1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
8 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997,
9 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
11 Permission is granted to copy, distribute and/or modify this document
12 under the terms of the GNU Free Documentation License, Version 1.2 or
13 any later version published by the Free Software Foundation; with the
14 Invariant Sections being ``GNU General Public License'' and ``Funding
15 Free Software'', the Front-Cover texts being (a) (see below), and with
16 the Back-Cover Texts being (b) (see below). A copy of the license is
17 included in the gfdl(7) man page.
19 (a) The FSF's Front-Cover Text is:
23 (b) The FSF's Back-Cover Text is:
25 You have freedom to copy and modify this GNU Manual, like GNU
26 software. Copies published by the Free Software Foundation raise
27 funds for GNU development.
29 @c Set file name and title for the man page.
31 @settitle GNU project C and C++ compiler
33 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
34 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
35 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
36 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
37 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
38 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
39 [@option{-o} @var{outfile}] @var{infile}@dots{}
41 Only the most useful options are listed here; see below for the
42 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
45 gpl(7), gfdl(7), fsf-funding(7),
46 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
47 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
48 @file{ld}, @file{binutils} and @file{gdb}.
51 For instructions on reporting bugs, see
52 @w{@uref{http://gcc.gnu.org/bugs.html}}.
55 See the Info entry for @command{gcc}, or
56 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
57 for contributors to GCC@.
62 @chapter GCC Command Options
63 @cindex GCC command options
64 @cindex command options
65 @cindex options, GCC command
67 @c man begin DESCRIPTION
68 When you invoke GCC, it normally does preprocessing, compilation,
69 assembly and linking. The ``overall options'' allow you to stop this
70 process at an intermediate stage. For example, the @option{-c} option
71 says not to run the linker. Then the output consists of object files
72 output by the assembler.
74 Other options are passed on to one stage of processing. Some options
75 control the preprocessor and others the compiler itself. Yet other
76 options control the assembler and linker; most of these are not
77 documented here, since you rarely need to use any of them.
79 @cindex C compilation options
80 Most of the command line options that you can use with GCC are useful
81 for C programs; when an option is only useful with another language
82 (usually C++), the explanation says so explicitly. If the description
83 for a particular option does not mention a source language, you can use
84 that option with all supported languages.
86 @cindex C++ compilation options
87 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
88 options for compiling C++ programs.
90 @cindex grouping options
91 @cindex options, grouping
92 The @command{gcc} program accepts options and file names as operands. Many
93 options have multi-letter names; therefore multiple single-letter options
94 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
97 @cindex order of options
98 @cindex options, order
99 You can mix options and other arguments. For the most part, the order
100 you use doesn't matter. Order does matter when you use several options
101 of the same kind; for example, if you specify @option{-L} more than once,
102 the directories are searched in the order specified.
104 Many options have long names starting with @samp{-f} or with
105 @samp{-W}---for example, @option{-fforce-mem},
106 @option{-fstrength-reduce}, @option{-Wformat} and so on. Most of
107 these have both positive and negative forms; the negative form of
108 @option{-ffoo} would be @option{-fno-foo}. This manual documents
109 only one of these two forms, whichever one is not the default.
113 @xref{Option Index}, for an index to GCC's options.
116 * Option Summary:: Brief list of all options, without explanations.
117 * Overall Options:: Controlling the kind of output:
118 an executable, object files, assembler files,
119 or preprocessed source.
120 * Invoking G++:: Compiling C++ programs.
121 * C Dialect Options:: Controlling the variant of C language compiled.
122 * C++ Dialect Options:: Variations on C++.
123 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
125 * Language Independent Options:: Controlling how diagnostics should be
127 * Warning Options:: How picky should the compiler be?
128 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
129 * Optimize Options:: How much optimization?
130 * Preprocessor Options:: Controlling header files and macro definitions.
131 Also, getting dependency information for Make.
132 * Assembler Options:: Passing options to the assembler.
133 * Link Options:: Specifying libraries and so on.
134 * Directory Options:: Where to find header files and libraries.
135 Where to find the compiler executable files.
136 * Spec Files:: How to pass switches to sub-processes.
137 * Target Options:: Running a cross-compiler, or an old version of GCC.
138 * Submodel Options:: Specifying minor hardware or convention variations,
139 such as 68010 vs 68020.
140 * Code Gen Options:: Specifying conventions for function calls, data layout
142 * Environment Variables:: Env vars that affect GCC.
143 * Precompiled Headers:: Compiling a header once, and using it many times.
144 * Running Protoize:: Automatically adding or removing function prototypes.
150 @section Option Summary
152 Here is a summary of all the options, grouped by type. Explanations are
153 in the following sections.
156 @item Overall Options
157 @xref{Overall Options,,Options Controlling the Kind of Output}.
158 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
159 -x @var{language} -v -### --help --target-help --version}
161 @item C Language Options
162 @xref{C Dialect Options,,Options Controlling C Dialect}.
163 @gccoptlist{-ansi -std=@var{standard} -aux-info @var{filename} @gol
164 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
165 -fhosted -ffreestanding -fms-extensions @gol
166 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
167 -fallow-single-precision -fcond-mismatch @gol
168 -fsigned-bitfields -fsigned-char @gol
169 -funsigned-bitfields -funsigned-char}
171 @item C++ Language Options
172 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
173 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
174 -fconserve-space -fno-const-strings @gol
175 -fno-elide-constructors @gol
176 -fno-enforce-eh-specs @gol
177 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
178 -fno-implicit-templates @gol
179 -fno-implicit-inline-templates @gol
180 -fno-implement-inlines -fms-extensions @gol
181 -fno-nonansi-builtins -fno-operator-names @gol
182 -fno-optional-diags -fpermissive @gol
183 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
184 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
185 -fno-default-inline -fvisibility-inlines-hidden @gol
186 -Wabi -Wctor-dtor-privacy @gol
187 -Wnon-virtual-dtor -Wreorder @gol
188 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel @gol
189 -Wno-non-template-friend -Wold-style-cast @gol
190 -Woverloaded-virtual -Wno-pmf-conversions @gol
193 @item Objective-C and Objective-C++ Language Options
194 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
195 Objective-C and Objective-C++ Dialects}.
197 -fconstant-string-class=@var{class-name} @gol
198 -fgnu-runtime -fnext-runtime @gol
199 -fno-nil-receivers @gol
200 -fobjc-call-cxx-cdtors @gol
201 -fobjc-direct-dispatch @gol
202 -fobjc-exceptions @gol
204 -freplace-objc-classes @gol
207 -Wassign-intercept @gol
208 -Wno-protocol -Wselector @gol
209 -Wstrict-selector-match @gol
210 -Wundeclared-selector}
212 @item Language Independent Options
213 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
214 @gccoptlist{-fmessage-length=@var{n} @gol
215 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}} @gol
216 -fdiagnostics-show-options
218 @item Warning Options
219 @xref{Warning Options,,Options to Request or Suppress Warnings}.
220 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
221 -w -Wextra -Wall -Waggregate-return @gol
222 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
223 -Wconversion -Wno-deprecated-declarations @gol
224 -Wdisabled-optimization -Wno-div-by-zero -Wno-endif-labels @gol
225 -Werror -Werror-implicit-function-declaration @gol
226 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
227 -Wno-format-extra-args -Wformat-nonliteral @gol
228 -Wformat-security -Wformat-y2k @gol
229 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
230 -Wimport -Wno-import -Winit-self -Winline @gol
231 -Wno-int-to-pointer-cast @gol
232 -Wno-invalid-offsetof -Winvalid-pch @gol
233 -Wlarger-than-@var{len} -Wlong-long @gol
234 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
235 -Wmissing-format-attribute -Wmissing-include-dirs @gol
236 -Wmissing-noreturn @gol
237 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
238 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
239 -Wredundant-decls @gol
240 -Wreturn-type -Wsequence-point -Wshadow @gol
241 -Wsign-compare -Wstrict-aliasing -Wstrict-aliasing=2 @gol
242 -Wswitch -Wswitch-default -Wswitch-enum @gol
243 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
244 -Wunknown-pragmas -Wunreachable-code @gol
245 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
246 -Wunused-value -Wunused-variable -Wwrite-strings @gol
249 @item C-only Warning Options
250 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
251 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
252 -Wstrict-prototypes -Wtraditional @gol
253 -Wdeclaration-after-statement -Wno-pointer-sign}
255 @item Debugging Options
256 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
257 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
258 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
259 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
260 -fdump-ipa-all -fdump-ipa-cgraph @gol
262 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
263 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
264 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
267 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
268 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
269 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
270 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
271 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
272 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
273 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
274 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
275 -fdump-tree-nrv -fdump-tree-vect @gol
276 -fdump-tree-sink @gol
277 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
278 -fdump-tree-salias @gol
279 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
280 -ftree-vectorizer-verbose=@var{n} @gol
281 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
282 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
283 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs -ftree-based-profiling @gol
284 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
285 -ftest-coverage -ftime-report -fvar-tracking @gol
286 -g -g@var{level} -gcoff -gdwarf-2 @gol
287 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
288 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
289 -print-multi-directory -print-multi-lib @gol
290 -print-prog-name=@var{program} -print-search-dirs -Q @gol
293 @item Optimization Options
294 @xref{Optimize Options,,Options that Control Optimization}.
295 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
296 -falign-labels=@var{n} -falign-loops=@var{n} @gol
297 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
298 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
299 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
300 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
301 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
302 -fdelayed-branch -fdelete-null-pointer-checks @gol
303 -fexpensive-optimizations -ffast-math -ffloat-store @gol
304 -fforce-addr -fforce-mem -ffunction-sections @gol
305 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
306 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2 @gol
307 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
308 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
309 -fmodulo-sched -fno-branch-count-reg @gol
310 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants @gol
311 -fno-function-cse -fno-guess-branch-probability @gol
312 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
313 -funsafe-math-optimizations -ffinite-math-only @gol
314 -fno-trapping-math -fno-zero-initialized-in-bss @gol
315 -fomit-frame-pointer -foptimize-register-move @gol
316 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
317 -fprofile-generate -fprofile-use @gol
318 -fregmove -frename-registers @gol
319 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
320 -frerun-cse-after-loop -frerun-loop-opt @gol
321 -frounding-math -fschedule-insns -fschedule-insns2 @gol
322 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
323 -fsched-spec-load-dangerous @gol
324 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
325 -fsched2-use-superblocks @gol
326 -fsched2-use-traces -freschedule-modulo-scheduled-loops @gol
327 -fsignaling-nans -fsingle-precision-constant -fspeculative-prefetching @gol
328 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
329 -funroll-all-loops -funroll-loops -fpeel-loops @gol
330 -fsplit-ivs-in-unroller -funswitch-loops @gol
331 -fvariable-expansion-in-unroller @gol
332 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
333 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
334 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
335 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
336 -ftree-salias -fweb @gol
337 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop @gol
338 --param @var{name}=@var{value}
339 -O -O0 -O1 -O2 -O3 -Os}
341 @item Preprocessor Options
342 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
343 @gccoptlist{-A@var{question}=@var{answer} @gol
344 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
345 -C -dD -dI -dM -dN @gol
346 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
347 -idirafter @var{dir} @gol
348 -include @var{file} -imacros @var{file} @gol
349 -iprefix @var{file} -iwithprefix @var{dir} @gol
350 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
351 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
352 -P -fworking-directory -remap @gol
353 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
354 -Xpreprocessor @var{option}}
356 @item Assembler Option
357 @xref{Assembler Options,,Passing Options to the Assembler}.
358 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
361 @xref{Link Options,,Options for Linking}.
362 @gccoptlist{@var{object-file-name} -l@var{library} @gol
363 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
364 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
365 -Wl,@var{option} -Xlinker @var{option} @gol
368 @item Directory Options
369 @xref{Directory Options,,Options for Directory Search}.
370 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-}
373 @c I wrote this xref this way to avoid overfull hbox. -- rms
374 @xref{Target Options}.
375 @gccoptlist{-V @var{version} -b @var{machine}}
377 @item Machine Dependent Options
378 @xref{Submodel Options,,Hardware Models and Configurations}.
379 @c This list is ordered alphanumerically by subsection name.
380 @c Try and put the significant identifier (CPU or system) first,
381 @c so users have a clue at guessing where the ones they want will be.
384 @gccoptlist{-EB -EL @gol
385 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
386 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
389 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
390 -mabi=@var{name} @gol
391 -mapcs-stack-check -mno-apcs-stack-check @gol
392 -mapcs-float -mno-apcs-float @gol
393 -mapcs-reentrant -mno-apcs-reentrant @gol
394 -msched-prolog -mno-sched-prolog @gol
395 -mlittle-endian -mbig-endian -mwords-little-endian @gol
396 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
397 -mthumb-interwork -mno-thumb-interwork @gol
398 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
399 -mstructure-size-boundary=@var{n} @gol
400 -mabort-on-noreturn @gol
401 -mlong-calls -mno-long-calls @gol
402 -msingle-pic-base -mno-single-pic-base @gol
403 -mpic-register=@var{reg} @gol
404 -mnop-fun-dllimport @gol
405 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
406 -mpoke-function-name @gol
408 -mtpcs-frame -mtpcs-leaf-frame @gol
409 -mcaller-super-interworking -mcallee-super-interworking}
412 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
413 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
415 @emph{Blackfin Options}
416 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer -mcsync @gol
417 -mno-csync -mlow-64k -mno-low64k -mid-shared-library @gol
418 -mno-id-shared-library -mshared-library-id=@var{n} @gol}
421 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
422 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
423 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
424 -mstack-align -mdata-align -mconst-align @gol
425 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
426 -melf -maout -melinux -mlinux -sim -sim2 @gol
427 -mmul-bug-workaround -mno-mul-bug-workaround}
429 @emph{Darwin Options}
430 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
431 -arch_only -bind_at_load -bundle -bundle_loader @gol
432 -client_name -compatibility_version -current_version @gol
434 -dependency-file -dylib_file -dylinker_install_name @gol
435 -dynamic -dynamiclib -exported_symbols_list @gol
436 -filelist -flat_namespace -force_cpusubtype_ALL @gol
437 -force_flat_namespace -headerpad_max_install_names @gol
438 -image_base -init -install_name -keep_private_externs @gol
439 -multi_module -multiply_defined -multiply_defined_unused @gol
440 -noall_load -no_dead_strip_inits_and_terms @gol
441 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
442 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
443 -private_bundle -read_only_relocs -sectalign @gol
444 -sectobjectsymbols -whyload -seg1addr @gol
445 -sectcreate -sectobjectsymbols -sectorder @gol
446 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
447 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
448 -segprot -segs_read_only_addr -segs_read_write_addr @gol
449 -single_module -static -sub_library -sub_umbrella @gol
450 -twolevel_namespace -umbrella -undefined @gol
451 -unexported_symbols_list -weak_reference_mismatches @gol
452 -whatsloaded -F -gused -gfull -mone-byte-bool}
454 @emph{DEC Alpha Options}
455 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
456 -mieee -mieee-with-inexact -mieee-conformant @gol
457 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
458 -mtrap-precision=@var{mode} -mbuild-constants @gol
459 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
460 -mbwx -mmax -mfix -mcix @gol
461 -mfloat-vax -mfloat-ieee @gol
462 -mexplicit-relocs -msmall-data -mlarge-data @gol
463 -msmall-text -mlarge-text @gol
464 -mmemory-latency=@var{time}}
466 @emph{DEC Alpha/VMS Options}
467 @gccoptlist{-mvms-return-codes}
470 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
471 -mhard-float -msoft-float @gol
472 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
473 -mdouble -mno-double @gol
474 -mmedia -mno-media -mmuladd -mno-muladd @gol
475 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
476 -mlinked-fp -mlong-calls -malign-labels @gol
477 -mlibrary-pic -macc-4 -macc-8 @gol
478 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
479 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
480 -mvliw-branch -mno-vliw-branch @gol
481 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
482 -mno-nested-cond-exec -mtomcat-stats @gol
486 @emph{H8/300 Options}
487 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
490 @gccoptlist{-march=@var{architecture-type} @gol
491 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
492 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
493 -mfixed-range=@var{register-range} @gol
494 -mjump-in-delay -mlinker-opt -mlong-calls @gol
495 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
496 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
497 -mno-jump-in-delay -mno-long-load-store @gol
498 -mno-portable-runtime -mno-soft-float @gol
499 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
500 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
501 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
502 -munix=@var{unix-std} -nolibdld -static -threads}
504 @emph{i386 and x86-64 Options}
505 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
506 -mfpmath=@var{unit} @gol
507 -masm=@var{dialect} -mno-fancy-math-387 @gol
508 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
509 -mno-wide-multiply -mrtd -malign-double @gol
510 -mpreferred-stack-boundary=@var{num} @gol
511 -mmmx -msse -msse2 -msse3 -m3dnow @gol
512 -mthreads -mno-align-stringops -minline-all-stringops @gol
513 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
514 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
515 -mno-red-zone -mno-tls-direct-seg-refs @gol
516 -mcmodel=@var{code-model} @gol
520 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
521 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
522 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
523 -minline-float-divide-max-throughput @gol
524 -minline-int-divide-min-latency @gol
525 -minline-int-divide-max-throughput @gol
526 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
527 -mno-dwarf2-asm -mearly-stop-bits @gol
528 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
529 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64}
531 @emph{M32R/D Options}
532 @gccoptlist{-m32r2 -m32rx -m32r @gol
534 -malign-loops -mno-align-loops @gol
535 -missue-rate=@var{number} @gol
536 -mbranch-cost=@var{number} @gol
537 -mmodel=@var{code-size-model-type} @gol
538 -msdata=@var{sdata-type} @gol
539 -mno-flush-func -mflush-func=@var{name} @gol
540 -mno-flush-trap -mflush-trap=@var{number} @gol
543 @emph{M680x0 Options}
544 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
545 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
546 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
547 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
548 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
550 @emph{M68hc1x Options}
551 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
552 -mauto-incdec -minmax -mlong-calls -mshort @gol
553 -msoft-reg-count=@var{count}}
556 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
557 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
558 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
559 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
560 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
563 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
564 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
565 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
566 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
567 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
568 -mpaired-single -mips3d @gol
569 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
570 -G@var{num} -membedded-data -mno-embedded-data @gol
571 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
572 -msplit-addresses -mno-split-addresses @gol
573 -mexplicit-relocs -mno-explicit-relocs @gol
574 -mcheck-zero-division -mno-check-zero-division @gol
575 -mdivide-traps -mdivide-breaks @gol
576 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
577 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
578 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
579 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
580 -mfix-sb1 -mno-fix-sb1 @gol
581 -mflush-func=@var{func} -mno-flush-func @gol
582 -mbranch-likely -mno-branch-likely @gol
583 -mfp-exceptions -mno-fp-exceptions @gol
584 -mvr4130-align -mno-vr4130-align}
587 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
588 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
589 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
590 -mno-base-addresses -msingle-exit -mno-single-exit}
592 @emph{MN10300 Options}
593 @gccoptlist{-mmult-bug -mno-mult-bug @gol
594 -mam33 -mno-am33 @gol
595 -mam33-2 -mno-am33-2 @gol
599 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
600 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
601 -mregparam -mnoregparam -msb -mnosb @gol
602 -mbitfield -mnobitfield -mhimem -mnohimem}
604 @emph{PDP-11 Options}
605 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
606 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
607 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
608 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
609 -mbranch-expensive -mbranch-cheap @gol
610 -msplit -mno-split -munix-asm -mdec-asm}
612 @emph{PowerPC Options}
613 See RS/6000 and PowerPC Options.
615 @emph{RS/6000 and PowerPC Options}
616 @gccoptlist{-mcpu=@var{cpu-type} @gol
617 -mtune=@var{cpu-type} @gol
618 -mpower -mno-power -mpower2 -mno-power2 @gol
619 -mpowerpc -mpowerpc64 -mno-powerpc @gol
620 -maltivec -mno-altivec @gol
621 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
622 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
623 -mnew-mnemonics -mold-mnemonics @gol
624 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
625 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
626 -malign-power -malign-natural @gol
627 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
628 -mstring -mno-string -mupdate -mno-update @gol
629 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
630 -mstrict-align -mno-strict-align -mrelocatable @gol
631 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
632 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
633 -mdynamic-no-pic @gol
634 -mprioritize-restricted-insns=@var{priority} @gol
635 -msched-costly-dep=@var{dependence_type} @gol
636 -minsert-sched-nops=@var{scheme} @gol
637 -mcall-sysv -mcall-netbsd @gol
638 -maix-struct-return -msvr4-struct-return @gol
639 -mabi=@var{abi-type} @gol
640 -misel -mno-isel @gol
641 -misel=yes -misel=no @gol
643 -mspe=yes -mspe=no @gol
644 -mvrsave -mno-vrsave @gol
645 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
646 -mprototype -mno-prototype @gol
647 -msim -mmvme -mads -myellowknife -memb -msdata @gol
648 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
650 @emph{S/390 and zSeries Options}
651 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
652 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
653 -mpacked-stack -mno-packed-stack @gol
654 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
655 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
656 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
657 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
660 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
661 -m4-nofpu -m4-single-only -m4-single -m4 @gol
662 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
663 -m5-64media -m5-64media-nofpu @gol
664 -m5-32media -m5-32media-nofpu @gol
665 -m5-compact -m5-compact-nofpu @gol
666 -mb -ml -mdalign -mrelax @gol
667 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
668 -mieee -misize -mpadstruct -mspace @gol
669 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
670 -mdivsi3_libfunc=@var{name} @gol
671 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
675 @gccoptlist{-mcpu=@var{cpu-type} @gol
676 -mtune=@var{cpu-type} @gol
677 -mcmodel=@var{code-model} @gol
678 -m32 -m64 -mapp-regs -mno-app-regs @gol
679 -mfaster-structs -mno-faster-structs @gol
680 -mfpu -mno-fpu -mhard-float -msoft-float @gol
681 -mhard-quad-float -msoft-quad-float @gol
682 -mimpure-text -mno-impure-text -mlittle-endian @gol
683 -mstack-bias -mno-stack-bias @gol
684 -munaligned-doubles -mno-unaligned-doubles @gol
685 -mv8plus -mno-v8plus -mvis -mno-vis
688 @emph{System V Options}
689 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
691 @emph{TMS320C3x/C4x Options}
692 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
693 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
694 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
695 -mparallel-insns -mparallel-mpy -mpreserve-float}
698 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
699 -mprolog-function -mno-prolog-function -mspace @gol
700 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
701 -mapp-regs -mno-app-regs @gol
702 -mdisable-callt -mno-disable-callt @gol
708 @gccoptlist{-mg -mgnu -munix}
710 @emph{x86-64 Options}
711 See i386 and x86-64 Options.
713 @emph{Xstormy16 Options}
716 @emph{Xtensa Options}
717 @gccoptlist{-mconst16 -mno-const16 @gol
718 -mfused-madd -mno-fused-madd @gol
719 -mtext-section-literals -mno-text-section-literals @gol
720 -mtarget-align -mno-target-align @gol
721 -mlongcalls -mno-longcalls}
723 @emph{zSeries Options}
724 See S/390 and zSeries Options.
726 @item Code Generation Options
727 @xref{Code Gen Options,,Options for Code Generation Conventions}.
728 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
729 -ffixed-@var{reg} -fexceptions @gol
730 -fnon-call-exceptions -funwind-tables @gol
731 -fasynchronous-unwind-tables @gol
732 -finhibit-size-directive -finstrument-functions @gol
733 -fno-common -fno-ident @gol
734 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
735 -freg-struct-return -fshared-data -fshort-enums @gol
736 -fshort-double -fshort-wchar @gol
737 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
738 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
739 -fargument-alias -fargument-noalias @gol
740 -fargument-noalias-global -fleading-underscore @gol
741 -ftls-model=@var{model} @gol
742 -ftrapv -fwrapv -fbounds-check @gol
747 * Overall Options:: Controlling the kind of output:
748 an executable, object files, assembler files,
749 or preprocessed source.
750 * C Dialect Options:: Controlling the variant of C language compiled.
751 * C++ Dialect Options:: Variations on C++.
752 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
754 * Language Independent Options:: Controlling how diagnostics should be
756 * Warning Options:: How picky should the compiler be?
757 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
758 * Optimize Options:: How much optimization?
759 * Preprocessor Options:: Controlling header files and macro definitions.
760 Also, getting dependency information for Make.
761 * Assembler Options:: Passing options to the assembler.
762 * Link Options:: Specifying libraries and so on.
763 * Directory Options:: Where to find header files and libraries.
764 Where to find the compiler executable files.
765 * Spec Files:: How to pass switches to sub-processes.
766 * Target Options:: Running a cross-compiler, or an old version of GCC.
