This is the default.
10333 Use Unix assembler syntax. This is the default when configured for
10334 @samp{pdp11-*-bsd}.
10338 Use DEC assembler syntax. This is the default when configured for any
10339 PDP-11 target other than @samp{pdp11-*-bsd}.
10342 @node PowerPC Options
10343 @subsection PowerPC Options
10344 @cindex PowerPC options
10346 These are listed under @xref{RS/6000 and PowerPC Options}.
10348 @node RS/6000 and PowerPC Options
10349 @subsection IBM RS/6000 and PowerPC Options
10350 @cindex RS/6000 and PowerPC Options
10351 @cindex IBM RS/6000 and PowerPC Options
10353 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
10360 @itemx -mno-powerpc
10361 @itemx -mpowerpc-gpopt
10362 @itemx -mno-powerpc-gpopt
10363 @itemx -mpowerpc-gfxopt
10364 @itemx -mno-powerpc-gfxopt
10366 @itemx -mno-powerpc64
10370 @opindex mno-power2
10372 @opindex mno-powerpc
10373 @opindex mpowerpc-gpopt
10374 @opindex mno-powerpc-gpopt
10375 @opindex mpowerpc-gfxopt
10376 @opindex mno-powerpc-gfxopt
10377 @opindex mpowerpc64
10378 @opindex mno-powerpc64
10379 GCC supports two related instruction set architectures for the
10380 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
10381 instructions supported by the @samp{rios} chip set used in the original
10382 RS/6000 systems and the @dfn{PowerPC} instruction set is the
10383 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
10384 the IBM 4xx microprocessors.
10386 Neither architecture is a subset of the other. However there is a
10387 large common subset of instructions supported by both. An MQ
10388 register is included in processors supporting the POWER architecture.
10390 You use these options to specify which instructions are available on the
10391 processor you are using. The default value of these options is
10392 determined when configuring GCC@. Specifying the
10393 @option{-mcpu=@var{cpu_type}} overrides the specification of these
10394 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
10395 rather than the options listed above.
10397 The @option{-mpower} option allows GCC to generate instructions that
10398 are found only in the POWER architecture and to use the MQ register.
10399 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
10400 to generate instructions that are present in the POWER2 architecture but
10401 not the original POWER architecture.
10403 The @option{-mpowerpc} option allows GCC to generate instructions that
10404 are found only in the 32-bit subset of the PowerPC architecture.
10405 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
10406 GCC to use the optional PowerPC architecture instructions in the
10407 General Purpose group, including floating-point square root. Specifying
10408 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
10409 use the optional PowerPC architecture instructions in the Graphics
10410 group, including floating-point select.
10412 The @option{-mpowerpc64} option allows GCC to generate the additional
10413 64-bit instructions that are found in the full PowerPC64 architecture
10414 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
10415 @option{-mno-powerpc64}.
10417 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
10418 will use only the instructions in the common subset of both
10419 architectures plus some special AIX common-mode calls, and will not use
10420 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
10421 permits GCC to use any instruction from either architecture and to
10422 allow use of the MQ register; specify this for the Motorola MPC601.
10424 @item -mnew-mnemonics
10425 @itemx -mold-mnemonics
10426 @opindex mnew-mnemonics
10427 @opindex mold-mnemonics
10428 Select which mnemonics to use in the generated assembler code. With
10429 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
10430 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
10431 assembler mnemonics defined for the POWER architecture. Instructions
10432 defined in only one architecture have only one mnemonic; GCC uses that
10433 mnemonic irrespective of which of these options is specified.
10435 GCC defaults to the mnemonics appropriate for the architecture in
10436 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
10437 value of these option. Unless you are building a cross-compiler, you
10438 should normally not specify either @option{-mnew-mnemonics} or
10439 @option{-mold-mnemonics}, but should instead accept the default.
10441 @item -mcpu=@var{cpu_type}
10443 Set architecture type, register usage, choice of mnemonics, and
10444 instruction scheduling parameters for machine type @var{cpu_type}.
10445 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
10446 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
10447 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
10448 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
10449 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
10450 @samp{860}, @samp{970}, @samp{8540}, @samp{common}, @samp{ec603e}, @samp{G3},
10451 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
10452 @samp{power4}, @samp{power5}, @samp{powerpc}, @samp{powerpc64},
10453 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64a}.
10455 @option{-mcpu=common} selects a completely generic processor. Code
10456 generated under this option will run on any POWER or PowerPC processor.
10457 GCC will use only the instructions in the common subset of both
10458 architectures, and will not use the MQ register. GCC assumes a generic
10459 processor model for scheduling purposes.
10461 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
10462 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
10463 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
10464 types, with an appropriate, generic processor model assumed for
10465 scheduling purposes.
10467 The other options specify a specific processor. Code generated under
10468 those options will run best on that processor, and may not run at all on
10471 The @option{-mcpu} options automatically enable or disable the
10472 following options: @option{-maltivec}, @option{-mhard-float},
10473 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
10474 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
10475 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
10476 @option{-mstring}. The particular options set for any particular CPU
10477 will vary between compiler versions, depending on what setting seems
10478 to produce optimal code for that CPU; it doesn't necessarily reflect
10479 the actual hardware's capabilities. If you wish to set an individual
10480 option to a particular value, you may specify it after the
10481 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
10483 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
10484 not enabled or disabled by the @option{-mcpu} option at present, since
10485 AIX does not have full support for these options. You may still
10486 enable or disable them individually if you're sure it'll work in your
10489 @item -mtune=@var{cpu_type}
10491 Set the instruction scheduling parameters for machine type
10492 @var{cpu_type}, but do not set the architecture type, register usage, or
10493 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
10494 values for @var{cpu_type} are used for @option{-mtune} as for
10495 @option{-mcpu}. If both are specified, the code generated will use the
10496 architecture, registers, and mnemonics set by @option{-mcpu}, but the
10497 scheduling parameters set by @option{-mtune}.
10500 @itemx -mno-altivec
10502 @opindex mno-altivec
10503 Generate code that uses (does not use) AltiVec instructions, and also
10504 enable the use of built-in functions that allow more direct access to
10505 the AltiVec instruction set. You may also need to set
10506 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
10511 Extend the current ABI with SPE ABI extensions. This does not change
10512 the default ABI, instead it adds the SPE ABI extensions to the current
10516 @opindex mabi=no-spe
10517 Disable Booke SPE ABI extensions for the current ABI@.
10519 @item -misel=@var{yes/no}
10522 This switch enables or disables the generation of ISEL instructions.
10524 @item -mspe=@var{yes/no}
10527 This switch enables or disables the generation of SPE simd
10530 @item -mfloat-gprs=@var{yes/single/double/no}
10531 @itemx -mfloat-gprs
10532 @opindex mfloat-gprs
10533 This switch enables or disables the generation of floating point
10534 operations on the general purpose registers for architectures that
10537 The argument @var{yes} or @var{single} enables the use of
10538 single-precision floating point operations.
10540 The argument @var{double} enables the use of single and
10541 double-precision floating point operations.
10543 The argument @var{no} disables floating point operations on the
10544 general purpose registers.
10546 This option is currently only available on the MPC854x.
10552 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
10553 targets (including GNU/Linux). The 32-bit environment sets int, long
10554 and pointer to 32 bits and generates code that runs on any PowerPC
10555 variant. The 64-bit environment sets int to 32 bits and long and
10556 pointer to 64 bits, and generates code for PowerPC64, as for
10557 @option{-mpowerpc64}.
10560 @itemx -mno-fp-in-toc
10561 @itemx -mno-sum-in-toc
10562 @itemx -mminimal-toc
10564 @opindex mno-fp-in-toc
10565 @opindex mno-sum-in-toc
10566 @opindex mminimal-toc
10567 Modify generation of the TOC (Table Of Contents), which is created for
10568 every executable file. The @option{-mfull-toc} option is selected by
10569 default. In that case, GCC will allocate at least one TOC entry for
10570 each unique non-automatic variable reference in your program. GCC
10571 will also place floating-point constants in the TOC@. However, only
10572 16,384 entries are available in the TOC@.
10574 If you receive a linker error message that saying you have overflowed
10575 the available TOC space, you can reduce the amount of TOC space used
10576 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
10577 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
10578 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
10579 generate code to calculate the sum of an address and a constant at
10580 run-time instead of putting that sum into the TOC@. You may specify one
10581 or both of these options. Each causes GCC to produce very slightly
10582 slower and larger code at the expense of conserving TOC space.
10584 If you still run out of space in the TOC even when you specify both of
10585 these options, specify @option{-mminimal-toc} instead. This option causes
10586 GCC to make only one TOC entry for every file. When you specify this
10587 option, GCC will produce code that is slower and larger but which
10588 uses extremely little TOC space. You may wish to use this option
10589 only on files that contain less frequently executed code.
10595 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
10596 @code{long} type, and the infrastructure needed to support them.
10597 Specifying @option{-maix64} implies @option{-mpowerpc64} and
10598 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
10599 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
10602 @itemx -mno-xl-compat
10603 @opindex mxl-compat
10604 @opindex mno-xl-compat
10605 Produce code that conforms more closely to IBM XLC semantics when using
10606 AIX-compatible ABI. Pass floating-point arguments to prototyped
10607 functions beyond the register save area (RSA) on the stack in addition
10608 to argument FPRs. Do not assume that most significant double in 128
10609 bit long double value is properly rounded when comparing values.
10611 The AIX calling convention was extended but not initially documented to
10612 handle an obscure K&R C case of calling a function that takes the
10613 address of its arguments with fewer arguments than declared. AIX XL
10614 compilers access floating point arguments which do not fit in the
10615 RSA from the stack when a subroutine is compiled without
10616 optimization. Because always storing floating-point arguments on the
10617 stack is inefficient and rarely needed, this option is not enabled by
10618 default and only is necessary when calling subroutines compiled by AIX
10619 XL compilers without optimization.
10623 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
10624 application written to use message passing with special startup code to
10625 enable the application to run. The system must have PE installed in the
10626 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
10627 must be overridden with the @option{-specs=} option to specify the
10628 appropriate directory location. The Parallel Environment does not
10629 support threads, so the @option{-mpe} option and the @option{-pthread}
10630 option are incompatible.
10632 @item -malign-natural
10633 @itemx -malign-power
10634 @opindex malign-natural
10635 @opindex malign-power
10636 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
10637 @option{-malign-natural} overrides the ABI-defined alignment of larger
10638 types, such as floating-point doubles, on their natural size-based boundary.
