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 Dialect Options:: Variations on Objective-C.
125 * Language Independent Options:: Controlling how diagnostics should be
127 * Warning Options:: How picky should the compiler be?
128 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
129 * Optimize Options:: How much optimization?
130 * Preprocessor Options:: Controlling header files and macro definitions.
131 Also, getting dependency information for Make.
132 * Assembler Options:: Passing options to the assembler.
133 * Link Options:: Specifying libraries and so on.
134 * Directory Options:: Where to find header files and libraries.
135 Where to find the compiler executable files.
136 * Spec Files:: How to pass switches to sub-processes.
137 * Target Options:: Running a cross-compiler, or an old version of GCC.
138 * Submodel Options:: Specifying minor hardware or convention variations,
139 such as 68010 vs 68020.
140 * Code Gen Options:: Specifying conventions for function calls, data layout
142 * Environment Variables:: Env vars that affect GCC.
143 * Precompiled Headers:: Compiling a header once, and using it many times.
144 * Running Protoize:: Automatically adding or removing function prototypes.
150 @section Option Summary
152 Here is a summary of all the options, grouped by type. Explanations are
153 in the following sections.
156 @item Overall Options
157 @xref{Overall Options,,Options Controlling the Kind of Output}.
158 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
159 -x @var{language} -v -### --help --target-help --version}
161 @item C Language Options
162 @xref{C Dialect Options,,Options Controlling C Dialect}.
163 @gccoptlist{-ansi -std=@var{standard} -aux-info @var{filename} @gol
164 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
165 -fhosted -ffreestanding -fms-extensions @gol
166 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
167 -fallow-single-precision -fcond-mismatch @gol
168 -fsigned-bitfields -fsigned-char @gol
169 -funsigned-bitfields -funsigned-char}
171 @item C++ Language Options
172 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
173 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
174 -fconserve-space -fno-const-strings @gol
175 -fno-elide-constructors @gol
176 -fno-enforce-eh-specs @gol
177 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
178 -fno-implicit-templates @gol
179 -fno-implicit-inline-templates @gol
180 -fno-implement-inlines -fms-extensions @gol
181 -fno-nonansi-builtins -fno-operator-names @gol
182 -fno-optional-diags -fpermissive @gol
183 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
184 -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
185 -fno-default-inline -fvisibility-inlines-hidden @gol
186 -Wabi -Wctor-dtor-privacy @gol
187 -Wnon-virtual-dtor -Wreorder @gol
188 -Weffc++ -Wno-deprecated @gol
189 -Wno-non-template-friend -Wold-style-cast @gol
190 -Woverloaded-virtual -Wno-pmf-conversions @gol
191 -Wsign-promo -Wsynth}
193 @item Objective-C Language Options
194 @xref{Objective-C Dialect Options,,Options Controlling Objective-C Dialect}.
196 -fconstant-string-class=@var{class-name} @gol
197 -fgnu-runtime -fnext-runtime @gol
198 -fno-nil-receivers @gol
199 -fobjc-exceptions @gol
200 -freplace-objc-classes @gol
203 -Wno-protocol -Wselector -Wundeclared-selector}
205 @item Language Independent Options
206 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
207 @gccoptlist{-fmessage-length=@var{n} @gol
208 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}}
210 @item Warning Options
211 @xref{Warning Options,,Options to Request or Suppress Warnings}.
212 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
213 -w -Wextra -Wall -Waggregate-return @gol
214 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
215 -Wconversion -Wno-deprecated-declarations @gol
216 -Wdisabled-optimization -Wno-div-by-zero -Wendif-labels @gol
217 -Werror -Werror-implicit-function-declaration @gol
218 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
219 -Wno-format-extra-args -Wformat-nonliteral @gol
220 -Wformat-security -Wformat-y2k @gol
221 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
222 -Wimport -Wno-import -Winit-self -Winline @gol
223 -Wno-invalid-offsetof -Winvalid-pch @gol
224 -Wlarger-than-@var{len} -Wlong-long @gol
225 -Wmain -Wmissing-braces @gol
226 -Wmissing-format-attribute -Wmissing-include-dirs @gol
227 -Wmissing-noreturn @gol
228 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
229 -Wparentheses -Wpointer-arith -Wredundant-decls @gol
230 -Wreturn-type -Wsequence-point -Wshadow @gol
231 -Wsign-compare -Wstrict-aliasing -Wstrict-aliasing=2 @gol
232 -Wswitch -Wswitch-default -Wswitch-enum @gol
233 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
234 -Wunknown-pragmas -Wunreachable-code @gol
235 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
236 -Wunused-value -Wunused-variable -Wwrite-strings @gol
239 @item C-only Warning Options
240 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
241 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
242 -Wstrict-prototypes -Wtraditional @gol
243 -Wdeclaration-after-statement}
245 @item Debugging Options
246 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
247 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
248 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
249 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
251 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
252 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
253 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
254 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
256 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
257 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
258 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
259 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
260 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
261 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
262 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
263 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
266 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
267 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
268 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs -ftree-based-profiling @gol
269 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
270 -ftest-coverage -ftime-report -fvar-tracking @gol
271 -g -g@var{level} -gcoff -gdwarf-2 @gol
272 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
273 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
274 -print-multi-directory -print-multi-lib @gol
275 -print-prog-name=@var{program} -print-search-dirs -Q @gol
278 @item Optimization Options
279 @xref{Optimize Options,,Options that Control Optimization}.
280 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
281 -falign-labels=@var{n} -falign-loops=@var{n} @gol
282 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
283 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
284 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
285 -fcaller-saves -fcprop-registers @gol
286 -fcse-follow-jumps -fcse-skip-blocks -fdata-sections @gol
287 -fdelayed-branch -fdelete-null-pointer-checks @gol
288 -fexpensive-optimizations -ffast-math -ffloat-store @gol
289 -fforce-addr -fforce-mem -ffunction-sections @gol
290 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
291 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2 @gol
292 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
293 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
294 -fmodulo-sched -fmove-all-movables -fnew-ra -fno-branch-count-reg @gol
295 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants @gol
296 -fno-function-cse -fno-guess-branch-probability @gol
297 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
298 -funsafe-math-optimizations -ffinite-math-only @gol
299 -fno-trapping-math -fno-zero-initialized-in-bss @gol
300 -fomit-frame-pointer -foptimize-register-move @gol
301 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
302 -fprofile-generate -fprofile-use @gol
303 -freduce-all-givs -fregmove -frename-registers @gol
304 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
305 -frerun-cse-after-loop -frerun-loop-opt @gol
306 -frounding-math -fschedule-insns -fschedule-insns2 @gol
307 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
308 -fsched-spec-load-dangerous @gol
309 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
310 -fsched2-use-superblocks @gol
311 -fsched2-use-traces -fsignaling-nans @gol
312 -fsingle-precision-constant @gol
313 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
314 -funroll-all-loops -funroll-loops -fpeel-loops @gol
315 -funswitch-loops -fold-unroll-loops -fold-unroll-all-loops @gol
316 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
318 -ftree-dominator-opts -ftree-dse -ftree-copyrename @gol
319 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre @gol
320 --param @var{name}=@var{value}
321 -O -O0 -O1 -O2 -O3 -Os}
323 @item Preprocessor Options
324 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
325 @gccoptlist{-A@var{question}=@var{answer} @gol
326 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
327 -C -dD -dI -dM -dN @gol
328 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
329 -idirafter @var{dir} @gol
330 -include @var{file} -imacros @var{file} @gol
331 -iprefix @var{file} -iwithprefix @var{dir} @gol
332 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
333 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
334 -P -fworking-directory -remap @gol
335 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
336 -Xpreprocessor @var{option}}
338 @item Assembler Option
339 @xref{Assembler Options,,Passing Options to the Assembler}.
340 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
343 @xref{Link Options,,Options for Linking}.
344 @gccoptlist{@var{object-file-name} -l@var{library} @gol
345 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
346 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
347 -Wl,@var{option} -Xlinker @var{option} @gol
350 @item Directory Options
351 @xref{Directory Options,,Options for Directory Search}.
352 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-}
355 @c I wrote this xref this way to avoid overfull hbox. -- rms
356 @xref{Target Options}.
357 @gccoptlist{-V @var{version} -b @var{machine}}
359 @item Machine Dependent Options
360 @xref{Submodel Options,,Hardware Models and Configurations}.
361 @c This list is ordered alphanumerically by subsection name.
362 @c Try and put the significant identifier (CPU or system) first,
363 @c so users have a clue at guessing where the ones they want will be.
366 @gccoptlist{-EB -EL @gol
367 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
368 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
371 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
372 -mabi=@var{name} @gol
373 -mapcs-stack-check -mno-apcs-stack-check @gol
374 -mapcs-float -mno-apcs-float @gol
375 -mapcs-reentrant -mno-apcs-reentrant @gol
376 -msched-prolog -mno-sched-prolog @gol
377 -mlittle-endian -mbig-endian -mwords-little-endian @gol
378 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
379 -mthumb-interwork -mno-thumb-interwork @gol
380 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
381 -mstructure-size-boundary=@var{n} @gol
382 -mabort-on-noreturn @gol
383 -mlong-calls -mno-long-calls @gol
384 -msingle-pic-base -mno-single-pic-base @gol
385 -mpic-register=@var{reg} @gol
386 -mnop-fun-dllimport @gol
387 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
388 -mpoke-function-name @gol
390 -mtpcs-frame -mtpcs-leaf-frame @gol
391 -mcaller-super-interworking -mcallee-super-interworking}
394 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
395 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
398 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
399 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
400 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
401 -mstack-align -mdata-align -mconst-align @gol
402 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
403 -melf -maout -melinux -mlinux -sim -sim2 @gol
404 -mmul-bug-workaround -mno-mul-bug-workaround}
406 @emph{Darwin Options}
407 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
408 -arch_only -bind_at_load -bundle -bundle_loader @gol
409 -client_name -compatibility_version -current_version @gol
410 -dependency-file -dylib_file -dylinker_install_name @gol
411 -dynamic -dynamiclib -exported_symbols_list @gol
412 -filelist -flat_namespace -force_cpusubtype_ALL @gol
413 -force_flat_namespace -headerpad_max_install_names @gol
414 -image_base -init -install_name -keep_private_externs @gol
415 -multi_module -multiply_defined -multiply_defined_unused @gol
416 -noall_load -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
417 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
418 -private_bundle -read_only_relocs -sectalign @gol
419 -sectobjectsymbols -whyload -seg1addr @gol
420 -sectcreate -sectobjectsymbols -sectorder @gol
421 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
422 -segprot -segs_read_only_addr -segs_read_write_addr @gol
423 -single_module -static -sub_library -sub_umbrella @gol
424 -twolevel_namespace -umbrella -undefined @gol
425 -unexported_symbols_list -weak_reference_mismatches @gol
426 -whatsloaded -F -gused -gfull -mone-byte-bool}
428 @emph{DEC Alpha Options}
429 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
430 -mieee -mieee-with-inexact -mieee-conformant @gol
431 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
432 -mtrap-precision=@var{mode} -mbuild-constants @gol
433 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
434 -mbwx -mmax -mfix -mcix @gol
435 -mfloat-vax -mfloat-ieee @gol
436 -mexplicit-relocs -msmall-data -mlarge-data @gol
437 -msmall-text -mlarge-text @gol
438 -mmemory-latency=@var{time}}
440 @emph{DEC Alpha/VMS Options}
441 @gccoptlist{-mvms-return-codes}
444 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
445 -mhard-float -msoft-float @gol
446 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
447 -mdouble -mno-double @gol
448 -mmedia -mno-media -mmuladd -mno-muladd @gol
449 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic -mlinked-fp @gol
450 -mlibrary-pic -macc-4 -macc-8 @gol
451 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
452 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
453 -mvliw-branch -mno-vliw-branch @gol
454 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
455 -mno-nested-cond-exec -mtomcat-stats @gol
458 @emph{H8/300 Options}
459 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
462 @gccoptlist{-march=@var{architecture-type} @gol
463 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
464 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
465 -mfixed-range=@var{register-range} @gol
466 -mjump-in-delay -mlinker-opt -mlong-calls @gol
467 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
468 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
469 -mno-jump-in-delay -mno-long-load-store @gol
470 -mno-portable-runtime -mno-soft-float @gol
471 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
472 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
473 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
474 -nolibdld -static -threads}
476 @emph{i386 and x86-64 Options}
477 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
478 -mfpmath=@var{unit} @gol
479 -masm=@var{dialect} -mno-fancy-math-387 @gol
480 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
481 -mno-wide-multiply -mrtd -malign-double @gol
482 -mpreferred-stack-boundary=@var{num} @gol
483 -mmmx -msse -msse2 -msse3 -m3dnow @gol
484 -mthreads -mno-align-stringops -minline-all-stringops @gol
485 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
486 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
487 -mno-red-zone -mno-tls-direct-seg-refs @gol
488 -mcmodel=@var{code-model} @gol
492 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
493 -mvolatile-asm-stop -mb-step -mregister-names -mno-sdata @gol
494 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
495 -minline-float-divide-max-throughput @gol
496 -minline-int-divide-min-latency @gol
497 -minline-int-divide-max-throughput -mno-dwarf2-asm @gol
498 -mfixed-range=@var{register-range}}
500 @emph{M32R/D Options}
501 @gccoptlist{-m32r2 -m32rx -m32r @gol
503 -malign-loops -mno-align-loops @gol
504 -missue-rate=@var{number} @gol
505 -mbranch-cost=@var{number} @gol
506 -mmodel=@var{code-size-model-type} @gol
507 -msdata=@var{sdata-type} @gol
508 -mno-flush-func -mflush-func=@var{name} @gol
509 -mno-flush-trap -mflush-trap=@var{number} @gol
512 @emph{M680x0 Options}
513 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
514 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
515 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
516 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
517 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
519 @emph{M68hc1x Options}
520 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
521 -mauto-incdec -minmax -mlong-calls -mshort @gol
522 -msoft-reg-count=@var{count}}
525 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
526 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
527 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
528 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
529 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
532 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
533 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
534 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
535 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
536 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
537 -mint64 -mlong64 -mlong32 @gol
538 -G@var{num} -membedded-data -mno-embedded-data @gol
539 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
540 -msplit-addresses -mno-split-addresses @gol
541 -mexplicit-relocs -mno-explicit-relocs @gol
542 -mcheck-zero-division -mno-check-zero-division @gol
543 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
544 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
545 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
546 -mfix-vr4120 -mno-fix-vr4120 -mfix-sb1 -mno-fix-sb1 @gol
547 -mflush-func=@var{func} -mno-flush-func @gol
548 -mbranch-likely -mno-branch-likely @gol
549 -mfp-exceptions -mno-fp-exceptions @gol
550 -mvr4130-align -mno-vr4130-align}
553 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
554 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
555 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
556 -mno-base-addresses -msingle-exit -mno-single-exit}
558 @emph{MN10300 Options}
559 @gccoptlist{-mmult-bug -mno-mult-bug @gol
560 -mam33 -mno-am33 @gol
561 -mam33-2 -mno-am33-2 @gol
565 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
566 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
567 -mregparam -mnoregparam -msb -mnosb @gol
568 -mbitfield -mnobitfield -mhimem -mnohimem}
570 @emph{PDP-11 Options}
571 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
572 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
573 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
574 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
575 -mbranch-expensive -mbranch-cheap @gol
576 -msplit -mno-split -munix-asm -mdec-asm}
578 @emph{PowerPC Options}
579 See RS/6000 and PowerPC Options.
581 @emph{RS/6000 and PowerPC Options}
582 @gccoptlist{-mcpu=@var{cpu-type} @gol
583 -mtune=@var{cpu-type} @gol
584 -mpower -mno-power -mpower2 -mno-power2 @gol
585 -mpowerpc -mpowerpc64 -mno-powerpc @gol
586 -maltivec -mno-altivec @gol
587 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
588 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
589 -mnew-mnemonics -mold-mnemonics @gol
590 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
591 -m64 -m32 -mxl-call -mno-xl-call -mpe @gol
592 -malign-power -malign-natural @gol
593 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
594 -mstring -mno-string -mupdate -mno-update @gol
595 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
596 -mstrict-align -mno-strict-align -mrelocatable @gol
597 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
598 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
599 -mdynamic-no-pic @gol
600 -mprioritize-restricted-insns=@var{priority} @gol
601 -msched-costly-dep=@var{dependence_type} @gol
602 -minsert-sched-nops=@var{scheme} @gol
603 -mcall-sysv -mcall-netbsd @gol
604 -maix-struct-return -msvr4-struct-return @gol
605 -mabi=altivec -mabi=no-altivec @gol
606 -mabi=spe -mabi=no-spe @gol
607 -misel=yes -misel=no @gol
608 -mspe=yes -mspe=no @gol
609 -mfloat-gprs=yes -mfloat-gprs=no @gol
610 -mprototype -mno-prototype @gol
611 -msim -mmvme -mads -myellowknife -memb -msdata @gol
612 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
614 @emph{S/390 and zSeries Options}
615 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
616 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
617 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
618 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
619 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd}
622 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
623 -m4-nofpu -m4-single-only -m4-single -m4 @gol
624 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
625 -m5-64media -m5-64media-nofpu @gol
626 -m5-32media -m5-32media-nofpu @gol
627 -m5-compact -m5-compact-nofpu @gol
628 -mb -ml -mdalign -mrelax @gol
629 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
630 -mieee -misize -mpadstruct -mspace @gol
631 -mprefergot -musermode}
634 @gccoptlist{-mcpu=@var{cpu-type} @gol
635 -mtune=@var{cpu-type} @gol
636 -mcmodel=@var{code-model} @gol
637 -m32 -m64 -mapp-regs -mno-app-regs @gol
638 -mfaster-structs -mno-faster-structs @gol
639 -mfpu -mno-fpu -mhard-float -msoft-float @gol
640 -mhard-quad-float -msoft-quad-float @gol
641 -mimpure-text -mno-impure-text -mlittle-endian @gol
642 -mstack-bias -mno-stack-bias @gol
643 -munaligned-doubles -mno-unaligned-doubles @gol
644 -mv8plus -mno-v8plus -mvis -mno-vis
647 @emph{System V Options}
648 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
650 @emph{TMS320C3x/C4x Options}
651 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
652 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
653 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
654 -mparallel-insns -mparallel-mpy -mpreserve-float}
657 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
658 -mprolog-function -mno-prolog-function -mspace @gol
659 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
660 -mapp-regs -mno-app-regs @gol
661 -mdisable-callt -mno-disable-callt @gol
667 @gccoptlist{-mg -mgnu -munix}
669 @emph{x86-64 Options}
670 See i386 and x86-64 Options.
672 @emph{Xstormy16 Options}
675 @emph{Xtensa Options}
676 @gccoptlist{-mconst16 -mno-const16 @gol
677 -mfused-madd -mno-fused-madd @gol
678 -mtext-section-literals -mno-text-section-literals @gol
679 -mtarget-align -mno-target-align @gol
680 -mlongcalls -mno-longcalls}
682 @emph{zSeries Options}
683 See S/390 and zSeries Options.
685 @item Code Generation Options
686 @xref{Code Gen Options,,Options for Code Generation Conventions}.
687 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
688 -ffixed-@var{reg} -fexceptions @gol
689 -fnon-call-exceptions -funwind-tables @gol
690 -fasynchronous-unwind-tables @gol
691 -finhibit-size-directive -finstrument-functions @gol
692 -fno-common -fno-ident @gol
693 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
694 -freg-struct-return -fshared-data -fshort-enums @gol
695 -fshort-double -fshort-wchar @gol
696 -fverbose-asm -fpack-struct -fstack-check @gol
697 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
698 -fargument-alias -fargument-noalias @gol
699 -fargument-noalias-global -fleading-underscore @gol
700 -ftls-model=@var{model} @gol
701 -ftrapv -fwrapv -fbounds-check @gol
706 * Overall Options:: Controlling the kind of output:
707 an executable, object files, assembler files,
708 or preprocessed source.
709 * C Dialect Options:: Controlling the variant of C language compiled.
710 * C++ Dialect Options:: Variations on C++.
711 * Objective-C Dialect Options:: Variations on Objective-C.
712 * Language Independent Options:: Controlling how diagnostics should be
714 * Warning Options:: How picky should the compiler be?
715 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
716 * Optimize Options:: How much optimization?
717 * Preprocessor Options:: Controlling header files and macro definitions.
718 Also, getting dependency information for Make.
719 * Assembler Options:: Passing options to the assembler.
720 * Link Options:: Specifying libraries and so on.
721 * Directory Options:: Where to find header files and libraries.
722 Where to find the compiler executable files.
723 * Spec Files:: How to pass switches to sub-processes.
724 * Target Options:: Running a cross-compiler, or an old version of GCC.
727 @node Overall Options
728 @section Options Controlling the Kind of Output
730 Compilation can involve up to four stages: preprocessing, compilation
731 proper, assembly and linking, always in that order. GCC is capable of
732 preprocessing and compiling several files either into several
733 assembler input files, or into one assembler input file; then each
734 assembler input file produces an object file, and linking combines all
735 the object files (those newly compiled, and those specified as input)
736 into an executable file.
738 @cindex file name suffix
739 For any given input file, the file name suffix determines what kind of
744 C source code which must be preprocessed.
747 C source code which should not be preprocessed.
750 C++ source code which should not be preprocessed.
753 Objective-C source code. Note that you must link with the library
754 @file{libobjc.a} to make an Objective-C program work.
757 Objective-C source code which should not be preprocessed.
760 C or C++ header file to be turned into a precompiled header.
764 @itemx @var{file}.cxx
765 @itemx @var{file}.cpp
766 @itemx @var{file}.CPP
767 @itemx @var{file}.c++
769 C++ source code which must be preprocessed. Note that in @samp{.cxx},
770 the last two letters must both be literally @samp{x}. Likewise,
771 @samp{.C} refers to a literal capital C@.
775 C++ header file to be turned into a precompiled header.
778 @itemx @var{file}.for
779 @itemx @var{file}.FOR
780 Fortran source code which should not be preprocessed.
783 @itemx @var{file}.fpp
784 @itemx @var{file}.FPP
785 Fortran source code which must be preprocessed (with the traditional
789 Fortran source code which must be preprocessed with a RATFOR
790 preprocessor (not included with GCC)@.
793 @itemx @var{file}.f95
794 Fortran 90/95 source code which should not be preprocessed.
796 @c FIXME: Descriptions of Java file types.
803 Ada source code file which contains a library unit declaration (a
804 declaration of a package, subprogram, or generic, or a generic
805 instantiation), or a library unit renaming declaration (a package,
806 generic, or subprogram renaming declaration). Such files are also
809 @itemx @var{file}.adb
810 Ada source code file containing a library unit body (a subprogram or
811 package body). Such files are also called @dfn{bodies}.
813 @c GCC also knows about some suffixes for languages not yet included:
822 Assembler code which must be preprocessed.
825 An object file to be fed straight into linking.
826 Any file name with no recognized suffix is treated this way.
830 You can specify the input language explicitly with the @option{-x} option:
833 @item -x @var{language}
834 Specify explicitly the @var{language} for the following input files
835 (rather than letting the compiler choose a default based on the file
836 name suffix). This option applies to all following input files until
837 the next @option{-x} option. Possible values for @var{language} are:
839 c c-header cpp-output
840 c++ c++-header c++-cpp-output
841 objective-c objective-c-header objc-cpp-output
842 assembler assembler-with-cpp
844 f77 f77-cpp-input ratfor
851 Turn off any specification of a language, so that subsequent files are
852 handled according to their file name suffixes (as they are if @option{-x}
853 has not been used at all).
855 @item -pass-exit-codes
856 @opindex pass-exit-codes
857 Normally the @command{gcc} program will exit with the code of 1 if any
858 phase of the compiler returns a non-success return code. If you specify
859 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
860 numerically highest error produced by any phase that returned an error
864 If you only want some of the stages of compilation, you can use
865 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
866 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
867 @command{gcc} is to stop. Note that some combinations (for example,
868 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
873 Compile or assemble the source files, but do not link. The linking
874 stage simply is not done. The ultimate output is in the form of an
875 object file for each source file.
877 By default, the object file name for a source file is made by replacing
878 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
880 Unrecognized input files, not requiring compilation or assembly, are
885 Stop after the stage of compilation proper; do not assemble. The output
886 is in the form of an assembler code file for each non-assembler input
889 By default, the assembler file name for a source file is made by
890 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
892 Input files that don't require compilation are ignored.
896 Stop after the preprocessing stage; do not run the compiler proper. The
897 output is in the form of preprocessed source code, which is sent to the
900 Input files which don't require preprocessing are ignored.
902 @cindex output file option
905 Place output in file @var{file}. This applies regardless to whatever
906 sort of output is being produced, whether it be an executable file,
907 an object file, an assembler file or preprocessed C code.
909 If @option{-o} is not specified, the default is to put an executable
910 file in @file{a.out}, the object file for
911 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
912 assembler file in @file{@var{source}.s}, a precompiled header file in
913 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
918 Print (on standard error output) the commands executed to run the stages
919 of compilation. Also print the version number of the compiler driver
920 program and of the preprocessor and the compiler proper.
924 Like @option{-v} except the commands are not executed and all command
925 arguments are quoted. This is useful for shell scripts to capture the
926 driver-generated command lines.
930 Use pipes rather than temporary files for communication between the
931 various stages of compilation. This fails to work on some systems where
932 the assembler is unable to read from a pipe; but the GNU assembler has
937 If you are compiling multiple source files, this option tells the driver
938 to pass all the source files to the compiler at once (for those
939 languages for which the compiler can handle this). This will allow
940 intermodule analysis (IMA) to be performed by the compiler. Currently the only
941 language for which this is supported is C. If you pass source files for
942 multiple languages to the driver, using this option, the driver will invoke
943 the compiler(s) that support IMA once each, passing each compiler all the
944 source files appropriate for it. For those languages that do not support
945 IMA this option will be ignored, and the compiler will be invoked once for
946 each source file in that language. If you use this option in conjunction
947 with -save-temps, the compiler will generate multiple pre-processed files
948 (one for each source file), but only one (combined) .o or .s file.
952 Print (on the standard output) a description of the command line options
953 understood by @command{gcc}. If the @option{-v} option is also specified
954 then @option{--help} will also be passed on to the various processes
955 invoked by @command{gcc}, so that they can display the command line options
956 they accept. If the @option{-Wextra} option is also specified then command
957 line options which have no documentation associated with them will also
962 Print (on the standard output) a description of target specific command
963 line options for each tool.
967 Display the version number and copyrights of the invoked GCC.
971 @section Compiling C++ Programs
973 @cindex suffixes for C++ source
974 @cindex C++ source file suffixes
975 C++ source files conventionally use one of the suffixes @samp{.C},
976 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
977 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
978 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
979 files with these names and compiles them as C++ programs even if you
980 call the compiler the same way as for compiling C programs (usually
981 with the name @command{gcc}).
985 However, C++ programs often require class libraries as well as a
986 compiler that understands the C++ language---and under some
987 circumstances, you might want to compile programs or header files from
988 standard input, or otherwise without a suffix that flags them as C++
989 programs. You might also like to precompile a C header file with a
990 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
991 program that calls GCC with the default language set to C++, and
992 automatically specifies linking against the C++ library. On many
993 systems, @command{g++} is also installed with the name @command{c++}.
995 @cindex invoking @command{g++}
996 When you compile C++ programs, you may specify many of the same
997 command-line options that you use for compiling programs in any
998 language; or command-line options meaningful for C and related
999 languages; or options that are meaningful only for C++ programs.
1000 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1001 explanations of options for languages related to C@.
1002 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1003 explanations of options that are meaningful only for C++ programs.
1005 @node C Dialect Options
1006 @section Options Controlling C Dialect
1007 @cindex dialect options
1008 @cindex language dialect options
1009 @cindex options, dialect
1011 The following options control the dialect of C (or languages derived
1012 from C, such as C++ and Objective-C) that the compiler accepts:
1015 @cindex ANSI support
1019 In C mode, support all ISO C90 programs. In C++ mode,
1020 remove GNU extensions that conflict with ISO C++.
1022 This turns off certain features of GCC that are incompatible with ISO
1023 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1024 such as the @code{asm} and @code{typeof} keywords, and
1025 predefined macros such as @code{unix} and @code{vax} that identify the
1026 type of system you are using. It also enables the undesirable and
1027 rarely used ISO trigraph feature. For the C compiler,
1028 it disables recognition of C++ style @samp{//} comments as well as
1029 the @code{inline} keyword.
1031 The alternate keywords @code{__asm__}, @code{__extension__},
1032 @code{__inline__} and @code{__typeof__} continue to work despite
1033 @option{-ansi}. You would not want to use them in an ISO C program, of
1034 course, but it is useful to put them in header files that might be included
1035 in compilations done with @option{-ansi}. Alternate predefined macros
1036 such as @code{__unix__} and @code{__vax__} are also available, with or
1037 without @option{-ansi}.
1039 The @option{-ansi} option does not cause non-ISO programs to be
1040 rejected gratuitously. For that, @option{-pedantic} is required in
1041 addition to @option{-ansi}. @xref{Warning Options}.
1043 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1044 option is used. Some header files may notice this macro and refrain
1045 from declaring certain functions or defining certain macros that the
1046 ISO standard doesn't call for; this is to avoid interfering with any
1047 programs that might use these names for other things.
1049 Functions which would normally be built in but do not have semantics
1050 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1051 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1052 built-in functions provided by GCC}, for details of the functions
1057 Determine the language standard. This option is currently only
1058 supported when compiling C or C++. A value for this option must be
1059 provided; possible values are
1064 ISO C90 (same as @option{-ansi}).
1066 @item iso9899:199409
1067 ISO C90 as modified in amendment 1.
1073 ISO C99. Note that this standard is not yet fully supported; see
1074 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1075 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1078 Default, ISO C90 plus GNU extensions (including some C99 features).
1082 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1083 this will become the default. The name @samp{gnu9x} is deprecated.
1086 The 1998 ISO C++ standard plus amendments.
1089 The same as @option{-std=c++98} plus GNU extensions. This is the
1090 default for C++ code.
1093 Even when this option is not specified, you can still use some of the
1094 features of newer standards in so far as they do not conflict with
1095 previous C standards. For example, you may use @code{__restrict__} even
1096 when @option{-std=c99} is not specified.
1098 The @option{-std} options specifying some version of ISO C have the same
1099 effects as @option{-ansi}, except that features that were not in ISO C90
1100 but are in the specified version (for example, @samp{//} comments and
1101 the @code{inline} keyword in ISO C99) are not disabled.
1103 @xref{Standards,,Language Standards Supported by GCC}, for details of
1104 these standard versions.
1106 @item -aux-info @var{filename}
1108 Output to the given filename prototyped declarations for all functions
1109 declared and/or defined in a translation unit, including those in header
1110 files. This option is silently ignored in any language other than C@.
1112 Besides declarations, the file indicates, in comments, the origin of
1113 each declaration (source file and line), whether the declaration was
1114 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1115 @samp{O} for old, respectively, in the first character after the line
1116 number and the colon), and whether it came from a declaration or a
1117 definition (@samp{C} or @samp{F}, respectively, in the following
1118 character). In the case of function definitions, a K&R-style list of
1119 arguments followed by their declarations is also provided, inside
1120 comments, after the declaration.
1124 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1125 keyword, so that code can use these words as identifiers. You can use
1126 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1127 instead. @option{-ansi} implies @option{-fno-asm}.
1129 In C++, this switch only affects the @code{typeof} keyword, since
1130 @code{asm} and @code{inline} are standard keywords. You may want to
1131 use the @option{-fno-gnu-keywords} flag instead, which has the same
1132 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1133 switch only affects the @code{asm} and @code{typeof} keywords, since
1134 @code{inline} is a standard keyword in ISO C99.
1137 @itemx -fno-builtin-@var{function}
1138 @opindex fno-builtin
1139 @cindex built-in functions
1140 Don't recognize built-in functions that do not begin with
1141 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1142 functions provided by GCC}, for details of the functions affected,
1143 including those which are not built-in functions when @option{-ansi} or
1144 @option{-std} options for strict ISO C conformance are used because they
1145 do not have an ISO standard meaning.
1147 GCC normally generates special code to handle certain built-in functions
1148 more efficiently; for instance, calls to @code{alloca} may become single
1149 instructions that adjust the stack directly, and calls to @code{memcpy}
1150 may become inline copy loops. The resulting code is often both smaller
1151 and faster, but since the function calls no longer appear as such, you
1152 cannot set a breakpoint on those calls, nor can you change the behavior
1153 of the functions by linking with a different library.
1155 With the @option{-fno-builtin-@var{function}} option
1156 only the built-in function @var{function} is
1157 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1158 function is named this is not built-in in this version of GCC, this
1159 option is ignored. There is no corresponding
1160 @option{-fbuiltin-@var{function}} option; if you wish to enable
1161 built-in functions selectively when using @option{-fno-builtin} or
1162 @option{-ffreestanding}, you may define macros such as:
1165 #define abs(n) __builtin_abs ((n))
1166 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1171 @cindex hosted environment
1173 Assert that compilation takes place in a hosted environment. This implies
1174 @option{-fbuiltin}. A hosted environment is one in which the
1175 entire standard library is available, and in which @code{main} has a return
1176 type of @code{int}. Examples are nearly everything except a kernel.
1177 This is equivalent to @option{-fno-freestanding}.
1179 @item -ffreestanding
1180 @opindex ffreestanding
1181 @cindex hosted environment
1183 Assert that compilation takes place in a freestanding environment. This
1184 implies @option{-fno-builtin}. A freestanding environment
1185 is one in which the standard library may not exist, and program startup may
1186 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1187 This is equivalent to @option{-fno-hosted}.
1189 @xref{Standards,,Language Standards Supported by GCC}, for details of
1190 freestanding and hosted environments.
1192 @item -fms-extensions
1193 @opindex fms-extensions
1194 Accept some non-standard constructs used in Microsoft header files.
1198 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1199 options for strict ISO C conformance) implies @option{-trigraphs}.
1201 @item -no-integrated-cpp
1202 @opindex no-integrated-cpp
1203 Performs a compilation in two passes: preprocessing and compiling. This
1204 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1205 @option{-B} option. The user supplied compilation step can then add in
1206 an additional preprocessing step after normal preprocessing but before
1207 compiling. The default is to use the integrated cpp (internal cpp)
1209 The semantics of this option will change if "cc1", "cc1plus", and
1210 "cc1obj" are merged.
1212 @cindex traditional C language
1213 @cindex C language, traditional
1215 @itemx -traditional-cpp
1216 @opindex traditional-cpp
1217 @opindex traditional
1218 Formerly, these options caused GCC to attempt to emulate a pre-standard
1219 C compiler. They are now only supported with the @option{-E} switch.
1220 The preprocessor continues to support a pre-standard mode. See the GNU
1221 CPP manual for details.
1223 @item -fcond-mismatch
1224 @opindex fcond-mismatch
1225 Allow conditional expressions with mismatched types in the second and
1226 third arguments. The value of such an expression is void. This option
1227 is not supported for C++.
1229 @item -funsigned-char
1230 @opindex funsigned-char
1231 Let the type @code{char} be unsigned, like @code{unsigned char}.
1233 Each kind of machine has a default for what @code{char} should
1234 be. It is either like @code{unsigned char} by default or like
1235 @code{signed char} by default.
1237 Ideally, a portable program should always use @code{signed char} or
1238 @code{unsigned char} when it depends on the signedness of an object.
1239 But many programs have been written to use plain @code{char} and
1240 expect it to be signed, or expect it to be unsigned, depending on the
1241 machines they were written for. This option, and its inverse, let you
1242 make such a program work with the opposite default.
1244 The type @code{char} is always a distinct type from each of
1245 @code{signed char} or @code{unsigned char}, even though its behavior
1246 is always just like one of those two.
1249 @opindex fsigned-char
1250 Let the type @code{char} be signed, like @code{signed char}.
1252 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1253 the negative form of @option{-funsigned-char}. Likewise, the option
1254 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1256 @item -fsigned-bitfields
1257 @itemx -funsigned-bitfields
1258 @itemx -fno-signed-bitfields
1259 @itemx -fno-unsigned-bitfields
1260 @opindex fsigned-bitfields
1261 @opindex funsigned-bitfields
1262 @opindex fno-signed-bitfields
1263 @opindex fno-unsigned-bitfields
1264 These options control whether a bit-field is signed or unsigned, when the
1265 declaration does not use either @code{signed} or @code{unsigned}. By
1266 default, such a bit-field is signed, because this is consistent: the
1267 basic integer types such as @code{int} are signed types.
1270 @node C++ Dialect Options
1271 @section Options Controlling C++ Dialect
1273 @cindex compiler options, C++
1274 @cindex C++ options, command line
1275 @cindex options, C++
1276 This section describes the command-line options that are only meaningful
1277 for C++ programs; but you can also use most of the GNU compiler options
1278 regardless of what language your program is in. For example, you
1279 might compile a file @code{firstClass.C} like this:
1282 g++ -g -frepo -O -c firstClass.C
1286 In this example, only @option{-frepo} is an option meant
1287 only for C++ programs; you can use the other options with any
1288 language supported by GCC@.
1290 Here is a list of options that are @emph{only} for compiling C++ programs:
1294 @item -fabi-version=@var{n}
1295 @opindex fabi-version
1296 Use version @var{n} of the C++ ABI. Version 2 is the version of the
1297 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1298 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1299 the version that conforms most closely to the C++ ABI specification.
1300 Therefore, the ABI obtained using version 0 will change as ABI bugs
1303 The default is version 2.
1305 @item -fno-access-control
1306 @opindex fno-access-control
1307 Turn off all access checking. This switch is mainly useful for working
1308 around bugs in the access control code.
1312 Check that the pointer returned by @code{operator new} is non-null
1313 before attempting to modify the storage allocated. This check is
1314 normally unnecessary because the C++ standard specifies that
1315 @code{operator new} will only return @code{0} if it is declared
1316 @samp{throw()}, in which case the compiler will always check the
1317 return value even without this option. In all other cases, when
1318 @code{operator new} has a non-empty exception specification, memory
1319 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1320 @samp{new (nothrow)}.
1322 @item -fconserve-space
1323 @opindex fconserve-space
1324 Put uninitialized or runtime-initialized global variables into the
1325 common segment, as C does. This saves space in the executable at the
1326 cost of not diagnosing duplicate definitions. If you compile with this
1327 flag and your program mysteriously crashes after @code{main()} has
1328 completed, you may have an object that is being destroyed twice because
1329 two definitions were merged.
1331 This option is no longer useful on most targets, now that support has
1332 been added for putting variables into BSS without making them common.
1334 @item -fno-const-strings
1335 @opindex fno-const-strings
1336 Give string constants type @code{char *} instead of type @code{const
1337 char *}. By default, G++ uses type @code{const char *} as required by
1338 the standard. Even if you use @option{-fno-const-strings}, you cannot
1339 actually modify the value of a string constant.
1341 This option might be removed in a future release of G++. For maximum
1342 portability, you should structure your code so that it works with
1343 string constants that have type @code{const char *}.
1345 @item -fno-elide-constructors
1346 @opindex fno-elide-constructors
1347 The C++ standard allows an implementation to omit creating a temporary
1348 which is only used to initialize another object of the same type.
1349 Specifying this option disables that optimization, and forces G++ to
1350 call the copy constructor in all cases.
1352 @item -fno-enforce-eh-specs
1353 @opindex fno-enforce-eh-specs
1354 Don't check for violation of exception specifications at runtime. This
1355 option violates the C++ standard, but may be useful for reducing code
1356 size in production builds, much like defining @samp{NDEBUG}. The compiler
1357 will still optimize based on the exception specifications.
1360 @itemx -fno-for-scope
1362 @opindex fno-for-scope
1363 If @option{-ffor-scope} is specified, the scope of variables declared in
1364 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1365 as specified by the C++ standard.
1366 If @option{-fno-for-scope} is specified, the scope of variables declared in
1367 a @i{for-init-statement} extends to the end of the enclosing scope,
1368 as was the case in old versions of G++, and other (traditional)
1369 implementations of C++.
1371 The default if neither flag is given to follow the standard,
1372 but to allow and give a warning for old-style code that would
1373 otherwise be invalid, or have different behavior.
1375 @item -fno-gnu-keywords
1376 @opindex fno-gnu-keywords
1377 Do not recognize @code{typeof} as a keyword, so that code can use this
1378 word as an identifier. You can use the keyword @code{__typeof__} instead.
1379 @option{-ansi} implies @option{-fno-gnu-keywords}.
1381 @item -fno-implicit-templates
1382 @opindex fno-implicit-templates
1383 Never emit code for non-inline templates which are instantiated
1384 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1385 @xref{Template Instantiation}, for more information.
1387 @item -fno-implicit-inline-templates
1388 @opindex fno-implicit-inline-templates
1389 Don't emit code for implicit instantiations of inline templates, either.
1390 The default is to handle inlines differently so that compiles with and
1391 without optimization will need the same set of explicit instantiations.
1393 @item -fno-implement-inlines
1394 @opindex fno-implement-inlines
1395 To save space, do not emit out-of-line copies of inline functions
1396 controlled by @samp{#pragma implementation}. This will cause linker
1397 errors if these functions are not inlined everywhere they are called.
1399 @item -fms-extensions
1400 @opindex fms-extensions
1401 Disable pedantic warnings about constructs used in MFC, such as implicit
1402 int and getting a pointer to member function via non-standard syntax.
1404 @item -fno-nonansi-builtins
1405 @opindex fno-nonansi-builtins
1406 Disable built-in declarations of functions that are not mandated by
1407 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1408 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1410 @item -fno-operator-names
1411 @opindex fno-operator-names
1412 Do not treat the operator name keywords @code{and}, @code{bitand},
1413 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1414 synonyms as keywords.
1416 @item -fno-optional-diags
1417 @opindex fno-optional-diags
1418 Disable diagnostics that the standard says a compiler does not need to
1419 issue. Currently, the only such diagnostic issued by G++ is the one for
1420 a name having multiple meanings within a class.
1423 @opindex fpermissive
1424 Downgrade some diagnostics about nonconformant code from errors to
1425 warnings. Thus, using @option{-fpermissive} will allow some
1426 nonconforming code to compile.
1430 Enable automatic template instantiation at link time. This option also
1431 implies @option{-fno-implicit-templates}. @xref{Template
1432 Instantiation}, for more information.
1436 Disable generation of information about every class with virtual
1437 functions for use by the C++ runtime type identification features
1438 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1439 of the language, you can save some space by using this flag. Note that
1440 exception handling uses the same information, but it will generate it as
1445 Emit statistics about front-end processing at the end of the compilation.
1446 This information is generally only useful to the G++ development team.
1448 @item -ftemplate-depth-@var{n}
1449 @opindex ftemplate-depth
1450 Set the maximum instantiation depth for template classes to @var{n}.
1451 A limit on the template instantiation depth is needed to detect
1452 endless recursions during template class instantiation. ANSI/ISO C++
1453 conforming programs must not rely on a maximum depth greater than 17.
1455 @item -fuse-cxa-atexit
1456 @opindex fuse-cxa-atexit
1457 Register destructors for objects with static storage duration with the
1458 @code{__cxa_atexit} function rather than the @code{atexit} function.
1459 This option is required for fully standards-compliant handling of static
1460 destructors, but will only work if your C library supports
1461 @code{__cxa_atexit}.
1463 @item -fvisibility-inlines-hidden
1464 @opindex fvisibility-inlines-hidden
1465 Causes all inlined methods to be marked with
1466 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1467 appear in the export table of a DSO and do not require a PLT indirection
1468 when used within the DSO. Enabling this option can have a dramatic effect
1469 on load and link times of a DSO as it massively reduces the size of the
1470 dynamic export table when the library makes heavy use of templates. While
1471 it can cause bloating through duplication of code within each DSO where
1472 it is used, often the wastage is less than the considerable space occupied
1473 by a long symbol name in the export table which is typical when using
1474 templates and namespaces. For even more savings, combine with the
1475 @code{-fvisibility=hidden} switch.
1479 Do not use weak symbol support, even if it is provided by the linker.
1480 By default, G++ will use weak symbols if they are available. This
1481 option exists only for testing, and should not be used by end-users;
1482 it will result in inferior code and has no benefits. This option may
1483 be removed in a future release of G++.
1487 Do not search for header files in the standard directories specific to
1488 C++, but do still search the other standard directories. (This option
1489 is used when building the C++ library.)
1492 In addition, these optimization, warning, and code generation options
1493 have meanings only for C++ programs:
1496 @item -fno-default-inline
1497 @opindex fno-default-inline
1498 Do not assume @samp{inline} for functions defined inside a class scope.
1499 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1500 functions will have linkage like inline functions; they just won't be
1503 @item -Wabi @r{(C++ only)}
1505 Warn when G++ generates code that is probably not compatible with the
1506 vendor-neutral C++ ABI. Although an effort has been made to warn about
1507 all such cases, there are probably some cases that are not warned about,
1508 even though G++ is generating incompatible code. There may also be
1509 cases where warnings are emitted even though the code that is generated
1512 You should rewrite your code to avoid these warnings if you are
1513 concerned about the fact that code generated by G++ may not be binary
1514 compatible with code generated by other compilers.
1516 The known incompatibilities at this point include:
1521 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1522 pack data into the same byte as a base class. For example:
1525 struct A @{ virtual void f(); int f1 : 1; @};
1526 struct B : public A @{ int f2 : 1; @};
1530 In this case, G++ will place @code{B::f2} into the same byte
1531 as@code{A::f1}; other compilers will not. You can avoid this problem
1532 by explicitly padding @code{A} so that its size is a multiple of the
1533 byte size on your platform; that will cause G++ and other compilers to
1534 layout @code{B} identically.
1537 Incorrect handling of tail-padding for virtual bases. G++ does not use
1538 tail padding when laying out virtual bases. For example:
1541 struct A @{ virtual void f(); char c1; @};
1542 struct B @{ B(); char c2; @};
1543 struct C : public A, public virtual B @{@};
1547 In this case, G++ will not place @code{B} into the tail-padding for
1548 @code{A}; other compilers will. You can avoid this problem by
1549 explicitly padding @code{A} so that its size is a multiple of its
1550 alignment (ignoring virtual base classes); that will cause G++ and other
1551 compilers to layout @code{C} identically.
