1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004 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 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}}. Use of the @command{gccbug}
53 script to report bugs is recommended.
56 See the Info entry for @command{gcc}, or
57 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
58 for contributors to GCC@.
63 @chapter GCC Command Options
64 @cindex GCC command options
65 @cindex command options
66 @cindex options, GCC command
68 @c man begin DESCRIPTION
69 When you invoke GCC, it normally does preprocessing, compilation,
70 assembly and linking. The ``overall options'' allow you to stop this
71 process at an intermediate stage. For example, the @option{-c} option
72 says not to run the linker. Then the output consists of object files
73 output by the assembler.
75 Other options are passed on to one stage of processing. Some options
76 control the preprocessor and others the compiler itself. Yet other
77 options control the assembler and linker; most of these are not
78 documented here, since you rarely need to use any of them.
80 @cindex C compilation options
81 Most of the command line options that you can use with GCC are useful
82 for C programs; when an option is only useful with another language
83 (usually C++), the explanation says so explicitly. If the description
84 for a particular option does not mention a source language, you can use
85 that option with all supported languages.
87 @cindex C++ compilation options
88 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
89 options for compiling C++ programs.
91 @cindex grouping options
92 @cindex options, grouping
93 The @command{gcc} program accepts options and file names as operands. Many
94 options have multi-letter names; therefore multiple single-letter options
95 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
98 @cindex order of options
99 @cindex options, order
100 You can mix options and other arguments. For the most part, the order
101 you use doesn't matter. Order does matter when you use several options
102 of the same kind; for example, if you specify @option{-L} more than once,
103 the directories are searched in the order specified.
105 Many options have long names starting with @samp{-f} or with
106 @samp{-W}---for example, @option{-fforce-mem},
107 @option{-fstrength-reduce}, @option{-Wformat} and so on. Most of
108 these have both positive and negative forms; the negative form of
109 @option{-ffoo} would be @option{-fno-foo}. This manual documents
110 only one of these two forms, whichever one is not the default.
114 @xref{Option Index}, for an index to GCC's options.
117 * Option Summary:: Brief list of all options, without explanations.
118 * Overall Options:: Controlling the kind of output:
119 an executable, object files, assembler files,
120 or preprocessed source.
121 * Invoking G++:: Compiling C++ programs.
122 * C Dialect Options:: Controlling the variant of C language compiled.
123 * C++ Dialect Options:: Variations on C++.
124 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
126 * Language Independent Options:: Controlling how diagnostics should be
128 * Warning Options:: How picky should the compiler be?
129 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
130 * Optimize Options:: How much optimization?
131 * Preprocessor Options:: Controlling header files and macro definitions.
132 Also, getting dependency information for Make.
133 * Assembler Options:: Passing options to the assembler.
134 * Link Options:: Specifying libraries and so on.
135 * Directory Options:: Where to find header files and libraries.
136 Where to find the compiler executable files.
137 * Spec Files:: How to pass switches to sub-processes.
138 * Target Options:: Running a cross-compiler, or an old version of GCC.
139 * Submodel Options:: Specifying minor hardware or convention variations,
140 such as 68010 vs 68020.
141 * Code Gen Options:: Specifying conventions for function calls, data layout
143 * Environment Variables:: Env vars that affect GCC.
144 * Precompiled Headers:: Compiling a header once, and using it many times.
145 * Running Protoize:: Automatically adding or removing function prototypes.
151 @section Option Summary
153 Here is a summary of all the options, grouped by type. Explanations are
154 in the following sections.
157 @item Overall Options
158 @xref{Overall Options,,Options Controlling the Kind of Output}.
159 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
160 -x @var{language} -v -### --help --target-help --version}
162 @item C Language Options
163 @xref{C Dialect Options,,Options Controlling C Dialect}.
164 @gccoptlist{-ansi -std=@var{standard} -aux-info @var{filename} @gol
165 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
166 -fhosted -ffreestanding -fms-extensions @gol
167 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
168 -fallow-single-precision -fcond-mismatch @gol
169 -fsigned-bitfields -fsigned-char @gol
170 -funsigned-bitfields -funsigned-char}
172 @item C++ Language Options
173 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
174 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
175 -fconserve-space -fno-const-strings @gol
176 -fno-elide-constructors @gol
177 -fno-enforce-eh-specs @gol
178 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
179 -fno-implicit-templates @gol
180 -fno-implicit-inline-templates @gol
181 -fno-implement-inlines -fms-extensions @gol
182 -fno-nonansi-builtins -fno-operator-names @gol
183 -fno-optional-diags -fpermissive @gol
184 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
185 -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
186 -fno-default-inline -fvisibility-inlines-hidden @gol
187 -Wabi -Wctor-dtor-privacy @gol
188 -Wnon-virtual-dtor -Wreorder @gol
189 -Weffc++ -Wno-deprecated @gol
190 -Wno-non-template-friend -Wold-style-cast @gol
191 -Woverloaded-virtual -Wno-pmf-conversions @gol
192 -Wsign-promo -Wsynth}
194 @item Objective-C and Objective-C++ Language Options
195 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
196 Objective-C and Objective-C++ Dialects}.
198 -fconstant-string-class=@var{class-name} @gol
199 -fgnu-runtime -fnext-runtime @gol
200 -fno-nil-receivers @gol
201 -fobjc-exceptions @gol
202 -freplace-objc-classes @gol
205 -Wno-protocol -Wselector -Wundeclared-selector}
207 @item Language Independent Options
208 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
209 @gccoptlist{-fmessage-length=@var{n} @gol
210 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}}
212 @item Warning Options
213 @xref{Warning Options,,Options to Request or Suppress Warnings}.
214 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
215 -w -Wextra -Wall -Waggregate-return @gol
216 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
217 -Wconversion -Wno-deprecated-declarations @gol
218 -Wdisabled-optimization -Wno-div-by-zero -Wendif-labels @gol
219 -Werror -Werror-implicit-function-declaration @gol
220 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
221 -Wno-format-extra-args -Wformat-nonliteral @gol
222 -Wformat-security -Wformat-y2k @gol
223 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
224 -Wimport -Wno-import -Winit-self -Winline @gol
225 -Wno-invalid-offsetof -Winvalid-pch @gol
226 -Wlarger-than-@var{len} -Wlong-long @gol
227 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
228 -Wmissing-format-attribute -Wmissing-include-dirs @gol
229 -Wmissing-noreturn @gol
230 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
231 -Wparentheses -Wpointer-arith -Wredundant-decls @gol
232 -Wreturn-type -Wsequence-point -Wshadow @gol
233 -Wsign-compare -Wstrict-aliasing -Wstrict-aliasing=2 @gol
234 -Wswitch -Wswitch-default -Wswitch-enum @gol
235 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
236 -Wunknown-pragmas -Wunreachable-code @gol
237 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
238 -Wunused-value -Wunused-variable -Wwrite-strings @gol
241 @item C-only Warning Options
242 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
243 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
244 -Wstrict-prototypes -Wtraditional @gol
245 -Wdeclaration-after-statement}
247 @item Debugging Options
248 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
249 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
250 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
251 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @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-sra@r{[}-@var{n}@r{]} @gol
268 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
269 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
270 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs -ftree-based-profiling @gol
271 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
272 -ftest-coverage -ftime-report -fvar-tracking @gol
273 -g -g@var{level} -gcoff -gdwarf-2 @gol
274 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
275 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
276 -print-multi-directory -print-multi-lib @gol
277 -print-prog-name=@var{program} -print-search-dirs -Q @gol
280 @item Optimization Options
281 @xref{Optimize Options,,Options that Control Optimization}.
282 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
283 -falign-labels=@var{n} -falign-loops=@var{n} @gol
284 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
285 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
286 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
287 -fcaller-saves -fcprop-registers @gol
288 -fcse-follow-jumps -fcse-skip-blocks -fdata-sections @gol
289 -fdelayed-branch -fdelete-null-pointer-checks @gol
290 -fexpensive-optimizations -ffast-math -ffloat-store @gol
291 -fforce-addr -fforce-mem -ffunction-sections @gol
292 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
293 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2 @gol
294 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
295 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
296 -fmodulo-sched -fmove-all-movables -fnew-ra -fno-branch-count-reg @gol
297 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants @gol
298 -fno-function-cse -fno-guess-branch-probability @gol
299 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
300 -funsafe-math-optimizations -ffinite-math-only @gol
301 -fno-trapping-math -fno-zero-initialized-in-bss @gol
302 -fomit-frame-pointer -foptimize-register-move @gol
303 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
304 -fprofile-generate -fprofile-use @gol
305 -freduce-all-givs -fregmove -frename-registers @gol
306 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
307 -frerun-cse-after-loop -frerun-loop-opt @gol
308 -frounding-math -fschedule-insns -fschedule-insns2 @gol
309 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
310 -fsched-spec-load-dangerous @gol
311 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
312 -fsched2-use-superblocks @gol
313 -fsched2-use-traces -freschedule-modulo-scheduled-loops @gol
314 -fsignaling-nans -fsingle-precision-constant @gol
315 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
316 -funroll-all-loops -funroll-loops -fpeel-loops @gol
317 -funswitch-loops -fold-unroll-loops -fold-unroll-all-loops @gol
318 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
320 -ftree-dominator-opts -ftree-dse -ftree-copyrename @gol
321 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
322 --param @var{name}=@var{value}
323 -O -O0 -O1 -O2 -O3 -Os}
325 @item Preprocessor Options
326 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
327 @gccoptlist{-A@var{question}=@var{answer} @gol
328 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
329 -C -dD -dI -dM -dN @gol
330 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
331 -idirafter @var{dir} @gol
332 -include @var{file} -imacros @var{file} @gol
333 -iprefix @var{file} -iwithprefix @var{dir} @gol
334 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
335 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
336 -P -fworking-directory -remap @gol
337 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
338 -Xpreprocessor @var{option}}
340 @item Assembler Option
341 @xref{Assembler Options,,Passing Options to the Assembler}.
342 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
345 @xref{Link Options,,Options for Linking}.
346 @gccoptlist{@var{object-file-name} -l@var{library} @gol
347 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
348 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
349 -Wl,@var{option} -Xlinker @var{option} @gol
352 @item Directory Options
353 @xref{Directory Options,,Options for Directory Search}.
354 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-}
357 @c I wrote this xref this way to avoid overfull hbox. -- rms
358 @xref{Target Options}.
359 @gccoptlist{-V @var{version} -b @var{machine}}
361 @item Machine Dependent Options
362 @xref{Submodel Options,,Hardware Models and Configurations}.
363 @c This list is ordered alphanumerically by subsection name.
364 @c Try and put the significant identifier (CPU or system) first,
365 @c so users have a clue at guessing where the ones they want will be.
368 @gccoptlist{-EB -EL @gol
369 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
370 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
373 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
374 -mabi=@var{name} @gol
375 -mapcs-stack-check -mno-apcs-stack-check @gol
376 -mapcs-float -mno-apcs-float @gol
377 -mapcs-reentrant -mno-apcs-reentrant @gol
378 -msched-prolog -mno-sched-prolog @gol
379 -mlittle-endian -mbig-endian -mwords-little-endian @gol
380 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
381 -mthumb-interwork -mno-thumb-interwork @gol
382 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
383 -mstructure-size-boundary=@var{n} @gol
384 -mabort-on-noreturn @gol
385 -mlong-calls -mno-long-calls @gol
386 -msingle-pic-base -mno-single-pic-base @gol
387 -mpic-register=@var{reg} @gol
388 -mnop-fun-dllimport @gol
389 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
390 -mpoke-function-name @gol
392 -mtpcs-frame -mtpcs-leaf-frame @gol
393 -mcaller-super-interworking -mcallee-super-interworking}
396 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
397 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
400 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
401 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
402 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
403 -mstack-align -mdata-align -mconst-align @gol
404 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
405 -melf -maout -melinux -mlinux -sim -sim2 @gol
406 -mmul-bug-workaround -mno-mul-bug-workaround}
408 @emph{Darwin Options}
409 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
410 -arch_only -bind_at_load -bundle -bundle_loader @gol
411 -client_name -compatibility_version -current_version @gol
413 -dependency-file -dylib_file -dylinker_install_name @gol
414 -dynamic -dynamiclib -exported_symbols_list @gol
415 -filelist -flat_namespace -force_cpusubtype_ALL @gol
416 -force_flat_namespace -headerpad_max_install_names @gol
417 -image_base -init -install_name -keep_private_externs @gol
418 -multi_module -multiply_defined -multiply_defined_unused @gol
419 -noall_load -no_dead_strip_inits_and_terms @gol
420 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
421 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
422 -private_bundle -read_only_relocs -sectalign @gol
423 -sectobjectsymbols -whyload -seg1addr @gol
424 -sectcreate -sectobjectsymbols -sectorder @gol
425 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
426 -segprot -segs_read_only_addr -segs_read_write_addr @gol
427 -single_module -static -sub_library -sub_umbrella @gol
428 -twolevel_namespace -umbrella -undefined @gol
429 -unexported_symbols_list -weak_reference_mismatches @gol
430 -whatsloaded -F -gused -gfull -mone-byte-bool}
432 @emph{DEC Alpha Options}
433 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
434 -mieee -mieee-with-inexact -mieee-conformant @gol
435 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
436 -mtrap-precision=@var{mode} -mbuild-constants @gol
437 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
438 -mbwx -mmax -mfix -mcix @gol
439 -mfloat-vax -mfloat-ieee @gol
440 -mexplicit-relocs -msmall-data -mlarge-data @gol
441 -msmall-text -mlarge-text @gol
442 -mmemory-latency=@var{time}}
444 @emph{DEC Alpha/VMS Options}
445 @gccoptlist{-mvms-return-codes}
448 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
449 -mhard-float -msoft-float @gol
450 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
451 -mdouble -mno-double @gol
452 -mmedia -mno-media -mmuladd -mno-muladd @gol
453 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic -mlinked-fp @gol
454 -mlibrary-pic -macc-4 -macc-8 @gol
455 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
456 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
457 -mvliw-branch -mno-vliw-branch @gol
458 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
459 -mno-nested-cond-exec -mtomcat-stats @gol
462 @emph{H8/300 Options}
463 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
466 @gccoptlist{-march=@var{architecture-type} @gol
467 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
468 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
469 -mfixed-range=@var{register-range} @gol
470 -mjump-in-delay -mlinker-opt -mlong-calls @gol
471 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
472 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
473 -mno-jump-in-delay -mno-long-load-store @gol
474 -mno-portable-runtime -mno-soft-float @gol
475 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
476 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
477 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
478 -nolibdld -static -threads}
480 @emph{i386 and x86-64 Options}
481 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
482 -mfpmath=@var{unit} @gol
483 -masm=@var{dialect} -mno-fancy-math-387 @gol
484 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
485 -mno-wide-multiply -mrtd -malign-double @gol
486 -mpreferred-stack-boundary=@var{num} @gol
487 -mmmx -msse -msse2 -msse3 -m3dnow @gol
488 -mthreads -mno-align-stringops -minline-all-stringops @gol
489 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
490 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
491 -mno-red-zone -mno-tls-direct-seg-refs @gol
492 -mcmodel=@var{code-model} @gol
496 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
497 -mvolatile-asm-stop -mb-step -mregister-names -mno-sdata @gol
498 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
499 -minline-float-divide-max-throughput @gol
500 -minline-int-divide-min-latency @gol
501 -minline-int-divide-max-throughput -mno-dwarf2-asm @gol
502 -mfixed-range=@var{register-range}}
504 @emph{M32R/D Options}
505 @gccoptlist{-m32r2 -m32rx -m32r @gol
507 -malign-loops -mno-align-loops @gol
508 -missue-rate=@var{number} @gol
509 -mbranch-cost=@var{number} @gol
510 -mmodel=@var{code-size-model-type} @gol
511 -msdata=@var{sdata-type} @gol
512 -mno-flush-func -mflush-func=@var{name} @gol
513 -mno-flush-trap -mflush-trap=@var{number} @gol
516 @emph{M680x0 Options}
517 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
518 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
519 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
520 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
521 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
523 @emph{M68hc1x Options}
524 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
525 -mauto-incdec -minmax -mlong-calls -mshort @gol
526 -msoft-reg-count=@var{count}}
529 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
530 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
531 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
532 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
533 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
536 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
537 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
538 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
539 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
540 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
541 -mint64 -mlong64 -mlong32 @gol
542 -G@var{num} -membedded-data -mno-embedded-data @gol
543 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
544 -msplit-addresses -mno-split-addresses @gol
545 -mexplicit-relocs -mno-explicit-relocs @gol
546 -mcheck-zero-division -mno-check-zero-division @gol
547 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
548 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
549 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
550 -mfix-vr4120 -mno-fix-vr4120 -mfix-sb1 -mno-fix-sb1 @gol
551 -mflush-func=@var{func} -mno-flush-func @gol
552 -mbranch-likely -mno-branch-likely @gol
553 -mfp-exceptions -mno-fp-exceptions @gol
554 -mvr4130-align -mno-vr4130-align}
557 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
558 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
559 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
560 -mno-base-addresses -msingle-exit -mno-single-exit}
562 @emph{MN10300 Options}
563 @gccoptlist{-mmult-bug -mno-mult-bug @gol
564 -mam33 -mno-am33 @gol
565 -mam33-2 -mno-am33-2 @gol
569 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
570 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
571 -mregparam -mnoregparam -msb -mnosb @gol
572 -mbitfield -mnobitfield -mhimem -mnohimem}
574 @emph{PDP-11 Options}
575 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
576 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
577 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
578 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
579 -mbranch-expensive -mbranch-cheap @gol
580 -msplit -mno-split -munix-asm -mdec-asm}
582 @emph{PowerPC Options}
583 See RS/6000 and PowerPC Options.
585 @emph{RS/6000 and PowerPC Options}
586 @gccoptlist{-mcpu=@var{cpu-type} @gol
587 -mtune=@var{cpu-type} @gol
588 -mpower -mno-power -mpower2 -mno-power2 @gol
589 -mpowerpc -mpowerpc64 -mno-powerpc @gol
590 -maltivec -mno-altivec @gol
591 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
592 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
593 -mnew-mnemonics -mold-mnemonics @gol
594 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
595 -m64 -m32 -mxl-call -mno-xl-call -mpe @gol
596 -malign-power -malign-natural @gol
597 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
598 -mstring -mno-string -mupdate -mno-update @gol
599 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
600 -mstrict-align -mno-strict-align -mrelocatable @gol
601 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
602 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
603 -mdynamic-no-pic @gol
604 -mprioritize-restricted-insns=@var{priority} @gol
605 -msched-costly-dep=@var{dependence_type} @gol
606 -minsert-sched-nops=@var{scheme} @gol
607 -mcall-sysv -mcall-netbsd @gol
608 -maix-struct-return -msvr4-struct-return @gol
609 -mabi=altivec -mabi=no-altivec @gol
610 -mabi=spe -mabi=no-spe @gol
611 -misel=yes -misel=no @gol
612 -mspe=yes -mspe=no @gol
613 -mfloat-gprs=yes -mfloat-gprs=no @gol
614 -mprototype -mno-prototype @gol
615 -msim -mmvme -mads -myellowknife -memb -msdata @gol
616 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
618 @emph{S/390 and zSeries Options}
619 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
620 -mhard-float -msoft-float -mbackchain -mno-backchain -mkernel-backchain @gol
621 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
622 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
623 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd}
626 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
627 -m4-nofpu -m4-single-only -m4-single -m4 @gol
628 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
629 -m5-64media -m5-64media-nofpu @gol
630 -m5-32media -m5-32media-nofpu @gol
631 -m5-compact -m5-compact-nofpu @gol
632 -mb -ml -mdalign -mrelax @gol
633 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
634 -mieee -misize -mpadstruct -mspace @gol
635 -mprefergot -musermode}
638 @gccoptlist{-mcpu=@var{cpu-type} @gol
639 -mtune=@var{cpu-type} @gol
640 -mcmodel=@var{code-model} @gol
641 -m32 -m64 -mapp-regs -mno-app-regs @gol
642 -mfaster-structs -mno-faster-structs @gol
643 -mfpu -mno-fpu -mhard-float -msoft-float @gol
644 -mhard-quad-float -msoft-quad-float @gol
645 -mimpure-text -mno-impure-text -mlittle-endian @gol
646 -mstack-bias -mno-stack-bias @gol
647 -munaligned-doubles -mno-unaligned-doubles @gol
648 -mv8plus -mno-v8plus -mvis -mno-vis
651 @emph{System V Options}
652 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
654 @emph{TMS320C3x/C4x Options}
655 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
656 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
657 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
658 -mparallel-insns -mparallel-mpy -mpreserve-float}
661 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
662 -mprolog-function -mno-prolog-function -mspace @gol
663 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
664 -mapp-regs -mno-app-regs @gol
665 -mdisable-callt -mno-disable-callt @gol
671 @gccoptlist{-mg -mgnu -munix}
673 @emph{x86-64 Options}
674 See i386 and x86-64 Options.
676 @emph{Xstormy16 Options}
679 @emph{Xtensa Options}
680 @gccoptlist{-mconst16 -mno-const16 @gol
681 -mfused-madd -mno-fused-madd @gol
682 -mtext-section-literals -mno-text-section-literals @gol
683 -mtarget-align -mno-target-align @gol
684 -mlongcalls -mno-longcalls}
686 @emph{zSeries Options}
687 See S/390 and zSeries Options.
689 @item Code Generation Options
690 @xref{Code Gen Options,,Options for Code Generation Conventions}.
691 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
692 -ffixed-@var{reg} -fexceptions @gol
693 -fnon-call-exceptions -funwind-tables @gol
694 -fasynchronous-unwind-tables @gol
695 -finhibit-size-directive -finstrument-functions @gol
696 -fno-common -fno-ident @gol
697 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
698 -freg-struct-return -fshared-data -fshort-enums @gol
699 -fshort-double -fshort-wchar @gol
700 -fverbose-asm -fpack-struct -fstack-check @gol
701 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
702 -fargument-alias -fargument-noalias @gol
703 -fargument-noalias-global -fleading-underscore @gol
704 -ftls-model=@var{model} @gol
705 -ftrapv -fwrapv -fbounds-check @gol
710 * Overall Options:: Controlling the kind of output:
711 an executable, object files, assembler files,
712 or preprocessed source.
713 * C Dialect Options:: Controlling the variant of C language compiled.
714 * C++ Dialect Options:: Variations on C++.
715 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
717 * Language Independent Options:: Controlling how diagnostics should be
719 * Warning Options:: How picky should the compiler be?
720 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
721 * Optimize Options:: How much optimization?
722 * Preprocessor Options:: Controlling header files and macro definitions.
723 Also, getting dependency information for Make.
724 * Assembler Options:: Passing options to the assembler.
725 * Link Options:: Specifying libraries and so on.
726 * Directory Options:: Where to find header files and libraries.
727 Where to find the compiler executable files.
728 * Spec Files:: How to pass switches to sub-processes.
729 * Target Options:: Running a cross-compiler, or an old version of GCC.
732 @node Overall Options
733 @section Options Controlling the Kind of Output
735 Compilation can involve up to four stages: preprocessing, compilation
736 proper, assembly and linking, always in that order. GCC is capable of
737 preprocessing and compiling several files either into several
738 assembler input files, or into one assembler input file; then each
739 assembler input file produces an object file, and linking combines all
740 the object files (those newly compiled, and those specified as input)
741 into an executable file.
743 @cindex file name suffix
744 For any given input file, the file name suffix determines what kind of
749 C source code which must be preprocessed.
752 C source code which should not be preprocessed.
755 C++ source code which should not be preprocessed.
758 Objective-C source code. Note that you must link with the @file{libobjc}
759 library to make an Objective-C program work.
762 Objective-C source code which should not be preprocessed.
766 Objective-C++ source code. Note that you must link with the @file{libobjc}
767 library to make an Objective-C++ program work. Note that @samp{.M} refers
768 to a literal capital M@.
771 Objective-C++ source code which should not be preprocessed.
774 C, C++, Objective-C or Objective-C++ header file to be turned into a
779 @itemx @var{file}.cxx
780 @itemx @var{file}.cpp
781 @itemx @var{file}.CPP
782 @itemx @var{file}.c++
784 C++ source code which must be preprocessed. Note that in @samp{.cxx},
785 the last two letters must both be literally @samp{x}. Likewise,
786 @samp{.C} refers to a literal capital C@.
790 C++ header file to be turned into a precompiled header.
793 @itemx @var{file}.for
794 @itemx @var{file}.FOR
795 Fortran source code which should not be preprocessed.
798 @itemx @var{file}.fpp
799 @itemx @var{file}.FPP
800 Fortran source code which must be preprocessed (with the traditional
804 Fortran source code which must be preprocessed with a RATFOR
805 preprocessor (not included with GCC)@.
808 @itemx @var{file}.f95
809 Fortran 90/95 source code which should not be preprocessed.
811 @c FIXME: Descriptions of Java file types.
818 Ada source code file which contains a library unit declaration (a
819 declaration of a package, subprogram, or generic, or a generic
820 instantiation), or a library unit renaming declaration (a package,
821 generic, or subprogram renaming declaration). Such files are also
824 @itemx @var{file}.adb
825 Ada source code file containing a library unit body (a subprogram or
826 package body). Such files are also called @dfn{bodies}.
828 @c GCC also knows about some suffixes for languages not yet included:
837 Assembler code which must be preprocessed.
840 An object file to be fed straight into linking.
841 Any file name with no recognized suffix is treated this way.
845 You can specify the input language explicitly with the @option{-x} option:
848 @item -x @var{language}
849 Specify explicitly the @var{language} for the following input files
850 (rather than letting the compiler choose a default based on the file
851 name suffix). This option applies to all following input files until
852 the next @option{-x} option. Possible values for @var{language} are:
854 c c-header c-cpp-output
855 c++ c++-header c++-cpp-output
856 objective-c objective-c-header objective-c-cpp-output
857 objective-c++ objective-c++-header objective-c++-cpp-output
858 assembler assembler-with-cpp
860 f77 f77-cpp-input ratfor
867 Turn off any specification of a language, so that subsequent files are
868 handled according to their file name suffixes (as they are if @option{-x}
869 has not been used at all).
871 @item -pass-exit-codes
872 @opindex pass-exit-codes
873 Normally the @command{gcc} program will exit with the code of 1 if any
874 phase of the compiler returns a non-success return code. If you specify
875 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
876 numerically highest error produced by any phase that returned an error
880 If you only want some of the stages of compilation, you can use
881 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
882 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
883 @command{gcc} is to stop. Note that some combinations (for example,
884 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
889 Compile or assemble the source files, but do not link. The linking
890 stage simply is not done. The ultimate output is in the form of an
891 object file for each source file.
893 By default, the object file name for a source file is made by replacing
894 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
896 Unrecognized input files, not requiring compilation or assembly, are
901 Stop after the stage of compilation proper; do not assemble. The output
902 is in the form of an assembler code file for each non-assembler input
905 By default, the assembler file name for a source file is made by
906 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
908 Input files that don't require compilation are ignored.
912 Stop after the preprocessing stage; do not run the compiler proper. The
913 output is in the form of preprocessed source code, which is sent to the
916 Input files which don't require preprocessing are ignored.
918 @cindex output file option
921 Place output in file @var{file}. This applies regardless to whatever
922 sort of output is being produced, whether it be an executable file,
923 an object file, an assembler file or preprocessed C code.
925 If @option{-o} is not specified, the default is to put an executable
926 file in @file{a.out}, the object file for
927 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
928 assembler file in @file{@var{source}.s}, a precompiled header file in
929 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
934 Print (on standard error output) the commands executed to run the stages
935 of compilation. Also print the version number of the compiler driver
936 program and of the preprocessor and the compiler proper.
940 Like @option{-v} except the commands are not executed and all command
941 arguments are quoted. This is useful for shell scripts to capture the
942 driver-generated command lines.
946 Use pipes rather than temporary files for communication between the
947 various stages of compilation. This fails to work on some systems where
948 the assembler is unable to read from a pipe; but the GNU assembler has
953 If you are compiling multiple source files, this option tells the driver
954 to pass all the source files to the compiler at once (for those
955 languages for which the compiler can handle this). This will allow
956 intermodule analysis (IMA) to be performed by the compiler. Currently the only
957 language for which this is supported is C. If you pass source files for
958 multiple languages to the driver, using this option, the driver will invoke
959 the compiler(s) that support IMA once each, passing each compiler all the
960 source files appropriate for it. For those languages that do not support
961 IMA this option will be ignored, and the compiler will be invoked once for
962 each source file in that language. If you use this option in conjunction
963 with -save-temps, the compiler will generate multiple pre-processed files
964 (one for each source file), but only one (combined) .o or .s file.
968 Print (on the standard output) a description of the command line options
969 understood by @command{gcc}. If the @option{-v} option is also specified
970 then @option{--help} will also be passed on to the various processes
971 invoked by @command{gcc}, so that they can display the command line options
972 they accept. If the @option{-Wextra} option is also specified then command
973 line options which have no documentation associated with them will also
978 Print (on the standard output) a description of target specific command
979 line options for each tool.
983 Display the version number and copyrights of the invoked GCC.
987 @section Compiling C++ Programs
989 @cindex suffixes for C++ source
990 @cindex C++ source file suffixes
991 C++ source files conventionally use one of the suffixes @samp{.C},
992 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
993 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
994 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
995 files with these names and compiles them as C++ programs even if you
996 call the compiler the same way as for compiling C programs (usually
997 with the name @command{gcc}).
1001 However, C++ programs often require class libraries as well as a
1002 compiler that understands the C++ language---and under some
1003 circumstances, you might want to compile programs or header files from
1004 standard input, or otherwise without a suffix that flags them as C++
1005 programs. You might also like to precompile a C header file with a
1006 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
1007 program that calls GCC with the default language set to C++, and
1008 automatically specifies linking against the C++ library. On many
1009 systems, @command{g++} is also installed with the name @command{c++}.
1011 @cindex invoking @command{g++}
1012 When you compile C++ programs, you may specify many of the same
1013 command-line options that you use for compiling programs in any
1014 language; or command-line options meaningful for C and related
1015 languages; or options that are meaningful only for C++ programs.
1016 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1017 explanations of options for languages related to C@.
1018 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1019 explanations of options that are meaningful only for C++ programs.
1021 @node C Dialect Options
1022 @section Options Controlling C Dialect
1023 @cindex dialect options
1024 @cindex language dialect options
1025 @cindex options, dialect
1027 The following options control the dialect of C (or languages derived
1028 from C, such as C++, Objective-C and Objective-C++) that the compiler
1032 @cindex ANSI support
1036 In C mode, support all ISO C90 programs. In C++ mode,
1037 remove GNU extensions that conflict with ISO C++.
1039 This turns off certain features of GCC that are incompatible with ISO
1040 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1041 such as the @code{asm} and @code{typeof} keywords, and
1042 predefined macros such as @code{unix} and @code{vax} that identify the
1043 type of system you are using. It also enables the undesirable and
1044 rarely used ISO trigraph feature. For the C compiler,
1045 it disables recognition of C++ style @samp{//} comments as well as
1046 the @code{inline} keyword.
1048 The alternate keywords @code{__asm__}, @code{__extension__},
1049 @code{__inline__} and @code{__typeof__} continue to work despite
1050 @option{-ansi}. You would not want to use them in an ISO C program, of
1051 course, but it is useful to put them in header files that might be included
1052 in compilations done with @option{-ansi}. Alternate predefined macros
1053 such as @code{__unix__} and @code{__vax__} are also available, with or
1054 without @option{-ansi}.
1056 The @option{-ansi} option does not cause non-ISO programs to be
1057 rejected gratuitously. For that, @option{-pedantic} is required in
1058 addition to @option{-ansi}. @xref{Warning Options}.
1060 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1061 option is used. Some header files may notice this macro and refrain
1062 from declaring certain functions or defining certain macros that the
1063 ISO standard doesn't call for; this is to avoid interfering with any
1064 programs that might use these names for other things.
1066 Functions which would normally be built in but do not have semantics
1067 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1068 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1069 built-in functions provided by GCC}, for details of the functions
1074 Determine the language standard. This option is currently only
1075 supported when compiling C or C++. A value for this option must be
1076 provided; possible values are
1081 ISO C90 (same as @option{-ansi}).
1083 @item iso9899:199409
1084 ISO C90 as modified in amendment 1.
1090 ISO C99. Note that this standard is not yet fully supported; see
1091 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1092 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1095 Default, ISO C90 plus GNU extensions (including some C99 features).
1099 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1100 this will become the default. The name @samp{gnu9x} is deprecated.
1103 The 1998 ISO C++ standard plus amendments.
1106 The same as @option{-std=c++98} plus GNU extensions. This is the
1107 default for C++ code.
1110 Even when this option is not specified, you can still use some of the
1111 features of newer standards in so far as they do not conflict with
1112 previous C standards. For example, you may use @code{__restrict__} even
1113 when @option{-std=c99} is not specified.
1115 The @option{-std} options specifying some version of ISO C have the same
1116 effects as @option{-ansi}, except that features that were not in ISO C90
1117 but are in the specified version (for example, @samp{//} comments and
1118 the @code{inline} keyword in ISO C99) are not disabled.
1120 @xref{Standards,,Language Standards Supported by GCC}, for details of
1121 these standard versions.
1123 @item -aux-info @var{filename}
1125 Output to the given filename prototyped declarations for all functions
1126 declared and/or defined in a translation unit, including those in header
1127 files. This option is silently ignored in any language other than C@.
1129 Besides declarations, the file indicates, in comments, the origin of
1130 each declaration (source file and line), whether the declaration was
1131 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1132 @samp{O} for old, respectively, in the first character after the line
1133 number and the colon), and whether it came from a declaration or a
1134 definition (@samp{C} or @samp{F}, respectively, in the following
1135 character). In the case of function definitions, a K&R-style list of
1136 arguments followed by their declarations is also provided, inside
1137 comments, after the declaration.
1141 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1142 keyword, so that code can use these words as identifiers. You can use
1143 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1144 instead. @option{-ansi} implies @option{-fno-asm}.
1146 In C++, this switch only affects the @code{typeof} keyword, since
1147 @code{asm} and @code{inline} are standard keywords. You may want to
1148 use the @option{-fno-gnu-keywords} flag instead, which has the same
1149 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1150 switch only affects the @code{asm} and @code{typeof} keywords, since
1151 @code{inline} is a standard keyword in ISO C99.
1154 @itemx -fno-builtin-@var{function}
1155 @opindex fno-builtin
1156 @cindex built-in functions
1157 Don't recognize built-in functions that do not begin with
1158 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1159 functions provided by GCC}, for details of the functions affected,
1160 including those which are not built-in functions when @option{-ansi} or
1161 @option{-std} options for strict ISO C conformance are used because they
1162 do not have an ISO standard meaning.
1164 GCC normally generates special code to handle certain built-in functions
1165 more efficiently; for instance, calls to @code{alloca} may become single
1166 instructions that adjust the stack directly, and calls to @code{memcpy}
1167 may become inline copy loops. The resulting code is often both smaller
1168 and faster, but since the function calls no longer appear as such, you
1169 cannot set a breakpoint on those calls, nor can you change the behavior
1170 of the functions by linking with a different library.
1172 With the @option{-fno-builtin-@var{function}} option
1173 only the built-in function @var{function} is
1174 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1175 function is named this is not built-in in this version of GCC, this
1176 option is ignored. There is no corresponding
1177 @option{-fbuiltin-@var{function}} option; if you wish to enable
1178 built-in functions selectively when using @option{-fno-builtin} or
1179 @option{-ffreestanding}, you may define macros such as:
1182 #define abs(n) __builtin_abs ((n))
1183 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1188 @cindex hosted environment
1190 Assert that compilation takes place in a hosted environment. This implies
1191 @option{-fbuiltin}. A hosted environment is one in which the
1192 entire standard library is available, and in which @code{main} has a return
1193 type of @code{int}. Examples are nearly everything except a kernel.
1194 This is equivalent to @option{-fno-freestanding}.
1196 @item -ffreestanding
1197 @opindex ffreestanding
1198 @cindex hosted environment
1200 Assert that compilation takes place in a freestanding environment. This
1201 implies @option{-fno-builtin}. A freestanding environment
1202 is one in which the standard library may not exist, and program startup may
1203 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1204 This is equivalent to @option{-fno-hosted}.
1206 @xref{Standards,,Language Standards Supported by GCC}, for details of
1207 freestanding and hosted environments.
1209 @item -fms-extensions
1210 @opindex fms-extensions
1211 Accept some non-standard constructs used in Microsoft header files.
1215 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1216 options for strict ISO C conformance) implies @option{-trigraphs}.
1218 @item -no-integrated-cpp
1219 @opindex no-integrated-cpp
1220 Performs a compilation in two passes: preprocessing and compiling. This
1221 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1222 @option{-B} option. The user supplied compilation step can then add in
1223 an additional preprocessing step after normal preprocessing but before
1224 compiling. The default is to use the integrated cpp (internal cpp)
1226 The semantics of this option will change if "cc1", "cc1plus", and
1227 "cc1obj" are merged.
1229 @cindex traditional C language
1230 @cindex C language, traditional
1232 @itemx -traditional-cpp
1233 @opindex traditional-cpp
1234 @opindex traditional
1235 Formerly, these options caused GCC to attempt to emulate a pre-standard
1236 C compiler. They are now only supported with the @option{-E} switch.
1237 The preprocessor continues to support a pre-standard mode. See the GNU
1238 CPP manual for details.
1240 @item -fcond-mismatch
1241 @opindex fcond-mismatch
1242 Allow conditional expressions with mismatched types in the second and
1243 third arguments. The value of such an expression is void. This option
1244 is not supported for C++.
1246 @item -funsigned-char
1247 @opindex funsigned-char
1248 Let the type @code{char} be unsigned, like @code{unsigned char}.
1250 Each kind of machine has a default for what @code{char} should
1251 be. It is either like @code{unsigned char} by default or like
1252 @code{signed char} by default.
1254 Ideally, a portable program should always use @code{signed char} or
1255 @code{unsigned char} when it depends on the signedness of an object.
1256 But many programs have been written to use plain @code{char} and
1257 expect it to be signed, or expect it to be unsigned, depending on the
1258 machines they were written for. This option, and its inverse, let you
1259 make such a program work with the opposite default.
1261 The type @code{char} is always a distinct type from each of
1262 @code{signed char} or @code{unsigned char}, even though its behavior
1263 is always just like one of those two.
1266 @opindex fsigned-char
1267 Let the type @code{char} be signed, like @code{signed char}.
1269 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1270 the negative form of @option{-funsigned-char}. Likewise, the option
1271 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1273 @item -fsigned-bitfields
1274 @itemx -funsigned-bitfields
1275 @itemx -fno-signed-bitfields
1276 @itemx -fno-unsigned-bitfields
1277 @opindex fsigned-bitfields
1278 @opindex funsigned-bitfields
1279 @opindex fno-signed-bitfields
1280 @opindex fno-unsigned-bitfields
1281 These options control whether a bit-field is signed or unsigned, when the
1282 declaration does not use either @code{signed} or @code{unsigned}. By
1283 default, such a bit-field is signed, because this is consistent: the
1284 basic integer types such as @code{int} are signed types.
1287 @node C++ Dialect Options
1288 @section Options Controlling C++ Dialect
1290 @cindex compiler options, C++
1291 @cindex C++ options, command line
1292 @cindex options, C++
1293 This section describes the command-line options that are only meaningful
1294 for C++ programs; but you can also use most of the GNU compiler options
1295 regardless of what language your program is in. For example, you
1296 might compile a file @code{firstClass.C} like this:
1299 g++ -g -frepo -O -c firstClass.C
1303 In this example, only @option{-frepo} is an option meant
1304 only for C++ programs; you can use the other options with any
1305 language supported by GCC@.
1307 Here is a list of options that are @emph{only} for compiling C++ programs:
1311 @item -fabi-version=@var{n}
1312 @opindex fabi-version
1313 Use version @var{n} of the C++ ABI. Version 2 is the version of the
1314 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1315 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1316 the version that conforms most closely to the C++ ABI specification.
1317 Therefore, the ABI obtained using version 0 will change as ABI bugs
1320 The default is version 2.
1322 @item -fno-access-control
1323 @opindex fno-access-control
1324 Turn off all access checking. This switch is mainly useful for working
1325 around bugs in the access control code.
1329 Check that the pointer returned by @code{operator new} is non-null
1330 before attempting to modify the storage allocated. This check is
1331 normally unnecessary because the C++ standard specifies that
1332 @code{operator new} will only return @code{0} if it is declared
1333 @samp{throw()}, in which case the compiler will always check the
1334 return value even without this option. In all other cases, when
1335 @code{operator new} has a non-empty exception specification, memory
1336 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1337 @samp{new (nothrow)}.
1339 @item -fconserve-space
1340 @opindex fconserve-space
1341 Put uninitialized or runtime-initialized global variables into the
1342 common segment, as C does. This saves space in the executable at the
1343 cost of not diagnosing duplicate definitions. If you compile with this
1344 flag and your program mysteriously crashes after @code{main()} has
1345 completed, you may have an object that is being destroyed twice because
1346 two definitions were merged.
1348 This option is no longer useful on most targets, now that support has
1349 been added for putting variables into BSS without making them common.
1351 @item -fno-const-strings
1352 @opindex fno-const-strings
1353 Give string constants type @code{char *} instead of type @code{const
1354 char *}. By default, G++ uses type @code{const char *} as required by
1355 the standard. Even if you use @option{-fno-const-strings}, you cannot
1356 actually modify the value of a string constant.
1358 This option might be removed in a future release of G++. For maximum
1359 portability, you should structure your code so that it works with
1360 string constants that have type @code{const char *}.
1362 @item -fno-elide-constructors
1363 @opindex fno-elide-constructors
1364 The C++ standard allows an implementation to omit creating a temporary
1365 which is only used to initialize another object of the same type.
1366 Specifying this option disables that optimization, and forces G++ to
1367 call the copy constructor in all cases.
1369 @item -fno-enforce-eh-specs
1370 @opindex fno-enforce-eh-specs
1371 Don't check for violation of exception specifications at runtime. This
1372 option violates the C++ standard, but may be useful for reducing code
1373 size in production builds, much like defining @samp{NDEBUG}. The compiler
1374 will still optimize based on the exception specifications.
1377 @itemx -fno-for-scope
1379 @opindex fno-for-scope
1380 If @option{-ffor-scope} is specified, the scope of variables declared in
1381 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1382 as specified by the C++ standard.
1383 If @option{-fno-for-scope} is specified, the scope of variables declared in
1384 a @i{for-init-statement} extends to the end of the enclosing scope,
1385 as was the case in old versions of G++, and other (traditional)
1386 implementations of C++.
1388 The default if neither flag is given to follow the standard,
1389 but to allow and give a warning for old-style code that would
1390 otherwise be invalid, or have different behavior.
1392 @item -fno-gnu-keywords
1393 @opindex fno-gnu-keywords
1394 Do not recognize @code{typeof} as a keyword, so that code can use this
1395 word as an identifier. You can use the keyword @code{__typeof__} instead.
1396 @option{-ansi} implies @option{-fno-gnu-keywords}.
1398 @item -fno-implicit-templates
1399 @opindex fno-implicit-templates
1400 Never emit code for non-inline templates which are instantiated
1401 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1402 @xref{Template Instantiation}, for more information.
1404 @item -fno-implicit-inline-templates
1405 @opindex fno-implicit-inline-templates
1406 Don't emit code for implicit instantiations of inline templates, either.
1407 The default is to handle inlines differently so that compiles with and
1408 without optimization will need the same set of explicit instantiations.
1410 @item -fno-implement-inlines
1411 @opindex fno-implement-inlines
1412 To save space, do not emit out-of-line copies of inline functions
1413 controlled by @samp{#pragma implementation}. This will cause linker
1414 errors if these functions are not inlined everywhere they are called.
1416 @item -fms-extensions
1417 @opindex fms-extensions
1418 Disable pedantic warnings about constructs used in MFC, such as implicit
1419 int and getting a pointer to member function via non-standard syntax.
1421 @item -fno-nonansi-builtins
1422 @opindex fno-nonansi-builtins
1423 Disable built-in declarations of functions that are not mandated by
1424 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1425 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1427 @item -fno-operator-names
1428 @opindex fno-operator-names
1429 Do not treat the operator name keywords @code{and}, @code{bitand},
1430 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1431 synonyms as keywords.
1433 @item -fno-optional-diags
1434 @opindex fno-optional-diags
1435 Disable diagnostics that the standard says a compiler does not need to
1436 issue. Currently, the only such diagnostic issued by G++ is the one for
1437 a name having multiple meanings within a class.
1440 @opindex fpermissive
1441 Downgrade some diagnostics about nonconformant code from errors to
1442 warnings. Thus, using @option{-fpermissive} will allow some
1443 nonconforming code to compile.
1447 Enable automatic template instantiation at link time. This option also
1448 implies @option{-fno-implicit-templates}. @xref{Template
1449 Instantiation}, for more information.
1453 Disable generation of information about every class with virtual
1454 functions for use by the C++ runtime type identification features
1455 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1456 of the language, you can save some space by using this flag. Note that
1457 exception handling uses the same information, but it will generate it as
1462 Emit statistics about front-end processing at the end of the compilation.
1463 This information is generally only useful to the G++ development team.
1465 @item -ftemplate-depth-@var{n}
1466 @opindex ftemplate-depth
1467 Set the maximum instantiation depth for template classes to @var{n}.
1468 A limit on the template instantiation depth is needed to detect
1469 endless recursions during template class instantiation. ANSI/ISO C++
1470 conforming programs must not rely on a maximum depth greater than 17.
1472 @item -fuse-cxa-atexit
1473 @opindex fuse-cxa-atexit
1474 Register destructors for objects with static storage duration with the
1475 @code{__cxa_atexit} function rather than the @code{atexit} function.
1476 This option is required for fully standards-compliant handling of static
1477 destructors, but will only work if your C library supports
1478 @code{__cxa_atexit}.
1480 @item -fvisibility-inlines-hidden
1481 @opindex fvisibility-inlines-hidden
1482 Causes all inlined methods to be marked with
1483 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1484 appear in the export table of a DSO and do not require a PLT indirection
1485 when used within the DSO. Enabling this option can have a dramatic effect
1486 on load and link times of a DSO as it massively reduces the size of the
1487 dynamic export table when the library makes heavy use of templates. While
1488 it can cause bloating through duplication of code within each DSO where
1489 it is used, often the wastage is less than the considerable space occupied
1490 by a long symbol name in the export table which is typical when using
1491 templates and namespaces. For even more savings, combine with the
1492 @code{-fvisibility=hidden} switch.
1496 Do not use weak symbol support, even if it is provided by the linker.
1497 By default, G++ will use weak symbols if they are available. This
1498 option exists only for testing, and should not be used by end-users;
1499 it will result in inferior code and has no benefits. This option may
1500 be removed in a future release of G++.
1504 Do not search for header files in the standard directories specific to
1505 C++, but do still search the other standard directories. (This option
1506 is used when building the C++ library.)