10639 The option @option{-malign-power} instructs GCC to follow the ABI-specified
10640 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
10642 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
10646 @itemx -mhard-float
10647 @opindex msoft-float
10648 @opindex mhard-float
10649 Generate code that does not use (uses) the floating-point register set.
10650 Software floating point emulation is provided if you use the
10651 @option{-msoft-float} option, and pass the option to GCC when linking.
10654 @itemx -mno-multiple
10656 @opindex mno-multiple
10657 Generate code that uses (does not use) the load multiple word
10658 instructions and the store multiple word instructions. These
10659 instructions are generated by default on POWER systems, and not
10660 generated on PowerPC systems. Do not use @option{-mmultiple} on little
10661 endian PowerPC systems, since those instructions do not work when the
10662 processor is in little endian mode. The exceptions are PPC740 and
10663 PPC750 which permit the instructions usage in little endian mode.
10668 @opindex mno-string
10669 Generate code that uses (does not use) the load string instructions
10670 and the store string word instructions to save multiple registers and
10671 do small block moves. These instructions are generated by default on
10672 POWER systems, and not generated on PowerPC systems. Do not use
10673 @option{-mstring} on little endian PowerPC systems, since those
10674 instructions do not work when the processor is in little endian mode.
10675 The exceptions are PPC740 and PPC750 which permit the instructions
10676 usage in little endian mode.
10681 @opindex mno-update
10682 Generate code that uses (does not use) the load or store instructions
10683 that update the base register to the address of the calculated memory
10684 location. These instructions are generated by default. If you use
10685 @option{-mno-update}, there is a small window between the time that the
10686 stack pointer is updated and the address of the previous frame is
10687 stored, which means code that walks the stack frame across interrupts or
10688 signals may get corrupted data.
10691 @itemx -mno-fused-madd
10692 @opindex mfused-madd
10693 @opindex mno-fused-madd
10694 Generate code that uses (does not use) the floating point multiply and
10695 accumulate instructions. These instructions are generated by default if
10696 hardware floating is used.
10698 @item -mno-bit-align
10700 @opindex mno-bit-align
10701 @opindex mbit-align
10702 On System V.4 and embedded PowerPC systems do not (do) force structures
10703 and unions that contain bit-fields to be aligned to the base type of the
10706 For example, by default a structure containing nothing but 8
10707 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
10708 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
10709 the structure would be aligned to a 1 byte boundary and be one byte in
10712 @item -mno-strict-align
10713 @itemx -mstrict-align
10714 @opindex mno-strict-align
10715 @opindex mstrict-align
10716 On System V.4 and embedded PowerPC systems do not (do) assume that
10717 unaligned memory references will be handled by the system.
10719 @item -mrelocatable
10720 @itemx -mno-relocatable
10721 @opindex mrelocatable
10722 @opindex mno-relocatable
10723 On embedded PowerPC systems generate code that allows (does not allow)
10724 the program to be relocated to a different address at runtime. If you
10725 use @option{-mrelocatable} on any module, all objects linked together must
10726 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
10728 @item -mrelocatable-lib
10729 @itemx -mno-relocatable-lib
10730 @opindex mrelocatable-lib
10731 @opindex mno-relocatable-lib
10732 On embedded PowerPC systems generate code that allows (does not allow)
10733 the program to be relocated to a different address at runtime. Modules
10734 compiled with @option{-mrelocatable-lib} can be linked with either modules
10735 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
10736 with modules compiled with the @option{-mrelocatable} options.
10742 On System V.4 and embedded PowerPC systems do not (do) assume that
10743 register 2 contains a pointer to a global area pointing to the addresses
10744 used in the program.
10747 @itemx -mlittle-endian
10749 @opindex mlittle-endian
10750 On System V.4 and embedded PowerPC systems compile code for the
10751 processor in little endian mode. The @option{-mlittle-endian} option is
10752 the same as @option{-mlittle}.
10755 @itemx -mbig-endian
10757 @opindex mbig-endian
10758 On System V.4 and embedded PowerPC systems compile code for the
10759 processor in big endian mode. The @option{-mbig-endian} option is
10760 the same as @option{-mbig}.
10762 @item -mdynamic-no-pic
10763 @opindex mdynamic-no-pic
10764 On Darwin and Mac OS X systems, compile code so that it is not
10765 relocatable, but that its external references are relocatable. The
10766 resulting code is suitable for applications, but not shared
10769 @item -mprioritize-restricted-insns=@var{priority}
10770 @opindex mprioritize-restricted-insns
10771 This option controls the priority that is assigned to
10772 dispatch-slot restricted instructions during the second scheduling
10773 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
10774 @var{no/highest/second-highest} priority to dispatch slot restricted
10777 @item -msched-costly-dep=@var{dependence_type}
10778 @opindex msched-costly-dep
10779 This option controls which dependences are considered costly
10780 by the target during instruction scheduling. The argument
10781 @var{dependence_type} takes one of the following values:
10782 @var{no}: no dependence is costly,
10783 @var{all}: all dependences are costly,
10784 @var{true_store_to_load}: a true dependence from store to load is costly,
10785 @var{store_to_load}: any dependence from store to load is costly,
10786 @var{number}: any dependence which latency >= @var{number} is costly.
10788 @item -minsert-sched-nops=@var{scheme}
10789 @opindex minsert-sched-nops
10790 This option controls which nop insertion scheme will be used during
10791 the second scheduling pass. The argument @var{scheme} takes one of the
10793 @var{no}: Don't insert nops.
10794 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
10795 according to the scheduler's grouping.
10796 @var{regroup_exact}: Insert nops to force costly dependent insns into
10797 separate groups. Insert exactly as many nops as needed to force an insn
10798 to a new group, according to the estimated processor grouping.
10799 @var{number}: Insert nops to force costly dependent insns into
10800 separate groups. Insert @var{number} nops to force an insn to a new group.
10803 @opindex mcall-sysv
10804 On System V.4 and embedded PowerPC systems compile code using calling
10805 conventions that adheres to the March 1995 draft of the System V
10806 Application Binary Interface, PowerPC processor supplement. This is the
10807 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
10809 @item -mcall-sysv-eabi
10810 @opindex mcall-sysv-eabi
10811 Specify both @option{-mcall-sysv} and @option{-meabi} options.
10813 @item -mcall-sysv-noeabi
10814 @opindex mcall-sysv-noeabi
10815 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
10817 @item -mcall-solaris
10818 @opindex mcall-solaris
10819 On System V.4 and embedded PowerPC systems compile code for the Solaris
10823 @opindex mcall-linux
10824 On System V.4 and embedded PowerPC systems compile code for the
10825 Linux-based GNU system.
10829 On System V.4 and embedded PowerPC systems compile code for the
10830 Hurd-based GNU system.
10832 @item -mcall-netbsd
10833 @opindex mcall-netbsd
10834 On System V.4 and embedded PowerPC systems compile code for the
10835 NetBSD operating system.
10837 @item -maix-struct-return
10838 @opindex maix-struct-return
10839 Return all structures in memory (as specified by the AIX ABI)@.
10841 @item -msvr4-struct-return
10842 @opindex msvr4-struct-return
10843 Return structures smaller than 8 bytes in registers (as specified by the
10846 @item -mabi=altivec
10847 @opindex mabi=altivec
10848 Extend the current ABI with AltiVec ABI extensions. This does not
10849 change the default ABI, instead it adds the AltiVec ABI extensions to
10852 @item -mabi=no-altivec
10853 @opindex mabi=no-altivec
10854 Disable AltiVec ABI extensions for the current ABI@.
10857 @itemx -mno-prototype
10858 @opindex mprototype
10859 @opindex mno-prototype
10860 On System V.4 and embedded PowerPC systems assume that all calls to
10861 variable argument functions are properly prototyped. Otherwise, the
10862 compiler must insert an instruction before every non prototyped call to
10863 set or clear bit 6 of the condition code register (@var{CR}) to
10864 indicate whether floating point values were passed in the floating point
10865 registers in case the function takes a variable arguments. With
10866 @option{-mprototype}, only calls to prototyped variable argument functions
10867 will set or clear the bit.
10871 On embedded PowerPC systems, assume that the startup module is called
10872 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
10873 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
10878 On embedded PowerPC systems, assume that the startup module is called
10879 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
10884 On embedded PowerPC systems, assume that the startup module is called
10885 @file{crt0.o} and the standard C libraries are @file{libads.a} and
10888 @item -myellowknife
10889 @opindex myellowknife
10890 On embedded PowerPC systems, assume that the startup module is called
10891 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
10896 On System V.4 and embedded PowerPC systems, specify that you are
10897 compiling for a VxWorks system.
10901 Specify that you are compiling for the WindISS simulation environment.
10905 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
10906 header to indicate that @samp{eabi} extended relocations are used.
10912 On System V.4 and embedded PowerPC systems do (do not) adhere to the
10913 Embedded Applications Binary Interface (eabi) which is a set of
10914 modifications to the System V.4 specifications. Selecting @option{-meabi}
10915 means that the stack is aligned to an 8 byte boundary, a function
10916 @code{__eabi} is called to from @code{main} to set up the eabi
10917 environment, and the @option{-msdata} option can use both @code{r2} and
10918 @code{r13} to point to two separate small data areas. Selecting
10919 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
10920 do not call an initialization function from @code{main}, and the
10921 @option{-msdata} option will only use @code{r13} to point to a single
10922 small data area. The @option{-meabi} option is on by default if you
10923 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
10926 @opindex msdata=eabi
10927 On System V.4 and embedded PowerPC systems, put small initialized
10928 @code{const} global and static data in the @samp{.sdata2} section, which
10929 is pointed to by register @code{r2}. Put small initialized
10930 non-@code{const} global and static data in the @samp{.sdata} section,
10931 which is pointed to by register @code{r13}. Put small uninitialized
10932 global and static data in the @samp{.sbss} section, which is adjacent to
10933 the @samp{.sdata} section. The @option{-msdata=eabi} option is
10934 incompatible with the @option{-mrelocatable} option. The
10935 @option{-msdata=eabi} option also sets the @option{-memb} option.
10938 @opindex msdata=sysv
10939 On System V.4 and embedded PowerPC systems, put small global and static
10940 data in the @samp{.sdata} section, which is pointed to by register
10941 @code{r13}. Put small uninitialized global and static data in the
10942 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
10943 The @option{-msdata=sysv} option is incompatible with the
10944 @option{-mrelocatable} option.
10946 @item -msdata=default
10948 @opindex msdata=default
10950 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
10951 compile code the same as @option{-msdata=eabi}, otherwise compile code the
10952 same as @option{-msdata=sysv}.
10955 @opindex msdata-data
10956 On System V.4 and embedded PowerPC systems, put small global and static
10957 data in the @samp{.sdata} section. Put small uninitialized global and
10958 static data in the @samp{.sbss} section. Do not use register @code{r13}
10959 to address small data however. This is the default behavior unless
10960 other @option{-msdata} options are used.