1554 Incorrect handling of bit-fields with declared widths greater than that
1555 of their underlying types, when the bit-fields appear in a union. For
1559 union U @{ int i : 4096; @};
1563 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1564 union too small by the number of bits in an @code{int}.
1567 Empty classes can be placed at incorrect offsets. For example:
1577 struct C : public B, public A @{@};
1581 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1582 it should be placed at offset zero. G++ mistakenly believes that the
1583 @code{A} data member of @code{B} is already at offset zero.
1586 Names of template functions whose types involve @code{typename} or
1587 template template parameters can be mangled incorrectly.
1590 template <typename Q>
1591 void f(typename Q::X) @{@}
1593 template <template <typename> class Q>
1594 void f(typename Q<int>::X) @{@}
1598 Instantiations of these templates may be mangled incorrectly.
1602 @item -Wctor-dtor-privacy @r{(C++ only)}
1603 @opindex Wctor-dtor-privacy
1604 Warn when a class seems unusable because all the constructors or
1605 destructors in that class are private, and it has neither friends nor
1606 public static member functions.
1608 @item -Wnon-virtual-dtor @r{(C++ only)}
1609 @opindex Wnon-virtual-dtor
1610 Warn when a class appears to be polymorphic, thereby requiring a virtual
1611 destructor, yet it declares a non-virtual one.
1612 This warning is enabled by @option{-Wall}.
1614 @item -Wreorder @r{(C++ only)}
1616 @cindex reordering, warning
1617 @cindex warning for reordering of member initializers
1618 Warn when the order of member initializers given in the code does not
1619 match the order in which they must be executed. For instance:
1625 A(): j (0), i (1) @{ @}
1629 The compiler will rearrange the member initializers for @samp{i}
1630 and @samp{j} to match the declaration order of the members, emitting
1631 a warning to that effect. This warning is enabled by @option{-Wall}.
1634 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1637 @item -Weffc++ @r{(C++ only)}
1639 Warn about violations of the following style guidelines from Scott Meyers'
1640 @cite{Effective C++} book:
1644 Item 11: Define a copy constructor and an assignment operator for classes
1645 with dynamically allocated memory.
1648 Item 12: Prefer initialization to assignment in constructors.
1651 Item 14: Make destructors virtual in base classes.
1654 Item 15: Have @code{operator=} return a reference to @code{*this}.
1657 Item 23: Don't try to return a reference when you must return an object.
1661 Also warn about violations of the following style guidelines from
1662 Scott Meyers' @cite{More Effective C++} book:
1666 Item 6: Distinguish between prefix and postfix forms of increment and
1667 decrement operators.
1670 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1674 When selecting this option, be aware that the standard library
1675 headers do not obey all of these guidelines; use @samp{grep -v}
1676 to filter out those warnings.
1678 @item -Wno-deprecated @r{(C++ only)}
1679 @opindex Wno-deprecated
1680 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1682 @item -Wno-non-template-friend @r{(C++ only)}
1683 @opindex Wno-non-template-friend
1684 Disable warnings when non-templatized friend functions are declared
1685 within a template. Since the advent of explicit template specification
1686 support in G++, if the name of the friend is an unqualified-id (i.e.,
1687 @samp{friend foo(int)}), the C++ language specification demands that the
1688 friend declare or define an ordinary, nontemplate function. (Section
1689 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1690 could be interpreted as a particular specialization of a templatized
1691 function. Because this non-conforming behavior is no longer the default
1692 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1693 check existing code for potential trouble spots and is on by default.
1694 This new compiler behavior can be turned off with
1695 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1696 but disables the helpful warning.
1698 @item -Wold-style-cast @r{(C++ only)}
1699 @opindex Wold-style-cast
1700 Warn if an old-style (C-style) cast to a non-void type is used within
1701 a C++ program. The new-style casts (@samp{static_cast},
1702 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1703 unintended effects and much easier to search for.
1705 @item -Woverloaded-virtual @r{(C++ only)}
1706 @opindex Woverloaded-virtual
1707 @cindex overloaded virtual fn, warning
1708 @cindex warning for overloaded virtual fn
1709 Warn when a function declaration hides virtual functions from a
1710 base class. For example, in:
1717 struct B: public A @{
1722 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1730 will fail to compile.
1732 @item -Wno-pmf-conversions @r{(C++ only)}
1733 @opindex Wno-pmf-conversions
1734 Disable the diagnostic for converting a bound pointer to member function
1737 @item -Wsign-promo @r{(C++ only)}
1738 @opindex Wsign-promo
1739 Warn when overload resolution chooses a promotion from unsigned or
1740 enumerated type to a signed type, over a conversion to an unsigned type of
1741 the same size. Previous versions of G++ would try to preserve
1742 unsignedness, but the standard mandates the current behavior.
1744 @item -Wsynth @r{(C++ only)}
1746 @cindex warning for synthesized methods
1747 @cindex synthesized methods, warning
1748 Warn when G++'s synthesis behavior does not match that of cfront. For
1754 A& operator = (int);
1764 In this example, G++ will synthesize a default @samp{A& operator =
1765 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1768 @node Objective-C Dialect Options
1769 @section Options Controlling Objective-C Dialect
1771 @cindex compiler options, Objective-C
1772 @cindex Objective-C options, command line
1773 @cindex options, Objective-C
1774 (NOTE: This manual does not describe the Objective-C language itself. See
1775 @w{@uref{http://gcc.gnu.org/readings.html}} for references.)
1777 This section describes the command-line options that are only meaningful
1778 for Objective-C programs, but you can also use most of the GNU compiler
1779 options regardless of what language your program is in. For example,
1780 you might compile a file @code{some_class.m} like this:
1783 gcc -g -fgnu-runtime -O -c some_class.m
1787 In this example, @option{-fgnu-runtime} is an option meant only for
1788 Objective-C programs; you can use the other options with any language
1791 Here is a list of options that are @emph{only} for compiling Objective-C
1795 @item -fconstant-string-class=@var{class-name}
1796 @opindex fconstant-string-class
1797 Use @var{class-name} as the name of the class to instantiate for each
1798 literal string specified with the syntax @code{@@"@dots{}"}. The default
1799 class name is @code{NXConstantString} if the GNU runtime is being used, and
1800 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1801 @option{-fconstant-cfstrings} option, if also present, will override the
1802 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1803 to be laid out as constant CoreFoundation strings.
1806 @opindex fgnu-runtime
1807 Generate object code compatible with the standard GNU Objective-C
1808 runtime. This is the default for most types of systems.
1810 @item -fnext-runtime
1811 @opindex fnext-runtime
1812 Generate output compatible with the NeXT runtime. This is the default
1813 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1814 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1817 @item -fno-nil-receivers
1818 @opindex fno-nil-receivers
1819 Assume that all Objective-C message dispatches (e.g.,
1820 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1821 is not @code{nil}. This allows for more efficient entry points in the runtime to be
1822 used. Currently, this option is only available in conjunction with
1823 the NeXT runtime on Mac OS X 10.3 and later.
1825 @item -fobjc-exceptions
1826 @opindex fobjc-exceptions
1827 Enable syntactic support for structured exception handling in Objective-C,
1828 similar to what is offered by C++ and Java. Currently, this option is only
1829 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1837 @@catch (AnObjCClass *exc) @{
1844 @@catch (AnotherClass *exc) @{
1847 @@catch (id allOthers) @{
1857 The @code{@@throw} statement may appear anywhere in an Objective-C or
1858 Objective-C++ program; when used inside of a @code{@@catch} block, the
1859 @code{@@throw} may appear without an argument (as shown above), in which case
1860 the object caught by the @code{@@catch} will be rethrown.
1862 Note that only (pointers to) Objective-C objects may be thrown and
1863 caught using this scheme. When an object is thrown, it will be caught
1864 by the nearest @code{@@catch} clause capable of handling objects of that type,
1865 analogously to how @code{catch} blocks work in C++ and Java. A
1866 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1867 any and all Objective-C exceptions not caught by previous @code{@@catch}
1870 The @code{@@finally} clause, if present, will be executed upon exit from the
1871 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1872 regardless of whether any exceptions are thrown, caught or rethrown
1873 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1874 of the @code{finally} clause in Java.
1876 There are several caveats to using the new exception mechanism:
1880 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
1881 idioms provided by the @code{NSException} class, the new
1882 exceptions can only be used on Mac OS X 10.3 (Panther) and later
1883 systems, due to additional functionality needed in the (NeXT) Objective-C
1887 As mentioned above, the new exceptions do not support handling
1888 types other than Objective-C objects. Furthermore, when used from
1889 Objective-C++, the Objective-C exception model does not interoperate with C++
1890 exceptions at this time. This means you cannot @code{@@throw} an exception
1891 from Objective-C and @code{catch} it in C++, or vice versa
1892 (i.e., @code{throw @dots{} @@catch}).
1895 The @option{-fobjc-exceptions} switch also enables the use of synchronization
1896 blocks for thread-safe execution:
1899 @@synchronized (ObjCClass *guard) @{
1904 Upon entering the @code{@@synchronized} block, a thread of execution shall
1905 first check whether a lock has been placed on the corresponding @code{guard}
1906 object by another thread. If it has, the current thread shall wait until
1907 the other thread relinquishes its lock. Once @code{guard} becomes available,
1908 the current thread will place its own lock on it, execute the code contained in
1909 the @code{@@synchronized} block, and finally relinquish the lock (thereby
1910 making @code{guard} available to other threads).
1912 Unlike Java, Objective-C does not allow for entire methods to be marked
1913 @code{@@synchronized}. Note that throwing exceptions out of
1914 @code{@@synchronized} blocks is allowed, and will cause the guarding object
1915 to be unlocked properly.
1917 @item -freplace-objc-classes
1918 @opindex freplace-objc-classes
1919 Emit a special marker instructing @command{ld(1)} not to statically link in
1920 the resulting object file, and allow @command{dyld(1)} to load it in at
1921 run time instead. This is used in conjunction with the Fix-and-Continue
1922 debugging mode, where the object file in question may be recompiled and
1923 dynamically reloaded in the course of program execution, without the need
1924 to restart the program itself. Currently, Fix-and-Continue functionality
1925 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
1930 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
1931 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
1932 compile time) with static class references that get initialized at load time,
1933 which improves run-time performance. Specifying the @option{-fzero-link} flag
1934 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
1935 to be retained. This is useful in Zero-Link debugging mode, since it allows
1936 for individual class implementations to be modified during program execution.
1940 Dump interface declarations for all classes seen in the source file to a
1941 file named @file{@var{sourcename}.decl}.
1944 @opindex Wno-protocol
1945 If a class is declared to implement a protocol, a warning is issued for
1946 every method in the protocol that is not implemented by the class. The
1947 default behavior is to issue a warning for every method not explicitly
1948 implemented in the class, even if a method implementation is inherited
1949 from the superclass. If you use the @code{-Wno-protocol} option, then
1950 methods inherited from the superclass are considered to be implemented,
1951 and no warning is issued for them.
1955 Warn if multiple methods of different types for the same selector are
1956 found during compilation. The check is performed on the list of methods
1957 in the final stage of compilation. Additionally, a check is performed
1958 for each selector appearing in a @code{@@selector(@dots{})}
1959 expression, and a corresponding method for that selector has been found
1960 during compilation. Because these checks scan the method table only at
1961 the end of compilation, these warnings are not produced if the final
1962 stage of compilation is not reached, for example because an error is
1963 found during compilation, or because the @code{-fsyntax-only} option is
1966 @item -Wundeclared-selector
1967 @opindex Wundeclared-selector
1968 Warn if a @code{@@selector(@dots{})} expression referring to an
1969 undeclared selector is found. A selector is considered undeclared if no
1970 method with that name has been declared before the
1971 @code{@@selector(@dots{})} expression, either explicitly in an
1972 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
1973 an @code{@@implementation} section. This option always performs its
1974 checks as soon as a @code{@@selector(@dots{})} expression is found,
1975 while @code{-Wselector} only performs its checks in the final stage of
1976 compilation. This also enforces the coding style convention
1977 that methods and selectors must be declared before being used.
1979 @item -print-objc-runtime-info
1980 @opindex print-objc-runtime-info
1981 Generate C header describing the largest structure that is passed by
1986 @node Language Independent Options
1987 @section Options to Control Diagnostic Messages Formatting
1988 @cindex options to control diagnostics formatting
1989 @cindex diagnostic messages
1990 @cindex message formatting
1992 Traditionally, diagnostic messages have been formatted irrespective of
1993 the output device's aspect (e.g.@: its width, @dots{}). The options described
1994 below can be used to control the diagnostic messages formatting
1995 algorithm, e.g.@: how many characters per line, how often source location
1996 information should be reported. Right now, only the C++ front end can
1997 honor these options. However it is expected, in the near future, that
1998 the remaining front ends would be able to digest them correctly.
2001 @item -fmessage-length=@var{n}
2002 @opindex fmessage-length
2003 Try to format error messages so that they fit on lines of about @var{n}
2004 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2005 the front ends supported by GCC@. If @var{n} is zero, then no
2006 line-wrapping will be done; each error message will appear on a single
2009 @opindex fdiagnostics-show-location
2010 @item -fdiagnostics-show-location=once
2011 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2012 reporter to emit @emph{once} source location information; that is, in
2013 case the message is too long to fit on a single physical line and has to
2014 be wrapped, the source location won't be emitted (as prefix) again,
2015 over and over, in subsequent continuation lines. This is the default
2018 @item -fdiagnostics-show-location=every-line
2019 Only meaningful in line-wrapping mode. Instructs the diagnostic
2020 messages reporter to emit the same source location information (as
2021 prefix) for physical lines that result from the process of breaking
2022 a message which is too long to fit on a single line.
2026 @node Warning Options
2027 @section Options to Request or Suppress Warnings
2028 @cindex options to control warnings
2029 @cindex warning messages
2030 @cindex messages, warning
2031 @cindex suppressing warnings
2033 Warnings are diagnostic messages that report constructions which
2034 are not inherently erroneous but which are risky or suggest there
2035 may have been an error.
2037 You can request many specific warnings with options beginning @samp{-W},
2038 for example @option{-Wimplicit} to request warnings on implicit
2039 declarations. Each of these specific warning options also has a
2040 negative form beginning @samp{-Wno-} to turn off warnings;
2041 for example, @option{-Wno-implicit}. This manual lists only one of the
2042 two forms, whichever is not the default.
2044 The following options control the amount and kinds of warnings produced
2045 by GCC; for further, language-specific options also refer to
2046 @ref{C++ Dialect Options} and @ref{Objective-C Dialect Options}.
2049 @cindex syntax checking
2051 @opindex fsyntax-only
2052 Check the code for syntax errors, but don't do anything beyond that.
2056 Issue all the warnings demanded by strict ISO C and ISO C++;
2057 reject all programs that use forbidden extensions, and some other
2058 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2059 version of the ISO C standard specified by any @option{-std} option used.
2061 Valid ISO C and ISO C++ programs should compile properly with or without
2062 this option (though a rare few will require @option{-ansi} or a
2063 @option{-std} option specifying the required version of ISO C)@. However,
2064 without this option, certain GNU extensions and traditional C and C++
2065 features are supported as well. With this option, they are rejected.
2067 @option{-pedantic} does not cause warning messages for use of the
2068 alternate keywords whose names begin and end with @samp{__}. Pedantic
2069 warnings are also disabled in the expression that follows
2070 @code{__extension__}. However, only system header files should use
2071 these escape routes; application programs should avoid them.
2072 @xref{Alternate Keywords}.
2074 Some users try to use @option{-pedantic} to check programs for strict ISO
2075 C conformance. They soon find that it does not do quite what they want:
2076 it finds some non-ISO practices, but not all---only those for which
2077 ISO C @emph{requires} a diagnostic, and some others for which
2078 diagnostics have been added.
2080 A feature to report any failure to conform to ISO C might be useful in
2081 some instances, but would require considerable additional work and would
2082 be quite different from @option{-pedantic}. We don't have plans to
2083 support such a feature in the near future.
2085 Where the standard specified with @option{-std} represents a GNU
2086 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2087 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2088 extended dialect is based. Warnings from @option{-pedantic} are given
2089 where they are required by the base standard. (It would not make sense
2090 for such warnings to be given only for features not in the specified GNU
2091 C dialect, since by definition the GNU dialects of C include all
2092 features the compiler supports with the given option, and there would be
2093 nothing to warn about.)
2095 @item -pedantic-errors
2096 @opindex pedantic-errors
2097 Like @option{-pedantic}, except that errors are produced rather than
2102 Inhibit all warning messages.
2106 Inhibit warning messages about the use of @samp{#import}.
2108 @item -Wchar-subscripts
2109 @opindex Wchar-subscripts
2110 Warn if an array subscript has type @code{char}. This is a common cause
2111 of error, as programmers often forget that this type is signed on some
2116 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2117 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2119 @item -Wfatal-errors
2120 @opindex Wfatal-errors
2121 This option causes the compiler to abort compilation on the first error
2122 occurred rather than trying to keep going and printing further error
2127 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2128 the arguments supplied have types appropriate to the format string
2129 specified, and that the conversions specified in the format string make
2130 sense. This includes standard functions, and others specified by format
2131 attributes (@pxref{Function Attributes}), in the @code{printf},
2132 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2133 not in the C standard) families (or other target-specific families).
2135 The formats are checked against the format features supported by GNU
2136 libc version 2.2. These include all ISO C90 and C99 features, as well
2137 as features from the Single Unix Specification and some BSD and GNU
2138 extensions. Other library implementations may not support all these
2139 features; GCC does not support warning about features that go beyond a
2140 particular library's limitations. However, if @option{-pedantic} is used
2141 with @option{-Wformat}, warnings will be given about format features not
2142 in the selected standard version (but not for @code{strfmon} formats,
2143 since those are not in any version of the C standard). @xref{C Dialect
2144 Options,,Options Controlling C Dialect}.
2146 Since @option{-Wformat} also checks for null format arguments for
2147 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2149 @option{-Wformat} is included in @option{-Wall}. For more control over some
2150 aspects of format checking, the options @option{-Wformat-y2k},
2151 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2152 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2153 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2156 @opindex Wformat-y2k
2157 If @option{-Wformat} is specified, also warn about @code{strftime}
2158 formats which may yield only a two-digit year.
2160 @item -Wno-format-extra-args
2161 @opindex Wno-format-extra-args
2162 If @option{-Wformat} is specified, do not warn about excess arguments to a
2163 @code{printf} or @code{scanf} format function. The C standard specifies
2164 that such arguments are ignored.
2166 Where the unused arguments lie between used arguments that are
2167 specified with @samp{$} operand number specifications, normally
2168 warnings are still given, since the implementation could not know what
2169 type to pass to @code{va_arg} to skip the unused arguments. However,
2170 in the case of @code{scanf} formats, this option will suppress the
2171 warning if the unused arguments are all pointers, since the Single
2172 Unix Specification says that such unused arguments are allowed.
2174 @item -Wno-format-zero-length
2175 @opindex Wno-format-zero-length
2176 If @option{-Wformat} is specified, do not warn about zero-length formats.
2177 The C standard specifies that zero-length formats are allowed.
2179 @item -Wformat-nonliteral
2180 @opindex Wformat-nonliteral
2181 If @option{-Wformat} is specified, also warn if the format string is not a
2182 string literal and so cannot be checked, unless the format function
2183 takes its format arguments as a @code{va_list}.
2185 @item -Wformat-security
2186 @opindex Wformat-security
2187 If @option{-Wformat} is specified, also warn about uses of format
2188 functions that represent possible security problems. At present, this
2189 warns about calls to @code{printf} and @code{scanf} functions where the
2190 format string is not a string literal and there are no format arguments,
2191 as in @code{printf (foo);}. This may be a security hole if the format
2192 string came from untrusted input and contains @samp{%n}. (This is
2193 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2194 in future warnings may be added to @option{-Wformat-security} that are not
2195 included in @option{-Wformat-nonliteral}.)
2199 Enable @option{-Wformat} plus format checks not included in
2200 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2201 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2205 Warn about passing a null pointer for arguments marked as
2206 requiring a non-null value by the @code{nonnull} function attribute.
2208 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2209 can be disabled with the @option{-Wno-nonnull} option.
2211 @item -Winit-self @r{(C, C++, and Objective-C only)}
2213 Warn about uninitialized variables which are initialized with themselves.
2214 Note this option can only be used with the @option{-Wuninitialized} option,
2215 which in turn only works with @option{-O1} and above.
2217 For example, GCC will warn about @code{i} being uninitialized in the
2218 following snippet only when @option{-Winit-self} has been specified:
2229 @item -Wimplicit-int
2230 @opindex Wimplicit-int
2231 Warn when a declaration does not specify a type.
2233 @item -Wimplicit-function-declaration
2234 @itemx -Werror-implicit-function-declaration
2235 @opindex Wimplicit-function-declaration
2236 @opindex Werror-implicit-function-declaration
2237 Give a warning (or error) whenever a function is used before being
2242 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2246 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2247 function with external linkage, returning int, taking either zero
2248 arguments, two, or three arguments of appropriate types.
2250 @item -Wmissing-braces
2251 @opindex Wmissing-braces
2252 Warn if an aggregate or union initializer is not fully bracketed. In
2253 the following example, the initializer for @samp{a} is not fully
2254 bracketed, but that for @samp{b} is fully bracketed.
2257 int a[2][2] = @{ 0, 1, 2, 3 @};
2258 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2261 @item -Wmissing-include-dirs @r{(C, C++, and Objective-C only)}
2262 @opindex Wmissing-include-dirs
2263 Warn if a user-supplied include directory does not exist.
2266 @opindex Wparentheses
2267 Warn if parentheses are omitted in certain contexts, such
2268 as when there is an assignment in a context where a truth value
2269 is expected, or when operators are nested whose precedence people
2270 often get confused about. Only the warning for an assignment used as
2271 a truth value is supported when compiling C++; the other warnings are
2272 only supported when compiling C@.
2274 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2275 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2276 interpretation from that of ordinary mathematical notation.
2278 Also warn about constructions where there may be confusion to which
2279 @code{if} statement an @code{else} branch belongs. Here is an example of
2294 In C, every @code{else} branch belongs to the innermost possible @code{if}
2295 statement, which in this example is @code{if (b)}. This is often not
2296 what the programmer expected, as illustrated in the above example by
2297 indentation the programmer chose. When there is the potential for this
2298 confusion, GCC will issue a warning when this flag is specified.
2299 To eliminate the warning, add explicit braces around the innermost
2300 @code{if} statement so there is no way the @code{else} could belong to
2301 the enclosing @code{if}. The resulting code would look like this:
2317 @item -Wsequence-point
2318 @opindex Wsequence-point
2319 Warn about code that may have undefined semantics because of violations
2320 of sequence point rules in the C standard.
2322 The C standard defines the order in which expressions in a C program are
2323 evaluated in terms of @dfn{sequence points}, which represent a partial
2324 ordering between the execution of parts of the program: those executed
2325 before the sequence point, and those executed after it. These occur
2326 after the evaluation of a full expression (one which is not part of a
2327 larger expression), after the evaluation of the first operand of a
2328 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2329 function is called (but after the evaluation of its arguments and the
2330 expression denoting the called function), and in certain other places.
2331 Other than as expressed by the sequence point rules, the order of
2332 evaluation of subexpressions of an expression is not specified. All
2333 these rules describe only a partial order rather than a total order,
2334 since, for example, if two functions are called within one expression
2335 with no sequence point between them, the order in which the functions
2336 are called is not specified. However, the standards committee have
2337 ruled that function calls do not overlap.
2339 It is not specified when between sequence points modifications to the
2340 values of objects take effect. Programs whose behavior depends on this
2341 have undefined behavior; the C standard specifies that ``Between the
2342 previous and next sequence point an object shall have its stored value
2343 modified at most once by the evaluation of an expression. Furthermore,
2344 the prior value shall be read only to determine the value to be
2345 stored.''. If a program breaks these rules, the results on any
2346 particular implementation are entirely unpredictable.
2348 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2349 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2350 diagnosed by this option, and it may give an occasional false positive
2351 result, but in general it has been found fairly effective at detecting
2352 this sort of problem in programs.
2354 The present implementation of this option only works for C programs. A
2355 future implementation may also work for C++ programs.
2357 The C standard is worded confusingly, therefore there is some debate
2358 over the precise meaning of the sequence point rules in subtle cases.
2359 Links to discussions of the problem, including proposed formal
2360 definitions, may be found on the GCC readings page, at
2361 @w{@uref{http://gcc.gnu.org/readings.html}}.
2364 @opindex Wreturn-type
2365 Warn whenever a function is defined with a return-type that defaults to
2366 @code{int}. Also warn about any @code{return} statement with no
2367 return-value in a function whose return-type is not @code{void}.
2369 For C, also warn if the return type of a function has a type qualifier
2370 such as @code{const}. Such a type qualifier has no effect, since the
2371 value returned by a function is not an lvalue. ISO C prohibits
2372 qualified @code{void} return types on function definitions, so such
2373 return types always receive a warning even without this option.
2375 For C++, a function without return type always produces a diagnostic
2376 message, even when @option{-Wno-return-type} is specified. The only
2377 exceptions are @samp{main} and functions defined in system headers.
2381 Warn whenever a @code{switch} statement has an index of enumerated type
2382 and lacks a @code{case} for one or more of the named codes of that
2383 enumeration. (The presence of a @code{default} label prevents this
2384 warning.) @code{case} labels outside the enumeration range also
2385 provoke warnings when this option is used.
2387 @item -Wswitch-default
2388 @opindex Wswitch-switch
2389 Warn whenever a @code{switch} statement does not have a @code{default}
2393 @opindex Wswitch-enum
2394 Warn whenever a @code{switch} statement has an index of enumerated type
2395 and lacks a @code{case} for one or more of the named codes of that
2396 enumeration. @code{case} labels outside the enumeration range also
2397 provoke warnings when this option is used.
2401 Warn if any trigraphs are encountered that might change the meaning of
2402 the program (trigraphs within comments are not warned about).
2404 @item -Wunused-function
2405 @opindex Wunused-function
2406 Warn whenever a static function is declared but not defined or a
2407 non\-inline static function is unused.
2409 @item -Wunused-label
2410 @opindex Wunused-label
2411 Warn whenever a label is declared but not used.
2413 To suppress this warning use the @samp{unused} attribute
2414 (@pxref{Variable Attributes}).
2416 @item -Wunused-parameter
2417 @opindex Wunused-parameter
2418 Warn whenever a function parameter is unused aside from its declaration.
2420 To suppress this warning use the @samp{unused} attribute
2421 (@pxref{Variable Attributes}).
2423 @item -Wunused-variable
2424 @opindex Wunused-variable
2425 Warn whenever a local variable or non-constant static variable is unused
2426 aside from its declaration
2428 To suppress this warning use the @samp{unused} attribute
2429 (@pxref{Variable Attributes}).
2431 @item -Wunused-value
2432 @opindex Wunused-value
2433 Warn whenever a statement computes a result that is explicitly not used.
2435 To suppress this warning cast the expression to @samp{void}.
2439 All the above @option{-Wunused} options combined.
2441 In order to get a warning about an unused function parameter, you must
2442 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2443 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2445 @item -Wuninitialized
2446 @opindex Wuninitialized
2447 Warn if an automatic variable is used without first being initialized or
2448 if a variable may be clobbered by a @code{setjmp} call.
2450 These warnings are possible only in optimizing compilation,
2451 because they require data flow information that is computed only
2452 when optimizing. If you don't specify @option{-O}, you simply won't
2455 If you want to warn about code which uses the uninitialized value of the
2456 variable in its own initializer, use the @option{-Winit-self} option.
2458 These warnings occur only for variables that are candidates for
2459 register allocation. Therefore, they do not occur for a variable that
2460 is declared @code{volatile}, or whose address is taken, or whose size
2461 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
2462 structures, unions or arrays, even when they are in registers.
2464 Note that there may be no warning about a variable that is used only
2465 to compute a value that itself is never used, because such
2466 computations may be deleted by data flow analysis before the warnings
2469 These warnings are made optional because GCC is not smart
2470 enough to see all the reasons why the code might be correct
2471 despite appearing to have an error. Here is one example of how
2492 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2493 always initialized, but GCC doesn't know this. Here is
2494 another common case:
2499 if (change_y) save_y = y, y = new_y;
2501 if (change_y) y = save_y;
2506 This has no bug because @code{save_y} is used only if it is set.
2508 @cindex @code{longjmp} warnings
2509 This option also warns when a non-volatile automatic variable might be
2510 changed by a call to @code{longjmp}. These warnings as well are possible
2511 only in optimizing compilation.
2513 The compiler sees only the calls to @code{setjmp}. It cannot know
2514 where @code{longjmp} will be called; in fact, a signal handler could
2515 call it at any point in the code. As a result, you may get a warning
2516 even when there is in fact no problem because @code{longjmp} cannot
2517 in fact be called at the place which would cause a problem.
2519 Some spurious warnings can be avoided if you declare all the functions
2520 you use that never return as @code{noreturn}. @xref{Function
2523 @item -Wunknown-pragmas
2524 @opindex Wunknown-pragmas
2525 @cindex warning for unknown pragmas
2526 @cindex unknown pragmas, warning
2527 @cindex pragmas, warning of unknown
2528 Warn when a #pragma directive is encountered which is not understood by
2529 GCC@. If this command line option is used, warnings will even be issued
2530 for unknown pragmas in system header files. This is not the case if
2531 the warnings were only enabled by the @option{-Wall} command line option.
2533 @item -Wstrict-aliasing
2534 @opindex Wstrict-aliasing
2535 This option is only active when @option{-fstrict-aliasing} is active.
2536 It warns about code which might break the strict aliasing rules that the
2537 compiler is using for optimization. The warning does not catch all
2538 cases, but does attempt to catch the more common pitfalls. It is
2539 included in @option{-Wall}.
2541 @item -Wstrict-aliasing=2
2542 @opindex Wstrict-aliasing=2
2543 This option is only active when @option{-fstrict-aliasing} is active.
2544 It warns about all code which might break the strict aliasing rules that the
2545 compiler is using for optimization. This warning catches all cases, but
2546 it will also give a warning for some ambiguous cases that are safe.
2550 All of the above @samp{-W} options combined. This enables all the
2551 warnings about constructions that some users consider questionable, and
2552 that are easy to avoid (or modify to prevent the warning), even in
2553 conjunction with macros. This also enables some language-specific
2554 warnings described in @ref{C++ Dialect Options} and
2555 @ref{Objective-C Dialect Options}.
2558 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2559 Some of them warn about constructions that users generally do not
2560 consider questionable, but which occasionally you might wish to check
2561 for; others warn about constructions that are necessary or hard to avoid
2562 in some cases, and there is no simple way to modify the code to suppress
2569 (This option used to be called @option{-W}. The older name is still
2570 supported, but the newer name is more descriptive.) Print extra warning
2571 messages for these events:
2575 A function can return either with or without a value. (Falling
2576 off the end of the function body is considered returning without
2577 a value.) For example, this function would evoke such a
2591 An expression-statement or the left-hand side of a comma expression
2592 contains no side effects.
2593 To suppress the warning, cast the unused expression to void.
2594 For example, an expression such as @samp{x[i,j]} will cause a warning,
2595 but @samp{x[(void)i,j]} will not.
2598 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2601 Storage-class specifiers like @code{static} are not the first things in
2602 a declaration. According to the C Standard, this usage is obsolescent.
2605 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2609 A comparison between signed and unsigned values could produce an
2610 incorrect result when the signed value is converted to unsigned.
2611 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2614 An aggregate has an initializer which does not initialize all members.
2615 For example, the following code would cause such a warning, because
2616 @code{x.h} would be implicitly initialized to zero:
2619 struct s @{ int f, g, h; @};
2620 struct s x = @{ 3, 4 @};
2624 A function parameter is declared without a type specifier in K&R-style
2632 An empty body occurs in an @samp{if} or @samp{else} statement.
2635 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2636 @samp{>}, or @samp{>=}.
2639 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2642 Any of several floating-point events that often indicate errors, such as
2643 overflow, underflow, loss of precision, etc.
2645 @item @r{(C++ only)}
2646 An enumerator and a non-enumerator both appear in a conditional expression.
2648 @item @r{(C++ only)}
2649 A non-static reference or non-static @samp{const} member appears in a
2650 class without constructors.
2652 @item @r{(C++ only)}
2653 Ambiguous virtual bases.
2655 @item @r{(C++ only)}
2656 Subscripting an array which has been declared @samp{register}.
2658 @item @r{(C++ only)}
2659 Taking the address of a variable which has been declared @samp{register}.
2661 @item @r{(C++ only)}
2662 A base class is not initialized in a derived class' copy constructor.
2665 @item -Wno-div-by-zero
2666 @opindex Wno-div-by-zero
2667 @opindex Wdiv-by-zero
2668 Do not warn about compile-time integer division by zero. Floating point
2669 division by zero is not warned about, as it can be a legitimate way of
2670 obtaining infinities and NaNs.
2672 @item -Wsystem-headers
2673 @opindex Wsystem-headers
2674 @cindex warnings from system headers
2675 @cindex system headers, warnings from
2676 Print warning messages for constructs found in system header files.
2677 Warnings from system headers are normally suppressed, on the assumption
2678 that they usually do not indicate real problems and would only make the
2679 compiler output harder to read. Using this command line option tells
2680 GCC to emit warnings from system headers as if they occurred in user
2681 code. However, note that using @option{-Wall} in conjunction with this
2682 option will @emph{not} warn about unknown pragmas in system
2683 headers---for that, @option{-Wunknown-pragmas} must also be used.
2686 @opindex Wfloat-equal
2687 Warn if floating point values are used in equality comparisons.
2689 The idea behind this is that sometimes it is convenient (for the
2690 programmer) to consider floating-point values as approximations to
2691 infinitely precise real numbers. If you are doing this, then you need
2692 to compute (by analyzing the code, or in some other way) the maximum or
2693 likely maximum error that the computation introduces, and allow for it
2694 when performing comparisons (and when producing output, but that's a
2695 different problem). In particular, instead of testing for equality, you
2696 would check to see whether the two values have ranges that overlap; and
2697 this is done with the relational operators, so equality comparisons are
2700 @item -Wtraditional @r{(C only)}
2701 @opindex Wtraditional
2702 Warn about certain constructs that behave differently in traditional and
2703 ISO C@. Also warn about ISO C constructs that have no traditional C
2704 equivalent, and/or problematic constructs which should be avoided.
2708 Macro parameters that appear within string literals in the macro body.
2709 In traditional C macro replacement takes place within string literals,
2710 but does not in ISO C@.
2713 In traditional C, some preprocessor directives did not exist.
2714 Traditional preprocessors would only consider a line to be a directive
2715 if the @samp{#} appeared in column 1 on the line. Therefore
2716 @option{-Wtraditional} warns about directives that traditional C
2717 understands but would ignore because the @samp{#} does not appear as the
2718 first character on the line. It also suggests you hide directives like
2719 @samp{#pragma} not understood by traditional C by indenting them. Some
2720 traditional implementations would not recognize @samp{#elif}, so it
2721 suggests avoiding it altogether.
2724 A function-like macro that appears without arguments.
2727 The unary plus operator.
2730 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2731 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2732 constants.) Note, these suffixes appear in macros defined in the system
2733 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2734 Use of these macros in user code might normally lead to spurious
2735 warnings, however GCC's integrated preprocessor has enough context to
2736 avoid warning in these cases.
2739 A function declared external in one block and then used after the end of
2743 A @code{switch} statement has an operand of type @code{long}.
2746 A non-@code{static} function declaration follows a @code{static} one.
2747 This construct is not accepted by some traditional C compilers.
2750 The ISO type of an integer constant has a different width or
2751 signedness from its traditional type. This warning is only issued if
2752 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2753 typically represent bit patterns, are not warned about.
2756 Usage of ISO string concatenation is detected.
2759 Initialization of automatic aggregates.
2762 Identifier conflicts with labels. Traditional C lacks a separate
2763 namespace for labels.
2766 Initialization of unions. If the initializer is zero, the warning is
2767 omitted. This is done under the assumption that the zero initializer in
2768 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2769 initializer warnings and relies on default initialization to zero in the
2773 Conversions by prototypes between fixed/floating point values and vice
2774 versa. The absence of these prototypes when compiling with traditional
2775 C would cause serious problems. This is a subset of the possible
2776 conversion warnings, for the full set use @option{-Wconversion}.
2779 Use of ISO C style function definitions. This warning intentionally is
2780 @emph{not} issued for prototype declarations or variadic functions
2781 because these ISO C features will appear in your code when using
2782 libiberty's traditional C compatibility macros, @code{PARAMS} and
2783 @code{VPARAMS}. This warning is also bypassed for nested functions
2784 because that feature is already a GCC extension and thus not relevant to
2785 traditional C compatibility.
2788 @item -Wdeclaration-after-statement @r{(C only)}
2789 @opindex Wdeclaration-after-statement
2790 Warn when a declaration is found after a statement in a block. This
2791 construct, known from C++, was introduced with ISO C99 and is by default
2792 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2793 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2797 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2799 @item -Wendif-labels
2800 @opindex Wendif-labels
2801 Warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2805 Warn whenever a local variable shadows another local variable, parameter or
2806 global variable or whenever a built-in function is shadowed.
2808 @item -Wlarger-than-@var{len}
2809 @opindex Wlarger-than
2810 Warn whenever an object of larger than @var{len} bytes is defined.
2812 @item -Wpointer-arith
2813 @opindex Wpointer-arith
2814 Warn about anything that depends on the ``size of'' a function type or
2815 of @code{void}. GNU C assigns these types a size of 1, for
2816 convenience in calculations with @code{void *} pointers and pointers
2819 @item -Wbad-function-cast @r{(C only)}
2820 @opindex Wbad-function-cast
2821 Warn whenever a function call is cast to a non-matching type.
2822 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2826 Warn whenever a pointer is cast so as to remove a type qualifier from
2827 the target type. For example, warn if a @code{const char *} is cast
2828 to an ordinary @code{char *}.
2831 @opindex Wcast-align
2832 Warn whenever a pointer is cast such that the required alignment of the
2833 target is increased. For example, warn if a @code{char *} is cast to
2834 an @code{int *} on machines where integers can only be accessed at
2835 two- or four-byte boundaries.
2837 @item -Wwrite-strings
2838 @opindex Wwrite-strings
2839 When compiling C, give string constants the type @code{const
2840 char[@var{length}]} so that
2841 copying the address of one into a non-@code{const} @code{char *}
2842 pointer will get a warning; when compiling C++, warn about the
2843 deprecated conversion from string constants to @code{char *}.
2844 These warnings will help you find at
2845 compile time code that can try to write into a string constant, but
2846 only if you have been very careful about using @code{const} in
2847 declarations and prototypes. Otherwise, it will just be a nuisance;
2848 this is why we did not make @option{-Wall} request these warnings.
2851 @opindex Wconversion
2852 Warn if a prototype causes a type conversion that is different from what
2853 would happen to the same argument in the absence of a prototype. This
2854 includes conversions of fixed point to floating and vice versa, and
2855 conversions changing the width or signedness of a fixed point argument
2856 except when the same as the default promotion.
2858 Also, warn if a negative integer constant expression is implicitly
2859 converted to an unsigned type. For example, warn about the assignment
2860 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2861 casts like @code{(unsigned) -1}.
2863 @item -Wsign-compare
2864 @opindex Wsign-compare
2865 @cindex warning for comparison of signed and unsigned values
2866 @cindex comparison of signed and unsigned values, warning
2867 @cindex signed and unsigned values, comparison warning
2868 Warn when a comparison between signed and unsigned values could produce
2869 an incorrect result when the signed value is converted to unsigned.
2870 This warning is also enabled by @option{-Wextra}; to get the other warnings
2871 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2873 @item -Waggregate-return
2874 @opindex Waggregate-return
2875 Warn if any functions that return structures or unions are defined or
2876 called. (In languages where you can return an array, this also elicits
2879 @item -Wstrict-prototypes @r{(C only)}
2880 @opindex Wstrict-prototypes
2881 Warn if a function is declared or defined without specifying the
2882 argument types. (An old-style function definition is permitted without
2883 a warning if preceded by a declaration which specifies the argument
2886 @item -Wold-style-definition @r{(C only)}
2887 @opindex Wold-style-definition
2888 Warn if an old-style function definition is used. A warning is given
2889 even if there is a previous prototype.
2891 @item -Wmissing-prototypes @r{(C only)}
2892 @opindex Wmissing-prototypes
2893 Warn if a global function is defined without a previous prototype
2894 declaration. This warning is issued even if the definition itself
2895 provides a prototype. The aim is to detect global functions that fail
2896 to be declared in header files.
2898 @item -Wmissing-declarations @r{(C only)}
2899 @opindex Wmissing-declarations
2900 Warn if a global function is defined without a previous declaration.
2901 Do so even if the definition itself provides a prototype.
2902 Use this option to detect global functions that are not declared in
2905 @item -Wmissing-noreturn
2906 @opindex Wmissing-noreturn
2907 Warn about functions which might be candidates for attribute @code{noreturn}.
2908 Note these are only possible candidates, not absolute ones. Care should
2909 be taken to manually verify functions actually do not ever return before
2910 adding the @code{noreturn} attribute, otherwise subtle code generation
2911 bugs could be introduced. You will not get a warning for @code{main} in
2912 hosted C environments.
2914 @item -Wmissing-format-attribute
2915 @opindex Wmissing-format-attribute
2917 If @option{-Wformat} is enabled, also warn about functions which might be
2918 candidates for @code{format} attributes. Note these are only possible
2919 candidates, not absolute ones. GCC will guess that @code{format}
2920 attributes might be appropriate for any function that calls a function
2921 like @code{vprintf} or @code{vscanf}, but this might not always be the
2922 case, and some functions for which @code{format} attributes are
2923 appropriate may not be detected. This option has no effect unless
2924 @option{-Wformat} is enabled (possibly by @option{-Wall}).
2926 @item -Wno-multichar
2927 @opindex Wno-multichar
2929 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
2930 Usually they indicate a typo in the user's code, as they have
2931 implementation-defined values, and should not be used in portable code.
2933 @item -Wno-deprecated-declarations
2934 @opindex Wno-deprecated-declarations
2935 Do not warn about uses of functions, variables, and types marked as
2936 deprecated by using the @code{deprecated} attribute.
2937 (@pxref{Function Attributes}, @pxref{Variable Attributes},
2938 @pxref{Type Attributes}.)
2942 Warn if a structure is given the packed attribute, but the packed
2943 attribute has no effect on the layout or size of the structure.
2944 Such structures may be mis-aligned for little benefit. For
2945 instance, in this code, the variable @code{f.x} in @code{struct bar}
2946 will be misaligned even though @code{struct bar} does not itself
2947 have the packed attribute:
2954 @} __attribute__((packed));
2964 Warn if padding is included in a structure, either to align an element
2965 of the structure or to align the whole structure. Sometimes when this
2966 happens it is possible to rearrange the fields of the structure to
2967 reduce the padding and so make the structure smaller.
2969 @item -Wredundant-decls
2970 @opindex Wredundant-decls
2971 Warn if anything is declared more than once in the same scope, even in
2972 cases where multiple declaration is valid and changes nothing.
2974 @item -Wnested-externs @r{(C only)}
2975 @opindex Wnested-externs
2976 Warn if an @code{extern} declaration is encountered within a function.
2978 @item -Wunreachable-code
2979 @opindex Wunreachable-code
2980 Warn if the compiler detects that code will never be executed.
2982 This option is intended to warn when the compiler detects that at
2983 least a whole line of source code will never be executed, because
2984 some condition is never satisfied or because it is after a
2985 procedure that never returns.
2987 It is possible for this option to produce a warning even though there
2988 are circumstances under which part of the affected line can be executed,
2989 so care should be taken when removing apparently-unreachable code.
2991 For instance, when a function is inlined, a warning may mean that the
2992 line is unreachable in only one inlined copy of the function.
2994 This option is not made part of @option{-Wall} because in a debugging
2995 version of a program there is often substantial code which checks
2996 correct functioning of the program and is, hopefully, unreachable
2997 because the program does work. Another common use of unreachable
2998 code is to provide behavior which is selectable at compile-time.
3002 Warn if a function can not be inlined and it was declared as inline.
3003 Even with this option, the compiler will not warn about failures to
3004 inline functions declared in system headers.
3006 The compiler uses a variety of heuristics to determine whether or not
3007 to inline a function. For example, the compiler takes into account
3008 the size of the function being inlined and the the amount of inlining
3009 that has already been done in the current function. Therefore,
3010 seemingly insignificant changes in the source program can cause the
3011 warnings produced by @option{-Winline} to appear or disappear.
3013 @item -Wno-invalid-offsetof @r{(C++ only)}
3014 @opindex Wno-invalid-offsetof
3015 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3016 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3017 to a non-POD type is undefined. In existing C++ implementations,
3018 however, @samp{offsetof} typically gives meaningful results even when
3019 applied to certain kinds of non-POD types. (Such as a simple
3020 @samp{struct} that fails to be a POD type only by virtue of having a
3021 constructor.) This flag is for users who are aware that they are
3022 writing nonportable code and who have deliberately chosen to ignore the
3025 The restrictions on @samp{offsetof} may be relaxed in a future version
3026 of the C++ standard.
3029 @opindex Winvalid-pch
3030 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3031 the search path but can't be used.
3035 @opindex Wno-long-long
3036 Warn if @samp{long long} type is used. This is default. To inhibit
3037 the warning messages, use @option{-Wno-long-long}. Flags
3038 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3039 only when @option{-pedantic} flag is used.
3041 @item -Wvariadic-macros
3042 @opindex Wvariadic-macros
3043 @opindex Wno-variadic-macros
3044 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3045 alternate syntax when in pedantic ISO C99 mode. This is default.
3046 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3048 @item -Wdisabled-optimization
3049 @opindex Wdisabled-optimization
3050 Warn if a requested optimization pass is disabled. This warning does
3051 not generally indicate that there is anything wrong with your code; it
3052 merely indicates that GCC's optimizers were unable to handle the code
3053 effectively. Often, the problem is that your code is too big or too
3054 complex; GCC will refuse to optimize programs when the optimization
3055 itself is likely to take inordinate amounts of time.