1509 In addition, these optimization, warning, and code generation options
1510 have meanings only for C++ programs:
1513 @item -fno-default-inline
1514 @opindex fno-default-inline
1515 Do not assume @samp{inline} for functions defined inside a class scope.
1516 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1517 functions will have linkage like inline functions; they just won't be
1520 @item -Wabi @r{(C++ only)}
1522 Warn when G++ generates code that is probably not compatible with the
1523 vendor-neutral C++ ABI. Although an effort has been made to warn about
1524 all such cases, there are probably some cases that are not warned about,
1525 even though G++ is generating incompatible code. There may also be
1526 cases where warnings are emitted even though the code that is generated
1529 You should rewrite your code to avoid these warnings if you are
1530 concerned about the fact that code generated by G++ may not be binary
1531 compatible with code generated by other compilers.
1533 The known incompatibilities at this point include:
1538 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1539 pack data into the same byte as a base class. For example:
1542 struct A @{ virtual void f(); int f1 : 1; @};
1543 struct B : public A @{ int f2 : 1; @};
1547 In this case, G++ will place @code{B::f2} into the same byte
1548 as@code{A::f1}; other compilers will not. You can avoid this problem
1549 by explicitly padding @code{A} so that its size is a multiple of the
1550 byte size on your platform; that will cause G++ and other compilers to
1551 layout @code{B} identically.
1554 Incorrect handling of tail-padding for virtual bases. G++ does not use
1555 tail padding when laying out virtual bases. For example:
1558 struct A @{ virtual void f(); char c1; @};
1559 struct B @{ B(); char c2; @};
1560 struct C : public A, public virtual B @{@};
1564 In this case, G++ will not place @code{B} into the tail-padding for
1565 @code{A}; other compilers will. You can avoid this problem by
1566 explicitly padding @code{A} so that its size is a multiple of its
1567 alignment (ignoring virtual base classes); that will cause G++ and other
1568 compilers to layout @code{C} identically.
1571 Incorrect handling of bit-fields with declared widths greater than that
1572 of their underlying types, when the bit-fields appear in a union. For
1576 union U @{ int i : 4096; @};
1580 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1581 union too small by the number of bits in an @code{int}.
1584 Empty classes can be placed at incorrect offsets. For example:
1594 struct C : public B, public A @{@};
1598 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1599 it should be placed at offset zero. G++ mistakenly believes that the
1600 @code{A} data member of @code{B} is already at offset zero.
1603 Names of template functions whose types involve @code{typename} or
1604 template template parameters can be mangled incorrectly.
1607 template <typename Q>
1608 void f(typename Q::X) @{@}
1610 template <template <typename> class Q>
1611 void f(typename Q<int>::X) @{@}
1615 Instantiations of these templates may be mangled incorrectly.
1619 @item -Wctor-dtor-privacy @r{(C++ only)}
1620 @opindex Wctor-dtor-privacy
1621 Warn when a class seems unusable because all the constructors or
1622 destructors in that class are private, and it has neither friends nor
1623 public static member functions.
1625 @item -Wnon-virtual-dtor @r{(C++ only)}
1626 @opindex Wnon-virtual-dtor
1627 Warn when a class appears to be polymorphic, thereby requiring a virtual
1628 destructor, yet it declares a non-virtual one.
1629 This warning is enabled by @option{-Wall}.
1631 @item -Wreorder @r{(C++ only)}
1633 @cindex reordering, warning
1634 @cindex warning for reordering of member initializers
1635 Warn when the order of member initializers given in the code does not
1636 match the order in which they must be executed. For instance:
1642 A(): j (0), i (1) @{ @}
1646 The compiler will rearrange the member initializers for @samp{i}
1647 and @samp{j} to match the declaration order of the members, emitting
1648 a warning to that effect. This warning is enabled by @option{-Wall}.
1651 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1654 @item -Weffc++ @r{(C++ only)}
1656 Warn about violations of the following style guidelines from Scott Meyers'
1657 @cite{Effective C++} book:
1661 Item 11: Define a copy constructor and an assignment operator for classes
1662 with dynamically allocated memory.
1665 Item 12: Prefer initialization to assignment in constructors.
1668 Item 14: Make destructors virtual in base classes.
1671 Item 15: Have @code{operator=} return a reference to @code{*this}.
1674 Item 23: Don't try to return a reference when you must return an object.
1678 Also warn about violations of the following style guidelines from
1679 Scott Meyers' @cite{More Effective C++} book:
1683 Item 6: Distinguish between prefix and postfix forms of increment and
1684 decrement operators.
1687 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1691 When selecting this option, be aware that the standard library
1692 headers do not obey all of these guidelines; use @samp{grep -v}
1693 to filter out those warnings.
1695 @item -Wno-deprecated @r{(C++ only)}
1696 @opindex Wno-deprecated
1697 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1699 @item -Wno-non-template-friend @r{(C++ only)}
1700 @opindex Wno-non-template-friend
1701 Disable warnings when non-templatized friend functions are declared
1702 within a template. Since the advent of explicit template specification
1703 support in G++, if the name of the friend is an unqualified-id (i.e.,
1704 @samp{friend foo(int)}), the C++ language specification demands that the
1705 friend declare or define an ordinary, nontemplate function. (Section
1706 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1707 could be interpreted as a particular specialization of a templatized
1708 function. Because this non-conforming behavior is no longer the default
1709 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1710 check existing code for potential trouble spots and is on by default.
1711 This new compiler behavior can be turned off with
1712 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1713 but disables the helpful warning.
1715 @item -Wold-style-cast @r{(C++ only)}
1716 @opindex Wold-style-cast
1717 Warn if an old-style (C-style) cast to a non-void type is used within
1718 a C++ program. The new-style casts (@samp{static_cast},
1719 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1720 unintended effects and much easier to search for.
1722 @item -Woverloaded-virtual @r{(C++ only)}
1723 @opindex Woverloaded-virtual
1724 @cindex overloaded virtual fn, warning
1725 @cindex warning for overloaded virtual fn
1726 Warn when a function declaration hides virtual functions from a
1727 base class. For example, in:
1734 struct B: public A @{
1739 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1747 will fail to compile.
1749 @item -Wno-pmf-conversions @r{(C++ only)}
1750 @opindex Wno-pmf-conversions
1751 Disable the diagnostic for converting a bound pointer to member function
1754 @item -Wsign-promo @r{(C++ only)}
1755 @opindex Wsign-promo
1756 Warn when overload resolution chooses a promotion from unsigned or
1757 enumerated type to a signed type, over a conversion to an unsigned type of
1758 the same size. Previous versions of G++ would try to preserve
1759 unsignedness, but the standard mandates the current behavior.
1761 @item -Wsynth @r{(C++ only)}
1763 @cindex warning for synthesized methods
1764 @cindex synthesized methods, warning
1765 Warn when G++'s synthesis behavior does not match that of cfront. For
1771 A& operator = (int);
1781 In this example, G++ will synthesize a default @samp{A& operator =
1782 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1785 @node Objective-C and Objective-C++ Dialect Options
1786 @section Options Controlling Objective-C and Objective-C++ Dialects
1788 @cindex compiler options, Objective-C and Objective-C++
1789 @cindex Objective-C and Objective-C++ options, command line
1790 @cindex options, Objective-C and Objective-C++
1791 (NOTE: This manual does not describe the Objective-C and Objective-C++
1792 languages themselves. See @xref{Standards,,Language Standards
1793 Supported by GCC}, for references.)
1795 This section describes the command-line options that are only meaningful
1796 for Objective-C and Objective-C++ programs, but you can also use most of
1797 the language-independent GNU compiler options.
1798 For example, you might compile a file @code{some_class.m} like this:
1801 gcc -g -fgnu-runtime -O -c some_class.m
1805 In this example, @option{-fgnu-runtime} is an option meant only for
1806 Objective-C and Objective-C++ programs; you can use the other options with
1807 any language supported by GCC@.
1809 Note that since Objective-C is an extension of the C language, Objective-C
1810 compilations may also use options specific to the C front-end (e.g.,
1811 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1812 C++-specific options (e.g., @option{-Wabi}).
1814 Here is a list of options that are @emph{only} for compiling Objective-C
1815 and Objective-C++ programs:
1818 @item -fconstant-string-class=@var{class-name}
1819 @opindex fconstant-string-class
1820 Use @var{class-name} as the name of the class to instantiate for each
1821 literal string specified with the syntax @code{@@"@dots{}"}. The default
1822 class name is @code{NXConstantString} if the GNU runtime is being used, and
1823 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1824 @option{-fconstant-cfstrings} option, if also present, will override the
1825 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1826 to be laid out as constant CoreFoundation strings.
1829 @opindex fgnu-runtime
1830 Generate object code compatible with the standard GNU Objective-C
1831 runtime. This is the default for most types of systems.
1833 @item -fnext-runtime
1834 @opindex fnext-runtime
1835 Generate output compatible with the NeXT runtime. This is the default
1836 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1837 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1840 @item -fno-nil-receivers
1841 @opindex fno-nil-receivers
1842 Assume that all Objective-C message dispatches (e.g.,
1843 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1844 is not @code{nil}. This allows for more efficient entry points in the runtime
1845 to be used. Currently, this option is only available in conjunction with
1846 the NeXT runtime on Mac OS X 10.3 and later.
1848 @item -fobjc-exceptions
1849 @opindex fobjc-exceptions
1850 Enable syntactic support for structured exception handling in Objective-C,
1851 similar to what is offered by C++ and Java. Currently, this option is only
1852 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1860 @@catch (AnObjCClass *exc) @{
1867 @@catch (AnotherClass *exc) @{
1870 @@catch (id allOthers) @{
1880 The @code{@@throw} statement may appear anywhere in an Objective-C or
1881 Objective-C++ program; when used inside of a @code{@@catch} block, the
1882 @code{@@throw} may appear without an argument (as shown above), in which case
1883 the object caught by the @code{@@catch} will be rethrown.
1885 Note that only (pointers to) Objective-C objects may be thrown and
1886 caught using this scheme. When an object is thrown, it will be caught
1887 by the nearest @code{@@catch} clause capable of handling objects of that type,
1888 analogously to how @code{catch} blocks work in C++ and Java. A
1889 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1890 any and all Objective-C exceptions not caught by previous @code{@@catch}
1893 The @code{@@finally} clause, if present, will be executed upon exit from the
1894 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1895 regardless of whether any exceptions are thrown, caught or rethrown
1896 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1897 of the @code{finally} clause in Java.
1899 There are several caveats to using the new exception mechanism:
1903 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
1904 idioms provided by the @code{NSException} class, the new
1905 exceptions can only be used on Mac OS X 10.3 (Panther) and later
1906 systems, due to additional functionality needed in the (NeXT) Objective-C
1910 As mentioned above, the new exceptions do not support handling
1911 types other than Objective-C objects. Furthermore, when used from
1912 Objective-C++, the Objective-C exception model does not interoperate with C++
1913 exceptions at this time. This means you cannot @code{@@throw} an exception
1914 from Objective-C and @code{catch} it in C++, or vice versa
1915 (i.e., @code{throw @dots{} @@catch}).
1918 The @option{-fobjc-exceptions} switch also enables the use of synchronization
1919 blocks for thread-safe execution:
1922 @@synchronized (ObjCClass *guard) @{
1927 Upon entering the @code{@@synchronized} block, a thread of execution shall
1928 first check whether a lock has been placed on the corresponding @code{guard}
1929 object by another thread. If it has, the current thread shall wait until
1930 the other thread relinquishes its lock. Once @code{guard} becomes available,
1931 the current thread will place its own lock on it, execute the code contained in
1932 the @code{@@synchronized} block, and finally relinquish the lock (thereby
1933 making @code{guard} available to other threads).
1935 Unlike Java, Objective-C does not allow for entire methods to be marked
1936 @code{@@synchronized}. Note that throwing exceptions out of
1937 @code{@@synchronized} blocks is allowed, and will cause the guarding object
1938 to be unlocked properly.
1940 @item -freplace-objc-classes
1941 @opindex freplace-objc-classes
1942 Emit a special marker instructing @command{ld(1)} not to statically link in
1943 the resulting object file, and allow @command{dyld(1)} to load it in at
1944 run time instead. This is used in conjunction with the Fix-and-Continue
1945 debugging mode, where the object file in question may be recompiled and
1946 dynamically reloaded in the course of program execution, without the need
1947 to restart the program itself. Currently, Fix-and-Continue functionality
1948 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
1953 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
1954 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
1955 compile time) with static class references that get initialized at load time,
1956 which improves run-time performance. Specifying the @option{-fzero-link} flag
1957 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
1958 to be retained. This is useful in Zero-Link debugging mode, since it allows
1959 for individual class implementations to be modified during program execution.
1963 Dump interface declarations for all classes seen in the source file to a
1964 file named @file{@var{sourcename}.decl}.
1967 @opindex Wno-protocol
1968 If a class is declared to implement a protocol, a warning is issued for
1969 every method in the protocol that is not implemented by the class. The
1970 default behavior is to issue a warning for every method not explicitly
1971 implemented in the class, even if a method implementation is inherited
1972 from the superclass. If you use the @code{-Wno-protocol} option, then
1973 methods inherited from the superclass are considered to be implemented,
1974 and no warning is issued for them.
1978 Warn if multiple methods of different types for the same selector are
1979 found during compilation. The check is performed on the list of methods
1980 in the final stage of compilation. Additionally, a check is performed
1981 for each selector appearing in a @code{@@selector(@dots{})}
1982 expression, and a corresponding method for that selector has been found
1983 during compilation. Because these checks scan the method table only at
1984 the end of compilation, these warnings are not produced if the final
1985 stage of compilation is not reached, for example because an error is
1986 found during compilation, or because the @code{-fsyntax-only} option is
1989 @item -Wundeclared-selector
1990 @opindex Wundeclared-selector
1991 Warn if a @code{@@selector(@dots{})} expression referring to an
1992 undeclared selector is found. A selector is considered undeclared if no
1993 method with that name has been declared before the
1994 @code{@@selector(@dots{})} expression, either explicitly in an
1995 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
1996 an @code{@@implementation} section. This option always performs its
1997 checks as soon as a @code{@@selector(@dots{})} expression is found,
1998 while @code{-Wselector} only performs its checks in the final stage of
1999 compilation. This also enforces the coding style convention
2000 that methods and selectors must be declared before being used.
2002 @item -print-objc-runtime-info
2003 @opindex print-objc-runtime-info
2004 Generate C header describing the largest structure that is passed by
2009 @node Language Independent Options
2010 @section Options to Control Diagnostic Messages Formatting
2011 @cindex options to control diagnostics formatting
2012 @cindex diagnostic messages
2013 @cindex message formatting
2015 Traditionally, diagnostic messages have been formatted irrespective of
2016 the output device's aspect (e.g.@: its width, @dots{}). The options described
2017 below can be used to control the diagnostic messages formatting
2018 algorithm, e.g.@: how many characters per line, how often source location
2019 information should be reported. Right now, only the C++ front end can
2020 honor these options. However it is expected, in the near future, that
2021 the remaining front ends would be able to digest them correctly.
2024 @item -fmessage-length=@var{n}
2025 @opindex fmessage-length
2026 Try to format error messages so that they fit on lines of about @var{n}
2027 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2028 the front ends supported by GCC@. If @var{n} is zero, then no
2029 line-wrapping will be done; each error message will appear on a single
2032 @opindex fdiagnostics-show-location
2033 @item -fdiagnostics-show-location=once
2034 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2035 reporter to emit @emph{once} source location information; that is, in
2036 case the message is too long to fit on a single physical line and has to
2037 be wrapped, the source location won't be emitted (as prefix) again,
2038 over and over, in subsequent continuation lines. This is the default
2041 @item -fdiagnostics-show-location=every-line
2042 Only meaningful in line-wrapping mode. Instructs the diagnostic
2043 messages reporter to emit the same source location information (as
2044 prefix) for physical lines that result from the process of breaking
2045 a message which is too long to fit on a single line.
2049 @node Warning Options
2050 @section Options to Request or Suppress Warnings
2051 @cindex options to control warnings
2052 @cindex warning messages
2053 @cindex messages, warning
2054 @cindex suppressing warnings
2056 Warnings are diagnostic messages that report constructions which
2057 are not inherently erroneous but which are risky or suggest there
2058 may have been an error.
2060 You can request many specific warnings with options beginning @samp{-W},
2061 for example @option{-Wimplicit} to request warnings on implicit
2062 declarations. Each of these specific warning options also has a
2063 negative form beginning @samp{-Wno-} to turn off warnings;
2064 for example, @option{-Wno-implicit}. This manual lists only one of the
2065 two forms, whichever is not the default.
2067 The following options control the amount and kinds of warnings produced
2068 by GCC; for further, language-specific options also refer to
2069 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2073 @cindex syntax checking
2075 @opindex fsyntax-only
2076 Check the code for syntax errors, but don't do anything beyond that.
2080 Issue all the warnings demanded by strict ISO C and ISO C++;
2081 reject all programs that use forbidden extensions, and some other
2082 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2083 version of the ISO C standard specified by any @option{-std} option used.
2085 Valid ISO C and ISO C++ programs should compile properly with or without
2086 this option (though a rare few will require @option{-ansi} or a
2087 @option{-std} option specifying the required version of ISO C)@. However,
2088 without this option, certain GNU extensions and traditional C and C++
2089 features are supported as well. With this option, they are rejected.
2091 @option{-pedantic} does not cause warning messages for use of the
2092 alternate keywords whose names begin and end with @samp{__}. Pedantic
2093 warnings are also disabled in the expression that follows
2094 @code{__extension__}. However, only system header files should use
2095 these escape routes; application programs should avoid them.
2096 @xref{Alternate Keywords}.
2098 Some users try to use @option{-pedantic} to check programs for strict ISO
2099 C conformance. They soon find that it does not do quite what they want:
2100 it finds some non-ISO practices, but not all---only those for which
2101 ISO C @emph{requires} a diagnostic, and some others for which
2102 diagnostics have been added.
2104 A feature to report any failure to conform to ISO C might be useful in
2105 some instances, but would require considerable additional work and would
2106 be quite different from @option{-pedantic}. We don't have plans to
2107 support such a feature in the near future.
2109 Where the standard specified with @option{-std} represents a GNU
2110 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2111 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2112 extended dialect is based. Warnings from @option{-pedantic} are given
2113 where they are required by the base standard. (It would not make sense
2114 for such warnings to be given only for features not in the specified GNU
2115 C dialect, since by definition the GNU dialects of C include all
2116 features the compiler supports with the given option, and there would be
2117 nothing to warn about.)
2119 @item -pedantic-errors
2120 @opindex pedantic-errors
2121 Like @option{-pedantic}, except that errors are produced rather than
2126 Inhibit all warning messages.
2130 Inhibit warning messages about the use of @samp{#import}.
2132 @item -Wchar-subscripts
2133 @opindex Wchar-subscripts
2134 Warn if an array subscript has type @code{char}. This is a common cause
2135 of error, as programmers often forget that this type is signed on some
2140 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2141 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2143 @item -Wfatal-errors
2144 @opindex Wfatal-errors
2145 This option causes the compiler to abort compilation on the first error
2146 occurred rather than trying to keep going and printing further error
2151 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2152 the arguments supplied have types appropriate to the format string
2153 specified, and that the conversions specified in the format string make
2154 sense. This includes standard functions, and others specified by format
2155 attributes (@pxref{Function Attributes}), in the @code{printf},
2156 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2157 not in the C standard) families (or other target-specific families).
2159 The formats are checked against the format features supported by GNU
2160 libc version 2.2. These include all ISO C90 and C99 features, as well
2161 as features from the Single Unix Specification and some BSD and GNU
2162 extensions. Other library implementations may not support all these
2163 features; GCC does not support warning about features that go beyond a
2164 particular library's limitations. However, if @option{-pedantic} is used
2165 with @option{-Wformat}, warnings will be given about format features not
2166 in the selected standard version (but not for @code{strfmon} formats,
2167 since those are not in any version of the C standard). @xref{C Dialect
2168 Options,,Options Controlling C Dialect}.
2170 Since @option{-Wformat} also checks for null format arguments for
2171 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2173 @option{-Wformat} is included in @option{-Wall}. For more control over some
2174 aspects of format checking, the options @option{-Wformat-y2k},
2175 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2176 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2177 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2180 @opindex Wformat-y2k
2181 If @option{-Wformat} is specified, also warn about @code{strftime}
2182 formats which may yield only a two-digit year.
2184 @item -Wno-format-extra-args
2185 @opindex Wno-format-extra-args
2186 If @option{-Wformat} is specified, do not warn about excess arguments to a
2187 @code{printf} or @code{scanf} format function. The C standard specifies
2188 that such arguments are ignored.
2190 Where the unused arguments lie between used arguments that are
2191 specified with @samp{$} operand number specifications, normally
2192 warnings are still given, since the implementation could not know what
2193 type to pass to @code{va_arg} to skip the unused arguments. However,
2194 in the case of @code{scanf} formats, this option will suppress the
2195 warning if the unused arguments are all pointers, since the Single
2196 Unix Specification says that such unused arguments are allowed.
2198 @item -Wno-format-zero-length
2199 @opindex Wno-format-zero-length
2200 If @option{-Wformat} is specified, do not warn about zero-length formats.
2201 The C standard specifies that zero-length formats are allowed.
2203 @item -Wformat-nonliteral
2204 @opindex Wformat-nonliteral
2205 If @option{-Wformat} is specified, also warn if the format string is not a
2206 string literal and so cannot be checked, unless the format function
2207 takes its format arguments as a @code{va_list}.
2209 @item -Wformat-security
2210 @opindex Wformat-security
2211 If @option{-Wformat} is specified, also warn about uses of format
2212 functions that represent possible security problems. At present, this
2213 warns about calls to @code{printf} and @code{scanf} functions where the
2214 format string is not a string literal and there are no format arguments,
2215 as in @code{printf (foo);}. This may be a security hole if the format
2216 string came from untrusted input and contains @samp{%n}. (This is
2217 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2218 in future warnings may be added to @option{-Wformat-security} that are not
2219 included in @option{-Wformat-nonliteral}.)
2223 Enable @option{-Wformat} plus format checks not included in
2224 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2225 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2229 Warn about passing a null pointer for arguments marked as
2230 requiring a non-null value by the @code{nonnull} function attribute.
2232 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2233 can be disabled with the @option{-Wno-nonnull} option.
2235 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2237 Warn about uninitialized variables which are initialized with themselves.
2238 Note this option can only be used with the @option{-Wuninitialized} option,
2239 which in turn only works with @option{-O1} and above.
2241 For example, GCC will warn about @code{i} being uninitialized in the
2242 following snippet only when @option{-Winit-self} has been specified:
2253 @item -Wimplicit-int
2254 @opindex Wimplicit-int
2255 Warn when a declaration does not specify a type.
2257 @item -Wimplicit-function-declaration
2258 @itemx -Werror-implicit-function-declaration
2259 @opindex Wimplicit-function-declaration
2260 @opindex Werror-implicit-function-declaration
2261 Give a warning (or error) whenever a function is used before being
2266 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2270 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2271 function with external linkage, returning int, taking either zero
2272 arguments, two, or three arguments of appropriate types.
2274 @item -Wmissing-braces
2275 @opindex Wmissing-braces
2276 Warn if an aggregate or union initializer is not fully bracketed. In
2277 the following example, the initializer for @samp{a} is not fully
2278 bracketed, but that for @samp{b} is fully bracketed.
2281 int a[2][2] = @{ 0, 1, 2, 3 @};
2282 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2285 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2286 @opindex Wmissing-include-dirs
2287 Warn if a user-supplied include directory does not exist.
2290 @opindex Wparentheses
2291 Warn if parentheses are omitted in certain contexts, such
2292 as when there is an assignment in a context where a truth value
2293 is expected, or when operators are nested whose precedence people
2294 often get confused about. Only the warning for an assignment used as
2295 a truth value is supported when compiling C++; the other warnings are
2296 only supported when compiling C@.
2298 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2299 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2300 interpretation from that of ordinary mathematical notation.
2302 Also warn about constructions where there may be confusion to which
2303 @code{if} statement an @code{else} branch belongs. Here is an example of
2318 In C, every @code{else} branch belongs to the innermost possible @code{if}
2319 statement, which in this example is @code{if (b)}. This is often not
2320 what the programmer expected, as illustrated in the above example by
2321 indentation the programmer chose. When there is the potential for this
2322 confusion, GCC will issue a warning when this flag is specified.
2323 To eliminate the warning, add explicit braces around the innermost
2324 @code{if} statement so there is no way the @code{else} could belong to
2325 the enclosing @code{if}. The resulting code would look like this:
2341 @item -Wsequence-point
2342 @opindex Wsequence-point
2343 Warn about code that may have undefined semantics because of violations
2344 of sequence point rules in the C standard.
2346 The C standard defines the order in which expressions in a C program are
2347 evaluated in terms of @dfn{sequence points}, which represent a partial
2348 ordering between the execution of parts of the program: those executed
2349 before the sequence point, and those executed after it. These occur
2350 after the evaluation of a full expression (one which is not part of a
2351 larger expression), after the evaluation of the first operand of a
2352 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2353 function is called (but after the evaluation of its arguments and the
2354 expression denoting the called function), and in certain other places.
2355 Other than as expressed by the sequence point rules, the order of
2356 evaluation of subexpressions of an expression is not specified. All
2357 these rules describe only a partial order rather than a total order,
2358 since, for example, if two functions are called within one expression
2359 with no sequence point between them, the order in which the functions
2360 are called is not specified. However, the standards committee have
2361 ruled that function calls do not overlap.
2363 It is not specified when between sequence points modifications to the
2364 values of objects take effect. Programs whose behavior depends on this
2365 have undefined behavior; the C standard specifies that ``Between the
2366 previous and next sequence point an object shall have its stored value
2367 modified at most once by the evaluation of an expression. Furthermore,
2368 the prior value shall be read only to determine the value to be
2369 stored.''. If a program breaks these rules, the results on any
2370 particular implementation are entirely unpredictable.
2372 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2373 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2374 diagnosed by this option, and it may give an occasional false positive
2375 result, but in general it has been found fairly effective at detecting
2376 this sort of problem in programs.
2378 The present implementation of this option only works for C programs. A
2379 future implementation may also work for C++ programs.
2381 The C standard is worded confusingly, therefore there is some debate
2382 over the precise meaning of the sequence point rules in subtle cases.
2383 Links to discussions of the problem, including proposed formal
2384 definitions, may be found on the GCC readings page, at
2385 @w{@uref{http://gcc.gnu.org/readings.html}}.
2388 @opindex Wreturn-type
2389 Warn whenever a function is defined with a return-type that defaults to
2390 @code{int}. Also warn about any @code{return} statement with no
2391 return-value in a function whose return-type is not @code{void}.
2393 For C, also warn if the return type of a function has a type qualifier
2394 such as @code{const}. Such a type qualifier has no effect, since the
2395 value returned by a function is not an lvalue. ISO C prohibits
2396 qualified @code{void} return types on function definitions, so such
2397 return types always receive a warning even without this option.
2399 For C++, a function without return type always produces a diagnostic
2400 message, even when @option{-Wno-return-type} is specified. The only
2401 exceptions are @samp{main} and functions defined in system headers.
2405 Warn whenever a @code{switch} statement has an index of enumerated type
2406 and lacks a @code{case} for one or more of the named codes of that
2407 enumeration. (The presence of a @code{default} label prevents this
2408 warning.) @code{case} labels outside the enumeration range also
2409 provoke warnings when this option is used.
2411 @item -Wswitch-default
2412 @opindex Wswitch-switch
2413 Warn whenever a @code{switch} statement does not have a @code{default}
2417 @opindex Wswitch-enum
2418 Warn whenever a @code{switch} statement has an index of enumerated type
2419 and lacks a @code{case} for one or more of the named codes of that
2420 enumeration. @code{case} labels outside the enumeration range also
2421 provoke warnings when this option is used.
2425 Warn if any trigraphs are encountered that might change the meaning of
2426 the program (trigraphs within comments are not warned about).
2428 @item -Wunused-function
2429 @opindex Wunused-function
2430 Warn whenever a static function is declared but not defined or a
2431 non\-inline static function is unused.
2433 @item -Wunused-label
2434 @opindex Wunused-label
2435 Warn whenever a label is declared but not used.
2437 To suppress this warning use the @samp{unused} attribute
2438 (@pxref{Variable Attributes}).
2440 @item -Wunused-parameter
2441 @opindex Wunused-parameter
2442 Warn whenever a function parameter is unused aside from its declaration.
2444 To suppress this warning use the @samp{unused} attribute
2445 (@pxref{Variable Attributes}).
2447 @item -Wunused-variable
2448 @opindex Wunused-variable
2449 Warn whenever a local variable or non-constant static variable is unused
2450 aside from its declaration
2452 To suppress this warning use the @samp{unused} attribute
2453 (@pxref{Variable Attributes}).
2455 @item -Wunused-value
2456 @opindex Wunused-value
2457 Warn whenever a statement computes a result that is explicitly not used.
2459 To suppress this warning cast the expression to @samp{void}.
2463 All the above @option{-Wunused} options combined.
2465 In order to get a warning about an unused function parameter, you must
2466 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2467 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2469 @item -Wuninitialized
2470 @opindex Wuninitialized
2471 Warn if an automatic variable is used without first being initialized or
2472 if a variable may be clobbered by a @code{setjmp} call.
2474 These warnings are possible only in optimizing compilation,
2475 because they require data flow information that is computed only
2476 when optimizing. If you don't specify @option{-O}, you simply won't
2479 If you want to warn about code which uses the uninitialized value of the
2480 variable in its own initializer, use the @option{-Winit-self} option.
2482 These warnings occur only for variables that are candidates for
2483 register allocation. Therefore, they do not occur for a variable that
2484 is declared @code{volatile}, or whose address is taken, or whose size
2485 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
2486 structures, unions or arrays, even when they are in registers.
2488 Note that there may be no warning about a variable that is used only
2489 to compute a value that itself is never used, because such
2490 computations may be deleted by data flow analysis before the warnings
2493 These warnings are made optional because GCC is not smart
2494 enough to see all the reasons why the code might be correct
2495 despite appearing to have an error. Here is one example of how
2516 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2517 always initialized, but GCC doesn't know this. Here is
2518 another common case:
2523 if (change_y) save_y = y, y = new_y;
2525 if (change_y) y = save_y;
2530 This has no bug because @code{save_y} is used only if it is set.
2532 @cindex @code{longjmp} warnings
2533 This option also warns when a non-volatile automatic variable might be
2534 changed by a call to @code{longjmp}. These warnings as well are possible
2535 only in optimizing compilation.
2537 The compiler sees only the calls to @code{setjmp}. It cannot know
2538 where @code{longjmp} will be called; in fact, a signal handler could
2539 call it at any point in the code. As a result, you may get a warning
2540 even when there is in fact no problem because @code{longjmp} cannot
2541 in fact be called at the place which would cause a problem.
2543 Some spurious warnings can be avoided if you declare all the functions
2544 you use that never return as @code{noreturn}. @xref{Function
2547 @item -Wunknown-pragmas
2548 @opindex Wunknown-pragmas
2549 @cindex warning for unknown pragmas
2550 @cindex unknown pragmas, warning
2551 @cindex pragmas, warning of unknown
2552 Warn when a #pragma directive is encountered which is not understood by
2553 GCC@. If this command line option is used, warnings will even be issued
2554 for unknown pragmas in system header files. This is not the case if
2555 the warnings were only enabled by the @option{-Wall} command line option.
2557 @item -Wstrict-aliasing
2558 @opindex Wstrict-aliasing
2559 This option is only active when @option{-fstrict-aliasing} is active.
2560 It warns about code which might break the strict aliasing rules that the
2561 compiler is using for optimization. The warning does not catch all
2562 cases, but does attempt to catch the more common pitfalls. It is
2563 included in @option{-Wall}.
2565 @item -Wstrict-aliasing=2
2566 @opindex Wstrict-aliasing=2
2567 This option is only active when @option{-fstrict-aliasing} is active.
2568 It warns about all code which might break the strict aliasing rules that the
2569 compiler is using for optimization. This warning catches all cases, but
2570 it will also give a warning for some ambiguous cases that are safe.
2574 All of the above @samp{-W} options combined. This enables all the
2575 warnings about constructions that some users consider questionable, and
2576 that are easy to avoid (or modify to prevent the warning), even in
2577 conjunction with macros. This also enables some language-specific
2578 warnings described in @ref{C++ Dialect Options} and
2579 @ref{Objective-C and Objective-C++ Dialect Options}.
2582 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2583 Some of them warn about constructions that users generally do not
2584 consider questionable, but which occasionally you might wish to check
2585 for; others warn about constructions that are necessary or hard to avoid
2586 in some cases, and there is no simple way to modify the code to suppress
2593 (This option used to be called @option{-W}. The older name is still
2594 supported, but the newer name is more descriptive.) Print extra warning
2595 messages for these events:
2599 A function can return either with or without a value. (Falling
2600 off the end of the function body is considered returning without
2601 a value.) For example, this function would evoke such a
2615 An expression-statement or the left-hand side of a comma expression
2616 contains no side effects.
2617 To suppress the warning, cast the unused expression to void.
2618 For example, an expression such as @samp{x[i,j]} will cause a warning,
2619 but @samp{x[(void)i,j]} will not.
2622 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2625 Storage-class specifiers like @code{static} are not the first things in
2626 a declaration. According to the C Standard, this usage is obsolescent.
2629 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2633 A comparison between signed and unsigned values could produce an
2634 incorrect result when the signed value is converted to unsigned.
2635 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2638 An aggregate has an initializer which does not initialize all members.
2639 This warning can be independently controlled by
2640 @option{-Wmissing-field-initializers}.
2643 A function parameter is declared without a type specifier in K&R-style
2651 An empty body occurs in an @samp{if} or @samp{else} statement.
2654 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2655 @samp{>}, or @samp{>=}.
2658 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2661 Any of several floating-point events that often indicate errors, such as
2662 overflow, underflow, loss of precision, etc.
2664 @item @r{(C++ only)}
2665 An enumerator and a non-enumerator both appear in a conditional expression.
2667 @item @r{(C++ only)}
2668 A non-static reference or non-static @samp{const} member appears in a
2669 class without constructors.
2671 @item @r{(C++ only)}
2672 Ambiguous virtual bases.
2674 @item @r{(C++ only)}
2675 Subscripting an array which has been declared @samp{register}.
2677 @item @r{(C++ only)}
2678 Taking the address of a variable which has been declared @samp{register}.
2680 @item @r{(C++ only)}
2681 A base class is not initialized in a derived class' copy constructor.
2684 @item -Wno-div-by-zero
2685 @opindex Wno-div-by-zero
2686 @opindex Wdiv-by-zero
2687 Do not warn about compile-time integer division by zero. Floating point
2688 division by zero is not warned about, as it can be a legitimate way of
2689 obtaining infinities and NaNs.
2691 @item -Wsystem-headers
2692 @opindex Wsystem-headers
2693 @cindex warnings from system headers
2694 @cindex system headers, warnings from
2695 Print warning messages for constructs found in system header files.
2696 Warnings from system headers are normally suppressed, on the assumption
2697 that they usually do not indicate real problems and would only make the
2698 compiler output harder to read. Using this command line option tells
2699 GCC to emit warnings from system headers as if they occurred in user
2700 code. However, note that using @option{-Wall} in conjunction with this
2701 option will @emph{not} warn about unknown pragmas in system
2702 headers---for that, @option{-Wunknown-pragmas} must also be used.
2705 @opindex Wfloat-equal
2706 Warn if floating point values are used in equality comparisons.
2708 The idea behind this is that sometimes it is convenient (for the
2709 programmer) to consider floating-point values as approximations to
2710 infinitely precise real numbers. If you are doing this, then you need
2711 to compute (by analyzing the code, or in some other way) the maximum or
2712 likely maximum error that the computation introduces, and allow for it
2713 when performing comparisons (and when producing output, but that's a
2714 different problem). In particular, instead of testing for equality, you
2715 would check to see whether the two values have ranges that overlap; and
2716 this is done with the relational operators, so equality comparisons are
2719 @item -Wtraditional @r{(C only)}
2720 @opindex Wtraditional
2721 Warn about certain constructs that behave differently in traditional and
2722 ISO C@. Also warn about ISO C constructs that have no traditional C
2723 equivalent, and/or problematic constructs which should be avoided.
2727 Macro parameters that appear within string literals in the macro body.
2728 In traditional C macro replacement takes place within string literals,
2729 but does not in ISO C@.
2732 In traditional C, some preprocessor directives did not exist.
2733 Traditional preprocessors would only consider a line to be a directive
2734 if the @samp{#} appeared in column 1 on the line. Therefore
2735 @option{-Wtraditional} warns about directives that traditional C
2736 understands but would ignore because the @samp{#} does not appear as the
2737 first character on the line. It also suggests you hide directives like
2738 @samp{#pragma} not understood by traditional C by indenting them. Some
2739 traditional implementations would not recognize @samp{#elif}, so it
2740 suggests avoiding it altogether.
2743 A function-like macro that appears without arguments.
2746 The unary plus operator.
2749 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2750 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2751 constants.) Note, these suffixes appear in macros defined in the system
2752 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2753 Use of these macros in user code might normally lead to spurious
2754 warnings, however GCC's integrated preprocessor has enough context to
2755 avoid warning in these cases.
2758 A function declared external in one block and then used after the end of
2762 A @code{switch} statement has an operand of type @code{long}.
2765 A non-@code{static} function declaration follows a @code{static} one.
2766 This construct is not accepted by some traditional C compilers.
2769 The ISO type of an integer constant has a different width or
2770 signedness from its traditional type. This warning is only issued if
2771 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2772 typically represent bit patterns, are not warned about.
2775 Usage of ISO string concatenation is detected.
2778 Initialization of automatic aggregates.
2781 Identifier conflicts with labels. Traditional C lacks a separate
2782 namespace for labels.
2785 Initialization of unions. If the initializer is zero, the warning is
2786 omitted. This is done under the assumption that the zero initializer in
2787 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2788 initializer warnings and relies on default initialization to zero in the
2792 Conversions by prototypes between fixed/floating point values and vice
2793 versa. The absence of these prototypes when compiling with traditional
2794 C would cause serious problems. This is a subset of the possible
2795 conversion warnings, for the full set use @option{-Wconversion}.
2798 Use of ISO C style function definitions. This warning intentionally is
2799 @emph{not} issued for prototype declarations or variadic functions
2800 because these ISO C features will appear in your code when using
2801 libiberty's traditional C compatibility macros, @code{PARAMS} and
2802 @code{VPARAMS}. This warning is also bypassed for nested functions
2803 because that feature is already a GCC extension and thus not relevant to
2804 traditional C compatibility.
2807 @item -Wdeclaration-after-statement @r{(C only)}
2808 @opindex Wdeclaration-after-statement
2809 Warn when a declaration is found after a statement in a block. This
2810 construct, known from C++, was introduced with ISO C99 and is by default
2811 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2812 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2816 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2818 @item -Wendif-labels
2819 @opindex Wendif-labels
2820 Warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2824 Warn whenever a local variable shadows another local variable, parameter or
2825 global variable or whenever a built-in function is shadowed.
2827 @item -Wlarger-than-@var{len}
2828 @opindex Wlarger-than
2829 Warn whenever an object of larger than @var{len} bytes is defined.
2831 @item -Wpointer-arith
2832 @opindex Wpointer-arith
2833 Warn about anything that depends on the ``size of'' a function type or
2834 of @code{void}. GNU C assigns these types a size of 1, for
2835 convenience in calculations with @code{void *} pointers and pointers
2838 @item -Wbad-function-cast @r{(C only)}
2839 @opindex Wbad-function-cast
2840 Warn whenever a function call is cast to a non-matching type.
2841 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2845 Warn whenever a pointer is cast so as to remove a type qualifier from
2846 the target type. For example, warn if a @code{const char *} is cast
2847 to an ordinary @code{char *}.
2850 @opindex Wcast-align
2851 Warn whenever a pointer is cast such that the required alignment of the
2852 target is increased. For example, warn if a @code{char *} is cast to
2853 an @code{int *} on machines where integers can only be accessed at
2854 two- or four-byte boundaries.
2856 @item -Wwrite-strings
2857 @opindex Wwrite-strings
2858 When compiling C, give string constants the type @code{const
2859 char[@var{length}]} so that
2860 copying the address of one into a non-@code{const} @code{char *}
2861 pointer will get a warning; when compiling C++, warn about the
2862 deprecated conversion from string constants to @code{char *}.
2863 These warnings will help you find at
2864 compile time code that can try to write into a string constant, but
2865 only if you have been very careful about using @code{const} in
2866 declarations and prototypes. Otherwise, it will just be a nuisance;
2867 this is why we did not make @option{-Wall} request these warnings.
2870 @opindex Wconversion
2871 Warn if a prototype causes a type conversion that is different from what
2872 would happen to the same argument in the absence of a prototype. This
2873 includes conversions of fixed point to floating and vice versa, and
2874 conversions changing the width or signedness of a fixed point argument
2875 except when the same as the default promotion.
2877 Also, warn if a negative integer constant expression is implicitly
2878 converted to an unsigned type. For example, warn about the assignment
2879 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2880 casts like @code{(unsigned) -1}.
2882 @item -Wsign-compare
2883 @opindex Wsign-compare
2884 @cindex warning for comparison of signed and unsigned values
2885 @cindex comparison of signed and unsigned values, warning
2886 @cindex signed and unsigned values, comparison warning
2887 Warn when a comparison between signed and unsigned values could produce
2888 an incorrect result when the signed value is converted to unsigned.
2889 This warning is also enabled by @option{-Wextra}; to get the other warnings
2890 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2892 @item -Waggregate-return
2893 @opindex Waggregate-return
2894 Warn if any functions that return structures or unions are defined or
2895 called. (In languages where you can return an array, this also elicits
2898 @item -Wstrict-prototypes @r{(C only)}
2899 @opindex Wstrict-prototypes
2900 Warn if a function is declared or defined without specifying the
2901 argument types. (An old-style function definition is permitted without
2902 a warning if preceded by a declaration which specifies the argument
2905 @item -Wold-style-definition @r{(C only)}
2906 @opindex Wold-style-definition
2907 Warn if an old-style function definition is used. A warning is given
2908 even if there is a previous prototype.
2910 @item -Wmissing-prototypes @r{(C only)}
2911 @opindex Wmissing-prototypes
2912 Warn if a global function is defined without a previous prototype
2913 declaration. This warning is issued even if the definition itself
2914 provides a prototype. The aim is to detect global functions that fail
2915 to be declared in header files.
2917 @item -Wmissing-declarations @r{(C only)}
2918 @opindex Wmissing-declarations
2919 Warn if a global function is defined without a previous declaration.
2920 Do so even if the definition itself provides a prototype.
2921 Use this option to detect global functions that are not declared in
2924 @item -Wmissing-field-initializers
2925 @opindex Wmissing-field-initializers
2928 Warn if a structure's initializer has some fields missing. For
2929 example, the following code would cause such a warning, because
2930 @code{x.h} is implicitly zero:
2933 struct s @{ int f, g, h; @};
2934 struct s x = @{ 3, 4 @};
2937 This option does not warn about designated initializers, so the following
2938 modification would not trigger a warning:
2941 struct s @{ int f, g, h; @};
2942 struct s x = @{ .f = 3, .g = 4 @};
2945 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
2946 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
2948 @item -Wmissing-noreturn
2949 @opindex Wmissing-noreturn
2950 Warn about functions which might be candidates for attribute @code{noreturn}.
2951 Note these are only possible candidates, not absolute ones. Care should
2952 be taken to manually verify functions actually do not ever return before
2953 adding the @code{noreturn} attribute, otherwise subtle code generation
2954 bugs could be introduced. You will not get a warning for @code{main} in
2955 hosted C environments.
2957 @item -Wmissing-format-attribute
2958 @opindex Wmissing-format-attribute
2960 If @option{-Wformat} is enabled, also warn about functions which might be
2961 candidates for @code{format} attributes. Note these are only possible
2962 candidates, not absolute ones. GCC will guess that @code{format}
2963 attributes might be appropriate for any function that calls a function
2964 like @code{vprintf} or @code{vscanf}, but this might not always be the
2965 case, and some functions for which @code{format} attributes are
2966 appropriate may not be detected. This option has no effect unless
2967 @option{-Wformat} is enabled (possibly by @option{-Wall}).
2969 @item -Wno-multichar
2970 @opindex Wno-multichar
2972 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
2973 Usually they indicate a typo in the user's code, as they have
2974 implementation-defined values, and should not be used in portable code.
2976 @item -Wno-deprecated-declarations
2977 @opindex Wno-deprecated-declarations
2978 Do not warn about uses of functions, variables, and types marked as
2979 deprecated by using the @code{deprecated} attribute.
2980 (@pxref{Function Attributes}, @pxref{Variable Attributes},
2981 @pxref{Type Attributes}.)
2985 Warn if a structure is given the packed attribute, but the packed
2986 attribute has no effect on the layout or size of the structure.
2987 Such structures may be mis-aligned for little benefit. For
2988 instance, in this code, the variable @code{f.x} in @code{struct bar}
2989 will be misaligned even though @code{struct bar} does not itself
2990 have the packed attribute:
2997 @} __attribute__((packed));
3007 Warn if padding is included in a structure, either to align an element
3008 of the structure or to align the whole structure. Sometimes when this
3009 happens it is possible to rearrange the fields of the structure to
3010 reduce the padding and so make the structure smaller.
3012 @item -Wredundant-decls
3013 @opindex Wredundant-decls
3014 Warn if anything is declared more than once in the same scope, even in
3015 cases where multiple declaration is valid and changes nothing.
3017 @item -Wnested-externs @r{(C only)}
3018 @opindex Wnested-externs
3019 Warn if an @code{extern} declaration is encountered within a function.
3021 @item -Wunreachable-code
3022 @opindex Wunreachable-code
3023 Warn if the compiler detects that code will never be executed.
3025 This option is intended to warn when the compiler detects that at
3026 least a whole line of source code will never be executed, because
3027 some condition is never satisfied or because it is after a
3028 procedure that never returns.
3030 It is possible for this option to produce a warning even though there
3031 are circumstances under which part of the affected line can be executed,
3032 so care should be taken when removing apparently-unreachable code.
3034 For instance, when a function is inlined, a warning may mean that the
3035 line is unreachable in only one inlined copy of the function.
3037 This option is not made part of @option{-Wall} because in a debugging
3038 version of a program there is often substantial code which checks
3039 correct functioning of the program and is, hopefully, unreachable
3040 because the program does work. Another common use of unreachable
3041 code is to provide behavior which is selectable at compile-time.
3045 Warn if a function can not be inlined and it was declared as inline.
3046 Even with this option, the compiler will not warn about failures to
3047 inline functions declared in system headers.