10964 @opindex msdata=none
10966 On embedded PowerPC systems, put all initialized global and static data
10967 in the @samp{.data} section, and all uninitialized data in the
10968 @samp{.bss} section.
10972 @cindex smaller data references (PowerPC)
10973 @cindex .sdata/.sdata2 references (PowerPC)
10974 On embedded PowerPC systems, put global and static items less than or
10975 equal to @var{num} bytes into the small data or bss sections instead of
10976 the normal data or bss section. By default, @var{num} is 8. The
10977 @option{-G @var{num}} switch is also passed to the linker.
10978 All modules should be compiled with the same @option{-G @var{num}} value.
10981 @itemx -mno-regnames
10983 @opindex mno-regnames
10984 On System V.4 and embedded PowerPC systems do (do not) emit register
10985 names in the assembly language output using symbolic forms.
10988 @itemx -mno-longcall
10990 @opindex mno-longcall
10991 Default to making all function calls indirectly, using a register, so
10992 that functions which reside further than 32 megabytes (33,554,432
10993 bytes) from the current location can be called. This setting can be
10994 overridden by the @code{shortcall} function attribute, or by
10995 @code{#pragma longcall(0)}.
10997 Some linkers are capable of detecting out-of-range calls and generating
10998 glue code on the fly. On these systems, long calls are unnecessary and
10999 generate slower code. As of this writing, the AIX linker can do this,
11000 as can the GNU linker for PowerPC/64. It is planned to add this feature
11001 to the GNU linker for 32-bit PowerPC systems as well.
11003 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11004 callee, L42'', plus a ``branch island'' (glue code). The two target
11005 addresses represent the callee and the ``branch island''. The
11006 Darwin/PPC linker will prefer the first address and generate a ``bl
11007 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11008 otherwise, the linker will generate ``bl L42'' to call the ``branch
11009 island''. The ``branch island'' is appended to the body of the
11010 calling function; it computes the full 32-bit address of the callee
11013 On Mach-O (Darwin) systems, this option directs the compiler emit to
11014 the glue for every direct call, and the Darwin linker decides whether
11015 to use or discard it.
11017 In the future, we may cause GCC to ignore all longcall specifications
11018 when the linker is known to generate glue.
11022 Adds support for multithreading with the @dfn{pthreads} library.
11023 This option sets flags for both the preprocessor and linker.
11027 @node S/390 and zSeries Options
11028 @subsection S/390 and zSeries Options
11029 @cindex S/390 and zSeries Options
11031 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11035 @itemx -msoft-float
11036 @opindex mhard-float
11037 @opindex msoft-float
11038 Use (do not use) the hardware floating-point instructions and registers
11039 for floating-point operations. When @option{-msoft-float} is specified,
11040 functions in @file{libgcc.a} will be used to perform floating-point
11041 operations. When @option{-mhard-float} is specified, the compiler
11042 generates IEEE floating-point instructions. This is the default.
11045 @itemx -mno-backchain
11046 @opindex mbackchain
11047 @opindex mno-backchain
11048 Store (do not store) the address of the caller's frame as backchain pointer
11049 into the callee's stack frame.
11050 A backchain may be needed to allow debugging using tools that do not understand
11051 DWARF-2 call frame information.
11052 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
11053 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
11054 the backchain is placed into the topmost word of the 96/160 byte register
11057 In general, code compiled with @option{-mbackchain} is call-compatible with
11058 code compiled with @option{-mmo-backchain}; however, use of the backchain
11059 for debugging purposes usually requires that the whole binary is built with
11060 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
11061 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11062 to build a linux kernel use @option{-msoft-float}.
11064 The default is to not maintain the backchain.
11066 @item -mpacked-stack
11067 @item -mno-packed-stack
11068 @opindex mpacked-stack
11069 @opindex mno-packed-stack
11070 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
11071 specified, the compiler uses the all fields of the 96/160 byte register save
11072 area only for their default purpose; unused fields still take up stack space.
11073 When @option{-mpacked-stack} is specified, register save slots are densely
11074 packed at the top of the register save area; unused space is reused for other
11075 purposes, allowing for more efficient use of the available stack space.
11076 However, when @option{-mbackchain} is also in effect, the topmost word of
11077 the save area is always used to store the backchain, and the return address
11078 register is always saved two words below the backchain.
11080 As long as the stack frame backchain is not used, code generated with
11081 @option{-mpacked-stack} is call-compatible with code generated with
11082 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
11083 S/390 or zSeries generated code that uses the stack frame backchain at run
11084 time, not just for debugging purposes. Such code is not call-compatible
11085 with code compiled with @option{-mpacked-stack}. Also, note that the
11086 combination of @option{-mbackchain},
11087 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11088 to build a linux kernel use @option{-msoft-float}.
11090 The default is to not use the packed stack layout.
11093 @itemx -mno-small-exec
11094 @opindex msmall-exec
11095 @opindex mno-small-exec
11096 Generate (or do not generate) code using the @code{bras} instruction
11097 to do subroutine calls.
11098 This only works reliably if the total executable size does not
11099 exceed 64k. The default is to use the @code{basr} instruction instead,
11100 which does not have this limitation.
11106 When @option{-m31} is specified, generate code compliant to the
11107 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
11108 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
11109 particular to generate 64-bit instructions. For the @samp{s390}
11110 targets, the default is @option{-m31}, while the @samp{s390x}
11111 targets default to @option{-m64}.
11117 When @option{-mzarch} is specified, generate code using the
11118 instructions available on z/Architecture.
11119 When @option{-mesa} is specified, generate code using the
11120 instructions available on ESA/390. Note that @option{-mesa} is
11121 not possible with @option{-m64}.
11122 When generating code compliant to the GNU/Linux for S/390 ABI,
11123 the default is @option{-mesa}. When generating code compliant
11124 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
11130 Generate (or do not generate) code using the @code{mvcle} instruction
11131 to perform block moves. When @option{-mno-mvcle} is specified,
11132 use a @code{mvc} loop instead. This is the default.
11138 Print (or do not print) additional debug information when compiling.
11139 The default is to not print debug information.
11141 @item -march=@var{cpu-type}
11143 Generate code that will run on @var{cpu-type}, which is the name of a system
11144 representing a certain processor type. Possible values for
11145 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
11146 When generating code using the instructions available on z/Architecture,
11147 the default is @option{-march=z900}. Otherwise, the default is
11148 @option{-march=g5}.
11150 @item -mtune=@var{cpu-type}
11152 Tune to @var{cpu-type} everything applicable about the generated code,
11153 except for the ABI and the set of available instructions.
11154 The list of @var{cpu-type} values is the same as for @option{-march}.
11155 The default is the value used for @option{-march}.
11158 @itemx -mno-tpf-trace
11159 @opindex mtpf-trace
11160 @opindex mno-tpf-trace
11161 Generate code that adds (does not add) in TPF OS specific branches to trace
11162 routines in the operating system. This option is off by default, even
11163 when compiling for the TPF OS@.
11166 @itemx -mno-fused-madd
11167 @opindex mfused-madd
11168 @opindex mno-fused-madd
11169 Generate code that uses (does not use) the floating point multiply and
11170 accumulate instructions. These instructions are generated by default if
11171 hardware floating point is used.
11173 @item -mwarn-framesize=@var{framesize}
11174 @opindex mwarn-framesize
11175 Emit a warning if the current function exceeds the given frame size. Because
11176 this is a compile time check it doesn't need to be a real problem when the program
11177 runs. It is intended to identify functions which most probably cause
11178 a stack overflow. It is useful to be used in an environment with limited stack
11179 size e.g.@: the linux kernel.
11181 @item -mwarn-dynamicstack
11182 @opindex mwarn-dynamicstack
11183 Emit a warning if the function calls alloca or uses dynamically
11184 sized arrays. This is generally a bad idea with a limited stack size.
11186 @item -mstack-guard=@var{stack-guard}
11187 @item -mstack-size=@var{stack-size}
11188 @opindex mstack-guard
11189 @opindex mstack-size
11190 These arguments always have to be used in conjunction. If they are present the s390
11191 back end emits additional instructions in the function prologue which trigger a trap
11192 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
11193 (remember that the stack on s390 grows downward). These options are intended to
11194 be used to help debugging stack overflow problems. The additionally emitted code
11195 cause only little overhead and hence can also be used in production like systems
11196 without greater performance degradation. The given values have to be exact
11197 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard}.
11198 In order to be efficient the extra code makes the assumption that the stack starts
11199 at an address aligned to the value given by @var{stack-size}.
11203 @subsection SH Options
11205 These @samp{-m} options are defined for the SH implementations:
11210 Generate code for the SH1.
11214 Generate code for the SH2.
11217 Generate code for the SH2e.
11221 Generate code for the SH3.
11225 Generate code for the SH3e.
11229 Generate code for the SH4 without a floating-point unit.
11231 @item -m4-single-only
11232 @opindex m4-single-only
11233 Generate code for the SH4 with a floating-point unit that only
11234 supports single-precision arithmetic.
11238 Generate code for the SH4 assuming the floating-point unit is in
11239 single-precision mode by default.
11243 Generate code for the SH4.
11247 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
11248 floating-point unit is not used.
11250 @item -m4a-single-only
11251 @opindex m4a-single-only
11252 Generate code for the SH4a, in such a way that no double-precision
11253 floating point operations are used.
11256 @opindex m4a-single
11257 Generate code for the SH4a assuming the floating-point unit is in
11258 single-precision mode by default.
11262 Generate code for the SH4a.
11266 Same as @option{-m4a-nofpu}, except that it implicitly passes
11267 @option{-dsp} to the assembler. GCC doesn't generate any DSP
11268 instructions at the moment.
11272 Compile code for the processor in big endian mode.
11276 Compile code for the processor in little endian mode.
11280 Align doubles at 64-bit boundaries. Note that this changes the calling
11281 conventions, and thus some functions from the standard C library will
11282 not work unless you recompile it first with @option{-mdalign}.
11286 Shorten some address references at link time, when possible; uses the
11287 linker option @option{-relax}.
11291 Use 32-bit offsets in @code{switch} tables. The default is to use
11296 Enable the use of the instruction @code{fmovd}.
11300 Comply with the calling conventions defined by Renesas.
11304 Comply with the calling conventions defined by Renesas.
11308 Comply with the calling conventions defined for GCC before the Renesas
11309 conventions were available. This option is the default for all
11310 targets of the SH toolchain except for @samp{sh-symbianelf}.
11313 @opindex mnomacsave
11314 Mark the @code{MAC} register as call-clobbered, even if
11315 @option{-mhitachi} is given.