3059 Make all warnings into errors.
3062 @node Debugging Options
3063 @section Options for Debugging Your Program or GCC
3064 @cindex options, debugging
3065 @cindex debugging information options
3067 GCC has various special options that are used for debugging
3068 either your program or GCC:
3073 Produce debugging information in the operating system's native format
3074 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
3077 On most systems that use stabs format, @option{-g} enables use of extra
3078 debugging information that only GDB can use; this extra information
3079 makes debugging work better in GDB but will probably make other debuggers
3081 refuse to read the program. If you want to control for certain whether
3082 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3083 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3085 Unlike most other C compilers, GCC allows you to use @option{-g} with
3086 @option{-O}. The shortcuts taken by optimized code may occasionally
3087 produce surprising results: some variables you declared may not exist
3088 at all; flow of control may briefly move where you did not expect it;
3089 some statements may not be executed because they compute constant
3090 results or their values were already at hand; some statements may
3091 execute in different places because they were moved out of loops.
3093 Nevertheless it proves possible to debug optimized output. This makes
3094 it reasonable to use the optimizer for programs that might have bugs.
3096 The following options are useful when GCC is generated with the
3097 capability for more than one debugging format.
3101 Produce debugging information for use by GDB@. This means to use the
3102 most expressive format available (DWARF 2, stabs, or the native format
3103 if neither of those are supported), including GDB extensions if at all
3108 Produce debugging information in stabs format (if that is supported),
3109 without GDB extensions. This is the format used by DBX on most BSD
3110 systems. On MIPS, Alpha and System V Release 4 systems this option
3111 produces stabs debugging output which is not understood by DBX or SDB@.
3112 On System V Release 4 systems this option requires the GNU assembler.
3114 @item -feliminate-unused-debug-symbols
3115 @opindex feliminate-unused-debug-symbols
3116 Produce debugging information in stabs format (if that is supported),
3117 for only symbols that are actually used.
3121 Produce debugging information in stabs format (if that is supported),
3122 using GNU extensions understood only by the GNU debugger (GDB)@. The
3123 use of these extensions is likely to make other debuggers crash or
3124 refuse to read the program.
3128 Produce debugging information in COFF format (if that is supported).
3129 This is the format used by SDB on most System V systems prior to
3134 Produce debugging information in XCOFF format (if that is supported).
3135 This is the format used by the DBX debugger on IBM RS/6000 systems.
3139 Produce debugging information in XCOFF format (if that is supported),
3140 using GNU extensions understood only by the GNU debugger (GDB)@. The
3141 use of these extensions is likely to make other debuggers crash or
3142 refuse to read the program, and may cause assemblers other than the GNU
3143 assembler (GAS) to fail with an error.
3147 Produce debugging information in DWARF version 2 format (if that is
3148 supported). This is the format used by DBX on IRIX 6.
3152 Produce debugging information in VMS debug format (if that is
3153 supported). This is the format used by DEBUG on VMS systems.
3156 @itemx -ggdb@var{level}
3157 @itemx -gstabs@var{level}
3158 @itemx -gcoff@var{level}
3159 @itemx -gxcoff@var{level}
3160 @itemx -gvms@var{level}
3161 Request debugging information and also use @var{level} to specify how
3162 much information. The default level is 2.
3164 Level 1 produces minimal information, enough for making backtraces in
3165 parts of the program that you don't plan to debug. This includes
3166 descriptions of functions and external variables, but no information
3167 about local variables and no line numbers.
3169 Level 3 includes extra information, such as all the macro definitions
3170 present in the program. Some debuggers support macro expansion when
3171 you use @option{-g3}.
3173 Note that in order to avoid confusion between DWARF1 debug level 2,
3174 and DWARF2 @option{-gdwarf-2} does not accept a concatenated debug
3175 level. Instead use an additional @option{-g@var{level}} option to
3176 change the debug level for DWARF2.
3178 @item -feliminate-dwarf2-dups
3179 @opindex feliminate-dwarf2-dups
3180 Compress DWARF2 debugging information by eliminating duplicated
3181 information about each symbol. This option only makes sense when
3182 generating DWARF2 debugging information with @option{-gdwarf-2}.
3184 @cindex @command{prof}
3187 Generate extra code to write profile information suitable for the
3188 analysis program @command{prof}. You must use this option when compiling
3189 the source files you want data about, and you must also use it when
3192 @cindex @command{gprof}
3195 Generate extra code to write profile information suitable for the
3196 analysis program @command{gprof}. You must use this option when compiling
3197 the source files you want data about, and you must also use it when
3202 Makes the compiler print out each function name as it is compiled, and
3203 print some statistics about each pass when it finishes.
3206 @opindex ftime-report
3207 Makes the compiler print some statistics about the time consumed by each
3208 pass when it finishes.
3211 @opindex fmem-report
3212 Makes the compiler print some statistics about permanent memory
3213 allocation when it finishes.
3215 @item -fprofile-arcs
3216 @opindex fprofile-arcs
3217 Add code so that program flow @dfn{arcs} are instrumented. During
3218 execution the program records how many times each branch and call is
3219 executed and how many times it is taken or returns. When the compiled
3220 program exits it saves this data to a file called
3221 @file{@var{auxname}.gcda} for each source file. The data may be used for
3222 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3223 test coverage analysis (@option{-ftest-coverage}). Each object file's
3224 @var{auxname} is generated from the name of the output file, if
3225 explicitly specified and it is not the final executable, otherwise it is
3226 the basename of the source file. In both cases any suffix is removed
3227 (e.g. @file{foo.gcda} for input file @file{dir/foo.c}, or
3228 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3233 Compile the source files with @option{-fprofile-arcs} plus optimization
3234 and code generation options. For test coverage analysis, use the
3235 additional @option{-ftest-coverage} option. You do not need to profile
3236 every source file in a program.
3239 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3240 (the latter implies the former).
3243 Run the program on a representative workload to generate the arc profile
3244 information. This may be repeated any number of times. You can run
3245 concurrent instances of your program, and provided that the file system
3246 supports locking, the data files will be correctly updated. Also
3247 @code{fork} calls are detected and correctly handled (double counting
3251 For profile-directed optimizations, compile the source files again with
3252 the same optimization and code generation options plus
3253 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3254 Control Optimization}).
3257 For test coverage analysis, use @command{gcov} to produce human readable
3258 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3259 @command{gcov} documentation for further information.
3263 With @option{-fprofile-arcs}, for each function of your program GCC
3264 creates a program flow graph, then finds a spanning tree for the graph.
3265 Only arcs that are not on the spanning tree have to be instrumented: the
3266 compiler adds code to count the number of times that these arcs are
3267 executed. When an arc is the only exit or only entrance to a block, the
3268 instrumentation code can be added to the block; otherwise, a new basic
3269 block must be created to hold the instrumentation code.
3271 @item -ftree-based-profiling
3272 @opindex ftree-based-profiling
3273 This option is used in addition to @option{-fprofile-arcs} or
3274 @option{-fbranch-probabilities} to control whether those optimizations
3275 are performed on a tree-based or rtl-based internal representation.
3276 If you use this option when compiling with @option{-fprofile-arcs},
3277 you must also use it when compiling later with @option{-fbranch-probabilities}.
3278 Currently the tree-based optimization is in an early stage of
3279 development, and this option is recommended only for those people
3280 working on improving it.
3283 @item -ftest-coverage
3284 @opindex ftest-coverage
3285 Produce a notes file that the @command{gcov} code-coverage utility
3286 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3287 show program coverage. Each source file's note file is called
3288 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3289 above for a description of @var{auxname} and instructions on how to
3290 generate test coverage data. Coverage data will match the source files
3291 more closely, if you do not optimize.
3293 @item -d@var{letters}
3295 Says to make debugging dumps during compilation at times specified by
3296 @var{letters}. This is used for debugging the compiler. The file names
3297 for most of the dumps are made by appending a pass number and a word to
3298 the @var{dumpname}. @var{dumpname} is generated from the name of the
3299 output file, if explicitly specified and it is not an executable,
3300 otherwise it is the basename of the source file. In both cases any
3301 suffix is removed (e.g. @file{foo.01.rtl} or @file{foo.02.sibling}).
3302 Here are the possible letters for use in @var{letters}, and their
3308 Annotate the assembler output with miscellaneous debugging information.
3311 Dump after computing branch probabilities, to @file{@var{file}.12.bp}.
3314 Dump after block reordering, to @file{@var{file}.32.bbro}.
3317 Dump after instruction combination, to the file @file{@var{file}.20.combine}.
3320 Dump after the first if conversion, to the file @file{@var{file}.14.ce1}.
3321 Also dump after the second if conversion, to the file @file{@var{file}.21.ce2}.
3324 Dump after branch target load optimization, to to @file{@var{file}.33.btl}.
3325 Also dump after delayed branch scheduling, to @file{@var{file}.37.dbr}.
3328 Dump all macro definitions, at the end of preprocessing, in addition to
3332 Dump after the third if conversion, to @file{@var{file}.31.ce3}.
3335 Dump after control and data flow analysis, to @file{@var{file}.11.cfg}.
3336 Also dump after life analysis, to @file{@var{file}.19.life}.
3339 Dump after global register allocation, to @file{@var{file}.26.greg}.
3342 Dump after GCSE, to @file{@var{file}.08.gcse}.
3343 Also dump after jump bypassing and control flow optimizations, to
3344 @file{@var{file}.10.bypass}.
3347 Dump after finalization of EH handling code, to @file{@var{file}.03.eh}.
3350 Dump after sibling call optimizations, to @file{@var{file}.02.sibling}.
3353 Dump after the first jump optimization, to @file{@var{file}.04.jump}.
3356 Dump after conversion from registers to stack, to @file{@var{file}.35.stack}.
3359 Dump after local register allocation, to @file{@var{file}.25.lreg}.
3362 Dump after loop optimization passes, to @file{@var{file}.09.loop} and
3363 @file{@var{file}.16.loop2}.
3366 Dump after modulo scheduling, to @file{@var{file}.23.sms}.
3369 Dump after performing the machine dependent reorganization pass, to
3370 @file{@var{file}.36.mach}.
3373 Dump after register renumbering, to @file{@var{file}.30.rnreg}.
3376 Dump after the register move pass, to @file{@var{file}.22.regmove}.
3379 Dump after post-reload optimizations, to @file{@var{file}.27.postreload}.
3382 Dump after RTL generation, to @file{@var{file}.01.rtl}.
3385 Dump after the second scheduling pass, to @file{@var{file}.34.sched2}.
3388 Dump after CSE (including the jump optimization that sometimes follows
3389 CSE), to @file{@var{file}.06.cse}.
3392 Dump after the first scheduling pass, to @file{@var{file}.24.sched}.
3395 Dump after the second CSE pass (including the jump optimization that
3396 sometimes follows CSE), to @file{@var{file}.18.cse2}.
3399 Dump after running tracer, to @file{@var{file}.15.tracer}.
3402 Dump after null pointer elimination pass to @file{@var{file}.05.null}.
3405 Dump callgraph and unit-at-a-time optimization @file{@var{file}.00.unit}.
3408 Dump after the value profile transformations, to @file{@var{file}.13.vpt}.
3409 Also dump after variable tracking, to @file{@var{file}.35.vartrack}.
3412 Dump after the second flow pass, to @file{@var{file}.28.flow2}.
3415 Dump after the peephole pass, to @file{@var{file}.29.peephole2}.
3418 Dump after constructing the web, to @file{@var{file}.17.web}.
3421 Produce all the dumps listed above.
3424 Produce a core dump whenever an error occurs.
3427 Print statistics on memory usage, at the end of the run, to
3431 Annotate the assembler output with a comment indicating which
3432 pattern and alternative was used. The length of each instruction is
3436 Dump the RTL in the assembler output as a comment before each instruction.
3437 Also turns on @option{-dp} annotation.
3440 For each of the other indicated dump files (except for
3441 @file{@var{file}.01.rtl}), dump a representation of the control flow graph
3442 suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3445 Just generate RTL for a function instead of compiling it. Usually used
3449 Dump debugging information during parsing, to standard error.
3452 @item -fdump-unnumbered
3453 @opindex fdump-unnumbered
3454 When doing debugging dumps (see @option{-d} option above), suppress instruction
3455 numbers and line number note output. This makes it more feasible to
3456 use diff on debugging dumps for compiler invocations with different
3457 options, in particular with and without @option{-g}.
3459 @item -fdump-translation-unit @r{(C and C++ only)}
3460 @itemx -fdump-translation-unit-@var{options} @r{(C and C++ only)}
3461 @opindex fdump-translation-unit
3462 Dump a representation of the tree structure for the entire translation
3463 unit to a file. The file name is made by appending @file{.tu} to the
3464 source file name. If the @samp{-@var{options}} form is used, @var{options}
3465 controls the details of the dump as described for the
3466 @option{-fdump-tree} options.
3468 @item -fdump-class-hierarchy @r{(C++ only)}
3469 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3470 @opindex fdump-class-hierarchy
3471 Dump a representation of each class's hierarchy and virtual function
3472 table layout to a file. The file name is made by appending @file{.class}
3473 to the source file name. If the @samp{-@var{options}} form is used,
3474 @var{options} controls the details of the dump as described for the
3475 @option{-fdump-tree} options.
3477 @item -fdump-tree-@var{switch} @r{(C and C++ only)}
3478 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C and C++ only)}
3480 Control the dumping at various stages of processing the intermediate
3481 language tree to a file. The file name is generated by appending a switch
3482 specific suffix to the source file name. If the @samp{-@var{options}}
3483 form is used, @var{options} is a list of @samp{-} separated options that
3484 control the details of the dump. Not all options are applicable to all
3485 dumps, those which are not meaningful will be ignored. The following
3486 options are available
3490 Print the address of each node. Usually this is not meaningful as it
3491 changes according to the environment and source file. Its primary use
3492 is for tying up a dump file with a debug environment.
3494 Inhibit dumping of members of a scope or body of a function merely
3495 because that scope has been reached. Only dump such items when they
3496 are directly reachable by some other path. When dumping pretty-printed
3497 trees, this option inhibits dumping the bodies of control structures.
3499 Print a raw representation of the tree. By default, trees are
3500 pretty-printed into a C-like representation.
3502 Enable more detailed dumps (not honored by every dump option).
3504 Enable dumping various statistics about the pass (not honored by every dump
3507 Enable showing basic block boundaries (disabled in raw dumps).
3509 Enable showing virtual operands for every statement.
3511 Enable showing line numbers for statements.
3513 Enable showing the unique ID (@code{DECL_UID}) for each variable.
3515 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
3518 The following tree dumps are possible:
3522 Dump before any tree based optimization, to @file{@var{file}.original}.
3525 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3528 Dump after function inlining, to @file{@var{file}.inlined}.
3531 @opindex fdump-tree-gimple
3532 Dump each function before and after the gimplification pass to a file. The
3533 file name is made by appending @file{.gimple} to the source file name.
3536 @opindex fdump-tree-cfg
3537 Dump the control flow graph of each function to a file. The file name is
3538 made by appending @file{.cfg} to the source file name.
3541 @opindex fdump-tree-vcg
3542 Dump the control flow graph of each function to a file in VCG format. The
3543 file name is made by appending @file{.vcg} to the source file name. Note
3544 that if the file contains more than one function, the generated file cannot
3545 be used directly by VCG. You will need to cut and paste each function's
3546 graph into its own separate file first.
3549 @opindex fdump-tree-ch
3550 Dump each function after copying loop headers. The file name is made by
3551 appending @file{.ch} to the source file name.
3554 @opindex fdump-tree-ssa
3555 Dump SSA related information to a file. The file name is made by appending
3556 @file{.ssa} to the source file name.
3559 @opindex fdump-tree-alias
3560 Dump aliasing information for each function. The file name is made by
3561 appending @file{.alias} to the source file name.
3564 @opindex fdump-tree-ccp
3565 Dump each function after CCP. The file name is made by appending
3566 @file{.ccp} to the source file name.
3569 @opindex fdump-tree-pre
3570 Dump trees after partial redundancy elimination. The file name is made
3571 by appending @file{.pre} to the source file name.
3574 @opindex fdump-tree-fre
3575 Dump trees after full redundancy elimination. The file name is made
3576 by appending @file{.fre} to the source file name.
3579 @opindex fdump-tree-dce
3580 Dump each function after dead code elimination. The file name is made by
3581 appending @file{.dce} to the source file name.
3584 @opindex fdump-tree-mudflap
3585 Dump each function after adding mudflap instrumentation. The file name is
3586 made by appending @file{.mudflap} to the source file name.
3589 @opindex fdump-tree-sra
3590 Dump each function after performing scalar replacement of aggregates. The
3591 file name is made by appending @file{.sra} to the source file name.
3594 @opindex fdump-tree-dom
3595 Dump each function after applying dominator tree optimizations. The file
3596 name is made by appending @file{.dom} to the source file name.
3599 @opindex fdump-tree-dse
3600 Dump each function after applying dead store elimination. The file
3601 name is made by appending @file{.dse} to the source file name.
3604 @opindex fdump-tree-phiopt
3605 Dump each function after optimizing PHI nodes into straightline code. The file
3606 name is made by appending @file{.phiopt} to the source file name.
3609 @opindex fdump-tree-forwprop
3610 Dump each function after forward propagating single use variables. The file
3611 name is made by appending @file{.forwprop} to the source file name.
3614 @opindex fdump-tree-copyrename
3615 Dump each function after applying the copy rename optimization. The file
3616 name is made by appending @file{.copyrename} to the source file name.
3619 @opindex fdump-tree-nrv
3620 Dump each function after applying the named return value optimization on
3621 generic trees. The file name is made by appending @file{.nrv} to the source
3625 @opindex fdump-tree-all
3626 Enable all the available tree dumps with the flags provided in this option.
3629 @item -frandom-seed=@var{string}
3630 @opindex frandom-string
3631 This option provides a seed that GCC uses when it would otherwise use
3632 random numbers. It is used to generate certain symbol names
3633 that have to be different in every compiled file. It is also used to
3634 place unique stamps in coverage data files and the object files that
3635 produce them. You can use the @option{-frandom-seed} option to produce
3636 reproducibly identical object files.
3638 The @var{string} should be different for every file you compile.
3640 @item -fsched-verbose=@var{n}
3641 @opindex fsched-verbose
3642 On targets that use instruction scheduling, this option controls the
3643 amount of debugging output the scheduler prints. This information is
3644 written to standard error, unless @option{-dS} or @option{-dR} is
3645 specified, in which case it is output to the usual dump
3646 listing file, @file{.sched} or @file{.sched2} respectively. However
3647 for @var{n} greater than nine, the output is always printed to standard
3650 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
3651 same information as @option{-dRS}. For @var{n} greater than one, it
3652 also output basic block probabilities, detailed ready list information
3653 and unit/insn info. For @var{n} greater than two, it includes RTL
3654 at abort point, control-flow and regions info. And for @var{n} over
3655 four, @option{-fsched-verbose} also includes dependence info.
3659 Store the usual ``temporary'' intermediate files permanently; place them
3660 in the current directory and name them based on the source file. Thus,
3661 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
3662 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
3663 preprocessed @file{foo.i} output file even though the compiler now
3664 normally uses an integrated preprocessor.
3668 Report the CPU time taken by each subprocess in the compilation
3669 sequence. For C source files, this is the compiler proper and assembler
3670 (plus the linker if linking is done). The output looks like this:
3677 The first number on each line is the ``user time,'' that is time spent
3678 executing the program itself. The second number is ``system time,''
3679 time spent executing operating system routines on behalf of the program.
3680 Both numbers are in seconds.
3682 @item -fvar-tracking
3683 @opindex fvar-tracking
3684 Run variable tracking pass. It computes where variables are stored at each
3685 position in code. Better debugging information is then generated
3686 (if the debugging information format supports this information).
3688 It is enabled by default when compiling with optimization (@option{-Os},
3689 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
3690 the debug info format supports it.
3692 @item -print-file-name=@var{library}
3693 @opindex print-file-name
3694 Print the full absolute name of the library file @var{library} that
3695 would be used when linking---and don't do anything else. With this
3696 option, GCC does not compile or link anything; it just prints the
3699 @item -print-multi-directory
3700 @opindex print-multi-directory
3701 Print the directory name corresponding to the multilib selected by any
3702 other switches present in the command line. This directory is supposed
3703 to exist in @env{GCC_EXEC_PREFIX}.
3705 @item -print-multi-lib
3706 @opindex print-multi-lib
3707 Print the mapping from multilib directory names to compiler switches
3708 that enable them. The directory name is separated from the switches by
3709 @samp{;}, and each switch starts with an @samp{@@} instead of the
3710 @samp{-}, without spaces between multiple switches. This is supposed to
3711 ease shell-processing.
3713 @item -print-prog-name=@var{program}
3714 @opindex print-prog-name
3715 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
3717 @item -print-libgcc-file-name
3718 @opindex print-libgcc-file-name
3719 Same as @option{-print-file-name=libgcc.a}.
3721 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
3722 but you do want to link with @file{libgcc.a}. You can do
3725 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
3728 @item -print-search-dirs
3729 @opindex print-search-dirs
3730 Print the name of the configured installation directory and a list of
3731 program and library directories @command{gcc} will search---and don't do anything else.
3733 This is useful when @command{gcc} prints the error message
3734 @samp{installation problem, cannot exec cpp0: No such file or directory}.
3735 To resolve this you either need to put @file{cpp0} and the other compiler
3736 components where @command{gcc} expects to find them, or you can set the environment
3737 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
3738 Don't forget the trailing '/'.
3739 @xref{Environment Variables}.
3742 @opindex dumpmachine
3743 Print the compiler's target machine (for example,
3744 @samp{i686-pc-linux-gnu})---and don't do anything else.
3747 @opindex dumpversion
3748 Print the compiler version (for example, @samp{3.0})---and don't do
3753 Print the compiler's built-in specs---and don't do anything else. (This
3754 is used when GCC itself is being built.) @xref{Spec Files}.
3756 @item -feliminate-unused-debug-types
3757 @opindex feliminate-unused-debug-types
3758 Normally, when producing DWARF2 output, GCC will emit debugging
3759 information for all types declared in a compilation
3760 unit, regardless of whether or not they are actually used
3761 in that compilation unit. Sometimes this is useful, such as
3762 if, in the debugger, you want to cast a value to a type that is
3763 not actually used in your program (but is declared). More often,
3764 however, this results in a significant amount of wasted space.
3765 With this option, GCC will avoid producing debug symbol output
3766 for types that are nowhere used in the source file being compiled.
3769 @node Optimize Options
3770 @section Options That Control Optimization
3771 @cindex optimize options
3772 @cindex options, optimization
3774 These options control various sorts of optimizations.
3776 Without any optimization option, the compiler's goal is to reduce the
3777 cost of compilation and to make debugging produce the expected
3778 results. Statements are independent: if you stop the program with a
3779 breakpoint between statements, you can then assign a new value to any
3780 variable or change the program counter to any other statement in the
3781 function and get exactly the results you would expect from the source
3784 Turning on optimization flags makes the compiler attempt to improve
3785 the performance and/or code size at the expense of compilation time
3786 and possibly the ability to debug the program.
3788 The compiler performs optimization based on the knowledge it has of
3789 the program. Optimization levels @option{-O2} and above, in
3790 particular, enable @emph{unit-at-a-time} mode, which allows the
3791 compiler to consider information gained from later functions in
3792 the file when compiling a function. Compiling multiple files at
3793 once to a single output file in @emph{unit-at-a-time} mode allows
3794 the compiler to use information gained from all of the files when
3795 compiling each of them.
3797 Not all optimizations are controlled directly by a flag. Only
3798 optimizations that have a flag are listed.
3805 Optimize. Optimizing compilation takes somewhat more time, and a lot
3806 more memory for a large function.
3808 With @option{-O}, the compiler tries to reduce code size and execution
3809 time, without performing any optimizations that take a great deal of
3812 @option{-O} turns on the following optimization flags:
3813 @gccoptlist{-fdefer-pop @gol
3814 -fmerge-constants @gol
3816 -floop-optimize @gol
3817 -fif-conversion @gol
3818 -fif-conversion2 @gol
3819 -fdelayed-branch @gol
3820 -fguess-branch-probability @gol
3823 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
3824 where doing so does not interfere with debugging.
3828 Optimize even more. GCC performs nearly all supported optimizations
3829 that do not involve a space-speed tradeoff. The compiler does not
3830 perform loop unrolling or function inlining when you specify @option{-O2}.
3831 As compared to @option{-O}, this option increases both compilation time
3832 and the performance of the generated code.
3834 @option{-O2} turns on all optimization flags specified by @option{-O}. It
3835 also turns on the following optimization flags:
3836 @gccoptlist{-fforce-mem @gol
3837 -foptimize-sibling-calls @gol
3838 -fstrength-reduce @gol
3839 -fcse-follow-jumps -fcse-skip-blocks @gol
3840 -frerun-cse-after-loop -frerun-loop-opt @gol
3841 -fgcse -fgcse-lm -fgcse-sm -fgcse-las @gol
3842 -fdelete-null-pointer-checks @gol
3843 -fexpensive-optimizations @gol
3845 -fschedule-insns -fschedule-insns2 @gol
3846 -fsched-interblock -fsched-spec @gol
3849 -freorder-blocks -freorder-functions @gol
3850 -fstrict-aliasing @gol
3851 -funit-at-a-time @gol
3852 -falign-functions -falign-jumps @gol
3853 -falign-loops -falign-labels @gol
3856 Please note the warning under @option{-fgcse} about
3857 invoking @option{-O2} on programs that use computed gotos.
3861 Optimize yet more. @option{-O3} turns on all optimizations specified by
3862 @option{-O2} and also turns on the @option{-finline-functions},
3863 @option{-fweb} and @option{-fgcse-after-reload} options.
3867 Do not optimize. This is the default.
3871 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
3872 do not typically increase code size. It also performs further
3873 optimizations designed to reduce code size.
3875 @option{-Os} disables the following optimization flags:
3876 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
3877 -falign-labels -freorder-blocks -freorder-blocks-and-partition -fprefetch-loop-arrays}
3879 If you use multiple @option{-O} options, with or without level numbers,
3880 the last such option is the one that is effective.
3883 Options of the form @option{-f@var{flag}} specify machine-independent
3884 flags. Most flags have both positive and negative forms; the negative
3885 form of @option{-ffoo} would be @option{-fno-foo}. In the table
3886 below, only one of the forms is listed---the one you typically will
3887 use. You can figure out the other form by either removing @samp{no-}
3890 The following options control specific optimizations. They are either
3891 activated by @option{-O} options or are related to ones that are. You
3892 can use the following flags in the rare cases when ``fine-tuning'' of
3893 optimizations to be performed is desired.
3896 @item -fno-default-inline
3897 @opindex fno-default-inline
3898 Do not make member functions inline by default merely because they are
3899 defined inside the class scope (C++ only). Otherwise, when you specify
3900 @w{@option{-O}}, member functions defined inside class scope are compiled
3901 inline by default; i.e., you don't need to add @samp{inline} in front of
3902 the member function name.
3904 @item -fno-defer-pop
3905 @opindex fno-defer-pop
3906 Always pop the arguments to each function call as soon as that function
3907 returns. For machines which must pop arguments after a function call,
3908 the compiler normally lets arguments accumulate on the stack for several
3909 function calls and pops them all at once.
3911 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3915 Force memory operands to be copied into registers before doing
3916 arithmetic on them. This produces better code by making all memory
3917 references potential common subexpressions. When they are not common
3918 subexpressions, instruction combination should eliminate the separate
3921 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3924 @opindex fforce-addr
3925 Force memory address constants to be copied into registers before
3926 doing arithmetic on them. This may produce better code just as
3927 @option{-fforce-mem} may.
3929 @item -fomit-frame-pointer
3930 @opindex fomit-frame-pointer
3931 Don't keep the frame pointer in a register for functions that
3932 don't need one. This avoids the instructions to save, set up and
3933 restore frame pointers; it also makes an extra register available
3934 in many functions. @strong{It also makes debugging impossible on
3937 On some machines, such as the VAX, this flag has no effect, because
3938 the standard calling sequence automatically handles the frame pointer
3939 and nothing is saved by pretending it doesn't exist. The
3940 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
3941 whether a target machine supports this flag. @xref{Registers,,Register
3942 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
3944 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3946 @item -foptimize-sibling-calls
3947 @opindex foptimize-sibling-calls
3948 Optimize sibling and tail recursive calls.
3950 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3954 Don't pay attention to the @code{inline} keyword. Normally this option
3955 is used to keep the compiler from expanding any functions inline.
3956 Note that if you are not optimizing, no functions can be expanded inline.
3958 @item -finline-functions
3959 @opindex finline-functions
3960 Integrate all simple functions into their callers. The compiler
3961 heuristically decides which functions are simple enough to be worth
3962 integrating in this way.
3964 If all calls to a given function are integrated, and the function is
3965 declared @code{static}, then the function is normally not output as
3966 assembler code in its own right.
3968 Enabled at level @option{-O3}.
3970 @item -finline-limit=@var{n}
3971 @opindex finline-limit
3972 By default, GCC limits the size of functions that can be inlined. This flag
3973 allows the control of this limit for functions that are explicitly marked as
3974 inline (i.e., marked with the inline keyword or defined within the class
3975 definition in c++). @var{n} is the size of functions that can be inlined in
3976 number of pseudo instructions (not counting parameter handling). The default
3977 value of @var{n} is 600.
3978 Increasing this value can result in more inlined code at
3979 the cost of compilation time and memory consumption. Decreasing usually makes
3980 the compilation faster and less code will be inlined (which presumably
3981 means slower programs). This option is particularly useful for programs that
3982 use inlining heavily such as those based on recursive templates with C++.
3984 Inlining is actually controlled by a number of parameters, which may be
3985 specified individually by using @option{--param @var{name}=@var{value}}.
3986 The @option{-finline-limit=@var{n}} option sets some of these parameters
3990 @item max-inline-insns-single
3991 is set to @var{n}/2.
3992 @item max-inline-insns-auto
3993 is set to @var{n}/2.
3994 @item min-inline-insns
3995 is set to 130 or @var{n}/4, whichever is smaller.
3996 @item max-inline-insns-rtl
4000 See below for a documentation of the individual
4001 parameters controlling inlining.
4003 @emph{Note:} pseudo instruction represents, in this particular context, an
4004 abstract measurement of function's size. In no way, it represents a count
4005 of assembly instructions and as such its exact meaning might change from one
4006 release to an another.
4008 @item -fkeep-inline-functions
4009 @opindex fkeep-inline-functions
4010 Even if all calls to a given function are integrated, and the function
4011 is declared @code{static}, nevertheless output a separate run-time
4012 callable version of the function. This switch does not affect
4013 @code{extern inline} functions.
4015 @item -fkeep-static-consts
4016 @opindex fkeep-static-consts
4017 Emit variables declared @code{static const} when optimization isn't turned
4018 on, even if the variables aren't referenced.
4020 GCC enables this option by default. If you want to force the compiler to
4021 check if the variable was referenced, regardless of whether or not
4022 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4024 @item -fmerge-constants
4025 Attempt to merge identical constants (string constants and floating point
4026 constants) across compilation units.
4028 This option is the default for optimized compilation if the assembler and
4029 linker support it. Use @option{-fno-merge-constants} to inhibit this
4032 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4034 @item -fmerge-all-constants
4035 Attempt to merge identical constants and identical variables.
4037 This option implies @option{-fmerge-constants}. In addition to
4038 @option{-fmerge-constants} this considers e.g. even constant initialized
4039 arrays or initialized constant variables with integral or floating point
4040 types. Languages like C or C++ require each non-automatic variable to
4041 have distinct location, so using this option will result in non-conforming
4044 @item -fmodulo-sched
4045 @opindex fmodulo-sched
4046 Perform swing modulo scheduling immediately before the first scheduling
4047 pass. This pass looks at innermost loops and reorders their
4048 instructions by overlapping different iterations.
4052 Use a graph coloring register allocator. Currently this option is meant
4053 only for testing. Users should not specify this option, since it is not
4054 yet ready for production use.
4056 @item -fno-branch-count-reg
4057 @opindex fno-branch-count-reg
4058 Do not use ``decrement and branch'' instructions on a count register,
4059 but instead generate a sequence of instructions that decrement a
4060 register, compare it against zero, then branch based upon the result.
4061 This option is only meaningful on architectures that support such
4062 instructions, which include x86, PowerPC, IA-64 and S/390.
4064 The default is @option{-fbranch-count-reg}, enabled when
4065 @option{-fstrength-reduce} is enabled.
4067 @item -fno-function-cse
4068 @opindex fno-function-cse
4069 Do not put function addresses in registers; make each instruction that
4070 calls a constant function contain the function's address explicitly.
4072 This option results in less efficient code, but some strange hacks
4073 that alter the assembler output may be confused by the optimizations
4074 performed when this option is not used.
4076 The default is @option{-ffunction-cse}
4078 @item -fno-zero-initialized-in-bss
4079 @opindex fno-zero-initialized-in-bss
4080 If the target supports a BSS section, GCC by default puts variables that
4081 are initialized to zero into BSS@. This can save space in the resulting
4084 This option turns off this behavior because some programs explicitly
4085 rely on variables going to the data section. E.g., so that the
4086 resulting executable can find the beginning of that section and/or make
4087 assumptions based on that.
4089 The default is @option{-fzero-initialized-in-bss}.
4091 @item -fbounds-check
4092 @opindex fbounds-check
4093 For front-ends that support it, generate additional code to check that
4094 indices used to access arrays are within the declared range. This is
4095 currently only supported by the Java and Fortran front-ends, where
4096 this option defaults to true and false respectively.
4098 @item -fmudflap -fmudflapth -fmudflapir
4102 @cindex bounds checking
4104 For front-ends that support it (C and C++), instrument all risky
4105 pointer/array dereferencing operations, some standard library
4106 string/heap functions, and some other associated constructs with
4107 range/validity tests. Modules so instrumented should be immune to
4108 buffer overflows, invalid heap use, and some other classes of C/C++
4109 programming errors. The instrumentation relies on a separate runtime
4110 library (@file{libmudflap}), which will be linked into a program if
4111 @option{-fmudflap} is given at link time. Run-time behavior of the
4112 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4113 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4116 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4117 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4118 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4119 instrumentation should ignore pointer reads. This produces less
4120 instrumentation (and therefore faster execution) and still provides
4121 some protection against outright memory corrupting writes, but allows
4122 erroneously read data to propagate within a program.
4124 @item -fstrength-reduce
4125 @opindex fstrength-reduce
4126 Perform the optimizations of loop strength reduction and
4127 elimination of iteration variables.
4129 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4131 @item -fthread-jumps
4132 @opindex fthread-jumps
4133 Perform optimizations where we check to see if a jump branches to a
4134 location where another comparison subsumed by the first is found. If
4135 so, the first branch is redirected to either the destination of the
4136 second branch or a point immediately following it, depending on whether
4137 the condition is known to be true or false.
4139 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4141 @item -fcse-follow-jumps
4142 @opindex fcse-follow-jumps
4143 In common subexpression elimination, scan through jump instructions
4144 when the target of the jump is not reached by any other path. For
4145 example, when CSE encounters an @code{if} statement with an
4146 @code{else} clause, CSE will follow the jump when the condition
4149 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4151 @item -fcse-skip-blocks
4152 @opindex fcse-skip-blocks
4153 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4154 follow jumps which conditionally skip over blocks. When CSE
4155 encounters a simple @code{if} statement with no else clause,
4156 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4157 body of the @code{if}.
4159 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4161 @item -frerun-cse-after-loop
4162 @opindex frerun-cse-after-loop
4163 Re-run common subexpression elimination after loop optimizations has been
4166 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4168 @item -frerun-loop-opt
4169 @opindex frerun-loop-opt
4170 Run the loop optimizer twice.
4172 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4176 Perform a global common subexpression elimination pass.
4177 This pass also performs global constant and copy propagation.
4179 @emph{Note:} When compiling a program using computed gotos, a GCC
4180 extension, you may get better runtime performance if you disable
4181 the global common subexpression elimination pass by adding
4182 @option{-fno-gcse} to the command line.
4184 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4188 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4189 attempt to move loads which are only killed by stores into themselves. This
4190 allows a loop containing a load/store sequence to be changed to a load outside
4191 the loop, and a copy/store within the loop.
4193 Enabled by default when gcse is enabled.
4197 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4198 global common subexpression elimination. This pass will attempt to move
4199 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4200 loops containing a load/store sequence can be changed to a load before
4201 the loop and a store after the loop.
4203 Enabled by default when gcse is enabled.
4207 When @option{-fgcse-las} is enabled, the global common subexpression
4208 elimination pass eliminates redundant loads that come after stores to the
4209 same memory location (both partial and full redundancies).
4211 Enabled by default when gcse is enabled.
4213 @item -fgcse-after-reload
4214 @opindex fgcse-after-reload
4215 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4216 pass is performed after reload. The purpose of this pass is to cleanup
4219 @item -floop-optimize
4220 @opindex floop-optimize
4221 Perform loop optimizations: move constant expressions out of loops, simplify
4222 exit test conditions and optionally do strength-reduction and loop unrolling as
4225 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4227 @item -floop-optimize2
4228 @opindex floop-optimize2
4229 Perform loop optimizations using the new loop optimizer. The optimizations
4230 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4233 @item -fcrossjumping
4234 @opindex crossjumping
4235 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4236 resulting code may or may not perform better than without cross-jumping.
4238 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4240 @item -fif-conversion
4241 @opindex if-conversion
4242 Attempt to transform conditional jumps into branch-less equivalents. This
4243 include use of conditional moves, min, max, set flags and abs instructions, and
4244 some tricks doable by standard arithmetics. The use of conditional execution
4245 on chips where it is available is controlled by @code{if-conversion2}.
4247 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4249 @item -fif-conversion2
4250 @opindex if-conversion2
4251 Use conditional execution (where available) to transform conditional jumps into
4252 branch-less equivalents.
4254 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4256 @item -fdelete-null-pointer-checks
4257 @opindex fdelete-null-pointer-checks
4258 Use global dataflow analysis to identify and eliminate useless checks
4259 for null pointers. The compiler assumes that dereferencing a null
4260 pointer would have halted the program. If a pointer is checked after
4261 it has already been dereferenced, it cannot be null.
4263 In some environments, this assumption is not true, and programs can
4264 safely dereference null pointers. Use
4265 @option{-fno-delete-null-pointer-checks} to disable this optimization
4266 for programs which depend on that behavior.
4268 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4270 @item -fexpensive-optimizations
4271 @opindex fexpensive-optimizations
4272 Perform a number of minor optimizations that are relatively expensive.
4274 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4276 @item -foptimize-register-move
4278 @opindex foptimize-register-move
4280 Attempt to reassign register numbers in move instructions and as
4281 operands of other simple instructions in order to maximize the amount of
4282 register tying. This is especially helpful on machines with two-operand
4285 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4288 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4290 @item -fdelayed-branch
4291 @opindex fdelayed-branch
4292 If supported for the target machine, attempt to reorder instructions
4293 to exploit instruction slots available after delayed branch
4296 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4298 @item -fschedule-insns
4299 @opindex fschedule-insns
4300 If supported for the target machine, attempt to reorder instructions to
4301 eliminate execution stalls due to required data being unavailable. This
4302 helps machines that have slow floating point or memory load instructions
4303 by allowing other instructions to be issued until the result of the load
4304 or floating point instruction is required.
4306 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4308 @item -fschedule-insns2
4309 @opindex fschedule-insns2
4310 Similar to @option{-fschedule-insns}, but requests an additional pass of
4311 instruction scheduling after register allocation has been done. This is
4312 especially useful on machines with a relatively small number of
4313 registers and where memory load instructions take more than one cycle.
4315 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4317 @item -fno-sched-interblock
4318 @opindex fno-sched-interblock
4319 Don't schedule instructions across basic blocks. This is normally
4320 enabled by default when scheduling before register allocation, i.e.@:
4321 with @option{-fschedule-insns} or at @option{-O2} or higher.
4323 @item -fno-sched-spec
4324 @opindex fno-sched-spec
4325 Don't allow speculative motion of non-load instructions. This is normally
4326 enabled by default when scheduling before register allocation, i.e.@:
4327 with @option{-fschedule-insns} or at @option{-O2} or higher.
4329 @item -fsched-spec-load
4330 @opindex fsched-spec-load
4331 Allow speculative motion of some load instructions. This only makes
4332 sense when scheduling before register allocation, i.e.@: with
4333 @option{-fschedule-insns} or at @option{-O2} or higher.
4335 @item -fsched-spec-load-dangerous
4336 @opindex fsched-spec-load-dangerous
4337 Allow speculative motion of more load instructions. This only makes
4338 sense when scheduling before register allocation, i.e.@: with
4339 @option{-fschedule-insns} or at @option{-O2} or higher.
4341 @item -fsched-stalled-insns=@var{n}
4342 @opindex fsched-stalled-insns
4343 Define how many insns (if any) can be moved prematurely from the queue
4344 of stalled insns into the ready list, during the second scheduling pass.
4346 @item -fsched-stalled-insns-dep=@var{n}
4347 @opindex fsched-stalled-insns-dep
4348 Define how many insn groups (cycles) will be examined for a dependency
4349 on a stalled insn that is candidate for premature removal from the queue
4350 of stalled insns. Has an effect only during the second scheduling pass,
4351 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4353 @item -fsched2-use-superblocks
4354 @opindex fsched2-use-superblocks
4355 When scheduling after register allocation, do use superblock scheduling
4356 algorithm. Superblock scheduling allows motion across basic block boundaries
4357 resulting on faster schedules. This option is experimental, as not all machine
4358 descriptions used by GCC model the CPU closely enough to avoid unreliable
4359 results from the algorithm.
4361 This only makes sense when scheduling after register allocation, i.e.@: with
4362 @option{-fschedule-insns2} or at @option{-O2} or higher.
4364 @item -fsched2-use-traces
4365 @opindex fsched2-use-traces
4366 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4367 allocation and additionally perform code duplication in order to increase the
4368 size of superblocks using tracer pass. See @option{-ftracer} for details on
4371 This mode should produce faster but significantly longer programs. Also
4372 without @code{-fbranch-probabilities} the traces constructed may not match the
4373 reality and hurt the performance. This only makes
4374 sense when scheduling after register allocation, i.e.@: with
4375 @option{-fschedule-insns2} or at @option{-O2} or higher.
4377 @item -fcaller-saves
4378 @opindex fcaller-saves
4379 Enable values to be allocated in registers that will be clobbered by
4380 function calls, by emitting extra instructions to save and restore the
4381 registers around such calls. Such allocation is done only when it
4382 seems to result in better code than would otherwise be produced.
4384 This option is always enabled by default on certain machines, usually
4385 those which have no call-preserved registers to use instead.
4387 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4390 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4391 enabled by default at -O and higher.
4394 Perform Full Redundancy Elimination (FRE) on trees. The difference
4395 between FRE and PRE is that FRE only considers expressions
4396 that are computed on all paths leading to the redundant computation.
4397 This analysis faster than PRE, though it exposes fewer redundancies.
4398 This flag is enabled by default at -O and higher.
4401 Perform sparse conditional constant propagation (CCP) on trees. This flag
4402 is enabled by default at -O and higher.
4405 Perform dead code elimination (DCE) on trees. This flag is enabled by
4406 default at -O and higher.
4408 @item -ftree-dominator-opts
4409 Perform dead code elimination (DCE) on trees. This flag is enabled by
4410 default at -O and higher.
4413 Perform loop header copying on trees. This is beneficial since it increases
4414 effectivity of code motion optimizations. It also saves one jump. This flag
4415 is enabled by default at -O and higher. It is not enabled for -Os, since it
4416 usually increases code size.
4418 @item -ftree-loop-optimize
4419 Perform loop optimizations on trees. This flag is enabled by default at -O
4423 Perform loop invariant motion on trees. This pass moves only invartiants that
4424 would be hard to handle on rtl level (function calls, operations that expand to
4425 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
4426 operands of conditions that are invariant out of the loop, so that we can use
4427 just trivial invariantness analysis in loop unswitching. The pass also includes
4431 Perform scalar replacement of aggregates. This pass replaces structure
4432 references with scalars to prevent committing structures to memory too
4433 early. This flag is enabled by default at -O and higher.
4435 @item -ftree-copyrename
4436 Perform copy renaming on trees. This pass attempts to rename compiler
4437 temporaries to other variables at copy locations, usually resulting in
4438 variable names which more closely resemble the original variables. This flag
4439 is enabled by default at -O and higher.
4442 Perform temporary expression replacement during the SSA->normal phase. Single
4443 use/single def temporaries are replaced at their use location with their
4444 defining expression. This results in non-GIMPLE code, but gives the expanders
4445 much more complex trees to work on resulting in better RTL generation. This is
4446 enabled by default at -O and higher.
4449 Perform live range splitting during the SSA->normal phase. Distinct live
4450 ranges of a variable are split into unique variables, allowing for better
4451 optimization later. This is enabled by default at -O and higher.
4455 Perform tail duplication to enlarge superblock size. This transformation
4456 simplifies the control flow of the function allowing other optimizations to do
4459 @item -funroll-loops
4460 @opindex funroll-loops
4461 Unroll loops whose number of iterations can be determined at compile
4462 time or upon entry to the loop. @option{-funroll-loops} implies both
4463 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4464 option makes code larger, and may or may not make it run faster.
4466 @item -funroll-all-loops
4467 @opindex funroll-all-loops
4468 Unroll all loops, even if their number of iterations is uncertain when
4469 the loop is entered. This usually makes programs run more slowly.
4470 @option{-funroll-all-loops} implies the same options as
4471 @option{-funroll-loops},
4473 @item -fprefetch-loop-arrays
4474 @opindex fprefetch-loop-arrays
4475 If supported by the target machine, generate instructions to prefetch
4476 memory to improve the performance of loops that access large arrays.
4478 @item -fmove-all-movables
4479 @opindex fmove-all-movables
4480 Forces all invariant computations in loops to be moved
4483 @item -freduce-all-givs
4484 @opindex freduce-all-givs
4485 Forces all general-induction variables in loops to be
4488 @emph{Note:} When compiling programs written in Fortran,
4489 @option{-fmove-all-movables} and @option{-freduce-all-givs} are enabled
4490 by default when you use the optimizer.
4492 These options may generate better or worse code; results are highly
4493 dependent on the structure of loops within the source code.
4495 These two options are intended to be removed someday, once
4496 they have helped determine the efficacy of various
4497 approaches to improving loop optimizations.
4499 Please contact @w{@email{gcc@@gcc.gnu.org}}, and describe how use of
4500 these options affects the performance of your production code.