769 @node Overall Options
770 @section Options Controlling the Kind of Output
772 Compilation can involve up to four stages: preprocessing, compilation
773 proper, assembly and linking, always in that order. GCC is capable of
774 preprocessing and compiling several files either into several
775 assembler input files, or into one assembler input file; then each
776 assembler input file produces an object file, and linking combines all
777 the object files (those newly compiled, and those specified as input)
778 into an executable file.
780 @cindex file name suffix
781 For any given input file, the file name suffix determines what kind of
786 C source code which must be preprocessed.
789 C source code which should not be preprocessed.
792 C++ source code which should not be preprocessed.
795 Objective-C source code. Note that you must link with the @file{libobjc}
796 library to make an Objective-C program work.
799 Objective-C source code which should not be preprocessed.
803 Objective-C++ source code. Note that you must link with the @file{libobjc}
804 library to make an Objective-C++ program work. Note that @samp{.M} refers
805 to a literal capital M@.
808 Objective-C++ source code which should not be preprocessed.
811 C, C++, Objective-C or Objective-C++ header file to be turned into a
816 @itemx @var{file}.cxx
817 @itemx @var{file}.cpp
818 @itemx @var{file}.CPP
819 @itemx @var{file}.c++
821 C++ source code which must be preprocessed. Note that in @samp{.cxx},
822 the last two letters must both be literally @samp{x}. Likewise,
823 @samp{.C} refers to a literal capital C@.
827 Objective-C++ source code which must be preprocessed.
830 Objective-C++ source code which should not be preprocessed.
834 C++ header file to be turned into a precompiled header.
837 @itemx @var{file}.for
838 @itemx @var{file}.FOR
839 Fortran source code which should not be preprocessed.
842 @itemx @var{file}.fpp
843 @itemx @var{file}.FPP
844 Fortran source code which must be preprocessed (with the traditional
848 Fortran source code which must be preprocessed with a RATFOR
849 preprocessor (not included with GCC)@.
852 @itemx @var{file}.f95
853 Fortran 90/95 source code which should not be preprocessed.
855 @c FIXME: Descriptions of Java file types.
862 Ada source code file which contains a library unit declaration (a
863 declaration of a package, subprogram, or generic, or a generic
864 instantiation), or a library unit renaming declaration (a package,
865 generic, or subprogram renaming declaration). Such files are also
868 @itemx @var{file}.adb
869 Ada source code file containing a library unit body (a subprogram or
870 package body). Such files are also called @dfn{bodies}.
872 @c GCC also knows about some suffixes for languages not yet included:
881 Assembler code which must be preprocessed.
884 An object file to be fed straight into linking.
885 Any file name with no recognized suffix is treated this way.
889 You can specify the input language explicitly with the @option{-x} option:
892 @item -x @var{language}
893 Specify explicitly the @var{language} for the following input files
894 (rather than letting the compiler choose a default based on the file
895 name suffix). This option applies to all following input files until
896 the next @option{-x} option. Possible values for @var{language} are:
898 c c-header c-cpp-output
899 c++ c++-header c++-cpp-output
900 objective-c objective-c-header objective-c-cpp-output
901 objective-c++ objective-c++-header objective-c++-cpp-output
902 assembler assembler-with-cpp
904 f77 f77-cpp-input ratfor
911 Turn off any specification of a language, so that subsequent files are
912 handled according to their file name suffixes (as they are if @option{-x}
913 has not been used at all).
915 @item -pass-exit-codes
916 @opindex pass-exit-codes
917 Normally the @command{gcc} program will exit with the code of 1 if any
918 phase of the compiler returns a non-success return code. If you specify
919 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
920 numerically highest error produced by any phase that returned an error
924 If you only want some of the stages of compilation, you can use
925 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
926 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
927 @command{gcc} is to stop. Note that some combinations (for example,
928 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
933 Compile or assemble the source files, but do not link. The linking
934 stage simply is not done. The ultimate output is in the form of an
935 object file for each source file.
937 By default, the object file name for a source file is made by replacing
938 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
940 Unrecognized input files, not requiring compilation or assembly, are
945 Stop after the stage of compilation proper; do not assemble. The output
946 is in the form of an assembler code file for each non-assembler input
949 By default, the assembler file name for a source file is made by
950 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
952 Input files that don't require compilation are ignored.
956 Stop after the preprocessing stage; do not run the compiler proper. The
957 output is in the form of preprocessed source code, which is sent to the
960 Input files which don't require preprocessing are ignored.
962 @cindex output file option
965 Place output in file @var{file}. This applies regardless to whatever
966 sort of output is being produced, whether it be an executable file,
967 an object file, an assembler file or preprocessed C code.
969 If @option{-o} is not specified, the default is to put an executable
970 file in @file{a.out}, the object file for
971 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
972 assembler file in @file{@var{source}.s}, a precompiled header file in
973 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
978 Print (on standard error output) the commands executed to run the stages
979 of compilation. Also print the version number of the compiler driver
980 program and of the preprocessor and the compiler proper.
984 Like @option{-v} except the commands are not executed and all command
985 arguments are quoted. This is useful for shell scripts to capture the
986 driver-generated command lines.
990 Use pipes rather than temporary files for communication between the
991 various stages of compilation. This fails to work on some systems where
992 the assembler is unable to read from a pipe; but the GNU assembler has
997 If you are compiling multiple source files, this option tells the driver
998 to pass all the source files to the compiler at once (for those
999 languages for which the compiler can handle this). This will allow
1000 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1001 language for which this is supported is C@. If you pass source files for
1002 multiple languages to the driver, using this option, the driver will invoke
1003 the compiler(s) that support IMA once each, passing each compiler all the
1004 source files appropriate for it. For those languages that do not support
1005 IMA this option will be ignored, and the compiler will be invoked once for
1006 each source file in that language. If you use this option in conjunction
1007 with @option{-save-temps}, the compiler will generate multiple
1009 (one for each source file), but only one (combined) @file{.o} or
1014 Print (on the standard output) a description of the command line options
1015 understood by @command{gcc}. If the @option{-v} option is also specified
1016 then @option{--help} will also be passed on to the various processes
1017 invoked by @command{gcc}, so that they can display the command line options
1018 they accept. If the @option{-Wextra} option is also specified then command
1019 line options which have no documentation associated with them will also
1023 @opindex target-help
1024 Print (on the standard output) a description of target specific command
1025 line options for each tool.
1029 Display the version number and copyrights of the invoked GCC@.
1033 @section Compiling C++ Programs
1035 @cindex suffixes for C++ source
1036 @cindex C++ source file suffixes
1037 C++ source files conventionally use one of the suffixes @samp{.C},
1038 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1039 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1040 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1041 files with these names and compiles them as C++ programs even if you
1042 call the compiler the same way as for compiling C programs (usually
1043 with the name @command{gcc}).
1047 However, C++ programs often require class libraries as well as a
1048 compiler that understands the C++ language---and under some
1049 circumstances, you might want to compile programs or header files from
1050 standard input, or otherwise without a suffix that flags them as C++
1051 programs. You might also like to precompile a C header file with a
1052 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
1053 program that calls GCC with the default language set to C++, and
1054 automatically specifies linking against the C++ library. On many
1055 systems, @command{g++} is also installed with the name @command{c++}.
1057 @cindex invoking @command{g++}
1058 When you compile C++ programs, you may specify many of the same
1059 command-line options that you use for compiling programs in any
1060 language; or command-line options meaningful for C and related
1061 languages; or options that are meaningful only for C++ programs.
1062 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1063 explanations of options for languages related to C@.
1064 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1065 explanations of options that are meaningful only for C++ programs.
1067 @node C Dialect Options
1068 @section Options Controlling C Dialect
1069 @cindex dialect options
1070 @cindex language dialect options
1071 @cindex options, dialect
1073 The following options control the dialect of C (or languages derived
1074 from C, such as C++, Objective-C and Objective-C++) that the compiler
1078 @cindex ANSI support
1082 In C mode, support all ISO C90 programs. In C++ mode,
1083 remove GNU extensions that conflict with ISO C++.
1085 This turns off certain features of GCC that are incompatible with ISO
1086 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1087 such as the @code{asm} and @code{typeof} keywords, and
1088 predefined macros such as @code{unix} and @code{vax} that identify the
1089 type of system you are using. It also enables the undesirable and
1090 rarely used ISO trigraph feature. For the C compiler,
1091 it disables recognition of C++ style @samp{//} comments as well as
1092 the @code{inline} keyword.
1094 The alternate keywords @code{__asm__}, @code{__extension__},
1095 @code{__inline__} and @code{__typeof__} continue to work despite
1096 @option{-ansi}. You would not want to use them in an ISO C program, of
1097 course, but it is useful to put them in header files that might be included
1098 in compilations done with @option{-ansi}. Alternate predefined macros
1099 such as @code{__unix__} and @code{__vax__} are also available, with or
1100 without @option{-ansi}.
1102 The @option{-ansi} option does not cause non-ISO programs to be
1103 rejected gratuitously. For that, @option{-pedantic} is required in
1104 addition to @option{-ansi}. @xref{Warning Options}.
1106 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1107 option is used. Some header files may notice this macro and refrain
1108 from declaring certain functions or defining certain macros that the
1109 ISO standard doesn't call for; this is to avoid interfering with any
1110 programs that might use these names for other things.
1112 Functions which would normally be built in but do not have semantics
1113 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1114 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1115 built-in functions provided by GCC}, for details of the functions
1120 Determine the language standard. This option is currently only
1121 supported when compiling C or C++. A value for this option must be
1122 provided; possible values are
1127 ISO C90 (same as @option{-ansi}).
1129 @item iso9899:199409
1130 ISO C90 as modified in amendment 1.
1136 ISO C99. Note that this standard is not yet fully supported; see
1137 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1138 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1141 Default, ISO C90 plus GNU extensions (including some C99 features).
1145 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1146 this will become the default. The name @samp{gnu9x} is deprecated.
1149 The 1998 ISO C++ standard plus amendments.
1152 The same as @option{-std=c++98} plus GNU extensions. This is the
1153 default for C++ code.
1156 Even when this option is not specified, you can still use some of the
1157 features of newer standards in so far as they do not conflict with
1158 previous C standards. For example, you may use @code{__restrict__} even
1159 when @option{-std=c99} is not specified.
1161 The @option{-std} options specifying some version of ISO C have the same
1162 effects as @option{-ansi}, except that features that were not in ISO C90
1163 but are in the specified version (for example, @samp{//} comments and
1164 the @code{inline} keyword in ISO C99) are not disabled.
1166 @xref{Standards,,Language Standards Supported by GCC}, for details of
1167 these standard versions.
1169 @item -aux-info @var{filename}
1171 Output to the given filename prototyped declarations for all functions
1172 declared and/or defined in a translation unit, including those in header
1173 files. This option is silently ignored in any language other than C@.
1175 Besides declarations, the file indicates, in comments, the origin of
1176 each declaration (source file and line), whether the declaration was
1177 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1178 @samp{O} for old, respectively, in the first character after the line
1179 number and the colon), and whether it came from a declaration or a
1180 definition (@samp{C} or @samp{F}, respectively, in the following
1181 character). In the case of function definitions, a K&R-style list of
1182 arguments followed by their declarations is also provided, inside
1183 comments, after the declaration.
1187 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1188 keyword, so that code can use these words as identifiers. You can use
1189 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1190 instead. @option{-ansi} implies @option{-fno-asm}.
1192 In C++, this switch only affects the @code{typeof} keyword, since
1193 @code{asm} and @code{inline} are standard keywords. You may want to
1194 use the @option{-fno-gnu-keywords} flag instead, which has the same
1195 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1196 switch only affects the @code{asm} and @code{typeof} keywords, since
1197 @code{inline} is a standard keyword in ISO C99.
1200 @itemx -fno-builtin-@var{function}
1201 @opindex fno-builtin
1202 @cindex built-in functions
1203 Don't recognize built-in functions that do not begin with
1204 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1205 functions provided by GCC}, for details of the functions affected,
1206 including those which are not built-in functions when @option{-ansi} or
1207 @option{-std} options for strict ISO C conformance are used because they
1208 do not have an ISO standard meaning.
1210 GCC normally generates special code to handle certain built-in functions
1211 more efficiently; for instance, calls to @code{alloca} may become single
1212 instructions that adjust the stack directly, and calls to @code{memcpy}
1213 may become inline copy loops. The resulting code is often both smaller
1214 and faster, but since the function calls no longer appear as such, you
1215 cannot set a breakpoint on those calls, nor can you change the behavior
1216 of the functions by linking with a different library. In addition,
1217 when a function is recognized as a built-in function, GCC may use
1218 information about that function to warn about problems with calls to
1219 that function, or to generate more efficient code, even if the
1220 resulting code still contains calls to that function. For example,
1221 warnings are given with @option{-Wformat} for bad calls to
1222 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1223 known not to modify global memory.
1225 With the @option{-fno-builtin-@var{function}} option
1226 only the built-in function @var{function} is
1227 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1228 function is named this is not built-in in this version of GCC, this
1229 option is ignored. There is no corresponding
1230 @option{-fbuiltin-@var{function}} option; if you wish to enable
1231 built-in functions selectively when using @option{-fno-builtin} or
1232 @option{-ffreestanding}, you may define macros such as:
1235 #define abs(n) __builtin_abs ((n))
1236 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1241 @cindex hosted environment
1243 Assert that compilation takes place in a hosted environment. This implies
1244 @option{-fbuiltin}. A hosted environment is one in which the
1245 entire standard library is available, and in which @code{main} has a return
1246 type of @code{int}. Examples are nearly everything except a kernel.
1247 This is equivalent to @option{-fno-freestanding}.
1249 @item -ffreestanding
1250 @opindex ffreestanding
1251 @cindex hosted environment
1253 Assert that compilation takes place in a freestanding environment. This
1254 implies @option{-fno-builtin}. A freestanding environment
1255 is one in which the standard library may not exist, and program startup may
1256 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1257 This is equivalent to @option{-fno-hosted}.
1259 @xref{Standards,,Language Standards Supported by GCC}, for details of
1260 freestanding and hosted environments.
1262 @item -fms-extensions
1263 @opindex fms-extensions
1264 Accept some non-standard constructs used in Microsoft header files.
1266 Some cases of unnamed fields in structures and unions are only
1267 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1268 fields within structs/unions}, for details.
1272 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1273 options for strict ISO C conformance) implies @option{-trigraphs}.
1275 @item -no-integrated-cpp
1276 @opindex no-integrated-cpp
1277 Performs a compilation in two passes: preprocessing and compiling. This
1278 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1279 @option{-B} option. The user supplied compilation step can then add in
1280 an additional preprocessing step after normal preprocessing but before
1281 compiling. The default is to use the integrated cpp (internal cpp)
1283 The semantics of this option will change if "cc1", "cc1plus", and
1284 "cc1obj" are merged.
1286 @cindex traditional C language
1287 @cindex C language, traditional
1289 @itemx -traditional-cpp
1290 @opindex traditional-cpp
1291 @opindex traditional
1292 Formerly, these options caused GCC to attempt to emulate a pre-standard
1293 C compiler. They are now only supported with the @option{-E} switch.
1294 The preprocessor continues to support a pre-standard mode. See the GNU
1295 CPP manual for details.
1297 @item -fcond-mismatch
1298 @opindex fcond-mismatch
1299 Allow conditional expressions with mismatched types in the second and
1300 third arguments. The value of such an expression is void. This option
1301 is not supported for C++.
1303 @item -funsigned-char
1304 @opindex funsigned-char
1305 Let the type @code{char} be unsigned, like @code{unsigned char}.
1307 Each kind of machine has a default for what @code{char} should
1308 be. It is either like @code{unsigned char} by default or like
1309 @code{signed char} by default.
1311 Ideally, a portable program should always use @code{signed char} or
1312 @code{unsigned char} when it depends on the signedness of an object.
1313 But many programs have been written to use plain @code{char} and
1314 expect it to be signed, or expect it to be unsigned, depending on the
1315 machines they were written for. This option, and its inverse, let you
1316 make such a program work with the opposite default.
1318 The type @code{char} is always a distinct type from each of
1319 @code{signed char} or @code{unsigned char}, even though its behavior
1320 is always just like one of those two.
1323 @opindex fsigned-char
1324 Let the type @code{char} be signed, like @code{signed char}.
1326 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1327 the negative form of @option{-funsigned-char}. Likewise, the option
1328 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1330 @item -fsigned-bitfields
1331 @itemx -funsigned-bitfields
1332 @itemx -fno-signed-bitfields
1333 @itemx -fno-unsigned-bitfields
1334 @opindex fsigned-bitfields
1335 @opindex funsigned-bitfields
1336 @opindex fno-signed-bitfields
1337 @opindex fno-unsigned-bitfields
1338 These options control whether a bit-field is signed or unsigned, when the
1339 declaration does not use either @code{signed} or @code{unsigned}. By
1340 default, such a bit-field is signed, because this is consistent: the
1341 basic integer types such as @code{int} are signed types.
1344 @node C++ Dialect Options
1345 @section Options Controlling C++ Dialect
1347 @cindex compiler options, C++
1348 @cindex C++ options, command line
1349 @cindex options, C++
1350 This section describes the command-line options that are only meaningful
1351 for C++ programs; but you can also use most of the GNU compiler options
1352 regardless of what language your program is in. For example, you
1353 might compile a file @code{firstClass.C} like this:
1356 g++ -g -frepo -O -c firstClass.C
1360 In this example, only @option{-frepo} is an option meant
1361 only for C++ programs; you can use the other options with any
1362 language supported by GCC@.
1364 Here is a list of options that are @emph{only} for compiling C++ programs:
1368 @item -fabi-version=@var{n}
1369 @opindex fabi-version
1370 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1371 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1372 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1373 the version that conforms most closely to the C++ ABI specification.
1374 Therefore, the ABI obtained using version 0 will change as ABI bugs
1377 The default is version 2.
1379 @item -fno-access-control
1380 @opindex fno-access-control
1381 Turn off all access checking. This switch is mainly useful for working
1382 around bugs in the access control code.
1386 Check that the pointer returned by @code{operator new} is non-null
1387 before attempting to modify the storage allocated. This check is
1388 normally unnecessary because the C++ standard specifies that
1389 @code{operator new} will only return @code{0} if it is declared
1390 @samp{throw()}, in which case the compiler will always check the
1391 return value even without this option. In all other cases, when
1392 @code{operator new} has a non-empty exception specification, memory
1393 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1394 @samp{new (nothrow)}.
1396 @item -fconserve-space
1397 @opindex fconserve-space
1398 Put uninitialized or runtime-initialized global variables into the
1399 common segment, as C does. This saves space in the executable at the
1400 cost of not diagnosing duplicate definitions. If you compile with this
1401 flag and your program mysteriously crashes after @code{main()} has
1402 completed, you may have an object that is being destroyed twice because
1403 two definitions were merged.
1405 This option is no longer useful on most targets, now that support has
1406 been added for putting variables into BSS without making them common.
1408 @item -fno-const-strings
1409 @opindex fno-const-strings
1410 Give string constants type @code{char *} instead of type @code{const
1411 char *}. By default, G++ uses type @code{const char *} as required by
1412 the standard. Even if you use @option{-fno-const-strings}, you cannot
1413 actually modify the value of a string constant.
1415 This option might be removed in a future release of G++. For maximum
1416 portability, you should structure your code so that it works with
1417 string constants that have type @code{const char *}.
1419 @item -fno-elide-constructors
1420 @opindex fno-elide-constructors
1421 The C++ standard allows an implementation to omit creating a temporary
1422 which is only used to initialize another object of the same type.
1423 Specifying this option disables that optimization, and forces G++ to
1424 call the copy constructor in all cases.
1426 @item -fno-enforce-eh-specs
1427 @opindex fno-enforce-eh-specs
1428 Don't check for violation of exception specifications at runtime. This
1429 option violates the C++ standard, but may be useful for reducing code
1430 size in production builds, much like defining @samp{NDEBUG}. The compiler
1431 will still optimize based on the exception specifications.
1434 @itemx -fno-for-scope
1436 @opindex fno-for-scope
1437 If @option{-ffor-scope} is specified, the scope of variables declared in
1438 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1439 as specified by the C++ standard.
1440 If @option{-fno-for-scope} is specified, the scope of variables declared in
1441 a @i{for-init-statement} extends to the end of the enclosing scope,
1442 as was the case in old versions of G++, and other (traditional)
1443 implementations of C++.
1445 The default if neither flag is given to follow the standard,
1446 but to allow and give a warning for old-style code that would
1447 otherwise be invalid, or have different behavior.
1449 @item -fno-gnu-keywords
1450 @opindex fno-gnu-keywords
1451 Do not recognize @code{typeof} as a keyword, so that code can use this
1452 word as an identifier. You can use the keyword @code{__typeof__} instead.
1453 @option{-ansi} implies @option{-fno-gnu-keywords}.
1455 @item -fno-implicit-templates
1456 @opindex fno-implicit-templates
1457 Never emit code for non-inline templates which are instantiated
1458 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1459 @xref{Template Instantiation}, for more information.
1461 @item -fno-implicit-inline-templates
1462 @opindex fno-implicit-inline-templates
1463 Don't emit code for implicit instantiations of inline templates, either.
1464 The default is to handle inlines differently so that compiles with and
1465 without optimization will need the same set of explicit instantiations.
1467 @item -fno-implement-inlines
1468 @opindex fno-implement-inlines
1469 To save space, do not emit out-of-line copies of inline functions
1470 controlled by @samp{#pragma implementation}. This will cause linker
1471 errors if these functions are not inlined everywhere they are called.
1473 @item -fms-extensions
1474 @opindex fms-extensions
1475 Disable pedantic warnings about constructs used in MFC, such as implicit
1476 int and getting a pointer to member function via non-standard syntax.
1478 @item -fno-nonansi-builtins
1479 @opindex fno-nonansi-builtins
1480 Disable built-in declarations of functions that are not mandated by
1481 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1482 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1484 @item -fno-operator-names
1485 @opindex fno-operator-names
1486 Do not treat the operator name keywords @code{and}, @code{bitand},
1487 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1488 synonyms as keywords.
1490 @item -fno-optional-diags
1491 @opindex fno-optional-diags
1492 Disable diagnostics that the standard says a compiler does not need to
1493 issue. Currently, the only such diagnostic issued by G++ is the one for
1494 a name having multiple meanings within a class.
1497 @opindex fpermissive
1498 Downgrade some diagnostics about nonconformant code from errors to
1499 warnings. Thus, using @option{-fpermissive} will allow some
1500 nonconforming code to compile.
1504 Enable automatic template instantiation at link time. This option also
1505 implies @option{-fno-implicit-templates}. @xref{Template
1506 Instantiation}, for more information.
1510 Disable generation of information about every class with virtual
1511 functions for use by the C++ runtime type identification features
1512 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1513 of the language, you can save some space by using this flag. Note that
1514 exception handling uses the same information, but it will generate it as
1519 Emit statistics about front-end processing at the end of the compilation.
1520 This information is generally only useful to the G++ development team.
1522 @item -ftemplate-depth-@var{n}
1523 @opindex ftemplate-depth
1524 Set the maximum instantiation depth for template classes to @var{n}.
1525 A limit on the template instantiation depth is needed to detect
1526 endless recursions during template class instantiation. ANSI/ISO C++
1527 conforming programs must not rely on a maximum depth greater than 17.
1529 @item -fno-threadsafe-statics
1530 @opindex fno-threadsafe-statics
1531 Do not emit the extra code to use the routines specified in the C++
1532 ABI for thread-safe initialization of local statics. You can use this
1533 option to reduce code size slightly in code that doesn't need to be
1536 @item -fuse-cxa-atexit
1537 @opindex fuse-cxa-atexit
1538 Register destructors for objects with static storage duration with the
1539 @code{__cxa_atexit} function rather than the @code{atexit} function.
1540 This option is required for fully standards-compliant handling of static
1541 destructors, but will only work if your C library supports
1542 @code{__cxa_atexit}.
1544 @item -fvisibility-inlines-hidden
1545 @opindex fvisibility-inlines-hidden
1546 Causes all inlined methods to be marked with
1547 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1548 appear in the export table of a DSO and do not require a PLT indirection
1549 when used within the DSO@. Enabling this option can have a dramatic effect
1550 on load and link times of a DSO as it massively reduces the size of the
1551 dynamic export table when the library makes heavy use of templates. While
1552 it can cause bloating through duplication of code within each DSO where
1553 it is used, often the wastage is less than the considerable space occupied
1554 by a long symbol name in the export table which is typical when using
1555 templates and namespaces. For even more savings, combine with the
1556 @option{-fvisibility=hidden} switch.