11319 Increase IEEE-compliance of floating-point code.
11323 Dump instruction size and location in the assembly code.
11326 @opindex mpadstruct
11327 This option is deprecated. It pads structures to multiple of 4 bytes,
11328 which is incompatible with the SH ABI@.
11332 Optimize for space instead of speed. Implied by @option{-Os}.
11335 @opindex mprefergot
11336 When generating position-independent code, emit function calls using
11337 the Global Offset Table instead of the Procedure Linkage Table.
11341 Generate a library function call to invalidate instruction cache
11342 entries, after fixing up a trampoline. This library function call
11343 doesn't assume it can write to the whole memory address space. This
11344 is the default when the target is @code{sh-*-linux*}.
11347 @node SPARC Options
11348 @subsection SPARC Options
11349 @cindex SPARC options
11351 These @samp{-m} options are supported on the SPARC:
11354 @item -mno-app-regs
11356 @opindex mno-app-regs
11358 Specify @option{-mapp-regs} to generate output using the global registers
11359 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
11362 To be fully SVR4 ABI compliant at the cost of some performance loss,
11363 specify @option{-mno-app-regs}. You should compile libraries and system
11364 software with this option.
11367 @itemx -mhard-float
11369 @opindex mhard-float
11370 Generate output containing floating point instructions. This is the
11374 @itemx -msoft-float
11376 @opindex msoft-float
11377 Generate output containing library calls for floating point.
11378 @strong{Warning:} the requisite libraries are not available for all SPARC
11379 targets. Normally the facilities of the machine's usual C compiler are
11380 used, but this cannot be done directly in cross-compilation. You must make
11381 your own arrangements to provide suitable library functions for
11382 cross-compilation. The embedded targets @samp{sparc-*-aout} and
11383 @samp{sparclite-*-*} do provide software floating point support.
11385 @option{-msoft-float} changes the calling convention in the output file;
11386 therefore, it is only useful if you compile @emph{all} of a program with
11387 this option. In particular, you need to compile @file{libgcc.a}, the
11388 library that comes with GCC, with @option{-msoft-float} in order for
11391 @item -mhard-quad-float
11392 @opindex mhard-quad-float
11393 Generate output containing quad-word (long double) floating point
11396 @item -msoft-quad-float
11397 @opindex msoft-quad-float
11398 Generate output containing library calls for quad-word (long double)
11399 floating point instructions. The functions called are those specified
11400 in the SPARC ABI@. This is the default.
11402 As of this writing, there are no SPARC implementations that have hardware
11403 support for the quad-word floating point instructions. They all invoke
11404 a trap handler for one of these instructions, and then the trap handler
11405 emulates the effect of the instruction. Because of the trap handler overhead,
11406 this is much slower than calling the ABI library routines. Thus the
11407 @option{-msoft-quad-float} option is the default.
11409 @item -mno-unaligned-doubles
11410 @itemx -munaligned-doubles
11411 @opindex mno-unaligned-doubles
11412 @opindex munaligned-doubles
11413 Assume that doubles have 8 byte alignment. This is the default.
11415 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
11416 alignment only if they are contained in another type, or if they have an
11417 absolute address. Otherwise, it assumes they have 4 byte alignment.
11418 Specifying this option avoids some rare compatibility problems with code
11419 generated by other compilers. It is not the default because it results
11420 in a performance loss, especially for floating point code.
11422 @item -mno-faster-structs
11423 @itemx -mfaster-structs
11424 @opindex mno-faster-structs
11425 @opindex mfaster-structs
11426 With @option{-mfaster-structs}, the compiler assumes that structures
11427 should have 8 byte alignment. This enables the use of pairs of
11428 @code{ldd} and @code{std} instructions for copies in structure
11429 assignment, in place of twice as many @code{ld} and @code{st} pairs.
11430 However, the use of this changed alignment directly violates the SPARC
11431 ABI@. Thus, it's intended only for use on targets where the developer
11432 acknowledges that their resulting code will not be directly in line with
11433 the rules of the ABI@.
11435 @item -mimpure-text
11436 @opindex mimpure-text
11437 @option{-mimpure-text}, used in addition to @option{-shared}, tells
11438 the compiler to not pass @option{-z text} to the linker when linking a
11439 shared object. Using this option, you can link position-dependent
11440 code into a shared object.
11442 @option{-mimpure-text} suppresses the ``relocations remain against
11443 allocatable but non-writable sections'' linker error message.
11444 However, the necessary relocations will trigger copy-on-write, and the
11445 shared object is not actually shared across processes. Instead of
11446 using @option{-mimpure-text}, you should compile all source code with
11447 @option{-fpic} or @option{-fPIC}.
11449 This option is only available on SunOS and Solaris.
11451 @item -mcpu=@var{cpu_type}
11453 Set the instruction set, register set, and instruction scheduling parameters
11454 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
11455 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
11456 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
11457 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
11458 @samp{ultrasparc3}.
11460 Default instruction scheduling parameters are used for values that select
11461 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
11462 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
11464 Here is a list of each supported architecture and their supported
11469 v8: supersparc, hypersparc
11470 sparclite: f930, f934, sparclite86x
11472 v9: ultrasparc, ultrasparc3
11475 By default (unless configured otherwise), GCC generates code for the V7
11476 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
11477 additionally optimizes it for the Cypress CY7C602 chip, as used in the
11478 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
11479 SPARCStation 1, 2, IPX etc.
11481 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
11482 architecture. The only difference from V7 code is that the compiler emits
11483 the integer multiply and integer divide instructions which exist in SPARC-V8
11484 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
11485 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
11488 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
11489 the SPARC architecture. This adds the integer multiply, integer divide step
11490 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
11491 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
11492 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
11493 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
11494 MB86934 chip, which is the more recent SPARClite with FPU@.
11496 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
11497 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
11498 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
11499 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
11500 optimizes it for the TEMIC SPARClet chip.
11502 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
11503 architecture. This adds 64-bit integer and floating-point move instructions,
11504 3 additional floating-point condition code registers and conditional move
11505 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
11506 optimizes it for the Sun UltraSPARC I/II chips. With
11507 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
11508 Sun UltraSPARC III chip.
11510 @item -mtune=@var{cpu_type}
11512 Set the instruction scheduling parameters for machine type
11513 @var{cpu_type}, but do not set the instruction set or register set that the
11514 option @option{-mcpu=@var{cpu_type}} would.
11516 The same values for @option{-mcpu=@var{cpu_type}} can be used for
11517 @option{-mtune=@var{cpu_type}}, but the only useful values are those
11518 that select a particular cpu implementation. Those are @samp{cypress},
11519 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
11520 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
11521 @samp{ultrasparc3}.
11526 @opindex mno-v8plus
11527 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
11528 difference from the V8 ABI is that the global and out registers are
11529 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
11530 mode for all SPARC-V9 processors.
11536 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
11537 Visual Instruction Set extensions. The default is @option{-mno-vis}.
11540 These @samp{-m} options are supported in addition to the above
11541 on SPARC-V9 processors in 64-bit environments:
11544 @item -mlittle-endian
11545 @opindex mlittle-endian
11546 Generate code for a processor running in little-endian mode. It is only
11547 available for a few configurations and most notably not on Solaris.
11553 Generate code for a 32-bit or 64-bit environment.
11554 The 32-bit environment sets int, long and pointer to 32 bits.
11555 The 64-bit environment sets int to 32 bits and long and pointer
11558 @item -mcmodel=medlow
11559 @opindex mcmodel=medlow
11560 Generate code for the Medium/Low code model: 64-bit addresses, programs
11561 must be linked in the low 32 bits of memory. Programs can be statically
11562 or dynamically linked.
11564 @item -mcmodel=medmid
11565 @opindex mcmodel=medmid
11566 Generate code for the Medium/Middle code model: 64-bit addresses, programs
11567 must be linked in the low 44 bits of memory, the text and data segments must
11568 be less than 2GB in size and the data segment must be located within 2GB of
11571 @item -mcmodel=medany
11572 @opindex mcmodel=medany
11573 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
11574 may be linked anywhere in memory, the text and data segments must be less
11575 than 2GB in size and the data segment must be located within 2GB of the
11578 @item -mcmodel=embmedany
11579 @opindex mcmodel=embmedany
11580 Generate code for the Medium/Anywhere code model for embedded systems:
11581 64-bit addresses, the text and data segments must be less than 2GB in
11582 size, both starting anywhere in memory (determined at link time). The
11583 global register %g4 points to the base of the data segment. Programs
11584 are statically linked and PIC is not supported.
11587 @itemx -mno-stack-bias
11588 @opindex mstack-bias
11589 @opindex mno-stack-bias
11590 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
11591 frame pointer if present, are offset by @minus{}2047 which must be added back
11592 when making stack frame references. This is the default in 64-bit mode.
11593 Otherwise, assume no such offset is present.
11596 These switches are supported in addition to the above on Solaris:
11601 Add support for multithreading using the Solaris threads library. This
11602 option sets flags for both the preprocessor and linker. This option does
11603 not affect the thread safety of object code produced by the compiler or
11604 that of libraries supplied with it.
11608 Add support for multithreading using the POSIX threads library. This
11609 option sets flags for both the preprocessor and linker. This option does
11610 not affect the thread safety of object code produced by the compiler or
11611 that of libraries supplied with it.
11614 @node System V Options
11615 @subsection Options for System V
11617 These additional options are available on System V Release 4 for
11618 compatibility with other compilers on those systems:
11623 Create a shared object.
11624 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
11628 Identify the versions of each tool used by the compiler, in a
11629 @code{.ident} assembler directive in the output.
11633 Refrain from adding @code{.ident} directives to the output file (this is
11636 @item -YP,@var{dirs}
11638 Search the directories @var{dirs}, and no others, for libraries
11639 specified with @option{-l}.
11641 @item -Ym,@var{dir}
11643 Look in the directory @var{dir} to find the M4 preprocessor.
11644 The assembler uses this option.
11645 @c This is supposed to go with a -Yd for predefined M4 macro files, but
11646 @c the generic assembler that comes with Solaris takes just -Ym.
11649 @node TMS320C3x/C4x Options
11650 @subsection TMS320C3x/C4x Options
11651 @cindex TMS320C3x/C4x Options
11653 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
11657 @item -mcpu=@var{cpu_type}
11659 Set the instruction set, register set, and instruction scheduling
11660 parameters for machine type @var{cpu_type}. Supported values for
11661 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
11662 @samp{c44}. The default is @samp{c40} to generate code for the
11667 @itemx -msmall-memory
11669 @opindex mbig-memory
11671 @opindex msmall-memory
11673 Generates code for the big or small memory model. The small memory
11674 model assumed that all data fits into one 64K word page. At run-time
11675 the data page (DP) register must be set to point to the 64K page
11676 containing the .bss and .data program sections. The big memory model is
11677 the default and requires reloading of the DP register for every direct
11684 Allow (disallow) allocation of general integer operands into the block
11685 count register BK@.