3049 The compiler uses a variety of heuristics to determine whether or not
3050 to inline a function. For example, the compiler takes into account
3051 the size of the function being inlined and the the amount of inlining
3052 that has already been done in the current function. Therefore,
3053 seemingly insignificant changes in the source program can cause the
3054 warnings produced by @option{-Winline} to appear or disappear.
3056 @item -Wno-invalid-offsetof @r{(C++ only)}
3057 @opindex Wno-invalid-offsetof
3058 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3059 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3060 to a non-POD type is undefined. In existing C++ implementations,
3061 however, @samp{offsetof} typically gives meaningful results even when
3062 applied to certain kinds of non-POD types. (Such as a simple
3063 @samp{struct} that fails to be a POD type only by virtue of having a
3064 constructor.) This flag is for users who are aware that they are
3065 writing nonportable code and who have deliberately chosen to ignore the
3068 The restrictions on @samp{offsetof} may be relaxed in a future version
3069 of the C++ standard.
3072 @opindex Winvalid-pch
3073 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3074 the search path but can't be used.
3078 @opindex Wno-long-long
3079 Warn if @samp{long long} type is used. This is default. To inhibit
3080 the warning messages, use @option{-Wno-long-long}. Flags
3081 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3082 only when @option{-pedantic} flag is used.
3084 @item -Wvariadic-macros
3085 @opindex Wvariadic-macros
3086 @opindex Wno-variadic-macros
3087 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3088 alternate syntax when in pedantic ISO C99 mode. This is default.
3089 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3091 @item -Wdisabled-optimization
3092 @opindex Wdisabled-optimization
3093 Warn if a requested optimization pass is disabled. This warning does
3094 not generally indicate that there is anything wrong with your code; it
3095 merely indicates that GCC's optimizers were unable to handle the code
3096 effectively. Often, the problem is that your code is too big or too
3097 complex; GCC will refuse to optimize programs when the optimization
3098 itself is likely to take inordinate amounts of time.
3102 Make all warnings into errors.
3105 @node Debugging Options
3106 @section Options for Debugging Your Program or GCC
3107 @cindex options, debugging
3108 @cindex debugging information options
3110 GCC has various special options that are used for debugging
3111 either your program or GCC:
3116 Produce debugging information in the operating system's native format
3117 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
3120 On most systems that use stabs format, @option{-g} enables use of extra
3121 debugging information that only GDB can use; this extra information
3122 makes debugging work better in GDB but will probably make other debuggers
3124 refuse to read the program. If you want to control for certain whether
3125 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3126 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3128 Unlike most other C compilers, GCC allows you to use @option{-g} with
3129 @option{-O}. The shortcuts taken by optimized code may occasionally
3130 produce surprising results: some variables you declared may not exist
3131 at all; flow of control may briefly move where you did not expect it;
3132 some statements may not be executed because they compute constant
3133 results or their values were already at hand; some statements may
3134 execute in different places because they were moved out of loops.
3136 Nevertheless it proves possible to debug optimized output. This makes
3137 it reasonable to use the optimizer for programs that might have bugs.
3139 The following options are useful when GCC is generated with the
3140 capability for more than one debugging format.
3144 Produce debugging information for use by GDB@. This means to use the
3145 most expressive format available (DWARF 2, stabs, or the native format
3146 if neither of those are supported), including GDB extensions if at all
3151 Produce debugging information in stabs format (if that is supported),
3152 without GDB extensions. This is the format used by DBX on most BSD
3153 systems. On MIPS, Alpha and System V Release 4 systems this option
3154 produces stabs debugging output which is not understood by DBX or SDB@.
3155 On System V Release 4 systems this option requires the GNU assembler.
3157 @item -feliminate-unused-debug-symbols
3158 @opindex feliminate-unused-debug-symbols
3159 Produce debugging information in stabs format (if that is supported),
3160 for only symbols that are actually used.
3164 Produce debugging information in stabs format (if that is supported),
3165 using GNU extensions understood only by the GNU debugger (GDB)@. The
3166 use of these extensions is likely to make other debuggers crash or
3167 refuse to read the program.
3171 Produce debugging information in COFF format (if that is supported).
3172 This is the format used by SDB on most System V systems prior to
3177 Produce debugging information in XCOFF format (if that is supported).
3178 This is the format used by the DBX debugger on IBM RS/6000 systems.
3182 Produce debugging information in XCOFF format (if that is supported),
3183 using GNU extensions understood only by the GNU debugger (GDB)@. The
3184 use of these extensions is likely to make other debuggers crash or
3185 refuse to read the program, and may cause assemblers other than the GNU
3186 assembler (GAS) to fail with an error.
3190 Produce debugging information in DWARF version 2 format (if that is
3191 supported). This is the format used by DBX on IRIX 6.
3195 Produce debugging information in VMS debug format (if that is
3196 supported). This is the format used by DEBUG on VMS systems.
3199 @itemx -ggdb@var{level}
3200 @itemx -gstabs@var{level}
3201 @itemx -gcoff@var{level}
3202 @itemx -gxcoff@var{level}
3203 @itemx -gvms@var{level}
3204 Request debugging information and also use @var{level} to specify how
3205 much information. The default level is 2.
3207 Level 1 produces minimal information, enough for making backtraces in
3208 parts of the program that you don't plan to debug. This includes
3209 descriptions of functions and external variables, but no information
3210 about local variables and no line numbers.
3212 Level 3 includes extra information, such as all the macro definitions
3213 present in the program. Some debuggers support macro expansion when
3214 you use @option{-g3}.
3216 Note that in order to avoid confusion between DWARF1 debug level 2,
3217 and DWARF2 @option{-gdwarf-2} does not accept a concatenated debug
3218 level. Instead use an additional @option{-g@var{level}} option to
3219 change the debug level for DWARF2.
3221 @item -feliminate-dwarf2-dups
3222 @opindex feliminate-dwarf2-dups
3223 Compress DWARF2 debugging information by eliminating duplicated
3224 information about each symbol. This option only makes sense when
3225 generating DWARF2 debugging information with @option{-gdwarf-2}.
3227 @cindex @command{prof}
3230 Generate extra code to write profile information suitable for the
3231 analysis program @command{prof}. You must use this option when compiling
3232 the source files you want data about, and you must also use it when
3235 @cindex @command{gprof}
3238 Generate extra code to write profile information suitable for the
3239 analysis program @command{gprof}. You must use this option when compiling
3240 the source files you want data about, and you must also use it when
3245 Makes the compiler print out each function name as it is compiled, and
3246 print some statistics about each pass when it finishes.
3249 @opindex ftime-report
3250 Makes the compiler print some statistics about the time consumed by each
3251 pass when it finishes.
3254 @opindex fmem-report
3255 Makes the compiler print some statistics about permanent memory
3256 allocation when it finishes.
3258 @item -fprofile-arcs
3259 @opindex fprofile-arcs
3260 Add code so that program flow @dfn{arcs} are instrumented. During
3261 execution the program records how many times each branch and call is
3262 executed and how many times it is taken or returns. When the compiled
3263 program exits it saves this data to a file called
3264 @file{@var{auxname}.gcda} for each source file. The data may be used for
3265 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3266 test coverage analysis (@option{-ftest-coverage}). Each object file's
3267 @var{auxname} is generated from the name of the output file, if
3268 explicitly specified and it is not the final executable, otherwise it is
3269 the basename of the source file. In both cases any suffix is removed
3270 (e.g. @file{foo.gcda} for input file @file{dir/foo.c}, or
3271 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3276 Compile the source files with @option{-fprofile-arcs} plus optimization
3277 and code generation options. For test coverage analysis, use the
3278 additional @option{-ftest-coverage} option. You do not need to profile
3279 every source file in a program.
3282 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3283 (the latter implies the former).
3286 Run the program on a representative workload to generate the arc profile
3287 information. This may be repeated any number of times. You can run
3288 concurrent instances of your program, and provided that the file system
3289 supports locking, the data files will be correctly updated. Also
3290 @code{fork} calls are detected and correctly handled (double counting
3294 For profile-directed optimizations, compile the source files again with
3295 the same optimization and code generation options plus
3296 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3297 Control Optimization}).
3300 For test coverage analysis, use @command{gcov} to produce human readable
3301 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3302 @command{gcov} documentation for further information.
3306 With @option{-fprofile-arcs}, for each function of your program GCC
3307 creates a program flow graph, then finds a spanning tree for the graph.
3308 Only arcs that are not on the spanning tree have to be instrumented: the
3309 compiler adds code to count the number of times that these arcs are
3310 executed. When an arc is the only exit or only entrance to a block, the
3311 instrumentation code can be added to the block; otherwise, a new basic
3312 block must be created to hold the instrumentation code.
3314 @item -ftree-based-profiling
3315 @opindex ftree-based-profiling
3316 This option is used in addition to @option{-fprofile-arcs} or
3317 @option{-fbranch-probabilities} to control whether those optimizations
3318 are performed on a tree-based or rtl-based internal representation.
3319 If you use this option when compiling with @option{-fprofile-arcs},
3320 you must also use it when compiling later with @option{-fbranch-probabilities}.
3321 Currently the tree-based optimization is in an early stage of
3322 development, and this option is recommended only for those people
3323 working on improving it.
3326 @item -ftest-coverage
3327 @opindex ftest-coverage
3328 Produce a notes file that the @command{gcov} code-coverage utility
3329 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3330 show program coverage. Each source file's note file is called
3331 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3332 above for a description of @var{auxname} and instructions on how to
3333 generate test coverage data. Coverage data will match the source files
3334 more closely, if you do not optimize.
3336 @item -d@var{letters}
3338 Says to make debugging dumps during compilation at times specified by
3339 @var{letters}. This is used for debugging the compiler. The file names
3340 for most of the dumps are made by appending a pass number and a word to
3341 the @var{dumpname}. @var{dumpname} is generated from the name of the
3342 output file, if explicitly specified and it is not an executable,
3343 otherwise it is the basename of the source file. In both cases any
3344 suffix is removed (e.g. @file{foo.01.rtl} or @file{foo.02.sibling}).
3345 Here are the possible letters for use in @var{letters}, and their
3351 Annotate the assembler output with miscellaneous debugging information.
3354 Dump after computing branch probabilities, to @file{@var{file}.12.bp}.
3357 Dump after block reordering, to @file{@var{file}.32.bbro}.
3360 Dump after instruction combination, to the file @file{@var{file}.20.combine}.
3363 Dump after the first if conversion, to the file @file{@var{file}.14.ce1}.
3364 Also dump after the second if conversion, to the file @file{@var{file}.21.ce2}.
3367 Dump after branch target load optimization, to to @file{@var{file}.33.btl}.
3368 Also dump after delayed branch scheduling, to @file{@var{file}.37.dbr}.
3371 Dump all macro definitions, at the end of preprocessing, in addition to
3375 Dump after the third if conversion, to @file{@var{file}.31.ce3}.
3378 Dump after control and data flow analysis, to @file{@var{file}.11.cfg}.
3379 Also dump after life analysis, to @file{@var{file}.19.life}.
3382 Dump after global register allocation, to @file{@var{file}.26.greg}.
3385 Dump after GCSE, to @file{@var{file}.08.gcse}.
3386 Also dump after jump bypassing and control flow optimizations, to
3387 @file{@var{file}.10.bypass}.
3390 Dump after finalization of EH handling code, to @file{@var{file}.03.eh}.
3393 Dump after sibling call optimizations, to @file{@var{file}.02.sibling}.
3396 Dump after the first jump optimization, to @file{@var{file}.04.jump}.
3399 Dump after conversion from registers to stack, to @file{@var{file}.35.stack}.
3402 Dump after local register allocation, to @file{@var{file}.25.lreg}.
3405 Dump after loop optimization passes, to @file{@var{file}.09.loop} and
3406 @file{@var{file}.16.loop2}.
3409 Dump after modulo scheduling, to @file{@var{file}.23.sms}.
3412 Dump after performing the machine dependent reorganization pass, to
3413 @file{@var{file}.36.mach}.
3416 Dump after register renumbering, to @file{@var{file}.30.rnreg}.
3419 Dump after the register move pass, to @file{@var{file}.22.regmove}.
3422 Dump after post-reload optimizations, to @file{@var{file}.27.postreload}.
3425 Dump after RTL generation, to @file{@var{file}.01.rtl}.
3428 Dump after the second scheduling pass, to @file{@var{file}.34.sched2}.
3431 Dump after CSE (including the jump optimization that sometimes follows
3432 CSE), to @file{@var{file}.06.cse}.
3435 Dump after the first scheduling pass, to @file{@var{file}.24.sched}.
3438 Dump after the second CSE pass (including the jump optimization that
3439 sometimes follows CSE), to @file{@var{file}.18.cse2}.
3442 Dump after running tracer, to @file{@var{file}.15.tracer}.
3445 Dump after null pointer elimination pass to @file{@var{file}.05.null}.
3448 Dump callgraph and unit-at-a-time optimization @file{@var{file}.00.unit}.
3451 Dump after the value profile transformations, to @file{@var{file}.13.vpt}.
3452 Also dump after variable tracking, to @file{@var{file}.35.vartrack}.
3455 Dump after the second flow pass, to @file{@var{file}.28.flow2}.
3458 Dump after the peephole pass, to @file{@var{file}.29.peephole2}.
3461 Dump after constructing the web, to @file{@var{file}.17.web}.
3464 Produce all the dumps listed above.
3467 Produce a core dump whenever an error occurs.
3470 Print statistics on memory usage, at the end of the run, to
3474 Annotate the assembler output with a comment indicating which
3475 pattern and alternative was used. The length of each instruction is
3479 Dump the RTL in the assembler output as a comment before each instruction.
3480 Also turns on @option{-dp} annotation.
3483 For each of the other indicated dump files (except for
3484 @file{@var{file}.01.rtl}), dump a representation of the control flow graph
3485 suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3488 Just generate RTL for a function instead of compiling it. Usually used
3492 Dump debugging information during parsing, to standard error.
3495 @item -fdump-unnumbered
3496 @opindex fdump-unnumbered
3497 When doing debugging dumps (see @option{-d} option above), suppress instruction
3498 numbers and line number note output. This makes it more feasible to
3499 use diff on debugging dumps for compiler invocations with different
3500 options, in particular with and without @option{-g}.
3502 @item -fdump-translation-unit @r{(C and C++ only)}
3503 @itemx -fdump-translation-unit-@var{options} @r{(C and C++ only)}
3504 @opindex fdump-translation-unit
3505 Dump a representation of the tree structure for the entire translation
3506 unit to a file. The file name is made by appending @file{.tu} to the
3507 source file name. If the @samp{-@var{options}} form is used, @var{options}
3508 controls the details of the dump as described for the
3509 @option{-fdump-tree} options.
3511 @item -fdump-class-hierarchy @r{(C++ only)}
3512 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3513 @opindex fdump-class-hierarchy
3514 Dump a representation of each class's hierarchy and virtual function
3515 table layout to a file. The file name is made by appending @file{.class}
3516 to the source file name. If the @samp{-@var{options}} form is used,
3517 @var{options} controls the details of the dump as described for the
3518 @option{-fdump-tree} options.
3520 @item -fdump-tree-@var{switch} @r{(C and C++ only)}
3521 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C and C++ only)}
3523 Control the dumping at various stages of processing the intermediate
3524 language tree to a file. The file name is generated by appending a switch
3525 specific suffix to the source file name. If the @samp{-@var{options}}
3526 form is used, @var{options} is a list of @samp{-} separated options that
3527 control the details of the dump. Not all options are applicable to all
3528 dumps, those which are not meaningful will be ignored. The following
3529 options are available
3533 Print the address of each node. Usually this is not meaningful as it
3534 changes according to the environment and source file. Its primary use
3535 is for tying up a dump file with a debug environment.
3537 Inhibit dumping of members of a scope or body of a function merely
3538 because that scope has been reached. Only dump such items when they
3539 are directly reachable by some other path. When dumping pretty-printed
3540 trees, this option inhibits dumping the bodies of control structures.
3542 Print a raw representation of the tree. By default, trees are
3543 pretty-printed into a C-like representation.
3545 Enable more detailed dumps (not honored by every dump option).
3547 Enable dumping various statistics about the pass (not honored by every dump
3550 Enable showing basic block boundaries (disabled in raw dumps).
3552 Enable showing virtual operands for every statement.
3554 Enable showing line numbers for statements.
3556 Enable showing the unique ID (@code{DECL_UID}) for each variable.
3558 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
3561 The following tree dumps are possible:
3565 Dump before any tree based optimization, to @file{@var{file}.original}.
3568 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3571 Dump after function inlining, to @file{@var{file}.inlined}.
3574 @opindex fdump-tree-gimple
3575 Dump each function before and after the gimplification pass to a file. The
3576 file name is made by appending @file{.gimple} to the source file name.
3579 @opindex fdump-tree-cfg
3580 Dump the control flow graph of each function to a file. The file name is
3581 made by appending @file{.cfg} to the source file name.
3584 @opindex fdump-tree-vcg
3585 Dump the control flow graph of each function to a file in VCG format. The
3586 file name is made by appending @file{.vcg} to the source file name. Note
3587 that if the file contains more than one function, the generated file cannot
3588 be used directly by VCG. You will need to cut and paste each function's
3589 graph into its own separate file first.
3592 @opindex fdump-tree-ch
3593 Dump each function after copying loop headers. The file name is made by
3594 appending @file{.ch} to the source file name.
3597 @opindex fdump-tree-ssa
3598 Dump SSA related information to a file. The file name is made by appending
3599 @file{.ssa} to the source file name.
3602 @opindex fdump-tree-alias
3603 Dump aliasing information for each function. The file name is made by
3604 appending @file{.alias} to the source file name.
3607 @opindex fdump-tree-ccp
3608 Dump each function after CCP. The file name is made by appending
3609 @file{.ccp} to the source file name.
3612 @opindex fdump-tree-pre
3613 Dump trees after partial redundancy elimination. The file name is made
3614 by appending @file{.pre} to the source file name.
3617 @opindex fdump-tree-fre
3618 Dump trees after full redundancy elimination. The file name is made
3619 by appending @file{.fre} to the source file name.
3622 @opindex fdump-tree-dce
3623 Dump each function after dead code elimination. The file name is made by
3624 appending @file{.dce} to the source file name.
3627 @opindex fdump-tree-mudflap
3628 Dump each function after adding mudflap instrumentation. The file name is
3629 made by appending @file{.mudflap} to the source file name.
3632 @opindex fdump-tree-sra
3633 Dump each function after performing scalar replacement of aggregates. The
3634 file name is made by appending @file{.sra} to the source file name.
3637 @opindex fdump-tree-dom
3638 Dump each function after applying dominator tree optimizations. The file
3639 name is made by appending @file{.dom} to the source file name.
3642 @opindex fdump-tree-dse
3643 Dump each function after applying dead store elimination. The file
3644 name is made by appending @file{.dse} to the source file name.
3647 @opindex fdump-tree-phiopt
3648 Dump each function after optimizing PHI nodes into straightline code. The file
3649 name is made by appending @file{.phiopt} to the source file name.
3652 @opindex fdump-tree-forwprop
3653 Dump each function after forward propagating single use variables. The file
3654 name is made by appending @file{.forwprop} to the source file name.
3657 @opindex fdump-tree-copyrename
3658 Dump each function after applying the copy rename optimization. The file
3659 name is made by appending @file{.copyrename} to the source file name.
3662 @opindex fdump-tree-nrv
3663 Dump each function after applying the named return value optimization on
3664 generic trees. The file name is made by appending @file{.nrv} to the source
3668 @opindex fdump-tree-vect
3669 Dump each function after applying vectorization of loops. The file name is
3670 made by appending @file{.vect} to the source file name.
3673 @opindex fdump-tree-all
3674 Enable all the available tree dumps with the flags provided in this option.
3677 @item -frandom-seed=@var{string}
3678 @opindex frandom-string
3679 This option provides a seed that GCC uses when it would otherwise use
3680 random numbers. It is used to generate certain symbol names
3681 that have to be different in every compiled file. It is also used to
3682 place unique stamps in coverage data files and the object files that
3683 produce them. You can use the @option{-frandom-seed} option to produce
3684 reproducibly identical object files.
3686 The @var{string} should be different for every file you compile.
3688 @item -fsched-verbose=@var{n}
3689 @opindex fsched-verbose
3690 On targets that use instruction scheduling, this option controls the
3691 amount of debugging output the scheduler prints. This information is
3692 written to standard error, unless @option{-dS} or @option{-dR} is
3693 specified, in which case it is output to the usual dump
3694 listing file, @file{.sched} or @file{.sched2} respectively. However
3695 for @var{n} greater than nine, the output is always printed to standard
3698 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
3699 same information as @option{-dRS}. For @var{n} greater than one, it
3700 also output basic block probabilities, detailed ready list information
3701 and unit/insn info. For @var{n} greater than two, it includes RTL
3702 at abort point, control-flow and regions info. And for @var{n} over
3703 four, @option{-fsched-verbose} also includes dependence info.
3707 Store the usual ``temporary'' intermediate files permanently; place them
3708 in the current directory and name them based on the source file. Thus,
3709 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
3710 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
3711 preprocessed @file{foo.i} output file even though the compiler now
3712 normally uses an integrated preprocessor.
3716 Report the CPU time taken by each subprocess in the compilation
3717 sequence. For C source files, this is the compiler proper and assembler
3718 (plus the linker if linking is done). The output looks like this:
3725 The first number on each line is the ``user time,'' that is time spent
3726 executing the program itself. The second number is ``system time,''
3727 time spent executing operating system routines on behalf of the program.
3728 Both numbers are in seconds.
3730 @item -fvar-tracking
3731 @opindex fvar-tracking
3732 Run variable tracking pass. It computes where variables are stored at each
3733 position in code. Better debugging information is then generated
3734 (if the debugging information format supports this information).
3736 It is enabled by default when compiling with optimization (@option{-Os},
3737 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
3738 the debug info format supports it.
3740 @item -print-file-name=@var{library}
3741 @opindex print-file-name
3742 Print the full absolute name of the library file @var{library} that
3743 would be used when linking---and don't do anything else. With this
3744 option, GCC does not compile or link anything; it just prints the
3747 @item -print-multi-directory
3748 @opindex print-multi-directory
3749 Print the directory name corresponding to the multilib selected by any
3750 other switches present in the command line. This directory is supposed
3751 to exist in @env{GCC_EXEC_PREFIX}.
3753 @item -print-multi-lib
3754 @opindex print-multi-lib
3755 Print the mapping from multilib directory names to compiler switches
3756 that enable them. The directory name is separated from the switches by
3757 @samp{;}, and each switch starts with an @samp{@@} instead of the
3758 @samp{-}, without spaces between multiple switches. This is supposed to
3759 ease shell-processing.
3761 @item -print-prog-name=@var{program}
3762 @opindex print-prog-name
3763 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
3765 @item -print-libgcc-file-name
3766 @opindex print-libgcc-file-name
3767 Same as @option{-print-file-name=libgcc.a}.
3769 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
3770 but you do want to link with @file{libgcc.a}. You can do
3773 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
3776 @item -print-search-dirs
3777 @opindex print-search-dirs
3778 Print the name of the configured installation directory and a list of
3779 program and library directories @command{gcc} will search---and don't do anything else.
3781 This is useful when @command{gcc} prints the error message
3782 @samp{installation problem, cannot exec cpp0: No such file or directory}.
3783 To resolve this you either need to put @file{cpp0} and the other compiler
3784 components where @command{gcc} expects to find them, or you can set the environment
3785 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
3786 Don't forget the trailing '/'.
3787 @xref{Environment Variables}.
3790 @opindex dumpmachine
3791 Print the compiler's target machine (for example,
3792 @samp{i686-pc-linux-gnu})---and don't do anything else.
3795 @opindex dumpversion
3796 Print the compiler version (for example, @samp{3.0})---and don't do
3801 Print the compiler's built-in specs---and don't do anything else. (This
3802 is used when GCC itself is being built.) @xref{Spec Files}.
3804 @item -feliminate-unused-debug-types
3805 @opindex feliminate-unused-debug-types
3806 Normally, when producing DWARF2 output, GCC will emit debugging
3807 information for all types declared in a compilation
3808 unit, regardless of whether or not they are actually used
3809 in that compilation unit. Sometimes this is useful, such as
3810 if, in the debugger, you want to cast a value to a type that is
3811 not actually used in your program (but is declared). More often,
3812 however, this results in a significant amount of wasted space.
3813 With this option, GCC will avoid producing debug symbol output
3814 for types that are nowhere used in the source file being compiled.
3817 @node Optimize Options
3818 @section Options That Control Optimization
3819 @cindex optimize options
3820 @cindex options, optimization
3822 These options control various sorts of optimizations.
3824 Without any optimization option, the compiler's goal is to reduce the
3825 cost of compilation and to make debugging produce the expected
3826 results. Statements are independent: if you stop the program with a
3827 breakpoint between statements, you can then assign a new value to any
3828 variable or change the program counter to any other statement in the
3829 function and get exactly the results you would expect from the source
3832 Turning on optimization flags makes the compiler attempt to improve
3833 the performance and/or code size at the expense of compilation time
3834 and possibly the ability to debug the program.
3836 The compiler performs optimization based on the knowledge it has of
3837 the program. Optimization levels @option{-O2} and above, in
3838 particular, enable @emph{unit-at-a-time} mode, which allows the
3839 compiler to consider information gained from later functions in
3840 the file when compiling a function. Compiling multiple files at
3841 once to a single output file in @emph{unit-at-a-time} mode allows
3842 the compiler to use information gained from all of the files when
3843 compiling each of them.
3845 Not all optimizations are controlled directly by a flag. Only
3846 optimizations that have a flag are listed.
3853 Optimize. Optimizing compilation takes somewhat more time, and a lot
3854 more memory for a large function.
3856 With @option{-O}, the compiler tries to reduce code size and execution
3857 time, without performing any optimizations that take a great deal of
3860 @option{-O} turns on the following optimization flags:
3861 @gccoptlist{-fdefer-pop @gol
3862 -fmerge-constants @gol
3864 -floop-optimize @gol
3865 -fif-conversion @gol
3866 -fif-conversion2 @gol
3867 -fdelayed-branch @gol
3868 -fguess-branch-probability @gol
3871 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
3872 where doing so does not interfere with debugging.
3876 Optimize even more. GCC performs nearly all supported optimizations
3877 that do not involve a space-speed tradeoff. The compiler does not
3878 perform loop unrolling or function inlining when you specify @option{-O2}.
3879 As compared to @option{-O}, this option increases both compilation time
3880 and the performance of the generated code.
3882 @option{-O2} turns on all optimization flags specified by @option{-O}. It
3883 also turns on the following optimization flags:
3884 @gccoptlist{-fforce-mem @gol
3885 -foptimize-sibling-calls @gol
3886 -fstrength-reduce @gol
3887 -fcse-follow-jumps -fcse-skip-blocks @gol
3888 -frerun-cse-after-loop -frerun-loop-opt @gol
3889 -fgcse -fgcse-lm -fgcse-sm -fgcse-las @gol
3890 -fdelete-null-pointer-checks @gol
3891 -fexpensive-optimizations @gol
3893 -fschedule-insns -fschedule-insns2 @gol
3894 -fsched-interblock -fsched-spec @gol
3897 -freorder-blocks -freorder-functions @gol
3898 -fstrict-aliasing @gol
3899 -funit-at-a-time @gol
3900 -falign-functions -falign-jumps @gol
3901 -falign-loops -falign-labels @gol
3904 Please note the warning under @option{-fgcse} about
3905 invoking @option{-O2} on programs that use computed gotos.
3909 Optimize yet more. @option{-O3} turns on all optimizations specified by
3910 @option{-O2} and also turns on the @option{-finline-functions},
3911 @option{-fweb} and @option{-fgcse-after-reload} options.
3915 Do not optimize. This is the default.
3919 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
3920 do not typically increase code size. It also performs further
3921 optimizations designed to reduce code size.
3923 @option{-Os} disables the following optimization flags:
3924 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
3925 -falign-labels -freorder-blocks -freorder-blocks-and-partition -fprefetch-loop-arrays}
3927 If you use multiple @option{-O} options, with or without level numbers,
3928 the last such option is the one that is effective.
3931 Options of the form @option{-f@var{flag}} specify machine-independent
3932 flags. Most flags have both positive and negative forms; the negative
3933 form of @option{-ffoo} would be @option{-fno-foo}. In the table
3934 below, only one of the forms is listed---the one you typically will
3935 use. You can figure out the other form by either removing @samp{no-}
3938 The following options control specific optimizations. They are either
3939 activated by @option{-O} options or are related to ones that are. You
3940 can use the following flags in the rare cases when ``fine-tuning'' of
3941 optimizations to be performed is desired.
3944 @item -fno-default-inline
3945 @opindex fno-default-inline
3946 Do not make member functions inline by default merely because they are
3947 defined inside the class scope (C++ only). Otherwise, when you specify
3948 @w{@option{-O}}, member functions defined inside class scope are compiled
3949 inline by default; i.e., you don't need to add @samp{inline} in front of
3950 the member function name.
3952 @item -fno-defer-pop
3953 @opindex fno-defer-pop
3954 Always pop the arguments to each function call as soon as that function
3955 returns. For machines which must pop arguments after a function call,
3956 the compiler normally lets arguments accumulate on the stack for several
3957 function calls and pops them all at once.
3959 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3963 Force memory operands to be copied into registers before doing
3964 arithmetic on them. This produces better code by making all memory
3965 references potential common subexpressions. When they are not common
3966 subexpressions, instruction combination should eliminate the separate
3969 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3972 @opindex fforce-addr
3973 Force memory address constants to be copied into registers before
3974 doing arithmetic on them. This may produce better code just as
3975 @option{-fforce-mem} may.
3977 @item -fomit-frame-pointer
3978 @opindex fomit-frame-pointer
3979 Don't keep the frame pointer in a register for functions that
3980 don't need one. This avoids the instructions to save, set up and
3981 restore frame pointers; it also makes an extra register available
3982 in many functions. @strong{It also makes debugging impossible on
3985 On some machines, such as the VAX, this flag has no effect, because
3986 the standard calling sequence automatically handles the frame pointer
3987 and nothing is saved by pretending it doesn't exist. The
3988 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
3989 whether a target machine supports this flag. @xref{Registers,,Register
3990 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
3992 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3994 @item -foptimize-sibling-calls
3995 @opindex foptimize-sibling-calls
3996 Optimize sibling and tail recursive calls.
3998 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4002 Don't pay attention to the @code{inline} keyword. Normally this option
4003 is used to keep the compiler from expanding any functions inline.
4004 Note that if you are not optimizing, no functions can be expanded inline.
4006 @item -finline-functions
4007 @opindex finline-functions
4008 Integrate all simple functions into their callers. The compiler
4009 heuristically decides which functions are simple enough to be worth
4010 integrating in this way.
4012 If all calls to a given function are integrated, and the function is
4013 declared @code{static}, then the function is normally not output as
4014 assembler code in its own right.
4016 Enabled at level @option{-O3}.
4018 @item -finline-limit=@var{n}
4019 @opindex finline-limit
4020 By default, GCC limits the size of functions that can be inlined. This flag
4021 allows the control of this limit for functions that are explicitly marked as
4022 inline (i.e., marked with the inline keyword or defined within the class
4023 definition in c++). @var{n} is the size of functions that can be inlined in
4024 number of pseudo instructions (not counting parameter handling). The default
4025 value of @var{n} is 600.
4026 Increasing this value can result in more inlined code at
4027 the cost of compilation time and memory consumption. Decreasing usually makes
4028 the compilation faster and less code will be inlined (which presumably
4029 means slower programs). This option is particularly useful for programs that
4030 use inlining heavily such as those based on recursive templates with C++.
4032 Inlining is actually controlled by a number of parameters, which may be
4033 specified individually by using @option{--param @var{name}=@var{value}}.
4034 The @option{-finline-limit=@var{n}} option sets some of these parameters
4038 @item max-inline-insns-single
4039 is set to @var{n}/2.
4040 @item max-inline-insns-auto
4041 is set to @var{n}/2.
4042 @item min-inline-insns
4043 is set to 130 or @var{n}/4, whichever is smaller.
4044 @item max-inline-insns-rtl
4048 See below for a documentation of the individual
4049 parameters controlling inlining.
4051 @emph{Note:} pseudo instruction represents, in this particular context, an
4052 abstract measurement of function's size. In no way, it represents a count
4053 of assembly instructions and as such its exact meaning might change from one
4054 release to an another.
4056 @item -fkeep-inline-functions
4057 @opindex fkeep-inline-functions
4058 Even if all calls to a given function are integrated, and the function
4059 is declared @code{static}, nevertheless output a separate run-time
4060 callable version of the function. This switch does not affect
4061 @code{extern inline} functions.
4063 @item -fkeep-static-consts
4064 @opindex fkeep-static-consts
4065 Emit variables declared @code{static const} when optimization isn't turned
4066 on, even if the variables aren't referenced.
4068 GCC enables this option by default. If you want to force the compiler to
4069 check if the variable was referenced, regardless of whether or not
4070 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4072 @item -fmerge-constants
4073 Attempt to merge identical constants (string constants and floating point
4074 constants) across compilation units.
4076 This option is the default for optimized compilation if the assembler and
4077 linker support it. Use @option{-fno-merge-constants} to inhibit this
4080 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4082 @item -fmerge-all-constants
4083 Attempt to merge identical constants and identical variables.
4085 This option implies @option{-fmerge-constants}. In addition to
4086 @option{-fmerge-constants} this considers e.g. even constant initialized
4087 arrays or initialized constant variables with integral or floating point
4088 types. Languages like C or C++ require each non-automatic variable to
4089 have distinct location, so using this option will result in non-conforming
4092 @item -fmodulo-sched
4093 @opindex fmodulo-sched
4094 Perform swing modulo scheduling immediately before the first scheduling
4095 pass. This pass looks at innermost loops and reorders their
4096 instructions by overlapping different iterations.
4100 Use a graph coloring register allocator. Currently this option is meant
4101 only for testing. Users should not specify this option, since it is not
4102 yet ready for production use.
4104 @item -fno-branch-count-reg
4105 @opindex fno-branch-count-reg
4106 Do not use ``decrement and branch'' instructions on a count register,
4107 but instead generate a sequence of instructions that decrement a
4108 register, compare it against zero, then branch based upon the result.
4109 This option is only meaningful on architectures that support such
4110 instructions, which include x86, PowerPC, IA-64 and S/390.
4112 The default is @option{-fbranch-count-reg}, enabled when
4113 @option{-fstrength-reduce} is enabled.
4115 @item -fno-function-cse
4116 @opindex fno-function-cse
4117 Do not put function addresses in registers; make each instruction that
4118 calls a constant function contain the function's address explicitly.
4120 This option results in less efficient code, but some strange hacks
4121 that alter the assembler output may be confused by the optimizations
4122 performed when this option is not used.
4124 The default is @option{-ffunction-cse}
4126 @item -fno-zero-initialized-in-bss
4127 @opindex fno-zero-initialized-in-bss
4128 If the target supports a BSS section, GCC by default puts variables that
4129 are initialized to zero into BSS@. This can save space in the resulting
4132 This option turns off this behavior because some programs explicitly
4133 rely on variables going to the data section. E.g., so that the
4134 resulting executable can find the beginning of that section and/or make
4135 assumptions based on that.
4137 The default is @option{-fzero-initialized-in-bss}.
4139 @item -fbounds-check
4140 @opindex fbounds-check
4141 For front-ends that support it, generate additional code to check that
4142 indices used to access arrays are within the declared range. This is
4143 currently only supported by the Java and Fortran front-ends, where
4144 this option defaults to true and false respectively.
4146 @item -fmudflap -fmudflapth -fmudflapir
4150 @cindex bounds checking
4152 For front-ends that support it (C and C++), instrument all risky
4153 pointer/array dereferencing operations, some standard library
4154 string/heap functions, and some other associated constructs with
4155 range/validity tests. Modules so instrumented should be immune to
4156 buffer overflows, invalid heap use, and some other classes of C/C++
4157 programming errors. The instrumentation relies on a separate runtime
4158 library (@file{libmudflap}), which will be linked into a program if
4159 @option{-fmudflap} is given at link time. Run-time behavior of the
4160 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4161 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4164 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4165 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4166 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4167 instrumentation should ignore pointer reads. This produces less
4168 instrumentation (and therefore faster execution) and still provides
4169 some protection against outright memory corrupting writes, but allows
4170 erroneously read data to propagate within a program.
4172 @item -fstrength-reduce
4173 @opindex fstrength-reduce
4174 Perform the optimizations of loop strength reduction and
4175 elimination of iteration variables.
4177 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4179 @item -fthread-jumps
4180 @opindex fthread-jumps
4181 Perform optimizations where we check to see if a jump branches to a
4182 location where another comparison subsumed by the first is found. If
4183 so, the first branch is redirected to either the destination of the
4184 second branch or a point immediately following it, depending on whether
4185 the condition is known to be true or false.
4187 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4189 @item -fcse-follow-jumps
4190 @opindex fcse-follow-jumps
4191 In common subexpression elimination, scan through jump instructions
4192 when the target of the jump is not reached by any other path. For
4193 example, when CSE encounters an @code{if} statement with an
4194 @code{else} clause, CSE will follow the jump when the condition
4197 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4199 @item -fcse-skip-blocks
4200 @opindex fcse-skip-blocks
4201 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4202 follow jumps which conditionally skip over blocks. When CSE
4203 encounters a simple @code{if} statement with no else clause,
4204 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4205 body of the @code{if}.
4207 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4209 @item -frerun-cse-after-loop
4210 @opindex frerun-cse-after-loop
4211 Re-run common subexpression elimination after loop optimizations has been
4214 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4216 @item -frerun-loop-opt
4217 @opindex frerun-loop-opt
4218 Run the loop optimizer twice.
4220 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4224 Perform a global common subexpression elimination pass.
4225 This pass also performs global constant and copy propagation.
4227 @emph{Note:} When compiling a program using computed gotos, a GCC
4228 extension, you may get better runtime performance if you disable
4229 the global common subexpression elimination pass by adding
4230 @option{-fno-gcse} to the command line.
4232 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4236 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4237 attempt to move loads which are only killed by stores into themselves. This
4238 allows a loop containing a load/store sequence to be changed to a load outside
4239 the loop, and a copy/store within the loop.
4241 Enabled by default when gcse is enabled.
4245 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4246 global common subexpression elimination. This pass will attempt to move
4247 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4248 loops containing a load/store sequence can be changed to a load before
4249 the loop and a store after the loop.
4251 Enabled by default when gcse is enabled.
4255 When @option{-fgcse-las} is enabled, the global common subexpression
4256 elimination pass eliminates redundant loads that come after stores to the
4257 same memory location (both partial and full redundancies).
4259 Enabled by default when gcse is enabled.
4261 @item -fgcse-after-reload
4262 @opindex fgcse-after-reload
4263 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4264 pass is performed after reload. The purpose of this pass is to cleanup
4267 @item -floop-optimize
4268 @opindex floop-optimize
4269 Perform loop optimizations: move constant expressions out of loops, simplify
4270 exit test conditions and optionally do strength-reduction and loop unrolling as
4273 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4275 @item -floop-optimize2
4276 @opindex floop-optimize2
4277 Perform loop optimizations using the new loop optimizer. The optimizations
4278 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4281 @item -fcrossjumping
4282 @opindex crossjumping
4283 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4284 resulting code may or may not perform better than without cross-jumping.
4286 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4288 @item -fif-conversion
4289 @opindex if-conversion
4290 Attempt to transform conditional jumps into branch-less equivalents. This
4291 include use of conditional moves, min, max, set flags and abs instructions, and
4292 some tricks doable by standard arithmetics. The use of conditional execution
4293 on chips where it is available is controlled by @code{if-conversion2}.
4295 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4297 @item -fif-conversion2
4298 @opindex if-conversion2
4299 Use conditional execution (where available) to transform conditional jumps into
4300 branch-less equivalents.
4302 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4304 @item -fdelete-null-pointer-checks
4305 @opindex fdelete-null-pointer-checks
4306 Use global dataflow analysis to identify and eliminate useless checks
4307 for null pointers. The compiler assumes that dereferencing a null
4308 pointer would have halted the program. If a pointer is checked after
4309 it has already been dereferenced, it cannot be null.
4311 In some environments, this assumption is not true, and programs can
4312 safely dereference null pointers. Use
4313 @option{-fno-delete-null-pointer-checks} to disable this optimization
4314 for programs which depend on that behavior.
4316 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4318 @item -fexpensive-optimizations
4319 @opindex fexpensive-optimizations
4320 Perform a number of minor optimizations that are relatively expensive.
4322 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4324 @item -foptimize-register-move
4326 @opindex foptimize-register-move
4328 Attempt to reassign register numbers in move instructions and as
4329 operands of other simple instructions in order to maximize the amount of
4330 register tying. This is especially helpful on machines with two-operand
4333 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4336 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4338 @item -fdelayed-branch
4339 @opindex fdelayed-branch
4340 If supported for the target machine, attempt to reorder instructions
4341 to exploit instruction slots available after delayed branch
4344 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4346 @item -fschedule-insns
4347 @opindex fschedule-insns
4348 If supported for the target machine, attempt to reorder instructions to
4349 eliminate execution stalls due to required data being unavailable. This
4350 helps machines that have slow floating point or memory load instructions
4351 by allowing other instructions to be issued until the result of the load
4352 or floating point instruction is required.
4354 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4356 @item -fschedule-insns2
4357 @opindex fschedule-insns2
4358 Similar to @option{-fschedule-insns}, but requests an additional pass of
4359 instruction scheduling after register allocation has been done. This is
4360 especially useful on machines with a relatively small number of
4361 registers and where memory load instructions take more than one cycle.
4363 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4365 @item -fno-sched-interblock
4366 @opindex fno-sched-interblock
4367 Don't schedule instructions across basic blocks. This is normally
4368 enabled by default when scheduling before register allocation, i.e.@:
4369 with @option{-fschedule-insns} or at @option{-O2} or higher.
4371 @item -fno-sched-spec
4372 @opindex fno-sched-spec
4373 Don't allow speculative motion of non-load instructions. This is normally
4374 enabled by default when scheduling before register allocation, i.e.@:
4375 with @option{-fschedule-insns} or at @option{-O2} or higher.
4377 @item -fsched-spec-load
4378 @opindex fsched-spec-load
4379 Allow speculative motion of some load instructions. This only makes
4380 sense when scheduling before register allocation, i.e.@: with
4381 @option{-fschedule-insns} or at @option{-O2} or higher.
4383 @item -fsched-spec-load-dangerous
4384 @opindex fsched-spec-load-dangerous
4385 Allow speculative motion of more load instructions. This only makes
4386 sense when scheduling before register allocation, i.e.@: with
4387 @option{-fschedule-insns} or at @option{-O2} or higher.
4389 @item -fsched-stalled-insns=@var{n}
4390 @opindex fsched-stalled-insns
4391 Define how many insns (if any) can be moved prematurely from the queue
4392 of stalled insns into the ready list, during the second scheduling pass.
4394 @item -fsched-stalled-insns-dep=@var{n}
4395 @opindex fsched-stalled-insns-dep
4396 Define how many insn groups (cycles) will be examined for a dependency
4397 on a stalled insn that is candidate for premature removal from the queue
4398 of stalled insns. Has an effect only during the second scheduling pass,
4399 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4401 @item -fsched2-use-superblocks
4402 @opindex fsched2-use-superblocks
4403 When scheduling after register allocation, do use superblock scheduling
4404 algorithm. Superblock scheduling allows motion across basic block boundaries
4405 resulting on faster schedules. This option is experimental, as not all machine
4406 descriptions used by GCC model the CPU closely enough to avoid unreliable
4407 results from the algorithm.
4409 This only makes sense when scheduling after register allocation, i.e.@: with
4410 @option{-fschedule-insns2} or at @option{-O2} or higher.
4412 @item -fsched2-use-traces
4413 @opindex fsched2-use-traces
4414 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4415 allocation and additionally perform code duplication in order to increase the
4416 size of superblocks using tracer pass. See @option{-ftracer} for details on
4419 This mode should produce faster but significantly longer programs. Also
4420 without @code{-fbranch-probabilities} the traces constructed may not match the
4421 reality and hurt the performance. This only makes
4422 sense when scheduling after register allocation, i.e.@: with
4423 @option{-fschedule-insns2} or at @option{-O2} or higher.
4425 @item -freschedule-modulo-scheduled-loops
4426 @opindex fscheduling-in-modulo-scheduled-loops
4427 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4428 we may want to prevent the later scheduling passes from changing its schedule, we use this
4429 option to control that.
4431 @item -fcaller-saves
4432 @opindex fcaller-saves
4433 Enable values to be allocated in registers that will be clobbered by
4434 function calls, by emitting extra instructions to save and restore the
4435 registers around such calls. Such allocation is done only when it
4436 seems to result in better code than would otherwise be produced.
4438 This option is always enabled by default on certain machines, usually
4439 those which have no call-preserved registers to use instead.
4441 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4444 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4445 enabled by default at -O and higher.
4448 Perform Full Redundancy Elimination (FRE) on trees. The difference
4449 between FRE and PRE is that FRE only considers expressions
4450 that are computed on all paths leading to the redundant computation.
4451 This analysis faster than PRE, though it exposes fewer redundancies.
4452 This flag is enabled by default at -O and higher.
4455 Perform sparse conditional constant propagation (CCP) on trees. This flag
4456 is enabled by default at -O and higher.
4459 Perform dead code elimination (DCE) on trees. This flag is enabled by
4460 default at -O and higher.
4462 @item -ftree-dominator-opts
4463 Perform dead code elimination (DCE) on trees. This flag is enabled by
4464 default at -O and higher.
4467 Perform loop header copying on trees. This is beneficial since it increases
4468 effectivity of code motion optimizations. It also saves one jump. This flag
4469 is enabled by default at -O and higher. It is not enabled for -Os, since it
4470 usually increases code size.
4472 @item -ftree-loop-optimize
4473 Perform loop optimizations on trees. This flag is enabled by default at -O
4477 Perform loop invariant motion on trees. This pass moves only invartiants that
4478 would be hard to handle on rtl level (function calls, operations that expand to
4479 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
4480 operands of conditions that are invariant out of the loop, so that we can use
4481 just trivial invariantness analysis in loop unswitching. The pass also includes
4485 Perform scalar replacement of aggregates. This pass replaces structure
4486 references with scalars to prevent committing structures to memory too
4487 early. This flag is enabled by default at -O and higher.
4489 @item -ftree-copyrename
4490 Perform copy renaming on trees. This pass attempts to rename compiler
4491 temporaries to other variables at copy locations, usually resulting in
4492 variable names which more closely resemble the original variables. This flag
4493 is enabled by default at -O and higher.
4496 Perform temporary expression replacement during the SSA->normal phase. Single
4497 use/single def temporaries are replaced at their use location with their
4498 defining expression. This results in non-GIMPLE code, but gives the expanders
4499 much more complex trees to work on resulting in better RTL generation. This is
4500 enabled by default at -O and higher.