1560 Do not use weak symbol support, even if it is provided by the linker.
1561 By default, G++ will use weak symbols if they are available. This
1562 option exists only for testing, and should not be used by end-users;
1563 it will result in inferior code and has no benefits. This option may
1564 be removed in a future release of G++.
1568 Do not search for header files in the standard directories specific to
1569 C++, but do still search the other standard directories. (This option
1570 is used when building the C++ library.)
1573 In addition, these optimization, warning, and code generation options
1574 have meanings only for C++ programs:
1577 @item -fno-default-inline
1578 @opindex fno-default-inline
1579 Do not assume @samp{inline} for functions defined inside a class scope.
1580 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1581 functions will have linkage like inline functions; they just won't be
1584 @item -Wabi @r{(C++ only)}
1586 Warn when G++ generates code that is probably not compatible with the
1587 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1588 all such cases, there are probably some cases that are not warned about,
1589 even though G++ is generating incompatible code. There may also be
1590 cases where warnings are emitted even though the code that is generated
1593 You should rewrite your code to avoid these warnings if you are
1594 concerned about the fact that code generated by G++ may not be binary
1595 compatible with code generated by other compilers.
1597 The known incompatibilities at this point include:
1602 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1603 pack data into the same byte as a base class. For example:
1606 struct A @{ virtual void f(); int f1 : 1; @};
1607 struct B : public A @{ int f2 : 1; @};
1611 In this case, G++ will place @code{B::f2} into the same byte
1612 as@code{A::f1}; other compilers will not. You can avoid this problem
1613 by explicitly padding @code{A} so that its size is a multiple of the
1614 byte size on your platform; that will cause G++ and other compilers to
1615 layout @code{B} identically.
1618 Incorrect handling of tail-padding for virtual bases. G++ does not use
1619 tail padding when laying out virtual bases. For example:
1622 struct A @{ virtual void f(); char c1; @};
1623 struct B @{ B(); char c2; @};
1624 struct C : public A, public virtual B @{@};
1628 In this case, G++ will not place @code{B} into the tail-padding for
1629 @code{A}; other compilers will. You can avoid this problem by
1630 explicitly padding @code{A} so that its size is a multiple of its
1631 alignment (ignoring virtual base classes); that will cause G++ and other
1632 compilers to layout @code{C} identically.
1635 Incorrect handling of bit-fields with declared widths greater than that
1636 of their underlying types, when the bit-fields appear in a union. For
1640 union U @{ int i : 4096; @};
1644 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1645 union too small by the number of bits in an @code{int}.
1648 Empty classes can be placed at incorrect offsets. For example:
1658 struct C : public B, public A @{@};
1662 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1663 it should be placed at offset zero. G++ mistakenly believes that the
1664 @code{A} data member of @code{B} is already at offset zero.
1667 Names of template functions whose types involve @code{typename} or
1668 template template parameters can be mangled incorrectly.
1671 template <typename Q>
1672 void f(typename Q::X) @{@}
1674 template <template <typename> class Q>
1675 void f(typename Q<int>::X) @{@}
1679 Instantiations of these templates may be mangled incorrectly.
1683 @item -Wctor-dtor-privacy @r{(C++ only)}
1684 @opindex Wctor-dtor-privacy
1685 Warn when a class seems unusable because all the constructors or
1686 destructors in that class are private, and it has neither friends nor
1687 public static member functions.
1689 @item -Wnon-virtual-dtor @r{(C++ only)}
1690 @opindex Wnon-virtual-dtor
1691 Warn when a class appears to be polymorphic, thereby requiring a virtual
1692 destructor, yet it declares a non-virtual one.
1693 This warning is enabled by @option{-Wall}.
1695 @item -Wreorder @r{(C++ only)}
1697 @cindex reordering, warning
1698 @cindex warning for reordering of member initializers
1699 Warn when the order of member initializers given in the code does not
1700 match the order in which they must be executed. For instance:
1706 A(): j (0), i (1) @{ @}
1710 The compiler will rearrange the member initializers for @samp{i}
1711 and @samp{j} to match the declaration order of the members, emitting
1712 a warning to that effect. This warning is enabled by @option{-Wall}.
1715 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1718 @item -Weffc++ @r{(C++ only)}
1720 Warn about violations of the following style guidelines from Scott Meyers'
1721 @cite{Effective C++} book:
1725 Item 11: Define a copy constructor and an assignment operator for classes
1726 with dynamically allocated memory.
1729 Item 12: Prefer initialization to assignment in constructors.
1732 Item 14: Make destructors virtual in base classes.
1735 Item 15: Have @code{operator=} return a reference to @code{*this}.
1738 Item 23: Don't try to return a reference when you must return an object.
1742 Also warn about violations of the following style guidelines from
1743 Scott Meyers' @cite{More Effective C++} book:
1747 Item 6: Distinguish between prefix and postfix forms of increment and
1748 decrement operators.
1751 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1755 When selecting this option, be aware that the standard library
1756 headers do not obey all of these guidelines; use @samp{grep -v}
1757 to filter out those warnings.
1759 @item -Wno-deprecated @r{(C++ only)}
1760 @opindex Wno-deprecated
1761 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1763 @item -Wstrict-null-sentinel @r{(C++ only)}
1764 @opindex Wstrict-null-sentinel
1765 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1766 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1767 to @code{__null}. Although it is a null pointer constant not a null pointer,
1768 it is guaranteed to of the same size as a pointer. But this use is
1769 not portable across different compilers.
1771 @item -Wno-non-template-friend @r{(C++ only)}
1772 @opindex Wno-non-template-friend
1773 Disable warnings when non-templatized friend functions are declared
1774 within a template. Since the advent of explicit template specification
1775 support in G++, if the name of the friend is an unqualified-id (i.e.,
1776 @samp{friend foo(int)}), the C++ language specification demands that the
1777 friend declare or define an ordinary, nontemplate function. (Section
1778 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1779 could be interpreted as a particular specialization of a templatized
1780 function. Because this non-conforming behavior is no longer the default
1781 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1782 check existing code for potential trouble spots and is on by default.
1783 This new compiler behavior can be turned off with
1784 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1785 but disables the helpful warning.
1787 @item -Wold-style-cast @r{(C++ only)}
1788 @opindex Wold-style-cast
1789 Warn if an old-style (C-style) cast to a non-void type is used within
1790 a C++ program. The new-style casts (@samp{static_cast},
1791 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1792 unintended effects and much easier to search for.
1794 @item -Woverloaded-virtual @r{(C++ only)}
1795 @opindex Woverloaded-virtual
1796 @cindex overloaded virtual fn, warning
1797 @cindex warning for overloaded virtual fn
1798 Warn when a function declaration hides virtual functions from a
1799 base class. For example, in:
1806 struct B: public A @{
1811 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1819 will fail to compile.
1821 @item -Wno-pmf-conversions @r{(C++ only)}
1822 @opindex Wno-pmf-conversions
1823 Disable the diagnostic for converting a bound pointer to member function
1826 @item -Wsign-promo @r{(C++ only)}
1827 @opindex Wsign-promo
1828 Warn when overload resolution chooses a promotion from unsigned or
1829 enumerated type to a signed type, over a conversion to an unsigned type of
1830 the same size. Previous versions of G++ would try to preserve
1831 unsignedness, but the standard mandates the current behavior.
1836 A& operator = (int);
1846 In this example, G++ will synthesize a default @samp{A& operator =
1847 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1850 @node Objective-C and Objective-C++ Dialect Options
1851 @section Options Controlling Objective-C and Objective-C++ Dialects
1853 @cindex compiler options, Objective-C and Objective-C++
1854 @cindex Objective-C and Objective-C++ options, command line
1855 @cindex options, Objective-C and Objective-C++
1856 (NOTE: This manual does not describe the Objective-C and Objective-C++
1857 languages themselves. See @xref{Standards,,Language Standards
1858 Supported by GCC}, for references.)
1860 This section describes the command-line options that are only meaningful
1861 for Objective-C and Objective-C++ programs, but you can also use most of
1862 the language-independent GNU compiler options.
1863 For example, you might compile a file @code{some_class.m} like this:
1866 gcc -g -fgnu-runtime -O -c some_class.m
1870 In this example, @option{-fgnu-runtime} is an option meant only for
1871 Objective-C and Objective-C++ programs; you can use the other options with
1872 any language supported by GCC@.
1874 Note that since Objective-C is an extension of the C language, Objective-C
1875 compilations may also use options specific to the C front-end (e.g.,
1876 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1877 C++-specific options (e.g., @option{-Wabi}).
1879 Here is a list of options that are @emph{only} for compiling Objective-C
1880 and Objective-C++ programs:
1883 @item -fconstant-string-class=@var{class-name}
1884 @opindex fconstant-string-class
1885 Use @var{class-name} as the name of the class to instantiate for each
1886 literal string specified with the syntax @code{@@"@dots{}"}. The default
1887 class name is @code{NXConstantString} if the GNU runtime is being used, and
1888 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1889 @option{-fconstant-cfstrings} option, if also present, will override the
1890 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1891 to be laid out as constant CoreFoundation strings.
1894 @opindex fgnu-runtime
1895 Generate object code compatible with the standard GNU Objective-C
1896 runtime. This is the default for most types of systems.
1898 @item -fnext-runtime
1899 @opindex fnext-runtime
1900 Generate output compatible with the NeXT runtime. This is the default
1901 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1902 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1905 @item -fno-nil-receivers
1906 @opindex fno-nil-receivers
1907 Assume that all Objective-C message dispatches (e.g.,
1908 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1909 is not @code{nil}. This allows for more efficient entry points in the runtime
1910 to be used. Currently, this option is only available in conjunction with
1911 the NeXT runtime on Mac OS X 10.3 and later.
1913 @item -fobjc-call-cxx-cdtors
1914 @opindex fobjc-call-cxx-cdtors
1915 For each Objective-C class, check if any of its instance variables is a
1916 C++ object with a non-trivial default constructor. If so, synthesize a
1917 special @code{- (id) .cxx_construct} instance method that will run
1918 non-trivial default constructors on any such instance variables, in order,
1919 and then return @code{self}. Similarly, check if any instance variable
1920 is a C++ object with a non-trivial destructor, and if so, synthesize a
1921 special @code{- (void) .cxx_destruct} method that will run
1922 all such default destructors, in reverse order.
1924 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
1925 thusly generated will only operate on instance variables declared in the
1926 current Objective-C class, and not those inherited from superclasses. It
1927 is the responsibility of the Objective-C runtime to invoke all such methods
1928 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
1929 will be invoked by the runtime immediately after a new object
1930 instance is allocated; the @code{- (void) .cxx_destruct} methods will
1931 be invoked immediately before the runtime deallocates an object instance.
1933 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
1934 support for invoking the @code{- (id) .cxx_construct} and
1935 @code{- (void) .cxx_destruct} methods.
1937 @item -fobjc-direct-dispatch
1938 @opindex fobjc-direct-dispatch
1939 Allow fast jumps to the message dispatcher. On Darwin this is
1940 accomplished via the comm page.
1942 @item -fobjc-exceptions
1943 @opindex fobjc-exceptions
1944 Enable syntactic support for structured exception handling in Objective-C,
1945 similar to what is offered by C++ and Java. Currently, this option is only
1946 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1954 @@catch (AnObjCClass *exc) @{
1961 @@catch (AnotherClass *exc) @{
1964 @@catch (id allOthers) @{
1974 The @code{@@throw} statement may appear anywhere in an Objective-C or
1975 Objective-C++ program; when used inside of a @code{@@catch} block, the
1976 @code{@@throw} may appear without an argument (as shown above), in which case
1977 the object caught by the @code{@@catch} will be rethrown.
1979 Note that only (pointers to) Objective-C objects may be thrown and
1980 caught using this scheme. When an object is thrown, it will be caught
1981 by the nearest @code{@@catch} clause capable of handling objects of that type,
1982 analogously to how @code{catch} blocks work in C++ and Java. A
1983 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1984 any and all Objective-C exceptions not caught by previous @code{@@catch}
1987 The @code{@@finally} clause, if present, will be executed upon exit from the
1988 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1989 regardless of whether any exceptions are thrown, caught or rethrown
1990 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1991 of the @code{finally} clause in Java.
1993 There are several caveats to using the new exception mechanism:
1997 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
1998 idioms provided by the @code{NSException} class, the new
1999 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2000 systems, due to additional functionality needed in the (NeXT) Objective-C
2004 As mentioned above, the new exceptions do not support handling
2005 types other than Objective-C objects. Furthermore, when used from
2006 Objective-C++, the Objective-C exception model does not interoperate with C++
2007 exceptions at this time. This means you cannot @code{@@throw} an exception
2008 from Objective-C and @code{catch} it in C++, or vice versa
2009 (i.e., @code{throw @dots{} @@catch}).
2012 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2013 blocks for thread-safe execution:
2016 @@synchronized (ObjCClass *guard) @{
2021 Upon entering the @code{@@synchronized} block, a thread of execution shall
2022 first check whether a lock has been placed on the corresponding @code{guard}
2023 object by another thread. If it has, the current thread shall wait until
2024 the other thread relinquishes its lock. Once @code{guard} becomes available,
2025 the current thread will place its own lock on it, execute the code contained in
2026 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2027 making @code{guard} available to other threads).
2029 Unlike Java, Objective-C does not allow for entire methods to be marked
2030 @code{@@synchronized}. Note that throwing exceptions out of
2031 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2032 to be unlocked properly.
2036 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2038 @item -freplace-objc-classes
2039 @opindex freplace-objc-classes
2040 Emit a special marker instructing @command{ld(1)} not to statically link in
2041 the resulting object file, and allow @command{dyld(1)} to load it in at
2042 run time instead. This is used in conjunction with the Fix-and-Continue
2043 debugging mode, where the object file in question may be recompiled and
2044 dynamically reloaded in the course of program execution, without the need
2045 to restart the program itself. Currently, Fix-and-Continue functionality
2046 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2051 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2052 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2053 compile time) with static class references that get initialized at load time,
2054 which improves run-time performance. Specifying the @option{-fzero-link} flag
2055 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2056 to be retained. This is useful in Zero-Link debugging mode, since it allows
2057 for individual class implementations to be modified during program execution.
2061 Dump interface declarations for all classes seen in the source file to a
2062 file named @file{@var{sourcename}.decl}.
2064 @item -Wassign-intercept
2065 @opindex Wassign-intercept
2066 Warn whenever an Objective-C assignment is being intercepted by the
2070 @opindex Wno-protocol
2071 If a class is declared to implement a protocol, a warning is issued for
2072 every method in the protocol that is not implemented by the class. The
2073 default behavior is to issue a warning for every method not explicitly
2074 implemented in the class, even if a method implementation is inherited
2075 from the superclass. If you use the @option{-Wno-protocol} option, then
2076 methods inherited from the superclass are considered to be implemented,
2077 and no warning is issued for them.
2081 Warn if multiple methods of different types for the same selector are
2082 found during compilation. The check is performed on the list of methods
2083 in the final stage of compilation. Additionally, a check is performed
2084 for each selector appearing in a @code{@@selector(@dots{})}
2085 expression, and a corresponding method for that selector has been found
2086 during compilation. Because these checks scan the method table only at
2087 the end of compilation, these warnings are not produced if the final
2088 stage of compilation is not reached, for example because an error is
2089 found during compilation, or because the @option{-fsyntax-only} option is
2092 @item -Wstrict-selector-match
2093 @opindex Wstrict-selector-match
2094 Warn if multiple methods with differing argument and/or return types are
2095 found for a given selector when attempting to send a message using this
2096 selector to a receiver of type @code{id} or @code{Class}. When this flag
2097 is off (which is the default behavior), the compiler will omit such warnings
2098 if any differences found are confined to types which share the same size
2101 @item -Wundeclared-selector
2102 @opindex Wundeclared-selector
2103 Warn if a @code{@@selector(@dots{})} expression referring to an
2104 undeclared selector is found. A selector is considered undeclared if no
2105 method with that name has been declared before the
2106 @code{@@selector(@dots{})} expression, either explicitly in an
2107 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2108 an @code{@@implementation} section. This option always performs its
2109 checks as soon as a @code{@@selector(@dots{})} expression is found,
2110 while @option{-Wselector} only performs its checks in the final stage of
2111 compilation. This also enforces the coding style convention
2112 that methods and selectors must be declared before being used.
2114 @item -print-objc-runtime-info
2115 @opindex print-objc-runtime-info
2116 Generate C header describing the largest structure that is passed by
2121 @node Language Independent Options
2122 @section Options to Control Diagnostic Messages Formatting
2123 @cindex options to control diagnostics formatting
2124 @cindex diagnostic messages
2125 @cindex message formatting
2127 Traditionally, diagnostic messages have been formatted irrespective of
2128 the output device's aspect (e.g.@: its width, @dots{}). The options described
2129 below can be used to control the diagnostic messages formatting
2130 algorithm, e.g.@: how many characters per line, how often source location
2131 information should be reported. Right now, only the C++ front end can
2132 honor these options. However it is expected, in the near future, that
2133 the remaining front ends would be able to digest them correctly.
2136 @item -fmessage-length=@var{n}
2137 @opindex fmessage-length
2138 Try to format error messages so that they fit on lines of about @var{n}
2139 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2140 the front ends supported by GCC@. If @var{n} is zero, then no
2141 line-wrapping will be done; each error message will appear on a single
2144 @opindex fdiagnostics-show-location
2145 @item -fdiagnostics-show-location=once
2146 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2147 reporter to emit @emph{once} source location information; that is, in
2148 case the message is too long to fit on a single physical line and has to
2149 be wrapped, the source location won't be emitted (as prefix) again,
2150 over and over, in subsequent continuation lines. This is the default
2153 @item -fdiagnostics-show-location=every-line
2154 Only meaningful in line-wrapping mode. Instructs the diagnostic
2155 messages reporter to emit the same source location information (as
2156 prefix) for physical lines that result from the process of breaking
2157 a message which is too long to fit on a single line.
2159 @item -fdiagnostics-show-options
2160 @opindex fdiagnostics-show-options
2161 This option instructs the diagnostic machinery to add text to each
2162 diagnostic emitted, which indicates which command line option directly
2163 controls that diagnostic, when such an option is known to the
2164 diagnostic machinery.
2168 @node Warning Options
2169 @section Options to Request or Suppress Warnings
2170 @cindex options to control warnings
2171 @cindex warning messages
2172 @cindex messages, warning
2173 @cindex suppressing warnings
2175 Warnings are diagnostic messages that report constructions which
2176 are not inherently erroneous but which are risky or suggest there
2177 may have been an error.
2179 You can request many specific warnings with options beginning @samp{-W},
2180 for example @option{-Wimplicit} to request warnings on implicit
2181 declarations. Each of these specific warning options also has a
2182 negative form beginning @samp{-Wno-} to turn off warnings;
2183 for example, @option{-Wno-implicit}. This manual lists only one of the
2184 two forms, whichever is not the default.
2186 The following options control the amount and kinds of warnings produced
2187 by GCC; for further, language-specific options also refer to
2188 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2192 @cindex syntax checking
2194 @opindex fsyntax-only
2195 Check the code for syntax errors, but don't do anything beyond that.
2199 Issue all the warnings demanded by strict ISO C and ISO C++;
2200 reject all programs that use forbidden extensions, and some other
2201 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2202 version of the ISO C standard specified by any @option{-std} option used.
2204 Valid ISO C and ISO C++ programs should compile properly with or without
2205 this option (though a rare few will require @option{-ansi} or a
2206 @option{-std} option specifying the required version of ISO C)@. However,
2207 without this option, certain GNU extensions and traditional C and C++
2208 features are supported as well. With this option, they are rejected.
2210 @option{-pedantic} does not cause warning messages for use of the
2211 alternate keywords whose names begin and end with @samp{__}. Pedantic
2212 warnings are also disabled in the expression that follows
2213 @code{__extension__}. However, only system header files should use
2214 these escape routes; application programs should avoid them.
2215 @xref{Alternate Keywords}.
2217 Some users try to use @option{-pedantic} to check programs for strict ISO
2218 C conformance. They soon find that it does not do quite what they want:
2219 it finds some non-ISO practices, but not all---only those for which
2220 ISO C @emph{requires} a diagnostic, and some others for which
2221 diagnostics have been added.
2223 A feature to report any failure to conform to ISO C might be useful in
2224 some instances, but would require considerable additional work and would
2225 be quite different from @option{-pedantic}. We don't have plans to
2226 support such a feature in the near future.
2228 Where the standard specified with @option{-std} represents a GNU
2229 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2230 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2231 extended dialect is based. Warnings from @option{-pedantic} are given
2232 where they are required by the base standard. (It would not make sense
2233 for such warnings to be given only for features not in the specified GNU
2234 C dialect, since by definition the GNU dialects of C include all
2235 features the compiler supports with the given option, and there would be
2236 nothing to warn about.)
2238 @item -pedantic-errors
2239 @opindex pedantic-errors
2240 Like @option{-pedantic}, except that errors are produced rather than
2245 Inhibit all warning messages.
2249 Inhibit warning messages about the use of @samp{#import}.
2251 @item -Wchar-subscripts
2252 @opindex Wchar-subscripts
2253 Warn if an array subscript has type @code{char}. This is a common cause
2254 of error, as programmers often forget that this type is signed on some
2256 This warning is enabled by @option{-Wall}.
2260 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2261 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2262 This warning is enabled by @option{-Wall}.
2264 @item -Wfatal-errors
2265 @opindex Wfatal-errors
2266 This option causes the compiler to abort compilation on the first error
2267 occurred rather than trying to keep going and printing further error
2272 @opindex ffreestanding
2273 @opindex fno-builtin
2274 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2275 the arguments supplied have types appropriate to the format string
2276 specified, and that the conversions specified in the format string make
2277 sense. This includes standard functions, and others specified by format
2278 attributes (@pxref{Function Attributes}), in the @code{printf},
2279 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2280 not in the C standard) families (or other target-specific families).
2281 Which functions are checked without format attributes having been
2282 specified depends on the standard version selected, and such checks of
2283 functions without the attribute specified are disabled by
2284 @option{-ffreestanding} or @option{-fno-builtin}.
2286 The formats are checked against the format features supported by GNU
2287 libc version 2.2. These include all ISO C90 and C99 features, as well
2288 as features from the Single Unix Specification and some BSD and GNU
2289 extensions. Other library implementations may not support all these
2290 features; GCC does not support warning about features that go beyond a
2291 particular library's limitations. However, if @option{-pedantic} is used
2292 with @option{-Wformat}, warnings will be given about format features not
2293 in the selected standard version (but not for @code{strfmon} formats,
2294 since those are not in any version of the C standard). @xref{C Dialect
2295 Options,,Options Controlling C Dialect}.
2297 Since @option{-Wformat} also checks for null format arguments for
2298 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2300 @option{-Wformat} is included in @option{-Wall}. For more control over some
2301 aspects of format checking, the options @option{-Wformat-y2k},
2302 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2303 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2304 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2307 @opindex Wformat-y2k
2308 If @option{-Wformat} is specified, also warn about @code{strftime}
2309 formats which may yield only a two-digit year.
2311 @item -Wno-format-extra-args
2312 @opindex Wno-format-extra-args
2313 If @option{-Wformat} is specified, do not warn about excess arguments to a
2314 @code{printf} or @code{scanf} format function. The C standard specifies
2315 that such arguments are ignored.
2317 Where the unused arguments lie between used arguments that are
2318 specified with @samp{$} operand number specifications, normally
2319 warnings are still given, since the implementation could not know what
2320 type to pass to @code{va_arg} to skip the unused arguments. However,
2321 in the case of @code{scanf} formats, this option will suppress the
2322 warning if the unused arguments are all pointers, since the Single
2323 Unix Specification says that such unused arguments are allowed.
2325 @item -Wno-format-zero-length
2326 @opindex Wno-format-zero-length
2327 If @option{-Wformat} is specified, do not warn about zero-length formats.
2328 The C standard specifies that zero-length formats are allowed.
2330 @item -Wformat-nonliteral
2331 @opindex Wformat-nonliteral
2332 If @option{-Wformat} is specified, also warn if the format string is not a
2333 string literal and so cannot be checked, unless the format function
2334 takes its format arguments as a @code{va_list}.
2336 @item -Wformat-security
2337 @opindex Wformat-security
2338 If @option{-Wformat} is specified, also warn about uses of format
2339 functions that represent possible security problems. At present, this
2340 warns about calls to @code{printf} and @code{scanf} functions where the
2341 format string is not a string literal and there are no format arguments,
2342 as in @code{printf (foo);}. This may be a security hole if the format
2343 string came from untrusted input and contains @samp{%n}. (This is
2344 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2345 in future warnings may be added to @option{-Wformat-security} that are not
2346 included in @option{-Wformat-nonliteral}.)