11691 Enable (disable) generation of code using decrement and branch,
11692 DBcond(D), instructions. This is enabled by default for the C4x. To be
11693 on the safe side, this is disabled for the C3x, since the maximum
11694 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
11695 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
11696 that it can utilize the decrement and branch instruction, but will give
11697 up if there is more than one memory reference in the loop. Thus a loop
11698 where the loop counter is decremented can generate slightly more
11699 efficient code, in cases where the RPTB instruction cannot be utilized.
11701 @item -mdp-isr-reload
11703 @opindex mdp-isr-reload
11705 Force the DP register to be saved on entry to an interrupt service
11706 routine (ISR), reloaded to point to the data section, and restored on
11707 exit from the ISR@. This should not be required unless someone has
11708 violated the small memory model by modifying the DP register, say within
11715 For the C3x use the 24-bit MPYI instruction for integer multiplies
11716 instead of a library call to guarantee 32-bit results. Note that if one
11717 of the operands is a constant, then the multiplication will be performed
11718 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
11719 then squaring operations are performed inline instead of a library call.
11722 @itemx -mno-fast-fix
11724 @opindex mno-fast-fix
11725 The C3x/C4x FIX instruction to convert a floating point value to an
11726 integer value chooses the nearest integer less than or equal to the
11727 floating point value rather than to the nearest integer. Thus if the
11728 floating point number is negative, the result will be incorrectly
11729 truncated an additional code is necessary to detect and correct this
11730 case. This option can be used to disable generation of the additional
11731 code required to correct the result.
11737 Enable (disable) generation of repeat block sequences using the RPTB
11738 instruction for zero overhead looping. The RPTB construct is only used
11739 for innermost loops that do not call functions or jump across the loop
11740 boundaries. There is no advantage having nested RPTB loops due to the
11741 overhead required to save and restore the RC, RS, and RE registers.
11742 This is enabled by default with @option{-O2}.
11744 @item -mrpts=@var{count}
11748 Enable (disable) the use of the single instruction repeat instruction
11749 RPTS@. If a repeat block contains a single instruction, and the loop
11750 count can be guaranteed to be less than the value @var{count}, GCC will
11751 emit a RPTS instruction instead of a RPTB@. If no value is specified,
11752 then a RPTS will be emitted even if the loop count cannot be determined
11753 at compile time. Note that the repeated instruction following RPTS does
11754 not have to be reloaded from memory each iteration, thus freeing up the
11755 CPU buses for operands. However, since interrupts are blocked by this
11756 instruction, it is disabled by default.
11758 @item -mloop-unsigned
11759 @itemx -mno-loop-unsigned
11760 @opindex mloop-unsigned
11761 @opindex mno-loop-unsigned
11762 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
11763 is @math{2^{31} + 1} since these instructions test if the iteration count is
11764 negative to terminate the loop. If the iteration count is unsigned
11765 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
11766 exceeded. This switch allows an unsigned iteration count.
11770 Try to emit an assembler syntax that the TI assembler (asm30) is happy
11771 with. This also enforces compatibility with the API employed by the TI
11772 C3x C compiler. For example, long doubles are passed as structures
11773 rather than in floating point registers.
11779 Generate code that uses registers (stack) for passing arguments to functions.
11780 By default, arguments are passed in registers where possible rather
11781 than by pushing arguments on to the stack.
11783 @item -mparallel-insns
11784 @itemx -mno-parallel-insns
11785 @opindex mparallel-insns
11786 @opindex mno-parallel-insns
11787 Allow the generation of parallel instructions. This is enabled by
11788 default with @option{-O2}.
11790 @item -mparallel-mpy
11791 @itemx -mno-parallel-mpy
11792 @opindex mparallel-mpy
11793 @opindex mno-parallel-mpy
11794 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
11795 provided @option{-mparallel-insns} is also specified. These instructions have
11796 tight register constraints which can pessimize the code generation
11797 of large functions.
11802 @subsection V850 Options
11803 @cindex V850 Options
11805 These @samp{-m} options are defined for V850 implementations:
11809 @itemx -mno-long-calls
11810 @opindex mlong-calls
11811 @opindex mno-long-calls
11812 Treat all calls as being far away (near). If calls are assumed to be
11813 far away, the compiler will always load the functions address up into a
11814 register, and call indirect through the pointer.
11820 Do not optimize (do optimize) basic blocks that use the same index
11821 pointer 4 or more times to copy pointer into the @code{ep} register, and
11822 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
11823 option is on by default if you optimize.
11825 @item -mno-prolog-function
11826 @itemx -mprolog-function
11827 @opindex mno-prolog-function
11828 @opindex mprolog-function
11829 Do not use (do use) external functions to save and restore registers
11830 at the prologue and epilogue of a function. The external functions
11831 are slower, but use less code space if more than one function saves
11832 the same number of registers. The @option{-mprolog-function} option
11833 is on by default if you optimize.
11837 Try to make the code as small as possible. At present, this just turns
11838 on the @option{-mep} and @option{-mprolog-function} options.
11840 @item -mtda=@var{n}
11842 Put static or global variables whose size is @var{n} bytes or less into
11843 the tiny data area that register @code{ep} points to. The tiny data
11844 area can hold up to 256 bytes in total (128 bytes for byte references).
11846 @item -msda=@var{n}
11848 Put static or global variables whose size is @var{n} bytes or less into
11849 the small data area that register @code{gp} points to. The small data
11850 area can hold up to 64 kilobytes.
11852 @item -mzda=@var{n}
11854 Put static or global variables whose size is @var{n} bytes or less into
11855 the first 32 kilobytes of memory.
11859 Specify that the target processor is the V850.
11862 @opindex mbig-switch
11863 Generate code suitable for big switch tables. Use this option only if
11864 the assembler/linker complain about out of range branches within a switch
11869 This option will cause r2 and r5 to be used in the code generated by
11870 the compiler. This setting is the default.
11872 @item -mno-app-regs
11873 @opindex mno-app-regs
11874 This option will cause r2 and r5 to be treated as fixed registers.
11878 Specify that the target processor is the V850E1. The preprocessor
11879 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
11880 this option is used.
11884 Specify that the target processor is the V850E@. The preprocessor
11885 constant @samp{__v850e__} will be defined if this option is used.
11887 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
11888 are defined then a default target processor will be chosen and the
11889 relevant @samp{__v850*__} preprocessor constant will be defined.
11891 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
11892 defined, regardless of which processor variant is the target.
11894 @item -mdisable-callt
11895 @opindex mdisable-callt
11896 This option will suppress generation of the CALLT instruction for the
11897 v850e and v850e1 flavors of the v850 architecture. The default is
11898 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
11903 @subsection VAX Options
11904 @cindex VAX options
11906 These @samp{-m} options are defined for the VAX:
11911 Do not output certain jump instructions (@code{aobleq} and so on)
11912 that the Unix assembler for the VAX cannot handle across long
11917 Do output those jump instructions, on the assumption that you
11918 will assemble with the GNU assembler.
11922 Output code for g-format floating point numbers instead of d-format.
11925 @node x86-64 Options
11926 @subsection x86-64 Options
11927 @cindex x86-64 options
11929 These are listed under @xref{i386 and x86-64 Options}.
11931 @node Xstormy16 Options
11932 @subsection Xstormy16 Options
11933 @cindex Xstormy16 Options
11935 These options are defined for Xstormy16:
11940 Choose startup files and linker script suitable for the simulator.
11943 @node Xtensa Options
11944 @subsection Xtensa Options
11945 @cindex Xtensa Options
11947 These options are supported for Xtensa targets:
11951 @itemx -mno-const16
11953 @opindex mno-const16
11954 Enable or disable use of @code{CONST16} instructions for loading
11955 constant values. The @code{CONST16} instruction is currently not a
11956 standard option from Tensilica. When enabled, @code{CONST16}
11957 instructions are always used in place of the standard @code{L32R}
11958 instructions. The use of @code{CONST16} is enabled by default only if
11959 the @code{L32R} instruction is not available.
11962 @itemx -mno-fused-madd
11963 @opindex mfused-madd
11964 @opindex mno-fused-madd
11965 Enable or disable use of fused multiply/add and multiply/subtract
11966 instructions in the floating-point option. This has no effect if the
11967 floating-point option is not also enabled. Disabling fused multiply/add
11968 and multiply/subtract instructions forces the compiler to use separate
11969 instructions for the multiply and add/subtract operations. This may be
11970 desirable in some cases where strict IEEE 754-compliant results are
11971 required: the fused multiply add/subtract instructions do not round the
11972 intermediate result, thereby producing results with @emph{more} bits of
11973 precision than specified by the IEEE standard. Disabling fused multiply
11974 add/subtract instructions also ensures that the program output is not
11975 sensitive to the compiler's ability to combine multiply and add/subtract
11978 @item -mtext-section-literals
11979 @itemx -mno-text-section-literals
11980 @opindex mtext-section-literals
11981 @opindex mno-text-section-literals
11982 Control the treatment of literal pools. The default is
11983 @option{-mno-text-section-literals}, which places literals in a separate
11984 section in the output file. This allows the literal pool to be placed
11985 in a data RAM/ROM, and it also allows the linker to combine literal
11986 pools from separate object files to remove redundant literals and
11987 improve code size. With @option{-mtext-section-literals}, the literals
11988 are interspersed in the text section in order to keep them as close as
11989 possible to their references. This may be necessary for large assembly
11992 @item -mtarget-align
11993 @itemx -mno-target-align
11994 @opindex mtarget-align
11995 @opindex mno-target-align
11996 When this option is enabled, GCC instructs the assembler to
11997 automatically align instructions to reduce branch penalties at the
11998 expense of some code density. The assembler attempts to widen density
11999 instructions to align branch targets and the instructions following call
12000 instructions. If there are not enough preceding safe density
12001 instructions to align a target, no widening will be performed. The
12002 default is @option{-mtarget-align}. These options do not affect the
12003 treatment of auto-aligned instructions like @code{LOOP}, which the
12004 assembler will always align, either by widening density instructions or
12005 by inserting no-op instructions.