4503 Perform live range splitting during the SSA->normal phase. Distinct live
4504 ranges of a variable are split into unique variables, allowing for better
4505 optimization later. This is enabled by default at -O and higher.
4507 @item -ftree-vectorize
4508 Perform loop vectorization on trees.
4512 Perform tail duplication to enlarge superblock size. This transformation
4513 simplifies the control flow of the function allowing other optimizations to do
4516 @item -funroll-loops
4517 @opindex funroll-loops
4518 Unroll loops whose number of iterations can be determined at compile
4519 time or upon entry to the loop. @option{-funroll-loops} implies both
4520 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4521 option makes code larger, and may or may not make it run faster.
4523 @item -funroll-all-loops
4524 @opindex funroll-all-loops
4525 Unroll all loops, even if their number of iterations is uncertain when
4526 the loop is entered. This usually makes programs run more slowly.
4527 @option{-funroll-all-loops} implies the same options as
4528 @option{-funroll-loops},
4530 @item -fprefetch-loop-arrays
4531 @opindex fprefetch-loop-arrays
4532 If supported by the target machine, generate instructions to prefetch
4533 memory to improve the performance of loops that access large arrays.
4535 @item -fmove-all-movables
4536 @opindex fmove-all-movables
4537 Forces all invariant computations in loops to be moved
4540 @item -freduce-all-givs
4541 @opindex freduce-all-givs
4542 Forces all general-induction variables in loops to be
4545 @emph{Note:} When compiling programs written in Fortran,
4546 @option{-fmove-all-movables} and @option{-freduce-all-givs} are enabled
4547 by default when you use the optimizer.
4549 These options may generate better or worse code; results are highly
4550 dependent on the structure of loops within the source code.
4552 These two options are intended to be removed someday, once
4553 they have helped determine the efficacy of various
4554 approaches to improving loop optimizations.
4556 Please contact @w{@email{gcc@@gcc.gnu.org}}, and describe how use of
4557 these options affects the performance of your production code.
4558 Examples of code that runs @emph{slower} when these options are
4559 @emph{enabled} are very valuable.
4562 @itemx -fno-peephole2
4563 @opindex fno-peephole
4564 @opindex fno-peephole2
4565 Disable any machine-specific peephole optimizations. The difference
4566 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
4567 are implemented in the compiler; some targets use one, some use the
4568 other, a few use both.
4570 @option{-fpeephole} is enabled by default.
4571 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4573 @item -fno-guess-branch-probability
4574 @opindex fno-guess-branch-probability
4575 Do not guess branch probabilities using a randomized model.
4577 Sometimes GCC will opt to use a randomized model to guess branch
4578 probabilities, when none are available from either profiling feedback
4579 (@option{-fprofile-arcs}) or @samp{__builtin_expect}. This means that
4580 different runs of the compiler on the same program may produce different
4583 In a hard real-time system, people don't want different runs of the
4584 compiler to produce code that has different behavior; minimizing
4585 non-determinism is of paramount import. This switch allows users to
4586 reduce non-determinism, possibly at the expense of inferior
4589 The default is @option{-fguess-branch-probability} at levels
4590 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4592 @item -freorder-blocks
4593 @opindex freorder-blocks
4594 Reorder basic blocks in the compiled function in order to reduce number of
4595 taken branches and improve code locality.
4597 Enabled at levels @option{-O2}, @option{-O3}.
4599 @item -freorder-blocks-and-partition
4600 @opindex freorder-blocks-and-partition
4601 In addition to reordering basic blocks in the compiled function, in order
4602 to reduce number of taken branches, partitions hot and cold basic blocks
4603 into separate sections of the assembly and .o files, to improve
4604 paging and cache locality performance.
4606 @item -freorder-functions
4607 @opindex freorder-functions
4608 Reorder basic blocks in the compiled function in order to reduce number of
4609 taken branches and improve code locality. This is implemented by using special
4610 subsections @code{.text.hot} for most frequently executed functions and
4611 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
4612 the linker so object file format must support named sections and linker must
4613 place them in a reasonable way.
4615 Also profile feedback must be available in to make this option effective. See
4616 @option{-fprofile-arcs} for details.
4618 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4620 @item -fstrict-aliasing
4621 @opindex fstrict-aliasing
4622 Allows the compiler to assume the strictest aliasing rules applicable to
4623 the language being compiled. For C (and C++), this activates
4624 optimizations based on the type of expressions. In particular, an
4625 object of one type is assumed never to reside at the same address as an
4626 object of a different type, unless the types are almost the same. For
4627 example, an @code{unsigned int} can alias an @code{int}, but not a
4628 @code{void*} or a @code{double}. A character type may alias any other
4631 Pay special attention to code like this:
4644 The practice of reading from a different union member than the one most
4645 recently written to (called ``type-punning'') is common. Even with
4646 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
4647 is accessed through the union type. So, the code above will work as
4648 expected. However, this code might not:
4659 Every language that wishes to perform language-specific alias analysis
4660 should define a function that computes, given an @code{tree}
4661 node, an alias set for the node. Nodes in different alias sets are not
4662 allowed to alias. For an example, see the C front-end function
4663 @code{c_get_alias_set}.
4665 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4667 @item -falign-functions
4668 @itemx -falign-functions=@var{n}
4669 @opindex falign-functions
4670 Align the start of functions to the next power-of-two greater than
4671 @var{n}, skipping up to @var{n} bytes. For instance,
4672 @option{-falign-functions=32} aligns functions to the next 32-byte
4673 boundary, but @option{-falign-functions=24} would align to the next
4674 32-byte boundary only if this can be done by skipping 23 bytes or less.
4676 @option{-fno-align-functions} and @option{-falign-functions=1} are
4677 equivalent and mean that functions will not be aligned.
4679 Some assemblers only support this flag when @var{n} is a power of two;
4680 in that case, it is rounded up.
4682 If @var{n} is not specified or is zero, use a machine-dependent default.
4684 Enabled at levels @option{-O2}, @option{-O3}.
4686 @item -falign-labels
4687 @itemx -falign-labels=@var{n}
4688 @opindex falign-labels
4689 Align all branch targets to a power-of-two boundary, skipping up to
4690 @var{n} bytes like @option{-falign-functions}. This option can easily
4691 make code slower, because it must insert dummy operations for when the
4692 branch target is reached in the usual flow of the code.
4694 @option{-fno-align-labels} and @option{-falign-labels=1} are
4695 equivalent and mean that labels will not be aligned.
4697 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
4698 are greater than this value, then their values are used instead.
4700 If @var{n} is not specified or is zero, use a machine-dependent default
4701 which is very likely to be @samp{1}, meaning no alignment.
4703 Enabled at levels @option{-O2}, @option{-O3}.
4706 @itemx -falign-loops=@var{n}
4707 @opindex falign-loops
4708 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
4709 like @option{-falign-functions}. The hope is that the loop will be
4710 executed many times, which will make up for any execution of the dummy
4713 @option{-fno-align-loops} and @option{-falign-loops=1} are
4714 equivalent and mean that loops will not be aligned.
4716 If @var{n} is not specified or is zero, use a machine-dependent default.
4718 Enabled at levels @option{-O2}, @option{-O3}.
4721 @itemx -falign-jumps=@var{n}
4722 @opindex falign-jumps
4723 Align branch targets to a power-of-two boundary, for branch targets
4724 where the targets can only be reached by jumping, skipping up to @var{n}
4725 bytes like @option{-falign-functions}. In this case, no dummy operations
4728 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
4729 equivalent and mean that loops will not be aligned.
4731 If @var{n} is not specified or is zero, use a machine-dependent default.
4733 Enabled at levels @option{-O2}, @option{-O3}.
4735 @item -funit-at-a-time
4736 @opindex funit-at-a-time
4737 Parse the whole compilation unit before starting to produce code.
4738 This allows some extra optimizations to take place but consumes
4739 more memory (in general). There are some compatibility issues
4740 with @emph{unit-at-at-time} mode:
4743 enabling @emph{unit-at-a-time} mode may change the order
4744 in which functions, variables, and top-level @code{asm} statements
4745 are emitted, and will likely break code relying on some particular
4746 ordering. The majority of such top-level @code{asm} statements,
4747 though, can be replaced by @code{section} attributes.
4750 @emph{unit-at-a-time} mode removes unreferenced static variables
4751 and functions are removed. This may result in undefined references
4752 when an @code{asm} statement refers directly to variables or functions
4753 that are otherwise unused. In that case either the variable/function
4754 shall be listed as an operand of the @code{asm} statement operand or,
4755 in the case of top-level @code{asm} statements the attribute @code{used}
4756 shall be used on the declaration.
4759 Static functions now can use non-standard passing conventions that
4760 may break @code{asm} statements calling functions directly. Again,
4761 attribute @code{used} will prevent this behavior.
4764 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
4765 but this scheme may not be supported by future releases of GCC.
4767 Enabled at levels @option{-O2}, @option{-O3}.
4771 Constructs webs as commonly used for register allocation purposes and assign
4772 each web individual pseudo register. This allows the register allocation pass
4773 to operate on pseudos directly, but also strengthens several other optimization
4774 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
4775 however, make debugging impossible, since variables will no longer stay in a
4778 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os},
4779 on targets where the default format for debugging information supports
4782 @item -fno-cprop-registers
4783 @opindex fno-cprop-registers
4784 After register allocation and post-register allocation instruction splitting,
4785 we perform a copy-propagation pass to try to reduce scheduling dependencies
4786 and occasionally eliminate the copy.
4788 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4790 @item -fprofile-generate
4791 @opindex fprofile-generate
4793 Enable options usually used for instrumenting application to produce
4794 profile useful for later recompilation with profile feedback based
4795 optimization. You must use @code{-fprofile-generate} both when
4796 compiling and when linking your program.
4798 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
4801 @opindex fprofile-use
4802 Enable profile feedback directed optimizations, and optimizations
4803 generally profitable only with profile feedback available.
4805 The following options are enabled: @code{-fbranch-probabilities},
4806 @code{-fvpt}, @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}.
4810 The following options control compiler behavior regarding floating
4811 point arithmetic. These options trade off between speed and
4812 correctness. All must be specifically enabled.
4816 @opindex ffloat-store
4817 Do not store floating point variables in registers, and inhibit other
4818 options that might change whether a floating point value is taken from a
4821 @cindex floating point precision
4822 This option prevents undesirable excess precision on machines such as
4823 the 68000 where the floating registers (of the 68881) keep more
4824 precision than a @code{double} is supposed to have. Similarly for the
4825 x86 architecture. For most programs, the excess precision does only
4826 good, but a few programs rely on the precise definition of IEEE floating
4827 point. Use @option{-ffloat-store} for such programs, after modifying
4828 them to store all pertinent intermediate computations into variables.
4832 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
4833 @option{-fno-trapping-math}, @option{-ffinite-math-only},
4834 @option{-fno-rounding-math} and @option{-fno-signaling-nans}.
4836 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
4838 This option should never be turned on by any @option{-O} option since
4839 it can result in incorrect output for programs which depend on
4840 an exact implementation of IEEE or ISO rules/specifications for
4843 @item -fno-math-errno
4844 @opindex fno-math-errno
4845 Do not set ERRNO after calling math functions that are executed
4846 with a single instruction, e.g., sqrt. A program that relies on
4847 IEEE exceptions for math error handling may want to use this flag
4848 for speed while maintaining IEEE arithmetic compatibility.
4850 This option should never be turned on by any @option{-O} option since
4851 it can result in incorrect output for programs which depend on
4852 an exact implementation of IEEE or ISO rules/specifications for
4855 The default is @option{-fmath-errno}.
4857 @item -funsafe-math-optimizations
4858 @opindex funsafe-math-optimizations
4859 Allow optimizations for floating-point arithmetic that (a) assume
4860 that arguments and results are valid and (b) may violate IEEE or
4861 ANSI standards. When used at link-time, it may include libraries
4862 or startup files that change the default FPU control word or other
4863 similar optimizations.
4865 This option should never be turned on by any @option{-O} option since
4866 it can result in incorrect output for programs which depend on
4867 an exact implementation of IEEE or ISO rules/specifications for
4870 The default is @option{-fno-unsafe-math-optimizations}.
4872 @item -ffinite-math-only
4873 @opindex ffinite-math-only
4874 Allow optimizations for floating-point arithmetic that assume
4875 that arguments and results are not NaNs or +-Infs.
4877 This option should never be turned on by any @option{-O} option since
4878 it can result in incorrect output for programs which depend on
4879 an exact implementation of IEEE or ISO rules/specifications.
4881 The default is @option{-fno-finite-math-only}.
4883 @item -fno-trapping-math
4884 @opindex fno-trapping-math
4885 Compile code assuming that floating-point operations cannot generate
4886 user-visible traps. These traps include division by zero, overflow,
4887 underflow, inexact result and invalid operation. This option implies
4888 @option{-fno-signaling-nans}. Setting this option may allow faster
4889 code if one relies on ``non-stop'' IEEE arithmetic, for example.
4891 This option should never be turned on by any @option{-O} option since
4892 it can result in incorrect output for programs which depend on
4893 an exact implementation of IEEE or ISO rules/specifications for
4896 The default is @option{-ftrapping-math}.
4898 @item -frounding-math
4899 @opindex frounding-math
4900 Disable transformations and optimizations that assume default floating
4901 point rounding behavior. This is round-to-zero for all floating point
4902 to integer conversions, and round-to-nearest for all other arithmetic
4903 truncations. This option should be specified for programs that change
4904 the FP rounding mode dynamically, or that may be executed with a
4905 non-default rounding mode. This option disables constant folding of
4906 floating point expressions at compile-time (which may be affected by
4907 rounding mode) and arithmetic transformations that are unsafe in the
4908 presence of sign-dependent rounding modes.
4910 The default is @option{-fno-rounding-math}.
4912 This option is experimental and does not currently guarantee to
4913 disable all GCC optimizations that are affected by rounding mode.
4914 Future versions of GCC may provide finer control of this setting
4915 using C99's @code{FENV_ACCESS} pragma. This command line option
4916 will be used to specify the default state for @code{FENV_ACCESS}.
4918 @item -fsignaling-nans
4919 @opindex fsignaling-nans
4920 Compile code assuming that IEEE signaling NaNs may generate user-visible
4921 traps during floating-point operations. Setting this option disables
4922 optimizations that may change the number of exceptions visible with
4923 signaling NaNs. This option implies @option{-ftrapping-math}.
4925 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
4928 The default is @option{-fno-signaling-nans}.
4930 This option is experimental and does not currently guarantee to
4931 disable all GCC optimizations that affect signaling NaN behavior.
4933 @item -fsingle-precision-constant
4934 @opindex fsingle-precision-constant
4935 Treat floating point constant as single precision constant instead of
4936 implicitly converting it to double precision constant.
4941 The following options control optimizations that may improve
4942 performance, but are not enabled by any @option{-O} options. This
4943 section includes experimental options that may produce broken code.
4946 @item -fbranch-probabilities
4947 @opindex fbranch-probabilities
4948 After running a program compiled with @option{-fprofile-arcs}
4949 (@pxref{Debugging Options,, Options for Debugging Your Program or
4950 @command{gcc}}), you can compile it a second time using
4951 @option{-fbranch-probabilities}, to improve optimizations based on
4952 the number of times each branch was taken. When the program
4953 compiled with @option{-fprofile-arcs} exits it saves arc execution
4954 counts to a file called @file{@var{sourcename}.gcda} for each source
4955 file The information in this data file is very dependent on the
4956 structure of the generated code, so you must use the same source code
4957 and the same optimization options for both compilations.
4959 With @option{-fbranch-probabilities}, GCC puts a
4960 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
4961 These can be used to improve optimization. Currently, they are only
4962 used in one place: in @file{reorg.c}, instead of guessing which path a
4963 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
4964 exactly determine which path is taken more often.
4966 @item -fprofile-values
4967 @opindex fprofile-values
4968 If combined with @option{-fprofile-arcs}, it adds code so that some
4969 data about values of expressions in the program is gathered.
4971 With @option{-fbranch-probabilities}, it reads back the data gathered
4972 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
4973 notes to instructions for their later usage in optimizations.
4975 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
4979 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
4980 a code to gather information about values of expressions.
4982 With @option{-fbranch-probabilities}, it reads back the data gathered
4983 and actually performs the optimizations based on them.
4984 Currently the optimizations include specialization of division operation
4985 using the knowledge about the value of the denominator.
4987 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
4989 @item -frename-registers
4990 @opindex frename-registers
4991 Attempt to avoid false dependencies in scheduled code by making use
4992 of registers left over after register allocation. This optimization
4993 will most benefit processors with lots of registers. Depending on the
4994 debug information format adopted by the target, however, it can
4995 make debugging impossible, since variables will no longer stay in
4996 a ``home register''.
4998 Not enabled by default at any level because it has known bugs.
5002 Use a graph coloring register allocator. Currently this option is meant
5003 for testing, so we are interested to hear about miscompilations with
5008 Perform tail duplication to enlarge superblock size. This transformation
5009 simplifies the control flow of the function allowing other optimizations to do
5012 Enabled with @option{-fprofile-use}.
5014 @item -funroll-loops
5015 @opindex funroll-loops
5016 Unroll loops whose number of iterations can be determined at compile time or
5017 upon entry to the loop. @option{-funroll-loops} implies
5018 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
5019 (i.e. complete removal of loops with small constant number of iterations).
5020 This option makes code larger, and may or may not make it run faster.
5022 Enabled with @option{-fprofile-use}.
5024 @item -funroll-all-loops
5025 @opindex funroll-all-loops
5026 Unroll all loops, even if their number of iterations is uncertain when
5027 the loop is entered. This usually makes programs run more slowly.
5028 @option{-funroll-all-loops} implies the same options as
5029 @option{-funroll-loops}.
5032 @opindex fpeel-loops
5033 Peels the loops for that there is enough information that they do not
5034 roll much (from profile feedback). It also turns on complete loop peeling
5035 (i.e. complete removal of loops with small constant number of iterations).
5037 Enabled with @option{-fprofile-use}.
5039 @item -fmove-loop-invariants
5040 @opindex fmove-loop-invariants
5041 Enables the loop invariant motion pass in the new loop optimizer. Enabled
5042 at level @option{-O1}
5044 @item -funswitch-loops
5045 @opindex funswitch-loops
5046 Move branches with loop invariant conditions out of the loop, with duplicates
5047 of the loop on both branches (modified according to result of the condition).
5049 @item -fold-unroll-loops
5050 @opindex fold-unroll-loops
5051 Unroll loops whose number of iterations can be determined at compile
5052 time or upon entry to the loop, using the old loop unroller whose loop
5053 recognition is based on notes from frontend. @option{-fold-unroll-loops} implies
5054 both @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
5055 option makes code larger, and may or may not make it run faster.
5057 @item -fold-unroll-all-loops
5058 @opindex fold-unroll-all-loops
5059 Unroll all loops, even if their number of iterations is uncertain when
5060 the loop is entered. This is done using the old loop unroller whose loop
5061 recognition is based on notes from frontend. This usually makes programs run more slowly.
5062 @option{-fold-unroll-all-loops} implies the same options as
5063 @option{-fold-unroll-loops}.
5065 @item -fprefetch-loop-arrays
5066 @opindex fprefetch-loop-arrays
5067 If supported by the target machine, generate instructions to prefetch
5068 memory to improve the performance of loops that access large arrays.
5070 Disabled at level @option{-Os}.
5072 @item -ffunction-sections
5073 @itemx -fdata-sections
5074 @opindex ffunction-sections
5075 @opindex fdata-sections
5076 Place each function or data item into its own section in the output
5077 file if the target supports arbitrary sections. The name of the
5078 function or the name of the data item determines the section's name
5081 Use these options on systems where the linker can perform optimizations
5082 to improve locality of reference in the instruction space. Most systems
5083 using the ELF object format and SPARC processors running Solaris 2 have
5084 linkers with such optimizations. AIX may have these optimizations in
5087 Only use these options when there are significant benefits from doing
5088 so. When you specify these options, the assembler and linker will
5089 create larger object and executable files and will also be slower.
5090 You will not be able to use @code{gprof} on all systems if you
5091 specify this option and you may have problems with debugging if
5092 you specify both this option and @option{-g}.
5094 @item -fbranch-target-load-optimize
5095 @opindex fbranch-target-load-optimize
5096 Perform branch target register load optimization before prologue / epilogue
5098 The use of target registers can typically be exposed only during reload,
5099 thus hoisting loads out of loops and doing inter-block scheduling needs
5100 a separate optimization pass.
5102 @item -fbranch-target-load-optimize2
5103 @opindex fbranch-target-load-optimize2
5104 Perform branch target register load optimization after prologue / epilogue
5107 @item -fbtr-bb-exclusive
5108 @opindex fbtr-bb-exclusive
5109 When performing branch target register load optimization, don't reuse
5110 branch target registers in within any basic block.
5112 @item --param @var{name}=@var{value}
5114 In some places, GCC uses various constants to control the amount of
5115 optimization that is done. For example, GCC will not inline functions
5116 that contain more that a certain number of instructions. You can
5117 control some of these constants on the command-line using the
5118 @option{--param} option.
5120 The names of specific parameters, and the meaning of the values, are
5121 tied to the internals of the compiler, and are subject to change
5122 without notice in future releases.
5124 In each case, the @var{value} is an integer. The allowable choices for
5125 @var{name} are given in the following table:
5128 @item max-crossjump-edges
5129 The maximum number of incoming edges to consider for crossjumping.
5130 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5131 the number of edges incoming to each block. Increasing values mean
5132 more aggressive optimization, making the compile time increase with
5133 probably small improvement in executable size.
5135 @item min-crossjump-insns
5136 The minimum number of instructions which must be matched at the end
5137 of two blocks before crossjumping will be performed on them. This
5138 value is ignored in the case where all instructions in the block being
5139 crossjumped from are matched. The default value is 5.
5141 @item max-delay-slot-insn-search
5142 The maximum number of instructions to consider when looking for an
5143 instruction to fill a delay slot. If more than this arbitrary number of
5144 instructions is searched, the time savings from filling the delay slot
5145 will be minimal so stop searching. Increasing values mean more
5146 aggressive optimization, making the compile time increase with probably
5147 small improvement in executable run time.
5149 @item max-delay-slot-live-search
5150 When trying to fill delay slots, the maximum number of instructions to
5151 consider when searching for a block with valid live register
5152 information. Increasing this arbitrarily chosen value means more
5153 aggressive optimization, increasing the compile time. This parameter
5154 should be removed when the delay slot code is rewritten to maintain the
5157 @item max-gcse-memory
5158 The approximate maximum amount of memory that will be allocated in
5159 order to perform the global common subexpression elimination
5160 optimization. If more memory than specified is required, the
5161 optimization will not be done.
5163 @item max-gcse-passes
5164 The maximum number of passes of GCSE to run. The default is 1.
5166 @item max-pending-list-length
5167 The maximum number of pending dependencies scheduling will allow
5168 before flushing the current state and starting over. Large functions
5169 with few branches or calls can create excessively large lists which
5170 needlessly consume memory and resources.
5172 @item max-inline-insns-single
5173 Several parameters control the tree inliner used in gcc.
5174 This number sets the maximum number of instructions (counted in GCC's
5175 internal representation) in a single function that the tree inliner
5176 will consider for inlining. This only affects functions declared
5177 inline and methods implemented in a class declaration (C++).
5178 The default value is 500.
5180 @item max-inline-insns-auto
5181 When you use @option{-finline-functions} (included in @option{-O3}),
5182 a lot of functions that would otherwise not be considered for inlining
5183 by the compiler will be investigated. To those functions, a different
5184 (more restrictive) limit compared to functions declared inline can
5186 The default value is 120.
5188 @item large-function-insns
5189 The limit specifying really large functions. For functions greater than this
5190 limit inlining is constrained by @option{--param large-function-growth}.
5191 This parameter is useful primarily to avoid extreme compilation time caused by non-linear
5192 algorithms used by the backend.
5193 This parameter is ignored when @option{-funit-at-a-time} is not used.
5194 The default value is 3000.
5196 @item large-function-growth
5197 Specifies maximal growth of large function caused by inlining in percents.
5198 This parameter is ignored when @option{-funit-at-a-time} is not used.
5199 The default value is 200.
5201 @item inline-unit-growth
5202 Specifies maximal overall growth of the compilation unit caused by inlining.
5203 This parameter is ignored when @option{-funit-at-a-time} is not used.
5204 The default value is 150.
5206 @item max-inline-insns-recursive
5207 @itemx max-inline-insns-recursive-auto
5208 Specifies maximum number of instructions out-of-line copy of self recursive inline
5209 function can grow into by performing recursive inlining.
5211 For functions declared inline @option{--param max-inline-insns-recursive} is
5212 taken into acount. For function not declared inline, recursive inlining
5213 happens only when @option{-finline-functions} (included in @option{-O3}) is
5214 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5215 default value is 500.
5217 @item max-inline-recursive-depth
5218 @itemx max-inline-recursive-depth-auto
5219 Specifies maximum recursion depth used by the recursive inlining.
5221 For functions declared inline @option{--param max-inline-recursive-depth} is
5222 taken into acount. For function not declared inline, recursive inlining
5223 happens only when @option{-finline-functions} (included in @option{-O3}) is
5224 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5225 default value is 500.
5227 @item max-inline-insns-rtl
5228 For languages that use the RTL inliner (this happens at a later stage
5229 than tree inlining), you can set the maximum allowable size (counted
5230 in RTL instructions) for the RTL inliner with this parameter.
5231 The default value is 600.
5233 @item max-unrolled-insns
5234 The maximum number of instructions that a loop should have if that loop
5235 is unrolled, and if the loop is unrolled, it determines how many times
5236 the loop code is unrolled.
5238 @item max-average-unrolled-insns
5239 The maximum number of instructions biased by probabilities of their execution
5240 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5241 it determines how many times the loop code is unrolled.
5243 @item max-unroll-times
5244 The maximum number of unrollings of a single loop.
5246 @item max-peeled-insns
5247 The maximum number of instructions that a loop should have if that loop
5248 is peeled, and if the loop is peeled, it determines how many times
5249 the loop code is peeled.
5251 @item max-peel-times
5252 The maximum number of peelings of a single loop.
5254 @item max-completely-peeled-insns
5255 The maximum number of insns of a completely peeled loop.
5257 @item max-completely-peel-times
5258 The maximum number of iterations of a loop to be suitable for complete peeling.
5260 @item max-unswitch-insns
5261 The maximum number of insns of an unswitched loop.
5263 @item max-unswitch-level
5264 The maximum number of branches unswitched in a single loop.
5267 The minimum cost of an expensive expression in the loop invariant motion.
5269 @item max-iterations-to-track
5271 The maximum number of iterations of a loop the brute force algorithm
5272 for analysis of # of iterations of the loop tries to evaluate.
5274 @item hot-bb-count-fraction
5275 Select fraction of the maximal count of repetitions of basic block in program
5276 given basic block needs to have to be considered hot.
5278 @item hot-bb-frequency-fraction
5279 Select fraction of the maximal frequency of executions of basic block in
5280 function given basic block needs to have to be considered hot
5282 @item tracer-dynamic-coverage
5283 @itemx tracer-dynamic-coverage-feedback
5285 This value is used to limit superblock formation once the given percentage of
5286 executed instructions is covered. This limits unnecessary code size
5289 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5290 feedback is available. The real profiles (as opposed to statically estimated
5291 ones) are much less balanced allowing the threshold to be larger value.
5293 @item tracer-max-code-growth
5294 Stop tail duplication once code growth has reached given percentage. This is
5295 rather hokey argument, as most of the duplicates will be eliminated later in
5296 cross jumping, so it may be set to much higher values than is the desired code
5299 @item tracer-min-branch-ratio
5301 Stop reverse growth when the reverse probability of best edge is less than this
5302 threshold (in percent).
5304 @item tracer-min-branch-ratio
5305 @itemx tracer-min-branch-ratio-feedback
5307 Stop forward growth if the best edge do have probability lower than this
5310 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
5311 compilation for profile feedback and one for compilation without. The value
5312 for compilation with profile feedback needs to be more conservative (higher) in
5313 order to make tracer effective.
5315 @item max-cse-path-length
5317 Maximum number of basic blocks on path that cse considers. The default is 10.
5319 @item global-var-threshold
5321 Counts the number of function calls (N) and the number of
5322 call-clobbered variables (V). If NxV is larger than this limit, a
5323 single artificial variable will be created to represent all the
5324 call-clobbered variables at function call sites. This artificial
5325 variable will then be made to alias every call-clobbered variable.
5326 (done as int * size_t on the host machine; beware overflow).
5328 @item max-aliased-vops
5330 Maxiumum number of virtual operands allowed to represent aliases
5331 before triggering the alias grouping heuristic. Alias grouping
5332 reduces compile times and memory consumption needed for aliasing at
5333 the expense of precision loss in alias information.
5335 @item ggc-min-expand
5337 GCC uses a garbage collector to manage its own memory allocation. This
5338 parameter specifies the minimum percentage by which the garbage
5339 collector's heap should be allowed to expand between collections.
5340 Tuning this may improve compilation speed; it has no effect on code
5343 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
5344 RAM >= 1GB. If @code{getrlimit} is available, the notion of "RAM" is
5345 the smallest of actual RAM and RLIMIT_DATA or RLIMIT_AS. If
5346 GCC is not able to calculate RAM on a particular platform, the lower
5347 bound of 30% is used. Setting this parameter and
5348 @option{ggc-min-heapsize} to zero causes a full collection to occur at
5349 every opportunity. This is extremely slow, but can be useful for
5352 @item ggc-min-heapsize
5354 Minimum size of the garbage collector's heap before it begins bothering
5355 to collect garbage. The first collection occurs after the heap expands
5356 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
5357 tuning this may improve compilation speed, and has no effect on code
5360 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
5361 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
5362 with a lower bound of 4096 (four megabytes) and an upper bound of
5363 131072 (128 megabytes). If GCC is not able to calculate RAM on a
5364 particular platform, the lower bound is used. Setting this parameter
5365 very large effectively disables garbage collection. Setting this
5366 parameter and @option{ggc-min-expand} to zero causes a full collection
5367 to occur at every opportunity.
5369 @item max-reload-search-insns
5370 The maximum number of instruction reload should look backward for equivalent
5371 register. Increasing values mean more aggressive optimization, making the
5372 compile time increase with probably slightly better performance. The default
5375 @item max-cselib-memory-location
5376 The maximum number of memory locations cselib should take into acount.
5377 Increasing values mean more aggressive optimization, making the compile time
5378 increase with probably slightly better performance. The default value is 500.
5380 @item reorder-blocks-duplicate
5381 @itemx reorder-blocks-duplicate-feedback
5383 Used by basic block reordering pass to decide whether to use unconditional
5384 branch or duplicate the code on its destination. Code is duplicated when its
5385 estimated size is smaller than this value multiplied by the estimated size of
5386 unconditional jump in the hot spots of the program.
5388 The @option{reorder-block-duplicate-feedback} is used only when profile
5389 feedback is available and may be set to higher values than
5390 @option{reorder-block-duplicate} since information about the hot spots is more
5393 @item max-sched-region-blocks
5394 The maximum number of blocks in a region to be considered for
5395 interblock scheduling. The default value is 10.
5397 @item max-sched-region-insns
5398 The maximum number of insns in a region to be considered for
5399 interblock scheduling. The default value is 100.
5403 @node Preprocessor Options
5404 @section Options Controlling the Preprocessor
5405 @cindex preprocessor options
5406 @cindex options, preprocessor
5408 These options control the C preprocessor, which is run on each C source
5409 file before actual compilation.
5411 If you use the @option{-E} option, nothing is done except preprocessing.
5412 Some of these options make sense only together with @option{-E} because
5413 they cause the preprocessor output to be unsuitable for actual
5418 You can use @option{-Wp,@var{option}} to bypass the compiler driver
5419 and pass @var{option} directly through to the preprocessor. If
5420 @var{option} contains commas, it is split into multiple options at the
5421 commas. However, many options are modified, translated or interpreted
5422 by the compiler driver before being passed to the preprocessor, and
5423 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
5424 interface is undocumented and subject to change, so whenever possible
5425 you should avoid using @option{-Wp} and let the driver handle the
5428 @item -Xpreprocessor @var{option}
5429 @opindex preprocessor
5430 Pass @var{option} as an option to the preprocessor. You can use this to
5431 supply system-specific preprocessor options which GCC does not know how to
5434 If you want to pass an option that takes an argument, you must use
5435 @option{-Xpreprocessor} twice, once for the option and once for the argument.
5438 @include cppopts.texi
5440 @node Assembler Options
5441 @section Passing Options to the Assembler
5443 @c prevent bad page break with this line
5444 You can pass options to the assembler.
5447 @item -Wa,@var{option}
5449 Pass @var{option} as an option to the assembler. If @var{option}
5450 contains commas, it is split into multiple options at the commas.
5452 @item -Xassembler @var{option}
5454 Pass @var{option} as an option to the assembler. You can use this to
5455 supply system-specific assembler options which GCC does not know how to
5458 If you want to pass an option that takes an argument, you must use
5459 @option{-Xassembler} twice, once for the option and once for the argument.
5464 @section Options for Linking
5465 @cindex link options
5466 @cindex options, linking
5468 These options come into play when the compiler links object files into
5469 an executable output file. They are meaningless if the compiler is
5470 not doing a link step.
5474 @item @var{object-file-name}
5475 A file name that does not end in a special recognized suffix is
5476 considered to name an object file or library. (Object files are
5477 distinguished from libraries by the linker according to the file
5478 contents.) If linking is done, these object files are used as input
5487 If any of these options is used, then the linker is not run, and
5488 object file names should not be used as arguments. @xref{Overall
5492 @item -l@var{library}
5493 @itemx -l @var{library}
5495 Search the library named @var{library} when linking. (The second
5496 alternative with the library as a separate argument is only for
5497 POSIX compliance and is not recommended.)
5499 It makes a difference where in the command you write this option; the
5500 linker searches and processes libraries and object files in the order they
5501 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
5502 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
5503 to functions in @samp{z}, those functions may not be loaded.
5505 The linker searches a standard list of directories for the library,
5506 which is actually a file named @file{lib@var{library}.a}. The linker
5507 then uses this file as if it had been specified precisely by name.
5509 The directories searched include several standard system directories
5510 plus any that you specify with @option{-L}.
5512 Normally the files found this way are library files---archive files
5513 whose members are object files. The linker handles an archive file by
5514 scanning through it for members which define symbols that have so far
5515 been referenced but not defined. But if the file that is found is an
5516 ordinary object file, it is linked in the usual fashion. The only
5517 difference between using an @option{-l} option and specifying a file name
5518 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
5519 and searches several directories.
5523 You need this special case of the @option{-l} option in order to
5524 link an Objective-C or Objective-C++ program.
5527 @opindex nostartfiles
5528 Do not use the standard system startup files when linking.
5529 The standard system libraries are used normally, unless @option{-nostdlib}
5530 or @option{-nodefaultlibs} is used.
5532 @item -nodefaultlibs
5533 @opindex nodefaultlibs
5534 Do not use the standard system libraries when linking.
5535 Only the libraries you specify will be passed to the linker.
5536 The standard startup files are used normally, unless @option{-nostartfiles}
5537 is used. The compiler may generate calls to @code{memcmp},
5538 @code{memset}, @code{memcpy} and @code{memmove}.
5539 These entries are usually resolved by entries in
5540 libc. These entry points should be supplied through some other
5541 mechanism when this option is specified.
5545 Do not use the standard system startup files or libraries when linking.
5546 No startup files and only the libraries you specify will be passed to
5547 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
5548 @code{memcpy} and @code{memmove}.
5549 These entries are usually resolved by entries in
5550 libc. These entry points should be supplied through some other
5551 mechanism when this option is specified.
5553 @cindex @option{-lgcc}, use with @option{-nostdlib}
5554 @cindex @option{-nostdlib} and unresolved references
5555 @cindex unresolved references and @option{-nostdlib}
5556 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
5557 @cindex @option{-nodefaultlibs} and unresolved references
5558 @cindex unresolved references and @option{-nodefaultlibs}
5559 One of the standard libraries bypassed by @option{-nostdlib} and
5560 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
5561 that GCC uses to overcome shortcomings of particular machines, or special
5562 needs for some languages.
5563 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
5564 Collection (GCC) Internals},
5565 for more discussion of @file{libgcc.a}.)
5566 In most cases, you need @file{libgcc.a} even when you want to avoid
5567 other standard libraries. In other words, when you specify @option{-nostdlib}
5568 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
5569 This ensures that you have no unresolved references to internal GCC
5570 library subroutines. (For example, @samp{__main}, used to ensure C++
5571 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
5572 GNU Compiler Collection (GCC) Internals}.)
5576 Produce a position independent executable on targets which support it.
5577 For predictable results, you must also specify the same set of options
5578 that were used to generate code (@option{-fpie}, @option{-fPIE},
5579 or model suboptions) when you specify this option.
5583 Remove all symbol table and relocation information from the executable.
5587 On systems that support dynamic linking, this prevents linking with the shared
5588 libraries. On other systems, this option has no effect.
5592 Produce a shared object which can then be linked with other objects to
5593 form an executable. Not all systems support this option. For predictable
5594 results, you must also specify the same set of options that were used to
5595 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
5596 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
5597 needs to build supplementary stub code for constructors to work. On
5598 multi-libbed systems, @samp{gcc -shared} must select the correct support
5599 libraries to link against. Failing to supply the correct flags may lead
5600 to subtle defects. Supplying them in cases where they are not necessary
5603 @item -shared-libgcc
5604 @itemx -static-libgcc
5605 @opindex shared-libgcc
5606 @opindex static-libgcc
5607 On systems that provide @file{libgcc} as a shared library, these options
5608 force the use of either the shared or static version respectively.
5609 If no shared version of @file{libgcc} was built when the compiler was
5610 configured, these options have no effect.
5612 There are several situations in which an application should use the
5613 shared @file{libgcc} instead of the static version. The most common
5614 of these is when the application wishes to throw and catch exceptions
5615 across different shared libraries. In that case, each of the libraries
5616 as well as the application itself should use the shared @file{libgcc}.
5618 Therefore, the G++ and GCJ drivers automatically add
5619 @option{-shared-libgcc} whenever you build a shared library or a main
5620 executable, because C++ and Java programs typically use exceptions, so
5621 this is the right thing to do.
5623 If, instead, you use the GCC driver to create shared libraries, you may
5624 find that they will not always be linked with the shared @file{libgcc}.
5625 If GCC finds, at its configuration time, that you have a non-GNU linker
5626 or a GNU linker that does not support option @option{--eh-frame-hdr},
5627 it will link the shared version of @file{libgcc} into shared libraries
5628 by default. Otherwise, it will take advantage of the linker and optimize
5629 away the linking with the shared version of @file{libgcc}, linking with
5630 the static version of libgcc by default. This allows exceptions to
5631 propagate through such shared libraries, without incurring relocation
5632 costs at library load time.
5634 However, if a library or main executable is supposed to throw or catch
5635 exceptions, you must link it using the G++ or GCJ driver, as appropriate
5636 for the languages used in the program, or using the option
5637 @option{-shared-libgcc}, such that it is linked with the shared
5642 Bind references to global symbols when building a shared object. Warn
5643 about any unresolved references (unless overridden by the link editor
5644 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
5647 @item -Xlinker @var{option}
5649 Pass @var{option} as an option to the linker. You can use this to
5650 supply system-specific linker options which GCC does not know how to
5653 If you want to pass an option that takes an argument, you must use
5654 @option{-Xlinker} twice, once for the option and once for the argument.
5655 For example, to pass @option{-assert definitions}, you must write
5656 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
5657 @option{-Xlinker "-assert definitions"}, because this passes the entire
5658 string as a single argument, which is not what the linker expects.
5660 @item -Wl,@var{option}
5662 Pass @var{option} as an option to the linker. If @var{option} contains
5663 commas, it is split into multiple options at the commas.
5665 @item -u @var{symbol}
5667 Pretend the symbol @var{symbol} is undefined, to force linking of
5668 library modules to define it. You can use @option{-u} multiple times with
5669 different symbols to force loading of additional library modules.
5672 @node Directory Options
5673 @section Options for Directory Search
5674 @cindex directory options
5675 @cindex options, directory search
5678 These options specify directories to search for header files, for
5679 libraries and for parts of the compiler:
5684 Add the directory @var{dir} to the head of the list of directories to be
5685 searched for header files. This can be used to override a system header
5686 file, substituting your own version, since these directories are
5687 searched before the system header file directories. However, you should
5688 not use this option to add directories that contain vendor-supplied
5689 system header files (use @option{-isystem} for that). If you use more than
5690 one @option{-I} option, the directories are scanned in left-to-right
5691 order; the standard system directories come after.
5693 If a standard system include directory, or a directory specified with
5694 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
5695 option will be ignored. The directory will still be searched but as a
5696 system directory at its normal position in the system include chain.
5697 This is to ensure that GCC's procedure to fix buggy system headers and
5698 the ordering for the include_next directive are not inadvertently changed.
5699 If you really need to change the search order for system directories,
5700 use the @option{-nostdinc} and/or @option{-isystem} options.
5702 @item -iquote@var{dir}
5704 Add the directory @var{dir} to the head of the list of directories to
5705 be searched for header files only for the case of @samp{#include
5706 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
5707 otherwise just like @option{-I}.
5711 Add directory @var{dir} to the list of directories to be searched
5714 @item -B@var{prefix}
5716 This option specifies where to find the executables, libraries,
5717 include files, and data files of the compiler itself.
5719 The compiler driver program runs one or more of the subprograms
5720 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
5721 @var{prefix} as a prefix for each program it tries to run, both with and
5722 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
5724 For each subprogram to be run, the compiler driver first tries the
5725 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
5726 was not specified, the driver tries two standard prefixes, which are
5727 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
5728 those results in a file name that is found, the unmodified program
5729 name is searched for using the directories specified in your
5730 @env{PATH} environment variable.
5732 The compiler will check to see if the path provided by the @option{-B}
5733 refers to a directory, and if necessary it will add a directory
5734 separator character at the end of the path.
5736 @option{-B} prefixes that effectively specify directory names also apply
5737 to libraries in the linker, because the compiler translates these
5738 options into @option{-L} options for the linker. They also apply to
5739 includes files in the preprocessor, because the compiler translates these
5740 options into @option{-isystem} options for the preprocessor. In this case,
5741 the compiler appends @samp{include} to the prefix.
5743 The run-time support file @file{libgcc.a} can also be searched for using
5744 the @option{-B} prefix, if needed. If it is not found there, the two
5745 standard prefixes above are tried, and that is all. The file is left
5746 out of the link if it is not found by those means.
5748 Another way to specify a prefix much like the @option{-B} prefix is to use
5749 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
5752 As a special kludge, if the path provided by @option{-B} is
5753 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
5754 9, then it will be replaced by @file{[dir/]include}. This is to help
5755 with boot-strapping the compiler.
5757 @item -specs=@var{file}
5759 Process @var{file} after the compiler reads in the standard @file{specs}
5760 file, in order to override the defaults that the @file{gcc} driver
5761 program uses when determining what switches to pass to @file{cc1},
5762 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
5763 @option{-specs=@var{file}} can be specified on the command line, and they
5764 are processed in order, from left to right.
5768 This option has been deprecated. Please use @option{-iquote} instead for
5769 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
5770 Any directories you specify with @option{-I} options before the @option{-I-}
5771 option are searched only for the case of @samp{#include "@var{file}"};
5772 they are not searched for @samp{#include <@var{file}>}.
5774 If additional directories are specified with @option{-I} options after
5775 the @option{-I-}, these directories are searched for all @samp{#include}
5776 directives. (Ordinarily @emph{all} @option{-I} directories are used
5779 In addition, the @option{-I-} option inhibits the use of the current
5780 directory (where the current input file came from) as the first search
5781 directory for @samp{#include "@var{file}"}. There is no way to
5782 override this effect of @option{-I-}. With @option{-I.} you can specify
5783 searching the directory which was current when the compiler was
5784 invoked. That is not exactly the same as what the preprocessor does
5785 by default, but it is often satisfactory.
5787 @option{-I-} does not inhibit the use of the standard system directories
5788 for header files. Thus, @option{-I-} and @option{-nostdinc} are
5795 @section Specifying subprocesses and the switches to pass to them
5798 @command{gcc} is a driver program. It performs its job by invoking a
5799 sequence of other programs to do the work of compiling, assembling and
5800 linking. GCC interprets its command-line parameters and uses these to
5801 deduce which programs it should invoke, and which command-line options
5802 it ought to place on their command lines. This behavior is controlled
5803 by @dfn{spec strings}. In most cases there is one spec string for each
5804 program that GCC can invoke, but a few programs have multiple spec
5805 strings to control their behavior. The spec strings built into GCC can
5806 be overridden by using the @option{-specs=} command-line switch to specify
5809 @dfn{Spec files} are plaintext files that are used to construct spec
5810 strings. They consist of a sequence of directives separated by blank
5811 lines. The type of directive is determined by the first non-whitespace
5812 character on the line and it can be one of the following:
5815 @item %@var{command}
5816 Issues a @var{command} to the spec file processor. The commands that can
5820 @item %include <@var{file}>
5822 Search for @var{file} and insert its text at the current point in the
5825 @item %include_noerr <@var{file}>
5826 @cindex %include_noerr
5827 Just like @samp{%include}, but do not generate an error message if the include
5828 file cannot be found.
5830 @item %rename @var{old_name} @var{new_name}
5832 Rename the spec string @var{old_name} to @var{new_name}.
5836 @item *[@var{spec_name}]:
5837 This tells the compiler to create, override or delete the named spec
5838 string. All lines after this directive up to the next directive or
5839 blank line are considered to be the text for the spec string. If this
5840 results in an empty string then the spec will be deleted. (Or, if the
5841 spec did not exist, then nothing will happened.) Otherwise, if the spec
5842 does not currently exist a new spec will be created. If the spec does
5843 exist then its contents will be overridden by the text of this
5844 directive, unless the first character of that text is the @samp{+}
5845 character, in which case the text will be appended to the spec.
5847 @item [@var{suffix}]:
5848 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
5849 and up to the next directive or blank line are considered to make up the
5850 spec string for the indicated suffix. When the compiler encounters an
5851 input file with the named suffix, it will processes the spec string in
5852 order to work out how to compile that file. For example:
5859 This says that any input file whose name ends in @samp{.ZZ} should be
5860 passed to the program @samp{z-compile}, which should be invoked with the
5861 command-line switch @option{-input} and with the result of performing the
5862 @samp{%i} substitution. (See below.)
5864 As an alternative to providing a spec string, the text that follows a
5865 suffix directive can be one of the following:
5868 @item @@@var{language}
5869 This says that the suffix is an alias for a known @var{language}. This is
5870 similar to using the @option{-x} command-line switch to GCC to specify a
5871 language explicitly. For example:
5878 Says that .ZZ files are, in fact, C++ source files.
5881 This causes an error messages saying:
5884 @var{name} compiler not installed on this system.