2350 Enable @option{-Wformat} plus format checks not included in
2351 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2352 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2356 Warn about passing a null pointer for arguments marked as
2357 requiring a non-null value by the @code{nonnull} function attribute.
2359 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2360 can be disabled with the @option{-Wno-nonnull} option.
2362 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2364 Warn about uninitialized variables which are initialized with themselves.
2365 Note this option can only be used with the @option{-Wuninitialized} option,
2366 which in turn only works with @option{-O1} and above.
2368 For example, GCC will warn about @code{i} being uninitialized in the
2369 following snippet only when @option{-Winit-self} has been specified:
2380 @item -Wimplicit-int
2381 @opindex Wimplicit-int
2382 Warn when a declaration does not specify a type.
2383 This warning is enabled by @option{-Wall}.
2385 @item -Wimplicit-function-declaration
2386 @itemx -Werror-implicit-function-declaration
2387 @opindex Wimplicit-function-declaration
2388 @opindex Werror-implicit-function-declaration
2389 Give a warning (or error) whenever a function is used before being
2390 declared. The form @option{-Wno-error-implicit-function-declaration}
2392 This warning is enabled by @option{-Wall} (as a warning, not an error).
2396 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2397 This warning is enabled by @option{-Wall}.
2401 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2402 function with external linkage, returning int, taking either zero
2403 arguments, two, or three arguments of appropriate types.
2404 This warning is enabled by @option{-Wall}.
2406 @item -Wmissing-braces
2407 @opindex Wmissing-braces
2408 Warn if an aggregate or union initializer is not fully bracketed. In
2409 the following example, the initializer for @samp{a} is not fully
2410 bracketed, but that for @samp{b} is fully bracketed.
2413 int a[2][2] = @{ 0, 1, 2, 3 @};
2414 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2417 This warning is enabled by @option{-Wall}.
2419 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2420 @opindex Wmissing-include-dirs
2421 Warn if a user-supplied include directory does not exist.
2424 @opindex Wparentheses
2425 Warn if parentheses are omitted in certain contexts, such
2426 as when there is an assignment in a context where a truth value
2427 is expected, or when operators are nested whose precedence people
2428 often get confused about. Only the warning for an assignment used as
2429 a truth value is supported when compiling C++; the other warnings are
2430 only supported when compiling C@.
2432 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2433 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2434 interpretation from that of ordinary mathematical notation.
2436 Also warn about constructions where there may be confusion to which
2437 @code{if} statement an @code{else} branch belongs. Here is an example of
2452 In C, every @code{else} branch belongs to the innermost possible @code{if}
2453 statement, which in this example is @code{if (b)}. This is often not
2454 what the programmer expected, as illustrated in the above example by
2455 indentation the programmer chose. When there is the potential for this
2456 confusion, GCC will issue a warning when this flag is specified.
2457 To eliminate the warning, add explicit braces around the innermost
2458 @code{if} statement so there is no way the @code{else} could belong to
2459 the enclosing @code{if}. The resulting code would look like this:
2475 This warning is enabled by @option{-Wall}.
2477 @item -Wsequence-point
2478 @opindex Wsequence-point
2479 Warn about code that may have undefined semantics because of violations
2480 of sequence point rules in the C standard.
2482 The C standard defines the order in which expressions in a C program are
2483 evaluated in terms of @dfn{sequence points}, which represent a partial
2484 ordering between the execution of parts of the program: those executed
2485 before the sequence point, and those executed after it. These occur
2486 after the evaluation of a full expression (one which is not part of a
2487 larger expression), after the evaluation of the first operand of a
2488 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2489 function is called (but after the evaluation of its arguments and the
2490 expression denoting the called function), and in certain other places.
2491 Other than as expressed by the sequence point rules, the order of
2492 evaluation of subexpressions of an expression is not specified. All
2493 these rules describe only a partial order rather than a total order,
2494 since, for example, if two functions are called within one expression
2495 with no sequence point between them, the order in which the functions
2496 are called is not specified. However, the standards committee have
2497 ruled that function calls do not overlap.
2499 It is not specified when between sequence points modifications to the
2500 values of objects take effect. Programs whose behavior depends on this
2501 have undefined behavior; the C standard specifies that ``Between the
2502 previous and next sequence point an object shall have its stored value
2503 modified at most once by the evaluation of an expression. Furthermore,
2504 the prior value shall be read only to determine the value to be
2505 stored.''. If a program breaks these rules, the results on any
2506 particular implementation are entirely unpredictable.
2508 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2509 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2510 diagnosed by this option, and it may give an occasional false positive
2511 result, but in general it has been found fairly effective at detecting
2512 this sort of problem in programs.
2514 The present implementation of this option only works for C programs. A
2515 future implementation may also work for C++ programs.
2517 The C standard is worded confusingly, therefore there is some debate
2518 over the precise meaning of the sequence point rules in subtle cases.
2519 Links to discussions of the problem, including proposed formal
2520 definitions, may be found on the GCC readings page, at
2521 @w{@uref{http://gcc.gnu.org/readings.html}}.
2523 This warning is enabled by @option{-Wall}.
2526 @opindex Wreturn-type
2527 Warn whenever a function is defined with a return-type that defaults to
2528 @code{int}. Also warn about any @code{return} statement with no
2529 return-value in a function whose return-type is not @code{void}.
2531 For C, also warn if the return type of a function has a type qualifier
2532 such as @code{const}. Such a type qualifier has no effect, since the
2533 value returned by a function is not an lvalue. ISO C prohibits
2534 qualified @code{void} return types on function definitions, so such
2535 return types always receive a warning even without this option.
2537 For C++, a function without return type always produces a diagnostic
2538 message, even when @option{-Wno-return-type} is specified. The only
2539 exceptions are @samp{main} and functions defined in system headers.
2541 This warning is enabled by @option{-Wall}.
2545 Warn whenever a @code{switch} statement has an index of enumerated type
2546 and lacks a @code{case} for one or more of the named codes of that
2547 enumeration. (The presence of a @code{default} label prevents this
2548 warning.) @code{case} labels outside the enumeration range also
2549 provoke warnings when this option is used.
2550 This warning is enabled by @option{-Wall}.
2552 @item -Wswitch-default
2553 @opindex Wswitch-switch
2554 Warn whenever a @code{switch} statement does not have a @code{default}
2558 @opindex Wswitch-enum
2559 Warn whenever a @code{switch} statement has an index of enumerated type
2560 and lacks a @code{case} for one or more of the named codes of that
2561 enumeration. @code{case} labels outside the enumeration range also
2562 provoke warnings when this option is used.
2566 Warn if any trigraphs are encountered that might change the meaning of
2567 the program (trigraphs within comments are not warned about).
2568 This warning is enabled by @option{-Wall}.
2570 @item -Wunused-function
2571 @opindex Wunused-function
2572 Warn whenever a static function is declared but not defined or a
2573 non\-inline static function is unused.
2574 This warning is enabled by @option{-Wall}.
2576 @item -Wunused-label
2577 @opindex Wunused-label
2578 Warn whenever a label is declared but not used.
2579 This warning is enabled by @option{-Wall}.
2581 To suppress this warning use the @samp{unused} attribute
2582 (@pxref{Variable Attributes}).
2584 @item -Wunused-parameter
2585 @opindex Wunused-parameter
2586 Warn whenever a function parameter is unused aside from its declaration.
2588 To suppress this warning use the @samp{unused} attribute
2589 (@pxref{Variable Attributes}).
2591 @item -Wunused-variable
2592 @opindex Wunused-variable
2593 Warn whenever a local variable or non-constant static variable is unused
2594 aside from its declaration
2595 This warning is enabled by @option{-Wall}.
2597 To suppress this warning use the @samp{unused} attribute
2598 (@pxref{Variable Attributes}).
2600 @item -Wunused-value
2601 @opindex Wunused-value
2602 Warn whenever a statement computes a result that is explicitly not used.
2603 This warning is enabled by @option{-Wall}.
2605 To suppress this warning cast the expression to @samp{void}.
2609 All the above @option{-Wunused} options combined.
2611 In order to get a warning about an unused function parameter, you must
2612 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2613 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2615 @item -Wuninitialized
2616 @opindex Wuninitialized
2617 Warn if an automatic variable is used without first being initialized or
2618 if a variable may be clobbered by a @code{setjmp} call.
2620 These warnings are possible only in optimizing compilation,
2621 because they require data flow information that is computed only
2622 when optimizing. If you don't specify @option{-O}, you simply won't
2625 If you want to warn about code which uses the uninitialized value of the
2626 variable in its own initializer, use the @option{-Winit-self} option.
2628 These warnings occur for individual uninitialized or clobbered
2629 elements of structure, union or array variables as well as for
2630 variables which are uninitialized or clobbered as a whole. They do
2631 not occur for variables or elements declared @code{volatile}. Because
2632 these warnings depend on optimization, the exact variables or elements
2633 for which there are warnings will depend on the precise optimization
2634 options and version of GCC used.
2636 Note that there may be no warning about a variable that is used only
2637 to compute a value that itself is never used, because such
2638 computations may be deleted by data flow analysis before the warnings
2641 These warnings are made optional because GCC is not smart
2642 enough to see all the reasons why the code might be correct
2643 despite appearing to have an error. Here is one example of how
2664 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2665 always initialized, but GCC doesn't know this. Here is
2666 another common case:
2671 if (change_y) save_y = y, y = new_y;
2673 if (change_y) y = save_y;
2678 This has no bug because @code{save_y} is used only if it is set.
2680 @cindex @code{longjmp} warnings
2681 This option also warns when a non-volatile automatic variable might be
2682 changed by a call to @code{longjmp}. These warnings as well are possible
2683 only in optimizing compilation.
2685 The compiler sees only the calls to @code{setjmp}. It cannot know
2686 where @code{longjmp} will be called; in fact, a signal handler could
2687 call it at any point in the code. As a result, you may get a warning
2688 even when there is in fact no problem because @code{longjmp} cannot
2689 in fact be called at the place which would cause a problem.
2691 Some spurious warnings can be avoided if you declare all the functions
2692 you use that never return as @code{noreturn}. @xref{Function
2695 This warning is enabled by @option{-Wall}.
2697 @item -Wunknown-pragmas
2698 @opindex Wunknown-pragmas
2699 @cindex warning for unknown pragmas
2700 @cindex unknown pragmas, warning
2701 @cindex pragmas, warning of unknown
2702 Warn when a #pragma directive is encountered which is not understood by
2703 GCC@. If this command line option is used, warnings will even be issued
2704 for unknown pragmas in system header files. This is not the case if
2705 the warnings were only enabled by the @option{-Wall} command line option.
2707 @item -Wstrict-aliasing
2708 @opindex Wstrict-aliasing
2709 This option is only active when @option{-fstrict-aliasing} is active.
2710 It warns about code which might break the strict aliasing rules that the
2711 compiler is using for optimization. The warning does not catch all
2712 cases, but does attempt to catch the more common pitfalls. It is
2713 included in @option{-Wall}.
2715 @item -Wstrict-aliasing=2
2716 @opindex Wstrict-aliasing=2
2717 This option is only active when @option{-fstrict-aliasing} is active.
2718 It warns about all code which might break the strict aliasing rules that the
2719 compiler is using for optimization. This warning catches all cases, but
2720 it will also give a warning for some ambiguous cases that are safe.
2724 All of the above @samp{-W} options combined. This enables all the
2725 warnings about constructions that some users consider questionable, and
2726 that are easy to avoid (or modify to prevent the warning), even in
2727 conjunction with macros. This also enables some language-specific
2728 warnings described in @ref{C++ Dialect Options} and
2729 @ref{Objective-C and Objective-C++ Dialect Options}.
2732 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2733 Some of them warn about constructions that users generally do not
2734 consider questionable, but which occasionally you might wish to check
2735 for; others warn about constructions that are necessary or hard to avoid
2736 in some cases, and there is no simple way to modify the code to suppress
2743 (This option used to be called @option{-W}. The older name is still
2744 supported, but the newer name is more descriptive.) Print extra warning
2745 messages for these events:
2749 A function can return either with or without a value. (Falling
2750 off the end of the function body is considered returning without
2751 a value.) For example, this function would evoke such a
2765 An expression-statement or the left-hand side of a comma expression
2766 contains no side effects.
2767 To suppress the warning, cast the unused expression to void.
2768 For example, an expression such as @samp{x[i,j]} will cause a warning,
2769 but @samp{x[(void)i,j]} will not.
2772 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2775 Storage-class specifiers like @code{static} are not the first things in
2776 a declaration. According to the C Standard, this usage is obsolescent.
2779 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2783 A comparison between signed and unsigned values could produce an
2784 incorrect result when the signed value is converted to unsigned.
2785 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2788 An aggregate has an initializer which does not initialize all members.
2789 This warning can be independently controlled by
2790 @option{-Wmissing-field-initializers}.
2793 A function parameter is declared without a type specifier in K&R-style
2801 An empty body occurs in an @samp{if} or @samp{else} statement.
2804 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2805 @samp{>}, or @samp{>=}.
2808 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2811 Any of several floating-point events that often indicate errors, such as
2812 overflow, underflow, loss of precision, etc.
2814 @item @r{(C++ only)}
2815 An enumerator and a non-enumerator both appear in a conditional expression.
2817 @item @r{(C++ only)}
2818 A non-static reference or non-static @samp{const} member appears in a
2819 class without constructors.
2821 @item @r{(C++ only)}
2822 Ambiguous virtual bases.
2824 @item @r{(C++ only)}
2825 Subscripting an array which has been declared @samp{register}.
2827 @item @r{(C++ only)}
2828 Taking the address of a variable which has been declared @samp{register}.
2830 @item @r{(C++ only)}
2831 A base class is not initialized in a derived class' copy constructor.
2834 @item -Wno-div-by-zero
2835 @opindex Wno-div-by-zero
2836 @opindex Wdiv-by-zero
2837 Do not warn about compile-time integer division by zero. Floating point
2838 division by zero is not warned about, as it can be a legitimate way of
2839 obtaining infinities and NaNs.
2841 @item -Wsystem-headers
2842 @opindex Wsystem-headers
2843 @cindex warnings from system headers
2844 @cindex system headers, warnings from
2845 Print warning messages for constructs found in system header files.
2846 Warnings from system headers are normally suppressed, on the assumption
2847 that they usually do not indicate real problems and would only make the
2848 compiler output harder to read. Using this command line option tells
2849 GCC to emit warnings from system headers as if they occurred in user
2850 code. However, note that using @option{-Wall} in conjunction with this
2851 option will @emph{not} warn about unknown pragmas in system
2852 headers---for that, @option{-Wunknown-pragmas} must also be used.
2855 @opindex Wfloat-equal
2856 Warn if floating point values are used in equality comparisons.
2858 The idea behind this is that sometimes it is convenient (for the
2859 programmer) to consider floating-point values as approximations to
2860 infinitely precise real numbers. If you are doing this, then you need
2861 to compute (by analyzing the code, or in some other way) the maximum or
2862 likely maximum error that the computation introduces, and allow for it
2863 when performing comparisons (and when producing output, but that's a
2864 different problem). In particular, instead of testing for equality, you
2865 would check to see whether the two values have ranges that overlap; and
2866 this is done with the relational operators, so equality comparisons are
2869 @item -Wtraditional @r{(C only)}
2870 @opindex Wtraditional
2871 Warn about certain constructs that behave differently in traditional and
2872 ISO C@. Also warn about ISO C constructs that have no traditional C
2873 equivalent, and/or problematic constructs which should be avoided.
2877 Macro parameters that appear within string literals in the macro body.
2878 In traditional C macro replacement takes place within string literals,
2879 but does not in ISO C@.
2882 In traditional C, some preprocessor directives did not exist.
2883 Traditional preprocessors would only consider a line to be a directive
2884 if the @samp{#} appeared in column 1 on the line. Therefore
2885 @option{-Wtraditional} warns about directives that traditional C
2886 understands but would ignore because the @samp{#} does not appear as the
2887 first character on the line. It also suggests you hide directives like
2888 @samp{#pragma} not understood by traditional C by indenting them. Some
2889 traditional implementations would not recognize @samp{#elif}, so it
2890 suggests avoiding it altogether.
2893 A function-like macro that appears without arguments.
2896 The unary plus operator.
2899 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2900 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2901 constants.) Note, these suffixes appear in macros defined in the system
2902 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2903 Use of these macros in user code might normally lead to spurious
2904 warnings, however GCC's integrated preprocessor has enough context to
2905 avoid warning in these cases.
2908 A function declared external in one block and then used after the end of
2912 A @code{switch} statement has an operand of type @code{long}.
2915 A non-@code{static} function declaration follows a @code{static} one.
2916 This construct is not accepted by some traditional C compilers.
2919 The ISO type of an integer constant has a different width or
2920 signedness from its traditional type. This warning is only issued if
2921 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2922 typically represent bit patterns, are not warned about.
2925 Usage of ISO string concatenation is detected.
2928 Initialization of automatic aggregates.
2931 Identifier conflicts with labels. Traditional C lacks a separate
2932 namespace for labels.
2935 Initialization of unions. If the initializer is zero, the warning is
2936 omitted. This is done under the assumption that the zero initializer in
2937 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2938 initializer warnings and relies on default initialization to zero in the
2942 Conversions by prototypes between fixed/floating point values and vice
2943 versa. The absence of these prototypes when compiling with traditional
2944 C would cause serious problems. This is a subset of the possible
2945 conversion warnings, for the full set use @option{-Wconversion}.
2948 Use of ISO C style function definitions. This warning intentionally is
2949 @emph{not} issued for prototype declarations or variadic functions
2950 because these ISO C features will appear in your code when using
2951 libiberty's traditional C compatibility macros, @code{PARAMS} and
2952 @code{VPARAMS}. This warning is also bypassed for nested functions
2953 because that feature is already a GCC extension and thus not relevant to
2954 traditional C compatibility.
2957 @item -Wdeclaration-after-statement @r{(C only)}
2958 @opindex Wdeclaration-after-statement
2959 Warn when a declaration is found after a statement in a block. This
2960 construct, known from C++, was introduced with ISO C99 and is by default
2961 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2962 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2966 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2968 @item -Wno-endif-labels
2969 @opindex Wno-endif-labels
2970 @opindex Wendif-labels
2971 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2975 Warn whenever a local variable shadows another local variable, parameter or
2976 global variable or whenever a built-in function is shadowed.
2978 @item -Wlarger-than-@var{len}
2979 @opindex Wlarger-than
2980 Warn whenever an object of larger than @var{len} bytes is defined.
2982 @item -Wpointer-arith
2983 @opindex Wpointer-arith
2984 Warn about anything that depends on the ``size of'' a function type or
2985 of @code{void}. GNU C assigns these types a size of 1, for
2986 convenience in calculations with @code{void *} pointers and pointers
2989 @item -Wbad-function-cast @r{(C only)}
2990 @opindex Wbad-function-cast
2991 Warn whenever a function call is cast to a non-matching type.
2992 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2996 Warn whenever a pointer is cast so as to remove a type qualifier from
2997 the target type. For example, warn if a @code{const char *} is cast
2998 to an ordinary @code{char *}.
3001 @opindex Wcast-align
3002 Warn whenever a pointer is cast such that the required alignment of the
3003 target is increased. For example, warn if a @code{char *} is cast to
3004 an @code{int *} on machines where integers can only be accessed at
3005 two- or four-byte boundaries.
3007 @item -Wwrite-strings
3008 @opindex Wwrite-strings
3009 When compiling C, give string constants the type @code{const
3010 char[@var{length}]} so that
3011 copying the address of one into a non-@code{const} @code{char *}
3012 pointer will get a warning; when compiling C++, warn about the
3013 deprecated conversion from string constants to @code{char *}.
3014 These warnings will help you find at
3015 compile time code that can try to write into a string constant, but
3016 only if you have been very careful about using @code{const} in
3017 declarations and prototypes. Otherwise, it will just be a nuisance;
3018 this is why we did not make @option{-Wall} request these warnings.
3021 @opindex Wconversion
3022 Warn if a prototype causes a type conversion that is different from what
3023 would happen to the same argument in the absence of a prototype. This
3024 includes conversions of fixed point to floating and vice versa, and
3025 conversions changing the width or signedness of a fixed point argument
3026 except when the same as the default promotion.
3028 Also, warn if a negative integer constant expression is implicitly
3029 converted to an unsigned type. For example, warn about the assignment
3030 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
3031 casts like @code{(unsigned) -1}.
3033 @item -Wsign-compare
3034 @opindex Wsign-compare
3035 @cindex warning for comparison of signed and unsigned values
3036 @cindex comparison of signed and unsigned values, warning
3037 @cindex signed and unsigned values, comparison warning
3038 Warn when a comparison between signed and unsigned values could produce
3039 an incorrect result when the signed value is converted to unsigned.
3040 This warning is also enabled by @option{-Wextra}; to get the other warnings
3041 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3043 @item -Waggregate-return
3044 @opindex Waggregate-return
3045 Warn if any functions that return structures or unions are defined or
3046 called. (In languages where you can return an array, this also elicits
3049 @item -Wstrict-prototypes @r{(C only)}
3050 @opindex Wstrict-prototypes
3051 Warn if a function is declared or defined without specifying the
3052 argument types. (An old-style function definition is permitted without
3053 a warning if preceded by a declaration which specifies the argument
3056 @item -Wold-style-definition @r{(C only)}
3057 @opindex Wold-style-definition
3058 Warn if an old-style function definition is used. A warning is given
3059 even if there is a previous prototype.
3061 @item -Wmissing-prototypes @r{(C only)}
3062 @opindex Wmissing-prototypes
3063 Warn if a global function is defined without a previous prototype
3064 declaration. This warning is issued even if the definition itself
3065 provides a prototype. The aim is to detect global functions that fail
3066 to be declared in header files.
3068 @item -Wmissing-declarations @r{(C only)}
3069 @opindex Wmissing-declarations
3070 Warn if a global function is defined without a previous declaration.
3071 Do so even if the definition itself provides a prototype.
3072 Use this option to detect global functions that are not declared in
3075 @item -Wmissing-field-initializers
3076 @opindex Wmissing-field-initializers
3079 Warn if a structure's initializer has some fields missing. For
3080 example, the following code would cause such a warning, because
3081 @code{x.h} is implicitly zero:
3084 struct s @{ int f, g, h; @};
3085 struct s x = @{ 3, 4 @};
3088 This option does not warn about designated initializers, so the following
3089 modification would not trigger a warning:
3092 struct s @{ int f, g, h; @};
3093 struct s x = @{ .f = 3, .g = 4 @};
3096 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3097 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3099 @item -Wmissing-noreturn
3100 @opindex Wmissing-noreturn
3101 Warn about functions which might be candidates for attribute @code{noreturn}.
3102 Note these are only possible candidates, not absolute ones. Care should
3103 be taken to manually verify functions actually do not ever return before
3104 adding the @code{noreturn} attribute, otherwise subtle code generation
3105 bugs could be introduced. You will not get a warning for @code{main} in
3106 hosted C environments.
3108 @item -Wmissing-format-attribute
3109 @opindex Wmissing-format-attribute
3111 If @option{-Wformat} is enabled, also warn about functions which might be
3112 candidates for @code{format} attributes. Note these are only possible
3113 candidates, not absolute ones. GCC will guess that @code{format}
3114 attributes might be appropriate for any function that calls a function
3115 like @code{vprintf} or @code{vscanf}, but this might not always be the
3116 case, and some functions for which @code{format} attributes are
3117 appropriate may not be detected. This option has no effect unless
3118 @option{-Wformat} is enabled (possibly by @option{-Wall}).
3120 @item -Wno-multichar
3121 @opindex Wno-multichar
3123 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3124 Usually they indicate a typo in the user's code, as they have
3125 implementation-defined values, and should not be used in portable code.
3127 @item -Wnormalized=<none|id|nfc|nfkc>
3128 @opindex Wnormalized
3131 @cindex character set, input normalization
3132 In ISO C and ISO C++, two identifiers are different if they are
3133 different sequences of characters. However, sometimes when characters
3134 outside the basic ASCII character set are used, you can have two
3135 different character sequences that look the same. To avoid confusion,
3136 the ISO 10646 standard sets out some @dfn{normalization rules} which
3137 when applied ensure that two sequences that look the same are turned into
3138 the same sequence. GCC can warn you if you are using identifiers which
3139 have not been normalized; this option controls that warning.
3141 There are four levels of warning that GCC supports. The default is
3142 @option{-Wnormalized=nfc}, which warns about any identifier which is
3143 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3144 recommended form for most uses.