4501 Examples of code that runs @emph{slower} when these options are
4502 @emph{enabled} are very valuable.
4505 @itemx -fno-peephole2
4506 @opindex fno-peephole
4507 @opindex fno-peephole2
4508 Disable any machine-specific peephole optimizations. The difference
4509 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
4510 are implemented in the compiler; some targets use one, some use the
4511 other, a few use both.
4513 @option{-fpeephole} is enabled by default.
4514 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4516 @item -fno-guess-branch-probability
4517 @opindex fno-guess-branch-probability
4518 Do not guess branch probabilities using a randomized model.
4520 Sometimes GCC will opt to use a randomized model to guess branch
4521 probabilities, when none are available from either profiling feedback
4522 (@option{-fprofile-arcs}) or @samp{__builtin_expect}. This means that
4523 different runs of the compiler on the same program may produce different
4526 In a hard real-time system, people don't want different runs of the
4527 compiler to produce code that has different behavior; minimizing
4528 non-determinism is of paramount import. This switch allows users to
4529 reduce non-determinism, possibly at the expense of inferior
4532 The default is @option{-fguess-branch-probability} at levels
4533 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4535 @item -freorder-blocks
4536 @opindex freorder-blocks
4537 Reorder basic blocks in the compiled function in order to reduce number of
4538 taken branches and improve code locality.
4540 Enabled at levels @option{-O2}, @option{-O3}.
4542 @item -freorder-blocks-and-partition
4543 @opindex freorder-blocks-and-partition
4544 In addition to reordering basic blocks in the compiled function, in order
4545 to reduce number of taken branches, partitions hot and cold basic blocks
4546 into separate sections of the assembly and .o files, to improve
4547 paging and cache locality performance.
4549 @item -freorder-functions
4550 @opindex freorder-functions
4551 Reorder basic blocks in the compiled function in order to reduce number of
4552 taken branches and improve code locality. This is implemented by using special
4553 subsections @code{.text.hot} for most frequently executed functions and
4554 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
4555 the linker so object file format must support named sections and linker must
4556 place them in a reasonable way.
4558 Also profile feedback must be available in to make this option effective. See
4559 @option{-fprofile-arcs} for details.
4561 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4563 @item -fstrict-aliasing
4564 @opindex fstrict-aliasing
4565 Allows the compiler to assume the strictest aliasing rules applicable to
4566 the language being compiled. For C (and C++), this activates
4567 optimizations based on the type of expressions. In particular, an
4568 object of one type is assumed never to reside at the same address as an
4569 object of a different type, unless the types are almost the same. For
4570 example, an @code{unsigned int} can alias an @code{int}, but not a
4571 @code{void*} or a @code{double}. A character type may alias any other
4574 Pay special attention to code like this:
4587 The practice of reading from a different union member than the one most
4588 recently written to (called ``type-punning'') is common. Even with
4589 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
4590 is accessed through the union type. So, the code above will work as
4591 expected. However, this code might not:
4602 Every language that wishes to perform language-specific alias analysis
4603 should define a function that computes, given an @code{tree}
4604 node, an alias set for the node. Nodes in different alias sets are not
4605 allowed to alias. For an example, see the C front-end function
4606 @code{c_get_alias_set}.
4608 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4610 @item -falign-functions
4611 @itemx -falign-functions=@var{n}
4612 @opindex falign-functions
4613 Align the start of functions to the next power-of-two greater than
4614 @var{n}, skipping up to @var{n} bytes. For instance,
4615 @option{-falign-functions=32} aligns functions to the next 32-byte
4616 boundary, but @option{-falign-functions=24} would align to the next
4617 32-byte boundary only if this can be done by skipping 23 bytes or less.
4619 @option{-fno-align-functions} and @option{-falign-functions=1} are
4620 equivalent and mean that functions will not be aligned.
4622 Some assemblers only support this flag when @var{n} is a power of two;
4623 in that case, it is rounded up.
4625 If @var{n} is not specified or is zero, use a machine-dependent default.
4627 Enabled at levels @option{-O2}, @option{-O3}.
4629 @item -falign-labels
4630 @itemx -falign-labels=@var{n}
4631 @opindex falign-labels
4632 Align all branch targets to a power-of-two boundary, skipping up to
4633 @var{n} bytes like @option{-falign-functions}. This option can easily
4634 make code slower, because it must insert dummy operations for when the
4635 branch target is reached in the usual flow of the code.
4637 @option{-fno-align-labels} and @option{-falign-labels=1} are
4638 equivalent and mean that labels will not be aligned.
4640 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
4641 are greater than this value, then their values are used instead.
4643 If @var{n} is not specified or is zero, use a machine-dependent default
4644 which is very likely to be @samp{1}, meaning no alignment.
4646 Enabled at levels @option{-O2}, @option{-O3}.
4649 @itemx -falign-loops=@var{n}
4650 @opindex falign-loops
4651 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
4652 like @option{-falign-functions}. The hope is that the loop will be
4653 executed many times, which will make up for any execution of the dummy
4656 @option{-fno-align-loops} and @option{-falign-loops=1} are
4657 equivalent and mean that loops will not be aligned.
4659 If @var{n} is not specified or is zero, use a machine-dependent default.
4661 Enabled at levels @option{-O2}, @option{-O3}.
4664 @itemx -falign-jumps=@var{n}
4665 @opindex falign-jumps
4666 Align branch targets to a power-of-two boundary, for branch targets
4667 where the targets can only be reached by jumping, skipping up to @var{n}
4668 bytes like @option{-falign-functions}. In this case, no dummy operations
4671 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
4672 equivalent and mean that loops will not be aligned.
4674 If @var{n} is not specified or is zero, use a machine-dependent default.
4676 Enabled at levels @option{-O2}, @option{-O3}.
4678 @item -funit-at-a-time
4679 @opindex funit-at-a-time
4680 Parse the whole compilation unit before starting to produce code.
4681 This allows some extra optimizations to take place but consumes
4682 more memory (in general). There are some compatibility issues
4683 with @emph{unit-at-at-time} mode:
4686 enabling @emph{unit-at-a-time} mode may change the order
4687 in which functions, variables, and top-level @code{asm} statements
4688 are emitted, and will likely break code relying on some particular
4689 ordering. The majority of such top-level @code{asm} statements,
4690 though, can be replaced by @code{section} attributes.
4693 @emph{unit-at-a-time} mode removes unreferenced static variables
4694 and functions are removed. This may result in undefined references
4695 when an @code{asm} statement refers directly to variables or functions
4696 that are otherwise unused. In that case either the variable/function
4697 shall be listed as an operand of the @code{asm} statement operand or,
4698 in the case of top-level @code{asm} statements the attribute @code{used}
4699 shall be used on the declaration.
4702 Static functions now can use non-standard passing conventions that
4703 may break @code{asm} statements calling functions directly. Again,
4704 attribute @code{used} will prevent this behavior.
4707 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
4708 but this scheme may not be supported by future releases of GCC.
4710 Enabled at levels @option{-O2}, @option{-O3}.
4714 Constructs webs as commonly used for register allocation purposes and assign
4715 each web individual pseudo register. This allows the register allocation pass
4716 to operate on pseudos directly, but also strengthens several other optimization
4717 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
4718 however, make debugging impossible, since variables will no longer stay in a
4721 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os},
4722 on targets where the default format for debugging information supports
4725 @item -fno-cprop-registers
4726 @opindex fno-cprop-registers
4727 After register allocation and post-register allocation instruction splitting,
4728 we perform a copy-propagation pass to try to reduce scheduling dependencies
4729 and occasionally eliminate the copy.
4731 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4733 @item -fprofile-generate
4734 @opindex fprofile-generate
4736 Enable options usually used for instrumenting application to produce
4737 profile useful for later recompilation with profile feedback based
4738 optimization. You must use @code{-fprofile-generate} both when
4739 compiling and when linking your program.
4741 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
4744 @opindex fprofile-use
4745 Enable profile feedback directed optimizations, and optimizations
4746 generally profitable only with profile feedback available.
4748 The following options are enabled: @code{-fbranch-probabilities},
4749 @code{-fvpt}, @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}.
4753 The following options control compiler behavior regarding floating
4754 point arithmetic. These options trade off between speed and
4755 correctness. All must be specifically enabled.
4759 @opindex ffloat-store
4760 Do not store floating point variables in registers, and inhibit other
4761 options that might change whether a floating point value is taken from a
4764 @cindex floating point precision
4765 This option prevents undesirable excess precision on machines such as
4766 the 68000 where the floating registers (of the 68881) keep more
4767 precision than a @code{double} is supposed to have. Similarly for the
4768 x86 architecture. For most programs, the excess precision does only
4769 good, but a few programs rely on the precise definition of IEEE floating
4770 point. Use @option{-ffloat-store} for such programs, after modifying
4771 them to store all pertinent intermediate computations into variables.
4775 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
4776 @option{-fno-trapping-math}, @option{-ffinite-math-only},
4777 @option{-fno-rounding-math} and @option{-fno-signaling-nans}.
4779 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
4781 This option should never be turned on by any @option{-O} option since
4782 it can result in incorrect output for programs which depend on
4783 an exact implementation of IEEE or ISO rules/specifications for
4786 @item -fno-math-errno
4787 @opindex fno-math-errno
4788 Do not set ERRNO after calling math functions that are executed
4789 with a single instruction, e.g., sqrt. A program that relies on
4790 IEEE exceptions for math error handling may want to use this flag
4791 for speed while maintaining IEEE arithmetic compatibility.
4793 This option should never be turned on by any @option{-O} option since
4794 it can result in incorrect output for programs which depend on
4795 an exact implementation of IEEE or ISO rules/specifications for
4798 The default is @option{-fmath-errno}.
4800 @item -funsafe-math-optimizations
4801 @opindex funsafe-math-optimizations
4802 Allow optimizations for floating-point arithmetic that (a) assume
4803 that arguments and results are valid and (b) may violate IEEE or
4804 ANSI standards. When used at link-time, it may include libraries
4805 or startup files that change the default FPU control word or other
4806 similar optimizations.
4808 This option should never be turned on by any @option{-O} option since
4809 it can result in incorrect output for programs which depend on
4810 an exact implementation of IEEE or ISO rules/specifications for
4813 The default is @option{-fno-unsafe-math-optimizations}.
4815 @item -ffinite-math-only
4816 @opindex ffinite-math-only
4817 Allow optimizations for floating-point arithmetic that assume
4818 that arguments and results are not NaNs or +-Infs.
4820 This option should never be turned on by any @option{-O} option since
4821 it can result in incorrect output for programs which depend on
4822 an exact implementation of IEEE or ISO rules/specifications.
4824 The default is @option{-fno-finite-math-only}.
4826 @item -fno-trapping-math
4827 @opindex fno-trapping-math
4828 Compile code assuming that floating-point operations cannot generate
4829 user-visible traps. These traps include division by zero, overflow,
4830 underflow, inexact result and invalid operation. This option implies
4831 @option{-fno-signaling-nans}. Setting this option may allow faster
4832 code if one relies on ``non-stop'' IEEE arithmetic, for example.
4834 This option should never be turned on by any @option{-O} option since
4835 it can result in incorrect output for programs which depend on
4836 an exact implementation of IEEE or ISO rules/specifications for
4839 The default is @option{-ftrapping-math}.
4841 @item -frounding-math
4842 @opindex frounding-math
4843 Disable transformations and optimizations that assume default floating
4844 point rounding behavior. This is round-to-zero for all floating point
4845 to integer conversions, and round-to-nearest for all other arithmetic
4846 truncations. This option should be specified for programs that change
4847 the FP rounding mode dynamically, or that may be executed with a
4848 non-default rounding mode. This option disables constant folding of
4849 floating point expressions at compile-time (which may be affected by
4850 rounding mode) and arithmetic transformations that are unsafe in the
4851 presence of sign-dependent rounding modes.
4853 The default is @option{-fno-rounding-math}.
4855 This option is experimental and does not currently guarantee to
4856 disable all GCC optimizations that are affected by rounding mode.
4857 Future versions of GCC may provide finer control of this setting
4858 using C99's @code{FENV_ACCESS} pragma. This command line option
4859 will be used to specify the default state for @code{FENV_ACCESS}.
4861 @item -fsignaling-nans
4862 @opindex fsignaling-nans
4863 Compile code assuming that IEEE signaling NaNs may generate user-visible
4864 traps during floating-point operations. Setting this option disables
4865 optimizations that may change the number of exceptions visible with
4866 signaling NaNs. This option implies @option{-ftrapping-math}.
4868 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
4871 The default is @option{-fno-signaling-nans}.
4873 This option is experimental and does not currently guarantee to
4874 disable all GCC optimizations that affect signaling NaN behavior.
4876 @item -fsingle-precision-constant
4877 @opindex fsingle-precision-constant
4878 Treat floating point constant as single precision constant instead of
4879 implicitly converting it to double precision constant.
4884 The following options control optimizations that may improve
4885 performance, but are not enabled by any @option{-O} options. This
4886 section includes experimental options that may produce broken code.
4889 @item -fbranch-probabilities
4890 @opindex fbranch-probabilities
4891 After running a program compiled with @option{-fprofile-arcs}
4892 (@pxref{Debugging Options,, Options for Debugging Your Program or
4893 @command{gcc}}), you can compile it a second time using
4894 @option{-fbranch-probabilities}, to improve optimizations based on
4895 the number of times each branch was taken. When the program
4896 compiled with @option{-fprofile-arcs} exits it saves arc execution
4897 counts to a file called @file{@var{sourcename}.gcda} for each source
4898 file The information in this data file is very dependent on the
4899 structure of the generated code, so you must use the same source code
4900 and the same optimization options for both compilations.
4902 With @option{-fbranch-probabilities}, GCC puts a
4903 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
4904 These can be used to improve optimization. Currently, they are only
4905 used in one place: in @file{reorg.c}, instead of guessing which path a
4906 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
4907 exactly determine which path is taken more often.
4909 @item -fprofile-values
4910 @opindex fprofile-values
4911 If combined with @option{-fprofile-arcs}, it adds code so that some
4912 data about values of expressions in the program is gathered.
4914 With @option{-fbranch-probabilities}, it reads back the data gathered
4915 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
4916 notes to instructions for their later usage in optimizations.
4918 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
4922 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
4923 a code to gather information about values of expressions.
4925 With @option{-fbranch-probabilities}, it reads back the data gathered
4926 and actually performs the optimizations based on them.
4927 Currently the optimizations include specialization of division operation
4928 using the knowledge about the value of the denominator.
4930 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
4932 @item -frename-registers
4933 @opindex frename-registers
4934 Attempt to avoid false dependencies in scheduled code by making use
4935 of registers left over after register allocation. This optimization
4936 will most benefit processors with lots of registers. Depending on the
4937 debug information format adopted by the target, however, it can
4938 make debugging impossible, since variables will no longer stay in
4939 a ``home register''.
4941 Not enabled by default at any level because it has known bugs.
4945 Use a graph coloring register allocator. Currently this option is meant
4946 for testing, so we are interested to hear about miscompilations with
4951 Perform tail duplication to enlarge superblock size. This transformation
4952 simplifies the control flow of the function allowing other optimizations to do
4955 Enabled with @option{-fprofile-use}.
4957 @item -funroll-loops
4958 @opindex funroll-loops
4959 Unroll loops whose number of iterations can be determined at compile time or
4960 upon entry to the loop. @option{-funroll-loops} implies
4961 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
4962 (i.e. complete removal of loops with small constant number of iterations).
4963 This option makes code larger, and may or may not make it run faster.
4965 Enabled with @option{-fprofile-use}.
4967 @item -funroll-all-loops
4968 @opindex funroll-all-loops
4969 Unroll all loops, even if their number of iterations is uncertain when
4970 the loop is entered. This usually makes programs run more slowly.
4971 @option{-funroll-all-loops} implies the same options as
4972 @option{-funroll-loops}.
4975 @opindex fpeel-loops
4976 Peels the loops for that there is enough information that they do not
4977 roll much (from profile feedback). It also turns on complete loop peeling
4978 (i.e. complete removal of loops with small constant number of iterations).
4980 Enabled with @option{-fprofile-use}.
4982 @item -fmove-loop-invariants
4983 @opindex fmove-loop-invariants
4984 Enables the loop invariant motion pass in the new loop optimizer. Enabled
4985 at level @option{-O1}
4987 @item -funswitch-loops
4988 @opindex funswitch-loops
4989 Move branches with loop invariant conditions out of the loop, with duplicates
4990 of the loop on both branches (modified according to result of the condition).
4992 @item -fold-unroll-loops
4993 @opindex fold-unroll-loops
4994 Unroll loops whose number of iterations can be determined at compile
4995 time or upon entry to the loop, using the old loop unroller whose loop
4996 recognition is based on notes from frontend. @option{-fold-unroll-loops} implies
4997 both @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4998 option makes code larger, and may or may not make it run faster.
5000 @item -fold-unroll-all-loops
5001 @opindex fold-unroll-all-loops
5002 Unroll all loops, even if their number of iterations is uncertain when
5003 the loop is entered. This is done using the old loop unroller whose loop
5004 recognition is based on notes from frontend. This usually makes programs run more slowly.
5005 @option{-fold-unroll-all-loops} implies the same options as
5006 @option{-fold-unroll-loops}.
5008 @item -fprefetch-loop-arrays
5009 @opindex fprefetch-loop-arrays
5010 If supported by the target machine, generate instructions to prefetch
5011 memory to improve the performance of loops that access large arrays.
5013 Disabled at level @option{-Os}.
5015 @item -ffunction-sections
5016 @itemx -fdata-sections
5017 @opindex ffunction-sections
5018 @opindex fdata-sections
5019 Place each function or data item into its own section in the output
5020 file if the target supports arbitrary sections. The name of the
5021 function or the name of the data item determines the section's name
5024 Use these options on systems where the linker can perform optimizations
5025 to improve locality of reference in the instruction space. Most systems
5026 using the ELF object format and SPARC processors running Solaris 2 have
5027 linkers with such optimizations. AIX may have these optimizations in
5030 Only use these options when there are significant benefits from doing
5031 so. When you specify these options, the assembler and linker will
5032 create larger object and executable files and will also be slower.
5033 You will not be able to use @code{gprof} on all systems if you
5034 specify this option and you may have problems with debugging if
5035 you specify both this option and @option{-g}.
5037 @item -fbranch-target-load-optimize
5038 @opindex fbranch-target-load-optimize
5039 Perform branch target register load optimization before prologue / epilogue
5041 The use of target registers can typically be exposed only during reload,
5042 thus hoisting loads out of loops and doing inter-block scheduling needs
5043 a separate optimization pass.
5045 @item -fbranch-target-load-optimize2
5046 @opindex fbranch-target-load-optimize2
5047 Perform branch target register load optimization after prologue / epilogue
5050 @item -fbtr-bb-exclusive
5051 @opindex fbtr-bb-exclusive
5052 When performing branch target register load optimization, don't reuse
5053 branch target registers in within any basic block.
5055 @item --param @var{name}=@var{value}
5057 In some places, GCC uses various constants to control the amount of
5058 optimization that is done. For example, GCC will not inline functions
5059 that contain more that a certain number of instructions. You can
5060 control some of these constants on the command-line using the
5061 @option{--param} option.
5063 The names of specific parameters, and the meaning of the values, are
5064 tied to the internals of the compiler, and are subject to change
5065 without notice in future releases.
5067 In each case, the @var{value} is an integer. The allowable choices for
5068 @var{name} are given in the following table:
5071 @item max-crossjump-edges
5072 The maximum number of incoming edges to consider for crossjumping.
5073 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5074 the number of edges incoming to each block. Increasing values mean
5075 more aggressive optimization, making the compile time increase with
5076 probably small improvement in executable size.
5078 @item max-delay-slot-insn-search
5079 The maximum number of instructions to consider when looking for an
5080 instruction to fill a delay slot. If more than this arbitrary number of
5081 instructions is searched, the time savings from filling the delay slot
5082 will be minimal so stop searching. Increasing values mean more
5083 aggressive optimization, making the compile time increase with probably
5084 small improvement in executable run time.
5086 @item max-delay-slot-live-search
5087 When trying to fill delay slots, the maximum number of instructions to
5088 consider when searching for a block with valid live register
5089 information. Increasing this arbitrarily chosen value means more
5090 aggressive optimization, increasing the compile time. This parameter
5091 should be removed when the delay slot code is rewritten to maintain the
5094 @item max-gcse-memory
5095 The approximate maximum amount of memory that will be allocated in
5096 order to perform the global common subexpression elimination
5097 optimization. If more memory than specified is required, the
5098 optimization will not be done.
5100 @item max-gcse-passes
5101 The maximum number of passes of GCSE to run. The default is 1.
5103 @item max-pending-list-length
5104 The maximum number of pending dependencies scheduling will allow
5105 before flushing the current state and starting over. Large functions
5106 with few branches or calls can create excessively large lists which
5107 needlessly consume memory and resources.
5109 @item max-inline-insns-single
5110 Several parameters control the tree inliner used in gcc.
5111 This number sets the maximum number of instructions (counted in GCC's
5112 internal representation) in a single function that the tree inliner
5113 will consider for inlining. This only affects functions declared
5114 inline and methods implemented in a class declaration (C++).
5115 The default value is 500.
5117 @item max-inline-insns-auto
5118 When you use @option{-finline-functions} (included in @option{-O3}),
5119 a lot of functions that would otherwise not be considered for inlining
5120 by the compiler will be investigated. To those functions, a different
5121 (more restrictive) limit compared to functions declared inline can
5123 The default value is 120.
5125 @item large-function-insns
5126 The limit specifying really large functions. For functions greater than this
5127 limit inlining is constrained by @option{--param large-function-growth}.
5128 This parameter is useful primarily to avoid extreme compilation time caused by non-linear
5129 algorithms used by the backend.
5130 This parameter is ignored when @option{-funit-at-a-time} is not used.
5131 The default value is 3000.
5133 @item large-function-growth
5134 Specifies maximal growth of large function caused by inlining in percents.
5135 This parameter is ignored when @option{-funit-at-a-time} is not used.
5136 The default value is 200.
5138 @item inline-unit-growth
5139 Specifies maximal overall growth of the compilation unit caused by inlining.
5140 This parameter is ignored when @option{-funit-at-a-time} is not used.
5141 The default value is 150.
5143 @item max-inline-insns-recursive
5144 @itemx max-inline-insns-recursive-auto
5145 Specifies maximum number of instructions out-of-line copy of self recursive inline
5146 function can grow into by performing recursive inlining.
5148 For functions declared inline @option{--param max-inline-insns-recursive} is
5149 taken into acount. For function not declared inline, recursive inlining
5150 happens only when @option{-finline-functions} (included in @option{-O3}) is
5151 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5152 default value is 500.
5154 @item max-inline-recursive-depth
5155 @itemx max-inline-recursive-depth-auto
5156 Specifies maximum recursion depth used by the recursive inlining.
5158 For functions declared inline @option{--param max-inline-recursive-depth} is
5159 taken into acount. For function not declared inline, recursive inlining
5160 happens only when @option{-finline-functions} (included in @option{-O3}) is
5161 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5162 default value is 500.
5164 @item max-inline-insns-rtl
5165 For languages that use the RTL inliner (this happens at a later stage
5166 than tree inlining), you can set the maximum allowable size (counted
5167 in RTL instructions) for the RTL inliner with this parameter.
5168 The default value is 600.
5170 @item max-unrolled-insns
5171 The maximum number of instructions that a loop should have if that loop
5172 is unrolled, and if the loop is unrolled, it determines how many times
5173 the loop code is unrolled.
5175 @item max-average-unrolled-insns
5176 The maximum number of instructions biased by probabilities of their execution
5177 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5178 it determines how many times the loop code is unrolled.
5180 @item max-unroll-times
5181 The maximum number of unrollings of a single loop.
5183 @item max-peeled-insns
5184 The maximum number of instructions that a loop should have if that loop
5185 is peeled, and if the loop is peeled, it determines how many times
5186 the loop code is peeled.
5188 @item max-peel-times
5189 The maximum number of peelings of a single loop.
5191 @item max-completely-peeled-insns
5192 The maximum number of insns of a completely peeled loop.
5194 @item max-completely-peel-times
5195 The maximum number of iterations of a loop to be suitable for complete peeling.
5197 @item max-unswitch-insns
5198 The maximum number of insns of an unswitched loop.
5200 @item max-unswitch-level
5201 The maximum number of branches unswitched in a single loop.
5204 The minimum cost of an expensive expression in the loop invariant motion.
5206 @item max-iterations-to-track
5208 The maximum number of iterations of a loop the brute force algorithm
5209 for analysis of # of iterations of the loop tries to evaluate.
5211 @item hot-bb-count-fraction
5212 Select fraction of the maximal count of repetitions of basic block in program
5213 given basic block needs to have to be considered hot.
5215 @item hot-bb-frequency-fraction
5216 Select fraction of the maximal frequency of executions of basic block in
5217 function given basic block needs to have to be considered hot
5219 @item tracer-dynamic-coverage
5220 @itemx tracer-dynamic-coverage-feedback
5222 This value is used to limit superblock formation once the given percentage of
5223 executed instructions is covered. This limits unnecessary code size
5226 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5227 feedback is available. The real profiles (as opposed to statically estimated
5228 ones) are much less balanced allowing the threshold to be larger value.
5230 @item tracer-max-code-growth
5231 Stop tail duplication once code growth has reached given percentage. This is
5232 rather hokey argument, as most of the duplicates will be eliminated later in
5233 cross jumping, so it may be set to much higher values than is the desired code
5236 @item tracer-min-branch-ratio
5238 Stop reverse growth when the reverse probability of best edge is less than this
5239 threshold (in percent).
5241 @item tracer-min-branch-ratio
5242 @itemx tracer-min-branch-ratio-feedback
5244 Stop forward growth if the best edge do have probability lower than this
5247 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
5248 compilation for profile feedback and one for compilation without. The value
5249 for compilation with profile feedback needs to be more conservative (higher) in
5250 order to make tracer effective.
5252 @item max-cse-path-length
5254 Maximum number of basic blocks on path that cse considers. The default is 10.
5256 @item global-var-threshold
5258 Counts the number of function calls (N) and the number of
5259 call-clobbered variables (V). If NxV is larger than this limit, a
5260 single artificial variable will be created to represent all the
5261 call-clobbered variables at function call sites. This artificial
5262 variable will then be made to alias every call-clobbered variable.
5263 (done as int * size_t on the host machine; beware overflow).
5265 @item max-aliased-vops
5267 Maxiumum number of virtual operands allowed to represent aliases
5268 before triggering the alias grouping heuristic. Alias grouping
5269 reduces compile times and memory consumption needed for aliasing at
5270 the expense of precision loss in alias information.
5272 @item ggc-min-expand
5274 GCC uses a garbage collector to manage its own memory allocation. This
5275 parameter specifies the minimum percentage by which the garbage
5276 collector's heap should be allowed to expand between collections.
5277 Tuning this may improve compilation speed; it has no effect on code
5280 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
5281 RAM >= 1GB. If @code{getrlimit} is available, the notion of "RAM" is
5282 the smallest of actual RAM, RLIMIT_RSS, RLIMIT_DATA and RLIMIT_AS. If
5283 GCC is not able to calculate RAM on a particular platform, the lower
5284 bound of 30% is used. Setting this parameter and
5285 @option{ggc-min-heapsize} to zero causes a full collection to occur at
5286 every opportunity. This is extremely slow, but can be useful for
5289 @item ggc-min-heapsize
5291 Minimum size of the garbage collector's heap before it begins bothering
5292 to collect garbage. The first collection occurs after the heap expands
5293 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
5294 tuning this may improve compilation speed, and has no effect on code
5297 The default is RAM/8, with a lower bound of 4096 (four megabytes) and an
5298 upper bound of 131072 (128 megabytes). If @code{getrlimit} is
5299 available, the notion of "RAM" is the smallest of actual RAM,
5300 RLIMIT_RSS, RLIMIT_DATA and RLIMIT_AS. If GCC is not able to calculate
5301 RAM on a particular platform, the lower bound is used. Setting this
5302 parameter very large effectively disables garbage collection. Setting
5303 this parameter and @option{ggc-min-expand} to zero causes a full
5304 collection to occur at every opportunity.
5306 @item max-reload-search-insns
5307 The maximum number of instruction reload should look backward for equivalent
5308 register. Increasing values mean more aggressive optimization, making the
5309 compile time increase with probably slightly better performance. The default
5312 @item max-cselib-memory-location
5313 The maximum number of memory locations cselib should take into acount.
5314 Increasing values mean more aggressive optimization, making the compile time
5315 increase with probably slightly better performance. The default value is 500.
5317 @item reorder-blocks-duplicate
5318 @itemx reorder-blocks-duplicate-feedback
5320 Used by basic block reordering pass to decide whether to use unconditional
5321 branch or duplicate the code on its destination. Code is duplicated when its
5322 estimated size is smaller than this value multiplied by the estimated size of
5323 unconditional jump in the hot spots of the program.
5325 The @option{reorder-block-duplicate-feedback} is used only when profile
5326 feedback is available and may be set to higher values than
5327 @option{reorder-block-duplicate} since information about the hot spots is more
5330 @item max-sched-region-blocks
5331 The maximum number of blocks in a region to be considered for
5332 interblock scheduling. The default value is 10.
5334 @item max-sched-region-insns
5335 The maximum number of insns in a region to be considered for
5336 interblock scheduling. The default value is 100.
5340 @node Preprocessor Options
5341 @section Options Controlling the Preprocessor
5342 @cindex preprocessor options
5343 @cindex options, preprocessor
5345 These options control the C preprocessor, which is run on each C source
5346 file before actual compilation.
5348 If you use the @option{-E} option, nothing is done except preprocessing.
5349 Some of these options make sense only together with @option{-E} because
5350 they cause the preprocessor output to be unsuitable for actual
5355 You can use @option{-Wp,@var{option}} to bypass the compiler driver
5356 and pass @var{option} directly through to the preprocessor. If
5357 @var{option} contains commas, it is split into multiple options at the
5358 commas. However, many options are modified, translated or interpreted
5359 by the compiler driver before being passed to the preprocessor, and
5360 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
5361 interface is undocumented and subject to change, so whenever possible
5362 you should avoid using @option{-Wp} and let the driver handle the
5365 @item -Xpreprocessor @var{option}
5366 @opindex preprocessor
5367 Pass @var{option} as an option to the preprocessor. You can use this to
5368 supply system-specific preprocessor options which GCC does not know how to
5371 If you want to pass an option that takes an argument, you must use
5372 @option{-Xpreprocessor} twice, once for the option and once for the argument.
5375 @include cppopts.texi
5377 @node Assembler Options
5378 @section Passing Options to the Assembler
5380 @c prevent bad page break with this line
5381 You can pass options to the assembler.
5384 @item -Wa,@var{option}
5386 Pass @var{option} as an option to the assembler. If @var{option}
5387 contains commas, it is split into multiple options at the commas.
5389 @item -Xassembler @var{option}
5391 Pass @var{option} as an option to the assembler. You can use this to
5392 supply system-specific assembler options which GCC does not know how to
5395 If you want to pass an option that takes an argument, you must use
5396 @option{-Xassembler} twice, once for the option and once for the argument.
5401 @section Options for Linking
5402 @cindex link options
5403 @cindex options, linking
5405 These options come into play when the compiler links object files into
5406 an executable output file. They are meaningless if the compiler is
5407 not doing a link step.
5411 @item @var{object-file-name}
5412 A file name that does not end in a special recognized suffix is
5413 considered to name an object file or library. (Object files are
5414 distinguished from libraries by the linker according to the file
5415 contents.) If linking is done, these object files are used as input
5424 If any of these options is used, then the linker is not run, and
5425 object file names should not be used as arguments. @xref{Overall
5429 @item -l@var{library}
5430 @itemx -l @var{library}
5432 Search the library named @var{library} when linking. (The second
5433 alternative with the library as a separate argument is only for
5434 POSIX compliance and is not recommended.)
5436 It makes a difference where in the command you write this option; the
5437 linker searches and processes libraries and object files in the order they
5438 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
5439 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
5440 to functions in @samp{z}, those functions may not be loaded.
5442 The linker searches a standard list of directories for the library,
5443 which is actually a file named @file{lib@var{library}.a}. The linker
5444 then uses this file as if it had been specified precisely by name.
5446 The directories searched include several standard system directories
5447 plus any that you specify with @option{-L}.
5449 Normally the files found this way are library files---archive files
5450 whose members are object files. The linker handles an archive file by
5451 scanning through it for members which define symbols that have so far
5452 been referenced but not defined. But if the file that is found is an
5453 ordinary object file, it is linked in the usual fashion. The only
5454 difference between using an @option{-l} option and specifying a file name
5455 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
5456 and searches several directories.
5460 You need this special case of the @option{-l} option in order to
5461 link an Objective-C program.
5464 @opindex nostartfiles
5465 Do not use the standard system startup files when linking.
5466 The standard system libraries are used normally, unless @option{-nostdlib}
5467 or @option{-nodefaultlibs} is used.
5469 @item -nodefaultlibs
5470 @opindex nodefaultlibs
5471 Do not use the standard system libraries when linking.
5472 Only the libraries you specify will be passed to the linker.
5473 The standard startup files are used normally, unless @option{-nostartfiles}
5474 is used. The compiler may generate calls to @code{memcmp},
5475 @code{memset}, @code{memcpy} and @code{memmove}.
5476 These entries are usually resolved by entries in
5477 libc. These entry points should be supplied through some other
5478 mechanism when this option is specified.
5482 Do not use the standard system startup files or libraries when linking.
5483 No startup files and only the libraries you specify will be passed to
5484 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
5485 @code{memcpy} and @code{memmove}.
5486 These entries are usually resolved by entries in
5487 libc. These entry points should be supplied through some other
5488 mechanism when this option is specified.
5490 @cindex @option{-lgcc}, use with @option{-nostdlib}
5491 @cindex @option{-nostdlib} and unresolved references
5492 @cindex unresolved references and @option{-nostdlib}
5493 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
5494 @cindex @option{-nodefaultlibs} and unresolved references
5495 @cindex unresolved references and @option{-nodefaultlibs}
5496 One of the standard libraries bypassed by @option{-nostdlib} and
5497 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
5498 that GCC uses to overcome shortcomings of particular machines, or special
5499 needs for some languages.
5500 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
5501 Collection (GCC) Internals},
5502 for more discussion of @file{libgcc.a}.)
5503 In most cases, you need @file{libgcc.a} even when you want to avoid
5504 other standard libraries. In other words, when you specify @option{-nostdlib}
5505 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
5506 This ensures that you have no unresolved references to internal GCC
5507 library subroutines. (For example, @samp{__main}, used to ensure C++
5508 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
5509 GNU Compiler Collection (GCC) Internals}.)
5513 Produce a position independent executable on targets which support it.
5514 For predictable results, you must also specify the same set of options
5515 that were used to generate code (@option{-fpie}, @option{-fPIE},
5516 or model suboptions) when you specify this option.
5520 Remove all symbol table and relocation information from the executable.
5524 On systems that support dynamic linking, this prevents linking with the shared
5525 libraries. On other systems, this option has no effect.
5529 Produce a shared object which can then be linked with other objects to
5530 form an executable. Not all systems support this option. For predictable
5531 results, you must also specify the same set of options that were used to
5532 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
5533 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
5534 needs to build supplementary stub code for constructors to work. On
5535 multi-libbed systems, @samp{gcc -shared} must select the correct support
5536 libraries to link against. Failing to supply the correct flags may lead
5537 to subtle defects. Supplying them in cases where they are not necessary
5540 @item -shared-libgcc
5541 @itemx -static-libgcc
5542 @opindex shared-libgcc
5543 @opindex static-libgcc
5544 On systems that provide @file{libgcc} as a shared library, these options
5545 force the use of either the shared or static version respectively.
5546 If no shared version of @file{libgcc} was built when the compiler was
5547 configured, these options have no effect.
5549 There are several situations in which an application should use the
5550 shared @file{libgcc} instead of the static version. The most common
5551 of these is when the application wishes to throw and catch exceptions
5552 across different shared libraries. In that case, each of the libraries
5553 as well as the application itself should use the shared @file{libgcc}.
5555 Therefore, the G++ and GCJ drivers automatically add
5556 @option{-shared-libgcc} whenever you build a shared library or a main
5557 executable, because C++ and Java programs typically use exceptions, so
5558 this is the right thing to do.
5560 If, instead, you use the GCC driver to create shared libraries, you may
5561 find that they will not always be linked with the shared @file{libgcc}.
5562 If GCC finds, at its configuration time, that you have a non-GNU linker
5563 or a GNU linker that does not support option @option{--eh-frame-hdr},
5564 it will link the shared version of @file{libgcc} into shared libraries
5565 by default. Otherwise, it will take advantage of the linker and optimize
5566 away the linking with the shared version of @file{libgcc}, linking with
5567 the static version of libgcc by default. This allows exceptions to
5568 propagate through such shared libraries, without incurring relocation
5569 costs at library load time.
5571 However, if a library or main executable is supposed to throw or catch
5572 exceptions, you must link it using the G++ or GCJ driver, as appropriate
5573 for the languages used in the program, or using the option
5574 @option{-shared-libgcc}, such that it is linked with the shared
5579 Bind references to global symbols when building a shared object. Warn
5580 about any unresolved references (unless overridden by the link editor
5581 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
5584 @item -Xlinker @var{option}
5586 Pass @var{option} as an option to the linker. You can use this to
5587 supply system-specific linker options which GCC does not know how to
5590 If you want to pass an option that takes an argument, you must use
5591 @option{-Xlinker} twice, once for the option and once for the argument.
5592 For example, to pass @option{-assert definitions}, you must write
5593 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
5594 @option{-Xlinker "-assert definitions"}, because this passes the entire
5595 string as a single argument, which is not what the linker expects.
5597 @item -Wl,@var{option}
5599 Pass @var{option} as an option to the linker. If @var{option} contains
5600 commas, it is split into multiple options at the commas.
5602 @item -u @var{symbol}
5604 Pretend the symbol @var{symbol} is undefined, to force linking of
5605 library modules to define it. You can use @option{-u} multiple times with
5606 different symbols to force loading of additional library modules.
5609 @node Directory Options
5610 @section Options for Directory Search
5611 @cindex directory options
5612 @cindex options, directory search
5615 These options specify directories to search for header files, for
5616 libraries and for parts of the compiler:
5621 Add the directory @var{dir} to the head of the list of directories to be
5622 searched for header files. This can be used to override a system header
5623 file, substituting your own version, since these directories are
5624 searched before the system header file directories. However, you should
5625 not use this option to add directories that contain vendor-supplied
5626 system header files (use @option{-isystem} for that). If you use more than
5627 one @option{-I} option, the directories are scanned in left-to-right
5628 order; the standard system directories come after.
5630 If a standard system include directory, or a directory specified with
5631 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
5632 option will be ignored. The directory will still be searched but as a
5633 system directory at its normal position in the system include chain.
5634 This is to ensure that GCC's procedure to fix buggy system headers and
5635 the ordering for the include_next directive are not inadvertently changed.
5636 If you really need to change the search order for system directories,
5637 use the @option{-nostdinc} and/or @option{-isystem} options.
5639 @item -iquote@var{dir}
5641 Add the directory @var{dir} to the head of the list of directories to
5642 be searched for header files only for the case of @samp{#include
5643 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
5644 otherwise just like @option{-I}.
5648 Add directory @var{dir} to the list of directories to be searched
5651 @item -B@var{prefix}
5653 This option specifies where to find the executables, libraries,
5654 include files, and data files of the compiler itself.
5656 The compiler driver program runs one or more of the subprograms
5657 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
5658 @var{prefix} as a prefix for each program it tries to run, both with and
5659 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
5661 For each subprogram to be run, the compiler driver first tries the
5662 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
5663 was not specified, the driver tries two standard prefixes, which are
5664 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
5665 those results in a file name that is found, the unmodified program
5666 name is searched for using the directories specified in your
5667 @env{PATH} environment variable.
5669 The compiler will check to see if the path provided by the @option{-B}
5670 refers to a directory, and if necessary it will add a directory
5671 separator character at the end of the path.
5673 @option{-B} prefixes that effectively specify directory names also apply
5674 to libraries in the linker, because the compiler translates these
5675 options into @option{-L} options for the linker. They also apply to
5676 includes files in the preprocessor, because the compiler translates these
5677 options into @option{-isystem} options for the preprocessor. In this case,
5678 the compiler appends @samp{include} to the prefix.
5680 The run-time support file @file{libgcc.a} can also be searched for using
5681 the @option{-B} prefix, if needed. If it is not found there, the two
5682 standard prefixes above are tried, and that is all. The file is left
5683 out of the link if it is not found by those means.
5685 Another way to specify a prefix much like the @option{-B} prefix is to use
5686 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
5689 As a special kludge, if the path provided by @option{-B} is
5690 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
5691 9, then it will be replaced by @file{[dir/]include}. This is to help
5692 with boot-strapping the compiler.
5694 @item -specs=@var{file}
5696 Process @var{file} after the compiler reads in the standard @file{specs}
5697 file, in order to override the defaults that the @file{gcc} driver
5698 program uses when determining what switches to pass to @file{cc1},
5699 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
5700 @option{-specs=@var{file}} can be specified on the command line, and they
5701 are processed in order, from left to right.
5705 This option has been deprecated. Please use @option{-iquote} instead for
5706 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
5707 Any directories you specify with @option{-I} options before the @option{-I-}
5708 option are searched only for the case of @samp{#include "@var{file}"};
5709 they are not searched for @samp{#include <@var{file}>}.
5711 If additional directories are specified with @option{-I} options after
5712 the @option{-I-}, these directories are searched for all @samp{#include}
5713 directives. (Ordinarily @emph{all} @option{-I} directories are used
5716 In addition, the @option{-I-} option inhibits the use of the current
5717 directory (where the current input file came from) as the first search
5718 directory for @samp{#include "@var{file}"}. There is no way to
5719 override this effect of @option{-I-}. With @option{-I.} you can specify
5720 searching the directory which was current when the compiler was
5721 invoked. That is not exactly the same as what the preprocessor does
5722 by default, but it is often satisfactory.
5724 @option{-I-} does not inhibit the use of the standard system directories
5725 for header files. Thus, @option{-I-} and @option{-nostdinc} are
5732 @section Specifying subprocesses and the switches to pass to them
5735 @command{gcc} is a driver program. It performs its job by invoking a
5736 sequence of other programs to do the work of compiling, assembling and
5737 linking. GCC interprets its command-line parameters and uses these to
5738 deduce which programs it should invoke, and which command-line options
5739 it ought to place on their command lines. This behavior is controlled
5740 by @dfn{spec strings}. In most cases there is one spec string for each
5741 program that GCC can invoke, but a few programs have multiple spec
5742 strings to control their behavior. The spec strings built into GCC can
5743 be overridden by using the @option{-specs=} command-line switch to specify
5746 @dfn{Spec files} are plaintext files that are used to construct spec
5747 strings. They consist of a sequence of directives separated by blank
5748 lines. The type of directive is determined by the first non-whitespace
5749 character on the line and it can be one of the following:
5752 @item %@var{command}
5753 Issues a @var{command} to the spec file processor. The commands that can
5757 @item %include <@var{file}>
5759 Search for @var{file} and insert its text at the current point in the
5762 @item %include_noerr <@var{file}>
5763 @cindex %include_noerr
5764 Just like @samp{%include}, but do not generate an error message if the include
5765 file cannot be found.
5767 @item %rename @var{old_name} @var{new_name}
5769 Rename the spec string @var{old_name} to @var{new_name}.
5773 @item *[@var{spec_name}]:
5774 This tells the compiler to create, override or delete the named spec
5775 string. All lines after this directive up to the next directive or
5776 blank line are considered to be the text for the spec string. If this
5777 results in an empty string then the spec will be deleted. (Or, if the
5778 spec did not exist, then nothing will happened.) Otherwise, if the spec
5779 does not currently exist a new spec will be created. If the spec does
5780 exist then its contents will be overridden by the text of this
5781 directive, unless the first character of that text is the @samp{+}
5782 character, in which case the text will be appended to the spec.
5784 @item [@var{suffix}]:
5785 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
5786 and up to the next directive or blank line are considered to make up the
5787 spec string for the indicated suffix. When the compiler encounters an
5788 input file with the named suffix, it will processes the spec string in
5789 order to work out how to compile that file. For example:
5796 This says that any input file whose name ends in @samp{.ZZ} should be
5797 passed to the program @samp{z-compile}, which should be invoked with the
5798 command-line switch @option{-input} and with the result of performing the
5799 @samp{%i} substitution. (See below.)
5801 As an alternative to providing a spec string, the text that follows a
5802 suffix directive can be one of the following:
5805 @item @@@var{language}
5806 This says that the suffix is an alias for a known @var{language}. This is
5807 similar to using the @option{-x} command-line switch to GCC to specify a
5808 language explicitly. For example:
5815 Says that .ZZ files are, in fact, C++ source files.
5818 This causes an error messages saying:
5821 @var{name} compiler not installed on this system.
5825 GCC already has an extensive list of suffixes built into it.
5826 This directive will add an entry to the end of the list of suffixes, but
5827 since the list is searched from the end backwards, it is effectively
5828 possible to override earlier entries using this technique.
5832 GCC has the following spec strings built into it. Spec files can
5833 override these strings or create their own. Note that individual
5834 targets can also add their own spec strings to this list.
5837 asm Options to pass to the assembler
5838 asm_final Options to pass to the assembler post-processor
5839 cpp Options to pass to the C preprocessor
5840 cc1 Options to pass to the C compiler
5841 cc1plus Options to pass to the C++ compiler
5842 endfile Object files to include at the end of the link
5843 link Options to pass to the linker
5844 lib Libraries to include on the command line to the linker
5845 libgcc Decides which GCC support library to pass to the linker
5846 linker Sets the name of the linker
5847 predefines Defines to be passed to the C preprocessor
5848 signed_char Defines to pass to CPP to say whether @code{char} is signed
5850 startfile Object files to include at the start of the link
5853 Here is a small example of a spec file:
5859 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
5862 This example renames the spec called @samp{lib} to @samp{old_lib} and
5863 then overrides the previous definition of @samp{lib} with a new one.
5864 The new definition adds in some extra command-line options before
5865 including the text of the old definition.