5888 GCC already has an extensive list of suffixes built into it.
5889 This directive will add an entry to the end of the list of suffixes, but
5890 since the list is searched from the end backwards, it is effectively
5891 possible to override earlier entries using this technique.
5895 GCC has the following spec strings built into it. Spec files can
5896 override these strings or create their own. Note that individual
5897 targets can also add their own spec strings to this list.
5900 asm Options to pass to the assembler
5901 asm_final Options to pass to the assembler post-processor
5902 cpp Options to pass to the C preprocessor
5903 cc1 Options to pass to the C compiler
5904 cc1plus Options to pass to the C++ compiler
5905 endfile Object files to include at the end of the link
5906 link Options to pass to the linker
5907 lib Libraries to include on the command line to the linker
5908 libgcc Decides which GCC support library to pass to the linker
5909 linker Sets the name of the linker
5910 predefines Defines to be passed to the C preprocessor
5911 signed_char Defines to pass to CPP to say whether @code{char} is signed
5913 startfile Object files to include at the start of the link
5916 Here is a small example of a spec file:
5922 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
5925 This example renames the spec called @samp{lib} to @samp{old_lib} and
5926 then overrides the previous definition of @samp{lib} with a new one.
5927 The new definition adds in some extra command-line options before
5928 including the text of the old definition.
5930 @dfn{Spec strings} are a list of command-line options to be passed to their
5931 corresponding program. In addition, the spec strings can contain
5932 @samp{%}-prefixed sequences to substitute variable text or to
5933 conditionally insert text into the command line. Using these constructs
5934 it is possible to generate quite complex command lines.
5936 Here is a table of all defined @samp{%}-sequences for spec
5937 strings. Note that spaces are not generated automatically around the
5938 results of expanding these sequences. Therefore you can concatenate them
5939 together or combine them with constant text in a single argument.
5943 Substitute one @samp{%} into the program name or argument.
5946 Substitute the name of the input file being processed.
5949 Substitute the basename of the input file being processed.
5950 This is the substring up to (and not including) the last period
5951 and not including the directory.
5954 This is the same as @samp{%b}, but include the file suffix (text after
5958 Marks the argument containing or following the @samp{%d} as a
5959 temporary file name, so that that file will be deleted if GCC exits
5960 successfully. Unlike @samp{%g}, this contributes no text to the
5963 @item %g@var{suffix}
5964 Substitute a file name that has suffix @var{suffix} and is chosen
5965 once per compilation, and mark the argument in the same way as
5966 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
5967 name is now chosen in a way that is hard to predict even when previously
5968 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
5969 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
5970 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
5971 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
5972 was simply substituted with a file name chosen once per compilation,
5973 without regard to any appended suffix (which was therefore treated
5974 just like ordinary text), making such attacks more likely to succeed.
5976 @item %u@var{suffix}
5977 Like @samp{%g}, but generates a new temporary file name even if
5978 @samp{%u@var{suffix}} was already seen.
5980 @item %U@var{suffix}
5981 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
5982 new one if there is no such last file name. In the absence of any
5983 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
5984 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
5985 would involve the generation of two distinct file names, one
5986 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
5987 simply substituted with a file name chosen for the previous @samp{%u},
5988 without regard to any appended suffix.
5990 @item %j@var{suffix}
5991 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
5992 writable, and if save-temps is off; otherwise, substitute the name
5993 of a temporary file, just like @samp{%u}. This temporary file is not
5994 meant for communication between processes, but rather as a junk
5997 @item %|@var{suffix}
5998 @itemx %m@var{suffix}
5999 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6000 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6001 all. These are the two most common ways to instruct a program that it
6002 should read from standard input or write to standard output. If you
6003 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6004 construct: see for example @file{f/lang-specs.h}.
6006 @item %.@var{SUFFIX}
6007 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6008 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6009 terminated by the next space or %.
6012 Marks the argument containing or following the @samp{%w} as the
6013 designated output file of this compilation. This puts the argument
6014 into the sequence of arguments that @samp{%o} will substitute later.
6017 Substitutes the names of all the output files, with spaces
6018 automatically placed around them. You should write spaces
6019 around the @samp{%o} as well or the results are undefined.
6020 @samp{%o} is for use in the specs for running the linker.
6021 Input files whose names have no recognized suffix are not compiled
6022 at all, but they are included among the output files, so they will
6026 Substitutes the suffix for object files. Note that this is
6027 handled specially when it immediately follows @samp{%g, %u, or %U},
6028 because of the need for those to form complete file names. The
6029 handling is such that @samp{%O} is treated exactly as if it had already
6030 been substituted, except that @samp{%g, %u, and %U} do not currently
6031 support additional @var{suffix} characters following @samp{%O} as they would
6032 following, for example, @samp{.o}.
6035 Substitutes the standard macro predefinitions for the
6036 current target machine. Use this when running @code{cpp}.
6039 Like @samp{%p}, but puts @samp{__} before and after the name of each
6040 predefined macro, except for macros that start with @samp{__} or with
6041 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6045 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6046 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
6047 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6051 Current argument is the name of a library or startup file of some sort.
6052 Search for that file in a standard list of directories and substitute
6053 the full name found.
6056 Print @var{str} as an error message. @var{str} is terminated by a newline.
6057 Use this when inconsistent options are detected.
6060 Substitute the contents of spec string @var{name} at this point.
6063 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6065 @item %x@{@var{option}@}
6066 Accumulate an option for @samp{%X}.
6069 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6073 Output the accumulated assembler options specified by @option{-Wa}.
6076 Output the accumulated preprocessor options specified by @option{-Wp}.
6079 Process the @code{asm} spec. This is used to compute the
6080 switches to be passed to the assembler.
6083 Process the @code{asm_final} spec. This is a spec string for
6084 passing switches to an assembler post-processor, if such a program is
6088 Process the @code{link} spec. This is the spec for computing the
6089 command line passed to the linker. Typically it will make use of the
6090 @samp{%L %G %S %D and %E} sequences.
6093 Dump out a @option{-L} option for each directory that GCC believes might
6094 contain startup files. If the target supports multilibs then the
6095 current multilib directory will be prepended to each of these paths.
6098 Output the multilib directory with directory separators replaced with
6099 @samp{_}. If multilib directories are not set, or the multilib directory is
6100 @file{.} then this option emits nothing.
6103 Process the @code{lib} spec. This is a spec string for deciding which
6104 libraries should be included on the command line to the linker.
6107 Process the @code{libgcc} spec. This is a spec string for deciding
6108 which GCC support library should be included on the command line to the linker.
6111 Process the @code{startfile} spec. This is a spec for deciding which
6112 object files should be the first ones passed to the linker. Typically
6113 this might be a file named @file{crt0.o}.
6116 Process the @code{endfile} spec. This is a spec string that specifies
6117 the last object files that will be passed to the linker.
6120 Process the @code{cpp} spec. This is used to construct the arguments
6121 to be passed to the C preprocessor.
6124 Process the @code{cc1} spec. This is used to construct the options to be
6125 passed to the actual C compiler (@samp{cc1}).
6128 Process the @code{cc1plus} spec. This is used to construct the options to be
6129 passed to the actual C++ compiler (@samp{cc1plus}).
6132 Substitute the variable part of a matched option. See below.
6133 Note that each comma in the substituted string is replaced by
6137 Remove all occurrences of @code{-S} from the command line. Note---this
6138 command is position dependent. @samp{%} commands in the spec string
6139 before this one will see @code{-S}, @samp{%} commands in the spec string
6140 after this one will not.
6142 @item %:@var{function}(@var{args})
6143 Call the named function @var{function}, passing it @var{args}.
6144 @var{args} is first processed as a nested spec string, then split
6145 into an argument vector in the usual fashion. The function returns
6146 a string which is processed as if it had appeared literally as part
6147 of the current spec.
6149 The following built-in spec functions are provided:
6152 @item @code{if-exists}
6153 The @code{if-exists} spec function takes one argument, an absolute
6154 pathname to a file. If the file exists, @code{if-exists} returns the
6155 pathname. Here is a small example of its usage:
6159 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6162 @item @code{if-exists-else}
6163 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6164 spec function, except that it takes two arguments. The first argument is
6165 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6166 returns the pathname. If it does not exist, it returns the second argument.
6167 This way, @code{if-exists-else} can be used to select one file or another,
6168 based on the existence of the first. Here is a small example of its usage:
6172 crt0%O%s %:if-exists(crti%O%s) \
6173 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6178 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6179 If that switch was not specified, this substitutes nothing. Note that
6180 the leading dash is omitted when specifying this option, and it is
6181 automatically inserted if the substitution is performed. Thus the spec
6182 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6183 and would output the command line option @option{-foo}.
6185 @item %W@{@code{S}@}
6186 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6189 @item %@{@code{S}*@}
6190 Substitutes all the switches specified to GCC whose names start
6191 with @code{-S}, but which also take an argument. This is used for
6192 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6193 GCC considers @option{-o foo} as being
6194 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6195 text, including the space. Thus two arguments would be generated.
6197 @item %@{@code{S}*&@code{T}*@}
6198 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6199 (the order of @code{S} and @code{T} in the spec is not significant).
6200 There can be any number of ampersand-separated variables; for each the
6201 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6203 @item %@{@code{S}:@code{X}@}
6204 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6206 @item %@{!@code{S}:@code{X}@}
6207 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6209 @item %@{@code{S}*:@code{X}@}
6210 Substitutes @code{X} if one or more switches whose names start with
6211 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6212 once, no matter how many such switches appeared. However, if @code{%*}
6213 appears somewhere in @code{X}, then @code{X} will be substituted once
6214 for each matching switch, with the @code{%*} replaced by the part of
6215 that switch that matched the @code{*}.
6217 @item %@{.@code{S}:@code{X}@}
6218 Substitutes @code{X}, if processing a file with suffix @code{S}.
6220 @item %@{!.@code{S}:@code{X}@}
6221 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6223 @item %@{@code{S}|@code{P}:@code{X}@}
6224 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6225 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
6226 although they have a stronger binding than the @samp{|}. If @code{%*}
6227 appears in @code{X}, all of the alternatives must be starred, and only
6228 the first matching alternative is substituted.
6230 For example, a spec string like this:
6233 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
6236 will output the following command-line options from the following input
6237 command-line options:
6242 -d fred.c -foo -baz -boggle
6243 -d jim.d -bar -baz -boggle
6246 @item %@{S:X; T:Y; :D@}
6248 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
6249 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
6250 be as many clauses as you need. This may be combined with @code{.},
6251 @code{!}, @code{|}, and @code{*} as needed.
6256 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
6257 construct may contain other nested @samp{%} constructs or spaces, or
6258 even newlines. They are processed as usual, as described above.
6259 Trailing white space in @code{X} is ignored. White space may also
6260 appear anywhere on the left side of the colon in these constructs,
6261 except between @code{.} or @code{*} and the corresponding word.
6263 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
6264 handled specifically in these constructs. If another value of
6265 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
6266 @option{-W} switch is found later in the command line, the earlier
6267 switch value is ignored, except with @{@code{S}*@} where @code{S} is
6268 just one letter, which passes all matching options.
6270 The character @samp{|} at the beginning of the predicate text is used to
6271 indicate that a command should be piped to the following command, but
6272 only if @option{-pipe} is specified.
6274 It is built into GCC which switches take arguments and which do not.
6275 (You might think it would be useful to generalize this to allow each
6276 compiler's spec to say which switches take arguments. But this cannot
6277 be done in a consistent fashion. GCC cannot even decide which input
6278 files have been specified without knowing which switches take arguments,
6279 and it must know which input files to compile in order to tell which
6282 GCC also knows implicitly that arguments starting in @option{-l} are to be
6283 treated as compiler output files, and passed to the linker in their
6284 proper position among the other output files.
6286 @c man begin OPTIONS
6288 @node Target Options
6289 @section Specifying Target Machine and Compiler Version
6290 @cindex target options
6291 @cindex cross compiling
6292 @cindex specifying machine version
6293 @cindex specifying compiler version and target machine
6294 @cindex compiler version, specifying
6295 @cindex target machine, specifying
6297 The usual way to run GCC is to run the executable called @file{gcc}, or
6298 @file{<machine>-gcc} when cross-compiling, or
6299 @file{<machine>-gcc-<version>} to run a version other than the one that
6300 was installed last. Sometimes this is inconvenient, so GCC provides
6301 options that will switch to another cross-compiler or version.
6304 @item -b @var{machine}
6306 The argument @var{machine} specifies the target machine for compilation.
6308 The value to use for @var{machine} is the same as was specified as the
6309 machine type when configuring GCC as a cross-compiler. For
6310 example, if a cross-compiler was configured with @samp{configure
6311 i386v}, meaning to compile for an 80386 running System V, then you
6312 would specify @option{-b i386v} to run that cross compiler.
6314 @item -V @var{version}
6316 The argument @var{version} specifies which version of GCC to run.
6317 This is useful when multiple versions are installed. For example,
6318 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
6321 The @option{-V} and @option{-b} options work by running the
6322 @file{<machine>-gcc-<version>} executable, so there's no real reason to
6323 use them if you can just run that directly.
6325 @node Submodel Options
6326 @section Hardware Models and Configurations
6327 @cindex submodel options
6328 @cindex specifying hardware config
6329 @cindex hardware models and configurations, specifying
6330 @cindex machine dependent options
6332 Earlier we discussed the standard option @option{-b} which chooses among
6333 different installed compilers for completely different target
6334 machines, such as VAX vs.@: 68000 vs.@: 80386.
6336 In addition, each of these target machine types can have its own
6337 special options, starting with @samp{-m}, to choose among various
6338 hardware models or configurations---for example, 68010 vs 68020,
6339 floating coprocessor or none. A single installed version of the
6340 compiler can compile for any model or configuration, according to the
6343 Some configurations of the compiler also support additional special
6344 options, usually for compatibility with other compilers on the same
6347 These options are defined by the macro @code{TARGET_SWITCHES} in the
6348 machine description. The default for the options is also defined by
6349 that macro, which enables you to change the defaults.
6351 @c This list is ordered alphanumerically by subsection name.
6352 @c It should be the same order and spelling as these options are listed
6353 @c in Machine Dependent Options
6361 * DEC Alpha Options::
6362 * DEC Alpha/VMS Options::
6366 * i386 and x86-64 Options::
6378 * RS/6000 and PowerPC Options::
6379 * S/390 and zSeries Options::
6382 * System V Options::
6383 * TMS320C3x/C4x Options::
6387 * Xstormy16 Options::
6393 @subsection ARC Options
6396 These options are defined for ARC implementations:
6401 Compile code for little endian mode. This is the default.
6405 Compile code for big endian mode.
6408 @opindex mmangle-cpu
6409 Prepend the name of the cpu to all public symbol names.
6410 In multiple-processor systems, there are many ARC variants with different
6411 instruction and register set characteristics. This flag prevents code
6412 compiled for one cpu to be linked with code compiled for another.
6413 No facility exists for handling variants that are ``almost identical''.
6414 This is an all or nothing option.
6416 @item -mcpu=@var{cpu}
6418 Compile code for ARC variant @var{cpu}.
6419 Which variants are supported depend on the configuration.
6420 All variants support @option{-mcpu=base}, this is the default.
6422 @item -mtext=@var{text-section}
6423 @itemx -mdata=@var{data-section}
6424 @itemx -mrodata=@var{readonly-data-section}
6428 Put functions, data, and readonly data in @var{text-section},
6429 @var{data-section}, and @var{readonly-data-section} respectively
6430 by default. This can be overridden with the @code{section} attribute.
6431 @xref{Variable Attributes}.
6436 @subsection ARM Options
6439 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6443 @item -mabi=@var{name}
6445 Generate code for the specified ABI. Permissible values are: @samp{apcs-gnu},
6446 @samp{atpcs}, @samp{aapcs} and @samp{iwmmxt}.
6449 @opindex mapcs-frame
6450 Generate a stack frame that is compliant with the ARM Procedure Call
6451 Standard for all functions, even if this is not strictly necessary for
6452 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6453 with this option will cause the stack frames not to be generated for
6454 leaf functions. The default is @option{-mno-apcs-frame}.
6458 This is a synonym for @option{-mapcs-frame}.
6461 @c not currently implemented
6462 @item -mapcs-stack-check
6463 @opindex mapcs-stack-check
6464 Generate code to check the amount of stack space available upon entry to
6465 every function (that actually uses some stack space). If there is
6466 insufficient space available then either the function
6467 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
6468 called, depending upon the amount of stack space required. The run time
6469 system is required to provide these functions. The default is
6470 @option{-mno-apcs-stack-check}, since this produces smaller code.
6472 @c not currently implemented
6474 @opindex mapcs-float
6475 Pass floating point arguments using the float point registers. This is
6476 one of the variants of the APCS@. This option is recommended if the
6477 target hardware has a floating point unit or if a lot of floating point
6478 arithmetic is going to be performed by the code. The default is
6479 @option{-mno-apcs-float}, since integer only code is slightly increased in
6480 size if @option{-mapcs-float} is used.
6482 @c not currently implemented
6483 @item -mapcs-reentrant
6484 @opindex mapcs-reentrant
6485 Generate reentrant, position independent code. The default is
6486 @option{-mno-apcs-reentrant}.
6489 @item -mthumb-interwork
6490 @opindex mthumb-interwork
6491 Generate code which supports calling between the ARM and Thumb
6492 instruction sets. Without this option the two instruction sets cannot
6493 be reliably used inside one program. The default is
6494 @option{-mno-thumb-interwork}, since slightly larger code is generated
6495 when @option{-mthumb-interwork} is specified.
6497 @item -mno-sched-prolog
6498 @opindex mno-sched-prolog
6499 Prevent the reordering of instructions in the function prolog, or the
6500 merging of those instruction with the instructions in the function's
6501 body. This means that all functions will start with a recognizable set
6502 of instructions (or in fact one of a choice from a small set of
6503 different function prologues), and this information can be used to
6504 locate the start if functions inside an executable piece of code. The
6505 default is @option{-msched-prolog}.
6508 @opindex mhard-float
6509 Generate output containing floating point instructions. This is the
6513 @opindex msoft-float
6514 Generate output containing library calls for floating point.
6515 @strong{Warning:} the requisite libraries are not available for all ARM
6516 targets. Normally the facilities of the machine's usual C compiler are
6517 used, but this cannot be done directly in cross-compilation. You must make
6518 your own arrangements to provide suitable library functions for
6521 @option{-msoft-float} changes the calling convention in the output file;
6522 therefore, it is only useful if you compile @emph{all} of a program with
6523 this option. In particular, you need to compile @file{libgcc.a}, the
6524 library that comes with GCC, with @option{-msoft-float} in order for
6527 @item -mfloat-abi=@var{name}
6529 Specifies which ABI to use for floating point values. Permissible values
6530 are: @samp{soft}, @samp{softfp} and @samp{hard}.
6532 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
6533 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
6534 of floating point instructions, but still uses the soft-float calling
6537 @item -mlittle-endian
6538 @opindex mlittle-endian
6539 Generate code for a processor running in little-endian mode. This is
6540 the default for all standard configurations.
6543 @opindex mbig-endian
6544 Generate code for a processor running in big-endian mode; the default is
6545 to compile code for a little-endian processor.
6547 @item -mwords-little-endian
6548 @opindex mwords-little-endian
6549 This option only applies when generating code for big-endian processors.
6550 Generate code for a little-endian word order but a big-endian byte
6551 order. That is, a byte order of the form @samp{32107654}. Note: this
6552 option should only be used if you require compatibility with code for
6553 big-endian ARM processors generated by versions of the compiler prior to
6556 @item -mcpu=@var{name}
6558 This specifies the name of the target ARM processor. GCC uses this name
6559 to determine what kind of instructions it can emit when generating
6560 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
6561 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
6562 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
6563 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
6564 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
6565 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm8},
6566 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
6567 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
6568 @samp{arm920t}, @samp{arm922t}, @samp{arm946es}, @samp{arm966es},
6569 @samp{arm968es}, @samp{arm926ejs}, @samp{arm940t}, @samp{arm9tdmi},
6570 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ejs},
6571 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
6572 @samp{arm1136js}, @samp{arm1136jfs} ,@samp{xscale}, @samp{iwmmxt},
6575 @itemx -mtune=@var{name}
6577 This option is very similar to the @option{-mcpu=} option, except that
6578 instead of specifying the actual target processor type, and hence
6579 restricting which instructions can be used, it specifies that GCC should
6580 tune the performance of the code as if the target were of the type
6581 specified in this option, but still choosing the instructions that it
6582 will generate based on the cpu specified by a @option{-mcpu=} option.
6583 For some ARM implementations better performance can be obtained by using
6586 @item -march=@var{name}
6588 This specifies the name of the target ARM architecture. GCC uses this
6589 name to determine what kind of instructions it can emit when generating
6590 assembly code. This option can be used in conjunction with or instead
6591 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
6592 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
6593 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
6594 @samp{iwmmxt}, @samp{ep9312}.
6596 @item -mfpu=@var{name}
6597 @itemx -mfpe=@var{number}
6598 @itemx -mfp=@var{number}
6602 This specifies what floating point hardware (or hardware emulation) is
6603 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
6604 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
6605 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
6606 with older versions of GCC@.
6608 If @option{-msoft-float} is specified this specifies the format of
6609 floating point values.
6611 @item -mstructure-size-boundary=@var{n}
6612 @opindex mstructure-size-boundary
6613 The size of all structures and unions will be rounded up to a multiple
6614 of the number of bits set by this option. Permissible values are 8, 32
6615 and 64. The default value varies for different toolchains. For the COFF
6616 targeted toolchain the default value is 8. A value of 64 is only allowed
6617 if the underlying ABI supports it.
6619 Specifying the larger number can produce faster, more efficient code, but
6620 can also increase the size of the program. Different values are potentially
6621 incompatible. Code compiled with one value cannot necessarily expect to
6622 work with code or libraries compiled with another value, if they exchange
6623 information using structures or unions.
6625 @item -mabort-on-noreturn
6626 @opindex mabort-on-noreturn
6627 Generate a call to the function @code{abort} at the end of a
6628 @code{noreturn} function. It will be executed if the function tries to
6632 @itemx -mno-long-calls
6633 @opindex mlong-calls
6634 @opindex mno-long-calls
6635 Tells the compiler to perform function calls by first loading the
6636 address of the function into a register and then performing a subroutine
6637 call on this register. This switch is needed if the target function
6638 will lie outside of the 64 megabyte addressing range of the offset based
6639 version of subroutine call instruction.
6641 Even if this switch is enabled, not all function calls will be turned
6642 into long calls. The heuristic is that static functions, functions
6643 which have the @samp{short-call} attribute, functions that are inside
6644 the scope of a @samp{#pragma no_long_calls} directive and functions whose
6645 definitions have already been compiled within the current compilation
6646 unit, will not be turned into long calls. The exception to this rule is
6647 that weak function definitions, functions with the @samp{long-call}
6648 attribute or the @samp{section} attribute, and functions that are within
6649 the scope of a @samp{#pragma long_calls} directive, will always be
6650 turned into long calls.
6652 This feature is not enabled by default. Specifying
6653 @option{-mno-long-calls} will restore the default behavior, as will
6654 placing the function calls within the scope of a @samp{#pragma
6655 long_calls_off} directive. Note these switches have no effect on how
6656 the compiler generates code to handle function calls via function
6659 @item -mnop-fun-dllimport
6660 @opindex mnop-fun-dllimport
6661 Disable support for the @code{dllimport} attribute.
6663 @item -msingle-pic-base
6664 @opindex msingle-pic-base
6665 Treat the register used for PIC addressing as read-only, rather than
6666 loading it in the prologue for each function. The run-time system is
6667 responsible for initializing this register with an appropriate value
6668 before execution begins.
6670 @item -mpic-register=@var{reg}
6671 @opindex mpic-register
6672 Specify the register to be used for PIC addressing. The default is R10
6673 unless stack-checking is enabled, when R9 is used.
6675 @item -mcirrus-fix-invalid-insns
6676 @opindex mcirrus-fix-invalid-insns
6677 @opindex mno-cirrus-fix-invalid-insns
6678 Insert NOPs into the instruction stream to in order to work around
6679 problems with invalid Maverick instruction combinations. This option
6680 is only valid if the @option{-mcpu=ep9312} option has been used to
6681 enable generation of instructions for the Cirrus Maverick floating
6682 point co-processor. This option is not enabled by default, since the
6683 problem is only present in older Maverick implementations. The default
6684 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
6687 @item -mpoke-function-name
6688 @opindex mpoke-function-name
6689 Write the name of each function into the text section, directly
6690 preceding the function prologue. The generated code is similar to this:
6694 .ascii "arm_poke_function_name", 0
6697 .word 0xff000000 + (t1 - t0)
6698 arm_poke_function_name
6700 stmfd sp!, @{fp, ip, lr, pc@}
6704 When performing a stack backtrace, code can inspect the value of
6705 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
6706 location @code{pc - 12} and the top 8 bits are set, then we know that
6707 there is a function name embedded immediately preceding this location
6708 and has length @code{((pc[-3]) & 0xff000000)}.
6712 Generate code for the 16-bit Thumb instruction set. The default is to
6713 use the 32-bit ARM instruction set.
6716 @opindex mtpcs-frame
6717 Generate a stack frame that is compliant with the Thumb Procedure Call
6718 Standard for all non-leaf functions. (A leaf function is one that does
6719 not call any other functions.) The default is @option{-mno-tpcs-frame}.
6721 @item -mtpcs-leaf-frame
6722 @opindex mtpcs-leaf-frame
6723 Generate a stack frame that is compliant with the Thumb Procedure Call
6724 Standard for all leaf functions. (A leaf function is one that does
6725 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
6727 @item -mcallee-super-interworking
6728 @opindex mcallee-super-interworking
6729 Gives all externally visible functions in the file being compiled an ARM
6730 instruction set header which switches to Thumb mode before executing the
6731 rest of the function. This allows these functions to be called from
6732 non-interworking code.
6734 @item -mcaller-super-interworking
6735 @opindex mcaller-super-interworking
6736 Allows calls via function pointers (including virtual functions) to
6737 execute correctly regardless of whether the target code has been
6738 compiled for interworking or not. There is a small overhead in the cost
6739 of executing a function pointer if this option is enabled.
6744 @subsection AVR Options
6747 These options are defined for AVR implementations:
6750 @item -mmcu=@var{mcu}
6752 Specify ATMEL AVR instruction set or MCU type.
6754 Instruction set avr1 is for the minimal AVR core, not supported by the C
6755 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
6756 attiny11, attiny12, attiny15, attiny28).
6758 Instruction set avr2 (default) is for the classic AVR core with up to
6759 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
6760 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
6761 at90c8534, at90s8535).
6763 Instruction set avr3 is for the classic AVR core with up to 128K program
6764 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
6766 Instruction set avr4 is for the enhanced AVR core with up to 8K program
6767 memory space (MCU types: atmega8, atmega83, atmega85).
6769 Instruction set avr5 is for the enhanced AVR core with up to 128K program
6770 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
6771 atmega64, atmega128, at43usb355, at94k).
6775 Output instruction sizes to the asm file.
6777 @item -minit-stack=@var{N}
6778 @opindex minit-stack
6779 Specify the initial stack address, which may be a symbol or numeric value,
6780 @samp{__stack} is the default.
6782 @item -mno-interrupts
6783 @opindex mno-interrupts
6784 Generated code is not compatible with hardware interrupts.
6785 Code size will be smaller.
6787 @item -mcall-prologues
6788 @opindex mcall-prologues
6789 Functions prologues/epilogues expanded as call to appropriate
6790 subroutines. Code size will be smaller.
6792 @item -mno-tablejump
6793 @opindex mno-tablejump
6794 Do not generate tablejump insns which sometimes increase code size.
6797 @opindex mtiny-stack
6798 Change only the low 8 bits of the stack pointer.
6802 Assume int to be 8 bit integer. This affects the sizes of all types: A
6803 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
6804 and long long will be 4 bytes. Please note that this option does not
6805 comply to the C standards, but it will provide you with smaller code
6810 @subsection CRIS Options
6811 @cindex CRIS Options
6813 These options are defined specifically for the CRIS ports.
6816 @item -march=@var{architecture-type}
6817 @itemx -mcpu=@var{architecture-type}
6820 Generate code for the specified architecture. The choices for
6821 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
6822 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX.
6823 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
6826 @item -mtune=@var{architecture-type}
6828 Tune to @var{architecture-type} everything applicable about the generated
6829 code, except for the ABI and the set of available instructions. The
6830 choices for @var{architecture-type} are the same as for
6831 @option{-march=@var{architecture-type}}.
6833 @item -mmax-stack-frame=@var{n}
6834 @opindex mmax-stack-frame
6835 Warn when the stack frame of a function exceeds @var{n} bytes.
6837 @item -melinux-stacksize=@var{n}
6838 @opindex melinux-stacksize
6839 Only available with the @samp{cris-axis-aout} target. Arranges for
6840 indications in the program to the kernel loader that the stack of the
6841 program should be set to @var{n} bytes.
6847 The options @option{-metrax4} and @option{-metrax100} are synonyms for
6848 @option{-march=v3} and @option{-march=v8} respectively.
6850 @item -mmul-bug-workaround
6851 @itemx -mno-mul-bug-workaround
6852 @opindex mmul-bug-workaround
6853 @opindex mno-mul-bug-workaround
6854 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
6855 models where it applies. This option is active by default.
6859 Enable CRIS-specific verbose debug-related information in the assembly
6860 code. This option also has the effect to turn off the @samp{#NO_APP}
6861 formatted-code indicator to the assembler at the beginning of the
6866 Do not use condition-code results from previous instruction; always emit
6867 compare and test instructions before use of condition codes.
6869 @item -mno-side-effects
6870 @opindex mno-side-effects
6871 Do not emit instructions with side-effects in addressing modes other than
6875 @itemx -mno-stack-align
6877 @itemx -mno-data-align
6878 @itemx -mconst-align
6879 @itemx -mno-const-align
6880 @opindex mstack-align
6881 @opindex mno-stack-align
6882 @opindex mdata-align
6883 @opindex mno-data-align
6884 @opindex mconst-align
6885 @opindex mno-const-align
6886 These options (no-options) arranges (eliminate arrangements) for the
6887 stack-frame, individual data and constants to be aligned for the maximum
6888 single data access size for the chosen CPU model. The default is to
6889 arrange for 32-bit alignment. ABI details such as structure layout are
6890 not affected by these options.
6898 Similar to the stack- data- and const-align options above, these options
6899 arrange for stack-frame, writable data and constants to all be 32-bit,
6900 16-bit or 8-bit aligned. The default is 32-bit alignment.
6902 @item -mno-prologue-epilogue
6903 @itemx -mprologue-epilogue
6904 @opindex mno-prologue-epilogue
6905 @opindex mprologue-epilogue
6906 With @option{-mno-prologue-epilogue}, the normal function prologue and
6907 epilogue that sets up the stack-frame are omitted and no return
6908 instructions or return sequences are generated in the code. Use this
6909 option only together with visual inspection of the compiled code: no
6910 warnings or errors are generated when call-saved registers must be saved,
6911 or storage for local variable needs to be allocated.
6917 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
6918 instruction sequences that load addresses for functions from the PLT part
6919 of the GOT rather than (traditional on other architectures) calls to the
6920 PLT. The default is @option{-mgotplt}.
6924 Legacy no-op option only recognized with the cris-axis-aout target.
6928 Legacy no-op option only recognized with the cris-axis-elf and
6929 cris-axis-linux-gnu targets.
6933 Only recognized with the cris-axis-aout target, where it selects a
6934 GNU/linux-like multilib, include files and instruction set for
6939 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
6943 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
6944 to link with input-output functions from a simulator library. Code,
6945 initialized data and zero-initialized data are allocated consecutively.
6949 Like @option{-sim}, but pass linker options to locate initialized data at
6950 0x40000000 and zero-initialized data at 0x80000000.
6953 @node Darwin Options
6954 @subsection Darwin Options
6955 @cindex Darwin options
6957 These options are defined for all architectures running the Darwin operating
6958 system. They are useful for compatibility with other Mac OS compilers.
6963 Add the framework directory @var{dir} to the head of the list of
6964 directories to be searched for header files. These directories are
6965 interleaved with those specified by @option{-I} options and are
6966 scanned in a left-to-right order.
6968 A framework directory is a directory with frameworks in it. A
6969 framework is a directory with a @samp{"Headers"} and/or
6970 @samp{"PrivateHeaders"} directory contained directly in it that ends
6971 in @samp{".framework"}. The name of a framework is the name of this
6972 directory excluding the @samp{".framework"}. Headers associated with
6973 the framework are found in one of those two directories, with
6974 @samp{"Headers"} being searched first. A subframework is a framework
6975 directory that is in a framework's @samp{"Frameworks"} directory.
6976 Includes of subframework headers can only appear in a header of a
6977 framework that contains the subframework, or in a sibling subframework
6978 header. Two subframeworks are siblings if they occur in the same
6979 framework. A subframework should not have the same name as a
6980 framework, a warning will be issued if this is violated. Currently a
6981 subframework cannot have subframeworks, in the future, the mechanism
6982 may be extended to support this. The standard frameworks can be found
6983 in @samp{"/System/Library/Frameworks"} and
6984 @samp{"/Library/Frameworks"}. An example include looks like
6985 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
6986 the name of the framework and header.h is found in the
6987 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
6991 Emit debugging information for symbols that are used. For STABS
6992 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
6993 This is by default ON.
6997 Emit debugging information for all symbols and types.
6999 @item -mone-byte-bool
7000 @opindex -mone-byte-bool
7001 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7002 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7003 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7004 option has no effect on x86.
7006 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
7007 to generate code that is not binary compatible with code generated
7008 without that switch. Using this switch may require recompiling all
7009 other modules in a program, including system libraries. Use this
7010 switch to conform to a non-default data model.
7014 Loads all members of static archive libraries.
7015 See man ld(1) for more information.
7017 @item -arch_errors_fatal
7018 @opindex arch_errors_fatal
7019 Cause the errors having to do with files that have the wrong architecture
7023 @opindex bind_at_load
7024 Causes the output file to be marked such that the dynamic linker will
7025 bind all undefined references when the file is loaded or launched.
7029 Produce a Mach-o bundle format file.
7030 See man ld(1) for more information.
7032 @item -bundle_loader @var{executable}
7033 @opindex bundle_loader
7034 This specifies the @var{executable} that will be loading the build
7035 output file being linked. See man ld(1) for more information.
7037 @item -allowable_client @var{client_name}
7041 @itemx -compatibility_version
7042 @itemx -current_version
7044 @itemx -dependency-file
7046 @itemx -dylinker_install_name
7049 @itemx -exported_symbols_list
7051 @itemx -flat_namespace
7052 @itemx -force_cpusubtype_ALL
7053 @itemx -force_flat_namespace
7054 @itemx -headerpad_max_install_names
7057 @itemx -install_name
7058 @itemx -keep_private_externs
7059 @itemx -multi_module
7060 @itemx -multiply_defined
7061 @itemx -multiply_defined_unused
7063 @itemx -no_dead_strip_inits_and_terms
7064 @itemx -nofixprebinding
7067 @itemx -noseglinkedit
7068 @itemx -pagezero_size
7070 @itemx -prebind_all_twolevel_modules
7071 @itemx -private_bundle
7072 @itemx -read_only_relocs
7074 @itemx -sectobjectsymbols
7078 @itemx -sectobjectsymbols
7080 @itemx -seg_addr_table
7081 @itemx -seg_addr_table_filename
7084 @itemx -segs_read_only_addr
7085 @itemx -segs_read_write_addr
7086 @itemx -single_module
7089 @itemx -sub_umbrella
7090 @itemx -twolevel_namespace
7093 @itemx -unexported_symbols_list
7094 @itemx -weak_reference_mismatches
7097 @opindex allowable_client
7099 @opindex client_name
7100 @opindex compatibility_version
7101 @opindex current_version
7103 @opindex dependency-file
7105 @opindex dylinker_install_name
7108 @opindex exported_symbols_list
7110 @opindex flat_namespace
7111 @opindex force_cpusubtype_ALL
7112 @opindex force_flat_namespace
7113 @opindex headerpad_max_install_names
7116 @opindex install_name
7117 @opindex keep_private_externs
7118 @opindex multi_module
7119 @opindex multiply_defined
7120 @opindex multiply_defined_unused
7122 @opindex no_dead_strip_inits_and_terms
7123 @opindex nofixprebinding
7124 @opindex nomultidefs
7126 @opindex noseglinkedit
7127 @opindex pagezero_size
7129 @opindex prebind_all_twolevel_modules
7130 @opindex private_bundle
7131 @opindex read_only_relocs
7133 @opindex sectobjectsymbols
7137 @opindex sectobjectsymbols
7139 @opindex seg_addr_table
7140 @opindex seg_addr_table_filename
7141 @opindex seglinkedit
7143 @opindex segs_read_only_addr
7144 @opindex segs_read_write_addr
7145 @opindex single_module
7147 @opindex sub_library
7148 @opindex sub_umbrella
7149 @opindex twolevel_namespace
7152 @opindex unexported_symbols_list
7153 @opindex weak_reference_mismatches
7154 @opindex whatsloaded
7156 These options are available for Darwin linker. Darwin linker man page
7157 describes them in detail.
7160 @node DEC Alpha Options
7161 @subsection DEC Alpha Options
7163 These @samp{-m} options are defined for the DEC Alpha implementations:
7166 @item -mno-soft-float
7168 @opindex mno-soft-float
7169 @opindex msoft-float
7170 Use (do not use) the hardware floating-point instructions for
7171 floating-point operations. When @option{-msoft-float} is specified,
7172 functions in @file{libgcc.a} will be used to perform floating-point
7173 operations. Unless they are replaced by routines that emulate the
7174 floating-point operations, or compiled in such a way as to call such
7175 emulations routines, these routines will issue floating-point
7176 operations. If you are compiling for an Alpha without floating-point
7177 operations, you must ensure that the library is built so as not to call
7180 Note that Alpha implementations without floating-point operations are
7181 required to have floating-point registers.
7186 @opindex mno-fp-regs
7187 Generate code that uses (does not use) the floating-point register set.
7188 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
7189 register set is not used, floating point operands are passed in integer
7190 registers as if they were integers and floating-point results are passed
7191 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
7192 so any function with a floating-point argument or return value called by code
7193 compiled with @option{-mno-fp-regs} must also be compiled with that
7196 A typical use of this option is building a kernel that does not use,
7197 and hence need not save and restore, any floating-point registers.
7201 The Alpha architecture implements floating-point hardware optimized for
7202 maximum performance. It is mostly compliant with the IEEE floating
7203 point standard. However, for full compliance, software assistance is
7204 required. This option generates code fully IEEE compliant code
7205 @emph{except} that the @var{inexact-flag} is not maintained (see below).
7206 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
7207 defined during compilation. The resulting code is less efficient but is
7208 able to correctly support denormalized numbers and exceptional IEEE
7209 values such as not-a-number and plus/minus infinity. Other Alpha
7210 compilers call this option @option{-ieee_with_no_inexact}.
7212 @item -mieee-with-inexact
7213 @opindex mieee-with-inexact
7214 This is like @option{-mieee} except the generated code also maintains
7215 the IEEE @var{inexact-flag}. Turning on this option causes the
7216 generated code to implement fully-compliant IEEE math. In addition to
7217 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
7218 macro. On some Alpha implementations the resulting code may execute
7219 significantly slower than the code generated by default. Since there is
7220 very little code that depends on the @var{inexact-flag}, you should
7221 normally not specify this option. Other Alpha compilers call this
7222 option @option{-ieee_with_inexact}.
7224 @item -mfp-trap-mode=@var{trap-mode}
7225 @opindex mfp-trap-mode
7226 This option controls what floating-point related traps are enabled.
7227 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
7228 The trap mode can be set to one of four values:
7232 This is the default (normal) setting. The only traps that are enabled
7233 are the ones that cannot be disabled in software (e.g., division by zero
7237 In addition to the traps enabled by @samp{n}, underflow traps are enabled
7241 Like @samp{su}, but the instructions are marked to be safe for software
7242 completion (see Alpha architecture manual for details).
7245 Like @samp{su}, but inexact traps are enabled as well.
7248 @item -mfp-rounding-mode=@var{rounding-mode}
7249 @opindex mfp-rounding-mode
7250 Selects the IEEE rounding mode. Other Alpha compilers call this option
7251 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
7256 Normal IEEE rounding mode. Floating point numbers are rounded towards
7257 the nearest machine number or towards the even machine number in case
7261 Round towards minus infinity.
7264 Chopped rounding mode. Floating point numbers are rounded towards zero.
7267 Dynamic rounding mode. A field in the floating point control register
7268 (@var{fpcr}, see Alpha architecture reference manual) controls the
7269 rounding mode in effect. The C library initializes this register for
7270 rounding towards plus infinity. Thus, unless your program modifies the
7271 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
7274 @item -mtrap-precision=@var{trap-precision}
7275 @opindex mtrap-precision
7276 In the Alpha architecture, floating point traps are imprecise. This
7277 means without software assistance it is impossible to recover from a
7278 floating trap and program execution normally needs to be terminated.
7279 GCC can generate code that can assist operating system trap handlers
7280 in determining the exact location that caused a floating point trap.
7281 Depending on the requirements of an application, different levels of
7282 precisions can be selected:
7286 Program precision. This option is the default and means a trap handler
7287 can only identify which program caused a floating point exception.
7290 Function precision. The trap handler can determine the function that
7291 caused a floating point exception.
7294 Instruction precision. The trap handler can determine the exact
7295 instruction that caused a floating point exception.
7298 Other Alpha compilers provide the equivalent options called
7299 @option{-scope_safe} and @option{-resumption_safe}.
7301 @item -mieee-conformant
7302 @opindex mieee-conformant
7303 This option marks the generated code as IEEE conformant. You must not
7304 use this option unless you also specify @option{-mtrap-precision=i} and either
7305 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
7306 is to emit the line @samp{.eflag 48} in the function prologue of the
7307 generated assembly file. Under DEC Unix, this has the effect that
7308 IEEE-conformant math library routines will be linked in.
7310 @item -mbuild-constants
7311 @opindex mbuild-constants
7312 Normally GCC examines a 32- or 64-bit integer constant to
7313 see if it can construct it from smaller constants in two or three
7314 instructions. If it cannot, it will output the constant as a literal and
7315 generate code to load it from the data segment at runtime.
7317 Use this option to require GCC to construct @emph{all} integer constants
7318 using code, even if it takes more instructions (the maximum is six).
7320 You would typically use this option to build a shared library dynamic
7321 loader. Itself a shared library, it must relocate itself in memory
7322 before it can find the variables and constants in its own data segment.
7328 Select whether to generate code to be assembled by the vendor-supplied
7329 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
7347 Indicate whether GCC should generate code to use the optional BWX,
7348 CIX, FIX and MAX instruction sets. The default is to use the instruction
7349 sets supported by the CPU type specified via @option{-mcpu=} option or that
7350 of the CPU on which GCC was built if none was specified.
7355 @opindex mfloat-ieee
7356 Generate code that uses (does not use) VAX F and G floating point
7357 arithmetic instead of IEEE single and double precision.
7359 @item -mexplicit-relocs
7360 @itemx -mno-explicit-relocs
7361 @opindex mexplicit-relocs
7362 @opindex mno-explicit-relocs
7363 Older Alpha assemblers provided no way to generate symbol relocations
7364 except via assembler macros. Use of these macros does not allow
7365 optimal instruction scheduling. GNU binutils as of version 2.12
7366 supports a new syntax that allows the compiler to explicitly mark
7367 which relocations should apply to which instructions. This option
7368 is mostly useful for debugging, as GCC detects the capabilities of
7369 the assembler when it is built and sets the default accordingly.
7373 @opindex msmall-data
7374 @opindex mlarge-data
7375 When @option{-mexplicit-relocs} is in effect, static data is
7376 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
7377 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
7378 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
7379 16-bit relocations off of the @code{$gp} register. This limits the
7380 size of the small data area to 64KB, but allows the variables to be
7381 directly accessed via a single instruction.
7383 The default is @option{-mlarge-data}. With this option the data area
7384 is limited to just below 2GB. Programs that require more than 2GB of
7385 data must use @code{malloc} or @code{mmap} to allocate the data in the
7386 heap instead of in the program's data segment.
7388 When generating code for shared libraries, @option{-fpic} implies
7389 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
7393 @opindex msmall-text
7394 @opindex mlarge-text
7395 When @option{-msmall-text} is used, the compiler assumes that the
7396 code of the entire program (or shared library) fits in 4MB, and is
7397 thus reachable with a branch instruction. When @option{-msmall-data}
7398 is used, the compiler can assume that all local symbols share the
7399 same @code{$gp} value, and thus reduce the number of instructions
7400 required for a function call from 4 to 1.
7402 The default is @option{-mlarge-text}.
7404 @item -mcpu=@var{cpu_type}
7406 Set the instruction set and instruction scheduling parameters for
7407 machine type @var{cpu_type}. You can specify either the @samp{EV}
7408 style name or the corresponding chip number. GCC supports scheduling
7409 parameters for the EV4, EV5 and EV6 family of processors and will
7410 choose the default values for the instruction set from the processor
7411 you specify. If you do not specify a processor type, GCC will default
7412 to the processor on which the compiler was built.
7414 Supported values for @var{cpu_type} are
7420 Schedules as an EV4 and has no instruction set extensions.
7424 Schedules as an EV5 and has no instruction set extensions.
7428 Schedules as an EV5 and supports the BWX extension.
7433 Schedules as an EV5 and supports the BWX and MAX extensions.
7437 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
7441 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
7444 @item -mtune=@var{cpu_type}
7446 Set only the instruction scheduling parameters for machine type
7447 @var{cpu_type}. The instruction set is not changed.
7449 @item -mmemory-latency=@var{time}
7450 @opindex mmemory-latency
7451 Sets the latency the scheduler should assume for typical memory
7452 references as seen by the application. This number is highly
7453 dependent on the memory access patterns used by the application
7454 and the size of the external cache on the machine.
7456 Valid options for @var{time} are
7460 A decimal number representing clock cycles.
7466 The compiler contains estimates of the number of clock cycles for
7467 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
7468 (also called Dcache, Scache, and Bcache), as well as to main memory.
7469 Note that L3 is only valid for EV5.
7474 @node DEC Alpha/VMS Options
7475 @subsection DEC Alpha/VMS Options
7477 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
7480 @item -mvms-return-codes
7481 @opindex mvms-return-codes
7482 Return VMS condition codes from main. The default is to return POSIX
7483 style condition (e.g.@ error) codes.
7487 @subsection FRV Options
7494 Only use the first 32 general purpose registers.
7499 Use all 64 general purpose registers.
7504 Use only the first 32 floating point registers.
7509 Use all 64 floating point registers
7512 @opindex mhard-float
7514 Use hardware instructions for floating point operations.
7517 @opindex msoft-float
7519 Use library routines for floating point operations.
7524 Dynamically allocate condition code registers.
7529 Do not try to dynamically allocate condition code registers, only
7530 use @code{icc0} and @code{fcc0}.