12008 @itemx -mno-longcalls
12009 @opindex mlongcalls
12010 @opindex mno-longcalls
12011 When this option is enabled, GCC instructs the assembler to translate
12012 direct calls to indirect calls unless it can determine that the target
12013 of a direct call is in the range allowed by the call instruction. This
12014 translation typically occurs for calls to functions in other source
12015 files. Specifically, the assembler translates a direct @code{CALL}
12016 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
12017 The default is @option{-mno-longcalls}. This option should be used in
12018 programs where the call target can potentially be out of range. This
12019 option is implemented in the assembler, not the compiler, so the
12020 assembly code generated by GCC will still show direct call
12021 instructions---look at the disassembled object code to see the actual
12022 instructions. Note that the assembler will use an indirect call for
12023 every cross-file call, not just those that really will be out of range.
12026 @node zSeries Options
12027 @subsection zSeries Options
12028 @cindex zSeries options
12030 These are listed under @xref{S/390 and zSeries Options}.
12032 @node Code Gen Options
12033 @section Options for Code Generation Conventions
12034 @cindex code generation conventions
12035 @cindex options, code generation
12036 @cindex run-time options
12038 These machine-independent options control the interface conventions
12039 used in code generation.
12041 Most of them have both positive and negative forms; the negative form
12042 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
12043 one of the forms is listed---the one which is not the default. You
12044 can figure out the other form by either removing @samp{no-} or adding
12048 @item -fbounds-check
12049 @opindex fbounds-check
12050 For front-ends that support it, generate additional code to check that
12051 indices used to access arrays are within the declared range. This is
12052 currently only supported by the Java and Fortran 77 front-ends, where
12053 this option defaults to true and false respectively.
12057 This option generates traps for signed overflow on addition, subtraction,
12058 multiplication operations.
12062 This option instructs the compiler to assume that signed arithmetic
12063 overflow of addition, subtraction and multiplication wraps around
12064 using twos-complement representation. This flag enables some optimizations
12065 and disables other. This option is enabled by default for the Java
12066 front-end, as required by the Java language specification.
12069 @opindex fexceptions
12070 Enable exception handling. Generates extra code needed to propagate
12071 exceptions. For some targets, this implies GCC will generate frame
12072 unwind information for all functions, which can produce significant data
12073 size overhead, although it does not affect execution. If you do not
12074 specify this option, GCC will enable it by default for languages like
12075 C++ which normally require exception handling, and disable it for
12076 languages like C that do not normally require it. However, you may need
12077 to enable this option when compiling C code that needs to interoperate
12078 properly with exception handlers written in C++. You may also wish to
12079 disable this option if you are compiling older C++ programs that don't
12080 use exception handling.
12082 @item -fnon-call-exceptions
12083 @opindex fnon-call-exceptions
12084 Generate code that allows trapping instructions to throw exceptions.
12085 Note that this requires platform-specific runtime support that does
12086 not exist everywhere. Moreover, it only allows @emph{trapping}
12087 instructions to throw exceptions, i.e.@: memory references or floating
12088 point instructions. It does not allow exceptions to be thrown from
12089 arbitrary signal handlers such as @code{SIGALRM}.
12091 @item -funwind-tables
12092 @opindex funwind-tables
12093 Similar to @option{-fexceptions}, except that it will just generate any needed
12094 static data, but will not affect the generated code in any other way.
12095 You will normally not enable this option; instead, a language processor
12096 that needs this handling would enable it on your behalf.
12098 @item -fasynchronous-unwind-tables
12099 @opindex fasynchronous-unwind-tables
12100 Generate unwind table in dwarf2 format, if supported by target machine. The
12101 table is exact at each instruction boundary, so it can be used for stack
12102 unwinding from asynchronous events (such as debugger or garbage collector).
12104 @item -fpcc-struct-return
12105 @opindex fpcc-struct-return
12106 Return ``short'' @code{struct} and @code{union} values in memory like
12107 longer ones, rather than in registers. This convention is less
12108 efficient, but it has the advantage of allowing intercallability between
12109 GCC-compiled files and files compiled with other compilers, particularly
12110 the Portable C Compiler (pcc).
12112 The precise convention for returning structures in memory depends
12113 on the target configuration macros.
12115 Short structures and unions are those whose size and alignment match
12116 that of some integer type.
12118 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12119 switch is not binary compatible with code compiled with the
12120 @option{-freg-struct-return} switch.
12121 Use it to conform to a non-default application binary interface.
12123 @item -freg-struct-return
12124 @opindex freg-struct-return
12125 Return @code{struct} and @code{union} values in registers when possible.
12126 This is more efficient for small structures than
12127 @option{-fpcc-struct-return}.
12129 If you specify neither @option{-fpcc-struct-return} nor
12130 @option{-freg-struct-return}, GCC defaults to whichever convention is
12131 standard for the target. If there is no standard convention, GCC
12132 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12133 the principal compiler. In those cases, we can choose the standard, and
12134 we chose the more efficient register return alternative.
12136 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12137 switch is not binary compatible with code compiled with the
12138 @option{-fpcc-struct-return} switch.
12139 Use it to conform to a non-default application binary interface.
12141 @item -fshort-enums
12142 @opindex fshort-enums
12143 Allocate to an @code{enum} type only as many bytes as it needs for the
12144 declared range of possible values. Specifically, the @code{enum} type
12145 will be equivalent to the smallest integer type which has enough room.
12147 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12148 code that is not binary compatible with code generated without that switch.
12149 Use it to conform to a non-default application binary interface.
12151 @item -fshort-double
12152 @opindex fshort-double
12153 Use the same size for @code{double} as for @code{float}.
12155 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
12156 code that is not binary compatible with code generated without that switch.
12157 Use it to conform to a non-default application binary interface.
12159 @item -fshort-wchar
12160 @opindex fshort-wchar
12161 Override the underlying type for @samp{wchar_t} to be @samp{short
12162 unsigned int} instead of the default for the target. This option is
12163 useful for building programs to run under WINE@.
12165 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12166 code that is not binary compatible with code generated without that switch.
12167 Use it to conform to a non-default application binary interface.
12169 @item -fshared-data
12170 @opindex fshared-data
12171 Requests that the data and non-@code{const} variables of this
12172 compilation be shared data rather than private data. The distinction
12173 makes sense only on certain operating systems, where shared data is
12174 shared between processes running the same program, while private data
12175 exists in one copy per process.
12178 @opindex fno-common
12179 In C, allocate even uninitialized global variables in the data section of the
12180 object file, rather than generating them as common blocks. This has the
12181 effect that if the same variable is declared (without @code{extern}) in
12182 two different compilations, you will get an error when you link them.
12183 The only reason this might be useful is if you wish to verify that the
12184 program will work on other systems which always work this way.
12188 Ignore the @samp{#ident} directive.
12190 @item -finhibit-size-directive
12191 @opindex finhibit-size-directive
12192 Don't output a @code{.size} assembler directive, or anything else that
12193 would cause trouble if the function is split in the middle, and the
12194 two halves are placed at locations far apart in memory. This option is
12195 used when compiling @file{crtstuff.c}; you should not need to use it
12198 @item -fverbose-asm
12199 @opindex fverbose-asm
12200 Put extra commentary information in the generated assembly code to
12201 make it more readable. This option is generally only of use to those
12202 who actually need to read the generated assembly code (perhaps while
12203 debugging the compiler itself).
12205 @option{-fno-verbose-asm}, the default, causes the
12206 extra information to be omitted and is useful when comparing two assembler
12211 @cindex global offset table
12213 Generate position-independent code (PIC) suitable for use in a shared
12214 library, if supported for the target machine. Such code accesses all
12215 constant addresses through a global offset table (GOT)@. The dynamic
12216 loader resolves the GOT entries when the program starts (the dynamic
12217 loader is not part of GCC; it is part of the operating system). If
12218 the GOT size for the linked executable exceeds a machine-specific
12219 maximum size, you get an error message from the linker indicating that
12220 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12221 instead. (These maximums are 8k on the SPARC and 32k
12222 on the m68k and RS/6000. The 386 has no such limit.)
12224 Position-independent code requires special support, and therefore works
12225 only on certain machines. For the 386, GCC supports PIC for System V
12226 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12227 position-independent.
12231 If supported for the target machine, emit position-independent code,
12232 suitable for dynamic linking and avoiding any limit on the size of the
12233 global offset table. This option makes a difference on the m68k,
12234 PowerPC and SPARC@.
12236 Position-independent code requires special support, and therefore works
12237 only on certain machines.
12243 These options are similar to @option{-fpic} and @option{-fPIC}, but
12244 generated position independent code can be only linked into executables.
12245 Usually these options are used when @option{-pie} GCC option will be
12246 used during linking.
12248 @item -ffixed-@var{reg}
12250 Treat the register named @var{reg} as a fixed register; generated code
12251 should never refer to it (except perhaps as a stack pointer, frame
12252 pointer or in some other fixed role).
12254 @var{reg} must be the name of a register. The register names accepted
12255 are machine-specific and are defined in the @code{REGISTER_NAMES}
12256 macro in the machine description macro file.
12258 This flag does not have a negative form, because it specifies a
12261 @item -fcall-used-@var{reg}
12262 @opindex fcall-used
12263 Treat the register named @var{reg} as an allocable register that is
12264 clobbered by function calls. It may be allocated for temporaries or
12265 variables that do not live across a call. Functions compiled this way
12266 will not save and restore the register @var{reg}.
12268 It is an error to used this flag with the frame pointer or stack pointer.
12269 Use of this flag for other registers that have fixed pervasive roles in
12270 the machine's execution model will produce disastrous results.
12272 This flag does not have a negative form, because it specifies a
12275 @item -fcall-saved-@var{reg}
12276 @opindex fcall-saved
12277 Treat the register named @var{reg} as an allocable register saved by
12278 functions. It may be allocated even for temporaries or variables that
12279 live across a call. Functions compiled this way will save and restore
12280 the register @var{reg} if they use it.
12282 It is an error to used this flag with the frame pointer or stack pointer.
12283 Use of this flag for other registers that have fixed pervasive roles in
12284 the machine's execution model will produce disastrous results.
12286 A different sort of disaster will result from the use of this flag for
12287 a register in which function values may be returned.
12289 This flag does not have a negative form, because it specifies a
12292 @item -fpack-struct[=@var{n}]
12293 @opindex fpack-struct
12294 Without a value specified, pack all structure members together without
12295 holes. When a value is specified (which must be a small power of two), pack
12296 structure members according to this value, representing the maximum
12297 alignment (that is, objects with default alignment requirements larger than
12298 this will be output potentially unaligned at the next fitting location.
12300 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12301 code that is not binary compatible with code generated without that switch.
12302 Additionally, it makes the code suboptimal.
12303 Use it to conform to a non-default application binary interface.
12305 @item -finstrument-functions
12306 @opindex finstrument-functions
12307 Generate instrumentation calls for entry and exit to functions. Just
12308 after function entry and just before function exit, the following
12309 profiling functions will be called with the address of the current
12310 function and its call site. (On some platforms,
12311 @code{__builtin_return_address} does not work beyond the current
12312 function, so the call site information may not be available to the
12313 profiling functions otherwise.)