5867 @dfn{Spec strings} are a list of command-line options to be passed to their
5868 corresponding program. In addition, the spec strings can contain
5869 @samp{%}-prefixed sequences to substitute variable text or to
5870 conditionally insert text into the command line. Using these constructs
5871 it is possible to generate quite complex command lines.
5873 Here is a table of all defined @samp{%}-sequences for spec
5874 strings. Note that spaces are not generated automatically around the
5875 results of expanding these sequences. Therefore you can concatenate them
5876 together or combine them with constant text in a single argument.
5880 Substitute one @samp{%} into the program name or argument.
5883 Substitute the name of the input file being processed.
5886 Substitute the basename of the input file being processed.
5887 This is the substring up to (and not including) the last period
5888 and not including the directory.
5891 This is the same as @samp{%b}, but include the file suffix (text after
5895 Marks the argument containing or following the @samp{%d} as a
5896 temporary file name, so that that file will be deleted if GCC exits
5897 successfully. Unlike @samp{%g}, this contributes no text to the
5900 @item %g@var{suffix}
5901 Substitute a file name that has suffix @var{suffix} and is chosen
5902 once per compilation, and mark the argument in the same way as
5903 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
5904 name is now chosen in a way that is hard to predict even when previously
5905 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
5906 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
5907 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
5908 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
5909 was simply substituted with a file name chosen once per compilation,
5910 without regard to any appended suffix (which was therefore treated
5911 just like ordinary text), making such attacks more likely to succeed.
5913 @item %u@var{suffix}
5914 Like @samp{%g}, but generates a new temporary file name even if
5915 @samp{%u@var{suffix}} was already seen.
5917 @item %U@var{suffix}
5918 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
5919 new one if there is no such last file name. In the absence of any
5920 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
5921 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
5922 would involve the generation of two distinct file names, one
5923 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
5924 simply substituted with a file name chosen for the previous @samp{%u},
5925 without regard to any appended suffix.
5927 @item %j@var{suffix}
5928 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
5929 writable, and if save-temps is off; otherwise, substitute the name
5930 of a temporary file, just like @samp{%u}. This temporary file is not
5931 meant for communication between processes, but rather as a junk
5934 @item %|@var{suffix}
5935 @itemx %m@var{suffix}
5936 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
5937 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
5938 all. These are the two most common ways to instruct a program that it
5939 should read from standard input or write to standard output. If you
5940 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
5941 construct: see for example @file{f/lang-specs.h}.
5943 @item %.@var{SUFFIX}
5944 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
5945 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
5946 terminated by the next space or %.
5949 Marks the argument containing or following the @samp{%w} as the
5950 designated output file of this compilation. This puts the argument
5951 into the sequence of arguments that @samp{%o} will substitute later.
5954 Substitutes the names of all the output files, with spaces
5955 automatically placed around them. You should write spaces
5956 around the @samp{%o} as well or the results are undefined.
5957 @samp{%o} is for use in the specs for running the linker.
5958 Input files whose names have no recognized suffix are not compiled
5959 at all, but they are included among the output files, so they will
5963 Substitutes the suffix for object files. Note that this is
5964 handled specially when it immediately follows @samp{%g, %u, or %U},
5965 because of the need for those to form complete file names. The
5966 handling is such that @samp{%O} is treated exactly as if it had already
5967 been substituted, except that @samp{%g, %u, and %U} do not currently
5968 support additional @var{suffix} characters following @samp{%O} as they would
5969 following, for example, @samp{.o}.
5972 Substitutes the standard macro predefinitions for the
5973 current target machine. Use this when running @code{cpp}.
5976 Like @samp{%p}, but puts @samp{__} before and after the name of each
5977 predefined macro, except for macros that start with @samp{__} or with
5978 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
5982 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
5983 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
5984 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
5988 Current argument is the name of a library or startup file of some sort.
5989 Search for that file in a standard list of directories and substitute
5990 the full name found.
5993 Print @var{str} as an error message. @var{str} is terminated by a newline.
5994 Use this when inconsistent options are detected.
5997 Substitute the contents of spec string @var{name} at this point.
6000 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6002 @item %x@{@var{option}@}
6003 Accumulate an option for @samp{%X}.
6006 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6010 Output the accumulated assembler options specified by @option{-Wa}.
6013 Output the accumulated preprocessor options specified by @option{-Wp}.
6016 Process the @code{asm} spec. This is used to compute the
6017 switches to be passed to the assembler.
6020 Process the @code{asm_final} spec. This is a spec string for
6021 passing switches to an assembler post-processor, if such a program is
6025 Process the @code{link} spec. This is the spec for computing the
6026 command line passed to the linker. Typically it will make use of the
6027 @samp{%L %G %S %D and %E} sequences.
6030 Dump out a @option{-L} option for each directory that GCC believes might
6031 contain startup files. If the target supports multilibs then the
6032 current multilib directory will be prepended to each of these paths.
6035 Output the multilib directory with directory separators replaced with
6036 @samp{_}. If multilib directories are not set, or the multilib directory is
6037 @file{.} then this option emits nothing.
6040 Process the @code{lib} spec. This is a spec string for deciding which
6041 libraries should be included on the command line to the linker.
6044 Process the @code{libgcc} spec. This is a spec string for deciding
6045 which GCC support library should be included on the command line to the linker.
6048 Process the @code{startfile} spec. This is a spec for deciding which
6049 object files should be the first ones passed to the linker. Typically
6050 this might be a file named @file{crt0.o}.
6053 Process the @code{endfile} spec. This is a spec string that specifies
6054 the last object files that will be passed to the linker.
6057 Process the @code{cpp} spec. This is used to construct the arguments
6058 to be passed to the C preprocessor.
6061 Process the @code{cc1} spec. This is used to construct the options to be
6062 passed to the actual C compiler (@samp{cc1}).
6065 Process the @code{cc1plus} spec. This is used to construct the options to be
6066 passed to the actual C++ compiler (@samp{cc1plus}).
6069 Substitute the variable part of a matched option. See below.
6070 Note that each comma in the substituted string is replaced by
6074 Remove all occurrences of @code{-S} from the command line. Note---this
6075 command is position dependent. @samp{%} commands in the spec string
6076 before this one will see @code{-S}, @samp{%} commands in the spec string
6077 after this one will not.
6079 @item %:@var{function}(@var{args})
6080 Call the named function @var{function}, passing it @var{args}.
6081 @var{args} is first processed as a nested spec string, then split
6082 into an argument vector in the usual fashion. The function returns
6083 a string which is processed as if it had appeared literally as part
6084 of the current spec.
6086 The following built-in spec functions are provided:
6089 @item @code{if-exists}
6090 The @code{if-exists} spec function takes one argument, an absolute
6091 pathname to a file. If the file exists, @code{if-exists} returns the
6092 pathname. Here is a small example of its usage:
6096 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6099 @item @code{if-exists-else}
6100 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6101 spec function, except that it takes two arguments. The first argument is
6102 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6103 returns the pathname. If it does not exist, it returns the second argument.
6104 This way, @code{if-exists-else} can be used to select one file or another,
6105 based on the existence of the first. Here is a small example of its usage:
6109 crt0%O%s %:if-exists(crti%O%s) \
6110 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6115 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6116 If that switch was not specified, this substitutes nothing. Note that
6117 the leading dash is omitted when specifying this option, and it is
6118 automatically inserted if the substitution is performed. Thus the spec
6119 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6120 and would output the command line option @option{-foo}.
6122 @item %W@{@code{S}@}
6123 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6126 @item %@{@code{S}*@}
6127 Substitutes all the switches specified to GCC whose names start
6128 with @code{-S}, but which also take an argument. This is used for
6129 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6130 GCC considers @option{-o foo} as being
6131 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6132 text, including the space. Thus two arguments would be generated.
6134 @item %@{@code{S}*&@code{T}*@}
6135 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6136 (the order of @code{S} and @code{T} in the spec is not significant).
6137 There can be any number of ampersand-separated variables; for each the
6138 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6140 @item %@{@code{S}:@code{X}@}
6141 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6143 @item %@{!@code{S}:@code{X}@}
6144 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6146 @item %@{@code{S}*:@code{X}@}
6147 Substitutes @code{X} if one or more switches whose names start with
6148 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6149 once, no matter how many such switches appeared. However, if @code{%*}
6150 appears somewhere in @code{X}, then @code{X} will be substituted once
6151 for each matching switch, with the @code{%*} replaced by the part of
6152 that switch that matched the @code{*}.
6154 @item %@{.@code{S}:@code{X}@}
6155 Substitutes @code{X}, if processing a file with suffix @code{S}.
6157 @item %@{!.@code{S}:@code{X}@}
6158 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6160 @item %@{@code{S}|@code{P}:@code{X}@}
6161 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6162 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
6163 although they have a stronger binding than the @samp{|}. If @code{%*}
6164 appears in @code{X}, all of the alternatives must be starred, and only
6165 the first matching alternative is substituted.
6167 For example, a spec string like this:
6170 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
6173 will output the following command-line options from the following input
6174 command-line options:
6179 -d fred.c -foo -baz -boggle
6180 -d jim.d -bar -baz -boggle
6183 @item %@{S:X; T:Y; :D@}
6185 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
6186 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
6187 be as many clauses as you need. This may be combined with @code{.},
6188 @code{!}, @code{|}, and @code{*} as needed.
6193 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
6194 construct may contain other nested @samp{%} constructs or spaces, or
6195 even newlines. They are processed as usual, as described above.
6196 Trailing white space in @code{X} is ignored. White space may also
6197 appear anywhere on the left side of the colon in these constructs,
6198 except between @code{.} or @code{*} and the corresponding word.
6200 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
6201 handled specifically in these constructs. If another value of
6202 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
6203 @option{-W} switch is found later in the command line, the earlier
6204 switch value is ignored, except with @{@code{S}*@} where @code{S} is
6205 just one letter, which passes all matching options.
6207 The character @samp{|} at the beginning of the predicate text is used to
6208 indicate that a command should be piped to the following command, but
6209 only if @option{-pipe} is specified.
6211 It is built into GCC which switches take arguments and which do not.
6212 (You might think it would be useful to generalize this to allow each
6213 compiler's spec to say which switches take arguments. But this cannot
6214 be done in a consistent fashion. GCC cannot even decide which input
6215 files have been specified without knowing which switches take arguments,
6216 and it must know which input files to compile in order to tell which
6219 GCC also knows implicitly that arguments starting in @option{-l} are to be
6220 treated as compiler output files, and passed to the linker in their
6221 proper position among the other output files.
6223 @c man begin OPTIONS
6225 @node Target Options
6226 @section Specifying Target Machine and Compiler Version
6227 @cindex target options
6228 @cindex cross compiling
6229 @cindex specifying machine version
6230 @cindex specifying compiler version and target machine
6231 @cindex compiler version, specifying
6232 @cindex target machine, specifying
6234 The usual way to run GCC is to run the executable called @file{gcc}, or
6235 @file{<machine>-gcc} when cross-compiling, or
6236 @file{<machine>-gcc-<version>} to run a version other than the one that
6237 was installed last. Sometimes this is inconvenient, so GCC provides
6238 options that will switch to another cross-compiler or version.
6241 @item -b @var{machine}
6243 The argument @var{machine} specifies the target machine for compilation.
6245 The value to use for @var{machine} is the same as was specified as the
6246 machine type when configuring GCC as a cross-compiler. For
6247 example, if a cross-compiler was configured with @samp{configure
6248 i386v}, meaning to compile for an 80386 running System V, then you
6249 would specify @option{-b i386v} to run that cross compiler.
6251 @item -V @var{version}
6253 The argument @var{version} specifies which version of GCC to run.
6254 This is useful when multiple versions are installed. For example,
6255 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
6258 The @option{-V} and @option{-b} options work by running the
6259 @file{<machine>-gcc-<version>} executable, so there's no real reason to
6260 use them if you can just run that directly.
6262 @node Submodel Options
6263 @section Hardware Models and Configurations
6264 @cindex submodel options
6265 @cindex specifying hardware config
6266 @cindex hardware models and configurations, specifying
6267 @cindex machine dependent options
6269 Earlier we discussed the standard option @option{-b} which chooses among
6270 different installed compilers for completely different target
6271 machines, such as VAX vs.@: 68000 vs.@: 80386.
6273 In addition, each of these target machine types can have its own
6274 special options, starting with @samp{-m}, to choose among various
6275 hardware models or configurations---for example, 68010 vs 68020,
6276 floating coprocessor or none. A single installed version of the
6277 compiler can compile for any model or configuration, according to the
6280 Some configurations of the compiler also support additional special
6281 options, usually for compatibility with other compilers on the same
6284 These options are defined by the macro @code{TARGET_SWITCHES} in the
6285 machine description. The default for the options is also defined by
6286 that macro, which enables you to change the defaults.
6288 @c This list is ordered alphanumerically by subsection name.
6289 @c It should be the same order and spelling as these options are listed
6290 @c in Machine Dependent Options
6298 * DEC Alpha Options::
6299 * DEC Alpha/VMS Options::
6303 * i386 and x86-64 Options::
6315 * RS/6000 and PowerPC Options::
6316 * S/390 and zSeries Options::
6319 * System V Options::
6320 * TMS320C3x/C4x Options::
6324 * Xstormy16 Options::
6330 @subsection ARC Options
6333 These options are defined for ARC implementations:
6338 Compile code for little endian mode. This is the default.
6342 Compile code for big endian mode.
6345 @opindex mmangle-cpu
6346 Prepend the name of the cpu to all public symbol names.
6347 In multiple-processor systems, there are many ARC variants with different
6348 instruction and register set characteristics. This flag prevents code
6349 compiled for one cpu to be linked with code compiled for another.
6350 No facility exists for handling variants that are ``almost identical''.
6351 This is an all or nothing option.
6353 @item -mcpu=@var{cpu}
6355 Compile code for ARC variant @var{cpu}.
6356 Which variants are supported depend on the configuration.
6357 All variants support @option{-mcpu=base}, this is the default.
6359 @item -mtext=@var{text-section}
6360 @itemx -mdata=@var{data-section}
6361 @itemx -mrodata=@var{readonly-data-section}
6365 Put functions, data, and readonly data in @var{text-section},
6366 @var{data-section}, and @var{readonly-data-section} respectively
6367 by default. This can be overridden with the @code{section} attribute.
6368 @xref{Variable Attributes}.
6373 @subsection ARM Options
6376 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6380 @item -mabi=@var{name}
6382 Generate code for the specified ABI. Permissible values are: @samp{apcs-gnu},
6383 @samp{atpcs}, @samp{aapcs} and @samp{iwmmxt}.
6386 @opindex mapcs-frame
6387 Generate a stack frame that is compliant with the ARM Procedure Call
6388 Standard for all functions, even if this is not strictly necessary for
6389 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6390 with this option will cause the stack frames not to be generated for
6391 leaf functions. The default is @option{-mno-apcs-frame}.
6395 This is a synonym for @option{-mapcs-frame}.
6398 @c not currently implemented
6399 @item -mapcs-stack-check
6400 @opindex mapcs-stack-check
6401 Generate code to check the amount of stack space available upon entry to
6402 every function (that actually uses some stack space). If there is
6403 insufficient space available then either the function
6404 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
6405 called, depending upon the amount of stack space required. The run time
6406 system is required to provide these functions. The default is
6407 @option{-mno-apcs-stack-check}, since this produces smaller code.
6409 @c not currently implemented
6411 @opindex mapcs-float
6412 Pass floating point arguments using the float point registers. This is
6413 one of the variants of the APCS@. This option is recommended if the
6414 target hardware has a floating point unit or if a lot of floating point
6415 arithmetic is going to be performed by the code. The default is
6416 @option{-mno-apcs-float}, since integer only code is slightly increased in
6417 size if @option{-mapcs-float} is used.
6419 @c not currently implemented
6420 @item -mapcs-reentrant
6421 @opindex mapcs-reentrant
6422 Generate reentrant, position independent code. The default is
6423 @option{-mno-apcs-reentrant}.
6426 @item -mthumb-interwork
6427 @opindex mthumb-interwork
6428 Generate code which supports calling between the ARM and Thumb
6429 instruction sets. Without this option the two instruction sets cannot
6430 be reliably used inside one program. The default is
6431 @option{-mno-thumb-interwork}, since slightly larger code is generated
6432 when @option{-mthumb-interwork} is specified.
6434 @item -mno-sched-prolog
6435 @opindex mno-sched-prolog
6436 Prevent the reordering of instructions in the function prolog, or the
6437 merging of those instruction with the instructions in the function's
6438 body. This means that all functions will start with a recognizable set
6439 of instructions (or in fact one of a choice from a small set of
6440 different function prologues), and this information can be used to
6441 locate the start if functions inside an executable piece of code. The
6442 default is @option{-msched-prolog}.
6445 @opindex mhard-float
6446 Generate output containing floating point instructions. This is the
6450 @opindex msoft-float
6451 Generate output containing library calls for floating point.
6452 @strong{Warning:} the requisite libraries are not available for all ARM
6453 targets. Normally the facilities of the machine's usual C compiler are
6454 used, but this cannot be done directly in cross-compilation. You must make
6455 your own arrangements to provide suitable library functions for
6458 @option{-msoft-float} changes the calling convention in the output file;
6459 therefore, it is only useful if you compile @emph{all} of a program with
6460 this option. In particular, you need to compile @file{libgcc.a}, the
6461 library that comes with GCC, with @option{-msoft-float} in order for
6464 @item -mfloat-abi=@var{name}
6466 Specifies which ABI to use for floating point values. Permissible values
6467 are: @samp{soft}, @samp{softfp} and @samp{hard}.
6469 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
6470 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
6471 of floating point instructions, but still uses the soft-float calling
6474 @item -mlittle-endian
6475 @opindex mlittle-endian
6476 Generate code for a processor running in little-endian mode. This is
6477 the default for all standard configurations.
6480 @opindex mbig-endian
6481 Generate code for a processor running in big-endian mode; the default is
6482 to compile code for a little-endian processor.
6484 @item -mwords-little-endian
6485 @opindex mwords-little-endian
6486 This option only applies when generating code for big-endian processors.
6487 Generate code for a little-endian word order but a big-endian byte
6488 order. That is, a byte order of the form @samp{32107654}. Note: this
6489 option should only be used if you require compatibility with code for
6490 big-endian ARM processors generated by versions of the compiler prior to
6493 @item -mcpu=@var{name}
6495 This specifies the name of the target ARM processor. GCC uses this name
6496 to determine what kind of instructions it can emit when generating
6497 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
6498 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
6499 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
6500 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
6501 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
6502 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm8},
6503 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
6504 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
6505 @samp{arm920t}, @samp{arm922t}, @samp{arm946es}, @samp{arm966es},
6506 @samp{arm968es}, @samp{arm926ejs}, @samp{arm940t}, @samp{arm9tdmi},
6507 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ejs},
6508 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
6509 @samp{arm1136js}, @samp{arm1136jfs} ,@samp{xscale}, @samp{iwmmxt},
6512 @itemx -mtune=@var{name}
6514 This option is very similar to the @option{-mcpu=} option, except that
6515 instead of specifying the actual target processor type, and hence
6516 restricting which instructions can be used, it specifies that GCC should
6517 tune the performance of the code as if the target were of the type
6518 specified in this option, but still choosing the instructions that it
6519 will generate based on the cpu specified by a @option{-mcpu=} option.
6520 For some ARM implementations better performance can be obtained by using
6523 @item -march=@var{name}
6525 This specifies the name of the target ARM architecture. GCC uses this
6526 name to determine what kind of instructions it can emit when generating
6527 assembly code. This option can be used in conjunction with or instead
6528 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
6529 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
6530 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
6531 @samp{iwmmxt}, @samp{ep9312}.
6533 @item -mfpu=@var{name}
6534 @itemx -mfpe=@var{number}
6535 @itemx -mfp=@var{number}
6539 This specifies what floating point hardware (or hardware emulation) is
6540 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
6541 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
6542 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
6543 with older versions of GCC@.
6545 If @option{-msoft-float} is specified this specifies the format of
6546 floating point values.
6548 @item -mstructure-size-boundary=@var{n}
6549 @opindex mstructure-size-boundary
6550 The size of all structures and unions will be rounded up to a multiple
6551 of the number of bits set by this option. Permissible values are 8, 32
6552 and 64. The default value varies for different toolchains. For the COFF
6553 targeted toolchain the default value is 8. A value of 64 is only allowed
6554 if the underlying ABI supports it.
6556 Specifying the larger number can produce faster, more efficient code, but
6557 can also increase the size of the program. Different values are potentially
6558 incompatible. Code compiled with one value cannot necessarily expect to
6559 work with code or libraries compiled with another value, if they exchange
6560 information using structures or unions.
6562 @item -mabort-on-noreturn
6563 @opindex mabort-on-noreturn
6564 Generate a call to the function @code{abort} at the end of a
6565 @code{noreturn} function. It will be executed if the function tries to
6569 @itemx -mno-long-calls
6570 @opindex mlong-calls
6571 @opindex mno-long-calls
6572 Tells the compiler to perform function calls by first loading the
6573 address of the function into a register and then performing a subroutine
6574 call on this register. This switch is needed if the target function
6575 will lie outside of the 64 megabyte addressing range of the offset based
6576 version of subroutine call instruction.
6578 Even if this switch is enabled, not all function calls will be turned
6579 into long calls. The heuristic is that static functions, functions
6580 which have the @samp{short-call} attribute, functions that are inside
6581 the scope of a @samp{#pragma no_long_calls} directive and functions whose
6582 definitions have already been compiled within the current compilation
6583 unit, will not be turned into long calls. The exception to this rule is
6584 that weak function definitions, functions with the @samp{long-call}
6585 attribute or the @samp{section} attribute, and functions that are within
6586 the scope of a @samp{#pragma long_calls} directive, will always be
6587 turned into long calls.
6589 This feature is not enabled by default. Specifying
6590 @option{-mno-long-calls} will restore the default behavior, as will
6591 placing the function calls within the scope of a @samp{#pragma
6592 long_calls_off} directive. Note these switches have no effect on how
6593 the compiler generates code to handle function calls via function
6596 @item -mnop-fun-dllimport
6597 @opindex mnop-fun-dllimport
6598 Disable support for the @code{dllimport} attribute.
6600 @item -msingle-pic-base
6601 @opindex msingle-pic-base
6602 Treat the register used for PIC addressing as read-only, rather than
6603 loading it in the prologue for each function. The run-time system is
6604 responsible for initializing this register with an appropriate value
6605 before execution begins.
6607 @item -mpic-register=@var{reg}
6608 @opindex mpic-register
6609 Specify the register to be used for PIC addressing. The default is R10
6610 unless stack-checking is enabled, when R9 is used.
6612 @item -mcirrus-fix-invalid-insns
6613 @opindex mcirrus-fix-invalid-insns
6614 @opindex mno-cirrus-fix-invalid-insns
6615 Insert NOPs into the instruction stream to in order to work around
6616 problems with invalid Maverick instruction combinations. This option
6617 is only valid if the @option{-mcpu=ep9312} option has been used to
6618 enable generation of instructions for the Cirrus Maverick floating
6619 point co-processor. This option is not enabled by default, since the
6620 problem is only present in older Maverick implementations. The default
6621 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
6624 @item -mpoke-function-name
6625 @opindex mpoke-function-name
6626 Write the name of each function into the text section, directly
6627 preceding the function prologue. The generated code is similar to this:
6631 .ascii "arm_poke_function_name", 0
6634 .word 0xff000000 + (t1 - t0)
6635 arm_poke_function_name
6637 stmfd sp!, @{fp, ip, lr, pc@}
6641 When performing a stack backtrace, code can inspect the value of
6642 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
6643 location @code{pc - 12} and the top 8 bits are set, then we know that
6644 there is a function name embedded immediately preceding this location
6645 and has length @code{((pc[-3]) & 0xff000000)}.
6649 Generate code for the 16-bit Thumb instruction set. The default is to
6650 use the 32-bit ARM instruction set.
6653 @opindex mtpcs-frame
6654 Generate a stack frame that is compliant with the Thumb Procedure Call
6655 Standard for all non-leaf functions. (A leaf function is one that does
6656 not call any other functions.) The default is @option{-mno-tpcs-frame}.
6658 @item -mtpcs-leaf-frame
6659 @opindex mtpcs-leaf-frame
6660 Generate a stack frame that is compliant with the Thumb Procedure Call
6661 Standard for all leaf functions. (A leaf function is one that does
6662 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
6664 @item -mcallee-super-interworking
6665 @opindex mcallee-super-interworking
6666 Gives all externally visible functions in the file being compiled an ARM
6667 instruction set header which switches to Thumb mode before executing the
6668 rest of the function. This allows these functions to be called from
6669 non-interworking code.
6671 @item -mcaller-super-interworking
6672 @opindex mcaller-super-interworking
6673 Allows calls via function pointers (including virtual functions) to
6674 execute correctly regardless of whether the target code has been
6675 compiled for interworking or not. There is a small overhead in the cost
6676 of executing a function pointer if this option is enabled.
6681 @subsection AVR Options
6684 These options are defined for AVR implementations:
6687 @item -mmcu=@var{mcu}
6689 Specify ATMEL AVR instruction set or MCU type.
6691 Instruction set avr1 is for the minimal AVR core, not supported by the C
6692 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
6693 attiny11, attiny12, attiny15, attiny28).
6695 Instruction set avr2 (default) is for the classic AVR core with up to
6696 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
6697 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
6698 at90c8534, at90s8535).
6700 Instruction set avr3 is for the classic AVR core with up to 128K program
6701 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
6703 Instruction set avr4 is for the enhanced AVR core with up to 8K program
6704 memory space (MCU types: atmega8, atmega83, atmega85).
6706 Instruction set avr5 is for the enhanced AVR core with up to 128K program
6707 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
6708 atmega64, atmega128, at43usb355, at94k).
6712 Output instruction sizes to the asm file.
6714 @item -minit-stack=@var{N}
6715 @opindex minit-stack
6716 Specify the initial stack address, which may be a symbol or numeric value,
6717 @samp{__stack} is the default.
6719 @item -mno-interrupts
6720 @opindex mno-interrupts
6721 Generated code is not compatible with hardware interrupts.
6722 Code size will be smaller.
6724 @item -mcall-prologues
6725 @opindex mcall-prologues
6726 Functions prologues/epilogues expanded as call to appropriate
6727 subroutines. Code size will be smaller.
6729 @item -mno-tablejump
6730 @opindex mno-tablejump
6731 Do not generate tablejump insns which sometimes increase code size.
6734 @opindex mtiny-stack
6735 Change only the low 8 bits of the stack pointer.
6739 Assume int to be 8 bit integer. This affects the sizes of all types: A
6740 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
6741 and long long will be 4 bytes. Please note that this option does not
6742 comply to the C standards, but it will provide you with smaller code
6747 @subsection CRIS Options
6748 @cindex CRIS Options
6750 These options are defined specifically for the CRIS ports.
6753 @item -march=@var{architecture-type}
6754 @itemx -mcpu=@var{architecture-type}
6757 Generate code for the specified architecture. The choices for
6758 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
6759 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX.
6760 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
6763 @item -mtune=@var{architecture-type}
6765 Tune to @var{architecture-type} everything applicable about the generated
6766 code, except for the ABI and the set of available instructions. The
6767 choices for @var{architecture-type} are the same as for
6768 @option{-march=@var{architecture-type}}.
6770 @item -mmax-stack-frame=@var{n}
6771 @opindex mmax-stack-frame
6772 Warn when the stack frame of a function exceeds @var{n} bytes.
6774 @item -melinux-stacksize=@var{n}
6775 @opindex melinux-stacksize
6776 Only available with the @samp{cris-axis-aout} target. Arranges for
6777 indications in the program to the kernel loader that the stack of the
6778 program should be set to @var{n} bytes.
6784 The options @option{-metrax4} and @option{-metrax100} are synonyms for
6785 @option{-march=v3} and @option{-march=v8} respectively.
6787 @item -mmul-bug-workaround
6788 @itemx -mno-mul-bug-workaround
6789 @opindex mmul-bug-workaround
6790 @opindex mno-mul-bug-workaround
6791 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
6792 models where it applies. This option is active by default.
6796 Enable CRIS-specific verbose debug-related information in the assembly
6797 code. This option also has the effect to turn off the @samp{#NO_APP}
6798 formatted-code indicator to the assembler at the beginning of the
6803 Do not use condition-code results from previous instruction; always emit
6804 compare and test instructions before use of condition codes.
6806 @item -mno-side-effects
6807 @opindex mno-side-effects
6808 Do not emit instructions with side-effects in addressing modes other than
6812 @itemx -mno-stack-align
6814 @itemx -mno-data-align
6815 @itemx -mconst-align
6816 @itemx -mno-const-align
6817 @opindex mstack-align
6818 @opindex mno-stack-align
6819 @opindex mdata-align
6820 @opindex mno-data-align
6821 @opindex mconst-align
6822 @opindex mno-const-align
6823 These options (no-options) arranges (eliminate arrangements) for the
6824 stack-frame, individual data and constants to be aligned for the maximum
6825 single data access size for the chosen CPU model. The default is to
6826 arrange for 32-bit alignment. ABI details such as structure layout are
6827 not affected by these options.
6835 Similar to the stack- data- and const-align options above, these options
6836 arrange for stack-frame, writable data and constants to all be 32-bit,
6837 16-bit or 8-bit aligned. The default is 32-bit alignment.
6839 @item -mno-prologue-epilogue
6840 @itemx -mprologue-epilogue
6841 @opindex mno-prologue-epilogue
6842 @opindex mprologue-epilogue
6843 With @option{-mno-prologue-epilogue}, the normal function prologue and
6844 epilogue that sets up the stack-frame are omitted and no return
6845 instructions or return sequences are generated in the code. Use this
6846 option only together with visual inspection of the compiled code: no
6847 warnings or errors are generated when call-saved registers must be saved,
6848 or storage for local variable needs to be allocated.
6854 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
6855 instruction sequences that load addresses for functions from the PLT part
6856 of the GOT rather than (traditional on other architectures) calls to the
6857 PLT. The default is @option{-mgotplt}.
6861 Legacy no-op option only recognized with the cris-axis-aout target.
6865 Legacy no-op option only recognized with the cris-axis-elf and
6866 cris-axis-linux-gnu targets.
6870 Only recognized with the cris-axis-aout target, where it selects a
6871 GNU/linux-like multilib, include files and instruction set for
6876 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
6880 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
6881 to link with input-output functions from a simulator library. Code,
6882 initialized data and zero-initialized data are allocated consecutively.
6886 Like @option{-sim}, but pass linker options to locate initialized data at
6887 0x40000000 and zero-initialized data at 0x80000000.
6890 @node Darwin Options
6891 @subsection Darwin Options
6892 @cindex Darwin options
6894 These options are defined for all architectures running the Darwin operating
6895 system. They are useful for compatibility with other Mac OS compilers.
6900 Add the framework directory @var{dir} to the head of the list of
6901 directories to be searched for header files. These directories are
6902 interleaved with those specified by @option{-I} options and are
6903 scanned in a left-to-right order.
6905 A framework directory is a directory with frameworks in it. A
6906 framework is a directory with a @samp{"Headers"} and/or
6907 @samp{"PrivateHeaders"} directory contained directly in it that ends
6908 in @samp{".framework"}. The name of a framework is the name of this
6909 directory excluding the @samp{".framework"}. Headers associated with
6910 the framework are found in one of those two directories, with
6911 @samp{"Headers"} being searched first. A subframework is a framework
6912 directory that is in a framework's @samp{"Frameworks"} directory.
6913 Includes of subframework headers can only appear in a header of a
6914 framework that contains the subframework, or in a sibling subframework
6915 header. Two subframeworks are siblings if they occur in the same
6916 framework. A subframework should not have the same name as a
6917 framework, a warning will be issued if this is violated. Currently a
6918 subframework cannot have subframeworks, in the future, the mechanism
6919 may be extended to support this. The standard frameworks can be found
6920 in @samp{"/System/Library/Frameworks"} and
6921 @samp{"/Library/Frameworks"}. An example include looks like
6922 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
6923 the name of the framework and header.h is found in the
6924 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
6928 Emit debugging information for symbols that are used. For STABS
6929 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
6930 This is by default ON.
6934 Emit debugging information for all symbols and types.
6936 @item -mone-byte-bool
6937 @opindex -mone-byte-bool
6938 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
6939 By default @samp{sizeof(bool)} is @samp{4} when compiling for
6940 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
6941 option has no effect on x86.
6943 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
6944 to generate code that is not binary compatible with code generated
6945 without that switch. Using this switch may require recompiling all
6946 other modules in a program, including system libraries. Use this
6947 switch to conform to a non-default data model.
6951 Loads all members of static archive libraries.
6952 See man ld(1) for more information.
6954 @item -arch_errors_fatal
6955 @opindex arch_errors_fatal
6956 Cause the errors having to do with files that have the wrong architecture
6960 @opindex bind_at_load
6961 Causes the output file to be marked such that the dynamic linker will
6962 bind all undefined references when the file is loaded or launched.
6966 Produce a Mach-o bundle format file.
6967 See man ld(1) for more information.
6969 @item -bundle_loader @var{executable}
6970 @opindex bundle_loader
6971 This specifies the @var{executable} that will be loading the build
6972 output file being linked. See man ld(1) for more information.
6974 @item -allowable_client @var{client_name}
6978 @itemx -compatibility_version
6979 @itemx -current_version
6980 @itemx -dependency-file
6982 @itemx -dylinker_install_name
6985 @itemx -exported_symbols_list
6987 @itemx -flat_namespace
6988 @itemx -force_cpusubtype_ALL
6989 @itemx -force_flat_namespace
6990 @itemx -headerpad_max_install_names
6993 @itemx -install_name
6994 @itemx -keep_private_externs
6995 @itemx -multi_module
6996 @itemx -multiply_defined
6997 @itemx -multiply_defined_unused
6999 @itemx -nofixprebinding
7002 @itemx -noseglinkedit
7003 @itemx -pagezero_size
7005 @itemx -prebind_all_twolevel_modules
7006 @itemx -private_bundle
7007 @itemx -read_only_relocs
7009 @itemx -sectobjectsymbols
7013 @itemx -sectobjectsymbols
7015 @itemx -seg_addr_table
7016 @itemx -seg_addr_table_filename
7019 @itemx -segs_read_only_addr
7020 @itemx -segs_read_write_addr
7021 @itemx -single_module
7024 @itemx -sub_umbrella
7025 @itemx -twolevel_namespace
7028 @itemx -unexported_symbols_list
7029 @itemx -weak_reference_mismatches
7032 @opindex allowable_client
7034 @opindex client_name
7035 @opindex compatibility_version
7036 @opindex current_version
7037 @opindex dependency-file
7039 @opindex dylinker_install_name
7042 @opindex exported_symbols_list
7044 @opindex flat_namespace
7045 @opindex force_cpusubtype_ALL
7046 @opindex force_flat_namespace
7047 @opindex headerpad_max_install_names
7050 @opindex install_name
7051 @opindex keep_private_externs
7052 @opindex multi_module
7053 @opindex multiply_defined
7054 @opindex multiply_defined_unused
7056 @opindex nofixprebinding
7057 @opindex nomultidefs
7059 @opindex noseglinkedit
7060 @opindex pagezero_size
7062 @opindex prebind_all_twolevel_modules
7063 @opindex private_bundle
7064 @opindex read_only_relocs
7066 @opindex sectobjectsymbols
7070 @opindex sectobjectsymbols
7072 @opindex seg_addr_table
7073 @opindex seg_addr_table_filename
7074 @opindex seglinkedit
7076 @opindex segs_read_only_addr
7077 @opindex segs_read_write_addr
7078 @opindex single_module
7080 @opindex sub_library
7081 @opindex sub_umbrella
7082 @opindex twolevel_namespace
7085 @opindex unexported_symbols_list
7086 @opindex weak_reference_mismatches
7087 @opindex whatsloaded
7089 These options are available for Darwin linker. Darwin linker man page
7090 describes them in detail.
7093 @node DEC Alpha Options
7094 @subsection DEC Alpha Options
7096 These @samp{-m} options are defined for the DEC Alpha implementations:
7099 @item -mno-soft-float
7101 @opindex mno-soft-float
7102 @opindex msoft-float
7103 Use (do not use) the hardware floating-point instructions for
7104 floating-point operations. When @option{-msoft-float} is specified,
7105 functions in @file{libgcc.a} will be used to perform floating-point
7106 operations. Unless they are replaced by routines that emulate the
7107 floating-point operations, or compiled in such a way as to call such
7108 emulations routines, these routines will issue floating-point
7109 operations. If you are compiling for an Alpha without floating-point
7110 operations, you must ensure that the library is built so as not to call
7113 Note that Alpha implementations without floating-point operations are
7114 required to have floating-point registers.
7119 @opindex mno-fp-regs
7120 Generate code that uses (does not use) the floating-point register set.
7121 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
7122 register set is not used, floating point operands are passed in integer
7123 registers as if they were integers and floating-point results are passed
7124 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
7125 so any function with a floating-point argument or return value called by code
7126 compiled with @option{-mno-fp-regs} must also be compiled with that
7129 A typical use of this option is building a kernel that does not use,
7130 and hence need not save and restore, any floating-point registers.
7134 The Alpha architecture implements floating-point hardware optimized for
7135 maximum performance. It is mostly compliant with the IEEE floating
7136 point standard. However, for full compliance, software assistance is
7137 required. This option generates code fully IEEE compliant code
7138 @emph{except} that the @var{inexact-flag} is not maintained (see below).
7139 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
7140 defined during compilation. The resulting code is less efficient but is
7141 able to correctly support denormalized numbers and exceptional IEEE
7142 values such as not-a-number and plus/minus infinity. Other Alpha
7143 compilers call this option @option{-ieee_with_no_inexact}.
7145 @item -mieee-with-inexact
7146 @opindex mieee-with-inexact
7147 This is like @option{-mieee} except the generated code also maintains
7148 the IEEE @var{inexact-flag}. Turning on this option causes the
7149 generated code to implement fully-compliant IEEE math. In addition to
7150 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
7151 macro. On some Alpha implementations the resulting code may execute
7152 significantly slower than the code generated by default. Since there is
7153 very little code that depends on the @var{inexact-flag}, you should
7154 normally not specify this option. Other Alpha compilers call this
7155 option @option{-ieee_with_inexact}.
7157 @item -mfp-trap-mode=@var{trap-mode}
7158 @opindex mfp-trap-mode
7159 This option controls what floating-point related traps are enabled.
7160 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
7161 The trap mode can be set to one of four values:
7165 This is the default (normal) setting. The only traps that are enabled
7166 are the ones that cannot be disabled in software (e.g., division by zero
7170 In addition to the traps enabled by @samp{n}, underflow traps are enabled
7174 Like @samp{su}, but the instructions are marked to be safe for software
7175 completion (see Alpha architecture manual for details).
7178 Like @samp{su}, but inexact traps are enabled as well.
7181 @item -mfp-rounding-mode=@var{rounding-mode}
7182 @opindex mfp-rounding-mode
7183 Selects the IEEE rounding mode. Other Alpha compilers call this option
7184 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
7189 Normal IEEE rounding mode. Floating point numbers are rounded towards
7190 the nearest machine number or towards the even machine number in case
7194 Round towards minus infinity.
7197 Chopped rounding mode. Floating point numbers are rounded towards zero.
7200 Dynamic rounding mode. A field in the floating point control register
7201 (@var{fpcr}, see Alpha architecture reference manual) controls the
7202 rounding mode in effect. The C library initializes this register for
7203 rounding towards plus infinity. Thus, unless your program modifies the
7204 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
7207 @item -mtrap-precision=@var{trap-precision}
7208 @opindex mtrap-precision
7209 In the Alpha architecture, floating point traps are imprecise. This
7210 means without software assistance it is impossible to recover from a
7211 floating trap and program execution normally needs to be terminated.
7212 GCC can generate code that can assist operating system trap handlers
7213 in determining the exact location that caused a floating point trap.
7214 Depending on the requirements of an application, different levels of
7215 precisions can be selected:
7219 Program precision. This option is the default and means a trap handler
7220 can only identify which program caused a floating point exception.
7223 Function precision. The trap handler can determine the function that
7224 caused a floating point exception.
7227 Instruction precision. The trap handler can determine the exact
7228 instruction that caused a floating point exception.
7231 Other Alpha compilers provide the equivalent options called
7232 @option{-scope_safe} and @option{-resumption_safe}.
7234 @item -mieee-conformant
7235 @opindex mieee-conformant
7236 This option marks the generated code as IEEE conformant. You must not
7237 use this option unless you also specify @option{-mtrap-precision=i} and either
7238 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
7239 is to emit the line @samp{.eflag 48} in the function prologue of the
7240 generated assembly file. Under DEC Unix, this has the effect that
7241 IEEE-conformant math library routines will be linked in.
7243 @item -mbuild-constants
7244 @opindex mbuild-constants
7245 Normally GCC examines a 32- or 64-bit integer constant to
7246 see if it can construct it from smaller constants in two or three
7247 instructions. If it cannot, it will output the constant as a literal and
7248 generate code to load it from the data segment at runtime.
7250 Use this option to require GCC to construct @emph{all} integer constants
7251 using code, even if it takes more instructions (the maximum is six).
7253 You would typically use this option to build a shared library dynamic
7254 loader. Itself a shared library, it must relocate itself in memory
7255 before it can find the variables and constants in its own data segment.
7261 Select whether to generate code to be assembled by the vendor-supplied
7262 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
7280 Indicate whether GCC should generate code to use the optional BWX,
7281 CIX, FIX and MAX instruction sets. The default is to use the instruction
7282 sets supported by the CPU type specified via @option{-mcpu=} option or that
7283 of the CPU on which GCC was built if none was specified.
7288 @opindex mfloat-ieee
7289 Generate code that uses (does not use) VAX F and G floating point
7290 arithmetic instead of IEEE single and double precision.
7292 @item -mexplicit-relocs
7293 @itemx -mno-explicit-relocs
7294 @opindex mexplicit-relocs
7295 @opindex mno-explicit-relocs
7296 Older Alpha assemblers provided no way to generate symbol relocations
7297 except via assembler macros. Use of these macros does not allow
7298 optimal instruction scheduling. GNU binutils as of version 2.12
7299 supports a new syntax that allows the compiler to explicitly mark
7300 which relocations should apply to which instructions. This option
7301 is mostly useful for debugging, as GCC detects the capabilities of
7302 the assembler when it is built and sets the default accordingly.
7306 @opindex msmall-data
7307 @opindex mlarge-data
7308 When @option{-mexplicit-relocs} is in effect, static data is
7309 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
7310 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
7311 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
7312 16-bit relocations off of the @code{$gp} register. This limits the
7313 size of the small data area to 64KB, but allows the variables to be
7314 directly accessed via a single instruction.
7316 The default is @option{-mlarge-data}. With this option the data area
7317 is limited to just below 2GB. Programs that require more than 2GB of
7318 data must use @code{malloc} or @code{mmap} to allocate the data in the
7319 heap instead of in the program's data segment.
7321 When generating code for shared libraries, @option{-fpic} implies
7322 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
7326 @opindex msmall-text
7327 @opindex mlarge-text
7328 When @option{-msmall-text} is used, the compiler assumes that the
7329 code of the entire program (or shared library) fits in 4MB, and is
7330 thus reachable with a branch instruction. When @option{-msmall-data}
7331 is used, the compiler can assume that all local symbols share the
7332 same @code{$gp} value, and thus reduce the number of instructions
7333 required for a function call from 4 to 1.
7335 The default is @option{-mlarge-text}.
7337 @item -mcpu=@var{cpu_type}
7339 Set the instruction set and instruction scheduling parameters for
7340 machine type @var{cpu_type}. You can specify either the @samp{EV}
7341 style name or the corresponding chip number. GCC supports scheduling
7342 parameters for the EV4, EV5 and EV6 family of processors and will
7343 choose the default values for the instruction set from the processor
7344 you specify. If you do not specify a processor type, GCC will default
7345 to the processor on which the compiler was built.
7347 Supported values for @var{cpu_type} are
7353 Schedules as an EV4 and has no instruction set extensions.
7357 Schedules as an EV5 and has no instruction set extensions.
7361 Schedules as an EV5 and supports the BWX extension.
7366 Schedules as an EV5 and supports the BWX and MAX extensions.
7370 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
7374 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
7377 @item -mtune=@var{cpu_type}
7379 Set only the instruction scheduling parameters for machine type
7380 @var{cpu_type}. The instruction set is not changed.
7382 @item -mmemory-latency=@var{time}
7383 @opindex mmemory-latency
7384 Sets the latency the scheduler should assume for typical memory
7385 references as seen by the application. This number is highly
7386 dependent on the memory access patterns used by the application
7387 and the size of the external cache on the machine.
7389 Valid options for @var{time} are
7393 A decimal number representing clock cycles.
7399 The compiler contains estimates of the number of clock cycles for
7400 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
7401 (also called Dcache, Scache, and Bcache), as well as to main memory.
7402 Note that L3 is only valid for EV5.
7407 @node DEC Alpha/VMS Options
7408 @subsection DEC Alpha/VMS Options
7410 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
7413 @item -mvms-return-codes
7414 @opindex mvms-return-codes
7415 Return VMS condition codes from main. The default is to return POSIX
7416 style condition (e.g.@ error) codes.
7420 @subsection FRV Options
7427 Only use the first 32 general purpose registers.
7432 Use all 64 general purpose registers.
7437 Use only the first 32 floating point registers.
7442 Use all 64 floating point registers
7445 @opindex mhard-float
7447 Use hardware instructions for floating point operations.
7450 @opindex msoft-float
7452 Use library routines for floating point operations.
7457 Dynamically allocate condition code registers.
7462 Do not try to dynamically allocate condition code registers, only
7463 use @code{icc0} and @code{fcc0}.
7468 Change ABI to use double word insns.
7473 Do not use double word instructions.
7478 Use floating point double instructions.
7483 Do not use floating point double instructions.
7488 Use media instructions.
7493 Do not use media instructions.
7498 Use multiply and add/subtract instructions.
7503 Do not use multiply and add/subtract instructions.
7506 @opindex mlibrary-pic
7508 Generate position-independent EABI code.
7513 Use only the first four media accumulator registers.
7518 Use all eight media accumulator registers.
7523 Pack VLIW instructions.
7528 Do not pack VLIW instructions.
7533 Do not mark ABI switches in e_flags.
7538 Enable the use of conditional-move instructions (default).
7540 This switch is mainly for debugging the compiler and will likely be removed
7541 in a future version.