3146 Unfortunately, there are some characters which ISO C and ISO C++ allow
3147 in identifiers that when turned into NFC aren't allowable as
3148 identifiers. That is, there's no way to use these symbols in portable
3149 ISO C or C++ and have all your identifiers in NFC.
3150 @option{-Wnormalized=id} suppresses the warning for these characters.
3151 It is hoped that future versions of the standards involved will correct
3152 this, which is why this option is not the default.
3154 You can switch the warning off for all characters by writing
3155 @option{-Wnormalized=none}. You would only want to do this if you
3156 were using some other normalization scheme (like ``D''), because
3157 otherwise you can easily create bugs that are literally impossible to see.
3159 Some characters in ISO 10646 have distinct meanings but look identical
3160 in some fonts or display methodologies, especially once formatting has
3161 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3162 LETTER N'', will display just like a regular @code{n} which has been
3163 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3164 normalisation scheme to convert all these into a standard form as
3165 well, and GCC will warn if your code is not in NFKC if you use
3166 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3167 about every identifier that contains the letter O because it might be
3168 confused with the digit 0, and so is not the default, but may be
3169 useful as a local coding convention if the programming environment is
3170 unable to be fixed to display these characters distinctly.
3172 @item -Wno-deprecated-declarations
3173 @opindex Wno-deprecated-declarations
3174 Do not warn about uses of functions, variables, and types marked as
3175 deprecated by using the @code{deprecated} attribute.
3176 (@pxref{Function Attributes}, @pxref{Variable Attributes},
3177 @pxref{Type Attributes}.)
3181 Warn if a structure is given the packed attribute, but the packed
3182 attribute has no effect on the layout or size of the structure.
3183 Such structures may be mis-aligned for little benefit. For
3184 instance, in this code, the variable @code{f.x} in @code{struct bar}
3185 will be misaligned even though @code{struct bar} does not itself
3186 have the packed attribute:
3193 @} __attribute__((packed));
3203 Warn if padding is included in a structure, either to align an element
3204 of the structure or to align the whole structure. Sometimes when this
3205 happens it is possible to rearrange the fields of the structure to
3206 reduce the padding and so make the structure smaller.
3208 @item -Wredundant-decls
3209 @opindex Wredundant-decls
3210 Warn if anything is declared more than once in the same scope, even in
3211 cases where multiple declaration is valid and changes nothing.
3213 @item -Wnested-externs @r{(C only)}
3214 @opindex Wnested-externs
3215 Warn if an @code{extern} declaration is encountered within a function.
3217 @item -Wunreachable-code
3218 @opindex Wunreachable-code
3219 Warn if the compiler detects that code will never be executed.
3221 This option is intended to warn when the compiler detects that at
3222 least a whole line of source code will never be executed, because
3223 some condition is never satisfied or because it is after a
3224 procedure that never returns.
3226 It is possible for this option to produce a warning even though there
3227 are circumstances under which part of the affected line can be executed,
3228 so care should be taken when removing apparently-unreachable code.
3230 For instance, when a function is inlined, a warning may mean that the
3231 line is unreachable in only one inlined copy of the function.
3233 This option is not made part of @option{-Wall} because in a debugging
3234 version of a program there is often substantial code which checks
3235 correct functioning of the program and is, hopefully, unreachable
3236 because the program does work. Another common use of unreachable
3237 code is to provide behavior which is selectable at compile-time.
3241 Warn if a function can not be inlined and it was declared as inline.
3242 Even with this option, the compiler will not warn about failures to
3243 inline functions declared in system headers.
3245 The compiler uses a variety of heuristics to determine whether or not
3246 to inline a function. For example, the compiler takes into account
3247 the size of the function being inlined and the amount of inlining
3248 that has already been done in the current function. Therefore,
3249 seemingly insignificant changes in the source program can cause the
3250 warnings produced by @option{-Winline} to appear or disappear.
3252 @item -Wno-invalid-offsetof @r{(C++ only)}
3253 @opindex Wno-invalid-offsetof
3254 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3255 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3256 to a non-POD type is undefined. In existing C++ implementations,
3257 however, @samp{offsetof} typically gives meaningful results even when
3258 applied to certain kinds of non-POD types. (Such as a simple
3259 @samp{struct} that fails to be a POD type only by virtue of having a
3260 constructor.) This flag is for users who are aware that they are
3261 writing nonportable code and who have deliberately chosen to ignore the
3264 The restrictions on @samp{offsetof} may be relaxed in a future version
3265 of the C++ standard.
3267 @item -Wno-int-to-pointer-cast @r{(C only)}
3268 @opindex Wno-int-to-pointer-cast
3269 Suppress warnings from casts to pointer type of an integer of a
3272 @item -Wno-pointer-to-int-cast @r{(C only)}
3273 @opindex Wno-pointer-to-int-cast
3274 Suppress warnings from casts from a pointer to an integer type of a
3278 @opindex Winvalid-pch
3279 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3280 the search path but can't be used.
3284 @opindex Wno-long-long
3285 Warn if @samp{long long} type is used. This is default. To inhibit
3286 the warning messages, use @option{-Wno-long-long}. Flags
3287 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3288 only when @option{-pedantic} flag is used.
3290 @item -Wvariadic-macros
3291 @opindex Wvariadic-macros
3292 @opindex Wno-variadic-macros
3293 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3294 alternate syntax when in pedantic ISO C99 mode. This is default.
3295 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3297 @item -Wdisabled-optimization
3298 @opindex Wdisabled-optimization
3299 Warn if a requested optimization pass is disabled. This warning does
3300 not generally indicate that there is anything wrong with your code; it
3301 merely indicates that GCC's optimizers were unable to handle the code
3302 effectively. Often, the problem is that your code is too big or too
3303 complex; GCC will refuse to optimize programs when the optimization
3304 itself is likely to take inordinate amounts of time.
3306 @item -Wno-pointer-sign
3307 @opindex Wno-pointer-sign
3308 Don't warn for pointer argument passing or assignment with different signedness.
3309 Only useful in the negative form since this warning is enabled by default.
3310 This option is only supported for C and Objective-C@.
3314 Make all warnings into errors.
3317 @node Debugging Options
3318 @section Options for Debugging Your Program or GCC
3319 @cindex options, debugging
3320 @cindex debugging information options
3322 GCC has various special options that are used for debugging
3323 either your program or GCC:
3328 Produce debugging information in the operating system's native format
3329 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3332 On most systems that use stabs format, @option{-g} enables use of extra
3333 debugging information that only GDB can use; this extra information
3334 makes debugging work better in GDB but will probably make other debuggers
3336 refuse to read the program. If you want to control for certain whether
3337 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3338 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3340 GCC allows you to use @option{-g} with
3341 @option{-O}. The shortcuts taken by optimized code may occasionally
3342 produce surprising results: some variables you declared may not exist
3343 at all; flow of control may briefly move where you did not expect it;
3344 some statements may not be executed because they compute constant
3345 results or their values were already at hand; some statements may
3346 execute in different places because they were moved out of loops.
3348 Nevertheless it proves possible to debug optimized output. This makes
3349 it reasonable to use the optimizer for programs that might have bugs.
3351 The following options are useful when GCC is generated with the
3352 capability for more than one debugging format.
3356 Produce debugging information for use by GDB@. This means to use the
3357 most expressive format available (DWARF 2, stabs, or the native format
3358 if neither of those are supported), including GDB extensions if at all
3363 Produce debugging information in stabs format (if that is supported),
3364 without GDB extensions. This is the format used by DBX on most BSD
3365 systems. On MIPS, Alpha and System V Release 4 systems this option
3366 produces stabs debugging output which is not understood by DBX or SDB@.
3367 On System V Release 4 systems this option requires the GNU assembler.
3369 @item -feliminate-unused-debug-symbols
3370 @opindex feliminate-unused-debug-symbols
3371 Produce debugging information in stabs format (if that is supported),
3372 for only symbols that are actually used.
3376 Produce debugging information in stabs format (if that is supported),
3377 using GNU extensions understood only by the GNU debugger (GDB)@. The
3378 use of these extensions is likely to make other debuggers crash or
3379 refuse to read the program.
3383 Produce debugging information in COFF format (if that is supported).
3384 This is the format used by SDB on most System V systems prior to
3389 Produce debugging information in XCOFF format (if that is supported).
3390 This is the format used by the DBX debugger on IBM RS/6000 systems.
3394 Produce debugging information in XCOFF format (if that is supported),
3395 using GNU extensions understood only by the GNU debugger (GDB)@. The
3396 use of these extensions is likely to make other debuggers crash or
3397 refuse to read the program, and may cause assemblers other than the GNU
3398 assembler (GAS) to fail with an error.
3402 Produce debugging information in DWARF version 2 format (if that is
3403 supported). This is the format used by DBX on IRIX 6. With this
3404 option, GCC uses features of DWARF version 3 when they are useful;
3405 version 3 is upward compatible with version 2, but may still cause
3406 problems for older debuggers.
3410 Produce debugging information in VMS debug format (if that is
3411 supported). This is the format used by DEBUG on VMS systems.
3414 @itemx -ggdb@var{level}
3415 @itemx -gstabs@var{level}
3416 @itemx -gcoff@var{level}
3417 @itemx -gxcoff@var{level}
3418 @itemx -gvms@var{level}
3419 Request debugging information and also use @var{level} to specify how
3420 much information. The default level is 2.
3422 Level 1 produces minimal information, enough for making backtraces in
3423 parts of the program that you don't plan to debug. This includes
3424 descriptions of functions and external variables, but no information
3425 about local variables and no line numbers.
3427 Level 3 includes extra information, such as all the macro definitions
3428 present in the program. Some debuggers support macro expansion when
3429 you use @option{-g3}.
3431 @option{-gdwarf-2} does not accept a concatenated debug level, because
3432 GCC used to support an option @option{-gdwarf} that meant to generate
3433 debug information in version 1 of the DWARF format (which is very
3434 different from version 2), and it would have been too confusing. That
3435 debug format is long obsolete, but the option cannot be changed now.
3436 Instead use an additional @option{-g@var{level}} option to change the
3437 debug level for DWARF2.
3439 @item -feliminate-dwarf2-dups
3440 @opindex feliminate-dwarf2-dups
3441 Compress DWARF2 debugging information by eliminating duplicated
3442 information about each symbol. This option only makes sense when
3443 generating DWARF2 debugging information with @option{-gdwarf-2}.
3445 @cindex @command{prof}
3448 Generate extra code to write profile information suitable for the
3449 analysis program @command{prof}. You must use this option when compiling
3450 the source files you want data about, and you must also use it when
3453 @cindex @command{gprof}
3456 Generate extra code to write profile information suitable for the
3457 analysis program @command{gprof}. You must use this option when compiling
3458 the source files you want data about, and you must also use it when
3463 Makes the compiler print out each function name as it is compiled, and
3464 print some statistics about each pass when it finishes.
3467 @opindex ftime-report
3468 Makes the compiler print some statistics about the time consumed by each
3469 pass when it finishes.
3472 @opindex fmem-report
3473 Makes the compiler print some statistics about permanent memory
3474 allocation when it finishes.
3476 @item -fprofile-arcs
3477 @opindex fprofile-arcs
3478 Add code so that program flow @dfn{arcs} are instrumented. During
3479 execution the program records how many times each branch and call is
3480 executed and how many times it is taken or returns. When the compiled
3481 program exits it saves this data to a file called
3482 @file{@var{auxname}.gcda} for each source file. The data may be used for
3483 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3484 test coverage analysis (@option{-ftest-coverage}). Each object file's
3485 @var{auxname} is generated from the name of the output file, if
3486 explicitly specified and it is not the final executable, otherwise it is
3487 the basename of the source file. In both cases any suffix is removed
3488 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3489 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3490 @xref{Cross-profiling}.
3492 @cindex @command{gcov}
3496 This option is used to compile and link code instrumented for coverage
3497 analysis. The option is a synonym for @option{-fprofile-arcs}
3498 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3499 linking). See the documentation for those options for more details.
3504 Compile the source files with @option{-fprofile-arcs} plus optimization
3505 and code generation options. For test coverage analysis, use the
3506 additional @option{-ftest-coverage} option. You do not need to profile
3507 every source file in a program.
3510 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3511 (the latter implies the former).
3514 Run the program on a representative workload to generate the arc profile
3515 information. This may be repeated any number of times. You can run
3516 concurrent instances of your program, and provided that the file system
3517 supports locking, the data files will be correctly updated. Also
3518 @code{fork} calls are detected and correctly handled (double counting
3522 For profile-directed optimizations, compile the source files again with
3523 the same optimization and code generation options plus
3524 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3525 Control Optimization}).
3528 For test coverage analysis, use @command{gcov} to produce human readable
3529 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3530 @command{gcov} documentation for further information.
3534 With @option{-fprofile-arcs}, for each function of your program GCC
3535 creates a program flow graph, then finds a spanning tree for the graph.
3536 Only arcs that are not on the spanning tree have to be instrumented: the
3537 compiler adds code to count the number of times that these arcs are
3538 executed. When an arc is the only exit or only entrance to a block, the
3539 instrumentation code can be added to the block; otherwise, a new basic
3540 block must be created to hold the instrumentation code.
3542 @item -ftree-based-profiling
3543 @opindex ftree-based-profiling
3544 This option is used in addition to @option{-fprofile-arcs} or
3545 @option{-fbranch-probabilities} to control whether those optimizations
3546 are performed on a tree-based or rtl-based internal representation.
3547 If you use this option when compiling with @option{-fprofile-arcs},
3548 you must also use it when compiling later with @option{-fbranch-probabilities}.
3549 Currently the tree-based optimization is in an early stage of
3550 development, and this option is recommended only for those people
3551 working on improving it.
3554 @item -ftest-coverage
3555 @opindex ftest-coverage
3556 Produce a notes file that the @command{gcov} code-coverage utility
3557 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3558 show program coverage. Each source file's note file is called
3559 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3560 above for a description of @var{auxname} and instructions on how to
3561 generate test coverage data. Coverage data will match the source files
3562 more closely, if you do not optimize.
3564 @item -d@var{letters}
3565 @item -fdump-rtl-@var{pass}
3567 Says to make debugging dumps during compilation at times specified by
3568 @var{letters}. This is used for debugging the RTL-based passes of the
3569 compiler. The file names for most of the dumps are made by appending a
3570 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3571 from the name of the output file, if explicitly specified and it is not
3572 an executable, otherwise it is the basename of the source file.
3574 Most debug dumps can be enabled either passing a letter to the @option{-d}
3575 option, or with a long @option{-fdump-rtl} switch; here are the possible
3576 letters for use in @var{letters} and @var{pass}, and their meanings:
3581 Annotate the assembler output with miscellaneous debugging information.
3584 @itemx -fdump-rtl-bp
3586 @opindex fdump-rtl-bp
3587 Dump after computing branch probabilities, to @file{@var{file}.09.bp}.
3590 @itemx -fdump-rtl-bbro
3592 @opindex fdump-rtl-bbro
3593 Dump after block reordering, to @file{@var{file}.30.bbro}.
3596 @itemx -fdump-rtl-combine
3598 @opindex fdump-rtl-combine
3599 Dump after instruction combination, to the file @file{@var{file}.17.combine}.
3602 @itemx -fdump-rtl-ce1
3603 @itemx -fdump-rtl-ce2
3605 @opindex fdump-rtl-ce1
3606 @opindex fdump-rtl-ce2
3607 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3608 first if conversion, to the file @file{@var{file}.11.ce1}. @option{-dC}
3609 and @option{-fdump-rtl-ce2} enable dumping after the second if
3610 conversion, to the file @file{@var{file}.18.ce2}.
3613 @itemx -fdump-rtl-btl
3614 @itemx -fdump-rtl-dbr
3616 @opindex fdump-rtl-btl
3617 @opindex fdump-rtl-dbr
3618 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3619 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3620 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3621 scheduling, to @file{@var{file}.36.dbr}.
3625 Dump all macro definitions, at the end of preprocessing, in addition to
3629 @itemx -fdump-rtl-ce3
3631 @opindex fdump-rtl-ce3
3632 Dump after the third if conversion, to @file{@var{file}.28.ce3}.
3635 @itemx -fdump-rtl-cfg
3636 @itemx -fdump-rtl-life
3638 @opindex fdump-rtl-cfg
3639 @opindex fdump-rtl-life
3640 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3641 and data flow analysis, to @file{@var{file}.08.cfg}. @option{-df}
3642 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3643 to @file{@var{file}.16.life}.
3646 @itemx -fdump-rtl-greg
3648 @opindex fdump-rtl-greg
3649 Dump after global register allocation, to @file{@var{file}.23.greg}.
3652 @itemx -fdump-rtl-gcse
3653 @itemx -fdump-rtl-bypass
3655 @opindex fdump-rtl-gcse
3656 @opindex fdump-rtl-bypass
3657 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3658 @file{@var{file}.05.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3659 enable dumping after jump bypassing and control flow optimizations, to
3660 @file{@var{file}.07.bypass}.
3663 @itemx -fdump-rtl-eh
3665 @opindex fdump-rtl-eh
3666 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3669 @itemx -fdump-rtl-sibling
3671 @opindex fdump-rtl-sibling
3672 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
3675 @itemx -fdump-rtl-jump
3677 @opindex fdump-rtl-jump
3678 Dump after the first jump optimization, to @file{@var{file}.03.jump}.
3681 @itemx -fdump-rtl-stack
3683 @opindex fdump-rtl-stack
3684 Dump after conversion from registers to stack, to @file{@var{file}.33.stack}.
3687 @itemx -fdump-rtl-lreg
3689 @opindex fdump-rtl-lreg
3690 Dump after local register allocation, to @file{@var{file}.22.lreg}.
3693 @itemx -fdump-rtl-loop
3694 @itemx -fdump-rtl-loop2
3696 @opindex fdump-rtl-loop
3697 @opindex fdump-rtl-loop2
3698 @option{-dL} and @option{-fdump-rtl-loop} enable dumping after the first
3699 loop optimization pass, to @file{@var{file}.06.loop}. @option{-dL} and
3700 @option{-fdump-rtl-loop2} enable dumping after the second pass, to
3701 @file{@var{file}.13.loop2}.
3704 @itemx -fdump-rtl-sms
3706 @opindex fdump-rtl-sms
3707 Dump after modulo scheduling, to @file{@var{file}.20.sms}.
3710 @itemx -fdump-rtl-mach
3712 @opindex fdump-rtl-mach
3713 Dump after performing the machine dependent reorganization pass, to
3714 @file{@var{file}.35.mach}.
3717 @itemx -fdump-rtl-rnreg
3719 @opindex fdump-rtl-rnreg
3720 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3723 @itemx -fdump-rtl-regmove
3725 @opindex fdump-rtl-regmove
3726 Dump after the register move pass, to @file{@var{file}.19.regmove}.
3729 @itemx -fdump-rtl-postreload
3731 @opindex fdump-rtl-postreload
3732 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3735 @itemx -fdump-rtl-expand
3737 @opindex fdump-rtl-expand
3738 Dump after RTL generation, to @file{@var{file}.00.expand}.
3741 @itemx -fdump-rtl-sched2
3743 @opindex fdump-rtl-sched2
3744 Dump after the second scheduling pass, to @file{@var{file}.32.sched2}.
3747 @itemx -fdump-rtl-cse
3749 @opindex fdump-rtl-cse
3750 Dump after CSE (including the jump optimization that sometimes follows
3751 CSE), to @file{@var{file}.04.cse}.
3754 @itemx -fdump-rtl-sched
3756 @opindex fdump-rtl-sched
3757 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3760 @itemx -fdump-rtl-cse2
3762 @opindex fdump-rtl-cse2
3763 Dump after the second CSE pass (including the jump optimization that
3764 sometimes follows CSE), to @file{@var{file}.15.cse2}.
3767 @itemx -fdump-rtl-tracer
3769 @opindex fdump-rtl-tracer
3770 Dump after running tracer, to @file{@var{file}.12.tracer}.
3773 @itemx -fdump-rtl-vpt
3774 @itemx -fdump-rtl-vartrack
3776 @opindex fdump-rtl-vpt
3777 @opindex fdump-rtl-vartrack
3778 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3779 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3780 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3781 to @file{@var{file}.34.vartrack}.
3784 @itemx -fdump-rtl-flow2
3786 @opindex fdump-rtl-flow2
3787 Dump after the second flow pass, to @file{@var{file}.26.flow2}.
3790 @itemx -fdump-rtl-peephole2
3792 @opindex fdump-rtl-peephole2
3793 Dump after the peephole pass, to @file{@var{file}.27.peephole2}.
3796 @itemx -fdump-rtl-web
3798 @opindex fdump-rtl-web
3799 Dump after live range splitting, to @file{@var{file}.14.web}.
3802 @itemx -fdump-rtl-all
3804 @opindex fdump-rtl-all
3805 Produce all the dumps listed above.
3809 Produce a core dump whenever an error occurs.
3813 Print statistics on memory usage, at the end of the run, to
3818 Annotate the assembler output with a comment indicating which
3819 pattern and alternative was used. The length of each instruction is
3824 Dump the RTL in the assembler output as a comment before each instruction.
3825 Also turns on @option{-dp} annotation.
3829 For each of the other indicated dump files (either with @option{-d} or
3830 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3831 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3835 Just generate RTL for a function instead of compiling it. Usually used
3836 with @samp{r} (@option{-fdump-rtl-expand}).
3840 Dump debugging information during parsing, to standard error.
3843 @item -fdump-unnumbered
3844 @opindex fdump-unnumbered
3845 When doing debugging dumps (see @option{-d} option above), suppress instruction
3846 numbers and line number note output. This makes it more feasible to
3847 use diff on debugging dumps for compiler invocations with different
3848 options, in particular with and without @option{-g}.
3850 @item -fdump-translation-unit @r{(C++ only)}
3851 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
3852 @opindex fdump-translation-unit
3853 Dump a representation of the tree structure for the entire translation
3854 unit to a file. The file name is made by appending @file{.tu} to the
3855 source file name. If the @samp{-@var{options}} form is used, @var{options}
3856 controls the details of the dump as described for the
3857 @option{-fdump-tree} options.
3859 @item -fdump-class-hierarchy @r{(C++ only)}
3860 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3861 @opindex fdump-class-hierarchy
3862 Dump a representation of each class's hierarchy and virtual function
3863 table layout to a file. The file name is made by appending @file{.class}
3864 to the source file name. If the @samp{-@var{options}} form is used,
3865 @var{options} controls the details of the dump as described for the
3866 @option{-fdump-tree} options.
3868 @item -fdump-ipa-@var{switch}
3870 Control the dumping at various stages of inter-procedural analysis
3871 language tree to a file. The file name is generated by appending a switch
3872 specific suffix to the source file name. The following dumps are possible:
3876 Enables all inter-procedural analysis dumps; currently the only produced
3877 dump is the @samp{cgraph} dump.
3880 Dumps information about call-graph optimization, unused function removal,
3881 and inlining decisions.
3884 @item -fdump-tree-@var{switch} @r{(C and C++ only)}
3885 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C and C++ only)}
3887 Control the dumping at various stages of processing the intermediate
3888 language tree to a file. The file name is generated by appending a switch
3889 specific suffix to the source file name. If the @samp{-@var{options}}
3890 form is used, @var{options} is a list of @samp{-} separated options that
3891 control the details of the dump. Not all options are applicable to all
3892 dumps, those which are not meaningful will be ignored. The following
3893 options are available
3897 Print the address of each node. Usually this is not meaningful as it
3898 changes according to the environment and source file. Its primary use
3899 is for tying up a dump file with a debug environment.
3901 Inhibit dumping of members of a scope or body of a function merely
3902 because that scope has been reached. Only dump such items when they
3903 are directly reachable by some other path. When dumping pretty-printed
3904 trees, this option inhibits dumping the bodies of control structures.
3906 Print a raw representation of the tree. By default, trees are
3907 pretty-printed into a C-like representation.
3909 Enable more detailed dumps (not honored by every dump option).
3911 Enable dumping various statistics about the pass (not honored by every dump
3914 Enable showing basic block boundaries (disabled in raw dumps).
3916 Enable showing virtual operands for every statement.
3918 Enable showing line numbers for statements.
3920 Enable showing the unique ID (@code{DECL_UID}) for each variable.
3922 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
3925 The following tree dumps are possible:
3929 Dump before any tree based optimization, to @file{@var{file}.original}.
3932 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3935 Dump after function inlining, to @file{@var{file}.inlined}.
3938 @opindex fdump-tree-gimple
3939 Dump each function before and after the gimplification pass to a file. The
3940 file name is made by appending @file{.gimple} to the source file name.
3943 @opindex fdump-tree-cfg
3944 Dump the control flow graph of each function to a file. The file name is
3945 made by appending @file{.cfg} to the source file name.