7535 Change ABI to use double word insns.
7540 Do not use double word instructions.
7545 Use floating point double instructions.
7550 Do not use floating point double instructions.
7555 Use media instructions.
7560 Do not use media instructions.
7565 Use multiply and add/subtract instructions.
7570 Do not use multiply and add/subtract instructions.
7573 @opindex mlibrary-pic
7575 Generate position-independent EABI code.
7580 Use only the first four media accumulator registers.
7585 Use all eight media accumulator registers.
7590 Pack VLIW instructions.
7595 Do not pack VLIW instructions.
7600 Do not mark ABI switches in e_flags.
7605 Enable the use of conditional-move instructions (default).
7607 This switch is mainly for debugging the compiler and will likely be removed
7608 in a future version.
7610 @item -mno-cond-move
7611 @opindex mno-cond-move
7613 Disable the use of conditional-move instructions.
7615 This switch is mainly for debugging the compiler and will likely be removed
7616 in a future version.
7621 Enable the use of conditional set instructions (default).
7623 This switch is mainly for debugging the compiler and will likely be removed
7624 in a future version.
7629 Disable the use of conditional set instructions.
7631 This switch is mainly for debugging the compiler and will likely be removed
7632 in a future version.
7637 Enable the use of conditional execution (default).
7639 This switch is mainly for debugging the compiler and will likely be removed
7640 in a future version.
7642 @item -mno-cond-exec
7643 @opindex mno-cond-exec
7645 Disable the use of conditional execution.
7647 This switch is mainly for debugging the compiler and will likely be removed
7648 in a future version.
7651 @opindex mvliw-branch
7653 Run a pass to pack branches into VLIW instructions (default).
7655 This switch is mainly for debugging the compiler and will likely be removed
7656 in a future version.
7658 @item -mno-vliw-branch
7659 @opindex mno-vliw-branch
7661 Do not run a pass to pack branches into VLIW instructions.
7663 This switch is mainly for debugging the compiler and will likely be removed
7664 in a future version.
7666 @item -mmulti-cond-exec
7667 @opindex mmulti-cond-exec
7669 Enable optimization of @code{&&} and @code{||} in conditional execution
7672 This switch is mainly for debugging the compiler and will likely be removed
7673 in a future version.
7675 @item -mno-multi-cond-exec
7676 @opindex mno-multi-cond-exec
7678 Disable optimization of @code{&&} and @code{||} in conditional execution.
7680 This switch is mainly for debugging the compiler and will likely be removed
7681 in a future version.
7683 @item -mnested-cond-exec
7684 @opindex mnested-cond-exec
7686 Enable nested conditional execution optimizations (default).
7688 This switch is mainly for debugging the compiler and will likely be removed
7689 in a future version.
7691 @item -mno-nested-cond-exec
7692 @opindex mno-nested-cond-exec
7694 Disable nested conditional execution optimizations.
7696 This switch is mainly for debugging the compiler and will likely be removed
7697 in a future version.
7699 @item -mtomcat-stats
7700 @opindex mtomcat-stats
7702 Cause gas to print out tomcat statistics.
7704 @item -mcpu=@var{cpu}
7707 Select the processor type for which to generate code. Possible values are
7708 @samp{simple}, @samp{tomcat}, @samp{fr500}, @samp{fr400}, @samp{fr300},
7713 @node H8/300 Options
7714 @subsection H8/300 Options
7716 These @samp{-m} options are defined for the H8/300 implementations:
7721 Shorten some address references at link time, when possible; uses the
7722 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
7723 ld, Using ld}, for a fuller description.
7727 Generate code for the H8/300H@.
7731 Generate code for the H8S@.
7735 Generate code for the H8S and H8/300H in the normal mode. This switch
7736 must be used either with -mh or -ms.
7740 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
7744 Make @code{int} data 32 bits by default.
7748 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
7749 The default for the H8/300H and H8S is to align longs and floats on 4
7751 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
7752 This option has no effect on the H8/300.
7756 @subsection HPPA Options
7757 @cindex HPPA Options
7759 These @samp{-m} options are defined for the HPPA family of computers:
7762 @item -march=@var{architecture-type}
7764 Generate code for the specified architecture. The choices for
7765 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
7766 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
7767 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
7768 architecture option for your machine. Code compiled for lower numbered
7769 architectures will run on higher numbered architectures, but not the
7772 PA 2.0 support currently requires gas snapshot 19990413 or later. The
7773 next release of binutils (current is 2.9.1) will probably contain PA 2.0
7777 @itemx -mpa-risc-1-1
7778 @itemx -mpa-risc-2-0
7779 @opindex mpa-risc-1-0
7780 @opindex mpa-risc-1-1
7781 @opindex mpa-risc-2-0
7782 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
7785 @opindex mbig-switch
7786 Generate code suitable for big switch tables. Use this option only if
7787 the assembler/linker complain about out of range branches within a switch
7790 @item -mjump-in-delay
7791 @opindex mjump-in-delay
7792 Fill delay slots of function calls with unconditional jump instructions
7793 by modifying the return pointer for the function call to be the target
7794 of the conditional jump.
7796 @item -mdisable-fpregs
7797 @opindex mdisable-fpregs
7798 Prevent floating point registers from being used in any manner. This is
7799 necessary for compiling kernels which perform lazy context switching of
7800 floating point registers. If you use this option and attempt to perform
7801 floating point operations, the compiler will abort.
7803 @item -mdisable-indexing
7804 @opindex mdisable-indexing
7805 Prevent the compiler from using indexing address modes. This avoids some
7806 rather obscure problems when compiling MIG generated code under MACH@.
7808 @item -mno-space-regs
7809 @opindex mno-space-regs
7810 Generate code that assumes the target has no space registers. This allows
7811 GCC to generate faster indirect calls and use unscaled index address modes.
7813 Such code is suitable for level 0 PA systems and kernels.
7815 @item -mfast-indirect-calls
7816 @opindex mfast-indirect-calls
7817 Generate code that assumes calls never cross space boundaries. This
7818 allows GCC to emit code which performs faster indirect calls.
7820 This option will not work in the presence of shared libraries or nested
7823 @item -mfixed-range=@var{register-range}
7824 @opindex mfixed-range
7825 Generate code treating the given register range as fixed registers.
7826 A fixed register is one that the register allocator can not use. This is
7827 useful when compiling kernel code. A register range is specified as
7828 two registers separated by a dash. Multiple register ranges can be
7829 specified separated by a comma.
7831 @item -mlong-load-store
7832 @opindex mlong-load-store
7833 Generate 3-instruction load and store sequences as sometimes required by
7834 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
7837 @item -mportable-runtime
7838 @opindex mportable-runtime
7839 Use the portable calling conventions proposed by HP for ELF systems.
7843 Enable the use of assembler directives only GAS understands.
7845 @item -mschedule=@var{cpu-type}
7847 Schedule code according to the constraints for the machine type
7848 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
7849 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
7850 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
7851 proper scheduling option for your machine. The default scheduling is
7855 @opindex mlinker-opt
7856 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
7857 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
7858 linkers in which they give bogus error messages when linking some programs.
7861 @opindex msoft-float
7862 Generate output containing library calls for floating point.
7863 @strong{Warning:} the requisite libraries are not available for all HPPA
7864 targets. Normally the facilities of the machine's usual C compiler are
7865 used, but this cannot be done directly in cross-compilation. You must make
7866 your own arrangements to provide suitable library functions for
7867 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
7868 does provide software floating point support.
7870 @option{-msoft-float} changes the calling convention in the output file;
7871 therefore, it is only useful if you compile @emph{all} of a program with
7872 this option. In particular, you need to compile @file{libgcc.a}, the
7873 library that comes with GCC, with @option{-msoft-float} in order for
7878 Generate the predefine, @code{_SIO}, for server IO. The default is
7879 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
7880 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO. These
7881 options are available under HP-UX and HI-UX.
7885 Use GNU ld specific options. This passes @option{-shared} to ld when
7886 building a shared library. It is the default when GCC is configured,
7887 explicitly or implicitly, with the GNU linker. This option does not
7888 have any affect on which ld is called, it only changes what parameters
7889 are passed to that ld. The ld that is called is determined by the
7890 @option{--with-ld} configure option, GCC's program search path, and
7891 finally by the user's @env{PATH}. The linker used by GCC can be printed
7892 using @samp{which `gcc -print-prog-name=ld`}.
7896 Use HP ld specific options. This passes @option{-b} to ld when building
7897 a shared library and passes @option{+Accept TypeMismatch} to ld on all
7898 links. It is the default when GCC is configured, explicitly or
7899 implicitly, with the HP linker. This option does not have any affect on
7900 which ld is called, it only changes what parameters are passed to that
7901 ld. The ld that is called is determined by the @option{--with-ld}
7902 configure option, GCC's program search path, and finally by the user's
7903 @env{PATH}. The linker used by GCC can be printed using @samp{which
7904 `gcc -print-prog-name=ld`}.
7909 Select the FDPIC ABI, that uses function descriptors to represent
7910 pointers to functions. Without any PIC/PIE-related options, it
7911 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
7912 assumes GOT entries and small data are within a 12-bit range from the
7913 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
7914 are computed with 32 bits.
7917 @opindex minline-plt
7919 Enable inlining of PLT entries in function calls to functions that are
7920 not known to bind locally. It has no effect without @option{-mfdpic}.
7921 It's enabled by default if optimizing for speed and compiling for
7922 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
7923 optimization option such as @option{-O3} or above is present in the
7929 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
7930 that is known to be in read-only sections. It's enabled by default,
7931 except for @option{-fpic} or @option{-fpie}: even though it may help
7932 make the global offset table smaller, it trades 1 instruction for 4.
7933 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
7934 one of which may be shared by multiple symbols, and it avoids the need
7935 for a GOT entry for the referenced symbol, so it's more likely to be a
7936 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
7938 @item -multilib-library-pic
7939 @opindex multilib-library-pic
7941 Link with the (library, not FD) pic libraries. It's implied by
7942 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
7943 @option{-fpic} without @option{-mfdpic}. You should never have to use
7949 Follow the EABI requirement of always creating a frame pointer whenever
7950 a stack frame is allocated. This option is enabled by default and can
7951 be disabled with @option{-mno-linked-fp}.
7954 @opindex mno-long-calls
7955 Generate code that uses long call sequences. This ensures that a call
7956 is always able to reach linker generated stubs. The default is to generate
7957 long calls only when the distance from the call site to the beginning
7958 of the function or translation unit, as the case may be, exceeds a
7959 predefined limit set by the branch type being used. The limits for
7960 normal calls are 7,600,000 and 240,000 bytes, respectively for the
7961 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
7964 Distances are measured from the beginning of functions when using the
7965 @option{-ffunction-sections} option, or when using the @option{-mgas}
7966 and @option{-mno-portable-runtime} options together under HP-UX with
7969 It is normally not desirable to use this option as it will degrade
7970 performance. However, it may be useful in large applications,
7971 particularly when partial linking is used to build the application.
7973 The types of long calls used depends on the capabilities of the
7974 assembler and linker, and the type of code being generated. The
7975 impact on systems that support long absolute calls, and long pic
7976 symbol-difference or pc-relative calls should be relatively small.
7977 However, an indirect call is used on 32-bit ELF systems in pic code
7978 and it is quite long.
7982 Suppress the generation of link options to search libdld.sl when the
7983 @option{-static} option is specified on HP-UX 10 and later.
7987 The HP-UX implementation of setlocale in libc has a dependency on
7988 libdld.sl. There isn't an archive version of libdld.sl. Thus,
7989 when the @option{-static} option is specified, special link options
7990 are needed to resolve this dependency.
7992 On HP-UX 10 and later, the GCC driver adds the necessary options to
7993 link with libdld.sl when the @option{-static} option is specified.
7994 This causes the resulting binary to be dynamic. On the 64-bit port,
7995 the linkers generate dynamic binaries by default in any case. The
7996 @option{-nolibdld} option can be used to prevent the GCC driver from
7997 adding these link options.
8001 Add support for multithreading with the @dfn{dce thread} library
8002 under HP-UX. This option sets flags for both the preprocessor and
8006 @node i386 and x86-64 Options
8007 @subsection Intel 386 and AMD x86-64 Options
8008 @cindex i386 Options
8009 @cindex x86-64 Options
8010 @cindex Intel 386 Options
8011 @cindex AMD x86-64 Options
8013 These @samp{-m} options are defined for the i386 and x86-64 family of
8017 @item -mtune=@var{cpu-type}
8019 Tune to @var{cpu-type} everything applicable about the generated code, except
8020 for the ABI and the set of available instructions. The choices for
8024 Original Intel's i386 CPU.
8026 Intel's i486 CPU. (No scheduling is implemented for this chip.)
8028 Intel Pentium CPU with no MMX support.
8030 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
8031 @item i686, pentiumpro
8032 Intel PentiumPro CPU.
8034 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
8035 @item pentium3, pentium3m
8036 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
8039 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
8040 support. Used by Centrino notebooks.
8041 @item pentium4, pentium4m
8042 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
8044 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
8047 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
8048 SSE2 and SSE3 instruction set support.
8050 AMD K6 CPU with MMX instruction set support.
8052 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
8053 @item athlon, athlon-tbird
8054 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
8056 @item athlon-4, athlon-xp, athlon-mp
8057 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
8058 instruction set support.
8059 @item k8, opteron, athlon64, athlon-fx
8060 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
8061 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
8063 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
8066 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
8067 instruction set support.
8069 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
8070 implemented for this chip.)
8072 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
8073 implemented for this chip.)
8076 While picking a specific @var{cpu-type} will schedule things appropriately
8077 for that particular chip, the compiler will not generate any code that
8078 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8081 @item -march=@var{cpu-type}
8083 Generate instructions for the machine type @var{cpu-type}. The choices
8084 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8085 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8087 @item -mcpu=@var{cpu-type}
8089 A deprecated synonym for @option{-mtune}.
8098 @opindex mpentiumpro
8099 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8100 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8101 These synonyms are deprecated.
8103 @item -mfpmath=@var{unit}
8105 Generate floating point arithmetics for selected unit @var{unit}. The choices
8110 Use the standard 387 floating point coprocessor present majority of chips and
8111 emulated otherwise. Code compiled with this option will run almost everywhere.
8112 The temporary results are computed in 80bit precision instead of precision
8113 specified by the type resulting in slightly different results compared to most
8114 of other chips. See @option{-ffloat-store} for more detailed description.
8116 This is the default choice for i386 compiler.
8119 Use scalar floating point instructions present in the SSE instruction set.
8120 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8121 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8122 instruction set supports only single precision arithmetics, thus the double and
8123 extended precision arithmetics is still done using 387. Later version, present
8124 only in Pentium4 and the future AMD x86-64 chips supports double precision
8127 For i387 you need to use @option{-march=@var{cpu-type}}, @option{-msse} or
8128 @option{-msse2} switches to enable SSE extensions and make this option
8129 effective. For x86-64 compiler, these extensions are enabled by default.
8131 The resulting code should be considerably faster in the majority of cases and avoid
8132 the numerical instability problems of 387 code, but may break some existing
8133 code that expects temporaries to be 80bit.
8135 This is the default choice for the x86-64 compiler.
8138 Attempt to utilize both instruction sets at once. This effectively double the
8139 amount of available registers and on chips with separate execution units for
8140 387 and SSE the execution resources too. Use this option with care, as it is
8141 still experimental, because the GCC register allocator does not model separate
8142 functional units well resulting in instable performance.
8145 @item -masm=@var{dialect}
8146 @opindex masm=@var{dialect}
8147 Output asm instructions using selected @var{dialect}. Supported choices are
8148 @samp{intel} or @samp{att} (the default one).
8153 @opindex mno-ieee-fp
8154 Control whether or not the compiler uses IEEE floating point
8155 comparisons. These handle correctly the case where the result of a
8156 comparison is unordered.
8159 @opindex msoft-float
8160 Generate output containing library calls for floating point.
8161 @strong{Warning:} the requisite libraries are not part of GCC@.
8162 Normally the facilities of the machine's usual C compiler are used, but
8163 this can't be done directly in cross-compilation. You must make your
8164 own arrangements to provide suitable library functions for
8167 On machines where a function returns floating point results in the 80387
8168 register stack, some floating point opcodes may be emitted even if
8169 @option{-msoft-float} is used.
8171 @item -mno-fp-ret-in-387
8172 @opindex mno-fp-ret-in-387
8173 Do not use the FPU registers for return values of functions.
8175 The usual calling convention has functions return values of types
8176 @code{float} and @code{double} in an FPU register, even if there
8177 is no FPU@. The idea is that the operating system should emulate
8180 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8181 in ordinary CPU registers instead.
8183 @item -mno-fancy-math-387
8184 @opindex mno-fancy-math-387
8185 Some 387 emulators do not support the @code{sin}, @code{cos} and
8186 @code{sqrt} instructions for the 387. Specify this option to avoid
8187 generating those instructions. This option is the default on FreeBSD,
8188 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8189 indicates that the target cpu will always have an FPU and so the
8190 instruction will not need emulation. As of revision 2.6.1, these
8191 instructions are not generated unless you also use the
8192 @option{-funsafe-math-optimizations} switch.
8194 @item -malign-double
8195 @itemx -mno-align-double
8196 @opindex malign-double
8197 @opindex mno-align-double
8198 Control whether GCC aligns @code{double}, @code{long double}, and
8199 @code{long long} variables on a two word boundary or a one word
8200 boundary. Aligning @code{double} variables on a two word boundary will
8201 produce code that runs somewhat faster on a @samp{Pentium} at the
8202 expense of more memory.
8204 @strong{Warning:} if you use the @option{-malign-double} switch,
8205 structures containing the above types will be aligned differently than
8206 the published application binary interface specifications for the 386
8207 and will not be binary compatible with structures in code compiled
8208 without that switch.
8210 @item -m96bit-long-double
8211 @itemx -m128bit-long-double
8212 @opindex m96bit-long-double
8213 @opindex m128bit-long-double
8214 These switches control the size of @code{long double} type. The i386
8215 application binary interface specifies the size to be 96 bits,
8216 so @option{-m96bit-long-double} is the default in 32 bit mode.
8218 Modern architectures (Pentium and newer) would prefer @code{long double}
8219 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8220 conforming to the ABI, this would not be possible. So specifying a
8221 @option{-m128bit-long-double} will align @code{long double}
8222 to a 16 byte boundary by padding the @code{long double} with an additional
8225 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8226 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8228 Notice that neither of these options enable any extra precision over the x87
8229 standard of 80 bits for a @code{long double}.
8231 @strong{Warning:} if you override the default value for your target ABI, the
8232 structures and arrays containing @code{long double} variables will change
8233 their size as well as function calling convention for function taking
8234 @code{long double} will be modified. Hence they will not be binary
8235 compatible with arrays or structures in code compiled without that switch.
8239 @itemx -mno-svr3-shlib
8240 @opindex msvr3-shlib
8241 @opindex mno-svr3-shlib
8242 Control whether GCC places uninitialized local variables into the
8243 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8244 into @code{bss}. These options are meaningful only on System V Release 3.
8248 Use a different function-calling convention, in which functions that
8249 take a fixed number of arguments return with the @code{ret} @var{num}
8250 instruction, which pops their arguments while returning. This saves one
8251 instruction in the caller since there is no need to pop the arguments
8254 You can specify that an individual function is called with this calling
8255 sequence with the function attribute @samp{stdcall}. You can also
8256 override the @option{-mrtd} option by using the function attribute
8257 @samp{cdecl}. @xref{Function Attributes}.
8259 @strong{Warning:} this calling convention is incompatible with the one
8260 normally used on Unix, so you cannot use it if you need to call
8261 libraries compiled with the Unix compiler.
8263 Also, you must provide function prototypes for all functions that
8264 take variable numbers of arguments (including @code{printf});
8265 otherwise incorrect code will be generated for calls to those
8268 In addition, seriously incorrect code will result if you call a
8269 function with too many arguments. (Normally, extra arguments are
8270 harmlessly ignored.)
8272 @item -mregparm=@var{num}
8274 Control how many registers are used to pass integer arguments. By
8275 default, no registers are used to pass arguments, and at most 3
8276 registers can be used. You can control this behavior for a specific
8277 function by using the function attribute @samp{regparm}.
8278 @xref{Function Attributes}.
8280 @strong{Warning:} if you use this switch, and
8281 @var{num} is nonzero, then you must build all modules with the same
8282 value, including any libraries. This includes the system libraries and
8285 @item -mpreferred-stack-boundary=@var{num}
8286 @opindex mpreferred-stack-boundary
8287 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8288 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8289 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8290 size (@option{-Os}), in which case the default is the minimum correct
8291 alignment (4 bytes for x86, and 8 bytes for x86-64).
8293 On Pentium and PentiumPro, @code{double} and @code{long double} values
8294 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8295 suffer significant run time performance penalties. On Pentium III, the
8296 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8297 penalties if it is not 16 byte aligned.
8299 To ensure proper alignment of this values on the stack, the stack boundary
8300 must be as aligned as that required by any value stored on the stack.
8301 Further, every function must be generated such that it keeps the stack
8302 aligned. Thus calling a function compiled with a higher preferred
8303 stack boundary from a function compiled with a lower preferred stack
8304 boundary will most likely misalign the stack. It is recommended that
8305 libraries that use callbacks always use the default setting.
8307 This extra alignment does consume extra stack space, and generally
8308 increases code size. Code that is sensitive to stack space usage, such
8309 as embedded systems and operating system kernels, may want to reduce the
8310 preferred alignment to @option{-mpreferred-stack-boundary=2}.
8328 These switches enable or disable the use of built-in functions that allow
8329 direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow extensions of the
8332 @xref{X86 Built-in Functions}, for details of the functions enabled
8333 and disabled by these switches.
8335 To have SSE/SSE2 instructions generated automatically from floating-point
8336 code, see @option{-mfpmath=sse}.
8339 @itemx -mno-push-args
8341 @opindex mno-push-args
8342 Use PUSH operations to store outgoing parameters. This method is shorter
8343 and usually equally fast as method using SUB/MOV operations and is enabled
8344 by default. In some cases disabling it may improve performance because of
8345 improved scheduling and reduced dependencies.
8347 @item -maccumulate-outgoing-args
8348 @opindex maccumulate-outgoing-args
8349 If enabled, the maximum amount of space required for outgoing arguments will be
8350 computed in the function prologue. This is faster on most modern CPUs
8351 because of reduced dependencies, improved scheduling and reduced stack usage
8352 when preferred stack boundary is not equal to 2. The drawback is a notable
8353 increase in code size. This switch implies @option{-mno-push-args}.
8357 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8358 on thread-safe exception handling must compile and link all code with the
8359 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8360 @option{-D_MT}; when linking, it links in a special thread helper library
8361 @option{-lmingwthrd} which cleans up per thread exception handling data.
8363 @item -mno-align-stringops
8364 @opindex mno-align-stringops
8365 Do not align destination of inlined string operations. This switch reduces
8366 code size and improves performance in case the destination is already aligned,
8367 but GCC doesn't know about it.
8369 @item -minline-all-stringops
8370 @opindex minline-all-stringops
8371 By default GCC inlines string operations only when destination is known to be
8372 aligned at least to 4 byte boundary. This enables more inlining, increase code
8373 size, but may improve performance of code that depends on fast memcpy, strlen
8374 and memset for short lengths.
8376 @item -momit-leaf-frame-pointer
8377 @opindex momit-leaf-frame-pointer
8378 Don't keep the frame pointer in a register for leaf functions. This
8379 avoids the instructions to save, set up and restore frame pointers and
8380 makes an extra register available in leaf functions. The option
8381 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8382 which might make debugging harder.
8384 @item -mtls-direct-seg-refs
8385 @itemx -mno-tls-direct-seg-refs
8386 @opindex mtls-direct-seg-refs
8387 Controls whether TLS variables may be accessed with offsets from the
8388 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
8389 or whether the thread base pointer must be added. Whether or not this
8390 is legal depends on the operating system, and whether it maps the
8391 segment to cover the entire TLS area.
8393 For systems that use GNU libc, the default is on.
8396 These @samp{-m} switches are supported in addition to the above
8397 on AMD x86-64 processors in 64-bit environments.
8404 Generate code for a 32-bit or 64-bit environment.
8405 The 32-bit environment sets int, long and pointer to 32 bits and
8406 generates code that runs on any i386 system.
8407 The 64-bit environment sets int to 32 bits and long and pointer
8408 to 64 bits and generates code for AMD's x86-64 architecture.
8411 @opindex no-red-zone
8412 Do not use a so called red zone for x86-64 code. The red zone is mandated
8413 by the x86-64 ABI, it is a 128-byte area beyond the location of the
8414 stack pointer that will not be modified by signal or interrupt handlers
8415 and therefore can be used for temporary data without adjusting the stack
8416 pointer. The flag @option{-mno-red-zone} disables this red zone.
8418 @item -mcmodel=small
8419 @opindex mcmodel=small
8420 Generate code for the small code model: the program and its symbols must
8421 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
8422 Programs can be statically or dynamically linked. This is the default
8425 @item -mcmodel=kernel
8426 @opindex mcmodel=kernel
8427 Generate code for the kernel code model. The kernel runs in the
8428 negative 2 GB of the address space.
8429 This model has to be used for Linux kernel code.
8431 @item -mcmodel=medium
8432 @opindex mcmodel=medium
8433 Generate code for the medium model: The program is linked in the lower 2
8434 GB of the address space but symbols can be located anywhere in the
8435 address space. Programs can be statically or dynamically linked, but
8436 building of shared libraries are not supported with the medium model.
8438 @item -mcmodel=large
8439 @opindex mcmodel=large
8440 Generate code for the large model: This model makes no assumptions
8441 about addresses and sizes of sections. Currently GCC does not implement
8446 @subsection IA-64 Options
8447 @cindex IA-64 Options
8449 These are the @samp{-m} options defined for the Intel IA-64 architecture.
8453 @opindex mbig-endian
8454 Generate code for a big endian target. This is the default for HP-UX@.
8456 @item -mlittle-endian
8457 @opindex mlittle-endian
8458 Generate code for a little endian target. This is the default for AIX5
8465 Generate (or don't) code for the GNU assembler. This is the default.
8466 @c Also, this is the default if the configure option @option{--with-gnu-as}
8473 Generate (or don't) code for the GNU linker. This is the default.
8474 @c Also, this is the default if the configure option @option{--with-gnu-ld}
8479 Generate code that does not use a global pointer register. The result
8480 is not position independent code, and violates the IA-64 ABI@.
8482 @item -mvolatile-asm-stop
8483 @itemx -mno-volatile-asm-stop
8484 @opindex mvolatile-asm-stop
8485 @opindex mno-volatile-asm-stop
8486 Generate (or don't) a stop bit immediately before and after volatile asm
8491 Generate code that works around Itanium B step errata.
8493 @item -mregister-names
8494 @itemx -mno-register-names
8495 @opindex mregister-names
8496 @opindex mno-register-names
8497 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
8498 the stacked registers. This may make assembler output more readable.
8504 Disable (or enable) optimizations that use the small data section. This may
8505 be useful for working around optimizer bugs.
8508 @opindex mconstant-gp
8509 Generate code that uses a single constant global pointer value. This is
8510 useful when compiling kernel code.
8514 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
8515 This is useful when compiling firmware code.
8517 @item -minline-float-divide-min-latency
8518 @opindex minline-float-divide-min-latency
8519 Generate code for inline divides of floating point values
8520 using the minimum latency algorithm.
8522 @item -minline-float-divide-max-throughput
8523 @opindex minline-float-divide-max-throughput
8524 Generate code for inline divides of floating point values
8525 using the maximum throughput algorithm.
8527 @item -minline-int-divide-min-latency
8528 @opindex minline-int-divide-min-latency
8529 Generate code for inline divides of integer values
8530 using the minimum latency algorithm.
8532 @item -minline-int-divide-max-throughput
8533 @opindex minline-int-divide-max-throughput
8534 Generate code for inline divides of integer values
8535 using the maximum throughput algorithm.
8537 @item -mno-dwarf2-asm
8539 @opindex mno-dwarf2-asm
8540 @opindex mdwarf2-asm
8541 Don't (or do) generate assembler code for the DWARF2 line number debugging
8542 info. This may be useful when not using the GNU assembler.
8544 @item -mfixed-range=@var{register-range}
8545 @opindex mfixed-range
8546 Generate code treating the given register range as fixed registers.
8547 A fixed register is one that the register allocator can not use. This is
8548 useful when compiling kernel code. A register range is specified as
8549 two registers separated by a dash. Multiple register ranges can be
8550 specified separated by a comma.
8552 @item -mearly-stop-bits
8553 @itemx -mno-early-stop-bits
8554 @opindex mearly-stop-bits
8555 @opindex mno-early-stop-bits
8556 Allow stop bits to be placed earlier than immediately preceding the
8557 instruction that triggered the stop bit. This can improve instruction
8558 scheduling, but does not always do so.
8561 @node M32R/D Options
8562 @subsection M32R/D Options
8563 @cindex M32R/D options
8565 These @option{-m} options are defined for Renesas M32R/D architectures:
8570 Generate code for the M32R/2@.
8574 Generate code for the M32R/X@.
8578 Generate code for the M32R@. This is the default.
8581 @opindex mmodel=small
8582 Assume all objects live in the lower 16MB of memory (so that their addresses
8583 can be loaded with the @code{ld24} instruction), and assume all subroutines
8584 are reachable with the @code{bl} instruction.
8585 This is the default.
8587 The addressability of a particular object can be set with the
8588 @code{model} attribute.
8590 @item -mmodel=medium
8591 @opindex mmodel=medium
8592 Assume objects may be anywhere in the 32-bit address space (the compiler
8593 will generate @code{seth/add3} instructions to load their addresses), and
8594 assume all subroutines are reachable with the @code{bl} instruction.
8597 @opindex mmodel=large
8598 Assume objects may be anywhere in the 32-bit address space (the compiler
8599 will generate @code{seth/add3} instructions to load their addresses), and
8600 assume subroutines may not be reachable with the @code{bl} instruction
8601 (the compiler will generate the much slower @code{seth/add3/jl}
8602 instruction sequence).
8605 @opindex msdata=none
8606 Disable use of the small data area. Variables will be put into
8607 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
8608 @code{section} attribute has been specified).
8609 This is the default.
8611 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
8612 Objects may be explicitly put in the small data area with the
8613 @code{section} attribute using one of these sections.
8616 @opindex msdata=sdata
8617 Put small global and static data in the small data area, but do not
8618 generate special code to reference them.
8622 Put small global and static data in the small data area, and generate
8623 special instructions to reference them.
8627 @cindex smaller data references
8628 Put global and static objects less than or equal to @var{num} bytes
8629 into the small data or bss sections instead of the normal data or bss
8630 sections. The default value of @var{num} is 8.
8631 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
8632 for this option to have any effect.
8634 All modules should be compiled with the same @option{-G @var{num}} value.
8635 Compiling with different values of @var{num} may or may not work; if it
8636 doesn't the linker will give an error message---incorrect code will not be
8641 Makes the M32R specific code in the compiler display some statistics
8642 that might help in debugging programs.
8645 @opindex malign-loops
8646 Align all loops to a 32-byte boundary.
8648 @item -mno-align-loops
8649 @opindex mno-align-loops
8650 Do not enforce a 32-byte alignment for loops. This is the default.
8652 @item -missue-rate=@var{number}
8653 @opindex missue-rate=@var{number}
8654 Issue @var{number} instructions per cycle. @var{number} can only be 1
8657 @item -mbranch-cost=@var{number}
8658 @opindex mbranch-cost=@var{number}
8659 @var{number} can only be 1 or 2. If it is 1 then branches will be
8660 preferred over conditional code, if it is 2, then the opposite will
8663 @item -mflush-trap=@var{number}
8664 @opindex mflush-trap=@var{number}
8665 Specifies the trap number to use to flush the cache. The default is
8666 12. Valid numbers are between 0 and 15 inclusive.
8668 @item -mno-flush-trap
8669 @opindex mno-flush-trap
8670 Specifies that the cache cannot be flushed by using a trap.
8672 @item -mflush-func=@var{name}
8673 @opindex mflush-func=@var{name}
8674 Specifies the name of the operating system function to call to flush
8675 the cache. The default is @emph{_flush_cache}, but a function call
8676 will only be used if a trap is not available.
8678 @item -mno-flush-func
8679 @opindex mno-flush-func
8680 Indicates that there is no OS function for flushing the cache.
8684 @node M680x0 Options
8685 @subsection M680x0 Options
8686 @cindex M680x0 options
8688 These are the @samp{-m} options defined for the 68000 series. The default
8689 values for these options depends on which style of 68000 was selected when
8690 the compiler was configured; the defaults for the most common choices are
8698 Generate output for a 68000. This is the default
8699 when the compiler is configured for 68000-based systems.
8701 Use this option for microcontrollers with a 68000 or EC000 core,
8702 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
8708 Generate output for a 68020. This is the default
8709 when the compiler is configured for 68020-based systems.
8713 Generate output containing 68881 instructions for floating point.
8714 This is the default for most 68020 systems unless @option{--nfp} was
8715 specified when the compiler was configured.
8719 Generate output for a 68030. This is the default when the compiler is
8720 configured for 68030-based systems.
8724 Generate output for a 68040. This is the default when the compiler is
8725 configured for 68040-based systems.
8727 This option inhibits the use of 68881/68882 instructions that have to be
8728 emulated by software on the 68040. Use this option if your 68040 does not
8729 have code to emulate those instructions.
8733 Generate output for a 68060. This is the default when the compiler is
8734 configured for 68060-based systems.
8736 This option inhibits the use of 68020 and 68881/68882 instructions that
8737 have to be emulated by software on the 68060. Use this option if your 68060
8738 does not have code to emulate those instructions.
8742 Generate output for a CPU32. This is the default
8743 when the compiler is configured for CPU32-based systems.
8745 Use this option for microcontrollers with a
8746 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
8747 68336, 68340, 68341, 68349 and 68360.
8751 Generate output for a 520X ``coldfire'' family cpu. This is the default
8752 when the compiler is configured for 520X-based systems.
8754 Use this option for microcontroller with a 5200 core, including
8755 the MCF5202, MCF5203, MCF5204 and MCF5202.
8760 Generate output for a 68040, without using any of the new instructions.
8761 This results in code which can run relatively efficiently on either a
8762 68020/68881 or a 68030 or a 68040. The generated code does use the
8763 68881 instructions that are emulated on the 68040.
8767 Generate output for a 68060, without using any of the new instructions.
8768 This results in code which can run relatively efficiently on either a
8769 68020/68881 or a 68030 or a 68040. The generated code does use the
8770 68881 instructions that are emulated on the 68060.
8773 @opindex msoft-float
8774 Generate output containing library calls for floating point.
8775 @strong{Warning:} the requisite libraries are not available for all m68k
8776 targets. Normally the facilities of the machine's usual C compiler are
8777 used, but this can't be done directly in cross-compilation. You must
8778 make your own arrangements to provide suitable library functions for
8779 cross-compilation. The embedded targets @samp{m68k-*-aout} and
8780 @samp{m68k-*-coff} do provide software floating point support.
8784 Consider type @code{int} to be 16 bits wide, like @code{short int}.
8785 Additionally, parameters passed on the stack are also aligned to a
8786 16-bit boundary even on targets whose API mandates promotion to 32-bit.
8789 @opindex mnobitfield
8790 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
8791 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
8795 Do use the bit-field instructions. The @option{-m68020} option implies
8796 @option{-mbitfield}. This is the default if you use a configuration
8797 designed for a 68020.
8801 Use a different function-calling convention, in which functions
8802 that take a fixed number of arguments return with the @code{rtd}
8803 instruction, which pops their arguments while returning. This
8804 saves one instruction in the caller since there is no need to pop
8805 the arguments there.
8807 This calling convention is incompatible with the one normally
8808 used on Unix, so you cannot use it if you need to call libraries
8809 compiled with the Unix compiler.
8811 Also, you must provide function prototypes for all functions that
8812 take variable numbers of arguments (including @code{printf});
8813 otherwise incorrect code will be generated for calls to those
8816 In addition, seriously incorrect code will result if you call a
8817 function with too many arguments. (Normally, extra arguments are
8818 harmlessly ignored.)
8820 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
8821 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
8824 @itemx -mno-align-int
8826 @opindex mno-align-int
8827 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
8828 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
8829 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
8830 Aligning variables on 32-bit boundaries produces code that runs somewhat
8831 faster on processors with 32-bit busses at the expense of more memory.
8833 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
8834 align structures containing the above types differently than
8835 most published application binary interface specifications for the m68k.
8839 Use the pc-relative addressing mode of the 68000 directly, instead of
8840 using a global offset table. At present, this option implies @option{-fpic},
8841 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
8842 not presently supported with @option{-mpcrel}, though this could be supported for
8843 68020 and higher processors.
8845 @item -mno-strict-align
8846 @itemx -mstrict-align
8847 @opindex mno-strict-align
8848 @opindex mstrict-align
8849 Do not (do) assume that unaligned memory references will be handled by
8853 Generate code that allows the data segment to be located in a different
8854 area of memory from the text segment. This allows for execute in place in
8855 an environment without virtual memory management. This option implies -fPIC.
8858 Generate code that assumes that the data segment follows the text segment.
8859 This is the default.
8861 @item -mid-shared-library
8862 Generate code that supports shared libraries via the library ID method.
8863 This allows for execute in place and shared libraries in an environment
8864 without virtual memory management. This option implies -fPIC.
8866 @item -mno-id-shared-library
8867 Generate code that doesn't assume ID based shared libraries are being used.
8868 This is the default.
8870 @item -mshared-library-id=n
8871 Specified the identification number of the ID based shared library being
8872 compiled. Specifying a value of 0 will generate more compact code, specifying
8873 other values will force the allocation of that number to the current
8874 library but is no more space or time efficient than omitting this option.
8878 @node M68hc1x Options
8879 @subsection M68hc1x Options
8880 @cindex M68hc1x options
8882 These are the @samp{-m} options defined for the 68hc11 and 68hc12
8883 microcontrollers. The default values for these options depends on
8884 which style of microcontroller was selected when the compiler was configured;
8885 the defaults for the most common choices are given below.
8892 Generate output for a 68HC11. This is the default
8893 when the compiler is configured for 68HC11-based systems.
8899 Generate output for a 68HC12. This is the default
8900 when the compiler is configured for 68HC12-based systems.
8906 Generate output for a 68HCS12.
8909 @opindex mauto-incdec
8910 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
8917 Enable the use of 68HC12 min and max instructions.
8920 @itemx -mno-long-calls
8921 @opindex mlong-calls
8922 @opindex mno-long-calls
8923 Treat all calls as being far away (near). If calls are assumed to be
8924 far away, the compiler will use the @code{call} instruction to
8925 call a function and the @code{rtc} instruction for returning.
8929 Consider type @code{int} to be 16 bits wide, like @code{short int}.
8931 @item -msoft-reg-count=@var{count}
8932 @opindex msoft-reg-count
8933 Specify the number of pseudo-soft registers which are used for the
8934 code generation. The maximum number is 32. Using more pseudo-soft
8935 register may or may not result in better code depending on the program.
8936 The default is 4 for 68HC11 and 2 for 68HC12.
8941 @subsection MCore Options
8942 @cindex MCore options
8944 These are the @samp{-m} options defined for the Motorola M*Core
8952 @opindex mno-hardlit
8953 Inline constants into the code stream if it can be done in two
8954 instructions or less.
8960 Use the divide instruction. (Enabled by default).
8962 @item -mrelax-immediate
8963 @itemx -mno-relax-immediate
8964 @opindex mrelax-immediate
8965 @opindex mno-relax-immediate
8966 Allow arbitrary sized immediates in bit operations.
8968 @item -mwide-bitfields
8969 @itemx -mno-wide-bitfields
8970 @opindex mwide-bitfields
8971 @opindex mno-wide-bitfields
8972 Always treat bit-fields as int-sized.
8974 @item -m4byte-functions
8975 @itemx -mno-4byte-functions
8976 @opindex m4byte-functions
8977 @opindex mno-4byte-functions
8978 Force all functions to be aligned to a four byte boundary.
8980 @item -mcallgraph-data
8981 @itemx -mno-callgraph-data
8982 @opindex mcallgraph-data
8983 @opindex mno-callgraph-data
8984 Emit callgraph information.
8987 @itemx -mno-slow-bytes
8988 @opindex mslow-bytes
8989 @opindex mno-slow-bytes
8990 Prefer word access when reading byte quantities.
8992 @item -mlittle-endian
8994 @opindex mlittle-endian
8995 @opindex mbig-endian
8996 Generate code for a little endian target.
9002 Generate code for the 210 processor.
9006 @subsection MIPS Options
9007 @cindex MIPS options
9013 Generate big-endian code.
9017 Generate little-endian code. This is the default for @samp{mips*el-*-*}
9020 @item -march=@var{arch}
9022 Generate code that will run on @var{arch}, which can be the name of a
9023 generic MIPS ISA, or the name of a particular processor.
9025 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
9026 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
9027 The processor names are:
9028 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
9030 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
9031 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
9035 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
9036 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
9037 The special value @samp{from-abi} selects the
9038 most compatible architecture for the selected ABI (that is,
9039 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
9041 In processor names, a final @samp{000} can be abbreviated as @samp{k}
9042 (for example, @samp{-march=r2k}). Prefixes are optional, and
9043 @samp{vr} may be written @samp{r}.
9045 GCC defines two macros based on the value of this option. The first
9046 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
9047 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
9048 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
9049 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
9050 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
9052 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
9053 above. In other words, it will have the full prefix and will not
9054 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
9055 the macro names the resolved architecture (either @samp{"mips1"} or
9056 @samp{"mips3"}). It names the default architecture when no
9057 @option{-march} option is given.
9059 @item -mtune=@var{arch}
9061 Optimize for @var{arch}. Among other things, this option controls
9062 the way instructions are scheduled, and the perceived cost of arithmetic
9063 operations. The list of @var{arch} values is the same as for
9066 When this option is not used, GCC will optimize for the processor
9067 specified by @option{-march}. By using @option{-march} and
9068 @option{-mtune} together, it is possible to generate code that will
9069 run on a family of processors, but optimize the code for one
9070 particular member of that family.
9072 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
9073 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
9074 @samp{-march} ones described above.
9078 Equivalent to @samp{-march=mips1}.
9082 Equivalent to @samp{-march=mips2}.
9086 Equivalent to @samp{-march=mips3}.
9090 Equivalent to @samp{-march=mips4}.
9094 Equivalent to @samp{-march=mips32}.
9098 Equivalent to @samp{-march=mips32r2}.
9102 Equivalent to @samp{-march=mips64}.
9108 Use (do not use) the MIPS16 ISA.
9120 Generate code for the given ABI@.
9122 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
9123 generates 64-bit code when you select a 64-bit architecture, but you
9124 can use @option{-mgp32} to get 32-bit code instead.
9126 For information about the O64 ABI, see
9127 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
9130 @itemx -mno-abicalls
9132 @opindex mno-abicalls
9133 Generate (do not generate) SVR4-style position-independent code.
9134 @option{-mabicalls} is the default for SVR4-based systems.
9140 Lift (do not lift) the usual restrictions on the size of the global
9143 GCC normally uses a single instruction to load values from the GOT.
9144 While this is relatively efficient, it will only work if the GOT
9145 is smaller than about 64k. Anything larger will cause the linker
9146 to report an error such as:
9148 @cindex relocation truncated to fit (MIPS)
9150 relocation truncated to fit: R_MIPS_GOT16 foobar
9153 If this happens, you should recompile your code with @option{-mxgot}.
9154 It should then work with very large GOTs, although it will also be
9155 less efficient, since it will take three instructions to fetch the
9156 value of a global symbol.
9158 Note that some linkers can create multiple GOTs. If you have such a
9159 linker, you should only need to use @option{-mxgot} when a single object
9160 file accesses more than 64k's worth of GOT entries. Very few do.
9162 These options have no effect unless GCC is generating position
9167 Assume that general-purpose registers are 32 bits wide.
9171 Assume that general-purpose registers are 64 bits wide.
9175 Assume that floating-point registers are 32 bits wide.
9179 Assume that floating-point registers are 64 bits wide.
9182 @opindex mhard-float
9183 Use floating-point coprocessor instructions.
9186 @opindex msoft-float
9187 Do not use floating-point coprocessor instructions. Implement
9188 floating-point calculations using library calls instead.
9190 @item -msingle-float
9191 @opindex msingle-float
9192 Assume that the floating-point coprocessor only supports single-precision
9195 @itemx -mdouble-float
9196 @opindex mdouble-float
9197 Assume that the floating-point coprocessor supports double-precision
9198 operations. This is the default.
9202 Force @code{int} and @code{long} types to be 64 bits wide. See
9203 @option{-mlong32} for an explanation of the default and the way
9204 that the pointer size is determined.
9208 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
9209 an explanation of the default and the way that the pointer size is
9214 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
9216 The default size of @code{int}s, @code{long}s and pointers depends on
9217 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
9218 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
9219 32-bit @code{long}s. Pointers are the same size as @code{long}s,
9220 or the same size as integer registers, whichever is smaller.
9224 @cindex smaller data references (MIPS)
9225 @cindex gp-relative references (MIPS)
9226 Put global and static items less than or equal to @var{num} bytes into
9227 the small data or bss section instead of the normal data or bss section.
9228 This allows the data to be accessed using a single instruction.
9230 All modules should be compiled with the same @option{-G @var{num}}
9233 @item -membedded-data
9234 @itemx -mno-embedded-data
9235 @opindex membedded-data
9236 @opindex mno-embedded-data
9237 Allocate variables to the read-only data section first if possible, then
9238 next in the small data section if possible, otherwise in data. This gives
9239 slightly slower code than the default, but reduces the amount of RAM required
9240 when executing, and thus may be preferred for some embedded systems.
9242 @item -muninit-const-in-rodata
9243 @itemx -mno-uninit-const-in-rodata
9244 @opindex muninit-const-in-rodata
9245 @opindex mno-uninit-const-in-rodata
9246 Put uninitialized @code{const} variables in the read-only data section.
9247 This option is only meaningful in conjunction with @option{-membedded-data}.
9249 @item -msplit-addresses
9250 @itemx -mno-split-addresses
9251 @opindex msplit-addresses
9252 @opindex mno-split-addresses
9253 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
9254 relocation operators. This option has been superceded by
9255 @option{-mexplicit-relocs} but is retained for backwards compatibility.
9257 @item -mexplicit-relocs
9258 @itemx -mno-explicit-relocs
9259 @opindex mexplicit-relocs
9260 @opindex mno-explicit-relocs
9261 Use (do not use) assembler relocation operators when dealing with symbolic
9262 addresses. The alternative, selected by @option{-mno-explicit-relocs},
9263 is to use assembler macros instead.
9265 @option{-mexplicit-relocs} is the default if GCC was configured
9266 to use an assembler that supports relocation operators.