7543 @item -mno-cond-move
7544 @opindex mno-cond-move
7546 Disable the use of conditional-move instructions.
7548 This switch is mainly for debugging the compiler and will likely be removed
7549 in a future version.
7554 Enable the use of conditional set instructions (default).
7556 This switch is mainly for debugging the compiler and will likely be removed
7557 in a future version.
7562 Disable the use of conditional set instructions.
7564 This switch is mainly for debugging the compiler and will likely be removed
7565 in a future version.
7570 Enable the use of conditional execution (default).
7572 This switch is mainly for debugging the compiler and will likely be removed
7573 in a future version.
7575 @item -mno-cond-exec
7576 @opindex mno-cond-exec
7578 Disable the use of conditional execution.
7580 This switch is mainly for debugging the compiler and will likely be removed
7581 in a future version.
7584 @opindex mvliw-branch
7586 Run a pass to pack branches into VLIW instructions (default).
7588 This switch is mainly for debugging the compiler and will likely be removed
7589 in a future version.
7591 @item -mno-vliw-branch
7592 @opindex mno-vliw-branch
7594 Do not run a pass to pack branches into VLIW instructions.
7596 This switch is mainly for debugging the compiler and will likely be removed
7597 in a future version.
7599 @item -mmulti-cond-exec
7600 @opindex mmulti-cond-exec
7602 Enable optimization of @code{&&} and @code{||} in conditional execution
7605 This switch is mainly for debugging the compiler and will likely be removed
7606 in a future version.
7608 @item -mno-multi-cond-exec
7609 @opindex mno-multi-cond-exec
7611 Disable optimization of @code{&&} and @code{||} in conditional execution.
7613 This switch is mainly for debugging the compiler and will likely be removed
7614 in a future version.
7616 @item -mnested-cond-exec
7617 @opindex mnested-cond-exec
7619 Enable nested conditional execution optimizations (default).
7621 This switch is mainly for debugging the compiler and will likely be removed
7622 in a future version.
7624 @item -mno-nested-cond-exec
7625 @opindex mno-nested-cond-exec
7627 Disable nested conditional execution optimizations.
7629 This switch is mainly for debugging the compiler and will likely be removed
7630 in a future version.
7632 @item -mtomcat-stats
7633 @opindex mtomcat-stats
7635 Cause gas to print out tomcat statistics.
7637 @item -mcpu=@var{cpu}
7640 Select the processor type for which to generate code. Possible values are
7641 @samp{simple}, @samp{tomcat}, @samp{fr500}, @samp{fr400}, @samp{fr300},
7646 @node H8/300 Options
7647 @subsection H8/300 Options
7649 These @samp{-m} options are defined for the H8/300 implementations:
7654 Shorten some address references at link time, when possible; uses the
7655 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
7656 ld, Using ld}, for a fuller description.
7660 Generate code for the H8/300H@.
7664 Generate code for the H8S@.
7668 Generate code for the H8S and H8/300H in the normal mode. This switch
7669 must be used either with -mh or -ms.
7673 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
7677 Make @code{int} data 32 bits by default.
7681 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
7682 The default for the H8/300H and H8S is to align longs and floats on 4
7684 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
7685 This option has no effect on the H8/300.
7689 @subsection HPPA Options
7690 @cindex HPPA Options
7692 These @samp{-m} options are defined for the HPPA family of computers:
7695 @item -march=@var{architecture-type}
7697 Generate code for the specified architecture. The choices for
7698 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
7699 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
7700 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
7701 architecture option for your machine. Code compiled for lower numbered
7702 architectures will run on higher numbered architectures, but not the
7705 PA 2.0 support currently requires gas snapshot 19990413 or later. The
7706 next release of binutils (current is 2.9.1) will probably contain PA 2.0
7710 @itemx -mpa-risc-1-1
7711 @itemx -mpa-risc-2-0
7712 @opindex mpa-risc-1-0
7713 @opindex mpa-risc-1-1
7714 @opindex mpa-risc-2-0
7715 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
7718 @opindex mbig-switch
7719 Generate code suitable for big switch tables. Use this option only if
7720 the assembler/linker complain about out of range branches within a switch
7723 @item -mjump-in-delay
7724 @opindex mjump-in-delay
7725 Fill delay slots of function calls with unconditional jump instructions
7726 by modifying the return pointer for the function call to be the target
7727 of the conditional jump.
7729 @item -mdisable-fpregs
7730 @opindex mdisable-fpregs
7731 Prevent floating point registers from being used in any manner. This is
7732 necessary for compiling kernels which perform lazy context switching of
7733 floating point registers. If you use this option and attempt to perform
7734 floating point operations, the compiler will abort.
7736 @item -mdisable-indexing
7737 @opindex mdisable-indexing
7738 Prevent the compiler from using indexing address modes. This avoids some
7739 rather obscure problems when compiling MIG generated code under MACH@.
7741 @item -mno-space-regs
7742 @opindex mno-space-regs
7743 Generate code that assumes the target has no space registers. This allows
7744 GCC to generate faster indirect calls and use unscaled index address modes.
7746 Such code is suitable for level 0 PA systems and kernels.
7748 @item -mfast-indirect-calls
7749 @opindex mfast-indirect-calls
7750 Generate code that assumes calls never cross space boundaries. This
7751 allows GCC to emit code which performs faster indirect calls.
7753 This option will not work in the presence of shared libraries or nested
7756 @item -mfixed-range=@var{register-range}
7757 @opindex mfixed-range
7758 Generate code treating the given register range as fixed registers.
7759 A fixed register is one that the register allocator can not use. This is
7760 useful when compiling kernel code. A register range is specified as
7761 two registers separated by a dash. Multiple register ranges can be
7762 specified separated by a comma.
7764 @item -mlong-load-store
7765 @opindex mlong-load-store
7766 Generate 3-instruction load and store sequences as sometimes required by
7767 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
7770 @item -mportable-runtime
7771 @opindex mportable-runtime
7772 Use the portable calling conventions proposed by HP for ELF systems.
7776 Enable the use of assembler directives only GAS understands.
7778 @item -mschedule=@var{cpu-type}
7780 Schedule code according to the constraints for the machine type
7781 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
7782 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
7783 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
7784 proper scheduling option for your machine. The default scheduling is
7788 @opindex mlinker-opt
7789 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
7790 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
7791 linkers in which they give bogus error messages when linking some programs.
7794 @opindex msoft-float
7795 Generate output containing library calls for floating point.
7796 @strong{Warning:} the requisite libraries are not available for all HPPA
7797 targets. Normally the facilities of the machine's usual C compiler are
7798 used, but this cannot be done directly in cross-compilation. You must make
7799 your own arrangements to provide suitable library functions for
7800 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
7801 does provide software floating point support.
7803 @option{-msoft-float} changes the calling convention in the output file;
7804 therefore, it is only useful if you compile @emph{all} of a program with
7805 this option. In particular, you need to compile @file{libgcc.a}, the
7806 library that comes with GCC, with @option{-msoft-float} in order for
7811 Generate the predefine, @code{_SIO}, for server IO. The default is
7812 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
7813 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO. These
7814 options are available under HP-UX and HI-UX.
7818 Use GNU ld specific options. This passes @option{-shared} to ld when
7819 building a shared library. It is the default when GCC is configured,
7820 explicitly or implicitly, with the GNU linker. This option does not
7821 have any affect on which ld is called, it only changes what parameters
7822 are passed to that ld. The ld that is called is determined by the
7823 @option{--with-ld} configure option, GCC's program search path, and
7824 finally by the user's @env{PATH}. The linker used by GCC can be printed
7825 using @samp{which `gcc -print-prog-name=ld`}.
7829 Use HP ld specific options. This passes @option{-b} to ld when building
7830 a shared library and passes @option{+Accept TypeMismatch} to ld on all
7831 links. It is the default when GCC is configured, explicitly or
7832 implicitly, with the HP linker. This option does not have any affect on
7833 which ld is called, it only changes what parameters are passed to that
7834 ld. The ld that is called is determined by the @option{--with-ld}
7835 configure option, GCC's program search path, and finally by the user's
7836 @env{PATH}. The linker used by GCC can be printed using @samp{which
7837 `gcc -print-prog-name=ld`}.
7842 Select the FDPIC ABI, that uses function descriptors to represent
7843 pointers to functions. Without any PIC/PIE-related options, it
7844 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
7845 assumes GOT entries and small data are within a 12-bit range from the
7846 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
7847 are computed with 32 bits.
7850 @opindex minline-plt
7852 Enable inlining of PLT entries in function calls to functions that are
7853 not known to bind locally. It has no effect without @option{-mfdpic}.
7854 It's enabled by default if optimizing for speed and compiling for
7855 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
7856 optimization option such as @option{-O3} or above is present in the
7862 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
7863 that is known to be in read-only sections. It's enabled by default,
7864 except for @option{-fpic} or @option{-fpie}: even though it may help
7865 make the global offset table smaller, it trades 1 instruction for 4.
7866 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
7867 one of which may be shared by multiple symbols, and it avoids the need
7868 for a GOT entry for the referenced symbol, so it's more likely to be a
7869 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
7871 @item -multilib-library-pic
7872 @opindex multilib-library-pic
7874 Link with the (library, not FD) pic libraries. It's implied by
7875 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
7876 @option{-fpic} without @option{-mfdpic}. You should never have to use
7882 Follow the EABI requirement of always creating a frame pointer whenever
7883 a stack frame is allocated. This option is enabled by default and can
7884 be disabled with @option{-mno-linked-fp}.
7887 @opindex mno-long-calls
7888 Generate code that uses long call sequences. This ensures that a call
7889 is always able to reach linker generated stubs. The default is to generate
7890 long calls only when the distance from the call site to the beginning
7891 of the function or translation unit, as the case may be, exceeds a
7892 predefined limit set by the branch type being used. The limits for
7893 normal calls are 7,600,000 and 240,000 bytes, respectively for the
7894 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
7897 Distances are measured from the beginning of functions when using the
7898 @option{-ffunction-sections} option, or when using the @option{-mgas}
7899 and @option{-mno-portable-runtime} options together under HP-UX with
7902 It is normally not desirable to use this option as it will degrade
7903 performance. However, it may be useful in large applications,
7904 particularly when partial linking is used to build the application.
7906 The types of long calls used depends on the capabilities of the
7907 assembler and linker, and the type of code being generated. The
7908 impact on systems that support long absolute calls, and long pic
7909 symbol-difference or pc-relative calls should be relatively small.
7910 However, an indirect call is used on 32-bit ELF systems in pic code
7911 and it is quite long.
7915 Suppress the generation of link options to search libdld.sl when the
7916 @option{-static} option is specified on HP-UX 10 and later.
7920 The HP-UX implementation of setlocale in libc has a dependency on
7921 libdld.sl. There isn't an archive version of libdld.sl. Thus,
7922 when the @option{-static} option is specified, special link options
7923 are needed to resolve this dependency.
7925 On HP-UX 10 and later, the GCC driver adds the necessary options to
7926 link with libdld.sl when the @option{-static} option is specified.
7927 This causes the resulting binary to be dynamic. On the 64-bit port,
7928 the linkers generate dynamic binaries by default in any case. The
7929 @option{-nolibdld} option can be used to prevent the GCC driver from
7930 adding these link options.
7934 Add support for multithreading with the @dfn{dce thread} library
7935 under HP-UX. This option sets flags for both the preprocessor and
7939 @node i386 and x86-64 Options
7940 @subsection Intel 386 and AMD x86-64 Options
7941 @cindex i386 Options
7942 @cindex x86-64 Options
7943 @cindex Intel 386 Options
7944 @cindex AMD x86-64 Options
7946 These @samp{-m} options are defined for the i386 and x86-64 family of
7950 @item -mtune=@var{cpu-type}
7952 Tune to @var{cpu-type} everything applicable about the generated code, except
7953 for the ABI and the set of available instructions. The choices for
7957 Original Intel's i386 CPU.
7959 Intel's i486 CPU. (No scheduling is implemented for this chip.)
7961 Intel Pentium CPU with no MMX support.
7963 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
7964 @item i686, pentiumpro
7965 Intel PentiumPro CPU.
7967 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
7968 @item pentium3, pentium3m
7969 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
7972 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
7973 support. Used by Centrino notebooks.
7974 @item pentium4, pentium4m
7975 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
7977 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
7980 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
7981 SSE2 and SSE3 instruction set support.
7983 AMD K6 CPU with MMX instruction set support.
7985 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
7986 @item athlon, athlon-tbird
7987 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
7989 @item athlon-4, athlon-xp, athlon-mp
7990 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
7991 instruction set support.
7992 @item k8, opteron, athlon64, athlon-fx
7993 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
7994 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
7996 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
7999 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
8000 instruction set support.
8002 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
8003 implemented for this chip.)
8005 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
8006 implemented for this chip.)
8009 While picking a specific @var{cpu-type} will schedule things appropriately
8010 for that particular chip, the compiler will not generate any code that
8011 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8014 @item -march=@var{cpu-type}
8016 Generate instructions for the machine type @var{cpu-type}. The choices
8017 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8018 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8020 @item -mcpu=@var{cpu-type}
8022 A deprecated synonym for @option{-mtune}.
8031 @opindex mpentiumpro
8032 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8033 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8034 These synonyms are deprecated.
8036 @item -mfpmath=@var{unit}
8038 Generate floating point arithmetics for selected unit @var{unit}. The choices
8043 Use the standard 387 floating point coprocessor present majority of chips and
8044 emulated otherwise. Code compiled with this option will run almost everywhere.
8045 The temporary results are computed in 80bit precision instead of precision
8046 specified by the type resulting in slightly different results compared to most
8047 of other chips. See @option{-ffloat-store} for more detailed description.
8049 This is the default choice for i386 compiler.
8052 Use scalar floating point instructions present in the SSE instruction set.
8053 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8054 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8055 instruction set supports only single precision arithmetics, thus the double and
8056 extended precision arithmetics is still done using 387. Later version, present
8057 only in Pentium4 and the future AMD x86-64 chips supports double precision
8060 For i387 you need to use @option{-march=@var{cpu-type}}, @option{-msse} or
8061 @option{-msse2} switches to enable SSE extensions and make this option
8062 effective. For x86-64 compiler, these extensions are enabled by default.
8064 The resulting code should be considerably faster in the majority of cases and avoid
8065 the numerical instability problems of 387 code, but may break some existing
8066 code that expects temporaries to be 80bit.
8068 This is the default choice for the x86-64 compiler.
8071 Attempt to utilize both instruction sets at once. This effectively double the
8072 amount of available registers and on chips with separate execution units for
8073 387 and SSE the execution resources too. Use this option with care, as it is
8074 still experimental, because the GCC register allocator does not model separate
8075 functional units well resulting in instable performance.
8078 @item -masm=@var{dialect}
8079 @opindex masm=@var{dialect}
8080 Output asm instructions using selected @var{dialect}. Supported choices are
8081 @samp{intel} or @samp{att} (the default one).
8086 @opindex mno-ieee-fp
8087 Control whether or not the compiler uses IEEE floating point
8088 comparisons. These handle correctly the case where the result of a
8089 comparison is unordered.
8092 @opindex msoft-float
8093 Generate output containing library calls for floating point.
8094 @strong{Warning:} the requisite libraries are not part of GCC@.
8095 Normally the facilities of the machine's usual C compiler are used, but
8096 this can't be done directly in cross-compilation. You must make your
8097 own arrangements to provide suitable library functions for
8100 On machines where a function returns floating point results in the 80387
8101 register stack, some floating point opcodes may be emitted even if
8102 @option{-msoft-float} is used.
8104 @item -mno-fp-ret-in-387
8105 @opindex mno-fp-ret-in-387
8106 Do not use the FPU registers for return values of functions.
8108 The usual calling convention has functions return values of types
8109 @code{float} and @code{double} in an FPU register, even if there
8110 is no FPU@. The idea is that the operating system should emulate
8113 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8114 in ordinary CPU registers instead.
8116 @item -mno-fancy-math-387
8117 @opindex mno-fancy-math-387
8118 Some 387 emulators do not support the @code{sin}, @code{cos} and
8119 @code{sqrt} instructions for the 387. Specify this option to avoid
8120 generating those instructions. This option is the default on FreeBSD,
8121 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8122 indicates that the target cpu will always have an FPU and so the
8123 instruction will not need emulation. As of revision 2.6.1, these
8124 instructions are not generated unless you also use the
8125 @option{-funsafe-math-optimizations} switch.
8127 @item -malign-double
8128 @itemx -mno-align-double
8129 @opindex malign-double
8130 @opindex mno-align-double
8131 Control whether GCC aligns @code{double}, @code{long double}, and
8132 @code{long long} variables on a two word boundary or a one word
8133 boundary. Aligning @code{double} variables on a two word boundary will
8134 produce code that runs somewhat faster on a @samp{Pentium} at the
8135 expense of more memory.
8137 @strong{Warning:} if you use the @option{-malign-double} switch,
8138 structures containing the above types will be aligned differently than
8139 the published application binary interface specifications for the 386
8140 and will not be binary compatible with structures in code compiled
8141 without that switch.
8143 @item -m96bit-long-double
8144 @itemx -m128bit-long-double
8145 @opindex m96bit-long-double
8146 @opindex m128bit-long-double
8147 These switches control the size of @code{long double} type. The i386
8148 application binary interface specifies the size to be 96 bits,
8149 so @option{-m96bit-long-double} is the default in 32 bit mode.
8151 Modern architectures (Pentium and newer) would prefer @code{long double}
8152 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8153 conforming to the ABI, this would not be possible. So specifying a
8154 @option{-m128bit-long-double} will align @code{long double}
8155 to a 16 byte boundary by padding the @code{long double} with an additional
8158 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8159 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8161 Notice that neither of these options enable any extra precision over the x87
8162 standard of 80 bits for a @code{long double}.
8164 @strong{Warning:} if you override the default value for your target ABI, the
8165 structures and arrays containing @code{long double} variables will change
8166 their size as well as function calling convention for function taking
8167 @code{long double} will be modified. Hence they will not be binary
8168 compatible with arrays or structures in code compiled without that switch.
8172 @itemx -mno-svr3-shlib
8173 @opindex msvr3-shlib
8174 @opindex mno-svr3-shlib
8175 Control whether GCC places uninitialized local variables into the
8176 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8177 into @code{bss}. These options are meaningful only on System V Release 3.
8181 Use a different function-calling convention, in which functions that
8182 take a fixed number of arguments return with the @code{ret} @var{num}
8183 instruction, which pops their arguments while returning. This saves one
8184 instruction in the caller since there is no need to pop the arguments
8187 You can specify that an individual function is called with this calling
8188 sequence with the function attribute @samp{stdcall}. You can also
8189 override the @option{-mrtd} option by using the function attribute
8190 @samp{cdecl}. @xref{Function Attributes}.
8192 @strong{Warning:} this calling convention is incompatible with the one
8193 normally used on Unix, so you cannot use it if you need to call
8194 libraries compiled with the Unix compiler.
8196 Also, you must provide function prototypes for all functions that
8197 take variable numbers of arguments (including @code{printf});
8198 otherwise incorrect code will be generated for calls to those
8201 In addition, seriously incorrect code will result if you call a
8202 function with too many arguments. (Normally, extra arguments are
8203 harmlessly ignored.)
8205 @item -mregparm=@var{num}
8207 Control how many registers are used to pass integer arguments. By
8208 default, no registers are used to pass arguments, and at most 3
8209 registers can be used. You can control this behavior for a specific
8210 function by using the function attribute @samp{regparm}.
8211 @xref{Function Attributes}.
8213 @strong{Warning:} if you use this switch, and
8214 @var{num} is nonzero, then you must build all modules with the same
8215 value, including any libraries. This includes the system libraries and
8218 @item -mpreferred-stack-boundary=@var{num}
8219 @opindex mpreferred-stack-boundary
8220 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8221 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8222 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8223 size (@option{-Os}), in which case the default is the minimum correct
8224 alignment (4 bytes for x86, and 8 bytes for x86-64).
8226 On Pentium and PentiumPro, @code{double} and @code{long double} values
8227 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8228 suffer significant run time performance penalties. On Pentium III, the
8229 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8230 penalties if it is not 16 byte aligned.
8232 To ensure proper alignment of this values on the stack, the stack boundary
8233 must be as aligned as that required by any value stored on the stack.
8234 Further, every function must be generated such that it keeps the stack
8235 aligned. Thus calling a function compiled with a higher preferred
8236 stack boundary from a function compiled with a lower preferred stack
8237 boundary will most likely misalign the stack. It is recommended that
8238 libraries that use callbacks always use the default setting.
8240 This extra alignment does consume extra stack space, and generally
8241 increases code size. Code that is sensitive to stack space usage, such
8242 as embedded systems and operating system kernels, may want to reduce the
8243 preferred alignment to @option{-mpreferred-stack-boundary=2}.
8261 These switches enable or disable the use of built-in functions that allow
8262 direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow extensions of the
8265 @xref{X86 Built-in Functions}, for details of the functions enabled
8266 and disabled by these switches.
8268 To have SSE/SSE2 instructions generated automatically from floating-point
8269 code, see @option{-mfpmath=sse}.
8272 @itemx -mno-push-args
8274 @opindex mno-push-args
8275 Use PUSH operations to store outgoing parameters. This method is shorter
8276 and usually equally fast as method using SUB/MOV operations and is enabled
8277 by default. In some cases disabling it may improve performance because of
8278 improved scheduling and reduced dependencies.
8280 @item -maccumulate-outgoing-args
8281 @opindex maccumulate-outgoing-args
8282 If enabled, the maximum amount of space required for outgoing arguments will be
8283 computed in the function prologue. This is faster on most modern CPUs
8284 because of reduced dependencies, improved scheduling and reduced stack usage
8285 when preferred stack boundary is not equal to 2. The drawback is a notable
8286 increase in code size. This switch implies @option{-mno-push-args}.
8290 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8291 on thread-safe exception handling must compile and link all code with the
8292 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8293 @option{-D_MT}; when linking, it links in a special thread helper library
8294 @option{-lmingwthrd} which cleans up per thread exception handling data.
8296 @item -mno-align-stringops
8297 @opindex mno-align-stringops
8298 Do not align destination of inlined string operations. This switch reduces
8299 code size and improves performance in case the destination is already aligned,
8300 but GCC doesn't know about it.
8302 @item -minline-all-stringops
8303 @opindex minline-all-stringops
8304 By default GCC inlines string operations only when destination is known to be
8305 aligned at least to 4 byte boundary. This enables more inlining, increase code
8306 size, but may improve performance of code that depends on fast memcpy, strlen
8307 and memset for short lengths.
8309 @item -momit-leaf-frame-pointer
8310 @opindex momit-leaf-frame-pointer
8311 Don't keep the frame pointer in a register for leaf functions. This
8312 avoids the instructions to save, set up and restore frame pointers and
8313 makes an extra register available in leaf functions. The option
8314 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8315 which might make debugging harder.
8317 @item -mtls-direct-seg-refs
8318 @itemx -mno-tls-direct-seg-refs
8319 @opindex mtls-direct-seg-refs
8320 Controls whether TLS variables may be accessed with offsets from the
8321 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
8322 or whether the thread base pointer must be added. Whether or not this
8323 is legal depends on the operating system, and whether it maps the
8324 segment to cover the entire TLS area.
8326 For systems that use GNU libc, the default is on.
8329 These @samp{-m} switches are supported in addition to the above
8330 on AMD x86-64 processors in 64-bit environments.
8337 Generate code for a 32-bit or 64-bit environment.
8338 The 32-bit environment sets int, long and pointer to 32 bits and
8339 generates code that runs on any i386 system.
8340 The 64-bit environment sets int to 32 bits and long and pointer
8341 to 64 bits and generates code for AMD's x86-64 architecture.
8344 @opindex no-red-zone
8345 Do not use a so called red zone for x86-64 code. The red zone is mandated
8346 by the x86-64 ABI, it is a 128-byte area beyond the location of the
8347 stack pointer that will not be modified by signal or interrupt handlers
8348 and therefore can be used for temporary data without adjusting the stack
8349 pointer. The flag @option{-mno-red-zone} disables this red zone.
8351 @item -mcmodel=small
8352 @opindex mcmodel=small
8353 Generate code for the small code model: the program and its symbols must
8354 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
8355 Programs can be statically or dynamically linked. This is the default
8358 @item -mcmodel=kernel
8359 @opindex mcmodel=kernel
8360 Generate code for the kernel code model. The kernel runs in the
8361 negative 2 GB of the address space.
8362 This model has to be used for Linux kernel code.
8364 @item -mcmodel=medium
8365 @opindex mcmodel=medium
8366 Generate code for the medium model: The program is linked in the lower 2
8367 GB of the address space but symbols can be located anywhere in the
8368 address space. Programs can be statically or dynamically linked, but
8369 building of shared libraries are not supported with the medium model.
8371 @item -mcmodel=large
8372 @opindex mcmodel=large
8373 Generate code for the large model: This model makes no assumptions
8374 about addresses and sizes of sections. Currently GCC does not implement
8379 @subsection IA-64 Options
8380 @cindex IA-64 Options
8382 These are the @samp{-m} options defined for the Intel IA-64 architecture.
8386 @opindex mbig-endian
8387 Generate code for a big endian target. This is the default for HP-UX@.
8389 @item -mlittle-endian
8390 @opindex mlittle-endian
8391 Generate code for a little endian target. This is the default for AIX5
8398 Generate (or don't) code for the GNU assembler. This is the default.
8399 @c Also, this is the default if the configure option @option{--with-gnu-as}
8406 Generate (or don't) code for the GNU linker. This is the default.
8407 @c Also, this is the default if the configure option @option{--with-gnu-ld}
8412 Generate code that does not use a global pointer register. The result
8413 is not position independent code, and violates the IA-64 ABI@.
8415 @item -mvolatile-asm-stop
8416 @itemx -mno-volatile-asm-stop
8417 @opindex mvolatile-asm-stop
8418 @opindex mno-volatile-asm-stop
8419 Generate (or don't) a stop bit immediately before and after volatile asm
8424 Generate code that works around Itanium B step errata.
8426 @item -mregister-names
8427 @itemx -mno-register-names
8428 @opindex mregister-names
8429 @opindex mno-register-names
8430 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
8431 the stacked registers. This may make assembler output more readable.
8437 Disable (or enable) optimizations that use the small data section. This may
8438 be useful for working around optimizer bugs.
8441 @opindex mconstant-gp
8442 Generate code that uses a single constant global pointer value. This is
8443 useful when compiling kernel code.
8447 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
8448 This is useful when compiling firmware code.
8450 @item -minline-float-divide-min-latency
8451 @opindex minline-float-divide-min-latency
8452 Generate code for inline divides of floating point values
8453 using the minimum latency algorithm.
8455 @item -minline-float-divide-max-throughput
8456 @opindex minline-float-divide-max-throughput
8457 Generate code for inline divides of floating point values
8458 using the maximum throughput algorithm.
8460 @item -minline-int-divide-min-latency
8461 @opindex minline-int-divide-min-latency
8462 Generate code for inline divides of integer values
8463 using the minimum latency algorithm.
8465 @item -minline-int-divide-max-throughput
8466 @opindex minline-int-divide-max-throughput
8467 Generate code for inline divides of integer values
8468 using the maximum throughput algorithm.
8470 @item -mno-dwarf2-asm
8472 @opindex mno-dwarf2-asm
8473 @opindex mdwarf2-asm
8474 Don't (or do) generate assembler code for the DWARF2 line number debugging
8475 info. This may be useful when not using the GNU assembler.
8477 @item -mfixed-range=@var{register-range}
8478 @opindex mfixed-range
8479 Generate code treating the given register range as fixed registers.
8480 A fixed register is one that the register allocator can not use. This is
8481 useful when compiling kernel code. A register range is specified as
8482 two registers separated by a dash. Multiple register ranges can be
8483 specified separated by a comma.
8485 @item -mearly-stop-bits
8486 @itemx -mno-early-stop-bits
8487 @opindex mearly-stop-bits
8488 @opindex mno-early-stop-bits
8489 Allow stop bits to be placed earlier than immediately preceding the
8490 instruction that triggered the stop bit. This can improve instruction
8491 scheduling, but does not always do so.
8494 @node M32R/D Options
8495 @subsection M32R/D Options
8496 @cindex M32R/D options
8498 These @option{-m} options are defined for Renesas M32R/D architectures:
8503 Generate code for the M32R/2@.
8507 Generate code for the M32R/X@.
8511 Generate code for the M32R@. This is the default.
8514 @opindex mmodel=small
8515 Assume all objects live in the lower 16MB of memory (so that their addresses
8516 can be loaded with the @code{ld24} instruction), and assume all subroutines
8517 are reachable with the @code{bl} instruction.
8518 This is the default.
8520 The addressability of a particular object can be set with the
8521 @code{model} attribute.
8523 @item -mmodel=medium
8524 @opindex mmodel=medium
8525 Assume objects may be anywhere in the 32-bit address space (the compiler
8526 will generate @code{seth/add3} instructions to load their addresses), and
8527 assume all subroutines are reachable with the @code{bl} instruction.
8530 @opindex mmodel=large
8531 Assume objects may be anywhere in the 32-bit address space (the compiler
8532 will generate @code{seth/add3} instructions to load their addresses), and
8533 assume subroutines may not be reachable with the @code{bl} instruction
8534 (the compiler will generate the much slower @code{seth/add3/jl}
8535 instruction sequence).
8538 @opindex msdata=none
8539 Disable use of the small data area. Variables will be put into
8540 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
8541 @code{section} attribute has been specified).
8542 This is the default.
8544 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
8545 Objects may be explicitly put in the small data area with the
8546 @code{section} attribute using one of these sections.
8549 @opindex msdata=sdata
8550 Put small global and static data in the small data area, but do not
8551 generate special code to reference them.
8555 Put small global and static data in the small data area, and generate
8556 special instructions to reference them.
8560 @cindex smaller data references
8561 Put global and static objects less than or equal to @var{num} bytes
8562 into the small data or bss sections instead of the normal data or bss
8563 sections. The default value of @var{num} is 8.
8564 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
8565 for this option to have any effect.
8567 All modules should be compiled with the same @option{-G @var{num}} value.
8568 Compiling with different values of @var{num} may or may not work; if it
8569 doesn't the linker will give an error message---incorrect code will not be
8574 Makes the M32R specific code in the compiler display some statistics
8575 that might help in debugging programs.
8578 @opindex malign-loops
8579 Align all loops to a 32-byte boundary.
8581 @item -mno-align-loops
8582 @opindex mno-align-loops
8583 Do not enforce a 32-byte alignment for loops. This is the default.
8585 @item -missue-rate=@var{number}
8586 @opindex missue-rate=@var{number}
8587 Issue @var{number} instructions per cycle. @var{number} can only be 1
8590 @item -mbranch-cost=@var{number}
8591 @opindex mbranch-cost=@var{number}
8592 @var{number} can only be 1 or 2. If it is 1 then branches will be
8593 preferred over conditional code, if it is 2, then the opposite will
8596 @item -mflush-trap=@var{number}
8597 @opindex mflush-trap=@var{number}
8598 Specifies the trap number to use to flush the cache. The default is
8599 12. Valid numbers are between 0 and 15 inclusive.
8601 @item -mno-flush-trap
8602 @opindex mno-flush-trap
8603 Specifies that the cache cannot be flushed by using a trap.
8605 @item -mflush-func=@var{name}
8606 @opindex mflush-func=@var{name}
8607 Specifies the name of the operating system function to call to flush
8608 the cache. The default is @emph{_flush_cache}, but a function call
8609 will only be used if a trap is not available.
8611 @item -mno-flush-func
8612 @opindex mno-flush-func
8613 Indicates that there is no OS function for flushing the cache.
8617 @node M680x0 Options
8618 @subsection M680x0 Options
8619 @cindex M680x0 options
8621 These are the @samp{-m} options defined for the 68000 series. The default
8622 values for these options depends on which style of 68000 was selected when
8623 the compiler was configured; the defaults for the most common choices are
8631 Generate output for a 68000. This is the default
8632 when the compiler is configured for 68000-based systems.
8634 Use this option for microcontrollers with a 68000 or EC000 core,
8635 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
8641 Generate output for a 68020. This is the default
8642 when the compiler is configured for 68020-based systems.
8646 Generate output containing 68881 instructions for floating point.
8647 This is the default for most 68020 systems unless @option{--nfp} was
8648 specified when the compiler was configured.
8652 Generate output for a 68030. This is the default when the compiler is
8653 configured for 68030-based systems.
8657 Generate output for a 68040. This is the default when the compiler is
8658 configured for 68040-based systems.
8660 This option inhibits the use of 68881/68882 instructions that have to be
8661 emulated by software on the 68040. Use this option if your 68040 does not
8662 have code to emulate those instructions.
8666 Generate output for a 68060. This is the default when the compiler is
8667 configured for 68060-based systems.
8669 This option inhibits the use of 68020 and 68881/68882 instructions that
8670 have to be emulated by software on the 68060. Use this option if your 68060
8671 does not have code to emulate those instructions.
8675 Generate output for a CPU32. This is the default
8676 when the compiler is configured for CPU32-based systems.
8678 Use this option for microcontrollers with a
8679 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
8680 68336, 68340, 68341, 68349 and 68360.
8684 Generate output for a 520X ``coldfire'' family cpu. This is the default
8685 when the compiler is configured for 520X-based systems.
8687 Use this option for microcontroller with a 5200 core, including
8688 the MCF5202, MCF5203, MCF5204 and MCF5202.
8693 Generate output for a 68040, without using any of the new instructions.
8694 This results in code which can run relatively efficiently on either a
8695 68020/68881 or a 68030 or a 68040. The generated code does use the
8696 68881 instructions that are emulated on the 68040.
8700 Generate output for a 68060, without using any of the new instructions.
8701 This results in code which can run relatively efficiently on either a
8702 68020/68881 or a 68030 or a 68040. The generated code does use the
8703 68881 instructions that are emulated on the 68060.
8706 @opindex msoft-float
8707 Generate output containing library calls for floating point.
8708 @strong{Warning:} the requisite libraries are not available for all m68k
8709 targets. Normally the facilities of the machine's usual C compiler are
8710 used, but this can't be done directly in cross-compilation. You must
8711 make your own arrangements to provide suitable library functions for
8712 cross-compilation. The embedded targets @samp{m68k-*-aout} and
8713 @samp{m68k-*-coff} do provide software floating point support.
8717 Consider type @code{int} to be 16 bits wide, like @code{short int}.
8718 Additionally, parameters passed on the stack are also aligned to a
8719 16-bit boundary even on targets whose API mandates promotion to 32-bit.
8722 @opindex mnobitfield
8723 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
8724 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
8728 Do use the bit-field instructions. The @option{-m68020} option implies
8729 @option{-mbitfield}. This is the default if you use a configuration
8730 designed for a 68020.
8734 Use a different function-calling convention, in which functions
8735 that take a fixed number of arguments return with the @code{rtd}
8736 instruction, which pops their arguments while returning. This
8737 saves one instruction in the caller since there is no need to pop
8738 the arguments there.
8740 This calling convention is incompatible with the one normally
8741 used on Unix, so you cannot use it if you need to call libraries
8742 compiled with the Unix compiler.
8744 Also, you must provide function prototypes for all functions that
8745 take variable numbers of arguments (including @code{printf});
8746 otherwise incorrect code will be generated for calls to those
8749 In addition, seriously incorrect code will result if you call a
8750 function with too many arguments. (Normally, extra arguments are
8751 harmlessly ignored.)
8753 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
8754 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
8757 @itemx -mno-align-int
8759 @opindex mno-align-int
8760 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
8761 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
8762 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
8763 Aligning variables on 32-bit boundaries produces code that runs somewhat
8764 faster on processors with 32-bit busses at the expense of more memory.
8766 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
8767 align structures containing the above types differently than
8768 most published application binary interface specifications for the m68k.
8772 Use the pc-relative addressing mode of the 68000 directly, instead of
8773 using a global offset table. At present, this option implies @option{-fpic},
8774 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
8775 not presently supported with @option{-mpcrel}, though this could be supported for
8776 68020 and higher processors.
8778 @item -mno-strict-align
8779 @itemx -mstrict-align
8780 @opindex mno-strict-align
8781 @opindex mstrict-align
8782 Do not (do) assume that unaligned memory references will be handled by
8786 Generate code that allows the data segment to be located in a different
8787 area of memory from the text segment. This allows for execute in place in
8788 an environment without virtual memory management. This option implies -fPIC.
8791 Generate code that assumes that the data segment follows the text segment.
8792 This is the default.
8794 @item -mid-shared-library
8795 Generate code that supports shared libraries via the library ID method.
8796 This allows for execute in place and shared libraries in an environment
8797 without virtual memory management. This option implies -fPIC.
8799 @item -mno-id-shared-library
8800 Generate code that doesn't assume ID based shared libraries are being used.
8801 This is the default.
8803 @item -mshared-library-id=n
8804 Specified the identification number of the ID based shared library being
8805 compiled. Specifying a value of 0 will generate more compact code, specifying
8806 other values will force the allocation of that number to the current
8807 library but is no more space or time efficient than omitting this option.
8811 @node M68hc1x Options
8812 @subsection M68hc1x Options
8813 @cindex M68hc1x options
8815 These are the @samp{-m} options defined for the 68hc11 and 68hc12
8816 microcontrollers. The default values for these options depends on
8817 which style of microcontroller was selected when the compiler was configured;
8818 the defaults for the most common choices are given below.
8825 Generate output for a 68HC11. This is the default
8826 when the compiler is configured for 68HC11-based systems.
8832 Generate output for a 68HC12. This is the default
8833 when the compiler is configured for 68HC12-based systems.
8839 Generate output for a 68HCS12.
8842 @opindex mauto-incdec
8843 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
8850 Enable the use of 68HC12 min and max instructions.
8853 @itemx -mno-long-calls
8854 @opindex mlong-calls
8855 @opindex mno-long-calls
8856 Treat all calls as being far away (near). If calls are assumed to be
8857 far away, the compiler will use the @code{call} instruction to
8858 call a function and the @code{rtc} instruction for returning.
8862 Consider type @code{int} to be 16 bits wide, like @code{short int}.
8864 @item -msoft-reg-count=@var{count}
8865 @opindex msoft-reg-count
8866 Specify the number of pseudo-soft registers which are used for the
8867 code generation. The maximum number is 32. Using more pseudo-soft
8868 register may or may not result in better code depending on the program.
8869 The default is 4 for 68HC11 and 2 for 68HC12.
8874 @subsection MCore Options
8875 @cindex MCore options
8877 These are the @samp{-m} options defined for the Motorola M*Core
8885 @opindex mno-hardlit
8886 Inline constants into the code stream if it can be done in two
8887 instructions or less.
8893 Use the divide instruction. (Enabled by default).
8895 @item -mrelax-immediate
8896 @itemx -mno-relax-immediate
8897 @opindex mrelax-immediate
8898 @opindex mno-relax-immediate
8899 Allow arbitrary sized immediates in bit operations.
8901 @item -mwide-bitfields
8902 @itemx -mno-wide-bitfields
8903 @opindex mwide-bitfields
8904 @opindex mno-wide-bitfields
8905 Always treat bit-fields as int-sized.
8907 @item -m4byte-functions
8908 @itemx -mno-4byte-functions
8909 @opindex m4byte-functions
8910 @opindex mno-4byte-functions
8911 Force all functions to be aligned to a four byte boundary.
8913 @item -mcallgraph-data
8914 @itemx -mno-callgraph-data
8915 @opindex mcallgraph-data
8916 @opindex mno-callgraph-data
8917 Emit callgraph information.
8920 @itemx -mno-slow-bytes
8921 @opindex mslow-bytes
8922 @opindex mno-slow-bytes
8923 Prefer word access when reading byte quantities.
8925 @item -mlittle-endian
8927 @opindex mlittle-endian
8928 @opindex mbig-endian
8929 Generate code for a little endian target.
8935 Generate code for the 210 processor.
8939 @subsection MIPS Options
8940 @cindex MIPS options
8946 Generate big-endian code.
8950 Generate little-endian code. This is the default for @samp{mips*el-*-*}
8953 @item -march=@var{arch}
8955 Generate code that will run on @var{arch}, which can be the name of a
8956 generic MIPS ISA, or the name of a particular processor.
8958 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
8959 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
8960 The processor names are:
8961 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
8963 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
8964 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
8968 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
8969 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
8970 The special value @samp{from-abi} selects the
8971 most compatible architecture for the selected ABI (that is,
8972 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
8974 In processor names, a final @samp{000} can be abbreviated as @samp{k}
8975 (for example, @samp{-march=r2k}). Prefixes are optional, and
8976 @samp{vr} may be written @samp{r}.
8978 GCC defines two macros based on the value of this option. The first
8979 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
8980 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
8981 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
8982 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
8983 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
8985 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
8986 above. In other words, it will have the full prefix and will not
8987 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
8988 the macro names the resolved architecture (either @samp{"mips1"} or
8989 @samp{"mips3"}). It names the default architecture when no
8990 @option{-march} option is given.
8992 @item -mtune=@var{arch}
8994 Optimize for @var{arch}. Among other things, this option controls
8995 the way instructions are scheduled, and the perceived cost of arithmetic
8996 operations. The list of @var{arch} values is the same as for
8999 When this option is not used, GCC will optimize for the processor
9000 specified by @option{-march}. By using @option{-march} and
9001 @option{-mtune} together, it is possible to generate code that will
9002 run on a family of processors, but optimize the code for one
9003 particular member of that family.
9005 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
9006 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
9007 @samp{-march} ones described above.
9011 Equivalent to @samp{-march=mips1}.
9015 Equivalent to @samp{-march=mips2}.
9019 Equivalent to @samp{-march=mips3}.
9023 Equivalent to @samp{-march=mips4}.
9027 Equivalent to @samp{-march=mips32}.
9031 Equivalent to @samp{-march=mips32r2}.
9035 Equivalent to @samp{-march=mips64}.
9041 Use (do not use) the MIPS16 ISA.
9053 Generate code for the given ABI@.
9055 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
9056 generates 64-bit code when you select a 64-bit architecture, but you
9057 can use @option{-mgp32} to get 32-bit code instead.
9059 For information about the O64 ABI, see
9060 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
9063 @itemx -mno-abicalls
9065 @opindex mno-abicalls
9066 Generate (do not generate) SVR4-style position-independent code.
9067 @option{-mabicalls} is the default for SVR4-based systems.
9073 Lift (do not lift) the usual restrictions on the size of the global
9076 GCC normally uses a single instruction to load values from the GOT.
9077 While this is relatively efficient, it will only work if the GOT
9078 is smaller than about 64k. Anything larger will cause the linker
9079 to report an error such as:
9081 @cindex relocation truncated to fit (MIPS)
9083 relocation truncated to fit: R_MIPS_GOT16 foobar
9086 If this happens, you should recompile your code with @option{-mxgot}.
9087 It should then work with very large GOTs, although it will also be
9088 less efficient, since it will take three instructions to fetch the
9089 value of a global symbol.
9091 Note that some linkers can create multiple GOTs. If you have such a
9092 linker, you should only need to use @option{-mxgot} when a single object
9093 file accesses more than 64k's worth of GOT entries. Very few do.
9095 These options have no effect unless GCC is generating position
9100 Assume that general-purpose registers are 32 bits wide.
9104 Assume that general-purpose registers are 64 bits wide.
9108 Assume that floating-point registers are 32 bits wide.
9112 Assume that floating-point registers are 64 bits wide.
9115 @opindex mhard-float
9116 Use floating-point coprocessor instructions.
9119 @opindex msoft-float
9120 Do not use floating-point coprocessor instructions. Implement
9121 floating-point calculations using library calls instead.
9123 @item -msingle-float
9124 @opindex msingle-float
9125 Assume that the floating-point coprocessor only supports single-precision
9128 @itemx -mdouble-float
9129 @opindex mdouble-float
9130 Assume that the floating-point coprocessor supports double-precision
9131 operations. This is the default.
9135 Force @code{int} and @code{long} types to be 64 bits wide. See
9136 @option{-mlong32} for an explanation of the default and the way
9137 that the pointer size is determined.
9141 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
9142 an explanation of the default and the way that the pointer size is
9147 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
9149 The default size of @code{int}s, @code{long}s and pointers depends on
9150 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
9151 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
9152 32-bit @code{long}s. Pointers are the same size as @code{long}s,
9153 or the same size as integer registers, whichever is smaller.
9157 @cindex smaller data references (MIPS)
9158 @cindex gp-relative references (MIPS)
9159 Put global and static items less than or equal to @var{num} bytes into
9160 the small data or bss section instead of the normal data or bss section.
9161 This allows the data to be accessed using a single instruction.
9163 All modules should be compiled with the same @option{-G @var{num}}
9166 @item -membedded-data
9167 @itemx -mno-embedded-data
9168 @opindex membedded-data
9169 @opindex mno-embedded-data
9170 Allocate variables to the read-only data section first if possible, then
9171 next in the small data section if possible, otherwise in data. This gives
9172 slightly slower code than the default, but reduces the amount of RAM required
9173 when executing, and thus may be preferred for some embedded systems.
9175 @item -muninit-const-in-rodata
9176 @itemx -mno-uninit-const-in-rodata
9177 @opindex muninit-const-in-rodata
9178 @opindex mno-uninit-const-in-rodata
9179 Put uninitialized @code{const} variables in the read-only data section.
9180 This option is only meaningful in conjunction with @option{-membedded-data}.
9182 @item -msplit-addresses
9183 @itemx -mno-split-addresses
9184 @opindex msplit-addresses
9185 @opindex mno-split-addresses
9186 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
9187 relocation operators. This option has been superceded by
9188 @option{-mexplicit-relocs} but is retained for backwards compatibility.
9190 @item -mexplicit-relocs
9191 @itemx -mno-explicit-relocs
9192 @opindex mexplicit-relocs
9193 @opindex mno-explicit-relocs
9194 Use (do not use) assembler relocation operators when dealing with symbolic
9195 addresses. The alternative, selected by @option{-mno-explicit-relocs},
9196 is to use assembler macros instead.
9198 @option{-mexplicit-relocs} is the default if GCC was configured
9199 to use an assembler that supports relocation operators.
9201 @item -mcheck-zero-division
9202 @itemx -mno-check-zero-division
9203 @opindex mcheck-zero-division
9204 @opindex mno-check-zero-division
9205 Trap (do not trap) on integer division by zero. The default is
9206 @option{-mcheck-zero-division}.