3948 @opindex fdump-tree-vcg
3949 Dump the control flow graph of each function to a file in VCG format. The
3950 file name is made by appending @file{.vcg} to the source file name. Note
3951 that if the file contains more than one function, the generated file cannot
3952 be used directly by VCG@. You will need to cut and paste each function's
3953 graph into its own separate file first.
3956 @opindex fdump-tree-ch
3957 Dump each function after copying loop headers. The file name is made by
3958 appending @file{.ch} to the source file name.
3961 @opindex fdump-tree-ssa
3962 Dump SSA related information to a file. The file name is made by appending
3963 @file{.ssa} to the source file name.
3966 @opindex fdump-tree-salias
3967 Dump structure aliasing variable information to a file. This file name
3968 is made by appending @file{.salias} to the source file name.
3971 @opindex fdump-tree-alias
3972 Dump aliasing information for each function. The file name is made by
3973 appending @file{.alias} to the source file name.
3976 @opindex fdump-tree-ccp
3977 Dump each function after CCP@. The file name is made by appending
3978 @file{.ccp} to the source file name.
3981 @opindex fdump-tree-storeccp
3982 Dump each function after STORE-CCP. The file name is made by appending
3983 @file{.storeccp} to the source file name.
3986 @opindex fdump-tree-pre
3987 Dump trees after partial redundancy elimination. The file name is made
3988 by appending @file{.pre} to the source file name.
3991 @opindex fdump-tree-fre
3992 Dump trees after full redundancy elimination. The file name is made
3993 by appending @file{.fre} to the source file name.
3996 @opindex fdump-tree-copyprop
3997 Dump trees after copy propagation. The file name is made
3998 by appending @file{.copyprop} to the source file name.
4000 @item store_copyprop
4001 @opindex fdump-tree-store_copyprop
4002 Dump trees after store copy-propagation. The file name is made
4003 by appending @file{.store_copyprop} to the source file name.
4006 @opindex fdump-tree-dce
4007 Dump each function after dead code elimination. The file name is made by
4008 appending @file{.dce} to the source file name.
4011 @opindex fdump-tree-mudflap
4012 Dump each function after adding mudflap instrumentation. The file name is
4013 made by appending @file{.mudflap} to the source file name.
4016 @opindex fdump-tree-sra
4017 Dump each function after performing scalar replacement of aggregates. The
4018 file name is made by appending @file{.sra} to the source file name.
4021 @opindex fdump-tree-sink
4022 Dump each function after performing code sinking. The file name is made
4023 by appending @file{.sink} to the source file name.
4026 @opindex fdump-tree-dom
4027 Dump each function after applying dominator tree optimizations. The file
4028 name is made by appending @file{.dom} to the source file name.
4031 @opindex fdump-tree-dse
4032 Dump each function after applying dead store elimination. The file
4033 name is made by appending @file{.dse} to the source file name.
4036 @opindex fdump-tree-phiopt
4037 Dump each function after optimizing PHI nodes into straightline code. The file
4038 name is made by appending @file{.phiopt} to the source file name.
4041 @opindex fdump-tree-forwprop
4042 Dump each function after forward propagating single use variables. The file
4043 name is made by appending @file{.forwprop} to the source file name.
4046 @opindex fdump-tree-copyrename
4047 Dump each function after applying the copy rename optimization. The file
4048 name is made by appending @file{.copyrename} to the source file name.
4051 @opindex fdump-tree-nrv
4052 Dump each function after applying the named return value optimization on
4053 generic trees. The file name is made by appending @file{.nrv} to the source
4057 @opindex fdump-tree-vect
4058 Dump each function after applying vectorization of loops. The file name is
4059 made by appending @file{.vect} to the source file name.
4062 @opindex fdump-tree-all
4063 Enable all the available tree dumps with the flags provided in this option.
4066 @item -ftree-vectorizer-verbose=@var{n}
4067 @opindex ftree-vectorizer-verbose
4068 This option controls the amount of debugging output the vectorizer prints.
4069 This information is written to standard error, unless @option{-fdump-tree-all}
4070 or @option{-fdump-tree-vect} is specified, in which case it is output to the
4071 usual dump listing file, @file{.vect}.
4073 @item -frandom-seed=@var{string}
4074 @opindex frandom-string
4075 This option provides a seed that GCC uses when it would otherwise use
4076 random numbers. It is used to generate certain symbol names
4077 that have to be different in every compiled file. It is also used to
4078 place unique stamps in coverage data files and the object files that
4079 produce them. You can use the @option{-frandom-seed} option to produce
4080 reproducibly identical object files.
4082 The @var{string} should be different for every file you compile.
4084 @item -fsched-verbose=@var{n}
4085 @opindex fsched-verbose
4086 On targets that use instruction scheduling, this option controls the
4087 amount of debugging output the scheduler prints. This information is
4088 written to standard error, unless @option{-dS} or @option{-dR} is
4089 specified, in which case it is output to the usual dump
4090 listing file, @file{.sched} or @file{.sched2} respectively. However
4091 for @var{n} greater than nine, the output is always printed to standard
4094 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4095 same information as @option{-dRS}. For @var{n} greater than one, it
4096 also output basic block probabilities, detailed ready list information
4097 and unit/insn info. For @var{n} greater than two, it includes RTL
4098 at abort point, control-flow and regions info. And for @var{n} over
4099 four, @option{-fsched-verbose} also includes dependence info.
4103 Store the usual ``temporary'' intermediate files permanently; place them
4104 in the current directory and name them based on the source file. Thus,
4105 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4106 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4107 preprocessed @file{foo.i} output file even though the compiler now
4108 normally uses an integrated preprocessor.
4110 When used in combination with the @option{-x} command line option,
4111 @option{-save-temps} is sensible enough to avoid over writing an
4112 input source file with the same extension as an intermediate file.
4113 The corresponding intermediate file may be obtained by renaming the
4114 source file before using @option{-save-temps}.
4118 Report the CPU time taken by each subprocess in the compilation
4119 sequence. For C source files, this is the compiler proper and assembler
4120 (plus the linker if linking is done). The output looks like this:
4127 The first number on each line is the ``user time'', that is time spent
4128 executing the program itself. The second number is ``system time'',
4129 time spent executing operating system routines on behalf of the program.
4130 Both numbers are in seconds.
4132 @item -fvar-tracking
4133 @opindex fvar-tracking
4134 Run variable tracking pass. It computes where variables are stored at each
4135 position in code. Better debugging information is then generated
4136 (if the debugging information format supports this information).
4138 It is enabled by default when compiling with optimization (@option{-Os},
4139 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4140 the debug info format supports it.
4142 @item -print-file-name=@var{library}
4143 @opindex print-file-name
4144 Print the full absolute name of the library file @var{library} that
4145 would be used when linking---and don't do anything else. With this
4146 option, GCC does not compile or link anything; it just prints the
4149 @item -print-multi-directory
4150 @opindex print-multi-directory
4151 Print the directory name corresponding to the multilib selected by any
4152 other switches present in the command line. This directory is supposed
4153 to exist in @env{GCC_EXEC_PREFIX}.
4155 @item -print-multi-lib
4156 @opindex print-multi-lib
4157 Print the mapping from multilib directory names to compiler switches
4158 that enable them. The directory name is separated from the switches by
4159 @samp{;}, and each switch starts with an @samp{@@} instead of the
4160 @samp{-}, without spaces between multiple switches. This is supposed to
4161 ease shell-processing.
4163 @item -print-prog-name=@var{program}
4164 @opindex print-prog-name
4165 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4167 @item -print-libgcc-file-name
4168 @opindex print-libgcc-file-name
4169 Same as @option{-print-file-name=libgcc.a}.
4171 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4172 but you do want to link with @file{libgcc.a}. You can do
4175 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4178 @item -print-search-dirs
4179 @opindex print-search-dirs
4180 Print the name of the configured installation directory and a list of
4181 program and library directories @command{gcc} will search---and don't do anything else.
4183 This is useful when @command{gcc} prints the error message
4184 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4185 To resolve this you either need to put @file{cpp0} and the other compiler
4186 components where @command{gcc} expects to find them, or you can set the environment
4187 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4188 Don't forget the trailing @samp{/}.
4189 @xref{Environment Variables}.
4192 @opindex dumpmachine
4193 Print the compiler's target machine (for example,
4194 @samp{i686-pc-linux-gnu})---and don't do anything else.
4197 @opindex dumpversion
4198 Print the compiler version (for example, @samp{3.0})---and don't do
4203 Print the compiler's built-in specs---and don't do anything else. (This
4204 is used when GCC itself is being built.) @xref{Spec Files}.
4206 @item -feliminate-unused-debug-types
4207 @opindex feliminate-unused-debug-types
4208 Normally, when producing DWARF2 output, GCC will emit debugging
4209 information for all types declared in a compilation
4210 unit, regardless of whether or not they are actually used
4211 in that compilation unit. Sometimes this is useful, such as
4212 if, in the debugger, you want to cast a value to a type that is
4213 not actually used in your program (but is declared). More often,
4214 however, this results in a significant amount of wasted space.
4215 With this option, GCC will avoid producing debug symbol output
4216 for types that are nowhere used in the source file being compiled.
4219 @node Optimize Options
4220 @section Options That Control Optimization
4221 @cindex optimize options
4222 @cindex options, optimization
4224 These options control various sorts of optimizations.
4226 Without any optimization option, the compiler's goal is to reduce the
4227 cost of compilation and to make debugging produce the expected
4228 results. Statements are independent: if you stop the program with a
4229 breakpoint between statements, you can then assign a new value to any
4230 variable or change the program counter to any other statement in the
4231 function and get exactly the results you would expect from the source
4234 Turning on optimization flags makes the compiler attempt to improve
4235 the performance and/or code size at the expense of compilation time
4236 and possibly the ability to debug the program.
4238 The compiler performs optimization based on the knowledge it has of
4239 the program. Optimization levels @option{-O2} and above, in
4240 particular, enable @emph{unit-at-a-time} mode, which allows the
4241 compiler to consider information gained from later functions in
4242 the file when compiling a function. Compiling multiple files at
4243 once to a single output file in @emph{unit-at-a-time} mode allows
4244 the compiler to use information gained from all of the files when
4245 compiling each of them.
4247 Not all optimizations are controlled directly by a flag. Only
4248 optimizations that have a flag are listed.
4255 Optimize. Optimizing compilation takes somewhat more time, and a lot
4256 more memory for a large function.
4258 With @option{-O}, the compiler tries to reduce code size and execution
4259 time, without performing any optimizations that take a great deal of
4262 @option{-O} turns on the following optimization flags:
4263 @gccoptlist{-fdefer-pop @gol
4264 -fdelayed-branch @gol
4265 -fguess-branch-probability @gol
4266 -fcprop-registers @gol
4267 -floop-optimize @gol
4268 -fif-conversion @gol
4269 -fif-conversion2 @gol
4272 -ftree-dominator-opts @gol
4277 -ftree-copyrename @gol
4282 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4283 where doing so does not interfere with debugging.
4287 Optimize even more. GCC performs nearly all supported optimizations
4288 that do not involve a space-speed tradeoff. The compiler does not
4289 perform loop unrolling or function inlining when you specify @option{-O2}.
4290 As compared to @option{-O}, this option increases both compilation time
4291 and the performance of the generated code.
4293 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4294 also turns on the following optimization flags:
4295 @gccoptlist{-fthread-jumps @gol
4297 -foptimize-sibling-calls @gol
4298 -fcse-follow-jumps -fcse-skip-blocks @gol
4299 -fgcse -fgcse-lm @gol
4300 -fexpensive-optimizations @gol
4301 -fstrength-reduce @gol
4302 -frerun-cse-after-loop -frerun-loop-opt @gol
4306 -fschedule-insns -fschedule-insns2 @gol
4307 -fsched-interblock -fsched-spec @gol
4309 -fstrict-aliasing @gol
4310 -fdelete-null-pointer-checks @gol
4311 -freorder-blocks -freorder-functions @gol
4312 -funit-at-a-time @gol
4313 -falign-functions -falign-jumps @gol
4314 -falign-loops -falign-labels @gol
4317 Please note the warning under @option{-fgcse} about
4318 invoking @option{-O2} on programs that use computed gotos.
4322 Optimize yet more. @option{-O3} turns on all optimizations specified by
4323 @option{-O2} and also turns on the @option{-finline-functions},
4324 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4328 Do not optimize. This is the default.
4332 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4333 do not typically increase code size. It also performs further
4334 optimizations designed to reduce code size.
4336 @option{-Os} disables the following optimization flags:
4337 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4338 -falign-labels -freorder-blocks -freorder-blocks-and-partition -fprefetch-loop-arrays}
4340 If you use multiple @option{-O} options, with or without level numbers,
4341 the last such option is the one that is effective.
4344 Options of the form @option{-f@var{flag}} specify machine-independent
4345 flags. Most flags have both positive and negative forms; the negative
4346 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4347 below, only one of the forms is listed---the one you typically will
4348 use. You can figure out the other form by either removing @samp{no-}
4351 The following options control specific optimizations. They are either
4352 activated by @option{-O} options or are related to ones that are. You
4353 can use the following flags in the rare cases when ``fine-tuning'' of
4354 optimizations to be performed is desired.
4357 @item -fno-default-inline
4358 @opindex fno-default-inline
4359 Do not make member functions inline by default merely because they are
4360 defined inside the class scope (C++ only). Otherwise, when you specify
4361 @w{@option{-O}}, member functions defined inside class scope are compiled
4362 inline by default; i.e., you don't need to add @samp{inline} in front of
4363 the member function name.
4365 @item -fno-defer-pop
4366 @opindex fno-defer-pop
4367 Always pop the arguments to each function call as soon as that function
4368 returns. For machines which must pop arguments after a function call,
4369 the compiler normally lets arguments accumulate on the stack for several
4370 function calls and pops them all at once.
4372 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4376 Force memory operands to be copied into registers before doing
4377 arithmetic on them. This produces better code by making all memory
4378 references potential common subexpressions. When they are not common
4379 subexpressions, instruction combination should eliminate the separate
4382 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4385 @opindex fforce-addr
4386 Force memory address constants to be copied into registers before
4387 doing arithmetic on them. This may produce better code just as
4388 @option{-fforce-mem} may.
4390 @item -fomit-frame-pointer
4391 @opindex fomit-frame-pointer
4392 Don't keep the frame pointer in a register for functions that
4393 don't need one. This avoids the instructions to save, set up and
4394 restore frame pointers; it also makes an extra register available
4395 in many functions. @strong{It also makes debugging impossible on
4398 On some machines, such as the VAX, this flag has no effect, because
4399 the standard calling sequence automatically handles the frame pointer
4400 and nothing is saved by pretending it doesn't exist. The
4401 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4402 whether a target machine supports this flag. @xref{Registers,,Register
4403 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4405 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4407 @item -foptimize-sibling-calls
4408 @opindex foptimize-sibling-calls
4409 Optimize sibling and tail recursive calls.
4411 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4415 Don't pay attention to the @code{inline} keyword. Normally this option
4416 is used to keep the compiler from expanding any functions inline.
4417 Note that if you are not optimizing, no functions can be expanded inline.
4419 @item -finline-functions
4420 @opindex finline-functions
4421 Integrate all simple functions into their callers. The compiler
4422 heuristically decides which functions are simple enough to be worth
4423 integrating in this way.
4425 If all calls to a given function are integrated, and the function is
4426 declared @code{static}, then the function is normally not output as
4427 assembler code in its own right.
4429 Enabled at level @option{-O3}.
4431 @item -finline-limit=@var{n}
4432 @opindex finline-limit
4433 By default, GCC limits the size of functions that can be inlined. This flag
4434 allows the control of this limit for functions that are explicitly marked as
4435 inline (i.e., marked with the inline keyword or defined within the class
4436 definition in c++). @var{n} is the size of functions that can be inlined in
4437 number of pseudo instructions (not counting parameter handling). The default
4438 value of @var{n} is 600.
4439 Increasing this value can result in more inlined code at
4440 the cost of compilation time and memory consumption. Decreasing usually makes
4441 the compilation faster and less code will be inlined (which presumably
4442 means slower programs). This option is particularly useful for programs that
4443 use inlining heavily such as those based on recursive templates with C++.
4445 Inlining is actually controlled by a number of parameters, which may be
4446 specified individually by using @option{--param @var{name}=@var{value}}.
4447 The @option{-finline-limit=@var{n}} option sets some of these parameters
4451 @item max-inline-insns-single
4452 is set to @var{n}/2.
4453 @item max-inline-insns-auto
4454 is set to @var{n}/2.
4455 @item min-inline-insns
4456 is set to 130 or @var{n}/4, whichever is smaller.
4457 @item max-inline-insns-rtl
4461 See below for a documentation of the individual
4462 parameters controlling inlining.
4464 @emph{Note:} pseudo instruction represents, in this particular context, an
4465 abstract measurement of function's size. In no way, it represents a count
4466 of assembly instructions and as such its exact meaning might change from one
4467 release to an another.
4469 @item -fkeep-inline-functions
4470 @opindex fkeep-inline-functions
4471 In C, emit @code{static} functions that are declared @code{inline}
4472 into the object file, even if the function has been inlined into all
4473 of its callers. This switch does not affect functions using the
4474 @code{extern inline} extension in GNU C@. In C++, emit any and all
4475 inline functions into the object file.
4477 @item -fkeep-static-consts
4478 @opindex fkeep-static-consts
4479 Emit variables declared @code{static const} when optimization isn't turned
4480 on, even if the variables aren't referenced.
4482 GCC enables this option by default. If you want to force the compiler to
4483 check if the variable was referenced, regardless of whether or not
4484 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4486 @item -fmerge-constants
4487 Attempt to merge identical constants (string constants and floating point
4488 constants) across compilation units.
4490 This option is the default for optimized compilation if the assembler and
4491 linker support it. Use @option{-fno-merge-constants} to inhibit this
4494 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4496 @item -fmerge-all-constants
4497 Attempt to merge identical constants and identical variables.
4499 This option implies @option{-fmerge-constants}. In addition to
4500 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4501 arrays or initialized constant variables with integral or floating point
4502 types. Languages like C or C++ require each non-automatic variable to
4503 have distinct location, so using this option will result in non-conforming
4506 @item -fmodulo-sched
4507 @opindex fmodulo-sched
4508 Perform swing modulo scheduling immediately before the first scheduling
4509 pass. This pass looks at innermost loops and reorders their
4510 instructions by overlapping different iterations.
4512 @item -fno-branch-count-reg
4513 @opindex fno-branch-count-reg
4514 Do not use ``decrement and branch'' instructions on a count register,
4515 but instead generate a sequence of instructions that decrement a
4516 register, compare it against zero, then branch based upon the result.
4517 This option is only meaningful on architectures that support such
4518 instructions, which include x86, PowerPC, IA-64 and S/390.
4520 The default is @option{-fbranch-count-reg}, enabled when
4521 @option{-fstrength-reduce} is enabled.
4523 @item -fno-function-cse
4524 @opindex fno-function-cse
4525 Do not put function addresses in registers; make each instruction that
4526 calls a constant function contain the function's address explicitly.
4528 This option results in less efficient code, but some strange hacks
4529 that alter the assembler output may be confused by the optimizations
4530 performed when this option is not used.
4532 The default is @option{-ffunction-cse}
4534 @item -fno-zero-initialized-in-bss
4535 @opindex fno-zero-initialized-in-bss
4536 If the target supports a BSS section, GCC by default puts variables that
4537 are initialized to zero into BSS@. This can save space in the resulting
4540 This option turns off this behavior because some programs explicitly
4541 rely on variables going to the data section. E.g., so that the
4542 resulting executable can find the beginning of that section and/or make
4543 assumptions based on that.
4545 The default is @option{-fzero-initialized-in-bss}.
4547 @item -fbounds-check
4548 @opindex fbounds-check
4549 For front-ends that support it, generate additional code to check that
4550 indices used to access arrays are within the declared range. This is
4551 currently only supported by the Java and Fortran front-ends, where
4552 this option defaults to true and false respectively.
4554 @item -fmudflap -fmudflapth -fmudflapir
4558 @cindex bounds checking
4560 For front-ends that support it (C and C++), instrument all risky
4561 pointer/array dereferencing operations, some standard library
4562 string/heap functions, and some other associated constructs with
4563 range/validity tests. Modules so instrumented should be immune to
4564 buffer overflows, invalid heap use, and some other classes of C/C++
4565 programming errors. The instrumentation relies on a separate runtime
4566 library (@file{libmudflap}), which will be linked into a program if
4567 @option{-fmudflap} is given at link time. Run-time behavior of the
4568 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4569 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4572 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4573 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4574 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4575 instrumentation should ignore pointer reads. This produces less
4576 instrumentation (and therefore faster execution) and still provides
4577 some protection against outright memory corrupting writes, but allows
4578 erroneously read data to propagate within a program.
4580 @item -fstrength-reduce
4581 @opindex fstrength-reduce
4582 Perform the optimizations of loop strength reduction and
4583 elimination of iteration variables.
4585 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4587 @item -fthread-jumps
4588 @opindex fthread-jumps
4589 Perform optimizations where we check to see if a jump branches to a
4590 location where another comparison subsumed by the first is found. If
4591 so, the first branch is redirected to either the destination of the
4592 second branch or a point immediately following it, depending on whether
4593 the condition is known to be true or false.
4595 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4597 @item -fcse-follow-jumps
4598 @opindex fcse-follow-jumps
4599 In common subexpression elimination, scan through jump instructions
4600 when the target of the jump is not reached by any other path. For
4601 example, when CSE encounters an @code{if} statement with an
4602 @code{else} clause, CSE will follow the jump when the condition
4605 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4607 @item -fcse-skip-blocks
4608 @opindex fcse-skip-blocks
4609 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4610 follow jumps which conditionally skip over blocks. When CSE
4611 encounters a simple @code{if} statement with no else clause,
4612 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4613 body of the @code{if}.
4615 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4617 @item -frerun-cse-after-loop
4618 @opindex frerun-cse-after-loop
4619 Re-run common subexpression elimination after loop optimizations has been
4622 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4624 @item -frerun-loop-opt
4625 @opindex frerun-loop-opt
4626 Run the loop optimizer twice.
4628 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4632 Perform a global common subexpression elimination pass.
4633 This pass also performs global constant and copy propagation.
4635 @emph{Note:} When compiling a program using computed gotos, a GCC
4636 extension, you may get better runtime performance if you disable
4637 the global common subexpression elimination pass by adding
4638 @option{-fno-gcse} to the command line.
4640 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4644 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4645 attempt to move loads which are only killed by stores into themselves. This
4646 allows a loop containing a load/store sequence to be changed to a load outside
4647 the loop, and a copy/store within the loop.
4649 Enabled by default when gcse is enabled.
4653 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4654 global common subexpression elimination. This pass will attempt to move
4655 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4656 loops containing a load/store sequence can be changed to a load before
4657 the loop and a store after the loop.
4659 Not enabled at any optimization level.
4663 When @option{-fgcse-las} is enabled, the global common subexpression
4664 elimination pass eliminates redundant loads that come after stores to the
4665 same memory location (both partial and full redundancies).
4667 Not enabled at any optimization level.
4669 @item -fgcse-after-reload
4670 @opindex fgcse-after-reload
4671 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4672 pass is performed after reload. The purpose of this pass is to cleanup
4675 @item -floop-optimize
4676 @opindex floop-optimize
4677 Perform loop optimizations: move constant expressions out of loops, simplify
4678 exit test conditions and optionally do strength-reduction as well.
4680 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4682 @item -floop-optimize2
4683 @opindex floop-optimize2
4684 Perform loop optimizations using the new loop optimizer. The optimizations
4685 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4688 @item -fcrossjumping
4689 @opindex crossjumping
4690 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4691 resulting code may or may not perform better than without cross-jumping.
4693 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4695 @item -fif-conversion
4696 @opindex if-conversion
4697 Attempt to transform conditional jumps into branch-less equivalents. This
4698 include use of conditional moves, min, max, set flags and abs instructions, and
4699 some tricks doable by standard arithmetics. The use of conditional execution
4700 on chips where it is available is controlled by @code{if-conversion2}.
4702 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4704 @item -fif-conversion2
4705 @opindex if-conversion2
4706 Use conditional execution (where available) to transform conditional jumps into
4707 branch-less equivalents.
4709 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4711 @item -fdelete-null-pointer-checks
4712 @opindex fdelete-null-pointer-checks
4713 Use global dataflow analysis to identify and eliminate useless checks
4714 for null pointers. The compiler assumes that dereferencing a null
4715 pointer would have halted the program. If a pointer is checked after
4716 it has already been dereferenced, it cannot be null.