12316 void __cyg_profile_func_enter (void *this_fn,
12318 void __cyg_profile_func_exit (void *this_fn,
12322 The first argument is the address of the start of the current function,
12323 which may be looked up exactly in the symbol table.
12325 This instrumentation is also done for functions expanded inline in other
12326 functions. The profiling calls will indicate where, conceptually, the
12327 inline function is entered and exited. This means that addressable
12328 versions of such functions must be available. If all your uses of a
12329 function are expanded inline, this may mean an additional expansion of
12330 code size. If you use @samp{extern inline} in your C code, an
12331 addressable version of such functions must be provided. (This is
12332 normally the case anyways, but if you get lucky and the optimizer always
12333 expands the functions inline, you might have gotten away without
12334 providing static copies.)
12336 A function may be given the attribute @code{no_instrument_function}, in
12337 which case this instrumentation will not be done. This can be used, for
12338 example, for the profiling functions listed above, high-priority
12339 interrupt routines, and any functions from which the profiling functions
12340 cannot safely be called (perhaps signal handlers, if the profiling
12341 routines generate output or allocate memory).
12343 @item -fstack-check
12344 @opindex fstack-check
12345 Generate code to verify that you do not go beyond the boundary of the
12346 stack. You should specify this flag if you are running in an
12347 environment with multiple threads, but only rarely need to specify it in
12348 a single-threaded environment since stack overflow is automatically
12349 detected on nearly all systems if there is only one stack.
12351 Note that this switch does not actually cause checking to be done; the
12352 operating system must do that. The switch causes generation of code
12353 to ensure that the operating system sees the stack being extended.
12355 @item -fstack-limit-register=@var{reg}
12356 @itemx -fstack-limit-symbol=@var{sym}
12357 @itemx -fno-stack-limit
12358 @opindex fstack-limit-register
12359 @opindex fstack-limit-symbol
12360 @opindex fno-stack-limit
12361 Generate code to ensure that the stack does not grow beyond a certain value,
12362 either the value of a register or the address of a symbol. If the stack
12363 would grow beyond the value, a signal is raised. For most targets,
12364 the signal is raised before the stack overruns the boundary, so
12365 it is possible to catch the signal without taking special precautions.
12367 For instance, if the stack starts at absolute address @samp{0x80000000}
12368 and grows downwards, you can use the flags
12369 @option{-fstack-limit-symbol=__stack_limit} and
12370 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12371 of 128KB@. Note that this may only work with the GNU linker.
12373 @cindex aliasing of parameters
12374 @cindex parameters, aliased
12375 @item -fargument-alias
12376 @itemx -fargument-noalias
12377 @itemx -fargument-noalias-global
12378 @opindex fargument-alias
12379 @opindex fargument-noalias
12380 @opindex fargument-noalias-global
12381 Specify the possible relationships among parameters and between
12382 parameters and global data.
12384 @option{-fargument-alias} specifies that arguments (parameters) may
12385 alias each other and may alias global storage.@*
12386 @option{-fargument-noalias} specifies that arguments do not alias
12387 each other, but may alias global storage.@*
12388 @option{-fargument-noalias-global} specifies that arguments do not
12389 alias each other and do not alias global storage.
12391 Each language will automatically use whatever option is required by
12392 the language standard. You should not need to use these options yourself.
12394 @item -fleading-underscore
12395 @opindex fleading-underscore
12396 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12397 change the way C symbols are represented in the object file. One use
12398 is to help link with legacy assembly code.
12400 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12401 generate code that is not binary compatible with code generated without that
12402 switch. Use it to conform to a non-default application binary interface.
12403 Not all targets provide complete support for this switch.
12405 @item -ftls-model=@var{model}
12406 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12407 The @var{model} argument should be one of @code{global-dynamic},
12408 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
12410 The default without @option{-fpic} is @code{initial-exec}; with
12411 @option{-fpic} the default is @code{global-dynamic}.
12413 @item -fvisibility=@var{default|internal|hidden|protected}
12414 @opindex fvisibility
12415 Set the default ELF image symbol visibility to the specified option---all
12416 symbols will be marked with this unless overridden within the code.
12417 Using this feature can very substantially improve linking and
12418 load times of shared object libraries, produce more optimized
12419 code, provide near-perfect API export and prevent symbol clashes.
12420 It is @strong{strongly} recommended that you use this in any shared objects
12423 Despite the nomenclature, @code{default} always means public ie;
12424 available to be linked against from outside the shared object.
12425 @code{protected} and @code{internal} are pretty useless in real-world
12426 usage so the only other commonly used option will be @code{hidden}.
12427 The default if @option{-fvisibility} isn't specified is
12428 @code{default}, i.e., make every
12429 symbol public---this causes the same behavior as previous versions of
12432 A good explanation of the benefits offered by ensuring ELF
12433 symbols have the correct visibility is given by ``How To Write
12434 Shared Libraries'' by Ulrich Drepper (which can be found at
12435 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
12436 solution made possible by this option to marking things hidden when
12437 the default is public is to make the default hidden and mark things
12438 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
12439 and @code{__attribute__ ((visibility("default")))} instead of
12440 @code{__declspec(dllexport)} you get almost identical semantics with
12441 identical syntax. This is a great boon to those working with
12442 cross-platform projects.
12444 For those adding visibility support to existing code, you may find
12445 @samp{#pragma GCC visibility} of use. This works by you enclosing
12446 the declarations you wish to set visibility for with (for example)
12447 @samp{#pragma GCC visibility push(hidden)} and
12448 @samp{#pragma GCC visibility pop}. These can be nested up to sixteen
12449 times. Bear in mind that symbol visibility should be viewed @strong{as
12450 part of the API interface contract} and thus all new code should
12451 always specify visibility when it is not the default ie; declarations
12452 only for use within the local DSO should @strong{always} be marked explicitly
12453 as hidden as so to avoid PLT indirection overheads---making this
12454 abundantly clear also aids readability and self-documentation of the code.
12455 Note that due to ISO C++ specification requirements, operator new and
12456 operator delete must always be of default visibility.
12458 An overview of these techniques, their benefits and how to use them
12459 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
12465 @node Environment Variables
12466 @section Environment Variables Affecting GCC
12467 @cindex environment variables
12469 @c man begin ENVIRONMENT
12470 This section describes several environment variables that affect how GCC
12471 operates. Some of them work by specifying directories or prefixes to use
12472 when searching for various kinds of files. Some are used to specify other
12473 aspects of the compilation environment.
12475 Note that you can also specify places to search using options such as
12476 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
12477 take precedence over places specified using environment variables, which
12478 in turn take precedence over those specified by the configuration of GCC@.
12479 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
12480 GNU Compiler Collection (GCC) Internals}.
12485 @c @itemx LC_COLLATE
12487 @c @itemx LC_MONETARY
12488 @c @itemx LC_NUMERIC
12493 @c @findex LC_COLLATE
12494 @findex LC_MESSAGES
12495 @c @findex LC_MONETARY
12496 @c @findex LC_NUMERIC
12500 These environment variables control the way that GCC uses
12501 localization information that allow GCC to work with different
12502 national conventions. GCC inspects the locale categories
12503 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
12504 so. These locale categories can be set to any value supported by your
12505 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
12506 Kingdom encoded in UTF-8.
12508 The @env{LC_CTYPE} environment variable specifies character
12509 classification. GCC uses it to determine the character boundaries in
12510 a string; this is needed for some multibyte encodings that contain quote
12511 and escape characters that would otherwise be interpreted as a string
12514 The @env{LC_MESSAGES} environment variable specifies the language to
12515 use in diagnostic messages.
12517 If the @env{LC_ALL} environment variable is set, it overrides the value
12518 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
12519 and @env{LC_MESSAGES} default to the value of the @env{LANG}
12520 environment variable. If none of these variables are set, GCC
12521 defaults to traditional C English behavior.
12525 If @env{TMPDIR} is set, it specifies the directory to use for temporary
12526 files. GCC uses temporary files to hold the output of one stage of
12527 compilation which is to be used as input to the next stage: for example,
12528 the output of the preprocessor, which is the input to the compiler
12531 @item GCC_EXEC_PREFIX
12532 @findex GCC_EXEC_PREFIX
12533 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
12534 names of the subprograms executed by the compiler. No slash is added
12535 when this prefix is combined with the name of a subprogram, but you can
12536 specify a prefix that ends with a slash if you wish.
12538 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
12539 an appropriate prefix to use based on the pathname it was invoked with.
12541 If GCC cannot find the subprogram using the specified prefix, it
12542 tries looking in the usual places for the subprogram.
12544 The default value of @env{GCC_EXEC_PREFIX} is
12545 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
12546 of @code{prefix} when you ran the @file{configure} script.
12548 Other prefixes specified with @option{-B} take precedence over this prefix.
12550 This prefix is also used for finding files such as @file{crt0.o} that are
12553 In addition, the prefix is used in an unusual way in finding the
12554 directories to search for header files. For each of the standard
12555 directories whose name normally begins with @samp{/usr/local/lib/gcc}
12556 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
12557 replacing that beginning with the specified prefix to produce an
12558 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
12559 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
12560 These alternate directories are searched first; the standard directories
12563 @item COMPILER_PATH
12564 @findex COMPILER_PATH
12565 The value of @env{COMPILER_PATH} is a colon-separated list of
12566 directories, much like @env{PATH}. GCC tries the directories thus
12567 specified when searching for subprograms, if it can't find the
12568 subprograms using @env{GCC_EXEC_PREFIX}.
12571 @findex LIBRARY_PATH
12572 The value of @env{LIBRARY_PATH} is a colon-separated list of
12573 directories, much like @env{PATH}. When configured as a native compiler,
12574 GCC tries the directories thus specified when searching for special
12575 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
12576 using GCC also uses these directories when searching for ordinary
12577 libraries for the @option{-l} option (but directories specified with
12578 @option{-L} come first).
12582 @cindex locale definition
12583 This variable is used to pass locale information to the compiler. One way in
12584 which this information is used is to determine the character set to be used
12585 when character literals, string literals and comments are parsed in C and C++.
12586 When the compiler is configured to allow multibyte characters,
12587 the following values for @env{LANG} are recognized:
12591 Recognize JIS characters.
12593 Recognize SJIS characters.
12595 Recognize EUCJP characters.
12598 If @env{LANG} is not defined, or if it has some other value, then the
12599 compiler will use mblen and mbtowc as defined by the default locale to
12600 recognize and translate multibyte characters.