9212 Force (do not force) the use of @code{memcpy()} for non-trivial block
9213 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
9214 most constant-sized copies.
9217 @itemx -mno-long-calls
9218 @opindex mlong-calls
9219 @opindex mno-long-calls
9220 Disable (do not disable) use of the @code{jal} instruction. Calling
9221 functions using @code{jal} is more efficient but requires the caller
9222 and callee to be in the same 256 megabyte segment.
9224 This option has no effect on abicalls code. The default is
9225 @option{-mno-long-calls}.
9231 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
9232 instructions, as provided by the R4650 ISA.
9235 @itemx -mno-fused-madd
9236 @opindex mfused-madd
9237 @opindex mno-fused-madd
9238 Enable (disable) use of the floating point multiply-accumulate
9239 instructions, when they are available. The default is
9240 @option{-mfused-madd}.
9242 When multiply-accumulate instructions are used, the intermediate
9243 product is calculated to infinite precision and is not subject to
9244 the FCSR Flush to Zero bit. This may be undesirable in some
9249 Tell the MIPS assembler to not run its preprocessor over user
9250 assembler files (with a @samp{.s} suffix) when assembling them.
9253 @itemx -mno-fix-r4000
9255 @opindex mno-fix-r4000
9256 Work around certain R4000 CPU errata:
9259 A double-word or a variable shift may give an incorrect result if executed
9260 immediately after starting an integer division.
9262 A double-word or a variable shift may give an incorrect result if executed
9263 while an integer multiplication is in progress.
9265 An integer division may give an incorrect result if started in a delay slot
9266 of a taken branch or a jump.
9270 @itemx -mno-fix-r4400
9272 @opindex mno-fix-r4400
9273 Work around certain R4400 CPU errata:
9276 A double-word or a variable shift may give an incorrect result if executed
9277 immediately after starting an integer division.
9281 @itemx -mno-fix-vr4120
9282 @opindex mfix-vr4120
9283 Work around certain VR4120 errata:
9286 @code{dmultu} does not always produce the correct result.
9288 @code{div} and @code{ddiv} do not always produce the correct result if one
9289 of the operands is negative.
9291 The workarounds for the division errata rely on special functions in
9292 @file{libgcc.a}. At present, these functions are only provided by
9293 the @code{mips64vr*-elf} configurations.
9295 Other VR4120 errata require a nop to be inserted between certain pairs of
9296 instructions. These errata are handled by the assembler, not by GCC itself.
9301 Work around certain SB-1 CPU core errata.
9302 (This flag currently works around the SB-1 revision 2
9303 ``F1'' and ``F2'' floating point errata.)
9305 @item -mflush-func=@var{func}
9306 @itemx -mno-flush-func
9307 @opindex mflush-func
9308 Specifies the function to call to flush the I and D caches, or to not
9309 call any such function. If called, the function must take the same
9310 arguments as the common @code{_flush_func()}, that is, the address of the
9311 memory range for which the cache is being flushed, the size of the
9312 memory range, and the number 3 (to flush both caches). The default
9313 depends on the target GCC was configured for, but commonly is either
9314 @samp{_flush_func} or @samp{__cpu_flush}.
9316 @item -mbranch-likely
9317 @itemx -mno-branch-likely
9318 @opindex mbranch-likely
9319 @opindex mno-branch-likely
9320 Enable or disable use of Branch Likely instructions, regardless of the
9321 default for the selected architecture. By default, Branch Likely
9322 instructions may be generated if they are supported by the selected
9323 architecture. An exception is for the MIPS32 and MIPS64 architectures
9324 and processors which implement those architectures; for those, Branch
9325 Likely instructions will not be generated by default because the MIPS32
9326 and MIPS64 architectures specifically deprecate their use.
9328 @item -mfp-exceptions
9329 @itemx -mno-fp-exceptions
9330 @opindex mfp-exceptions
9331 Specifies whether FP exceptions are enabled. This affects how we schedule
9332 FP instructions for some processors. The default is that FP exceptions are
9335 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
9336 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
9339 @item -mvr4130-align
9340 @itemx -mno-vr4130-align
9341 @opindex mvr4130-align
9342 The VR4130 pipeline is two-way superscalar, but can only issue two
9343 instructions together if the first one is 8-byte aligned. When this
9344 option is enabled, GCC will align pairs of instructions that it
9345 thinks should execute in parallel.
9347 This option only has an effect when optimizing for the VR4130.
9348 It normally makes code faster, but at the expense of making it bigger.
9349 It is enabled by default at optimization level @option{-O3}.
9353 @subsection MMIX Options
9354 @cindex MMIX Options
9356 These options are defined for the MMIX:
9360 @itemx -mno-libfuncs
9362 @opindex mno-libfuncs
9363 Specify that intrinsic library functions are being compiled, passing all
9364 values in registers, no matter the size.
9369 @opindex mno-epsilon
9370 Generate floating-point comparison instructions that compare with respect
9371 to the @code{rE} epsilon register.
9373 @item -mabi=mmixware
9375 @opindex mabi-mmixware
9377 Generate code that passes function parameters and return values that (in
9378 the called function) are seen as registers @code{$0} and up, as opposed to
9379 the GNU ABI which uses global registers @code{$231} and up.
9382 @itemx -mno-zero-extend
9383 @opindex mzero-extend
9384 @opindex mno-zero-extend
9385 When reading data from memory in sizes shorter than 64 bits, use (do not
9386 use) zero-extending load instructions by default, rather than
9387 sign-extending ones.
9390 @itemx -mno-knuthdiv
9392 @opindex mno-knuthdiv
9393 Make the result of a division yielding a remainder have the same sign as
9394 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
9395 remainder follows the sign of the dividend. Both methods are
9396 arithmetically valid, the latter being almost exclusively used.
9398 @item -mtoplevel-symbols
9399 @itemx -mno-toplevel-symbols
9400 @opindex mtoplevel-symbols
9401 @opindex mno-toplevel-symbols
9402 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
9403 code can be used with the @code{PREFIX} assembly directive.
9407 Generate an executable in the ELF format, rather than the default
9408 @samp{mmo} format used by the @command{mmix} simulator.
9410 @item -mbranch-predict
9411 @itemx -mno-branch-predict
9412 @opindex mbranch-predict
9413 @opindex mno-branch-predict
9414 Use (do not use) the probable-branch instructions, when static branch
9415 prediction indicates a probable branch.
9417 @item -mbase-addresses
9418 @itemx -mno-base-addresses
9419 @opindex mbase-addresses
9420 @opindex mno-base-addresses
9421 Generate (do not generate) code that uses @emph{base addresses}. Using a
9422 base address automatically generates a request (handled by the assembler
9423 and the linker) for a constant to be set up in a global register. The
9424 register is used for one or more base address requests within the range 0
9425 to 255 from the value held in the register. The generally leads to short
9426 and fast code, but the number of different data items that can be
9427 addressed is limited. This means that a program that uses lots of static
9428 data may require @option{-mno-base-addresses}.
9431 @itemx -mno-single-exit
9432 @opindex msingle-exit
9433 @opindex mno-single-exit
9434 Force (do not force) generated code to have a single exit point in each
9438 @node MN10300 Options
9439 @subsection MN10300 Options
9440 @cindex MN10300 options
9442 These @option{-m} options are defined for Matsushita MN10300 architectures:
9447 Generate code to avoid bugs in the multiply instructions for the MN10300
9448 processors. This is the default.
9451 @opindex mno-mult-bug
9452 Do not generate code to avoid bugs in the multiply instructions for the
9457 Generate code which uses features specific to the AM33 processor.
9461 Do not generate code which uses features specific to the AM33 processor. This
9466 Do not link in the C run-time initialization object file.
9470 Indicate to the linker that it should perform a relaxation optimization pass
9471 to shorten branches, calls and absolute memory addresses. This option only
9472 has an effect when used on the command line for the final link step.
9474 This option makes symbolic debugging impossible.
9478 @subsection NS32K Options
9479 @cindex NS32K options
9481 These are the @samp{-m} options defined for the 32000 series. The default
9482 values for these options depends on which style of 32000 was selected when
9483 the compiler was configured; the defaults for the most common choices are
9491 Generate output for a 32032. This is the default
9492 when the compiler is configured for 32032 and 32016 based systems.
9498 Generate output for a 32332. This is the default
9499 when the compiler is configured for 32332-based systems.
9505 Generate output for a 32532. This is the default
9506 when the compiler is configured for 32532-based systems.
9510 Generate output containing 32081 instructions for floating point.
9511 This is the default for all systems.
9515 Generate output containing 32381 instructions for floating point. This
9516 also implies @option{-m32081}. The 32381 is only compatible with the 32332
9517 and 32532 cpus. This is the default for the pc532-netbsd configuration.
9521 Try and generate multiply-add floating point instructions @code{polyF}
9522 and @code{dotF}. This option is only available if the @option{-m32381}
9523 option is in effect. Using these instructions requires changes to
9524 register allocation which generally has a negative impact on
9525 performance. This option should only be enabled when compiling code
9526 particularly likely to make heavy use of multiply-add instructions.
9529 @opindex mnomulti-add
9530 Do not try and generate multiply-add floating point instructions
9531 @code{polyF} and @code{dotF}. This is the default on all platforms.
9534 @opindex msoft-float
9535 Generate output containing library calls for floating point.
9536 @strong{Warning:} the requisite libraries may not be available.
9538 @item -mieee-compare
9539 @itemx -mno-ieee-compare
9540 @opindex mieee-compare
9541 @opindex mno-ieee-compare
9542 Control whether or not the compiler uses IEEE floating point
9543 comparisons. These handle correctly the case where the result of a
9544 comparison is unordered.
9545 @strong{Warning:} the requisite kernel support may not be available.
9548 @opindex mnobitfield
9549 Do not use the bit-field instructions. On some machines it is faster to
9550 use shifting and masking operations. This is the default for the pc532.
9554 Do use the bit-field instructions. This is the default for all platforms
9559 Use a different function-calling convention, in which functions
9560 that take a fixed number of arguments return pop their
9561 arguments on return with the @code{ret} instruction.
9563 This calling convention is incompatible with the one normally
9564 used on Unix, so you cannot use it if you need to call libraries
9565 compiled with the Unix compiler.
9567 Also, you must provide function prototypes for all functions that
9568 take variable numbers of arguments (including @code{printf});
9569 otherwise incorrect code will be generated for calls to those
9572 In addition, seriously incorrect code will result if you call a
9573 function with too many arguments. (Normally, extra arguments are
9574 harmlessly ignored.)
9576 This option takes its name from the 680x0 @code{rtd} instruction.
9581 Use a different function-calling convention where the first two arguments
9582 are passed in registers.
9584 This calling convention is incompatible with the one normally
9585 used on Unix, so you cannot use it if you need to call libraries
9586 compiled with the Unix compiler.
9589 @opindex mnoregparam
9590 Do not pass any arguments in registers. This is the default for all
9595 It is OK to use the sb as an index register which is always loaded with
9596 zero. This is the default for the pc532-netbsd target.
9600 The sb register is not available for use or has not been initialized to
9601 zero by the run time system. This is the default for all targets except
9602 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
9603 @option{-fpic} is set.
9607 Many ns32000 series addressing modes use displacements of up to 512MB@.
9608 If an address is above 512MB then displacements from zero can not be used.
9609 This option causes code to be generated which can be loaded above 512MB@.
9610 This may be useful for operating systems or ROM code.
9614 Assume code will be loaded in the first 512MB of virtual address space.
9615 This is the default for all platforms.
9619 @node PDP-11 Options
9620 @subsection PDP-11 Options
9621 @cindex PDP-11 Options
9623 These options are defined for the PDP-11:
9628 Use hardware FPP floating point. This is the default. (FIS floating
9629 point on the PDP-11/40 is not supported.)
9632 @opindex msoft-float
9633 Do not use hardware floating point.
9637 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
9641 Return floating-point results in memory. This is the default.
9645 Generate code for a PDP-11/40.
9649 Generate code for a PDP-11/45. This is the default.
9653 Generate code for a PDP-11/10.
9655 @item -mbcopy-builtin
9656 @opindex bcopy-builtin
9657 Use inline @code{movmemhi} patterns for copying memory. This is the
9662 Do not use inline @code{movmemhi} patterns for copying memory.
9668 Use 16-bit @code{int}. This is the default.
9674 Use 32-bit @code{int}.
9679 @opindex mno-float32
9680 Use 64-bit @code{float}. This is the default.
9685 @opindex mno-float64
9686 Use 32-bit @code{float}.
9690 Use @code{abshi2} pattern. This is the default.
9694 Do not use @code{abshi2} pattern.
9696 @item -mbranch-expensive
9697 @opindex mbranch-expensive
9698 Pretend that branches are expensive. This is for experimenting with
9699 code generation only.
9701 @item -mbranch-cheap
9702 @opindex mbranch-cheap
9703 Do not pretend that branches are expensive. This is the default.
9707 Generate code for a system with split I&D.
9711 Generate code for a system without split I&D. This is the default.
9715 Use Unix assembler syntax. This is the default when configured for
9720 Use DEC assembler syntax. This is the default when configured for any
9721 PDP-11 target other than @samp{pdp11-*-bsd}.
9724 @node PowerPC Options
9725 @subsection PowerPC Options
9726 @cindex PowerPC options
9728 These are listed under @xref{RS/6000 and PowerPC Options}.
9730 @node RS/6000 and PowerPC Options
9731 @subsection IBM RS/6000 and PowerPC Options
9732 @cindex RS/6000 and PowerPC Options
9733 @cindex IBM RS/6000 and PowerPC Options
9735 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
9743 @itemx -mpowerpc-gpopt
9744 @itemx -mno-powerpc-gpopt
9745 @itemx -mpowerpc-gfxopt
9746 @itemx -mno-powerpc-gfxopt
9748 @itemx -mno-powerpc64
9754 @opindex mno-powerpc
9755 @opindex mpowerpc-gpopt
9756 @opindex mno-powerpc-gpopt
9757 @opindex mpowerpc-gfxopt
9758 @opindex mno-powerpc-gfxopt
9760 @opindex mno-powerpc64
9761 GCC supports two related instruction set architectures for the
9762 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
9763 instructions supported by the @samp{rios} chip set used in the original
9764 RS/6000 systems and the @dfn{PowerPC} instruction set is the
9765 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
9766 the IBM 4xx microprocessors.
9768 Neither architecture is a subset of the other. However there is a
9769 large common subset of instructions supported by both. An MQ
9770 register is included in processors supporting the POWER architecture.
9772 You use these options to specify which instructions are available on the
9773 processor you are using. The default value of these options is
9774 determined when configuring GCC@. Specifying the
9775 @option{-mcpu=@var{cpu_type}} overrides the specification of these
9776 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
9777 rather than the options listed above.
9779 The @option{-mpower} option allows GCC to generate instructions that
9780 are found only in the POWER architecture and to use the MQ register.
9781 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
9782 to generate instructions that are present in the POWER2 architecture but
9783 not the original POWER architecture.
9785 The @option{-mpowerpc} option allows GCC to generate instructions that
9786 are found only in the 32-bit subset of the PowerPC architecture.
9787 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
9788 GCC to use the optional PowerPC architecture instructions in the
9789 General Purpose group, including floating-point square root. Specifying
9790 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
9791 use the optional PowerPC architecture instructions in the Graphics
9792 group, including floating-point select.
9794 The @option{-mpowerpc64} option allows GCC to generate the additional
9795 64-bit instructions that are found in the full PowerPC64 architecture
9796 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
9797 @option{-mno-powerpc64}.
9799 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
9800 will use only the instructions in the common subset of both
9801 architectures plus some special AIX common-mode calls, and will not use
9802 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
9803 permits GCC to use any instruction from either architecture and to
9804 allow use of the MQ register; specify this for the Motorola MPC601.
9806 @item -mnew-mnemonics
9807 @itemx -mold-mnemonics
9808 @opindex mnew-mnemonics
9809 @opindex mold-mnemonics
9810 Select which mnemonics to use in the generated assembler code. With
9811 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
9812 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
9813 assembler mnemonics defined for the POWER architecture. Instructions
9814 defined in only one architecture have only one mnemonic; GCC uses that
9815 mnemonic irrespective of which of these options is specified.
9817 GCC defaults to the mnemonics appropriate for the architecture in
9818 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
9819 value of these option. Unless you are building a cross-compiler, you
9820 should normally not specify either @option{-mnew-mnemonics} or
9821 @option{-mold-mnemonics}, but should instead accept the default.
9823 @item -mcpu=@var{cpu_type}
9825 Set architecture type, register usage, choice of mnemonics, and
9826 instruction scheduling parameters for machine type @var{cpu_type}.
9827 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
9828 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
9829 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
9830 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
9831 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
9832 @samp{860}, @samp{970}, @samp{common}, @samp{ec603e}, @samp{G3},
9833 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
9834 @samp{power4}, @samp{power5}, @samp{powerpc}, @samp{powerpc64},
9835 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64a}.
9837 @option{-mcpu=common} selects a completely generic processor. Code
9838 generated under this option will run on any POWER or PowerPC processor.
9839 GCC will use only the instructions in the common subset of both
9840 architectures, and will not use the MQ register. GCC assumes a generic
9841 processor model for scheduling purposes.
9843 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
9844 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
9845 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
9846 types, with an appropriate, generic processor model assumed for
9847 scheduling purposes.
9849 The other options specify a specific processor. Code generated under
9850 those options will run best on that processor, and may not run at all on
9853 The @option{-mcpu} options automatically enable or disable the
9854 following options: @option{-maltivec}, @option{-mhard-float},
9855 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
9856 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
9857 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
9858 @option{-mstring}. The particular options set for any particular CPU
9859 will vary between compiler versions, depending on what setting seems
9860 to produce optimal code for that CPU; it doesn't necessarily reflect
9861 the actual hardware's capabilities. If you wish to set an individual
9862 option to a particular value, you may specify it after the
9863 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
9865 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
9866 not enabled or disabled by the @option{-mcpu} option at present, since
9867 AIX does not have full support for these options. You may still
9868 enable or disable them individually if you're sure it'll work in your
9871 @item -mtune=@var{cpu_type}
9873 Set the instruction scheduling parameters for machine type
9874 @var{cpu_type}, but do not set the architecture type, register usage, or
9875 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
9876 values for @var{cpu_type} are used for @option{-mtune} as for
9877 @option{-mcpu}. If both are specified, the code generated will use the
9878 architecture, registers, and mnemonics set by @option{-mcpu}, but the
9879 scheduling parameters set by @option{-mtune}.
9884 @opindex mno-altivec
9885 These switches enable or disable the use of built-in functions that
9886 allow access to the AltiVec instruction set. You may also need to set
9887 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
9892 Extend the current ABI with SPE ABI extensions. This does not change
9893 the default ABI, instead it adds the SPE ABI extensions to the current
9897 @opindex mabi=no-spe
9898 Disable Booke SPE ABI extensions for the current ABI.
9900 @item -misel=@var{yes/no}
9903 This switch enables or disables the generation of ISEL instructions.
9905 @item -mspe=@var{yes/no}
9908 This switch enables or disables the generation of SPE simd
9911 @item -mfloat-gprs=@var{yes/no}
9913 @opindex mfloat-gprs
9914 This switch enables or disables the generation of floating point
9915 operations on the general purpose registers for architectures that
9916 support it. This option is currently only available on the MPC8540.
9919 @itemx -mno-fp-in-toc
9920 @itemx -mno-sum-in-toc
9921 @itemx -mminimal-toc
9923 @opindex mno-fp-in-toc
9924 @opindex mno-sum-in-toc
9925 @opindex mminimal-toc
9926 Modify generation of the TOC (Table Of Contents), which is created for
9927 every executable file. The @option{-mfull-toc} option is selected by
9928 default. In that case, GCC will allocate at least one TOC entry for
9929 each unique non-automatic variable reference in your program. GCC
9930 will also place floating-point constants in the TOC@. However, only
9931 16,384 entries are available in the TOC@.
9933 If you receive a linker error message that saying you have overflowed
9934 the available TOC space, you can reduce the amount of TOC space used
9935 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
9936 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
9937 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
9938 generate code to calculate the sum of an address and a constant at
9939 run-time instead of putting that sum into the TOC@. You may specify one
9940 or both of these options. Each causes GCC to produce very slightly
9941 slower and larger code at the expense of conserving TOC space.
9943 If you still run out of space in the TOC even when you specify both of
9944 these options, specify @option{-mminimal-toc} instead. This option causes
9945 GCC to make only one TOC entry for every file. When you specify this
9946 option, GCC will produce code that is slower and larger but which
9947 uses extremely little TOC space. You may wish to use this option
9948 only on files that contain less frequently executed code.
9954 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
9955 @code{long} type, and the infrastructure needed to support them.
9956 Specifying @option{-maix64} implies @option{-mpowerpc64} and
9957 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
9958 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
9963 @opindex mno-xl-call
9964 On AIX, pass floating-point arguments to prototyped functions beyond the
9965 register save area (RSA) on the stack in addition to argument FPRs. The
9966 AIX calling convention was extended but not initially documented to
9967 handle an obscure K&R C case of calling a function that takes the
9968 address of its arguments with fewer arguments than declared. AIX XL
9969 compilers access floating point arguments which do not fit in the
9970 RSA from the stack when a subroutine is compiled without
9971 optimization. Because always storing floating-point arguments on the
9972 stack is inefficient and rarely needed, this option is not enabled by
9973 default and only is necessary when calling subroutines compiled by AIX
9974 XL compilers without optimization.
9978 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
9979 application written to use message passing with special startup code to
9980 enable the application to run. The system must have PE installed in the
9981 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
9982 must be overridden with the @option{-specs=} option to specify the
9983 appropriate directory location. The Parallel Environment does not
9984 support threads, so the @option{-mpe} option and the @option{-pthread}
9985 option are incompatible.
9987 @item -malign-natural
9988 @itemx -malign-power
9989 @opindex malign-natural
9990 @opindex malign-power
9991 On AIX, Darwin, and 64-bit PowerPC GNU/Linux, the option
9992 @option{-malign-natural} overrides the ABI-defined alignment of larger
9993 types, such as floating-point doubles, on their natural size-based boundary.
9994 The option @option{-malign-power} instructs GCC to follow the ABI-specified
9995 alignment rules. GCC defaults to the standard alignment defined in the ABI.
9999 @opindex msoft-float
10000 @opindex mhard-float
10001 Generate code that does not use (uses) the floating-point register set.
10002 Software floating point emulation is provided if you use the
10003 @option{-msoft-float} option, and pass the option to GCC when linking.
10006 @itemx -mno-multiple
10008 @opindex mno-multiple
10009 Generate code that uses (does not use) the load multiple word
10010 instructions and the store multiple word instructions. These
10011 instructions are generated by default on POWER systems, and not
10012 generated on PowerPC systems. Do not use @option{-mmultiple} on little
10013 endian PowerPC systems, since those instructions do not work when the
10014 processor is in little endian mode. The exceptions are PPC740 and
10015 PPC750 which permit the instructions usage in little endian mode.
10020 @opindex mno-string
10021 Generate code that uses (does not use) the load string instructions
10022 and the store string word instructions to save multiple registers and
10023 do small block moves. These instructions are generated by default on
10024 POWER systems, and not generated on PowerPC systems. Do not use
10025 @option{-mstring} on little endian PowerPC systems, since those
10026 instructions do not work when the processor is in little endian mode.
10027 The exceptions are PPC740 and PPC750 which permit the instructions
10028 usage in little endian mode.
10033 @opindex mno-update
10034 Generate code that uses (does not use) the load or store instructions
10035 that update the base register to the address of the calculated memory
10036 location. These instructions are generated by default. If you use
10037 @option{-mno-update}, there is a small window between the time that the
10038 stack pointer is updated and the address of the previous frame is
10039 stored, which means code that walks the stack frame across interrupts or
10040 signals may get corrupted data.
10043 @itemx -mno-fused-madd
10044 @opindex mfused-madd
10045 @opindex mno-fused-madd
10046 Generate code that uses (does not use) the floating point multiply and
10047 accumulate instructions. These instructions are generated by default if
10048 hardware floating is used.
10050 @item -mno-bit-align
10052 @opindex mno-bit-align
10053 @opindex mbit-align
10054 On System V.4 and embedded PowerPC systems do not (do) force structures
10055 and unions that contain bit-fields to be aligned to the base type of the
10058 For example, by default a structure containing nothing but 8
10059 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
10060 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
10061 the structure would be aligned to a 1 byte boundary and be one byte in
10064 @item -mno-strict-align
10065 @itemx -mstrict-align
10066 @opindex mno-strict-align
10067 @opindex mstrict-align
10068 On System V.4 and embedded PowerPC systems do not (do) assume that
10069 unaligned memory references will be handled by the system.
10071 @item -mrelocatable
10072 @itemx -mno-relocatable
10073 @opindex mrelocatable
10074 @opindex mno-relocatable
10075 On embedded PowerPC systems generate code that allows (does not allow)
10076 the program to be relocated to a different address at runtime. If you
10077 use @option{-mrelocatable} on any module, all objects linked together must
10078 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
10080 @item -mrelocatable-lib
10081 @itemx -mno-relocatable-lib
10082 @opindex mrelocatable-lib
10083 @opindex mno-relocatable-lib
10084 On embedded PowerPC systems generate code that allows (does not allow)
10085 the program to be relocated to a different address at runtime. Modules
10086 compiled with @option{-mrelocatable-lib} can be linked with either modules
10087 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
10088 with modules compiled with the @option{-mrelocatable} options.
10094 On System V.4 and embedded PowerPC systems do not (do) assume that
10095 register 2 contains a pointer to a global area pointing to the addresses
10096 used in the program.
10099 @itemx -mlittle-endian
10101 @opindex mlittle-endian
10102 On System V.4 and embedded PowerPC systems compile code for the
10103 processor in little endian mode. The @option{-mlittle-endian} option is
10104 the same as @option{-mlittle}.
10107 @itemx -mbig-endian
10109 @opindex mbig-endian
10110 On System V.4 and embedded PowerPC systems compile code for the
10111 processor in big endian mode. The @option{-mbig-endian} option is
10112 the same as @option{-mbig}.
10114 @item -mdynamic-no-pic
10115 @opindex mdynamic-no-pic
10116 On Darwin and Mac OS X systems, compile code so that it is not
10117 relocatable, but that its external references are relocatable. The
10118 resulting code is suitable for applications, but not shared
10121 @item -mprioritize-restricted-insns=@var{priority}
10122 @opindex mprioritize-restricted-insns
10123 This option controls the priority that is assigned to
10124 dispatch-slot restricted instructions during the second scheduling
10125 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
10126 @var{no/highest/second-highest} priority to dispatch slot restricted
10129 @item -msched-costly-dep=@var{dependence_type}
10130 @opindex msched-costly-dep
10131 This option controls which dependences are considered costly
10132 by the target during instruction scheduling. The argument
10133 @var{dependence_type} takes one of the following values:
10134 @var{no}: no dependence is costly,
10135 @var{all}: all dependences are costly,
10136 @var{true_store_to_load}: a true dependence from store to load is costly,
10137 @var{store_to_load}: any dependence from store to load is costly,
10138 @var{number}: any dependence which latency >= @var{number} is costly.
10140 @item -minsert-sched-nops=@var{scheme}
10141 @opindex minsert-sched-nops
10142 This option controls which nop insertion scheme will be used during
10143 the second scheduling pass. The argument @var{scheme} takes one of the
10145 @var{no}: Don't insert nops.
10146 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
10147 according to the scheduler's grouping.
10148 @var{regroup_exact}: Insert nops to force costly dependent insns into
10149 separate groups. Insert exactly as many nops as needed to force an insn
10150 to a new group, according to the estimated processor grouping.
10151 @var{number}: Insert nops to force costly dependent insns into
10152 separate groups. Insert @var{number} nops to force an insn to a new group.
10155 @opindex mcall-sysv
10156 On System V.4 and embedded PowerPC systems compile code using calling
10157 conventions that adheres to the March 1995 draft of the System V
10158 Application Binary Interface, PowerPC processor supplement. This is the
10159 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
10161 @item -mcall-sysv-eabi
10162 @opindex mcall-sysv-eabi
10163 Specify both @option{-mcall-sysv} and @option{-meabi} options.
10165 @item -mcall-sysv-noeabi
10166 @opindex mcall-sysv-noeabi
10167 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
10169 @item -mcall-solaris
10170 @opindex mcall-solaris
10171 On System V.4 and embedded PowerPC systems compile code for the Solaris
10175 @opindex mcall-linux
10176 On System V.4 and embedded PowerPC systems compile code for the
10177 Linux-based GNU system.
10181 On System V.4 and embedded PowerPC systems compile code for the
10182 Hurd-based GNU system.
10184 @item -mcall-netbsd
10185 @opindex mcall-netbsd
10186 On System V.4 and embedded PowerPC systems compile code for the
10187 NetBSD operating system.
10189 @item -maix-struct-return
10190 @opindex maix-struct-return
10191 Return all structures in memory (as specified by the AIX ABI)@.
10193 @item -msvr4-struct-return
10194 @opindex msvr4-struct-return
10195 Return structures smaller than 8 bytes in registers (as specified by the
10198 @item -mabi=altivec
10199 @opindex mabi=altivec
10200 Extend the current ABI with AltiVec ABI extensions. This does not
10201 change the default ABI, instead it adds the AltiVec ABI extensions to
10204 @item -mabi=no-altivec
10205 @opindex mabi=no-altivec
10206 Disable AltiVec ABI extensions for the current ABI.
10209 @itemx -mno-prototype
10210 @opindex mprototype
10211 @opindex mno-prototype
10212 On System V.4 and embedded PowerPC systems assume that all calls to
10213 variable argument functions are properly prototyped. Otherwise, the
10214 compiler must insert an instruction before every non prototyped call to
10215 set or clear bit 6 of the condition code register (@var{CR}) to
10216 indicate whether floating point values were passed in the floating point
10217 registers in case the function takes a variable arguments. With
10218 @option{-mprototype}, only calls to prototyped variable argument functions
10219 will set or clear the bit.
10223 On embedded PowerPC systems, assume that the startup module is called
10224 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
10225 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
10230 On embedded PowerPC systems, assume that the startup module is called
10231 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
10236 On embedded PowerPC systems, assume that the startup module is called
10237 @file{crt0.o} and the standard C libraries are @file{libads.a} and
10240 @item -myellowknife
10241 @opindex myellowknife
10242 On embedded PowerPC systems, assume that the startup module is called
10243 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
10248 On System V.4 and embedded PowerPC systems, specify that you are
10249 compiling for a VxWorks system.
10253 Specify that you are compiling for the WindISS simulation environment.
10257 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
10258 header to indicate that @samp{eabi} extended relocations are used.
10264 On System V.4 and embedded PowerPC systems do (do not) adhere to the
10265 Embedded Applications Binary Interface (eabi) which is a set of
10266 modifications to the System V.4 specifications. Selecting @option{-meabi}
10267 means that the stack is aligned to an 8 byte boundary, a function
10268 @code{__eabi} is called to from @code{main} to set up the eabi
10269 environment, and the @option{-msdata} option can use both @code{r2} and
10270 @code{r13} to point to two separate small data areas. Selecting
10271 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
10272 do not call an initialization function from @code{main}, and the
10273 @option{-msdata} option will only use @code{r13} to point to a single
10274 small data area. The @option{-meabi} option is on by default if you
10275 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
10278 @opindex msdata=eabi
10279 On System V.4 and embedded PowerPC systems, put small initialized
10280 @code{const} global and static data in the @samp{.sdata2} section, which
10281 is pointed to by register @code{r2}. Put small initialized
10282 non-@code{const} global and static data in the @samp{.sdata} section,
10283 which is pointed to by register @code{r13}. Put small uninitialized
10284 global and static data in the @samp{.sbss} section, which is adjacent to
10285 the @samp{.sdata} section. The @option{-msdata=eabi} option is
10286 incompatible with the @option{-mrelocatable} option. The
10287 @option{-msdata=eabi} option also sets the @option{-memb} option.
10290 @opindex msdata=sysv
10291 On System V.4 and embedded PowerPC systems, put small global and static
10292 data in the @samp{.sdata} section, which is pointed to by register
10293 @code{r13}. Put small uninitialized global and static data in the
10294 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
10295 The @option{-msdata=sysv} option is incompatible with the
10296 @option{-mrelocatable} option.
10298 @item -msdata=default
10300 @opindex msdata=default
10302 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
10303 compile code the same as @option{-msdata=eabi}, otherwise compile code the
10304 same as @option{-msdata=sysv}.
10307 @opindex msdata-data
10308 On System V.4 and embedded PowerPC systems, put small global and static
10309 data in the @samp{.sdata} section. Put small uninitialized global and
10310 static data in the @samp{.sbss} section. Do not use register @code{r13}
10311 to address small data however. This is the default behavior unless
10312 other @option{-msdata} options are used.
10316 @opindex msdata=none
10318 On embedded PowerPC systems, put all initialized global and static data
10319 in the @samp{.data} section, and all uninitialized data in the
10320 @samp{.bss} section.
10324 @cindex smaller data references (PowerPC)
10325 @cindex .sdata/.sdata2 references (PowerPC)
10326 On embedded PowerPC systems, put global and static items less than or
10327 equal to @var{num} bytes into the small data or bss sections instead of
10328 the normal data or bss section. By default, @var{num} is 8. The
10329 @option{-G @var{num}} switch is also passed to the linker.
10330 All modules should be compiled with the same @option{-G @var{num}} value.
10333 @itemx -mno-regnames
10335 @opindex mno-regnames
10336 On System V.4 and embedded PowerPC systems do (do not) emit register
10337 names in the assembly language output using symbolic forms.
10340 @itemx -mno-longcall
10342 @opindex mno-longcall
10343 Default to making all function calls indirectly, using a register, so
10344 that functions which reside further than 32 megabytes (33,554,432
10345 bytes) from the current location can be called. This setting can be
10346 overridden by the @code{shortcall} function attribute, or by
10347 @code{#pragma longcall(0)}.
10349 Some linkers are capable of detecting out-of-range calls and generating
10350 glue code on the fly. On these systems, long calls are unnecessary and
10351 generate slower code. As of this writing, the AIX linker can do this,
10352 as can the GNU linker for PowerPC/64. It is planned to add this feature
10353 to the GNU linker for 32-bit PowerPC systems as well.
10355 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
10356 callee, L42'', plus a ``branch island'' (glue code). The two target
10357 addresses represent the callee and the ``branch island.'' The
10358 Darwin/PPC linker will prefer the first address and generate a ``bl
10359 callee'' if the PPC ``bl'' instruction will reach the callee directly;
10360 otherwise, the linker will generate ``bl L42'' to call the ``branch
10361 island.'' The ``branch island'' is appended to the body of the
10362 calling function; it computes the full 32-bit address of the callee
10365 On Mach-O (Darwin) systems, this option directs the compiler emit to
10366 the glue for every direct call, and the Darwin linker decides whether
10367 to use or discard it.
10369 In the future, we may cause GCC to ignore all longcall specifications
10370 when the linker is known to generate glue.
10374 Adds support for multithreading with the @dfn{pthreads} library.
10375 This option sets flags for both the preprocessor and linker.
10379 @node S/390 and zSeries Options
10380 @subsection S/390 and zSeries Options
10381 @cindex S/390 and zSeries Options
10383 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
10387 @itemx -msoft-float
10388 @opindex mhard-float
10389 @opindex msoft-float
10390 Use (do not use) the hardware floating-point instructions and registers
10391 for floating-point operations. When @option{-msoft-float} is specified,
10392 functions in @file{libgcc.a} will be used to perform floating-point
10393 operations. When @option{-mhard-float} is specified, the compiler
10394 generates IEEE floating-point instructions. This is the default.
10397 @itemx -mno-backchain
10398 @opindex mbackchain
10399 @opindex mno-backchain
10400 Generate (or do not generate) code which maintains an explicit
10401 backchain within the stack frame that points to the caller's frame.
10402 This may be needed to allow debugging using tools that do not understand
10403 DWARF-2 call frame information. The default is not to generate the
10407 @itemx -mno-small-exec
10408 @opindex msmall-exec
10409 @opindex mno-small-exec
10410 Generate (or do not generate) code using the @code{bras} instruction
10411 to do subroutine calls.
10412 This only works reliably if the total executable size does not
10413 exceed 64k. The default is to use the @code{basr} instruction instead,
10414 which does not have this limitation.
10420 When @option{-m31} is specified, generate code compliant to the
10421 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
10422 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
10423 particular to generate 64-bit instructions. For the @samp{s390}
10424 targets, the default is @option{-m31}, while the @samp{s390x}
10425 targets default to @option{-m64}.
10431 When @option{-mzarch} is specified, generate code using the
10432 instructions available on z/Architecture.
10433 When @option{-mesa} is specified, generate code using the
10434 instructions available on ESA/390. Note that @option{-mesa} is
10435 not possible with @option{-m64}.
10436 When generating code compliant to the GNU/Linux for S/390 ABI,
10437 the default is @option{-mesa}. When generating code compliant
10438 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
10444 Generate (or do not generate) code using the @code{mvcle} instruction
10445 to perform block moves. When @option{-mno-mvcle} is specified,
10446 use a @code{mvc} loop instead. This is the default.
10452 Print (or do not print) additional debug information when compiling.
10453 The default is to not print debug information.
10455 @item -march=@var{cpu-type}
10457 Generate code that will run on @var{cpu-type}, which is the name of a system
10458 representing a certain processor type. Possible values for
10459 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
10460 When generating code using the instructions available on z/Architecture,
10461 the default is @option{-march=z900}. Otherwise, the default is
10462 @option{-march=g5}.
10464 @item -mtune=@var{cpu-type}
10466 Tune to @var{cpu-type} everything applicable about the generated code,
10467 except for the ABI and the set of available instructions.
10468 The list of @var{cpu-type} values is the same as for @option{-march}.
10469 The default is the value used for @option{-march}.
10472 @itemx -mno-tpf-trace
10473 @opindex mtpf-trace
10474 @opindex mno-tpf-trace
10475 Generate code that adds (does not add) in TPF OS specific branches to trace
10476 routines in the operating system. This option is off by default, even
10477 when compiling for the TPF OS.
10480 @itemx -mno-fused-madd
10481 @opindex mfused-madd
10482 @opindex mno-fused-madd
10483 Generate code that uses (does not use) the floating point multiply and
10484 accumulate instructions. These instructions are generated by default if
10485 hardware floating point is used.
10489 @subsection SH Options
10491 These @samp{-m} options are defined for the SH implementations:
10496 Generate code for the SH1.
10500 Generate code for the SH2.
10503 Generate code for the SH2e.
10507 Generate code for the SH3.
10511 Generate code for the SH3e.
10515 Generate code for the SH4 without a floating-point unit.
10517 @item -m4-single-only
10518 @opindex m4-single-only
10519 Generate code for the SH4 with a floating-point unit that only
10520 supports single-precision arithmetic.
10524 Generate code for the SH4 assuming the floating-point unit is in
10525 single-precision mode by default.
10529 Generate code for the SH4.
10533 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
10534 floating-point unit is not used.
10536 @item -m4a-single-only
10537 @opindex m4a-single-only
10538 Generate code for the SH4a, in such a way that no double-precision
10539 floating point operations are used.
10542 @opindex m4a-single
10543 Generate code for the SH4a assuming the floating-point unit is in
10544 single-precision mode by default.
10548 Generate code for the SH4a.
10552 Same as @option{-m4a-nofpu}, except that it implicitly passes
10553 @option{-dsp} to the assembler. GCC doesn't generate any DSP
10554 instructions at the moment.
10558 Compile code for the processor in big endian mode.
10562 Compile code for the processor in little endian mode.
10566 Align doubles at 64-bit boundaries. Note that this changes the calling
10567 conventions, and thus some functions from the standard C library will
10568 not work unless you recompile it first with @option{-mdalign}.
10572 Shorten some address references at link time, when possible; uses the
10573 linker option @option{-relax}.
10577 Use 32-bit offsets in @code{switch} tables. The default is to use
10582 Enable the use of the instruction @code{fmovd}.
10586 Comply with the calling conventions defined by Renesas.
10590 Comply with the calling conventions defined by Renesas.
10594 Comply with the calling conventions defined for GCC before the Renesas
10595 conventions were available. This option is the default for all
10596 targets of the SH toolchain except for @samp{sh-symbianelf}.
10599 @opindex mnomacsave
10600 Mark the @code{MAC} register as call-clobbered, even if
10601 @option{-mhitachi} is given.
10605 Increase IEEE-compliance of floating-point code.
10609 Dump instruction size and location in the assembly code.
10612 @opindex mpadstruct
10613 This option is deprecated. It pads structures to multiple of 4 bytes,
10614 which is incompatible with the SH ABI@.
10618 Optimize for space instead of speed. Implied by @option{-Os}.
10621 @opindex mprefergot
10622 When generating position-independent code, emit function calls using
10623 the Global Offset Table instead of the Procedure Linkage Table.
10627 Generate a library function call to invalidate instruction cache
10628 entries, after fixing up a trampoline. This library function call
10629 doesn't assume it can write to the whole memory address space. This
10630 is the default when the target is @code{sh-*-linux*}.
10633 @node SPARC Options
10634 @subsection SPARC Options
10635 @cindex SPARC options
10637 These @samp{-m} options are supported on the SPARC:
10640 @item -mno-app-regs
10642 @opindex mno-app-regs
10644 Specify @option{-mapp-regs} to generate output using the global registers
10645 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
10648 To be fully SVR4 ABI compliant at the cost of some performance loss,
10649 specify @option{-mno-app-regs}. You should compile libraries and system
10650 software with this option.
10653 @itemx -mhard-float
10655 @opindex mhard-float
10656 Generate output containing floating point instructions. This is the
10660 @itemx -msoft-float
10662 @opindex msoft-float
10663 Generate output containing library calls for floating point.
10664 @strong{Warning:} the requisite libraries are not available for all SPARC
10665 targets. Normally the facilities of the machine's usual C compiler are
10666 used, but this cannot be done directly in cross-compilation. You must make
10667 your own arrangements to provide suitable library functions for
10668 cross-compilation. The embedded targets @samp{sparc-*-aout} and
10669 @samp{sparclite-*-*} do provide software floating point support.
10671 @option{-msoft-float} changes the calling convention in the output file;
10672 therefore, it is only useful if you compile @emph{all} of a program with
10673 this option. In particular, you need to compile @file{libgcc.a}, the
10674 library that comes with GCC, with @option{-msoft-float} in order for
10677 @item -mhard-quad-float
10678 @opindex mhard-quad-float
10679 Generate output containing quad-word (long double) floating point
10682 @item -msoft-quad-float
10683 @opindex msoft-quad-float
10684 Generate output containing library calls for quad-word (long double)
10685 floating point instructions. The functions called are those specified
10686 in the SPARC ABI@. This is the default.
10688 As of this writing, there are no SPARC implementations that have hardware
10689 support for the quad-word floating point instructions. They all invoke
10690 a trap handler for one of these instructions, and then the trap handler
10691 emulates the effect of the instruction. Because of the trap handler overhead,
10692 this is much slower than calling the ABI library routines. Thus the
10693 @option{-msoft-quad-float} option is the default.
10695 @item -mno-unaligned-doubles
10696 @itemx -munaligned-doubles
10697 @opindex mno-unaligned-doubles
10698 @opindex munaligned-doubles
10699 Assume that doubles have 8 byte alignment. This is the default.
10701 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
10702 alignment only if they are contained in another type, or if they have an
10703 absolute address. Otherwise, it assumes they have 4 byte alignment.
10704 Specifying this option avoids some rare compatibility problems with code
10705 generated by other compilers. It is not the default because it results
10706 in a performance loss, especially for floating point code.
10708 @item -mno-faster-structs
10709 @itemx -mfaster-structs
10710 @opindex mno-faster-structs
10711 @opindex mfaster-structs
10712 With @option{-mfaster-structs}, the compiler assumes that structures
10713 should have 8 byte alignment. This enables the use of pairs of
10714 @code{ldd} and @code{std} instructions for copies in structure
10715 assignment, in place of twice as many @code{ld} and @code{st} pairs.
10716 However, the use of this changed alignment directly violates the SPARC
10717 ABI@. Thus, it's intended only for use on targets where the developer
10718 acknowledges that their resulting code will not be directly in line with
10719 the rules of the ABI@.
10721 @item -mimpure-text
10722 @opindex mimpure-text
10723 @option{-mimpure-text}, used in addition to @option{-shared}, tells
10724 the compiler to not pass @option{-z text} to the linker when linking a
10725 shared object. Using this option, you can link position-dependent
10726 code into a shared object.
10728 @option{-mimpure-text} suppresses the ``relocations remain against
10729 allocatable but non-writable sections'' linker error message.
10730 However, the necessary relocations will trigger copy-on-write, and the
10731 shared object is not actually shared across processes. Instead of
10732 using @option{-mimpure-text}, you should compile all source code with
10733 @option{-fpic} or @option{-fPIC}.
10735 This option is only available on SunOS and Solaris.
10737 @item -mcpu=@var{cpu_type}
10739 Set the instruction set, register set, and instruction scheduling parameters
10740 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
10741 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
10742 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
10743 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
10744 @samp{ultrasparc3}.
10746 Default instruction scheduling parameters are used for values that select
10747 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
10748 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
10750 Here is a list of each supported architecture and their supported
10755 v8: supersparc, hypersparc
10756 sparclite: f930, f934, sparclite86x
10758 v9: ultrasparc, ultrasparc3
10761 By default (unless configured otherwise), GCC generates code for the V7
10762 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
10763 additionally optimizes it for the Cypress CY7C602 chip, as used in the
10764 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
10765 SPARCStation 1, 2, IPX etc.
10767 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
10768 architecture. The only difference from V7 code is that the compiler emits
10769 the integer multiply and integer divide instructions which exist in SPARC-V8
10770 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
10771 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
10774 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
10775 the SPARC architecture. This adds the integer multiply, integer divide step
10776 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
10777 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
10778 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU. With
10779 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
10780 MB86934 chip, which is the more recent SPARClite with FPU.
10782 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
10783 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
10784 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
10785 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
10786 optimizes it for the TEMIC SPARClet chip.