9268 @item -mcheck-zero-division
9269 @itemx -mno-check-zero-division
9270 @opindex mcheck-zero-division
9271 @opindex mno-check-zero-division
9272 Trap (do not trap) on integer division by zero. The default is
9273 @option{-mcheck-zero-division}.
9279 Force (do not force) the use of @code{memcpy()} for non-trivial block
9280 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
9281 most constant-sized copies.
9284 @itemx -mno-long-calls
9285 @opindex mlong-calls
9286 @opindex mno-long-calls
9287 Disable (do not disable) use of the @code{jal} instruction. Calling
9288 functions using @code{jal} is more efficient but requires the caller
9289 and callee to be in the same 256 megabyte segment.
9291 This option has no effect on abicalls code. The default is
9292 @option{-mno-long-calls}.
9298 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
9299 instructions, as provided by the R4650 ISA.
9302 @itemx -mno-fused-madd
9303 @opindex mfused-madd
9304 @opindex mno-fused-madd
9305 Enable (disable) use of the floating point multiply-accumulate
9306 instructions, when they are available. The default is
9307 @option{-mfused-madd}.
9309 When multiply-accumulate instructions are used, the intermediate
9310 product is calculated to infinite precision and is not subject to
9311 the FCSR Flush to Zero bit. This may be undesirable in some
9316 Tell the MIPS assembler to not run its preprocessor over user
9317 assembler files (with a @samp{.s} suffix) when assembling them.
9320 @itemx -mno-fix-r4000
9322 @opindex mno-fix-r4000
9323 Work around certain R4000 CPU errata:
9326 A double-word or a variable shift may give an incorrect result if executed
9327 immediately after starting an integer division.
9329 A double-word or a variable shift may give an incorrect result if executed
9330 while an integer multiplication is in progress.
9332 An integer division may give an incorrect result if started in a delay slot
9333 of a taken branch or a jump.
9337 @itemx -mno-fix-r4400
9339 @opindex mno-fix-r4400
9340 Work around certain R4400 CPU errata:
9343 A double-word or a variable shift may give an incorrect result if executed
9344 immediately after starting an integer division.
9348 @itemx -mno-fix-vr4120
9349 @opindex mfix-vr4120
9350 Work around certain VR4120 errata:
9353 @code{dmultu} does not always produce the correct result.
9355 @code{div} and @code{ddiv} do not always produce the correct result if one
9356 of the operands is negative.
9358 The workarounds for the division errata rely on special functions in
9359 @file{libgcc.a}. At present, these functions are only provided by
9360 the @code{mips64vr*-elf} configurations.
9362 Other VR4120 errata require a nop to be inserted between certain pairs of
9363 instructions. These errata are handled by the assembler, not by GCC itself.
9368 Work around certain SB-1 CPU core errata.
9369 (This flag currently works around the SB-1 revision 2
9370 ``F1'' and ``F2'' floating point errata.)
9372 @item -mflush-func=@var{func}
9373 @itemx -mno-flush-func
9374 @opindex mflush-func
9375 Specifies the function to call to flush the I and D caches, or to not
9376 call any such function. If called, the function must take the same
9377 arguments as the common @code{_flush_func()}, that is, the address of the
9378 memory range for which the cache is being flushed, the size of the
9379 memory range, and the number 3 (to flush both caches). The default
9380 depends on the target GCC was configured for, but commonly is either
9381 @samp{_flush_func} or @samp{__cpu_flush}.
9383 @item -mbranch-likely
9384 @itemx -mno-branch-likely
9385 @opindex mbranch-likely
9386 @opindex mno-branch-likely
9387 Enable or disable use of Branch Likely instructions, regardless of the
9388 default for the selected architecture. By default, Branch Likely
9389 instructions may be generated if they are supported by the selected
9390 architecture. An exception is for the MIPS32 and MIPS64 architectures
9391 and processors which implement those architectures; for those, Branch
9392 Likely instructions will not be generated by default because the MIPS32
9393 and MIPS64 architectures specifically deprecate their use.
9395 @item -mfp-exceptions
9396 @itemx -mno-fp-exceptions
9397 @opindex mfp-exceptions
9398 Specifies whether FP exceptions are enabled. This affects how we schedule
9399 FP instructions for some processors. The default is that FP exceptions are
9402 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
9403 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
9406 @item -mvr4130-align
9407 @itemx -mno-vr4130-align
9408 @opindex mvr4130-align
9409 The VR4130 pipeline is two-way superscalar, but can only issue two
9410 instructions together if the first one is 8-byte aligned. When this
9411 option is enabled, GCC will align pairs of instructions that it
9412 thinks should execute in parallel.
9414 This option only has an effect when optimizing for the VR4130.
9415 It normally makes code faster, but at the expense of making it bigger.
9416 It is enabled by default at optimization level @option{-O3}.
9420 @subsection MMIX Options
9421 @cindex MMIX Options
9423 These options are defined for the MMIX:
9427 @itemx -mno-libfuncs
9429 @opindex mno-libfuncs
9430 Specify that intrinsic library functions are being compiled, passing all
9431 values in registers, no matter the size.
9436 @opindex mno-epsilon
9437 Generate floating-point comparison instructions that compare with respect
9438 to the @code{rE} epsilon register.
9440 @item -mabi=mmixware
9442 @opindex mabi-mmixware
9444 Generate code that passes function parameters and return values that (in
9445 the called function) are seen as registers @code{$0} and up, as opposed to
9446 the GNU ABI which uses global registers @code{$231} and up.
9449 @itemx -mno-zero-extend
9450 @opindex mzero-extend
9451 @opindex mno-zero-extend
9452 When reading data from memory in sizes shorter than 64 bits, use (do not
9453 use) zero-extending load instructions by default, rather than
9454 sign-extending ones.
9457 @itemx -mno-knuthdiv
9459 @opindex mno-knuthdiv
9460 Make the result of a division yielding a remainder have the same sign as
9461 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
9462 remainder follows the sign of the dividend. Both methods are
9463 arithmetically valid, the latter being almost exclusively used.
9465 @item -mtoplevel-symbols
9466 @itemx -mno-toplevel-symbols
9467 @opindex mtoplevel-symbols
9468 @opindex mno-toplevel-symbols
9469 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
9470 code can be used with the @code{PREFIX} assembly directive.
9474 Generate an executable in the ELF format, rather than the default
9475 @samp{mmo} format used by the @command{mmix} simulator.
9477 @item -mbranch-predict
9478 @itemx -mno-branch-predict
9479 @opindex mbranch-predict
9480 @opindex mno-branch-predict
9481 Use (do not use) the probable-branch instructions, when static branch
9482 prediction indicates a probable branch.
9484 @item -mbase-addresses
9485 @itemx -mno-base-addresses
9486 @opindex mbase-addresses
9487 @opindex mno-base-addresses
9488 Generate (do not generate) code that uses @emph{base addresses}. Using a
9489 base address automatically generates a request (handled by the assembler
9490 and the linker) for a constant to be set up in a global register. The
9491 register is used for one or more base address requests within the range 0
9492 to 255 from the value held in the register. The generally leads to short
9493 and fast code, but the number of different data items that can be
9494 addressed is limited. This means that a program that uses lots of static
9495 data may require @option{-mno-base-addresses}.
9498 @itemx -mno-single-exit
9499 @opindex msingle-exit
9500 @opindex mno-single-exit
9501 Force (do not force) generated code to have a single exit point in each
9505 @node MN10300 Options
9506 @subsection MN10300 Options
9507 @cindex MN10300 options
9509 These @option{-m} options are defined for Matsushita MN10300 architectures:
9514 Generate code to avoid bugs in the multiply instructions for the MN10300
9515 processors. This is the default.
9518 @opindex mno-mult-bug
9519 Do not generate code to avoid bugs in the multiply instructions for the
9524 Generate code which uses features specific to the AM33 processor.
9528 Do not generate code which uses features specific to the AM33 processor. This
9533 Do not link in the C run-time initialization object file.
9537 Indicate to the linker that it should perform a relaxation optimization pass
9538 to shorten branches, calls and absolute memory addresses. This option only
9539 has an effect when used on the command line for the final link step.
9541 This option makes symbolic debugging impossible.
9545 @subsection NS32K Options
9546 @cindex NS32K options
9548 These are the @samp{-m} options defined for the 32000 series. The default
9549 values for these options depends on which style of 32000 was selected when
9550 the compiler was configured; the defaults for the most common choices are
9558 Generate output for a 32032. This is the default
9559 when the compiler is configured for 32032 and 32016 based systems.
9565 Generate output for a 32332. This is the default
9566 when the compiler is configured for 32332-based systems.
9572 Generate output for a 32532. This is the default
9573 when the compiler is configured for 32532-based systems.
9577 Generate output containing 32081 instructions for floating point.
9578 This is the default for all systems.
9582 Generate output containing 32381 instructions for floating point. This
9583 also implies @option{-m32081}. The 32381 is only compatible with the 32332
9584 and 32532 cpus. This is the default for the pc532-netbsd configuration.
9588 Try and generate multiply-add floating point instructions @code{polyF}
9589 and @code{dotF}. This option is only available if the @option{-m32381}
9590 option is in effect. Using these instructions requires changes to
9591 register allocation which generally has a negative impact on
9592 performance. This option should only be enabled when compiling code
9593 particularly likely to make heavy use of multiply-add instructions.
9596 @opindex mnomulti-add
9597 Do not try and generate multiply-add floating point instructions
9598 @code{polyF} and @code{dotF}. This is the default on all platforms.
9601 @opindex msoft-float
9602 Generate output containing library calls for floating point.
9603 @strong{Warning:} the requisite libraries may not be available.
9605 @item -mieee-compare
9606 @itemx -mno-ieee-compare
9607 @opindex mieee-compare
9608 @opindex mno-ieee-compare
9609 Control whether or not the compiler uses IEEE floating point
9610 comparisons. These handle correctly the case where the result of a
9611 comparison is unordered.
9612 @strong{Warning:} the requisite kernel support may not be available.
9615 @opindex mnobitfield
9616 Do not use the bit-field instructions. On some machines it is faster to
9617 use shifting and masking operations. This is the default for the pc532.
9621 Do use the bit-field instructions. This is the default for all platforms
9626 Use a different function-calling convention, in which functions
9627 that take a fixed number of arguments return pop their
9628 arguments on return with the @code{ret} instruction.
9630 This calling convention is incompatible with the one normally
9631 used on Unix, so you cannot use it if you need to call libraries
9632 compiled with the Unix compiler.
9634 Also, you must provide function prototypes for all functions that
9635 take variable numbers of arguments (including @code{printf});
9636 otherwise incorrect code will be generated for calls to those
9639 In addition, seriously incorrect code will result if you call a
9640 function with too many arguments. (Normally, extra arguments are
9641 harmlessly ignored.)
9643 This option takes its name from the 680x0 @code{rtd} instruction.
9648 Use a different function-calling convention where the first two arguments
9649 are passed in registers.
9651 This calling convention is incompatible with the one normally
9652 used on Unix, so you cannot use it if you need to call libraries
9653 compiled with the Unix compiler.
9656 @opindex mnoregparam
9657 Do not pass any arguments in registers. This is the default for all
9662 It is OK to use the sb as an index register which is always loaded with
9663 zero. This is the default for the pc532-netbsd target.
9667 The sb register is not available for use or has not been initialized to
9668 zero by the run time system. This is the default for all targets except
9669 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
9670 @option{-fpic} is set.
9674 Many ns32000 series addressing modes use displacements of up to 512MB@.
9675 If an address is above 512MB then displacements from zero can not be used.
9676 This option causes code to be generated which can be loaded above 512MB@.
9677 This may be useful for operating systems or ROM code.
9681 Assume code will be loaded in the first 512MB of virtual address space.
9682 This is the default for all platforms.
9686 @node PDP-11 Options
9687 @subsection PDP-11 Options
9688 @cindex PDP-11 Options
9690 These options are defined for the PDP-11:
9695 Use hardware FPP floating point. This is the default. (FIS floating
9696 point on the PDP-11/40 is not supported.)
9699 @opindex msoft-float
9700 Do not use hardware floating point.
9704 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
9708 Return floating-point results in memory. This is the default.
9712 Generate code for a PDP-11/40.
9716 Generate code for a PDP-11/45. This is the default.
9720 Generate code for a PDP-11/10.
9722 @item -mbcopy-builtin
9723 @opindex bcopy-builtin
9724 Use inline @code{movmemhi} patterns for copying memory. This is the
9729 Do not use inline @code{movmemhi} patterns for copying memory.
9735 Use 16-bit @code{int}. This is the default.
9741 Use 32-bit @code{int}.
9746 @opindex mno-float32
9747 Use 64-bit @code{float}. This is the default.
9752 @opindex mno-float64
9753 Use 32-bit @code{float}.
9757 Use @code{abshi2} pattern. This is the default.
9761 Do not use @code{abshi2} pattern.
9763 @item -mbranch-expensive
9764 @opindex mbranch-expensive
9765 Pretend that branches are expensive. This is for experimenting with
9766 code generation only.
9768 @item -mbranch-cheap
9769 @opindex mbranch-cheap
9770 Do not pretend that branches are expensive. This is the default.
9774 Generate code for a system with split I&D.
9778 Generate code for a system without split I&D. This is the default.
9782 Use Unix assembler syntax. This is the default when configured for
9787 Use DEC assembler syntax. This is the default when configured for any
9788 PDP-11 target other than @samp{pdp11-*-bsd}.
9791 @node PowerPC Options
9792 @subsection PowerPC Options
9793 @cindex PowerPC options
9795 These are listed under @xref{RS/6000 and PowerPC Options}.
9797 @node RS/6000 and PowerPC Options
9798 @subsection IBM RS/6000 and PowerPC Options
9799 @cindex RS/6000 and PowerPC Options
9800 @cindex IBM RS/6000 and PowerPC Options
9802 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
9810 @itemx -mpowerpc-gpopt
9811 @itemx -mno-powerpc-gpopt
9812 @itemx -mpowerpc-gfxopt
9813 @itemx -mno-powerpc-gfxopt
9815 @itemx -mno-powerpc64
9821 @opindex mno-powerpc
9822 @opindex mpowerpc-gpopt
9823 @opindex mno-powerpc-gpopt
9824 @opindex mpowerpc-gfxopt
9825 @opindex mno-powerpc-gfxopt
9827 @opindex mno-powerpc64
9828 GCC supports two related instruction set architectures for the
9829 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
9830 instructions supported by the @samp{rios} chip set used in the original
9831 RS/6000 systems and the @dfn{PowerPC} instruction set is the
9832 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
9833 the IBM 4xx microprocessors.
9835 Neither architecture is a subset of the other. However there is a
9836 large common subset of instructions supported by both. An MQ
9837 register is included in processors supporting the POWER architecture.
9839 You use these options to specify which instructions are available on the
9840 processor you are using. The default value of these options is
9841 determined when configuring GCC@. Specifying the
9842 @option{-mcpu=@var{cpu_type}} overrides the specification of these
9843 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
9844 rather than the options listed above.
9846 The @option{-mpower} option allows GCC to generate instructions that
9847 are found only in the POWER architecture and to use the MQ register.
9848 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
9849 to generate instructions that are present in the POWER2 architecture but
9850 not the original POWER architecture.
9852 The @option{-mpowerpc} option allows GCC to generate instructions that
9853 are found only in the 32-bit subset of the PowerPC architecture.
9854 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
9855 GCC to use the optional PowerPC architecture instructions in the
9856 General Purpose group, including floating-point square root. Specifying
9857 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
9858 use the optional PowerPC architecture instructions in the Graphics
9859 group, including floating-point select.
9861 The @option{-mpowerpc64} option allows GCC to generate the additional
9862 64-bit instructions that are found in the full PowerPC64 architecture
9863 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
9864 @option{-mno-powerpc64}.
9866 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
9867 will use only the instructions in the common subset of both
9868 architectures plus some special AIX common-mode calls, and will not use
9869 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
9870 permits GCC to use any instruction from either architecture and to
9871 allow use of the MQ register; specify this for the Motorola MPC601.
9873 @item -mnew-mnemonics
9874 @itemx -mold-mnemonics
9875 @opindex mnew-mnemonics
9876 @opindex mold-mnemonics
9877 Select which mnemonics to use in the generated assembler code. With
9878 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
9879 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
9880 assembler mnemonics defined for the POWER architecture. Instructions
9881 defined in only one architecture have only one mnemonic; GCC uses that
9882 mnemonic irrespective of which of these options is specified.
9884 GCC defaults to the mnemonics appropriate for the architecture in
9885 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
9886 value of these option. Unless you are building a cross-compiler, you
9887 should normally not specify either @option{-mnew-mnemonics} or
9888 @option{-mold-mnemonics}, but should instead accept the default.
9890 @item -mcpu=@var{cpu_type}
9892 Set architecture type, register usage, choice of mnemonics, and
9893 instruction scheduling parameters for machine type @var{cpu_type}.
9894 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
9895 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
9896 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
9897 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
9898 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
9899 @samp{860}, @samp{970}, @samp{common}, @samp{ec603e}, @samp{G3},
9900 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
9901 @samp{power4}, @samp{power5}, @samp{powerpc}, @samp{powerpc64},
9902 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64a}.
9904 @option{-mcpu=common} selects a completely generic processor. Code
9905 generated under this option will run on any POWER or PowerPC processor.
9906 GCC will use only the instructions in the common subset of both
9907 architectures, and will not use the MQ register. GCC assumes a generic
9908 processor model for scheduling purposes.
9910 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
9911 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
9912 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
9913 types, with an appropriate, generic processor model assumed for
9914 scheduling purposes.
9916 The other options specify a specific processor. Code generated under
9917 those options will run best on that processor, and may not run at all on
9920 The @option{-mcpu} options automatically enable or disable the
9921 following options: @option{-maltivec}, @option{-mhard-float},
9922 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
9923 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
9924 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
9925 @option{-mstring}. The particular options set for any particular CPU
9926 will vary between compiler versions, depending on what setting seems
9927 to produce optimal code for that CPU; it doesn't necessarily reflect
9928 the actual hardware's capabilities. If you wish to set an individual
9929 option to a particular value, you may specify it after the
9930 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
9932 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
9933 not enabled or disabled by the @option{-mcpu} option at present, since
9934 AIX does not have full support for these options. You may still
9935 enable or disable them individually if you're sure it'll work in your
9938 @item -mtune=@var{cpu_type}
9940 Set the instruction scheduling parameters for machine type
9941 @var{cpu_type}, but do not set the architecture type, register usage, or
9942 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
9943 values for @var{cpu_type} are used for @option{-mtune} as for
9944 @option{-mcpu}. If both are specified, the code generated will use the
9945 architecture, registers, and mnemonics set by @option{-mcpu}, but the
9946 scheduling parameters set by @option{-mtune}.
9951 @opindex mno-altivec
9952 These switches enable or disable the use of built-in functions that
9953 allow access to the AltiVec instruction set. You may also need to set
9954 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
9959 Extend the current ABI with SPE ABI extensions. This does not change
9960 the default ABI, instead it adds the SPE ABI extensions to the current
9964 @opindex mabi=no-spe
9965 Disable Booke SPE ABI extensions for the current ABI.
9967 @item -misel=@var{yes/no}
9970 This switch enables or disables the generation of ISEL instructions.
9972 @item -mspe=@var{yes/no}
9975 This switch enables or disables the generation of SPE simd
9978 @item -mfloat-gprs=@var{yes/no}
9980 @opindex mfloat-gprs
9981 This switch enables or disables the generation of floating point
9982 operations on the general purpose registers for architectures that
9983 support it. This option is currently only available on the MPC8540.
9989 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
9990 targets (including GNU/Linux). The 32-bit environment sets int, long
9991 and pointer to 32 bits and generates code that runs on any PowerPC
9992 variant. The 64-bit environment sets int to 32 bits and long and
9993 pointer to 64 bits, and generates code for PowerPC64, as for
9994 @option{-mpowerpc64}.
9997 @itemx -mno-fp-in-toc
9998 @itemx -mno-sum-in-toc
9999 @itemx -mminimal-toc
10001 @opindex mno-fp-in-toc
10002 @opindex mno-sum-in-toc
10003 @opindex mminimal-toc
10004 Modify generation of the TOC (Table Of Contents), which is created for
10005 every executable file. The @option{-mfull-toc} option is selected by
10006 default. In that case, GCC will allocate at least one TOC entry for
10007 each unique non-automatic variable reference in your program. GCC
10008 will also place floating-point constants in the TOC@. However, only
10009 16,384 entries are available in the TOC@.
10011 If you receive a linker error message that saying you have overflowed
10012 the available TOC space, you can reduce the amount of TOC space used
10013 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
10014 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
10015 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
10016 generate code to calculate the sum of an address and a constant at
10017 run-time instead of putting that sum into the TOC@. You may specify one
10018 or both of these options. Each causes GCC to produce very slightly
10019 slower and larger code at the expense of conserving TOC space.
10021 If you still run out of space in the TOC even when you specify both of
10022 these options, specify @option{-mminimal-toc} instead. This option causes
10023 GCC to make only one TOC entry for every file. When you specify this
10024 option, GCC will produce code that is slower and larger but which
10025 uses extremely little TOC space. You may wish to use this option
10026 only on files that contain less frequently executed code.
10032 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
10033 @code{long} type, and the infrastructure needed to support them.
10034 Specifying @option{-maix64} implies @option{-mpowerpc64} and
10035 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
10036 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
10039 @itemx -mno-xl-call
10041 @opindex mno-xl-call
10042 On AIX, pass floating-point arguments to prototyped functions beyond the
10043 register save area (RSA) on the stack in addition to argument FPRs. The
10044 AIX calling convention was extended but not initially documented to
10045 handle an obscure K&R C case of calling a function that takes the
10046 address of its arguments with fewer arguments than declared. AIX XL
10047 compilers access floating point arguments which do not fit in the
10048 RSA from the stack when a subroutine is compiled without
10049 optimization. Because always storing floating-point arguments on the
10050 stack is inefficient and rarely needed, this option is not enabled by
10051 default and only is necessary when calling subroutines compiled by AIX
10052 XL compilers without optimization.
10056 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
10057 application written to use message passing with special startup code to
10058 enable the application to run. The system must have PE installed in the
10059 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
10060 must be overridden with the @option{-specs=} option to specify the
10061 appropriate directory location. The Parallel Environment does not
10062 support threads, so the @option{-mpe} option and the @option{-pthread}
10063 option are incompatible.
10065 @item -malign-natural
10066 @itemx -malign-power
10067 @opindex malign-natural
10068 @opindex malign-power
10069 On AIX, Darwin, and 64-bit PowerPC GNU/Linux, the option
10070 @option{-malign-natural} overrides the ABI-defined alignment of larger
10071 types, such as floating-point doubles, on their natural size-based boundary.
10072 The option @option{-malign-power} instructs GCC to follow the ABI-specified
10073 alignment rules. GCC defaults to the standard alignment defined in the ABI.
10076 @itemx -mhard-float
10077 @opindex msoft-float
10078 @opindex mhard-float
10079 Generate code that does not use (uses) the floating-point register set.
10080 Software floating point emulation is provided if you use the
10081 @option{-msoft-float} option, and pass the option to GCC when linking.
10084 @itemx -mno-multiple
10086 @opindex mno-multiple
10087 Generate code that uses (does not use) the load multiple word
10088 instructions and the store multiple word instructions. These
10089 instructions are generated by default on POWER systems, and not
10090 generated on PowerPC systems. Do not use @option{-mmultiple} on little
10091 endian PowerPC systems, since those instructions do not work when the
10092 processor is in little endian mode. The exceptions are PPC740 and
10093 PPC750 which permit the instructions usage in little endian mode.
10098 @opindex mno-string
10099 Generate code that uses (does not use) the load string instructions
10100 and the store string word instructions to save multiple registers and
10101 do small block moves. These instructions are generated by default on
10102 POWER systems, and not generated on PowerPC systems. Do not use
10103 @option{-mstring} on little endian PowerPC systems, since those
10104 instructions do not work when the processor is in little endian mode.
10105 The exceptions are PPC740 and PPC750 which permit the instructions
10106 usage in little endian mode.
10111 @opindex mno-update
10112 Generate code that uses (does not use) the load or store instructions
10113 that update the base register to the address of the calculated memory
10114 location. These instructions are generated by default. If you use
10115 @option{-mno-update}, there is a small window between the time that the
10116 stack pointer is updated and the address of the previous frame is
10117 stored, which means code that walks the stack frame across interrupts or
10118 signals may get corrupted data.
10121 @itemx -mno-fused-madd
10122 @opindex mfused-madd
10123 @opindex mno-fused-madd
10124 Generate code that uses (does not use) the floating point multiply and
10125 accumulate instructions. These instructions are generated by default if
10126 hardware floating is used.
10128 @item -mno-bit-align
10130 @opindex mno-bit-align
10131 @opindex mbit-align
10132 On System V.4 and embedded PowerPC systems do not (do) force structures
10133 and unions that contain bit-fields to be aligned to the base type of the
10136 For example, by default a structure containing nothing but 8
10137 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
10138 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
10139 the structure would be aligned to a 1 byte boundary and be one byte in
10142 @item -mno-strict-align
10143 @itemx -mstrict-align
10144 @opindex mno-strict-align
10145 @opindex mstrict-align
10146 On System V.4 and embedded PowerPC systems do not (do) assume that
10147 unaligned memory references will be handled by the system.
10149 @item -mrelocatable
10150 @itemx -mno-relocatable
10151 @opindex mrelocatable
10152 @opindex mno-relocatable
10153 On embedded PowerPC systems generate code that allows (does not allow)
10154 the program to be relocated to a different address at runtime. If you
10155 use @option{-mrelocatable} on any module, all objects linked together must
10156 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
10158 @item -mrelocatable-lib
10159 @itemx -mno-relocatable-lib
10160 @opindex mrelocatable-lib
10161 @opindex mno-relocatable-lib
10162 On embedded PowerPC systems generate code that allows (does not allow)
10163 the program to be relocated to a different address at runtime. Modules
10164 compiled with @option{-mrelocatable-lib} can be linked with either modules
10165 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
10166 with modules compiled with the @option{-mrelocatable} options.
10172 On System V.4 and embedded PowerPC systems do not (do) assume that
10173 register 2 contains a pointer to a global area pointing to the addresses
10174 used in the program.
10177 @itemx -mlittle-endian
10179 @opindex mlittle-endian
10180 On System V.4 and embedded PowerPC systems compile code for the
10181 processor in little endian mode. The @option{-mlittle-endian} option is
10182 the same as @option{-mlittle}.
10185 @itemx -mbig-endian
10187 @opindex mbig-endian
10188 On System V.4 and embedded PowerPC systems compile code for the
10189 processor in big endian mode. The @option{-mbig-endian} option is
10190 the same as @option{-mbig}.
10192 @item -mdynamic-no-pic
10193 @opindex mdynamic-no-pic
10194 On Darwin and Mac OS X systems, compile code so that it is not
10195 relocatable, but that its external references are relocatable. The
10196 resulting code is suitable for applications, but not shared
10199 @item -mprioritize-restricted-insns=@var{priority}
10200 @opindex mprioritize-restricted-insns
10201 This option controls the priority that is assigned to
10202 dispatch-slot restricted instructions during the second scheduling
10203 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
10204 @var{no/highest/second-highest} priority to dispatch slot restricted
10207 @item -msched-costly-dep=@var{dependence_type}
10208 @opindex msched-costly-dep
10209 This option controls which dependences are considered costly
10210 by the target during instruction scheduling. The argument
10211 @var{dependence_type} takes one of the following values:
10212 @var{no}: no dependence is costly,
10213 @var{all}: all dependences are costly,
10214 @var{true_store_to_load}: a true dependence from store to load is costly,
10215 @var{store_to_load}: any dependence from store to load is costly,
10216 @var{number}: any dependence which latency >= @var{number} is costly.
10218 @item -minsert-sched-nops=@var{scheme}
10219 @opindex minsert-sched-nops
10220 This option controls which nop insertion scheme will be used during
10221 the second scheduling pass. The argument @var{scheme} takes one of the
10223 @var{no}: Don't insert nops.
10224 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
10225 according to the scheduler's grouping.
10226 @var{regroup_exact}: Insert nops to force costly dependent insns into
10227 separate groups. Insert exactly as many nops as needed to force an insn
10228 to a new group, according to the estimated processor grouping.
10229 @var{number}: Insert nops to force costly dependent insns into
10230 separate groups. Insert @var{number} nops to force an insn to a new group.
10233 @opindex mcall-sysv
10234 On System V.4 and embedded PowerPC systems compile code using calling
10235 conventions that adheres to the March 1995 draft of the System V
10236 Application Binary Interface, PowerPC processor supplement. This is the
10237 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
10239 @item -mcall-sysv-eabi
10240 @opindex mcall-sysv-eabi
10241 Specify both @option{-mcall-sysv} and @option{-meabi} options.
10243 @item -mcall-sysv-noeabi
10244 @opindex mcall-sysv-noeabi
10245 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
10247 @item -mcall-solaris
10248 @opindex mcall-solaris
10249 On System V.4 and embedded PowerPC systems compile code for the Solaris
10253 @opindex mcall-linux
10254 On System V.4 and embedded PowerPC systems compile code for the
10255 Linux-based GNU system.
10259 On System V.4 and embedded PowerPC systems compile code for the
10260 Hurd-based GNU system.
10262 @item -mcall-netbsd
10263 @opindex mcall-netbsd
10264 On System V.4 and embedded PowerPC systems compile code for the
10265 NetBSD operating system.
10267 @item -maix-struct-return
10268 @opindex maix-struct-return
10269 Return all structures in memory (as specified by the AIX ABI)@.
10271 @item -msvr4-struct-return
10272 @opindex msvr4-struct-return
10273 Return structures smaller than 8 bytes in registers (as specified by the
10276 @item -mabi=altivec
10277 @opindex mabi=altivec
10278 Extend the current ABI with AltiVec ABI extensions. This does not
10279 change the default ABI, instead it adds the AltiVec ABI extensions to
10282 @item -mabi=no-altivec
10283 @opindex mabi=no-altivec
10284 Disable AltiVec ABI extensions for the current ABI.
10287 @itemx -mno-prototype
10288 @opindex mprototype
10289 @opindex mno-prototype
10290 On System V.4 and embedded PowerPC systems assume that all calls to
10291 variable argument functions are properly prototyped. Otherwise, the
10292 compiler must insert an instruction before every non prototyped call to
10293 set or clear bit 6 of the condition code register (@var{CR}) to
10294 indicate whether floating point values were passed in the floating point
10295 registers in case the function takes a variable arguments. With
10296 @option{-mprototype}, only calls to prototyped variable argument functions
10297 will set or clear the bit.
10301 On embedded PowerPC systems, assume that the startup module is called
10302 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
10303 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
10308 On embedded PowerPC systems, assume that the startup module is called
10309 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
10314 On embedded PowerPC systems, assume that the startup module is called
10315 @file{crt0.o} and the standard C libraries are @file{libads.a} and
10318 @item -myellowknife
10319 @opindex myellowknife
10320 On embedded PowerPC systems, assume that the startup module is called
10321 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
10326 On System V.4 and embedded PowerPC systems, specify that you are
10327 compiling for a VxWorks system.
10331 Specify that you are compiling for the WindISS simulation environment.
10335 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
10336 header to indicate that @samp{eabi} extended relocations are used.
10342 On System V.4 and embedded PowerPC systems do (do not) adhere to the
10343 Embedded Applications Binary Interface (eabi) which is a set of
10344 modifications to the System V.4 specifications. Selecting @option{-meabi}
10345 means that the stack is aligned to an 8 byte boundary, a function
10346 @code{__eabi} is called to from @code{main} to set up the eabi
10347 environment, and the @option{-msdata} option can use both @code{r2} and
10348 @code{r13} to point to two separate small data areas. Selecting
10349 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
10350 do not call an initialization function from @code{main}, and the
10351 @option{-msdata} option will only use @code{r13} to point to a single
10352 small data area. The @option{-meabi} option is on by default if you
10353 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
10356 @opindex msdata=eabi
10357 On System V.4 and embedded PowerPC systems, put small initialized
10358 @code{const} global and static data in the @samp{.sdata2} section, which
10359 is pointed to by register @code{r2}. Put small initialized
10360 non-@code{const} global and static data in the @samp{.sdata} section,
10361 which is pointed to by register @code{r13}. Put small uninitialized
10362 global and static data in the @samp{.sbss} section, which is adjacent to
10363 the @samp{.sdata} section. The @option{-msdata=eabi} option is
10364 incompatible with the @option{-mrelocatable} option. The
10365 @option{-msdata=eabi} option also sets the @option{-memb} option.
10368 @opindex msdata=sysv
10369 On System V.4 and embedded PowerPC systems, put small global and static
10370 data in the @samp{.sdata} section, which is pointed to by register
10371 @code{r13}. Put small uninitialized global and static data in the
10372 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
10373 The @option{-msdata=sysv} option is incompatible with the
10374 @option{-mrelocatable} option.
10376 @item -msdata=default
10378 @opindex msdata=default
10380 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
10381 compile code the same as @option{-msdata=eabi}, otherwise compile code the
10382 same as @option{-msdata=sysv}.
10385 @opindex msdata-data
10386 On System V.4 and embedded PowerPC systems, put small global and static
10387 data in the @samp{.sdata} section. Put small uninitialized global and
10388 static data in the @samp{.sbss} section. Do not use register @code{r13}
10389 to address small data however. This is the default behavior unless
10390 other @option{-msdata} options are used.
10394 @opindex msdata=none
10396 On embedded PowerPC systems, put all initialized global and static data
10397 in the @samp{.data} section, and all uninitialized data in the
10398 @samp{.bss} section.
10402 @cindex smaller data references (PowerPC)
10403 @cindex .sdata/.sdata2 references (PowerPC)
10404 On embedded PowerPC systems, put global and static items less than or
10405 equal to @var{num} bytes into the small data or bss sections instead of
10406 the normal data or bss section. By default, @var{num} is 8. The
10407 @option{-G @var{num}} switch is also passed to the linker.
10408 All modules should be compiled with the same @option{-G @var{num}} value.
10411 @itemx -mno-regnames
10413 @opindex mno-regnames
10414 On System V.4 and embedded PowerPC systems do (do not) emit register
10415 names in the assembly language output using symbolic forms.
10418 @itemx -mno-longcall
10420 @opindex mno-longcall
10421 Default to making all function calls indirectly, using a register, so
10422 that functions which reside further than 32 megabytes (33,554,432
10423 bytes) from the current location can be called. This setting can be
10424 overridden by the @code{shortcall} function attribute, or by
10425 @code{#pragma longcall(0)}.
10427 Some linkers are capable of detecting out-of-range calls and generating
10428 glue code on the fly. On these systems, long calls are unnecessary and
10429 generate slower code. As of this writing, the AIX linker can do this,
10430 as can the GNU linker for PowerPC/64. It is planned to add this feature
10431 to the GNU linker for 32-bit PowerPC systems as well.
10433 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
10434 callee, L42'', plus a ``branch island'' (glue code). The two target
10435 addresses represent the callee and the ``branch island.'' The
10436 Darwin/PPC linker will prefer the first address and generate a ``bl
10437 callee'' if the PPC ``bl'' instruction will reach the callee directly;
10438 otherwise, the linker will generate ``bl L42'' to call the ``branch
10439 island.'' The ``branch island'' is appended to the body of the
10440 calling function; it computes the full 32-bit address of the callee
10443 On Mach-O (Darwin) systems, this option directs the compiler emit to
10444 the glue for every direct call, and the Darwin linker decides whether
10445 to use or discard it.
10447 In the future, we may cause GCC to ignore all longcall specifications
10448 when the linker is known to generate glue.
10452 Adds support for multithreading with the @dfn{pthreads} library.
10453 This option sets flags for both the preprocessor and linker.
10457 @node S/390 and zSeries Options
10458 @subsection S/390 and zSeries Options
10459 @cindex S/390 and zSeries Options
10461 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
10465 @itemx -msoft-float
10466 @opindex mhard-float
10467 @opindex msoft-float
10468 Use (do not use) the hardware floating-point instructions and registers
10469 for floating-point operations. When @option{-msoft-float} is specified,
10470 functions in @file{libgcc.a} will be used to perform floating-point
10471 operations. When @option{-mhard-float} is specified, the compiler
10472 generates IEEE floating-point instructions. This is the default.
10475 @itemx -mno-backchain
10476 @itemx -mkernel-backchain
10477 @opindex mbackchain
10478 @opindex mno-backchain
10479 @opindex mkernel-backchain
10480 In order to provide a backchain the address of the caller's frame
10481 is stored within the callee's stack frame.
10482 A backchain may be needed to allow debugging using tools that do not understand
10483 DWARF-2 call frame information.
10484 For @option{-mno-backchain} no backchain is maintained at all which is the
10486 If one of the other options is present the backchain pointer is placed either
10487 on top of the stack frame (@option{-mkernel-backchain}) or on
10488 the bottom (@option{-mbackchain}).
10489 Beside the different backchain location @option{-mkernel-backchain}
10490 also changes stack frame layout breaking the ABI. This option
10491 is intended to be used for code which internally needs a backchain but has
10492 to get by with a limited stack size e.g. the linux kernel.
10493 Internal unwinding code not using DWARF-2 info has to be able to locate the
10494 return address of a function. That will be eased be the fact that
10495 the return address of a function is placed two words below the backchain
10499 @itemx -mno-small-exec
10500 @opindex msmall-exec
10501 @opindex mno-small-exec
10502 Generate (or do not generate) code using the @code{bras} instruction
10503 to do subroutine calls.
10504 This only works reliably if the total executable size does not
10505 exceed 64k. The default is to use the @code{basr} instruction instead,
10506 which does not have this limitation.
10512 When @option{-m31} is specified, generate code compliant to the
10513 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
10514 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
10515 particular to generate 64-bit instructions. For the @samp{s390}
10516 targets, the default is @option{-m31}, while the @samp{s390x}
10517 targets default to @option{-m64}.
10523 When @option{-mzarch} is specified, generate code using the
10524 instructions available on z/Architecture.
10525 When @option{-mesa} is specified, generate code using the
10526 instructions available on ESA/390. Note that @option{-mesa} is
10527 not possible with @option{-m64}.
10528 When generating code compliant to the GNU/Linux for S/390 ABI,
10529 the default is @option{-mesa}. When generating code compliant
10530 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
10536 Generate (or do not generate) code using the @code{mvcle} instruction
10537 to perform block moves. When @option{-mno-mvcle} is specified,
10538 use a @code{mvc} loop instead. This is the default.
10544 Print (or do not print) additional debug information when compiling.
10545 The default is to not print debug information.
10547 @item -march=@var{cpu-type}
10549 Generate code that will run on @var{cpu-type}, which is the name of a system
10550 representing a certain processor type. Possible values for
10551 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
10552 When generating code using the instructions available on z/Architecture,
10553 the default is @option{-march=z900}. Otherwise, the default is
10554 @option{-march=g5}.
10556 @item -mtune=@var{cpu-type}
10558 Tune to @var{cpu-type} everything applicable about the generated code,
10559 except for the ABI and the set of available instructions.
10560 The list of @var{cpu-type} values is the same as for @option{-march}.
10561 The default is the value used for @option{-march}.
10564 @itemx -mno-tpf-trace
10565 @opindex mtpf-trace
10566 @opindex mno-tpf-trace
10567 Generate code that adds (does not add) in TPF OS specific branches to trace
10568 routines in the operating system. This option is off by default, even
10569 when compiling for the TPF OS.
10572 @itemx -mno-fused-madd
10573 @opindex mfused-madd
10574 @opindex mno-fused-madd
10575 Generate code that uses (does not use) the floating point multiply and
10576 accumulate instructions. These instructions are generated by default if
10577 hardware floating point is used.
10581 @subsection SH Options
10583 These @samp{-m} options are defined for the SH implementations:
10588 Generate code for the SH1.
10592 Generate code for the SH2.
10595 Generate code for the SH2e.
10599 Generate code for the SH3.
10603 Generate code for the SH3e.
10607 Generate code for the SH4 without a floating-point unit.
10609 @item -m4-single-only
10610 @opindex m4-single-only
10611 Generate code for the SH4 with a floating-point unit that only
10612 supports single-precision arithmetic.
10616 Generate code for the SH4 assuming the floating-point unit is in
10617 single-precision mode by default.
10621 Generate code for the SH4.
10625 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
10626 floating-point unit is not used.
10628 @item -m4a-single-only
10629 @opindex m4a-single-only
10630 Generate code for the SH4a, in such a way that no double-precision
10631 floating point operations are used.
10634 @opindex m4a-single
10635 Generate code for the SH4a assuming the floating-point unit is in
10636 single-precision mode by default.
10640 Generate code for the SH4a.
10644 Same as @option{-m4a-nofpu}, except that it implicitly passes
10645 @option{-dsp} to the assembler. GCC doesn't generate any DSP
10646 instructions at the moment.
10650 Compile code for the processor in big endian mode.
10654 Compile code for the processor in little endian mode.
10658 Align doubles at 64-bit boundaries. Note that this changes the calling
10659 conventions, and thus some functions from the standard C library will
10660 not work unless you recompile it first with @option{-mdalign}.
10664 Shorten some address references at link time, when possible; uses the
10665 linker option @option{-relax}.
10669 Use 32-bit offsets in @code{switch} tables. The default is to use
10674 Enable the use of the instruction @code{fmovd}.
10678 Comply with the calling conventions defined by Renesas.
10682 Comply with the calling conventions defined by Renesas.
10686 Comply with the calling conventions defined for GCC before the Renesas
10687 conventions were available. This option is the default for all
10688 targets of the SH toolchain except for @samp{sh-symbianelf}.
10691 @opindex mnomacsave
10692 Mark the @code{MAC} register as call-clobbered, even if
10693 @option{-mhitachi} is given.
10697 Increase IEEE-compliance of floating-point code.
10701 Dump instruction size and location in the assembly code.
10704 @opindex mpadstruct
10705 This option is deprecated. It pads structures to multiple of 4 bytes,
10706 which is incompatible with the SH ABI@.
10710 Optimize for space instead of speed. Implied by @option{-Os}.
10713 @opindex mprefergot
10714 When generating position-independent code, emit function calls using
10715 the Global Offset Table instead of the Procedure Linkage Table.
10719 Generate a library function call to invalidate instruction cache
10720 entries, after fixing up a trampoline. This library function call
10721 doesn't assume it can write to the whole memory address space. This
10722 is the default when the target is @code{sh-*-linux*}.
10725 @node SPARC Options
10726 @subsection SPARC Options
10727 @cindex SPARC options
10729 These @samp{-m} options are supported on the SPARC:
10732 @item -mno-app-regs
10734 @opindex mno-app-regs
10736 Specify @option{-mapp-regs} to generate output using the global registers
10737 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
10740 To be fully SVR4 ABI compliant at the cost of some performance loss,
10741 specify @option{-mno-app-regs}. You should compile libraries and system
10742 software with this option.
10745 @itemx -mhard-float
10747 @opindex mhard-float
10748 Generate output containing floating point instructions. This is the
10752 @itemx -msoft-float
10754 @opindex msoft-float
10755 Generate output containing library calls for floating point.
10756 @strong{Warning:} the requisite libraries are not available for all SPARC
10757 targets. Normally the facilities of the machine's usual C compiler are
10758 used, but this cannot be done directly in cross-compilation. You must make
10759 your own arrangements to provide suitable library functions for
10760 cross-compilation. The embedded targets @samp{sparc-*-aout} and
10761 @samp{sparclite-*-*} do provide software floating point support.
10763 @option{-msoft-float} changes the calling convention in the output file;
10764 therefore, it is only useful if you compile @emph{all} of a program with
10765 this option. In particular, you need to compile @file{libgcc.a}, the
10766 library that comes with GCC, with @option{-msoft-float} in order for
10769 @item -mhard-quad-float
10770 @opindex mhard-quad-float
10771 Generate output containing quad-word (long double) floating point
10774 @item -msoft-quad-float
10775 @opindex msoft-quad-float
10776 Generate output containing library calls for quad-word (long double)
10777 floating point instructions. The functions called are those specified
10778 in the SPARC ABI@. This is the default.
10780 As of this writing, there are no SPARC implementations that have hardware
10781 support for the quad-word floating point instructions. They all invoke
10782 a trap handler for one of these instructions, and then the trap handler
10783 emulates the effect of the instruction. Because of the trap handler overhead,
10784 this is much slower than calling the ABI library routines. Thus the
10785 @option{-msoft-quad-float} option is the default.
10787 @item -mno-unaligned-doubles
10788 @itemx -munaligned-doubles
10789 @opindex mno-unaligned-doubles
10790 @opindex munaligned-doubles
10791 Assume that doubles have 8 byte alignment. This is the default.
10793 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
10794 alignment only if they are contained in another type, or if they have an
10795 absolute address. Otherwise, it assumes they have 4 byte alignment.
10796 Specifying this option avoids some rare compatibility problems with code
10797 generated by other compilers. It is not the default because it results
10798 in a performance loss, especially for floating point code.
10800 @item -mno-faster-structs
10801 @itemx -mfaster-structs
10802 @opindex mno-faster-structs
10803 @opindex mfaster-structs
10804 With @option{-mfaster-structs}, the compiler assumes that structures
10805 should have 8 byte alignment. This enables the use of pairs of
10806 @code{ldd} and @code{std} instructions for copies in structure
10807 assignment, in place of twice as many @code{ld} and @code{st} pairs.
10808 However, the use of this changed alignment directly violates the SPARC
10809 ABI@. Thus, it's intended only for use on targets where the developer
10810 acknowledges that their resulting code will not be directly in line with
10811 the rules of the ABI@.
10813 @item -mimpure-text
10814 @opindex mimpure-text
10815 @option{-mimpure-text}, used in addition to @option{-shared}, tells
10816 the compiler to not pass @option{-z text} to the linker when linking a
10817 shared object. Using this option, you can link position-dependent
10818 code into a shared object.
10820 @option{-mimpure-text} suppresses the ``relocations remain against
10821 allocatable but non-writable sections'' linker error message.
10822 However, the necessary relocations will trigger copy-on-write, and the
10823 shared object is not actually shared across processes. Instead of
10824 using @option{-mimpure-text}, you should compile all source code with
10825 @option{-fpic} or @option{-fPIC}.
10827 This option is only available on SunOS and Solaris.
10829 @item -mcpu=@var{cpu_type}
10831 Set the instruction set, register set, and instruction scheduling parameters
10832 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
10833 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
10834 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
10835 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
10836 @samp{ultrasparc3}.
10838 Default instruction scheduling parameters are used for values that select
10839 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
10840 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
10842 Here is a list of each supported architecture and their supported
10847 v8: supersparc, hypersparc
10848 sparclite: f930, f934, sparclite86x
10850 v9: ultrasparc, ultrasparc3
10853 By default (unless configured otherwise), GCC generates code for the V7
10854 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
10855 additionally optimizes it for the Cypress CY7C602 chip, as used in the
10856 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
10857 SPARCStation 1, 2, IPX etc.
10859 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
10860 architecture. The only difference from V7 code is that the compiler emits
10861 the integer multiply and integer divide instructions which exist in SPARC-V8
10862 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
10863 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
10866 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
10867 the SPARC architecture. This adds the integer multiply, integer divide step
10868 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
10869 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
10870 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU. With
10871 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
10872 MB86934 chip, which is the more recent SPARClite with FPU.
10874 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
10875 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
10876 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
10877 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
10878 optimizes it for the TEMIC SPARClet chip.