4718 In some environments, this assumption is not true, and programs can
4719 safely dereference null pointers. Use
4720 @option{-fno-delete-null-pointer-checks} to disable this optimization
4721 for programs which depend on that behavior.
4723 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4725 @item -fexpensive-optimizations
4726 @opindex fexpensive-optimizations
4727 Perform a number of minor optimizations that are relatively expensive.
4729 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4731 @item -foptimize-register-move
4733 @opindex foptimize-register-move
4735 Attempt to reassign register numbers in move instructions and as
4736 operands of other simple instructions in order to maximize the amount of
4737 register tying. This is especially helpful on machines with two-operand
4740 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4743 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4745 @item -fdelayed-branch
4746 @opindex fdelayed-branch
4747 If supported for the target machine, attempt to reorder instructions
4748 to exploit instruction slots available after delayed branch
4751 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4753 @item -fschedule-insns
4754 @opindex fschedule-insns
4755 If supported for the target machine, attempt to reorder instructions to
4756 eliminate execution stalls due to required data being unavailable. This
4757 helps machines that have slow floating point or memory load instructions
4758 by allowing other instructions to be issued until the result of the load
4759 or floating point instruction is required.
4761 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4763 @item -fschedule-insns2
4764 @opindex fschedule-insns2
4765 Similar to @option{-fschedule-insns}, but requests an additional pass of
4766 instruction scheduling after register allocation has been done. This is
4767 especially useful on machines with a relatively small number of
4768 registers and where memory load instructions take more than one cycle.
4770 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4772 @item -fno-sched-interblock
4773 @opindex fno-sched-interblock
4774 Don't schedule instructions across basic blocks. This is normally
4775 enabled by default when scheduling before register allocation, i.e.@:
4776 with @option{-fschedule-insns} or at @option{-O2} or higher.
4778 @item -fno-sched-spec
4779 @opindex fno-sched-spec
4780 Don't allow speculative motion of non-load instructions. This is normally
4781 enabled by default when scheduling before register allocation, i.e.@:
4782 with @option{-fschedule-insns} or at @option{-O2} or higher.
4784 @item -fsched-spec-load
4785 @opindex fsched-spec-load
4786 Allow speculative motion of some load instructions. This only makes
4787 sense when scheduling before register allocation, i.e.@: with
4788 @option{-fschedule-insns} or at @option{-O2} or higher.
4790 @item -fsched-spec-load-dangerous
4791 @opindex fsched-spec-load-dangerous
4792 Allow speculative motion of more load instructions. This only makes
4793 sense when scheduling before register allocation, i.e.@: with
4794 @option{-fschedule-insns} or at @option{-O2} or higher.
4796 @item -fsched-stalled-insns=@var{n}
4797 @opindex fsched-stalled-insns
4798 Define how many insns (if any) can be moved prematurely from the queue
4799 of stalled insns into the ready list, during the second scheduling pass.
4801 @item -fsched-stalled-insns-dep=@var{n}
4802 @opindex fsched-stalled-insns-dep
4803 Define how many insn groups (cycles) will be examined for a dependency
4804 on a stalled insn that is candidate for premature removal from the queue
4805 of stalled insns. Has an effect only during the second scheduling pass,
4806 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4808 @item -fsched2-use-superblocks
4809 @opindex fsched2-use-superblocks
4810 When scheduling after register allocation, do use superblock scheduling
4811 algorithm. Superblock scheduling allows motion across basic block boundaries
4812 resulting on faster schedules. This option is experimental, as not all machine
4813 descriptions used by GCC model the CPU closely enough to avoid unreliable
4814 results from the algorithm.
4816 This only makes sense when scheduling after register allocation, i.e.@: with
4817 @option{-fschedule-insns2} or at @option{-O2} or higher.
4819 @item -fsched2-use-traces
4820 @opindex fsched2-use-traces
4821 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4822 allocation and additionally perform code duplication in order to increase the
4823 size of superblocks using tracer pass. See @option{-ftracer} for details on
4826 This mode should produce faster but significantly longer programs. Also
4827 without @option{-fbranch-probabilities} the traces constructed may not
4828 match the reality and hurt the performance. This only makes
4829 sense when scheduling after register allocation, i.e.@: with
4830 @option{-fschedule-insns2} or at @option{-O2} or higher.
4832 @item -freschedule-modulo-scheduled-loops
4833 @opindex fscheduling-in-modulo-scheduled-loops
4834 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4835 we may want to prevent the later scheduling passes from changing its schedule, we use this
4836 option to control that.
4838 @item -fcaller-saves
4839 @opindex fcaller-saves
4840 Enable values to be allocated in registers that will be clobbered by
4841 function calls, by emitting extra instructions to save and restore the
4842 registers around such calls. Such allocation is done only when it
4843 seems to result in better code than would otherwise be produced.
4845 This option is always enabled by default on certain machines, usually
4846 those which have no call-preserved registers to use instead.
4848 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4851 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4852 enabled by default at @option{-O2} and @option{-O3}.
4855 Perform Full Redundancy Elimination (FRE) on trees. The difference
4856 between FRE and PRE is that FRE only considers expressions
4857 that are computed on all paths leading to the redundant computation.
4858 This analysis faster than PRE, though it exposes fewer redundancies.
4859 This flag is enabled by default at @option{-O} and higher.
4861 @item -ftree-copy-prop
4862 Perform copy propagation on trees. This pass eliminates unnecessary
4863 copy operations. This flag is enabled by default at @option{-O} and
4866 @item -ftree-store-copy-prop
4867 Perform copy propagation of memory loads and stores. This pass
4868 eliminates unnecessary copy operations in memory references
4869 (structures, global variables, arrays, etc). This flag is enabled by
4870 default at @option{-O2} and higher.
4873 Perform structural alias analysis on trees. This flag
4874 is enabled by default at @option{-O} and higher.
4877 Perform forward store motion on trees. This flag is
4878 enabled by default at @option{-O} and higher.
4881 Perform sparse conditional constant propagation (CCP) on trees. This
4882 pass only operates on local scalar variables and is enabled by default
4883 at @option{-O} and higher.
4885 @item -ftree-store-ccp
4886 Perform sparse conditional constant propagation (CCP) on trees. This
4887 pass operates on both local scalar variables and memory stores and
4888 loads (global variables, structures, arrays, etc). This flag is
4889 enabled by default at @option{-O2} and higher.
4892 Perform dead code elimination (DCE) on trees. This flag is enabled by
4893 default at @option{-O} and higher.
4895 @item -ftree-dominator-opts
4896 Perform dead code elimination (DCE) on trees. This flag is enabled by
4897 default at @option{-O} and higher.
4900 Perform loop header copying on trees. This is beneficial since it increases
4901 effectiveness of code motion optimizations. It also saves one jump. This flag
4902 is enabled by default at @option{-O} and higher. It is not enabled
4903 for @option{-Os}, since it usually increases code size.
4905 @item -ftree-loop-optimize
4906 Perform loop optimizations on trees. This flag is enabled by default
4907 at @option{-O} and higher.
4909 @item -ftree-loop-linear
4910 Perform linear loop transformations on tree. This flag can improve cache
4911 performance and allow further loop optimizations to take place.
4913 @item -ftree-loop-im
4914 Perform loop invariant motion on trees. This pass moves only invariants that
4915 would be hard to handle at RTL level (function calls, operations that expand to
4916 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
4917 operands of conditions that are invariant out of the loop, so that we can use
4918 just trivial invariantness analysis in loop unswitching. The pass also includes
4921 @item -ftree-loop-ivcanon
4922 Create a canonical counter for number of iterations in the loop for that
4923 determining number of iterations requires complicated analysis. Later
4924 optimizations then may determine the number easily. Useful especially
4925 in connection with unrolling.
4928 Perform induction variable optimizations (strength reduction, induction
4929 variable merging and induction variable elimination) on trees.
4932 Perform scalar replacement of aggregates. This pass replaces structure
4933 references with scalars to prevent committing structures to memory too
4934 early. This flag is enabled by default at @option{-O} and higher.
4936 @item -ftree-copyrename
4937 Perform copy renaming on trees. This pass attempts to rename compiler
4938 temporaries to other variables at copy locations, usually resulting in
4939 variable names which more closely resemble the original variables. This flag
4940 is enabled by default at @option{-O} and higher.
4943 Perform temporary expression replacement during the SSA->normal phase. Single
4944 use/single def temporaries are replaced at their use location with their
4945 defining expression. This results in non-GIMPLE code, but gives the expanders
4946 much more complex trees to work on resulting in better RTL generation. This is
4947 enabled by default at @option{-O} and higher.
4950 Perform live range splitting during the SSA->normal phase. Distinct live
4951 ranges of a variable are split into unique variables, allowing for better
4952 optimization later. This is enabled by default at @option{-O} and higher.
4954 @item -ftree-vectorize
4955 Perform loop vectorization on trees.
4959 Perform tail duplication to enlarge superblock size. This transformation
4960 simplifies the control flow of the function allowing other optimizations to do
4963 @item -funroll-loops
4964 @opindex funroll-loops
4965 Unroll loops whose number of iterations can be determined at compile
4966 time or upon entry to the loop. @option{-funroll-loops} implies both
4967 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4968 option makes code larger, and may or may not make it run faster.
4970 @item -funroll-all-loops
4971 @opindex funroll-all-loops
4972 Unroll all loops, even if their number of iterations is uncertain when
4973 the loop is entered. This usually makes programs run more slowly.
4974 @option{-funroll-all-loops} implies the same options as
4975 @option{-funroll-loops},
4977 @item -fsplit-ivs-in-unroller
4978 @opindex -fsplit-ivs-in-unroller
4979 Enables expressing of values of induction variables in later iterations
4980 of the unrolled loop using the value in the first iteration. This breaks
4981 long dependency chains, thus improving efficiency of the scheduling passes
4982 (for best results, @option{-fweb} should be used as well).
4984 Combination of @option{-fweb} and CSE is often sufficient to obtain the
4985 same effect. However in cases the loop body is more complicated than
4986 a single basic block, this is not reliable. It also does not work at all
4987 on some of the architectures due to restrictions in the CSE pass.
4989 This optimization is enabled by default.
4991 @item -fvariable-expansion-in-unroller
4992 @opindex -fvariable-expansion-in-unroller
4993 With this option, the compiler will create multiple copies of some
4994 local variables when unrolling a loop which can result in superior code.
4996 @item -fprefetch-loop-arrays
4997 @opindex fprefetch-loop-arrays
4998 If supported by the target machine, generate instructions to prefetch
4999 memory to improve the performance of loops that access large arrays.
5001 These options may generate better or worse code; results are highly
5002 dependent on the structure of loops within the source code.
5005 @itemx -fno-peephole2
5006 @opindex fno-peephole
5007 @opindex fno-peephole2
5008 Disable any machine-specific peephole optimizations. The difference
5009 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5010 are implemented in the compiler; some targets use one, some use the
5011 other, a few use both.
5013 @option{-fpeephole} is enabled by default.
5014 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5016 @item -fno-guess-branch-probability
5017 @opindex fno-guess-branch-probability
5018 Do not guess branch probabilities using heuristics.
5020 GCC will use heuristics to guess branch probabilities if they are
5021 not provided by profiling feedback (@option{-fprofile-arcs}). These
5022 heuristics are based on the control flow graph. If some branch probabilities
5023 are specified by @samp{__builtin_expect}, then the heuristics will be
5024 used to guess branch probabilities for the rest of the control flow graph,
5025 taking the @samp{__builtin_expect} info into account. The interactions
5026 between the heuristics and @samp{__builtin_expect} can be complex, and in
5027 some cases, it may be useful to disable the heuristics so that the effects
5028 of @samp{__builtin_expect} are easier to understand.
5030 The default is @option{-fguess-branch-probability} at levels
5031 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5033 @item -freorder-blocks
5034 @opindex freorder-blocks
5035 Reorder basic blocks in the compiled function in order to reduce number of
5036 taken branches and improve code locality.
5038 Enabled at levels @option{-O2}, @option{-O3}.
5040 @item -freorder-blocks-and-partition
5041 @opindex freorder-blocks-and-partition
5042 In addition to reordering basic blocks in the compiled function, in order
5043 to reduce number of taken branches, partitions hot and cold basic blocks
5044 into separate sections of the assembly and .o files, to improve
5045 paging and cache locality performance.
5047 This optimization is automatically turned off in the presence of
5048 exception handling, for linkonce sections, for functions with a user-defined
5049 section attribute and on any architecture that does not support named
5052 @item -freorder-functions
5053 @opindex freorder-functions
5054 Reorder functions in the object file in order to
5055 improve code locality. This is implemented by using special
5056 subsections @code{.text.hot} for most frequently executed functions and
5057 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5058 the linker so object file format must support named sections and linker must
5059 place them in a reasonable way.
5061 Also profile feedback must be available in to make this option effective. See
5062 @option{-fprofile-arcs} for details.
5064 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5066 @item -fstrict-aliasing
5067 @opindex fstrict-aliasing
5068 Allows the compiler to assume the strictest aliasing rules applicable to
5069 the language being compiled. For C (and C++), this activates
5070 optimizations based on the type of expressions. In particular, an
5071 object of one type is assumed never to reside at the same address as an
5072 object of a different type, unless the types are almost the same. For
5073 example, an @code{unsigned int} can alias an @code{int}, but not a
5074 @code{void*} or a @code{double}. A character type may alias any other
5077 Pay special attention to code like this:
5090 The practice of reading from a different union member than the one most
5091 recently written to (called ``type-punning'') is common. Even with
5092 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5093 is accessed through the union type. So, the code above will work as
5094 expected. However, this code might not:
5105 Every language that wishes to perform language-specific alias analysis
5106 should define a function that computes, given an @code{tree}
5107 node, an alias set for the node. Nodes in different alias sets are not
5108 allowed to alias. For an example, see the C front-end function
5109 @code{c_get_alias_set}.
5111 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5113 @item -falign-functions
5114 @itemx -falign-functions=@var{n}
5115 @opindex falign-functions
5116 Align the start of functions to the next power-of-two greater than
5117 @var{n}, skipping up to @var{n} bytes. For instance,
5118 @option{-falign-functions=32} aligns functions to the next 32-byte
5119 boundary, but @option{-falign-functions=24} would align to the next
5120 32-byte boundary only if this can be done by skipping 23 bytes or less.
5122 @option{-fno-align-functions} and @option{-falign-functions=1} are
5123 equivalent and mean that functions will not be aligned.
5125 Some assemblers only support this flag when @var{n} is a power of two;
5126 in that case, it is rounded up.
5128 If @var{n} is not specified or is zero, use a machine-dependent default.
5130 Enabled at levels @option{-O2}, @option{-O3}.
5132 @item -falign-labels
5133 @itemx -falign-labels=@var{n}
5134 @opindex falign-labels
5135 Align all branch targets to a power-of-two boundary, skipping up to
5136 @var{n} bytes like @option{-falign-functions}. This option can easily
5137 make code slower, because it must insert dummy operations for when the
5138 branch target is reached in the usual flow of the code.
5140 @option{-fno-align-labels} and @option{-falign-labels=1} are
5141 equivalent and mean that labels will not be aligned.
5143 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5144 are greater than this value, then their values are used instead.
5146 If @var{n} is not specified or is zero, use a machine-dependent default
5147 which is very likely to be @samp{1}, meaning no alignment.
5149 Enabled at levels @option{-O2}, @option{-O3}.
5152 @itemx -falign-loops=@var{n}
5153 @opindex falign-loops
5154 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5155 like @option{-falign-functions}. The hope is that the loop will be
5156 executed many times, which will make up for any execution of the dummy
5159 @option{-fno-align-loops} and @option{-falign-loops=1} are
5160 equivalent and mean that loops will not be aligned.
5162 If @var{n} is not specified or is zero, use a machine-dependent default.
5164 Enabled at levels @option{-O2}, @option{-O3}.
5167 @itemx -falign-jumps=@var{n}
5168 @opindex falign-jumps
5169 Align branch targets to a power-of-two boundary, for branch targets
5170 where the targets can only be reached by jumping, skipping up to @var{n}
5171 bytes like @option{-falign-functions}. In this case, no dummy operations
5174 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5175 equivalent and mean that loops will not be aligned.
5177 If @var{n} is not specified or is zero, use a machine-dependent default.
5179 Enabled at levels @option{-O2}, @option{-O3}.
5181 @item -funit-at-a-time
5182 @opindex funit-at-a-time
5183 Parse the whole compilation unit before starting to produce code.
5184 This allows some extra optimizations to take place but consumes
5185 more memory (in general). There are some compatibility issues
5186 with @emph{unit-at-at-time} mode:
5189 enabling @emph{unit-at-a-time} mode may change the order
5190 in which functions, variables, and top-level @code{asm} statements
5191 are emitted, and will likely break code relying on some particular
5192 ordering. The majority of such top-level @code{asm} statements,
5193 though, can be replaced by @code{section} attributes.
5196 @emph{unit-at-a-time} mode removes unreferenced static variables
5197 and functions are removed. This may result in undefined references
5198 when an @code{asm} statement refers directly to variables or functions
5199 that are otherwise unused. In that case either the variable/function
5200 shall be listed as an operand of the @code{asm} statement operand or,
5201 in the case of top-level @code{asm} statements the attribute @code{used}
5202 shall be used on the declaration.
5205 Static functions now can use non-standard passing conventions that
5206 may break @code{asm} statements calling functions directly. Again,
5207 attribute @code{used} will prevent this behavior.
5210 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5211 but this scheme may not be supported by future releases of GCC@.
5213 Enabled at levels @option{-O2}, @option{-O3}.
5217 Constructs webs as commonly used for register allocation purposes and assign
5218 each web individual pseudo register. This allows the register allocation pass
5219 to operate on pseudos directly, but also strengthens several other optimization
5220 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5221 however, make debugging impossible, since variables will no longer stay in a
5224 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os},
5225 on targets where the default format for debugging information supports
5228 @item -fno-cprop-registers
5229 @opindex fno-cprop-registers
5230 After register allocation and post-register allocation instruction splitting,
5231 we perform a copy-propagation pass to try to reduce scheduling dependencies
5232 and occasionally eliminate the copy.
5234 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5236 @item -fprofile-generate
5237 @opindex fprofile-generate
5239 Enable options usually used for instrumenting application to produce
5240 profile useful for later recompilation with profile feedback based
5241 optimization. You must use @option{-fprofile-generate} both when
5242 compiling and when linking your program.
5244 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5247 @opindex fprofile-use
5248 Enable profile feedback directed optimizations, and optimizations
5249 generally profitable only with profile feedback available.
5251 The following options are enabled: @code{-fbranch-probabilities},
5252 @code{-fvpt}, @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}.
5256 The following options control compiler behavior regarding floating
5257 point arithmetic. These options trade off between speed and
5258 correctness. All must be specifically enabled.
5262 @opindex ffloat-store
5263 Do not store floating point variables in registers, and inhibit other
5264 options that might change whether a floating point value is taken from a
5267 @cindex floating point precision
5268 This option prevents undesirable excess precision on machines such as
5269 the 68000 where the floating registers (of the 68881) keep more
5270 precision than a @code{double} is supposed to have. Similarly for the
5271 x86 architecture. For most programs, the excess precision does only
5272 good, but a few programs rely on the precise definition of IEEE floating
5273 point. Use @option{-ffloat-store} for such programs, after modifying
5274 them to store all pertinent intermediate computations into variables.
5278 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5279 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5280 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5281 and @option{fcx-limited-range}.
5283 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5285 This option should never be turned on by any @option{-O} option since
5286 it can result in incorrect output for programs which depend on
5287 an exact implementation of IEEE or ISO rules/specifications for
5290 @item -fno-math-errno
5291 @opindex fno-math-errno
5292 Do not set ERRNO after calling math functions that are executed
5293 with a single instruction, e.g., sqrt. A program that relies on
5294 IEEE exceptions for math error handling may want to use this flag
5295 for speed while maintaining IEEE arithmetic compatibility.
5297 This option should never be turned on by any @option{-O} option since
5298 it can result in incorrect output for programs which depend on
5299 an exact implementation of IEEE or ISO rules/specifications for
5302 The default is @option{-fmath-errno}.
5304 @item -funsafe-math-optimizations
5305 @opindex funsafe-math-optimizations
5306 Allow optimizations for floating-point arithmetic that (a) assume
5307 that arguments and results are valid and (b) may violate IEEE or
5308 ANSI standards. When used at link-time, it may include libraries
5309 or startup files that change the default FPU control word or other
5310 similar optimizations.
5312 This option should never be turned on by any @option{-O} option since
5313 it can result in incorrect output for programs which depend on
5314 an exact implementation of IEEE or ISO rules/specifications for
5317 The default is @option{-fno-unsafe-math-optimizations}.
5319 @item -ffinite-math-only
5320 @opindex ffinite-math-only
5321 Allow optimizations for floating-point arithmetic that assume
5322 that arguments and results are not NaNs or +-Infs.
5324 This option should never be turned on by any @option{-O} option since
5325 it can result in incorrect output for programs which depend on
5326 an exact implementation of IEEE or ISO rules/specifications.
5328 The default is @option{-fno-finite-math-only}.
5330 @item -fno-trapping-math
5331 @opindex fno-trapping-math
5332 Compile code assuming that floating-point operations cannot generate
5333 user-visible traps. These traps include division by zero, overflow,
5334 underflow, inexact result and invalid operation. This option implies
5335 @option{-fno-signaling-nans}. Setting this option may allow faster
5336 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5338 This option should never be turned on by any @option{-O} option since
5339 it can result in incorrect output for programs which depend on
5340 an exact implementation of IEEE or ISO rules/specifications for
5343 The default is @option{-ftrapping-math}.
5345 @item -frounding-math
5346 @opindex frounding-math
5347 Disable transformations and optimizations that assume default floating
5348 point rounding behavior. This is round-to-zero for all floating point
5349 to integer conversions, and round-to-nearest for all other arithmetic
5350 truncations. This option should be specified for programs that change
5351 the FP rounding mode dynamically, or that may be executed with a
5352 non-default rounding mode. This option disables constant folding of
5353 floating point expressions at compile-time (which may be affected by
5354 rounding mode) and arithmetic transformations that are unsafe in the
5355 presence of sign-dependent rounding modes.
5357 The default is @option{-fno-rounding-math}.
5359 This option is experimental and does not currently guarantee to
5360 disable all GCC optimizations that are affected by rounding mode.
5361 Future versions of GCC may provide finer control of this setting
5362 using C99's @code{FENV_ACCESS} pragma. This command line option
5363 will be used to specify the default state for @code{FENV_ACCESS}.
5365 @item -fsignaling-nans
5366 @opindex fsignaling-nans
5367 Compile code assuming that IEEE signaling NaNs may generate user-visible
5368 traps during floating-point operations. Setting this option disables
5369 optimizations that may change the number of exceptions visible with
5370 signaling NaNs. This option implies @option{-ftrapping-math}.
5372 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5375 The default is @option{-fno-signaling-nans}.
5377 This option is experimental and does not currently guarantee to
5378 disable all GCC optimizations that affect signaling NaN behavior.
5380 @item -fsingle-precision-constant
5381 @opindex fsingle-precision-constant
5382 Treat floating point constant as single precision constant instead of
5383 implicitly converting it to double precision constant.
5385 @item -fcx-limited-range
5386 @itemx -fno-cx-limited-range
5387 @opindex fcx-limited-range
5388 @opindex fno-cx-limited-range
5389 When enabled, this option states that a range reduction step is not
5390 needed when performing complex division. The default is
5391 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5393 This option controls the default setting of the ISO C99
5394 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5399 The following options control optimizations that may improve
5400 performance, but are not enabled by any @option{-O} options. This
5401 section includes experimental options that may produce broken code.
5404 @item -fbranch-probabilities
5405 @opindex fbranch-probabilities
5406 After running a program compiled with @option{-fprofile-arcs}
5407 (@pxref{Debugging Options,, Options for Debugging Your Program or
5408 @command{gcc}}), you can compile it a second time using
5409 @option{-fbranch-probabilities}, to improve optimizations based on
5410 the number of times each branch was taken. When the program
5411 compiled with @option{-fprofile-arcs} exits it saves arc execution
5412 counts to a file called @file{@var{sourcename}.gcda} for each source
5413 file The information in this data file is very dependent on the
5414 structure of the generated code, so you must use the same source code
5415 and the same optimization options for both compilations.
5417 With @option{-fbranch-probabilities}, GCC puts a
5418 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5419 These can be used to improve optimization. Currently, they are only
5420 used in one place: in @file{reorg.c}, instead of guessing which path a
5421 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5422 exactly determine which path is taken more often.