10788 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
10789 architecture. This adds 64-bit integer and floating-point move instructions,
10790 3 additional floating-point condition code registers and conditional move
10791 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
10792 optimizes it for the Sun UltraSPARC I/II chips. With
10793 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
10794 Sun UltraSPARC III chip.
10796 @item -mtune=@var{cpu_type}
10798 Set the instruction scheduling parameters for machine type
10799 @var{cpu_type}, but do not set the instruction set or register set that the
10800 option @option{-mcpu=@var{cpu_type}} would.
10802 The same values for @option{-mcpu=@var{cpu_type}} can be used for
10803 @option{-mtune=@var{cpu_type}}, but the only useful values are those
10804 that select a particular cpu implementation. Those are @samp{cypress},
10805 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
10806 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
10807 @samp{ultrasparc3}.
10812 @opindex mno-v8plus
10813 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI. The
10814 difference from the V8 ABI is that the global and out registers are
10815 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
10816 mode for all SPARC-V9 processors.
10822 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
10823 Visual Instruction Set extensions. The default is @option{-mno-vis}.
10826 These @samp{-m} options are supported in addition to the above
10827 on SPARC-V9 processors in 64-bit environments:
10830 @item -mlittle-endian
10831 @opindex mlittle-endian
10832 Generate code for a processor running in little-endian mode. It is only
10833 available for a few configurations and most notably not on Solaris.
10839 Generate code for a 32-bit or 64-bit environment.
10840 The 32-bit environment sets int, long and pointer to 32 bits.
10841 The 64-bit environment sets int to 32 bits and long and pointer
10844 @item -mcmodel=medlow
10845 @opindex mcmodel=medlow
10846 Generate code for the Medium/Low code model: 64-bit addresses, programs
10847 must be linked in the low 32 bits of memory. Programs can be statically
10848 or dynamically linked.
10850 @item -mcmodel=medmid
10851 @opindex mcmodel=medmid
10852 Generate code for the Medium/Middle code model: 64-bit addresses, programs
10853 must be linked in the low 44 bits of memory, the text and data segments must
10854 be less than 2GB in size and the data segment must be located within 2GB of
10857 @item -mcmodel=medany
10858 @opindex mcmodel=medany
10859 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
10860 may be linked anywhere in memory, the text and data segments must be less
10861 than 2GB in size and the data segment must be located within 2GB of the
10864 @item -mcmodel=embmedany
10865 @opindex mcmodel=embmedany
10866 Generate code for the Medium/Anywhere code model for embedded systems:
10867 64-bit addresses, the text and data segments must be less than 2GB in
10868 size, both starting anywhere in memory (determined at link time). The
10869 global register %g4 points to the base of the data segment. Programs
10870 are statically linked and PIC is not supported.
10873 @itemx -mno-stack-bias
10874 @opindex mstack-bias
10875 @opindex mno-stack-bias
10876 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
10877 frame pointer if present, are offset by @minus{}2047 which must be added back
10878 when making stack frame references. This is the default in 64-bit mode.
10879 Otherwise, assume no such offset is present.
10882 These switches are supported in addition to the above on Solaris:
10887 Add support for multithreading using the Solaris threads library. This
10888 option sets flags for both the preprocessor and linker. This option does
10889 not affect the thread safety of object code produced by the compiler or
10890 that of libraries supplied with it.
10894 Add support for multithreading using the POSIX threads library. This
10895 option sets flags for both the preprocessor and linker. This option does
10896 not affect the thread safety of object code produced by the compiler or
10897 that of libraries supplied with it.
10900 @node System V Options
10901 @subsection Options for System V
10903 These additional options are available on System V Release 4 for
10904 compatibility with other compilers on those systems:
10909 Create a shared object.
10910 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
10914 Identify the versions of each tool used by the compiler, in a
10915 @code{.ident} assembler directive in the output.
10919 Refrain from adding @code{.ident} directives to the output file (this is
10922 @item -YP,@var{dirs}
10924 Search the directories @var{dirs}, and no others, for libraries
10925 specified with @option{-l}.
10927 @item -Ym,@var{dir}
10929 Look in the directory @var{dir} to find the M4 preprocessor.
10930 The assembler uses this option.
10931 @c This is supposed to go with a -Yd for predefined M4 macro files, but
10932 @c the generic assembler that comes with Solaris takes just -Ym.
10935 @node TMS320C3x/C4x Options
10936 @subsection TMS320C3x/C4x Options
10937 @cindex TMS320C3x/C4x Options
10939 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
10943 @item -mcpu=@var{cpu_type}
10945 Set the instruction set, register set, and instruction scheduling
10946 parameters for machine type @var{cpu_type}. Supported values for
10947 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
10948 @samp{c44}. The default is @samp{c40} to generate code for the
10953 @itemx -msmall-memory
10955 @opindex mbig-memory
10957 @opindex msmall-memory
10959 Generates code for the big or small memory model. The small memory
10960 model assumed that all data fits into one 64K word page. At run-time
10961 the data page (DP) register must be set to point to the 64K page
10962 containing the .bss and .data program sections. The big memory model is
10963 the default and requires reloading of the DP register for every direct
10970 Allow (disallow) allocation of general integer operands into the block
10971 count register BK@.
10977 Enable (disable) generation of code using decrement and branch,
10978 DBcond(D), instructions. This is enabled by default for the C4x. To be
10979 on the safe side, this is disabled for the C3x, since the maximum
10980 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
10981 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
10982 that it can utilize the decrement and branch instruction, but will give
10983 up if there is more than one memory reference in the loop. Thus a loop
10984 where the loop counter is decremented can generate slightly more
10985 efficient code, in cases where the RPTB instruction cannot be utilized.
10987 @item -mdp-isr-reload
10989 @opindex mdp-isr-reload
10991 Force the DP register to be saved on entry to an interrupt service
10992 routine (ISR), reloaded to point to the data section, and restored on
10993 exit from the ISR@. This should not be required unless someone has
10994 violated the small memory model by modifying the DP register, say within
11001 For the C3x use the 24-bit MPYI instruction for integer multiplies
11002 instead of a library call to guarantee 32-bit results. Note that if one
11003 of the operands is a constant, then the multiplication will be performed
11004 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
11005 then squaring operations are performed inline instead of a library call.
11008 @itemx -mno-fast-fix
11010 @opindex mno-fast-fix
11011 The C3x/C4x FIX instruction to convert a floating point value to an
11012 integer value chooses the nearest integer less than or equal to the
11013 floating point value rather than to the nearest integer. Thus if the
11014 floating point number is negative, the result will be incorrectly
11015 truncated an additional code is necessary to detect and correct this
11016 case. This option can be used to disable generation of the additional
11017 code required to correct the result.
11023 Enable (disable) generation of repeat block sequences using the RPTB
11024 instruction for zero overhead looping. The RPTB construct is only used
11025 for innermost loops that do not call functions or jump across the loop
11026 boundaries. There is no advantage having nested RPTB loops due to the
11027 overhead required to save and restore the RC, RS, and RE registers.
11028 This is enabled by default with @option{-O2}.
11030 @item -mrpts=@var{count}
11034 Enable (disable) the use of the single instruction repeat instruction
11035 RPTS@. If a repeat block contains a single instruction, and the loop
11036 count can be guaranteed to be less than the value @var{count}, GCC will
11037 emit a RPTS instruction instead of a RPTB@. If no value is specified,
11038 then a RPTS will be emitted even if the loop count cannot be determined
11039 at compile time. Note that the repeated instruction following RPTS does
11040 not have to be reloaded from memory each iteration, thus freeing up the
11041 CPU buses for operands. However, since interrupts are blocked by this
11042 instruction, it is disabled by default.
11044 @item -mloop-unsigned
11045 @itemx -mno-loop-unsigned
11046 @opindex mloop-unsigned
11047 @opindex mno-loop-unsigned
11048 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
11049 is @math{2^{31} + 1} since these instructions test if the iteration count is
11050 negative to terminate the loop. If the iteration count is unsigned
11051 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
11052 exceeded. This switch allows an unsigned iteration count.
11056 Try to emit an assembler syntax that the TI assembler (asm30) is happy
11057 with. This also enforces compatibility with the API employed by the TI
11058 C3x C compiler. For example, long doubles are passed as structures
11059 rather than in floating point registers.
11065 Generate code that uses registers (stack) for passing arguments to functions.
11066 By default, arguments are passed in registers where possible rather
11067 than by pushing arguments on to the stack.
11069 @item -mparallel-insns
11070 @itemx -mno-parallel-insns
11071 @opindex mparallel-insns
11072 @opindex mno-parallel-insns
11073 Allow the generation of parallel instructions. This is enabled by
11074 default with @option{-O2}.
11076 @item -mparallel-mpy
11077 @itemx -mno-parallel-mpy
11078 @opindex mparallel-mpy
11079 @opindex mno-parallel-mpy
11080 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
11081 provided @option{-mparallel-insns} is also specified. These instructions have
11082 tight register constraints which can pessimize the code generation
11083 of large functions.
11088 @subsection V850 Options
11089 @cindex V850 Options
11091 These @samp{-m} options are defined for V850 implementations:
11095 @itemx -mno-long-calls
11096 @opindex mlong-calls
11097 @opindex mno-long-calls
11098 Treat all calls as being far away (near). If calls are assumed to be
11099 far away, the compiler will always load the functions address up into a
11100 register, and call indirect through the pointer.
11106 Do not optimize (do optimize) basic blocks that use the same index
11107 pointer 4 or more times to copy pointer into the @code{ep} register, and
11108 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
11109 option is on by default if you optimize.
11111 @item -mno-prolog-function
11112 @itemx -mprolog-function
11113 @opindex mno-prolog-function
11114 @opindex mprolog-function
11115 Do not use (do use) external functions to save and restore registers
11116 at the prologue and epilogue of a function. The external functions
11117 are slower, but use less code space if more than one function saves
11118 the same number of registers. The @option{-mprolog-function} option
11119 is on by default if you optimize.
11123 Try to make the code as small as possible. At present, this just turns
11124 on the @option{-mep} and @option{-mprolog-function} options.
11126 @item -mtda=@var{n}
11128 Put static or global variables whose size is @var{n} bytes or less into
11129 the tiny data area that register @code{ep} points to. The tiny data
11130 area can hold up to 256 bytes in total (128 bytes for byte references).
11132 @item -msda=@var{n}
11134 Put static or global variables whose size is @var{n} bytes or less into
11135 the small data area that register @code{gp} points to. The small data
11136 area can hold up to 64 kilobytes.
11138 @item -mzda=@var{n}
11140 Put static or global variables whose size is @var{n} bytes or less into
11141 the first 32 kilobytes of memory.
11145 Specify that the target processor is the V850.
11148 @opindex mbig-switch
11149 Generate code suitable for big switch tables. Use this option only if
11150 the assembler/linker complain about out of range branches within a switch
11155 This option will cause r2 and r5 to be used in the code generated by
11156 the compiler. This setting is the default.
11158 @item -mno-app-regs
11159 @opindex mno-app-regs
11160 This option will cause r2 and r5 to be treated as fixed registers.
11164 Specify that the target processor is the V850E1. The preprocessor
11165 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
11166 this option is used.
11170 Specify that the target processor is the V850E. The preprocessor
11171 constant @samp{__v850e__} will be defined if this option is used.
11173 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
11174 are defined then a default target processor will be chosen and the
11175 relevant @samp{__v850*__} preprocessor constant will be defined.
11177 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
11178 defined, regardless of which processor variant is the target.
11180 @item -mdisable-callt
11181 @opindex mdisable-callt
11182 This option will suppress generation of the CALLT instruction for the
11183 v850e and v850e1 flavors of the v850 architecture. The default is
11184 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
11189 @subsection VAX Options
11190 @cindex VAX options
11192 These @samp{-m} options are defined for the VAX:
11197 Do not output certain jump instructions (@code{aobleq} and so on)
11198 that the Unix assembler for the VAX cannot handle across long
11203 Do output those jump instructions, on the assumption that you
11204 will assemble with the GNU assembler.
11208 Output code for g-format floating point numbers instead of d-format.
11211 @node x86-64 Options
11212 @subsection x86-64 Options
11213 @cindex x86-64 options
11215 These are listed under @xref{i386 and x86-64 Options}.
11217 @node Xstormy16 Options
11218 @subsection Xstormy16 Options
11219 @cindex Xstormy16 Options
11221 These options are defined for Xstormy16:
11226 Choose startup files and linker script suitable for the simulator.
11229 @node Xtensa Options
11230 @subsection Xtensa Options
11231 @cindex Xtensa Options
11233 These options are supported for Xtensa targets:
11237 @itemx -mno-const16
11239 @opindex mno-const16
11240 Enable or disable use of @code{CONST16} instructions for loading
11241 constant values. The @code{CONST16} instruction is currently not a
11242 standard option from Tensilica. When enabled, @code{CONST16}
11243 instructions are always used in place of the standard @code{L32R}
11244 instructions. The use of @code{CONST16} is enabled by default only if
11245 the @code{L32R} instruction is not available.
11248 @itemx -mno-fused-madd
11249 @opindex mfused-madd
11250 @opindex mno-fused-madd
11251 Enable or disable use of fused multiply/add and multiply/subtract
11252 instructions in the floating-point option. This has no effect if the
11253 floating-point option is not also enabled. Disabling fused multiply/add
11254 and multiply/subtract instructions forces the compiler to use separate
11255 instructions for the multiply and add/subtract operations. This may be
11256 desirable in some cases where strict IEEE 754-compliant results are
11257 required: the fused multiply add/subtract instructions do not round the
11258 intermediate result, thereby producing results with @emph{more} bits of
11259 precision than specified by the IEEE standard. Disabling fused multiply
11260 add/subtract instructions also ensures that the program output is not
11261 sensitive to the compiler's ability to combine multiply and add/subtract
11264 @item -mtext-section-literals
11265 @itemx -mno-text-section-literals
11266 @opindex mtext-section-literals
11267 @opindex mno-text-section-literals
11268 Control the treatment of literal pools. The default is
11269 @option{-mno-text-section-literals}, which places literals in a separate
11270 section in the output file. This allows the literal pool to be placed
11271 in a data RAM/ROM, and it also allows the linker to combine literal
11272 pools from separate object files to remove redundant literals and
11273 improve code size. With @option{-mtext-section-literals}, the literals
11274 are interspersed in the text section in order to keep them as close as
11275 possible to their references. This may be necessary for large assembly
11278 @item -mtarget-align
11279 @itemx -mno-target-align
11280 @opindex mtarget-align
11281 @opindex mno-target-align
11282 When this option is enabled, GCC instructs the assembler to
11283 automatically align instructions to reduce branch penalties at the
11284 expense of some code density. The assembler attempts to widen density
11285 instructions to align branch targets and the instructions following call
11286 instructions. If there are not enough preceding safe density
11287 instructions to align a target, no widening will be performed. The
11288 default is @option{-mtarget-align}. These options do not affect the
11289 treatment of auto-aligned instructions like @code{LOOP}, which the
11290 assembler will always align, either by widening density instructions or
11291 by inserting no-op instructions.
11294 @itemx -mno-longcalls
11295 @opindex mlongcalls
11296 @opindex mno-longcalls
11297 When this option is enabled, GCC instructs the assembler to translate
11298 direct calls to indirect calls unless it can determine that the target
11299 of a direct call is in the range allowed by the call instruction. This
11300 translation typically occurs for calls to functions in other source
11301 files. Specifically, the assembler translates a direct @code{CALL}
11302 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
11303 The default is @option{-mno-longcalls}. This option should be used in
11304 programs where the call target can potentially be out of range. This
11305 option is implemented in the assembler, not the compiler, so the
11306 assembly code generated by GCC will still show direct call
11307 instructions---look at the disassembled object code to see the actual
11308 instructions. Note that the assembler will use an indirect call for
11309 every cross-file call, not just those that really will be out of range.
11312 @node zSeries Options
11313 @subsection zSeries Options
11314 @cindex zSeries options
11316 These are listed under @xref{S/390 and zSeries Options}.
11318 @node Code Gen Options
11319 @section Options for Code Generation Conventions
11320 @cindex code generation conventions
11321 @cindex options, code generation
11322 @cindex run-time options
11324 These machine-independent options control the interface conventions
11325 used in code generation.
11327 Most of them have both positive and negative forms; the negative form
11328 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
11329 one of the forms is listed---the one which is not the default. You
11330 can figure out the other form by either removing @samp{no-} or adding
11334 @item -fbounds-check
11335 @opindex fbounds-check
11336 For front-ends that support it, generate additional code to check that
11337 indices used to access arrays are within the declared range. This is
11338 currently only supported by the Java and Fortran 77 front-ends, where
11339 this option defaults to true and false respectively.
11343 This option generates traps for signed overflow on addition, subtraction,
11344 multiplication operations.
11348 This option instructs the compiler to assume that signed arithmetic
11349 overflow of addition, subtraction and multiplication wraps around
11350 using twos-complement representation. This flag enables some optimizations
11351 and disables other. This option is enabled by default for the Java
11352 front-end, as required by the Java language specification.
11355 @opindex fexceptions
11356 Enable exception handling. Generates extra code needed to propagate
11357 exceptions. For some targets, this implies GCC will generate frame
11358 unwind information for all functions, which can produce significant data
11359 size overhead, although it does not affect execution. If you do not
11360 specify this option, GCC will enable it by default for languages like
11361 C++ which normally require exception handling, and disable it for
11362 languages like C that do not normally require it. However, you may need
11363 to enable this option when compiling C code that needs to interoperate
11364 properly with exception handlers written in C++. You may also wish to
11365 disable this option if you are compiling older C++ programs that don't
11366 use exception handling.
11368 @item -fnon-call-exceptions
11369 @opindex fnon-call-exceptions
11370 Generate code that allows trapping instructions to throw exceptions.
11371 Note that this requires platform-specific runtime support that does
11372 not exist everywhere. Moreover, it only allows @emph{trapping}
11373 instructions to throw exceptions, i.e.@: memory references or floating
11374 point instructions. It does not allow exceptions to be thrown from
11375 arbitrary signal handlers such as @code{SIGALRM}.
11377 @item -funwind-tables
11378 @opindex funwind-tables
11379 Similar to @option{-fexceptions}, except that it will just generate any needed
11380 static data, but will not affect the generated code in any other way.
11381 You will normally not enable this option; instead, a language processor
11382 that needs this handling would enable it on your behalf.
11384 @item -fasynchronous-unwind-tables
11385 @opindex fasynchronous-unwind-tables
11386 Generate unwind table in dwarf2 format, if supported by target machine. The
11387 table is exact at each instruction boundary, so it can be used for stack
11388 unwinding from asynchronous events (such as debugger or garbage collector).
11390 @item -fpcc-struct-return
11391 @opindex fpcc-struct-return
11392 Return ``short'' @code{struct} and @code{union} values in memory like
11393 longer ones, rather than in registers. This convention is less
11394 efficient, but it has the advantage of allowing intercallability between
11395 GCC-compiled files and files compiled with other compilers, particularly
11396 the Portable C Compiler (pcc).
11398 The precise convention for returning structures in memory depends
11399 on the target configuration macros.
11401 Short structures and unions are those whose size and alignment match
11402 that of some integer type.
11404 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
11405 switch is not binary compatible with code compiled with the
11406 @option{-freg-struct-return} switch.
11407 Use it to conform to a non-default application binary interface.
11409 @item -freg-struct-return
11410 @opindex freg-struct-return
11411 Return @code{struct} and @code{union} values in registers when possible.
11412 This is more efficient for small structures than
11413 @option{-fpcc-struct-return}.
11415 If you specify neither @option{-fpcc-struct-return} nor
11416 @option{-freg-struct-return}, GCC defaults to whichever convention is
11417 standard for the target. If there is no standard convention, GCC
11418 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
11419 the principal compiler. In those cases, we can choose the standard, and
11420 we chose the more efficient register return alternative.
11422 @strong{Warning:} code compiled with the @option{-freg-struct-return}
11423 switch is not binary compatible with code compiled with the
11424 @option{-fpcc-struct-return} switch.
11425 Use it to conform to a non-default application binary interface.
11427 @item -fshort-enums
11428 @opindex fshort-enums
11429 Allocate to an @code{enum} type only as many bytes as it needs for the
11430 declared range of possible values. Specifically, the @code{enum} type
11431 will be equivalent to the smallest integer type which has enough room.
11433 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
11434 code that is not binary compatible with code generated without that switch.
11435 Use it to conform to a non-default application binary interface.
11437 @item -fshort-double
11438 @opindex fshort-double
11439 Use the same size for @code{double} as for @code{float}.
11441 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
11442 code that is not binary compatible with code generated without that switch.
11443 Use it to conform to a non-default application binary interface.
11445 @item -fshort-wchar
11446 @opindex fshort-wchar
11447 Override the underlying type for @samp{wchar_t} to be @samp{short
11448 unsigned int} instead of the default for the target. This option is
11449 useful for building programs to run under WINE@.
11451 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
11452 code that is not binary compatible with code generated without that switch.
11453 Use it to conform to a non-default application binary interface.
11455 @item -fshared-data
11456 @opindex fshared-data
11457 Requests that the data and non-@code{const} variables of this
11458 compilation be shared data rather than private data. The distinction
11459 makes sense only on certain operating systems, where shared data is
11460 shared between processes running the same program, while private data
11461 exists in one copy per process.
11464 @opindex fno-common
11465 In C, allocate even uninitialized global variables in the data section of the
11466 object file, rather than generating them as common blocks. This has the
11467 effect that if the same variable is declared (without @code{extern}) in
11468 two different compilations, you will get an error when you link them.
11469 The only reason this might be useful is if you wish to verify that the
11470 program will work on other systems which always work this way.
11474 Ignore the @samp{#ident} directive.
11476 @item -finhibit-size-directive
11477 @opindex finhibit-size-directive
11478 Don't output a @code{.size} assembler directive, or anything else that
11479 would cause trouble if the function is split in the middle, and the
11480 two halves are placed at locations far apart in memory. This option is
11481 used when compiling @file{crtstuff.c}; you should not need to use it
11484 @item -fverbose-asm
11485 @opindex fverbose-asm
11486 Put extra commentary information in the generated assembly code to
11487 make it more readable. This option is generally only of use to those
11488 who actually need to read the generated assembly code (perhaps while
11489 debugging the compiler itself).
11491 @option{-fno-verbose-asm}, the default, causes the
11492 extra information to be omitted and is useful when comparing two assembler
11497 @cindex global offset table
11499 Generate position-independent code (PIC) suitable for use in a shared
11500 library, if supported for the target machine. Such code accesses all
11501 constant addresses through a global offset table (GOT)@. The dynamic
11502 loader resolves the GOT entries when the program starts (the dynamic
11503 loader is not part of GCC; it is part of the operating system). If
11504 the GOT size for the linked executable exceeds a machine-specific
11505 maximum size, you get an error message from the linker indicating that
11506 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
11507 instead. (These maximums are 8k on the SPARC and 32k
11508 on the m68k and RS/6000. The 386 has no such limit.)
11510 Position-independent code requires special support, and therefore works
11511 only on certain machines. For the 386, GCC supports PIC for System V
11512 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
11513 position-independent.
11517 If supported for the target machine, emit position-independent code,
11518 suitable for dynamic linking and avoiding any limit on the size of the
11519 global offset table. This option makes a difference on the m68k
11522 Position-independent code requires special support, and therefore works
11523 only on certain machines.
11529 These options are similar to @option{-fpic} and @option{-fPIC}, but
11530 generated position independent code can be only linked into executables.
11531 Usually these options are used when @option{-pie} GCC option will be
11532 used during linking.
11534 @item -ffixed-@var{reg}
11536 Treat the register named @var{reg} as a fixed register; generated code
11537 should never refer to it (except perhaps as a stack pointer, frame
11538 pointer or in some other fixed role).
11540 @var{reg} must be the name of a register. The register names accepted
11541 are machine-specific and are defined in the @code{REGISTER_NAMES}
11542 macro in the machine description macro file.
11544 This flag does not have a negative form, because it specifies a
11547 @item -fcall-used-@var{reg}
11548 @opindex fcall-used
11549 Treat the register named @var{reg} as an allocable register that is
11550 clobbered by function calls. It may be allocated for temporaries or
11551 variables that do not live across a call. Functions compiled this way
11552 will not save and restore the register @var{reg}.
11554 It is an error to used this flag with the frame pointer or stack pointer.
11555 Use of this flag for other registers that have fixed pervasive roles in
11556 the machine's execution model will produce disastrous results.
11558 This flag does not have a negative form, because it specifies a
11561 @item -fcall-saved-@var{reg}
11562 @opindex fcall-saved
11563 Treat the register named @var{reg} as an allocable register saved by
11564 functions. It may be allocated even for temporaries or variables that
11565 live across a call. Functions compiled this way will save and restore
11566 the register @var{reg} if they use it.
11568 It is an error to used this flag with the frame pointer or stack pointer.
11569 Use of this flag for other registers that have fixed pervasive roles in
11570 the machine's execution model will produce disastrous results.
11572 A different sort of disaster will result from the use of this flag for
11573 a register in which function values may be returned.
11575 This flag does not have a negative form, because it specifies a
11578 @item -fpack-struct
11579 @opindex fpack-struct
11580 Pack all structure members together without holes.
11582 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
11583 code that is not binary compatible with code generated without that switch.
11584 Additionally, it makes the code suboptimal.
11585 Use it to conform to a non-default application binary interface.
11587 @item -finstrument-functions
11588 @opindex finstrument-functions
11589 Generate instrumentation calls for entry and exit to functions. Just
11590 after function entry and just before function exit, the following
11591 profiling functions will be called with the address of the current
11592 function and its call site. (On some platforms,
11593 @code{__builtin_return_address} does not work beyond the current
11594 function, so the call site information may not be available to the
11595 profiling functions otherwise.)
11598 void __cyg_profile_func_enter (void *this_fn,
11600 void __cyg_profile_func_exit (void *this_fn,
11604 The first argument is the address of the start of the current function,
11605 which may be looked up exactly in the symbol table.
11607 This instrumentation is also done for functions expanded inline in other
11608 functions. The profiling calls will indicate where, conceptually, the
11609 inline function is entered and exited. This means that addressable
11610 versions of such functions must be available. If all your uses of a
11611 function are expanded inline, this may mean an additional expansion of
11612 code size. If you use @samp{extern inline} in your C code, an
11613 addressable version of such functions must be provided. (This is
11614 normally the case anyways, but if you get lucky and the optimizer always
11615 expands the functions inline, you might have gotten away without
11616 providing static copies.)
11618 A function may be given the attribute @code{no_instrument_function}, in
11619 which case this instrumentation will not be done. This can be used, for
11620 example, for the profiling functions listed above, high-priority
11621 interrupt routines, and any functions from which the profiling functions
11622 cannot safely be called (perhaps signal handlers, if the profiling
11623 routines generate output or allocate memory).
11625 @item -fstack-check
11626 @opindex fstack-check
11627 Generate code to verify that you do not go beyond the boundary of the
11628 stack. You should specify this flag if you are running in an
11629 environment with multiple threads, but only rarely need to specify it in
11630 a single-threaded environment since stack overflow is automatically
11631 detected on nearly all systems if there is only one stack.
11633 Note that this switch does not actually cause checking to be done; the
11634 operating system must do that. The switch causes generation of code
11635 to ensure that the operating system sees the stack being extended.
11637 @item -fstack-limit-register=@var{reg}
11638 @itemx -fstack-limit-symbol=@var{sym}
11639 @itemx -fno-stack-limit
11640 @opindex fstack-limit-register
11641 @opindex fstack-limit-symbol
11642 @opindex fno-stack-limit
11643 Generate code to ensure that the stack does not grow beyond a certain value,
11644 either the value of a register or the address of a symbol. If the stack
11645 would grow beyond the value, a signal is raised. For most targets,
11646 the signal is raised before the stack overruns the boundary, so
11647 it is possible to catch the signal without taking special precautions.
11649 For instance, if the stack starts at absolute address @samp{0x80000000}
11650 and grows downwards, you can use the flags
11651 @option{-fstack-limit-symbol=__stack_limit} and
11652 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11653 of 128KB@. Note that this may only work with the GNU linker.
11655 @cindex aliasing of parameters
11656 @cindex parameters, aliased
11657 @item -fargument-alias
11658 @itemx -fargument-noalias
11659 @itemx -fargument-noalias-global
11660 @opindex fargument-alias
11661 @opindex fargument-noalias
11662 @opindex fargument-noalias-global
11663 Specify the possible relationships among parameters and between
11664 parameters and global data.
11666 @option{-fargument-alias} specifies that arguments (parameters) may
11667 alias each other and may alias global storage.@*
11668 @option{-fargument-noalias} specifies that arguments do not alias
11669 each other, but may alias global storage.@*
11670 @option{-fargument-noalias-global} specifies that arguments do not
11671 alias each other and do not alias global storage.
11673 Each language will automatically use whatever option is required by
11674 the language standard. You should not need to use these options yourself.
11676 @item -fleading-underscore
11677 @opindex fleading-underscore
11678 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
11679 change the way C symbols are represented in the object file. One use
11680 is to help link with legacy assembly code.
11682 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
11683 generate code that is not binary compatible with code generated without that
11684 switch. Use it to conform to a non-default application binary interface.
11685 Not all targets provide complete support for this switch.
11687 @item -ftls-model=@var{model}
11688 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
11689 The @var{model} argument should be one of @code{global-dynamic},
11690 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
11692 The default without @option{-fpic} is @code{initial-exec}; with
11693 @option{-fpic} the default is @code{global-dynamic}.
11695 @item -fvisibility=@var{default|internal|hidden|protected}
11696 @opindex fvisibility
11697 Set the default ELF image symbol visibility to the specified option - all
11698 symbols will be marked with this unless overrided within the code.
11699 Using this feature can very substantially improve linking and
11700 load times of shared object libraries, produce more optimised
11701 code, provide near-perfect API export and prevent symbol clashes.
11702 It is @strong{strongly} recommended that you use this in any shared objects
11705 Despite the nomenclature, @code{default} always means public ie;
11706 available to be linked against from outside the shared object.
11707 @code{protected} and @code{internal} are pretty useless in real-world
11708 usage so the only other commonly used option will be @code{hidden}.
11709 The default if -fvisibility isn't specified is @code{default} ie; make every
11710 symbol public - this causes the same behaviour as previous versions of
11713 A good explanation of the benefits offered by ensuring ELF
11714 symbols have the correct visibility is given by ``How To Write
11715 Shared Libraries'' by Ulrich Drepper (which can be found at
11716 @w{@uref{http://people.redhat.com/~drepper/}}) - however a superior
11717 solution made possible by this option to marking things hidden when
11718 the default is public is to make the default hidden and mark things
11719 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
11720 and @code{__attribute__ ((visibility("default")))} instead of
11721 @code{__declspec(dllexport)} you get almost identical semantics with
11722 identical syntax. This is a great boon to those working with
11723 cross-platform projects.
11725 For those adding visibility support to existing code, you may find
11726 @samp{#pragma GCC visibility} of use. This works by you enclosing
11727 the declarations you wish to set visibility for with (for example)
11728 @samp{#pragma GCC visibility push(hidden)} and
11729 @samp{#pragma GCC visibility pop}. These can be nested up to sixteen
11730 times. Bear in mind that symbol visibility should be viewed @strong{as
11731 part of the API interface contract} and thus all new code should
11732 always specify visibility when it is not the default ie; declarations
11733 only for use within the local DSO should @strong{always} be marked explicitly
11734 as hidden as so to avoid PLT indirection overheads - making this
11735 abundantly clear also aids readability and self-documentation of the code.
11736 Note that due to ISO C++ specification requirements, operator new and
11737 operator delete must always be of default visibility.
11739 An overview of these techniques, their benefits and how to use them
11740 is at @w{@uref{http://www.nedprod.com/programs/gccvisibility.html}}.
11746 @node Environment Variables
11747 @section Environment Variables Affecting GCC
11748 @cindex environment variables
11750 @c man begin ENVIRONMENT
11751 This section describes several environment variables that affect how GCC
11752 operates. Some of them work by specifying directories or prefixes to use
11753 when searching for various kinds of files. Some are used to specify other
11754 aspects of the compilation environment.
11756 Note that you can also specify places to search using options such as
11757 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
11758 take precedence over places specified using environment variables, which
11759 in turn take precedence over those specified by the configuration of GCC@.
11760 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
11761 GNU Compiler Collection (GCC) Internals}.
11766 @c @itemx LC_COLLATE
11768 @c @itemx LC_MONETARY
11769 @c @itemx LC_NUMERIC
11774 @c @findex LC_COLLATE
11775 @findex LC_MESSAGES
11776 @c @findex LC_MONETARY
11777 @c @findex LC_NUMERIC
11781 These environment variables control the way that GCC uses
11782 localization information that allow GCC to work with different
11783 national conventions. GCC inspects the locale categories
11784 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
11785 so. These locale categories can be set to any value supported by your
11786 installation. A typical value is @samp{en_UK} for English in the United
11789 The @env{LC_CTYPE} environment variable specifies character
11790 classification. GCC uses it to determine the character boundaries in
11791 a string; this is needed for some multibyte encodings that contain quote
11792 and escape characters that would otherwise be interpreted as a string
11795 The @env{LC_MESSAGES} environment variable specifies the language to
11796 use in diagnostic messages.
11798 If the @env{LC_ALL} environment variable is set, it overrides the value
11799 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
11800 and @env{LC_MESSAGES} default to the value of the @env{LANG}
11801 environment variable. If none of these variables are set, GCC
11802 defaults to traditional C English behavior.
11806 If @env{TMPDIR} is set, it specifies the directory to use for temporary
11807 files. GCC uses temporary files to hold the output of one stage of
11808 compilation which is to be used as input to the next stage: for example,
11809 the output of the preprocessor, which is the input to the compiler
11812 @item GCC_EXEC_PREFIX
11813 @findex GCC_EXEC_PREFIX
11814 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
11815 names of the subprograms executed by the compiler. No slash is added
11816 when this prefix is combined with the name of a subprogram, but you can
11817 specify a prefix that ends with a slash if you wish.
11819 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
11820 an appropriate prefix to use based on the pathname it was invoked with.
11822 If GCC cannot find the subprogram using the specified prefix, it
11823 tries looking in the usual places for the subprogram.
11825 The default value of @env{GCC_EXEC_PREFIX} is
11826 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
11827 of @code{prefix} when you ran the @file{configure} script.
11829 Other prefixes specified with @option{-B} take precedence over this prefix.
11831 This prefix is also used for finding files such as @file{crt0.o} that are
11834 In addition, the prefix is used in an unusual way in finding the
11835 directories to search for header files. For each of the standard
11836 directories whose name normally begins with @samp{/usr/local/lib/gcc}
11837 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
11838 replacing that beginning with the specified prefix to produce an
11839 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
11840 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
11841 These alternate directories are searched first; the standard directories
11844 @item COMPILER_PATH
11845 @findex COMPILER_PATH
11846 The value of @env{COMPILER_PATH} is a colon-separated list of
11847 directories, much like @env{PATH}. GCC tries the directories thus
11848 specified when searching for subprograms, if it can't find the
11849 subprograms using @env{GCC_EXEC_PREFIX}.
11852 @findex LIBRARY_PATH
11853 The value of @env{LIBRARY_PATH} is a colon-separated list of
11854 directories, much like @env{PATH}. When configured as a native compiler,
11855 GCC tries the directories thus specified when searching for special
11856 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
11857 using GCC also uses these directories when searching for ordinary
11858 libraries for the @option{-l} option (but directories specified with
11859 @option{-L} come first).
11863 @cindex locale definition
11864 This variable is used to pass locale information to the compiler. One way in
11865 which this information is used is to determine the character set to be used
11866 when character literals, string literals and comments are parsed in C and C++.
11867 When the compiler is configured to allow multibyte characters,
11868 the following values for @env{LANG} are recognized:
11872 Recognize JIS characters.
11874 Recognize SJIS characters.
11876 Recognize EUCJP characters.
11879 If @env{LANG} is not defined, or if it has some other value, then the
11880 compiler will use mblen and mbtowc as defined by the default locale to
11881 recognize and translate multibyte characters.
11885 Some additional environments variables affect the behavior of the
11888 @include cppenv.texi
11892 @node Precompiled Headers
11893 @section Using Precompiled Headers
11894 @cindex precompiled headers
11895 @cindex speed of compilation
11897 Often large projects have many header files that are included in every
11898 source file. The time the compiler takes to process these header files
11899 over and over again can account for nearly all of the time required to
11900 build the project. To make builds faster, GCC allows users to
11901 `precompile' a header file; then, if builds can use the precompiled
11902 header file they will be much faster.
11904 @strong{Caution:} There are a few known situations where GCC will
11905 crash when trying to use a precompiled header. If you have trouble
11906 with a precompiled header, you should remove the precompiled header
11907 and compile without it. In addition, please use GCC's on-line
11908 defect-tracking system to report any problems you encounter with
11909 precompiled headers. @xref{Bugs}.
11911 To create a precompiled header file, simply compile it as you would any
11912 other file, if necessary using the @option{-x} option to make the driver
11913 treat it as a C or C++ header file. You will probably want to use a
11914 tool like @command{make} to keep the precompiled header up-to-date when
11915 the headers it contains change.
11917 A precompiled header file will be searched for when @code{#include} is
11918 seen in the compilation. As it searches for the included file
11919 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
11920 compiler looks for a precompiled header in each directory just before it
11921 looks for the include file in that directory. The name searched for is
11922 the name specified in the @code{#include} with @samp{.gch} appended. If
11923 the precompiled header file can't be used, it is ignored.
11925 For instance, if you have @code{#include "all.h"}, and you have
11926 @file{all.h.gch} in the same directory as @file{all.h}, then the
11927 precompiled header file will be used if possible, and the original
11928 header will be used otherwise.
11930 Alternatively, you might decide to put the precompiled header file in a
11931 directory and use @option{-I} to ensure that directory is searched
11932 before (or instead of) the directory containing the original header.
11933 Then, if you want to check that the precompiled header file is always
11934 used, you can put a file of the same name as the original header in this
11935 directory containing an @code{#error} command.
11937 This also works with @option{-include}. So yet another way to use
11938 precompiled headers, good for projects not designed with precompiled
11939 header files in mind, is to simply take most of the header files used by
11940 a project, include them from another header file, precompile that header
11941 file, and @option{-include} the precompiled header. If the header files
11942 have guards against multiple inclusion, they will be skipped because
11943 they've already been included (in the precompiled header).
11945 If you need to precompile the same header file for different
11946 languages, targets, or compiler options, you can instead make a
11947 @emph{directory} named like @file{all.h.gch}, and put each precompiled
11948 header in the directory, perhaps using @option{-o}. It doesn't matter
11949 what you call the files in the directory, every precompiled header in
11950 the directory will be considered. The first precompiled header
11951 encountered in the directory that is valid for this compilation will
11952 be used; they're searched in no particular order.
11954 There are many other possibilities, limited only by your imagination,
11955 good sense, and the constraints of your build system.
11957 A precompiled header file can be used only when these conditions apply:
11961 Only one precompiled header can be used in a particular compilation.
11964 A precompiled header can't be used once the first C token is seen. You
11965 can have preprocessor directives before a precompiled header; you can
11966 even include a precompiled header from inside another header, so long as
11967 there are no C tokens before the @code{#include}.
11970 The precompiled header file must be produced for the same language as
11971 the current compilation. You can't use a C precompiled header for a C++
11975 The precompiled header file must be produced by the same compiler
11976 version and configuration as the current compilation is using.
11977 The easiest way to guarantee this is to use the same compiler binary
11978 for creating and using precompiled headers.
11981 Any macros defined before the precompiled header is included must
11982 either be defined in the same way as when the precompiled header was
11983 generated, or must not affect the precompiled header, which usually
11984 means that the they don't appear in the precompiled header at all.
11986 The @option{-D} option is one way to define a macro before a
11987 precompiled header is included; using a @code{#define} can also do it.
11988 There are also some options that define macros implicitly, like
11989 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
11992 @item If debugging information is output when using the precompiled
11993 header, using @option{-g} or similar, the same kind of debugging information
11994 must have been output when building the precompiled header. However,
11995 a precompiled header built using @option{-g} can be used in a compilation
11996 when no debugging information is being output.
11998 @item The same @option{-m} options must generally be used when building
11999 and using the precompiled header. @xref{Submodel Options},
12000 for any cases where this rule is relaxed.
12002 @item Each of the following options must be the same when building and using
12003 the precompiled header:
12005 @gccoptlist{-fexceptions -funit-at-a-time}
12008 Some other command-line options starting with @option{-f},
12009 @option{-p}, or @option{-O} must be defined in the same way as when
12010 the precompiled header was generated. At present, it's not clear
12011 which options are safe to change and which are not; the safest choice
12012 is to use exactly the same options when generating and using the
12013 precompiled header. The following are known to be safe:
12015 @gccoptlist{-fpreprocessed -pedantic-errors}
12019 For all of these except the last, the compiler will automatically
12020 ignore the precompiled header if the conditions aren't met. If you
12021 find an option combination that doesn't work and doesn't cause the
12022 precompiled header to be ignored, please consider filing a bug report,
12025 If you do use differing options when generating and using the
12026 precompiled header, the actual behaviour will be a mixture of the
12027 behaviour for the options. For instance, if you use @option{-g} to
12028 generate the precompiled header but not when using it, you may or may
12029 not get debugging information for routines in the precompiled header.
12031 @node Running Protoize
12032 @section Running Protoize
12034 The program @code{protoize} is an optional part of GCC@. You can use
12035 it to add prototypes to a program, thus converting the program to ISO
12036 C in one respect. The companion program @code{unprotoize} does the
12037 reverse: it removes argument types from any prototypes that are found.
12039 When you run these programs, you must specify a set of source files as
12040 command line arguments. The conversion programs start out by compiling
12041 these files to see what functions they define. The information gathered
12042 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
12044 After scanning comes actual conversion. The specified files are all
12045 eligible to be converted; any files they include (whether sources or
12046 just headers) are eligible as well.
12048 But not all the eligible files are converted. By default,
12049 @code{protoize} and @code{unprotoize} convert only source and header
12050 files in the current directory. You can specify additional directories
12051 whose files should be converted with the @option{-d @var{directory}}
12052 option. You can also specify particular files to exclude with the
12053 @option{-x @var{file}} option. A file is converted if it is eligible, its
12054 directory name matches one of the specified directory names, and its
12055 name within the directory has not been excluded.
12057 Basic conversion with @code{protoize} consists of rewriting most
12058 function definitions and function declarations to specify the types of
12059 the arguments. The only ones not rewritten are those for varargs
12062 @code{protoize} optionally inserts prototype declarations at the
12063 beginning of the source file, to make them available for any calls that
12064 precede the function's definition. Or it can insert prototype
12065 declarations with block scope in the blocks where undeclared functions
12068 Basic conversion with @code{unprotoize} consists of rewriting most
12069 function declarations to remove any argument types, and rewriting
12070 function definitions to the old-style pre-ISO form.
12072 Both conversion programs print a warning for any function declaration or
12073 definition that they can't convert. You can suppress these warnings
12076 The output from @code{protoize} or @code{unprotoize} replaces the
12077 original source file. The original file is renamed to a name ending
12078 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
12079 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
12080 for DOS) file already exists, then the source file is simply discarded.
12082 @code{protoize} and @code{unprotoize} both depend on GCC itself to
12083 scan the program and collect information about the functions it uses.
12084 So neither of these programs will work until GCC is installed.
12086 Here is a table of the options you can use with @code{protoize} and
12087 @code{unprotoize}. Each option works with both programs unless
12091 @item -B @var{directory}
12092 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
12093 usual directory (normally @file{/usr/local/lib}). This file contains
12094 prototype information about standard system functions. This option
12095 applies only to @code{protoize}.
12097 @item -c @var{compilation-options}
12098 Use @var{compilation-options} as the options when running @command{gcc} to
12099 produce the @samp{.X} files. The special option @option{-aux-info} is
12100 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
12102 Note that the compilation options must be given as a single argument to
12103 @code{protoize} or @code{unprotoize}. If you want to specify several
12104 @command{gcc} options, you must quote the entire set of compilation options
12105 to make them a single word in the shell.
12107 There are certain @command{gcc} arguments that you cannot use, because they
12108 would produce the wrong kind of output. These include @option{-g},
12109 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
12110 the @var{compilation-options}, they are ignored.
12113 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
12114 systems) instead of @samp{.c}. This is convenient if you are converting
12115 a C program to C++. This option applies only to @code{protoize}.
12118 Add explicit global declarations. This means inserting explicit
12119 declarations at the beginning of each source file for each function
12120 that is called in the file and was not declared. These declarations
12121 precede the first function definition that contains a call to an
12122 undeclared function. This option applies only to @code{protoize}.
12124 @item -i @var{string}
12125 Indent old-style parameter declarations with the string @var{string}.
12126 This option applies only to @code{protoize}.
12128 @code{unprotoize} converts prototyped function definitions to old-style
12129 function definitions, where the arguments are declared between the
12130 argument list and the initial @samp{@{}. By default, @code{unprotoize}
12131 uses five spaces as the indentation. If you want to indent with just
12132 one space instead, use @option{-i " "}.
12135 Keep the @samp{.X} files. Normally, they are deleted after conversion
12139 Add explicit local declarations. @code{protoize} with @option{-l} inserts
12140 a prototype declaration for each function in each block which calls the
12141 function without any declaration. This option applies only to
12145 Make no real changes. This mode just prints information about the conversions
12146 that would have been done without @option{-n}.
12149 Make no @samp{.save} files. The original files are simply deleted.
12150 Use this option with caution.
12152 @item -p @var{program}
12153 Use the program @var{program} as the compiler. Normally, the name
12154 @file{gcc} is used.
12157 Work quietly. Most warnings are suppressed.
12160 Print the version number, just like @option{-v} for @command{gcc}.
12163 If you need special compiler options to compile one of your program's
12164 source files, then you should generate that file's @samp{.X} file
12165 specially, by running @command{gcc} on that source file with the
12166 appropriate options and the option @option{-aux-info}. Then run
12167 @code{protoize} on the entire set of files. @code{protoize} will use
12168 the existing @samp{.X} file because it is newer than the source file.
12172 gcc -Dfoo=bar file1.c -aux-info file1.X
12177 You need to include the special files along with the rest in the
12178 @code{protoize} command, even though their @samp{.X} files already
12179 exist, because otherwise they won't get converted.
12181 @xref{Protoize Caveats}, for more information on how to use
12182 @code{protoize} successfully.