12604 Some additional environments variables affect the behavior of the
12607 @include cppenv.texi
12611 @node Precompiled Headers
12612 @section Using Precompiled Headers
12613 @cindex precompiled headers
12614 @cindex speed of compilation
12616 Often large projects have many header files that are included in every
12617 source file. The time the compiler takes to process these header files
12618 over and over again can account for nearly all of the time required to
12619 build the project. To make builds faster, GCC allows users to
12620 `precompile' a header file; then, if builds can use the precompiled
12621 header file they will be much faster.
12623 @strong{Caution:} There are a few known situations where GCC will
12624 crash when trying to use a precompiled header. If you have trouble
12625 with a precompiled header, you should remove the precompiled header
12626 and compile without it. In addition, please use GCC's on-line
12627 defect-tracking system to report any problems you encounter with
12628 precompiled headers. @xref{Bugs}.
12630 To create a precompiled header file, simply compile it as you would any
12631 other file, if necessary using the @option{-x} option to make the driver
12632 treat it as a C or C++ header file. You will probably want to use a
12633 tool like @command{make} to keep the precompiled header up-to-date when
12634 the headers it contains change.
12636 A precompiled header file will be searched for when @code{#include} is
12637 seen in the compilation. As it searches for the included file
12638 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
12639 compiler looks for a precompiled header in each directory just before it
12640 looks for the include file in that directory. The name searched for is
12641 the name specified in the @code{#include} with @samp{.gch} appended. If
12642 the precompiled header file can't be used, it is ignored.
12644 For instance, if you have @code{#include "all.h"}, and you have
12645 @file{all.h.gch} in the same directory as @file{all.h}, then the
12646 precompiled header file will be used if possible, and the original
12647 header will be used otherwise.
12649 Alternatively, you might decide to put the precompiled header file in a
12650 directory and use @option{-I} to ensure that directory is searched
12651 before (or instead of) the directory containing the original header.
12652 Then, if you want to check that the precompiled header file is always
12653 used, you can put a file of the same name as the original header in this
12654 directory containing an @code{#error} command.
12656 This also works with @option{-include}. So yet another way to use
12657 precompiled headers, good for projects not designed with precompiled
12658 header files in mind, is to simply take most of the header files used by
12659 a project, include them from another header file, precompile that header
12660 file, and @option{-include} the precompiled header. If the header files
12661 have guards against multiple inclusion, they will be skipped because
12662 they've already been included (in the precompiled header).
12664 If you need to precompile the same header file for different
12665 languages, targets, or compiler options, you can instead make a
12666 @emph{directory} named like @file{all.h.gch}, and put each precompiled
12667 header in the directory, perhaps using @option{-o}. It doesn't matter
12668 what you call the files in the directory, every precompiled header in
12669 the directory will be considered. The first precompiled header
12670 encountered in the directory that is valid for this compilation will
12671 be used; they're searched in no particular order.
12673 There are many other possibilities, limited only by your imagination,
12674 good sense, and the constraints of your build system.
12676 A precompiled header file can be used only when these conditions apply:
12680 Only one precompiled header can be used in a particular compilation.
12683 A precompiled header can't be used once the first C token is seen. You
12684 can have preprocessor directives before a precompiled header; you can
12685 even include a precompiled header from inside another header, so long as
12686 there are no C tokens before the @code{#include}.
12689 The precompiled header file must be produced for the same language as
12690 the current compilation. You can't use a C precompiled header for a C++
12694 The precompiled header file must be produced by the same compiler
12695 version and configuration as the current compilation is using.
12696 The easiest way to guarantee this is to use the same compiler binary
12697 for creating and using precompiled headers.
12700 Any macros defined before the precompiled header is included must
12701 either be defined in the same way as when the precompiled header was
12702 generated, or must not affect the precompiled header, which usually
12703 means that the they don't appear in the precompiled header at all.
12705 The @option{-D} option is one way to define a macro before a
12706 precompiled header is included; using a @code{#define} can also do it.
12707 There are also some options that define macros implicitly, like
12708 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
12711 @item If debugging information is output when using the precompiled
12712 header, using @option{-g} or similar, the same kind of debugging information
12713 must have been output when building the precompiled header. However,
12714 a precompiled header built using @option{-g} can be used in a compilation
12715 when no debugging information is being output.
12717 @item The same @option{-m} options must generally be used when building
12718 and using the precompiled header. @xref{Submodel Options},
12719 for any cases where this rule is relaxed.
12721 @item Each of the following options must be the same when building and using
12722 the precompiled header:
12724 @gccoptlist{-fexceptions -funit-at-a-time}
12727 Some other command-line options starting with @option{-f},
12728 @option{-p}, or @option{-O} must be defined in the same way as when
12729 the precompiled header was generated. At present, it's not clear
12730 which options are safe to change and which are not; the safest choice
12731 is to use exactly the same options when generating and using the
12732 precompiled header. The following are known to be safe:
12734 @gccoptlist{-fpreprocessed -pedantic-errors}
12738 For all of these except the last, the compiler will automatically
12739 ignore the precompiled header if the conditions aren't met. If you
12740 find an option combination that doesn't work and doesn't cause the
12741 precompiled header to be ignored, please consider filing a bug report,
12744 If you do use differing options when generating and using the
12745 precompiled header, the actual behavior will be a mixture of the
12746 behavior for the options. For instance, if you use @option{-g} to
12747 generate the precompiled header but not when using it, you may or may
12748 not get debugging information for routines in the precompiled header.
12750 @node Running Protoize
12751 @section Running Protoize
12753 The program @code{protoize} is an optional part of GCC@. You can use
12754 it to add prototypes to a program, thus converting the program to ISO
12755 C in one respect. The companion program @code{unprotoize} does the
12756 reverse: it removes argument types from any prototypes that are found.
12758 When you run these programs, you must specify a set of source files as
12759 command line arguments. The conversion programs start out by compiling
12760 these files to see what functions they define. The information gathered
12761 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
12763 After scanning comes actual conversion. The specified files are all
12764 eligible to be converted; any files they include (whether sources or
12765 just headers) are eligible as well.
12767 But not all the eligible files are converted. By default,
12768 @code{protoize} and @code{unprotoize} convert only source and header
12769 files in the current directory. You can specify additional directories
12770 whose files should be converted with the @option{-d @var{directory}}
12771 option. You can also specify particular files to exclude with the
12772 @option{-x @var{file}} option. A file is converted if it is eligible, its
12773 directory name matches one of the specified directory names, and its
12774 name within the directory has not been excluded.
12776 Basic conversion with @code{protoize} consists of rewriting most
12777 function definitions and function declarations to specify the types of
12778 the arguments. The only ones not rewritten are those for varargs
12781 @code{protoize} optionally inserts prototype declarations at the
12782 beginning of the source file, to make them available for any calls that
12783 precede the function's definition. Or it can insert prototype
12784 declarations with block scope in the blocks where undeclared functions
12787 Basic conversion with @code{unprotoize} consists of rewriting most
12788 function declarations to remove any argument types, and rewriting
12789 function definitions to the old-style pre-ISO form.
12791 Both conversion programs print a warning for any function declaration or
12792 definition that they can't convert. You can suppress these warnings
12795 The output from @code{protoize} or @code{unprotoize} replaces the
12796 original source file. The original file is renamed to a name ending
12797 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
12798 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
12799 for DOS) file already exists, then the source file is simply discarded.
12801 @code{protoize} and @code{unprotoize} both depend on GCC itself to
12802 scan the program and collect information about the functions it uses.
12803 So neither of these programs will work until GCC is installed.
12805 Here is a table of the options you can use with @code{protoize} and
12806 @code{unprotoize}. Each option works with both programs unless
12810 @item -B @var{directory}
12811 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
12812 usual directory (normally @file{/usr/local/lib}). This file contains
12813 prototype information about standard system functions. This option
12814 applies only to @code{protoize}.
12816 @item -c @var{compilation-options}
12817 Use @var{compilation-options} as the options when running @command{gcc} to
12818 produce the @samp{.X} files. The special option @option{-aux-info} is
12819 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
12821 Note that the compilation options must be given as a single argument to
12822 @code{protoize} or @code{unprotoize}. If you want to specify several
12823 @command{gcc} options, you must quote the entire set of compilation options
12824 to make them a single word in the shell.
12826 There are certain @command{gcc} arguments that you cannot use, because they
12827 would produce the wrong kind of output. These include @option{-g},
12828 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
12829 the @var{compilation-options}, they are ignored.
12832 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
12833 systems) instead of @samp{.c}. This is convenient if you are converting
12834 a C program to C++. This option applies only to @code{protoize}.
12837 Add explicit global declarations. This means inserting explicit
12838 declarations at the beginning of each source file for each function
12839 that is called in the file and was not declared. These declarations
12840 precede the first function definition that contains a call to an
12841 undeclared function. This option applies only to @code{protoize}.
12843 @item -i @var{string}
12844 Indent old-style parameter declarations with the string @var{string}.
12845 This option applies only to @code{protoize}.
12847 @code{unprotoize} converts prototyped function definitions to old-style
12848 function definitions, where the arguments are declared between the
12849 argument list and the initial @samp{@{}. By default, @code{unprotoize}
12850 uses five spaces as the indentation. If you want to indent with just
12851 one space instead, use @option{-i " "}.
12854 Keep the @samp{.X} files. Normally, they are deleted after conversion
12858 Add explicit local declarations. @code{protoize} with @option{-l} inserts
12859 a prototype declaration for each function in each block which calls the
12860 function without any declaration. This option applies only to
12864 Make no real changes. This mode just prints information about the conversions
12865 that would have been done without @option{-n}.
12868 Make no @samp{.save} files. The original files are simply deleted.
12869 Use this option with caution.
12871 @item -p @var{program}
12872 Use the program @var{program} as the compiler. Normally, the name
12873 @file{gcc} is used.
12876 Work quietly. Most warnings are suppressed.
12879 Print the version number, just like @option{-v} for @command{gcc}.
12882 If you need special compiler options to compile one of your program's
12883 source files, then you should generate that file's @samp{.X} file
12884 specially, by running @command{gcc} on that source file with the
12885 appropriate options and the option @option{-aux-info}. Then run
12886 @code{protoize} on the entire set of files. @code{protoize} will use
12887 the existing @samp{.X} file because it is newer than the source file.
12891 gcc -Dfoo=bar file1.c -aux-info file1.X
12896 You need to include the special files along with the rest in the
12897 @code{protoize} command, even though their @samp{.X} files already
12898 exist, because otherwise they won't get converted.
12900 @xref{Protoize Caveats}, for more information on how to use
12901 @code{protoize} successfully.