10880 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
10881 architecture. This adds 64-bit integer and floating-point move instructions,
10882 3 additional floating-point condition code registers and conditional move
10883 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
10884 optimizes it for the Sun UltraSPARC I/II chips. With
10885 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
10886 Sun UltraSPARC III chip.
10888 @item -mtune=@var{cpu_type}
10890 Set the instruction scheduling parameters for machine type
10891 @var{cpu_type}, but do not set the instruction set or register set that the
10892 option @option{-mcpu=@var{cpu_type}} would.
10894 The same values for @option{-mcpu=@var{cpu_type}} can be used for
10895 @option{-mtune=@var{cpu_type}}, but the only useful values are those
10896 that select a particular cpu implementation. Those are @samp{cypress},
10897 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
10898 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
10899 @samp{ultrasparc3}.
10904 @opindex mno-v8plus
10905 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI. The
10906 difference from the V8 ABI is that the global and out registers are
10907 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
10908 mode for all SPARC-V9 processors.
10914 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
10915 Visual Instruction Set extensions. The default is @option{-mno-vis}.
10918 These @samp{-m} options are supported in addition to the above
10919 on SPARC-V9 processors in 64-bit environments:
10922 @item -mlittle-endian
10923 @opindex mlittle-endian
10924 Generate code for a processor running in little-endian mode. It is only
10925 available for a few configurations and most notably not on Solaris.
10931 Generate code for a 32-bit or 64-bit environment.
10932 The 32-bit environment sets int, long and pointer to 32 bits.
10933 The 64-bit environment sets int to 32 bits and long and pointer
10936 @item -mcmodel=medlow
10937 @opindex mcmodel=medlow
10938 Generate code for the Medium/Low code model: 64-bit addresses, programs
10939 must be linked in the low 32 bits of memory. Programs can be statically
10940 or dynamically linked.
10942 @item -mcmodel=medmid
10943 @opindex mcmodel=medmid
10944 Generate code for the Medium/Middle code model: 64-bit addresses, programs
10945 must be linked in the low 44 bits of memory, the text and data segments must
10946 be less than 2GB in size and the data segment must be located within 2GB of
10949 @item -mcmodel=medany
10950 @opindex mcmodel=medany
10951 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
10952 may be linked anywhere in memory, the text and data segments must be less
10953 than 2GB in size and the data segment must be located within 2GB of the
10956 @item -mcmodel=embmedany
10957 @opindex mcmodel=embmedany
10958 Generate code for the Medium/Anywhere code model for embedded systems:
10959 64-bit addresses, the text and data segments must be less than 2GB in
10960 size, both starting anywhere in memory (determined at link time). The
10961 global register %g4 points to the base of the data segment. Programs
10962 are statically linked and PIC is not supported.
10965 @itemx -mno-stack-bias
10966 @opindex mstack-bias
10967 @opindex mno-stack-bias
10968 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
10969 frame pointer if present, are offset by @minus{}2047 which must be added back
10970 when making stack frame references. This is the default in 64-bit mode.
10971 Otherwise, assume no such offset is present.
10974 These switches are supported in addition to the above on Solaris:
10979 Add support for multithreading using the Solaris threads library. This
10980 option sets flags for both the preprocessor and linker. This option does
10981 not affect the thread safety of object code produced by the compiler or
10982 that of libraries supplied with it.
10986 Add support for multithreading using the POSIX threads library. This
10987 option sets flags for both the preprocessor and linker. This option does
10988 not affect the thread safety of object code produced by the compiler or
10989 that of libraries supplied with it.
10992 @node System V Options
10993 @subsection Options for System V
10995 These additional options are available on System V Release 4 for
10996 compatibility with other compilers on those systems:
11001 Create a shared object.
11002 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
11006 Identify the versions of each tool used by the compiler, in a
11007 @code{.ident} assembler directive in the output.
11011 Refrain from adding @code{.ident} directives to the output file (this is
11014 @item -YP,@var{dirs}
11016 Search the directories @var{dirs}, and no others, for libraries
11017 specified with @option{-l}.
11019 @item -Ym,@var{dir}
11021 Look in the directory @var{dir} to find the M4 preprocessor.
11022 The assembler uses this option.
11023 @c This is supposed to go with a -Yd for predefined M4 macro files, but
11024 @c the generic assembler that comes with Solaris takes just -Ym.
11027 @node TMS320C3x/C4x Options
11028 @subsection TMS320C3x/C4x Options
11029 @cindex TMS320C3x/C4x Options
11031 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
11035 @item -mcpu=@var{cpu_type}
11037 Set the instruction set, register set, and instruction scheduling
11038 parameters for machine type @var{cpu_type}. Supported values for
11039 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
11040 @samp{c44}. The default is @samp{c40} to generate code for the
11045 @itemx -msmall-memory
11047 @opindex mbig-memory
11049 @opindex msmall-memory
11051 Generates code for the big or small memory model. The small memory
11052 model assumed that all data fits into one 64K word page. At run-time
11053 the data page (DP) register must be set to point to the 64K page
11054 containing the .bss and .data program sections. The big memory model is
11055 the default and requires reloading of the DP register for every direct
11062 Allow (disallow) allocation of general integer operands into the block
11063 count register BK@.
11069 Enable (disable) generation of code using decrement and branch,
11070 DBcond(D), instructions. This is enabled by default for the C4x. To be
11071 on the safe side, this is disabled for the C3x, since the maximum
11072 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
11073 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
11074 that it can utilize the decrement and branch instruction, but will give
11075 up if there is more than one memory reference in the loop. Thus a loop
11076 where the loop counter is decremented can generate slightly more
11077 efficient code, in cases where the RPTB instruction cannot be utilized.
11079 @item -mdp-isr-reload
11081 @opindex mdp-isr-reload
11083 Force the DP register to be saved on entry to an interrupt service
11084 routine (ISR), reloaded to point to the data section, and restored on
11085 exit from the ISR@. This should not be required unless someone has
11086 violated the small memory model by modifying the DP register, say within
11093 For the C3x use the 24-bit MPYI instruction for integer multiplies
11094 instead of a library call to guarantee 32-bit results. Note that if one
11095 of the operands is a constant, then the multiplication will be performed
11096 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
11097 then squaring operations are performed inline instead of a library call.
11100 @itemx -mno-fast-fix
11102 @opindex mno-fast-fix
11103 The C3x/C4x FIX instruction to convert a floating point value to an
11104 integer value chooses the nearest integer less than or equal to the
11105 floating point value rather than to the nearest integer. Thus if the
11106 floating point number is negative, the result will be incorrectly
11107 truncated an additional code is necessary to detect and correct this
11108 case. This option can be used to disable generation of the additional
11109 code required to correct the result.
11115 Enable (disable) generation of repeat block sequences using the RPTB
11116 instruction for zero overhead looping. The RPTB construct is only used
11117 for innermost loops that do not call functions or jump across the loop
11118 boundaries. There is no advantage having nested RPTB loops due to the
11119 overhead required to save and restore the RC, RS, and RE registers.
11120 This is enabled by default with @option{-O2}.
11122 @item -mrpts=@var{count}
11126 Enable (disable) the use of the single instruction repeat instruction
11127 RPTS@. If a repeat block contains a single instruction, and the loop
11128 count can be guaranteed to be less than the value @var{count}, GCC will
11129 emit a RPTS instruction instead of a RPTB@. If no value is specified,
11130 then a RPTS will be emitted even if the loop count cannot be determined
11131 at compile time. Note that the repeated instruction following RPTS does
11132 not have to be reloaded from memory each iteration, thus freeing up the
11133 CPU buses for operands. However, since interrupts are blocked by this
11134 instruction, it is disabled by default.
11136 @item -mloop-unsigned
11137 @itemx -mno-loop-unsigned
11138 @opindex mloop-unsigned
11139 @opindex mno-loop-unsigned
11140 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
11141 is @math{2^{31} + 1} since these instructions test if the iteration count is
11142 negative to terminate the loop. If the iteration count is unsigned
11143 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
11144 exceeded. This switch allows an unsigned iteration count.
11148 Try to emit an assembler syntax that the TI assembler (asm30) is happy
11149 with. This also enforces compatibility with the API employed by the TI
11150 C3x C compiler. For example, long doubles are passed as structures
11151 rather than in floating point registers.
11157 Generate code that uses registers (stack) for passing arguments to functions.
11158 By default, arguments are passed in registers where possible rather
11159 than by pushing arguments on to the stack.
11161 @item -mparallel-insns
11162 @itemx -mno-parallel-insns
11163 @opindex mparallel-insns
11164 @opindex mno-parallel-insns
11165 Allow the generation of parallel instructions. This is enabled by
11166 default with @option{-O2}.
11168 @item -mparallel-mpy
11169 @itemx -mno-parallel-mpy
11170 @opindex mparallel-mpy
11171 @opindex mno-parallel-mpy
11172 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
11173 provided @option{-mparallel-insns} is also specified. These instructions have
11174 tight register constraints which can pessimize the code generation
11175 of large functions.
11180 @subsection V850 Options
11181 @cindex V850 Options
11183 These @samp{-m} options are defined for V850 implementations:
11187 @itemx -mno-long-calls
11188 @opindex mlong-calls
11189 @opindex mno-long-calls
11190 Treat all calls as being far away (near). If calls are assumed to be
11191 far away, the compiler will always load the functions address up into a
11192 register, and call indirect through the pointer.
11198 Do not optimize (do optimize) basic blocks that use the same index
11199 pointer 4 or more times to copy pointer into the @code{ep} register, and
11200 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
11201 option is on by default if you optimize.
11203 @item -mno-prolog-function
11204 @itemx -mprolog-function
11205 @opindex mno-prolog-function
11206 @opindex mprolog-function
11207 Do not use (do use) external functions to save and restore registers
11208 at the prologue and epilogue of a function. The external functions
11209 are slower, but use less code space if more than one function saves
11210 the same number of registers. The @option{-mprolog-function} option
11211 is on by default if you optimize.
11215 Try to make the code as small as possible. At present, this just turns
11216 on the @option{-mep} and @option{-mprolog-function} options.
11218 @item -mtda=@var{n}
11220 Put static or global variables whose size is @var{n} bytes or less into
11221 the tiny data area that register @code{ep} points to. The tiny data
11222 area can hold up to 256 bytes in total (128 bytes for byte references).
11224 @item -msda=@var{n}
11226 Put static or global variables whose size is @var{n} bytes or less into
11227 the small data area that register @code{gp} points to. The small data
11228 area can hold up to 64 kilobytes.
11230 @item -mzda=@var{n}
11232 Put static or global variables whose size is @var{n} bytes or less into
11233 the first 32 kilobytes of memory.
11237 Specify that the target processor is the V850.
11240 @opindex mbig-switch
11241 Generate code suitable for big switch tables. Use this option only if
11242 the assembler/linker complain about out of range branches within a switch
11247 This option will cause r2 and r5 to be used in the code generated by
11248 the compiler. This setting is the default.
11250 @item -mno-app-regs
11251 @opindex mno-app-regs
11252 This option will cause r2 and r5 to be treated as fixed registers.
11256 Specify that the target processor is the V850E1. The preprocessor
11257 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
11258 this option is used.
11262 Specify that the target processor is the V850E. The preprocessor
11263 constant @samp{__v850e__} will be defined if this option is used.
11265 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
11266 are defined then a default target processor will be chosen and the
11267 relevant @samp{__v850*__} preprocessor constant will be defined.
11269 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
11270 defined, regardless of which processor variant is the target.
11272 @item -mdisable-callt
11273 @opindex mdisable-callt
11274 This option will suppress generation of the CALLT instruction for the
11275 v850e and v850e1 flavors of the v850 architecture. The default is
11276 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
11281 @subsection VAX Options
11282 @cindex VAX options
11284 These @samp{-m} options are defined for the VAX:
11289 Do not output certain jump instructions (@code{aobleq} and so on)
11290 that the Unix assembler for the VAX cannot handle across long
11295 Do output those jump instructions, on the assumption that you
11296 will assemble with the GNU assembler.
11300 Output code for g-format floating point numbers instead of d-format.
11303 @node x86-64 Options
11304 @subsection x86-64 Options
11305 @cindex x86-64 options
11307 These are listed under @xref{i386 and x86-64 Options}.
11309 @node Xstormy16 Options
11310 @subsection Xstormy16 Options
11311 @cindex Xstormy16 Options
11313 These options are defined for Xstormy16:
11318 Choose startup files and linker script suitable for the simulator.
11321 @node Xtensa Options
11322 @subsection Xtensa Options
11323 @cindex Xtensa Options
11325 These options are supported for Xtensa targets:
11329 @itemx -mno-const16
11331 @opindex mno-const16
11332 Enable or disable use of @code{CONST16} instructions for loading
11333 constant values. The @code{CONST16} instruction is currently not a
11334 standard option from Tensilica. When enabled, @code{CONST16}
11335 instructions are always used in place of the standard @code{L32R}
11336 instructions. The use of @code{CONST16} is enabled by default only if
11337 the @code{L32R} instruction is not available.
11340 @itemx -mno-fused-madd
11341 @opindex mfused-madd
11342 @opindex mno-fused-madd
11343 Enable or disable use of fused multiply/add and multiply/subtract
11344 instructions in the floating-point option. This has no effect if the
11345 floating-point option is not also enabled. Disabling fused multiply/add
11346 and multiply/subtract instructions forces the compiler to use separate
11347 instructions for the multiply and add/subtract operations. This may be
11348 desirable in some cases where strict IEEE 754-compliant results are
11349 required: the fused multiply add/subtract instructions do not round the
11350 intermediate result, thereby producing results with @emph{more} bits of
11351 precision than specified by the IEEE standard. Disabling fused multiply
11352 add/subtract instructions also ensures that the program output is not
11353 sensitive to the compiler's ability to combine multiply and add/subtract
11356 @item -mtext-section-literals
11357 @itemx -mno-text-section-literals
11358 @opindex mtext-section-literals
11359 @opindex mno-text-section-literals
11360 Control the treatment of literal pools. The default is
11361 @option{-mno-text-section-literals}, which places literals in a separate
11362 section in the output file. This allows the literal pool to be placed
11363 in a data RAM/ROM, and it also allows the linker to combine literal
11364 pools from separate object files to remove redundant literals and
11365 improve code size. With @option{-mtext-section-literals}, the literals
11366 are interspersed in the text section in order to keep them as close as
11367 possible to their references. This may be necessary for large assembly
11370 @item -mtarget-align
11371 @itemx -mno-target-align
11372 @opindex mtarget-align
11373 @opindex mno-target-align
11374 When this option is enabled, GCC instructs the assembler to
11375 automatically align instructions to reduce branch penalties at the
11376 expense of some code density. The assembler attempts to widen density
11377 instructions to align branch targets and the instructions following call
11378 instructions. If there are not enough preceding safe density
11379 instructions to align a target, no widening will be performed. The
11380 default is @option{-mtarget-align}. These options do not affect the
11381 treatment of auto-aligned instructions like @code{LOOP}, which the
11382 assembler will always align, either by widening density instructions or
11383 by inserting no-op instructions.
11386 @itemx -mno-longcalls
11387 @opindex mlongcalls
11388 @opindex mno-longcalls
11389 When this option is enabled, GCC instructs the assembler to translate
11390 direct calls to indirect calls unless it can determine that the target
11391 of a direct call is in the range allowed by the call instruction. This
11392 translation typically occurs for calls to functions in other source
11393 files. Specifically, the assembler translates a direct @code{CALL}
11394 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
11395 The default is @option{-mno-longcalls}. This option should be used in
11396 programs where the call target can potentially be out of range. This
11397 option is implemented in the assembler, not the compiler, so the
11398 assembly code generated by GCC will still show direct call
11399 instructions---look at the disassembled object code to see the actual
11400 instructions. Note that the assembler will use an indirect call for
11401 every cross-file call, not just those that really will be out of range.
11404 @node zSeries Options
11405 @subsection zSeries Options
11406 @cindex zSeries options
11408 These are listed under @xref{S/390 and zSeries Options}.
11410 @node Code Gen Options
11411 @section Options for Code Generation Conventions
11412 @cindex code generation conventions
11413 @cindex options, code generation
11414 @cindex run-time options
11416 These machine-independent options control the interface conventions
11417 used in code generation.
11419 Most of them have both positive and negative forms; the negative form
11420 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
11421 one of the forms is listed---the one which is not the default. You
11422 can figure out the other form by either removing @samp{no-} or adding
11426 @item -fbounds-check
11427 @opindex fbounds-check
11428 For front-ends that support it, generate additional code to check that
11429 indices used to access arrays are within the declared range. This is
11430 currently only supported by the Java and Fortran 77 front-ends, where
11431 this option defaults to true and false respectively.
11435 This option generates traps for signed overflow on addition, subtraction,
11436 multiplication operations.
11440 This option instructs the compiler to assume that signed arithmetic
11441 overflow of addition, subtraction and multiplication wraps around
11442 using twos-complement representation. This flag enables some optimizations
11443 and disables other. This option is enabled by default for the Java
11444 front-end, as required by the Java language specification.
11447 @opindex fexceptions
11448 Enable exception handling. Generates extra code needed to propagate
11449 exceptions. For some targets, this implies GCC will generate frame
11450 unwind information for all functions, which can produce significant data
11451 size overhead, although it does not affect execution. If you do not
11452 specify this option, GCC will enable it by default for languages like
11453 C++ which normally require exception handling, and disable it for
11454 languages like C that do not normally require it. However, you may need
11455 to enable this option when compiling C code that needs to interoperate
11456 properly with exception handlers written in C++. You may also wish to
11457 disable this option if you are compiling older C++ programs that don't
11458 use exception handling.
11460 @item -fnon-call-exceptions
11461 @opindex fnon-call-exceptions
11462 Generate code that allows trapping instructions to throw exceptions.
11463 Note that this requires platform-specific runtime support that does
11464 not exist everywhere. Moreover, it only allows @emph{trapping}
11465 instructions to throw exceptions, i.e.@: memory references or floating
11466 point instructions. It does not allow exceptions to be thrown from
11467 arbitrary signal handlers such as @code{SIGALRM}.
11469 @item -funwind-tables
11470 @opindex funwind-tables
11471 Similar to @option{-fexceptions}, except that it will just generate any needed
11472 static data, but will not affect the generated code in any other way.
11473 You will normally not enable this option; instead, a language processor
11474 that needs this handling would enable it on your behalf.
11476 @item -fasynchronous-unwind-tables
11477 @opindex fasynchronous-unwind-tables
11478 Generate unwind table in dwarf2 format, if supported by target machine. The
11479 table is exact at each instruction boundary, so it can be used for stack
11480 unwinding from asynchronous events (such as debugger or garbage collector).
11482 @item -fpcc-struct-return
11483 @opindex fpcc-struct-return
11484 Return ``short'' @code{struct} and @code{union} values in memory like
11485 longer ones, rather than in registers. This convention is less
11486 efficient, but it has the advantage of allowing intercallability between
11487 GCC-compiled files and files compiled with other compilers, particularly
11488 the Portable C Compiler (pcc).
11490 The precise convention for returning structures in memory depends
11491 on the target configuration macros.
11493 Short structures and unions are those whose size and alignment match
11494 that of some integer type.
11496 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
11497 switch is not binary compatible with code compiled with the
11498 @option{-freg-struct-return} switch.
11499 Use it to conform to a non-default application binary interface.
11501 @item -freg-struct-return
11502 @opindex freg-struct-return
11503 Return @code{struct} and @code{union} values in registers when possible.
11504 This is more efficient for small structures than
11505 @option{-fpcc-struct-return}.
11507 If you specify neither @option{-fpcc-struct-return} nor
11508 @option{-freg-struct-return}, GCC defaults to whichever convention is
11509 standard for the target. If there is no standard convention, GCC
11510 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
11511 the principal compiler. In those cases, we can choose the standard, and
11512 we chose the more efficient register return alternative.
11514 @strong{Warning:} code compiled with the @option{-freg-struct-return}
11515 switch is not binary compatible with code compiled with the
11516 @option{-fpcc-struct-return} switch.
11517 Use it to conform to a non-default application binary interface.
11519 @item -fshort-enums
11520 @opindex fshort-enums
11521 Allocate to an @code{enum} type only as many bytes as it needs for the
11522 declared range of possible values. Specifically, the @code{enum} type
11523 will be equivalent to the smallest integer type which has enough room.
11525 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
11526 code that is not binary compatible with code generated without that switch.
11527 Use it to conform to a non-default application binary interface.
11529 @item -fshort-double
11530 @opindex fshort-double
11531 Use the same size for @code{double} as for @code{float}.
11533 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
11534 code that is not binary compatible with code generated without that switch.
11535 Use it to conform to a non-default application binary interface.
11537 @item -fshort-wchar
11538 @opindex fshort-wchar
11539 Override the underlying type for @samp{wchar_t} to be @samp{short
11540 unsigned int} instead of the default for the target. This option is
11541 useful for building programs to run under WINE@.
11543 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
11544 code that is not binary compatible with code generated without that switch.
11545 Use it to conform to a non-default application binary interface.
11547 @item -fshared-data
11548 @opindex fshared-data
11549 Requests that the data and non-@code{const} variables of this
11550 compilation be shared data rather than private data. The distinction
11551 makes sense only on certain operating systems, where shared data is
11552 shared between processes running the same program, while private data
11553 exists in one copy per process.
11556 @opindex fno-common
11557 In C, allocate even uninitialized global variables in the data section of the
11558 object file, rather than generating them as common blocks. This has the
11559 effect that if the same variable is declared (without @code{extern}) in
11560 two different compilations, you will get an error when you link them.
11561 The only reason this might be useful is if you wish to verify that the
11562 program will work on other systems which always work this way.
11566 Ignore the @samp{#ident} directive.
11568 @item -finhibit-size-directive
11569 @opindex finhibit-size-directive
11570 Don't output a @code{.size} assembler directive, or anything else that
11571 would cause trouble if the function is split in the middle, and the
11572 two halves are placed at locations far apart in memory. This option is
11573 used when compiling @file{crtstuff.c}; you should not need to use it
11576 @item -fverbose-asm
11577 @opindex fverbose-asm
11578 Put extra commentary information in the generated assembly code to
11579 make it more readable. This option is generally only of use to those
11580 who actually need to read the generated assembly code (perhaps while
11581 debugging the compiler itself).
11583 @option{-fno-verbose-asm}, the default, causes the
11584 extra information to be omitted and is useful when comparing two assembler
11589 @cindex global offset table
11591 Generate position-independent code (PIC) suitable for use in a shared
11592 library, if supported for the target machine. Such code accesses all
11593 constant addresses through a global offset table (GOT)@. The dynamic
11594 loader resolves the GOT entries when the program starts (the dynamic
11595 loader is not part of GCC; it is part of the operating system). If
11596 the GOT size for the linked executable exceeds a machine-specific
11597 maximum size, you get an error message from the linker indicating that
11598 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
11599 instead. (These maximums are 8k on the SPARC and 32k
11600 on the m68k and RS/6000. The 386 has no such limit.)
11602 Position-independent code requires special support, and therefore works
11603 only on certain machines. For the 386, GCC supports PIC for System V
11604 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
11605 position-independent.
11609 If supported for the target machine, emit position-independent code,
11610 suitable for dynamic linking and avoiding any limit on the size of the
11611 global offset table. This option makes a difference on the m68k
11614 Position-independent code requires special support, and therefore works
11615 only on certain machines.
11621 These options are similar to @option{-fpic} and @option{-fPIC}, but
11622 generated position independent code can be only linked into executables.
11623 Usually these options are used when @option{-pie} GCC option will be
11624 used during linking.
11626 @item -ffixed-@var{reg}
11628 Treat the register named @var{reg} as a fixed register; generated code
11629 should never refer to it (except perhaps as a stack pointer, frame
11630 pointer or in some other fixed role).
11632 @var{reg} must be the name of a register. The register names accepted
11633 are machine-specific and are defined in the @code{REGISTER_NAMES}
11634 macro in the machine description macro file.
11636 This flag does not have a negative form, because it specifies a
11639 @item -fcall-used-@var{reg}
11640 @opindex fcall-used
11641 Treat the register named @var{reg} as an allocable register that is
11642 clobbered by function calls. It may be allocated for temporaries or
11643 variables that do not live across a call. Functions compiled this way
11644 will not save and restore the register @var{reg}.
11646 It is an error to used this flag with the frame pointer or stack pointer.
11647 Use of this flag for other registers that have fixed pervasive roles in
11648 the machine's execution model will produce disastrous results.
11650 This flag does not have a negative form, because it specifies a
11653 @item -fcall-saved-@var{reg}
11654 @opindex fcall-saved
11655 Treat the register named @var{reg} as an allocable register saved by
11656 functions. It may be allocated even for temporaries or variables that
11657 live across a call. Functions compiled this way will save and restore
11658 the register @var{reg} if they use it.
11660 It is an error to used this flag with the frame pointer or stack pointer.
11661 Use of this flag for other registers that have fixed pervasive roles in
11662 the machine's execution model will produce disastrous results.
11664 A different sort of disaster will result from the use of this flag for
11665 a register in which function values may be returned.
11667 This flag does not have a negative form, because it specifies a
11670 @item -fpack-struct
11671 @opindex fpack-struct
11672 Pack all structure members together without holes.
11674 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
11675 code that is not binary compatible with code generated without that switch.
11676 Additionally, it makes the code suboptimal.
11677 Use it to conform to a non-default application binary interface.
11679 @item -finstrument-functions
11680 @opindex finstrument-functions
11681 Generate instrumentation calls for entry and exit to functions. Just
11682 after function entry and just before function exit, the following
11683 profiling functions will be called with the address of the current
11684 function and its call site. (On some platforms,
11685 @code{__builtin_return_address} does not work beyond the current
11686 function, so the call site information may not be available to the
11687 profiling functions otherwise.)
11690 void __cyg_profile_func_enter (void *this_fn,
11692 void __cyg_profile_func_exit (void *this_fn,
11696 The first argument is the address of the start of the current function,
11697 which may be looked up exactly in the symbol table.
11699 This instrumentation is also done for functions expanded inline in other
11700 functions. The profiling calls will indicate where, conceptually, the
11701 inline function is entered and exited. This means that addressable
11702 versions of such functions must be available. If all your uses of a
11703 function are expanded inline, this may mean an additional expansion of
11704 code size. If you use @samp{extern inline} in your C code, an
11705 addressable version of such functions must be provided. (This is
11706 normally the case anyways, but if you get lucky and the optimizer always
11707 expands the functions inline, you might have gotten away without
11708 providing static copies.)
11710 A function may be given the attribute @code{no_instrument_function}, in
11711 which case this instrumentation will not be done. This can be used, for
11712 example, for the profiling functions listed above, high-priority
11713 interrupt routines, and any functions from which the profiling functions
11714 cannot safely be called (perhaps signal handlers, if the profiling
11715 routines generate output or allocate memory).
11717 @item -fstack-check
11718 @opindex fstack-check
11719 Generate code to verify that you do not go beyond the boundary of the
11720 stack. You should specify this flag if you are running in an
11721 environment with multiple threads, but only rarely need to specify it in
11722 a single-threaded environment since stack overflow is automatically
11723 detected on nearly all systems if there is only one stack.
11725 Note that this switch does not actually cause checking to be done; the
11726 operating system must do that. The switch causes generation of code
11727 to ensure that the operating system sees the stack being extended.
11729 @item -fstack-limit-register=@var{reg}
11730 @itemx -fstack-limit-symbol=@var{sym}
11731 @itemx -fno-stack-limit
11732 @opindex fstack-limit-register
11733 @opindex fstack-limit-symbol
11734 @opindex fno-stack-limit
11735 Generate code to ensure that the stack does not grow beyond a certain value,
11736 either the value of a register or the address of a symbol. If the stack
11737 would grow beyond the value, a signal is raised. For most targets,
11738 the signal is raised before the stack overruns the boundary, so
11739 it is possible to catch the signal without taking special precautions.
11741 For instance, if the stack starts at absolute address @samp{0x80000000}
11742 and grows downwards, you can use the flags
11743 @option{-fstack-limit-symbol=__stack_limit} and
11744 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11745 of 128KB@. Note that this may only work with the GNU linker.
11747 @cindex aliasing of parameters
11748 @cindex parameters, aliased
11749 @item -fargument-alias
11750 @itemx -fargument-noalias
11751 @itemx -fargument-noalias-global
11752 @opindex fargument-alias
11753 @opindex fargument-noalias
11754 @opindex fargument-noalias-global
11755 Specify the possible relationships among parameters and between
11756 parameters and global data.
11758 @option{-fargument-alias} specifies that arguments (parameters) may
11759 alias each other and may alias global storage.@*
11760 @option{-fargument-noalias} specifies that arguments do not alias
11761 each other, but may alias global storage.@*
11762 @option{-fargument-noalias-global} specifies that arguments do not
11763 alias each other and do not alias global storage.
11765 Each language will automatically use whatever option is required by
11766 the language standard. You should not need to use these options yourself.
11768 @item -fleading-underscore
11769 @opindex fleading-underscore
11770 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
11771 change the way C symbols are represented in the object file. One use
11772 is to help link with legacy assembly code.
11774 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
11775 generate code that is not binary compatible with code generated without that
11776 switch. Use it to conform to a non-default application binary interface.
11777 Not all targets provide complete support for this switch.
11779 @item -ftls-model=@var{model}
11780 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
11781 The @var{model} argument should be one of @code{global-dynamic},
11782 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
11784 The default without @option{-fpic} is @code{initial-exec}; with
11785 @option{-fpic} the default is @code{global-dynamic}.
11787 @item -fvisibility=@var{default|internal|hidden|protected}
11788 @opindex fvisibility
11789 Set the default ELF image symbol visibility to the specified option - all
11790 symbols will be marked with this unless overrided within the code.
11791 Using this feature can very substantially improve linking and
11792 load times of shared object libraries, produce more optimised
11793 code, provide near-perfect API export and prevent symbol clashes.
11794 It is @strong{strongly} recommended that you use this in any shared objects
11797 Despite the nomenclature, @code{default} always means public ie;
11798 available to be linked against from outside the shared object.
11799 @code{protected} and @code{internal} are pretty useless in real-world
11800 usage so the only other commonly used option will be @code{hidden}.
11801 The default if -fvisibility isn't specified is @code{default} ie; make every
11802 symbol public - this causes the same behaviour as previous versions of
11805 A good explanation of the benefits offered by ensuring ELF
11806 symbols have the correct visibility is given by ``How To Write
11807 Shared Libraries'' by Ulrich Drepper (which can be found at
11808 @w{@uref{http://people.redhat.com/~drepper/}}) - however a superior
11809 solution made possible by this option to marking things hidden when
11810 the default is public is to make the default hidden and mark things
11811 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
11812 and @code{__attribute__ ((visibility("default")))} instead of
11813 @code{__declspec(dllexport)} you get almost identical semantics with
11814 identical syntax. This is a great boon to those working with
11815 cross-platform projects.
11817 For those adding visibility support to existing code, you may find
11818 @samp{#pragma GCC visibility} of use. This works by you enclosing
11819 the declarations you wish to set visibility for with (for example)
11820 @samp{#pragma GCC visibility push(hidden)} and
11821 @samp{#pragma GCC visibility pop}. These can be nested up to sixteen
11822 times. Bear in mind that symbol visibility should be viewed @strong{as
11823 part of the API interface contract} and thus all new code should
11824 always specify visibility when it is not the default ie; declarations
11825 only for use within the local DSO should @strong{always} be marked explicitly
11826 as hidden as so to avoid PLT indirection overheads - making this
11827 abundantly clear also aids readability and self-documentation of the code.
11828 Note that due to ISO C++ specification requirements, operator new and
11829 operator delete must always be of default visibility.
11831 An overview of these techniques, their benefits and how to use them
11832 is at @w{@uref{http://www.nedprod.com/programs/gccvisibility.html}}.
11838 @node Environment Variables
11839 @section Environment Variables Affecting GCC
11840 @cindex environment variables
11842 @c man begin ENVIRONMENT
11843 This section describes several environment variables that affect how GCC
11844 operates. Some of them work by specifying directories or prefixes to use
11845 when searching for various kinds of files. Some are used to specify other
11846 aspects of the compilation environment.
11848 Note that you can also specify places to search using options such as
11849 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
11850 take precedence over places specified using environment variables, which
11851 in turn take precedence over those specified by the configuration of GCC@.
11852 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
11853 GNU Compiler Collection (GCC) Internals}.
11858 @c @itemx LC_COLLATE
11860 @c @itemx LC_MONETARY
11861 @c @itemx LC_NUMERIC
11866 @c @findex LC_COLLATE
11867 @findex LC_MESSAGES
11868 @c @findex LC_MONETARY
11869 @c @findex LC_NUMERIC
11873 These environment variables control the way that GCC uses
11874 localization information that allow GCC to work with different
11875 national conventions. GCC inspects the locale categories
11876 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
11877 so. These locale categories can be set to any value supported by your
11878 installation. A typical value is @samp{en_UK} for English in the United
11881 The @env{LC_CTYPE} environment variable specifies character
11882 classification. GCC uses it to determine the character boundaries in
11883 a string; this is needed for some multibyte encodings that contain quote
11884 and escape characters that would otherwise be interpreted as a string
11887 The @env{LC_MESSAGES} environment variable specifies the language to
11888 use in diagnostic messages.
11890 If the @env{LC_ALL} environment variable is set, it overrides the value
11891 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
11892 and @env{LC_MESSAGES} default to the value of the @env{LANG}
11893 environment variable. If none of these variables are set, GCC
11894 defaults to traditional C English behavior.
11898 If @env{TMPDIR} is set, it specifies the directory to use for temporary
11899 files. GCC uses temporary files to hold the output of one stage of
11900 compilation which is to be used as input to the next stage: for example,
11901 the output of the preprocessor, which is the input to the compiler
11904 @item GCC_EXEC_PREFIX
11905 @findex GCC_EXEC_PREFIX
11906 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
11907 names of the subprograms executed by the compiler. No slash is added
11908 when this prefix is combined with the name of a subprogram, but you can
11909 specify a prefix that ends with a slash if you wish.
11911 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
11912 an appropriate prefix to use based on the pathname it was invoked with.
11914 If GCC cannot find the subprogram using the specified prefix, it
11915 tries looking in the usual places for the subprogram.
11917 The default value of @env{GCC_EXEC_PREFIX} is
11918 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
11919 of @code{prefix} when you ran the @file{configure} script.
11921 Other prefixes specified with @option{-B} take precedence over this prefix.
11923 This prefix is also used for finding files such as @file{crt0.o} that are
11926 In addition, the prefix is used in an unusual way in finding the
11927 directories to search for header files. For each of the standard
11928 directories whose name normally begins with @samp{/usr/local/lib/gcc}
11929 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
11930 replacing that beginning with the specified prefix to produce an
11931 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
11932 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
11933 These alternate directories are searched first; the standard directories
11936 @item COMPILER_PATH
11937 @findex COMPILER_PATH
11938 The value of @env{COMPILER_PATH} is a colon-separated list of
11939 directories, much like @env{PATH}. GCC tries the directories thus
11940 specified when searching for subprograms, if it can't find the
11941 subprograms using @env{GCC_EXEC_PREFIX}.
11944 @findex LIBRARY_PATH
11945 The value of @env{LIBRARY_PATH} is a colon-separated list of
11946 directories, much like @env{PATH}. When configured as a native compiler,
11947 GCC tries the directories thus specified when searching for special
11948 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
11949 using GCC also uses these directories when searching for ordinary
11950 libraries for the @option{-l} option (but directories specified with
11951 @option{-L} come first).
11955 @cindex locale definition
11956 This variable is used to pass locale information to the compiler. One way in
11957 which this information is used is to determine the character set to be used
11958 when character literals, string literals and comments are parsed in C and C++.
11959 When the compiler is configured to allow multibyte characters,
11960 the following values for @env{LANG} are recognized:
11964 Recognize JIS characters.
11966 Recognize SJIS characters.
11968 Recognize EUCJP characters.
11971 If @env{LANG} is not defined, or if it has some other value, then the
11972 compiler will use mblen and mbtowc as defined by the default locale to
11973 recognize and translate multibyte characters.
11977 Some additional environments variables affect the behavior of the
11980 @include cppenv.texi
11984 @node Precompiled Headers
11985 @section Using Precompiled Headers
11986 @cindex precompiled headers
11987 @cindex speed of compilation
11989 Often large projects have many header files that are included in every
11990 source file. The time the compiler takes to process these header files
11991 over and over again can account for nearly all of the time required to
11992 build the project. To make builds faster, GCC allows users to
11993 `precompile' a header file; then, if builds can use the precompiled
11994 header file they will be much faster.
11996 @strong{Caution:} There are a few known situations where GCC will
11997 crash when trying to use a precompiled header. If you have trouble
11998 with a precompiled header, you should remove the precompiled header
11999 and compile without it. In addition, please use GCC's on-line
12000 defect-tracking system to report any problems you encounter with
12001 precompiled headers. @xref{Bugs}.
12003 To create a precompiled header file, simply compile it as you would any
12004 other file, if necessary using the @option{-x} option to make the driver
12005 treat it as a C or C++ header file. You will probably want to use a
12006 tool like @command{make} to keep the precompiled header up-to-date when
12007 the headers it contains change.
12009 A precompiled header file will be searched for when @code{#include} is
12010 seen in the compilation. As it searches for the included file
12011 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
12012 compiler looks for a precompiled header in each directory just before it
12013 looks for the include file in that directory. The name searched for is
12014 the name specified in the @code{#include} with @samp{.gch} appended. If
12015 the precompiled header file can't be used, it is ignored.
12017 For instance, if you have @code{#include "all.h"}, and you have
12018 @file{all.h.gch} in the same directory as @file{all.h}, then the
12019 precompiled header file will be used if possible, and the original
12020 header will be used otherwise.
12022 Alternatively, you might decide to put the precompiled header file in a
12023 directory and use @option{-I} to ensure that directory is searched
12024 before (or instead of) the directory containing the original header.
12025 Then, if you want to check that the precompiled header file is always
12026 used, you can put a file of the same name as the original header in this
12027 directory containing an @code{#error} command.
12029 This also works with @option{-include}. So yet another way to use
12030 precompiled headers, good for projects not designed with precompiled
12031 header files in mind, is to simply take most of the header files used by
12032 a project, include them from another header file, precompile that header
12033 file, and @option{-include} the precompiled header. If the header files
12034 have guards against multiple inclusion, they will be skipped because
12035 they've already been included (in the precompiled header).
12037 If you need to precompile the same header file for different
12038 languages, targets, or compiler options, you can instead make a
12039 @emph{directory} named like @file{all.h.gch}, and put each precompiled
12040 header in the directory, perhaps using @option{-o}. It doesn't matter
12041 what you call the files in the directory, every precompiled header in
12042 the directory will be considered. The first precompiled header
12043 encountered in the directory that is valid for this compilation will
12044 be used; they're searched in no particular order.
12046 There are many other possibilities, limited only by your imagination,
12047 good sense, and the constraints of your build system.
12049 A precompiled header file can be used only when these conditions apply:
12053 Only one precompiled header can be used in a particular compilation.
12056 A precompiled header can't be used once the first C token is seen. You
12057 can have preprocessor directives before a precompiled header; you can
12058 even include a precompiled header from inside another header, so long as
12059 there are no C tokens before the @code{#include}.
12062 The precompiled header file must be produced for the same language as
12063 the current compilation. You can't use a C precompiled header for a C++
12067 The precompiled header file must be produced by the same compiler
12068 version and configuration as the current compilation is using.
12069 The easiest way to guarantee this is to use the same compiler binary
12070 for creating and using precompiled headers.
12073 Any macros defined before the precompiled header is included must
12074 either be defined in the same way as when the precompiled header was
12075 generated, or must not affect the precompiled header, which usually
12076 means that the they don't appear in the precompiled header at all.
12078 The @option{-D} option is one way to define a macro before a
12079 precompiled header is included; using a @code{#define} can also do it.
12080 There are also some options that define macros implicitly, like
12081 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
12084 @item If debugging information is output when using the precompiled
12085 header, using @option{-g} or similar, the same kind of debugging information
12086 must have been output when building the precompiled header. However,
12087 a precompiled header built using @option{-g} can be used in a compilation
12088 when no debugging information is being output.
12090 @item The same @option{-m} options must generally be used when building
12091 and using the precompiled header. @xref{Submodel Options},
12092 for any cases where this rule is relaxed.
12094 @item Each of the following options must be the same when building and using
12095 the precompiled header:
12097 @gccoptlist{-fexceptions -funit-at-a-time}
12100 Some other command-line options starting with @option{-f},
12101 @option{-p}, or @option{-O} must be defined in the same way as when
12102 the precompiled header was generated. At present, it's not clear
12103 which options are safe to change and which are not; the safest choice
12104 is to use exactly the same options when generating and using the
12105 precompiled header. The following are known to be safe:
12107 @gccoptlist{-fpreprocessed -pedantic-errors}
12111 For all of these except the last, the compiler will automatically
12112 ignore the precompiled header if the conditions aren't met. If you
12113 find an option combination that doesn't work and doesn't cause the
12114 precompiled header to be ignored, please consider filing a bug report,
12117 If you do use differing options when generating and using the
12118 precompiled header, the actual behaviour will be a mixture of the
12119 behaviour for the options. For instance, if you use @option{-g} to
12120 generate the precompiled header but not when using it, you may or may
12121 not get debugging information for routines in the precompiled header.
12123 @node Running Protoize
12124 @section Running Protoize
12126 The program @code{protoize} is an optional part of GCC@. You can use
12127 it to add prototypes to a program, thus converting the program to ISO
12128 C in one respect. The companion program @code{unprotoize} does the
12129 reverse: it removes argument types from any prototypes that are found.
12131 When you run these programs, you must specify a set of source files as
12132 command line arguments. The conversion programs start out by compiling
12133 these files to see what functions they define. The information gathered
12134 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
12136 After scanning comes actual conversion. The specified files are all
12137 eligible to be converted; any files they include (whether sources or
12138 just headers) are eligible as well.
12140 But not all the eligible files are converted. By default,
12141 @code{protoize} and @code{unprotoize} convert only source and header
12142 files in the current directory. You can specify additional directories
12143 whose files should be converted with the @option{-d @var{directory}}
12144 option. You can also specify particular files to exclude with the
12145 @option{-x @var{file}} option. A file is converted if it is eligible, its
12146 directory name matches one of the specified directory names, and its
12147 name within the directory has not been excluded.
12149 Basic conversion with @code{protoize} consists of rewriting most
12150 function definitions and function declarations to specify the types of
12151 the arguments. The only ones not rewritten are those for varargs
12154 @code{protoize} optionally inserts prototype declarations at the
12155 beginning of the source file, to make them available for any calls that
12156 precede the function's definition. Or it can insert prototype
12157 declarations with block scope in the blocks where undeclared functions
12160 Basic conversion with @code{unprotoize} consists of rewriting most
12161 function declarations to remove any argument types, and rewriting
12162 function definitions to the old-style pre-ISO form.
12164 Both conversion programs print a warning for any function declaration or
12165 definition that they can't convert. You can suppress these warnings
12168 The output from @code{protoize} or @code{unprotoize} replaces the
12169 original source file. The original file is renamed to a name ending
12170 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
12171 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
12172 for DOS) file already exists, then the source file is simply discarded.
12174 @code{protoize} and @code{unprotoize} both depend on GCC itself to
12175 scan the program and collect information about the functions it uses.
12176 So neither of these programs will work until GCC is installed.
12178 Here is a table of the options you can use with @code{protoize} and
12179 @code{unprotoize}. Each option works with both programs unless
12183 @item -B @var{directory}
12184 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
12185 usual directory (normally @file{/usr/local/lib}). This file contains
12186 prototype information about standard system functions. This option
12187 applies only to @code{protoize}.
12189 @item -c @var{compilation-options}
12190 Use @var{compilation-options} as the options when running @command{gcc} to
12191 produce the @samp{.X} files. The special option @option{-aux-info} is
12192 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
12194 Note that the compilation options must be given as a single argument to
12195 @code{protoize} or @code{unprotoize}. If you want to specify several
12196 @command{gcc} options, you must quote the entire set of compilation options
12197 to make them a single word in the shell.
12199 There are certain @command{gcc} arguments that you cannot use, because they
12200 would produce the wrong kind of output. These include @option{-g},
12201 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
12202 the @var{compilation-options}, they are ignored.
12205 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
12206 systems) instead of @samp{.c}. This is convenient if you are converting
12207 a C program to C++. This option applies only to @code{protoize}.
12210 Add explicit global declarations. This means inserting explicit
12211 declarations at the beginning of each source file for each function
12212 that is called in the file and was not declared. These declarations
12213 precede the first function definition that contains a call to an
12214 undeclared function. This option applies only to @code{protoize}.
12216 @item -i @var{string}
12217 Indent old-style parameter declarations with the string @var{string}.
12218 This option applies only to @code{protoize}.
12220 @code{unprotoize} converts prototyped function definitions to old-style
12221 function definitions, where the arguments are declared between the
12222 argument list and the initial @samp{@{}. By default, @code{unprotoize}
12223 uses five spaces as the indentation. If you want to indent with just
12224 one space instead, use @option{-i " "}.
12227 Keep the @samp{.X} files. Normally, they are deleted after conversion
12231 Add explicit local declarations. @code{protoize} with @option{-l} inserts
12232 a prototype declaration for each function in each block which calls the
12233 function without any declaration. This option applies only to
12237 Make no real changes. This mode just prints information about the conversions
12238 that would have been done without @option{-n}.
12241 Make no @samp{.save} files. The original files are simply deleted.
12242 Use this option with caution.
12244 @item -p @var{program}
12245 Use the program @var{program} as the compiler. Normally, the name
12246 @file{gcc} is used.
12249 Work quietly. Most warnings are suppressed.
12252 Print the version number, just like @option{-v} for @command{gcc}.
12255 If you need special compiler options to compile one of your program's
12256 source files, then you should generate that file's @samp{.X} file
12257 specially, by running @command{gcc} on that source file with the
12258 appropriate options and the option @option{-aux-info}. Then run
12259 @code{protoize} on the entire set of files. @code{protoize} will use
12260 the existing @samp{.X} file because it is newer than the source file.
12264 gcc -Dfoo=bar file1.c -aux-info file1.X
12269 You need to include the special files along with the rest in the
12270 @code{protoize} command, even though their @samp{.X} files already
12271 exist, because otherwise they won't get converted.
12273 @xref{Protoize Caveats}, for more information on how to use
12274 @code{protoize} successfully.