5424 @item -fprofile-values
5425 @opindex fprofile-values
5426 If combined with @option{-fprofile-arcs}, it adds code so that some
5427 data about values of expressions in the program is gathered.
5429 With @option{-fbranch-probabilities}, it reads back the data gathered
5430 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5431 notes to instructions for their later usage in optimizations.
5433 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5437 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5438 a code to gather information about values of expressions.
5440 With @option{-fbranch-probabilities}, it reads back the data gathered
5441 and actually performs the optimizations based on them.
5442 Currently the optimizations include specialization of division operation
5443 using the knowledge about the value of the denominator.
5445 @item -fspeculative-prefetching
5446 @opindex fspeculative-prefetching
5447 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5448 a code to gather information about addresses of memory references in the
5451 With @option{-fbranch-probabilities}, it reads back the data gathered
5452 and issues prefetch instructions according to them. In addition to the opportunities
5453 noticed by @option{-fprefetch-loop-arrays}, it also notices more complicated
5454 memory access patterns---for example accesses to the data stored in linked
5455 list whose elements are usually allocated sequentially.
5457 In order to prevent issuing double prefetches, usage of
5458 @option{-fspeculative-prefetching} implies @option{-fno-prefetch-loop-arrays}.
5460 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5462 @item -frename-registers
5463 @opindex frename-registers
5464 Attempt to avoid false dependencies in scheduled code by making use
5465 of registers left over after register allocation. This optimization
5466 will most benefit processors with lots of registers. Depending on the
5467 debug information format adopted by the target, however, it can
5468 make debugging impossible, since variables will no longer stay in
5469 a ``home register''.
5471 Not enabled by default at any level because it has known bugs.
5475 Perform tail duplication to enlarge superblock size. This transformation
5476 simplifies the control flow of the function allowing other optimizations to do
5479 Enabled with @option{-fprofile-use}.
5481 @item -funroll-loops
5482 @opindex funroll-loops
5483 Unroll loops whose number of iterations can be determined at compile time or
5484 upon entry to the loop. @option{-funroll-loops} implies
5485 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
5486 (i.e.@: complete removal of loops with small constant number of iterations).
5487 This option makes code larger, and may or may not make it run faster.
5489 Enabled with @option{-fprofile-use}.
5491 @item -funroll-all-loops
5492 @opindex funroll-all-loops
5493 Unroll all loops, even if their number of iterations is uncertain when
5494 the loop is entered. This usually makes programs run more slowly.
5495 @option{-funroll-all-loops} implies the same options as
5496 @option{-funroll-loops}.
5499 @opindex fpeel-loops
5500 Peels the loops for that there is enough information that they do not
5501 roll much (from profile feedback). It also turns on complete loop peeling
5502 (i.e.@: complete removal of loops with small constant number of iterations).
5504 Enabled with @option{-fprofile-use}.
5506 @item -fmove-loop-invariants
5507 @opindex fmove-loop-invariants
5508 Enables the loop invariant motion pass in the new loop optimizer. Enabled
5509 at level @option{-O1}
5511 @item -funswitch-loops
5512 @opindex funswitch-loops
5513 Move branches with loop invariant conditions out of the loop, with duplicates
5514 of the loop on both branches (modified according to result of the condition).
5516 @item -fprefetch-loop-arrays
5517 @opindex fprefetch-loop-arrays
5518 If supported by the target machine, generate instructions to prefetch
5519 memory to improve the performance of loops that access large arrays.
5521 Disabled at level @option{-Os}.
5523 @item -ffunction-sections
5524 @itemx -fdata-sections
5525 @opindex ffunction-sections
5526 @opindex fdata-sections
5527 Place each function or data item into its own section in the output
5528 file if the target supports arbitrary sections. The name of the
5529 function or the name of the data item determines the section's name
5532 Use these options on systems where the linker can perform optimizations
5533 to improve locality of reference in the instruction space. Most systems
5534 using the ELF object format and SPARC processors running Solaris 2 have
5535 linkers with such optimizations. AIX may have these optimizations in
5538 Only use these options when there are significant benefits from doing
5539 so. When you specify these options, the assembler and linker will
5540 create larger object and executable files and will also be slower.
5541 You will not be able to use @code{gprof} on all systems if you
5542 specify this option and you may have problems with debugging if
5543 you specify both this option and @option{-g}.
5545 @item -fbranch-target-load-optimize
5546 @opindex fbranch-target-load-optimize
5547 Perform branch target register load optimization before prologue / epilogue
5549 The use of target registers can typically be exposed only during reload,
5550 thus hoisting loads out of loops and doing inter-block scheduling needs
5551 a separate optimization pass.
5553 @item -fbranch-target-load-optimize2
5554 @opindex fbranch-target-load-optimize2
5555 Perform branch target register load optimization after prologue / epilogue
5558 @item -fbtr-bb-exclusive
5559 @opindex fbtr-bb-exclusive
5560 When performing branch target register load optimization, don't reuse
5561 branch target registers in within any basic block.
5563 @item --param @var{name}=@var{value}
5565 In some places, GCC uses various constants to control the amount of
5566 optimization that is done. For example, GCC will not inline functions
5567 that contain more that a certain number of instructions. You can
5568 control some of these constants on the command-line using the
5569 @option{--param} option.
5571 The names of specific parameters, and the meaning of the values, are
5572 tied to the internals of the compiler, and are subject to change
5573 without notice in future releases.
5575 In each case, the @var{value} is an integer. The allowable choices for
5576 @var{name} are given in the following table:
5579 @item salias-max-implicit-fields
5580 The maximum number of fields in a variable without direct
5581 structure accesses for which structure aliasing will consider trying
5582 to track each field. The default is 5
5584 @item sra-max-structure-size
5585 The maximum structure size, in bytes, at which the scalar replacement
5586 of aggregates (SRA) optimization will perform block copies. The
5587 default value, 0, implies that GCC will select the most appropriate
5590 @item sra-field-structure-ratio
5591 The threshold ratio (as a percentage) between instantiated fields and
5592 the complete structure size. We say that if the ratio of the number
5593 of bytes in instantiated fields to the number of bytes in the complete
5594 structure exceeds this parameter, then block copies are not used. The
5597 @item max-crossjump-edges
5598 The maximum number of incoming edges to consider for crossjumping.
5599 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5600 the number of edges incoming to each block. Increasing values mean
5601 more aggressive optimization, making the compile time increase with
5602 probably small improvement in executable size.
5604 @item min-crossjump-insns
5605 The minimum number of instructions which must be matched at the end
5606 of two blocks before crossjumping will be performed on them. This
5607 value is ignored in the case where all instructions in the block being
5608 crossjumped from are matched. The default value is 5.
5610 @item max-goto-duplication-insns
5611 The maximum number of instructions to duplicate to a block that jumps
5612 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5613 passes, GCC factors computed gotos early in the compilation process,
5614 and unfactors them as late as possible. Only computed jumps at the
5615 end of a basic blocks with no more than max-goto-duplication-insns are
5616 unfactored. The default value is 8.
5618 @item max-delay-slot-insn-search
5619 The maximum number of instructions to consider when looking for an
5620 instruction to fill a delay slot. If more than this arbitrary number of
5621 instructions is searched, the time savings from filling the delay slot
5622 will be minimal so stop searching. Increasing values mean more
5623 aggressive optimization, making the compile time increase with probably
5624 small improvement in executable run time.
5626 @item max-delay-slot-live-search
5627 When trying to fill delay slots, the maximum number of instructions to
5628 consider when searching for a block with valid live register
5629 information. Increasing this arbitrarily chosen value means more
5630 aggressive optimization, increasing the compile time. This parameter
5631 should be removed when the delay slot code is rewritten to maintain the
5634 @item max-gcse-memory
5635 The approximate maximum amount of memory that will be allocated in
5636 order to perform the global common subexpression elimination
5637 optimization. If more memory than specified is required, the
5638 optimization will not be done.
5640 @item max-gcse-passes
5641 The maximum number of passes of GCSE to run. The default is 1.
5643 @item max-pending-list-length
5644 The maximum number of pending dependencies scheduling will allow
5645 before flushing the current state and starting over. Large functions
5646 with few branches or calls can create excessively large lists which
5647 needlessly consume memory and resources.
5649 @item max-inline-insns-single
5650 Several parameters control the tree inliner used in gcc.
5651 This number sets the maximum number of instructions (counted in GCC's
5652 internal representation) in a single function that the tree inliner
5653 will consider for inlining. This only affects functions declared
5654 inline and methods implemented in a class declaration (C++).
5655 The default value is 450.
5657 @item max-inline-insns-auto
5658 When you use @option{-finline-functions} (included in @option{-O3}),
5659 a lot of functions that would otherwise not be considered for inlining
5660 by the compiler will be investigated. To those functions, a different
5661 (more restrictive) limit compared to functions declared inline can
5663 The default value is 90.
5665 @item large-function-insns
5666 The limit specifying really large functions. For functions larger than this
5667 limit after inlining inlining is constrained by
5668 @option{--param large-function-growth}. This parameter is useful primarily
5669 to avoid extreme compilation time caused by non-linear algorithms used by the
5671 This parameter is ignored when @option{-funit-at-a-time} is not used.
5672 The default value is 2700.
5674 @item large-function-growth
5675 Specifies maximal growth of large function caused by inlining in percents.
5676 This parameter is ignored when @option{-funit-at-a-time} is not used.
5677 The default value is 100 which limits large function growth to 2.0 times
5680 @item inline-unit-growth
5681 Specifies maximal overall growth of the compilation unit caused by inlining.
5682 This parameter is ignored when @option{-funit-at-a-time} is not used.
5683 The default value is 50 which limits unit growth to 1.5 times the original
5686 @item max-inline-insns-recursive
5687 @itemx max-inline-insns-recursive-auto
5688 Specifies maximum number of instructions out-of-line copy of self recursive inline
5689 function can grow into by performing recursive inlining.
5691 For functions declared inline @option{--param max-inline-insns-recursive} is
5692 taken into acount. For function not declared inline, recursive inlining
5693 happens only when @option{-finline-functions} (included in @option{-O3}) is
5694 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5695 default value is 450.
5697 @item max-inline-recursive-depth
5698 @itemx max-inline-recursive-depth-auto
5699 Specifies maximum recursion depth used by the recursive inlining.
5701 For functions declared inline @option{--param max-inline-recursive-depth} is
5702 taken into acount. For function not declared inline, recursive inlining
5703 happens only when @option{-finline-functions} (included in @option{-O3}) is
5704 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5705 default value is 450.
5707 @item inline-call-cost
5708 Specify cost of call instruction relative to simple arithmetics operations
5709 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
5710 functions and at the same time increases size of leaf function that is believed to
5711 reduce function size by being inlined. In effect it increases amount of
5712 inlining for code having large abstraction penalty (many functions that just
5713 pass the arguments to other functions) and decrease inlining for code with low
5714 abstraction penalty. The default value is 16.
5716 @item max-unrolled-insns
5717 The maximum number of instructions that a loop should have if that loop
5718 is unrolled, and if the loop is unrolled, it determines how many times
5719 the loop code is unrolled.
5721 @item max-average-unrolled-insns
5722 The maximum number of instructions biased by probabilities of their execution
5723 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5724 it determines how many times the loop code is unrolled.
5726 @item max-unroll-times
5727 The maximum number of unrollings of a single loop.
5729 @item max-peeled-insns
5730 The maximum number of instructions that a loop should have if that loop
5731 is peeled, and if the loop is peeled, it determines how many times
5732 the loop code is peeled.
5734 @item max-peel-times
5735 The maximum number of peelings of a single loop.
5737 @item max-completely-peeled-insns
5738 The maximum number of insns of a completely peeled loop.
5740 @item max-completely-peel-times
5741 The maximum number of iterations of a loop to be suitable for complete peeling.
5743 @item max-unswitch-insns
5744 The maximum number of insns of an unswitched loop.
5746 @item max-unswitch-level
5747 The maximum number of branches unswitched in a single loop.
5750 The minimum cost of an expensive expression in the loop invariant motion.
5752 @item iv-consider-all-candidates-bound
5753 Bound on number of candidates for induction variables below that
5754 all candidates are considered for each use in induction variable
5755 optimizations. Only the most relevant candidates are considered
5756 if there are more candidates, to avoid quadratic time complexity.
5758 @item iv-max-considered-uses
5759 The induction variable optimizations give up on loops that contain more
5760 induction variable uses.
5762 @item iv-always-prune-cand-set-bound
5763 If number of candidates in the set is smaller than this value,
5764 we always try to remove unnecessary ivs from the set during its
5765 optimization when a new iv is added to the set.
5767 @item scev-max-expr-size
5768 Bound on size of expressions used in the scalar evolutions analyzer.
5769 Large expressions slow the analyzer.
5771 @item max-iterations-to-track
5773 The maximum number of iterations of a loop the brute force algorithm
5774 for analysis of # of iterations of the loop tries to evaluate.
5776 @item hot-bb-count-fraction
5777 Select fraction of the maximal count of repetitions of basic block in program
5778 given basic block needs to have to be considered hot.
5780 @item hot-bb-frequency-fraction
5781 Select fraction of the maximal frequency of executions of basic block in
5782 function given basic block needs to have to be considered hot
5784 @item tracer-dynamic-coverage
5785 @itemx tracer-dynamic-coverage-feedback
5787 This value is used to limit superblock formation once the given percentage of
5788 executed instructions is covered. This limits unnecessary code size
5791 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5792 feedback is available. The real profiles (as opposed to statically estimated
5793 ones) are much less balanced allowing the threshold to be larger value.
5795 @item tracer-max-code-growth
5796 Stop tail duplication once code growth has reached given percentage. This is
5797 rather hokey argument, as most of the duplicates will be eliminated later in
5798 cross jumping, so it may be set to much higher values than is the desired code
5801 @item tracer-min-branch-ratio
5803 Stop reverse growth when the reverse probability of best edge is less than this
5804 threshold (in percent).
5806 @item tracer-min-branch-ratio
5807 @itemx tracer-min-branch-ratio-feedback
5809 Stop forward growth if the best edge do have probability lower than this
5812 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
5813 compilation for profile feedback and one for compilation without. The value
5814 for compilation with profile feedback needs to be more conservative (higher) in
5815 order to make tracer effective.
5817 @item max-cse-path-length
5819 Maximum number of basic blocks on path that cse considers. The default is 10.
5821 @item global-var-threshold
5823 Counts the number of function calls (@var{n}) and the number of
5824 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
5825 single artificial variable will be created to represent all the
5826 call-clobbered variables at function call sites. This artificial
5827 variable will then be made to alias every call-clobbered variable.
5828 (done as @code{int * size_t} on the host machine; beware overflow).
5830 @item max-aliased-vops
5832 Maximum number of virtual operands allowed to represent aliases
5833 before triggering the alias grouping heuristic. Alias grouping
5834 reduces compile times and memory consumption needed for aliasing at
5835 the expense of precision loss in alias information.
5837 @item ggc-min-expand
5839 GCC uses a garbage collector to manage its own memory allocation. This
5840 parameter specifies the minimum percentage by which the garbage
5841 collector's heap should be allowed to expand between collections.
5842 Tuning this may improve compilation speed; it has no effect on code
5845 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
5846 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
5847 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
5848 GCC is not able to calculate RAM on a particular platform, the lower
5849 bound of 30% is used. Setting this parameter and
5850 @option{ggc-min-heapsize} to zero causes a full collection to occur at
5851 every opportunity. This is extremely slow, but can be useful for
5854 @item ggc-min-heapsize
5856 Minimum size of the garbage collector's heap before it begins bothering
5857 to collect garbage. The first collection occurs after the heap expands
5858 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
5859 tuning this may improve compilation speed, and has no effect on code
5862 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
5863 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
5864 with a lower bound of 4096 (four megabytes) and an upper bound of
5865 131072 (128 megabytes). If GCC is not able to calculate RAM on a
5866 particular platform, the lower bound is used. Setting this parameter
5867 very large effectively disables garbage collection. Setting this
5868 parameter and @option{ggc-min-expand} to zero causes a full collection
5869 to occur at every opportunity.
5871 @item max-reload-search-insns
5872 The maximum number of instruction reload should look backward for equivalent
5873 register. Increasing values mean more aggressive optimization, making the
5874 compile time increase with probably slightly better performance. The default
5877 @item max-cselib-memory-location
5878 The maximum number of memory locations cselib should take into acount.
5879 Increasing values mean more aggressive optimization, making the compile time
5880 increase with probably slightly better performance. The default value is 500.
5882 @item reorder-blocks-duplicate
5883 @itemx reorder-blocks-duplicate-feedback
5885 Used by basic block reordering pass to decide whether to use unconditional
5886 branch or duplicate the code on its destination. Code is duplicated when its
5887 estimated size is smaller than this value multiplied by the estimated size of
5888 unconditional jump in the hot spots of the program.
5890 The @option{reorder-block-duplicate-feedback} is used only when profile
5891 feedback is available and may be set to higher values than
5892 @option{reorder-block-duplicate} since information about the hot spots is more
5895 @item max-sched-region-blocks
5896 The maximum number of blocks in a region to be considered for
5897 interblock scheduling. The default value is 10.
5899 @item max-sched-region-insns
5900 The maximum number of insns in a region to be considered for
5901 interblock scheduling. The default value is 100.
5903 @item max-last-value-rtl
5905 The maximum size measured as number of RTLs that can be recorded in an expression
5906 in combiner for a pseudo register as last known value of that register. The default
5909 @item integer-share-limit
5910 Small integer constants can use a shared data structure, reducing the
5911 compiler's memory usage and increasing its speed. This sets the maximum
5912 value of a shared integer constant's. The default value is 256.
5914 @item min-virtual-mappings
5915 Specifies the minimum number of virtual mappings in the incremental
5916 SSA updater that should be registered to trigger the virtual mappings
5917 heuristic defined by virtual-mappings-ratio. The default value is
5920 @item virtual-mappings-ratio
5921 If the number of virtual mappings is virtual-mappings-ratio bigger
5922 than the number of virtual symbols to be updated, then the incremental
5923 SSA updater switches to a full update for those symbols. The default
5929 @node Preprocessor Options
5930 @section Options Controlling the Preprocessor
5931 @cindex preprocessor options
5932 @cindex options, preprocessor
5934 These options control the C preprocessor, which is run on each C source
5935 file before actual compilation.
5937 If you use the @option{-E} option, nothing is done except preprocessing.
5938 Some of these options make sense only together with @option{-E} because
5939 they cause the preprocessor output to be unsuitable for actual
5944 You can use @option{-Wp,@var{option}} to bypass the compiler driver
5945 and pass @var{option} directly through to the preprocessor. If
5946 @var{option} contains commas, it is split into multiple options at the
5947 commas. However, many options are modified, translated or interpreted
5948 by the compiler driver before being passed to the preprocessor, and
5949 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
5950 interface is undocumented and subject to change, so whenever possible
5951 you should avoid using @option{-Wp} and let the driver handle the
5954 @item -Xpreprocessor @var{option}
5955 @opindex preprocessor
5956 Pass @var{option} as an option to the preprocessor. You can use this to
5957 supply system-specific preprocessor options which GCC does not know how to
5960 If you want to pass an option that takes an argument, you must use
5961 @option{-Xpreprocessor} twice, once for the option and once for the argument.
5964 @include cppopts.texi
5966 @node Assembler Options
5967 @section Passing Options to the Assembler
5969 @c prevent bad page break with this line
5970 You can pass options to the assembler.
5973 @item -Wa,@var{option}
5975 Pass @var{option} as an option to the assembler. If @var{option}
5976 contains commas, it is split into multiple options at the commas.
5978 @item -Xassembler @var{option}
5980 Pass @var{option} as an option to the assembler. You can use this to
5981 supply system-specific assembler options which GCC does not know how to
5984 If you want to pass an option that takes an argument, you must use
5985 @option{-Xassembler} twice, once for the option and once for the argument.
5990 @section Options for Linking
5991 @cindex link options
5992 @cindex options, linking
5994 These options come into play when the compiler links object files into
5995 an executable output file. They are meaningless if the compiler is
5996 not doing a link step.
6000 @item @var{object-file-name}
6001 A file name that does not end in a special recognized suffix is
6002 considered to name an object file or library. (Object files are
6003 distinguished from libraries by the linker according to the file
6004 contents.) If linking is done, these object files are used as input
6013 If any of these options is used, then the linker is not run, and
6014 object file names should not be used as arguments. @xref{Overall
6018 @item -l@var{library}
6019 @itemx -l @var{library}
6021 Search the library named @var{library} when linking. (The second
6022 alternative with the library as a separate argument is only for
6023 POSIX compliance and is not recommended.)
6025 It makes a difference where in the command you write this option; the
6026 linker searches and processes libraries and object files in the order they
6027 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6028 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6029 to functions in @samp{z}, those functions may not be loaded.
6031 The linker searches a standard list of directories for the library,
6032 which is actually a file named @file{lib@var{library}.a}. The linker
6033 then uses this file as if it had been specified precisely by name.
6035 The directories searched include several standard system directories
6036 plus any that you specify with @option{-L}.
6038 Normally the files found this way are library files---archive files
6039 whose members are object files. The linker handles an archive file by
6040 scanning through it for members which define symbols that have so far
6041 been referenced but not defined. But if the file that is found is an
6042 ordinary object file, it is linked in the usual fashion. The only
6043 difference between using an @option{-l} option and specifying a file name
6044 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6045 and searches several directories.
6049 You need this special case of the @option{-l} option in order to
6050 link an Objective-C or Objective-C++ program.
6053 @opindex nostartfiles
6054 Do not use the standard system startup files when linking.
6055 The standard system libraries are used normally, unless @option{-nostdlib}
6056 or @option{-nodefaultlibs} is used.
6058 @item -nodefaultlibs
6059 @opindex nodefaultlibs
6060 Do not use the standard system libraries when linking.
6061 Only the libraries you specify will be passed to the linker.
6062 The standard startup files are used normally, unless @option{-nostartfiles}
6063 is used. The compiler may generate calls to @code{memcmp},
6064 @code{memset}, @code{memcpy} and @code{memmove}.
6065 These entries are usually resolved by entries in
6066 libc. These entry points should be supplied through some other
6067 mechanism when this option is specified.
6071 Do not use the standard system startup files or libraries when linking.
6072 No startup files and only the libraries you specify will be passed to
6073 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6074 @code{memcpy} and @code{memmove}.
6075 These entries are usually resolved by entries in
6076 libc. These entry points should be supplied through some other
6077 mechanism when this option is specified.
6079 @cindex @option{-lgcc}, use with @option{-nostdlib}
6080 @cindex @option{-nostdlib} and unresolved references
6081 @cindex unresolved references and @option{-nostdlib}
6082 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6083 @cindex @option{-nodefaultlibs} and unresolved references
6084 @cindex unresolved references and @option{-nodefaultlibs}
6085 One of the standard libraries bypassed by @option{-nostdlib} and
6086 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6087 that GCC uses to overcome shortcomings of particular machines, or special
6088 needs for some languages.
6089 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6090 Collection (GCC) Internals},
6091 for more discussion of @file{libgcc.a}.)
6092 In most cases, you need @file{libgcc.a} even when you want to avoid
6093 other standard libraries. In other words, when you specify @option{-nostdlib}
6094 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6095 This ensures that you have no unresolved references to internal GCC
6096 library subroutines. (For example, @samp{__main}, used to ensure C++
6097 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6098 GNU Compiler Collection (GCC) Internals}.)
6102 Produce a position independent executable on targets which support it.
6103 For predictable results, you must also specify the same set of options
6104 that were used to generate code (@option{-fpie}, @option{-fPIE},
6105 or model suboptions) when you specify this option.
6109 Remove all symbol table and relocation information from the executable.
6113 On systems that support dynamic linking, this prevents linking with the shared
6114 libraries. On other systems, this option has no effect.
6118 Produce a shared object which can then be linked with other objects to
6119 form an executable. Not all systems support this option. For predictable
6120 results, you must also specify the same set of options that were used to
6121 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6122 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6123 needs to build supplementary stub code for constructors to work. On
6124 multi-libbed systems, @samp{gcc -shared} must select the correct support
6125 libraries to link against. Failing to supply the correct flags may lead
6126 to subtle defects. Supplying them in cases where they are not necessary
6129 @item -shared-libgcc
6130 @itemx -static-libgcc
6131 @opindex shared-libgcc
6132 @opindex static-libgcc
6133 On systems that provide @file{libgcc} as a shared library, these options
6134 force the use of either the shared or static version respectively.
6135 If no shared version of @file{libgcc} was built when the compiler was
6136 configured, these options have no effect.
6138 There are several situations in which an application should use the
6139 shared @file{libgcc} instead of the static version. The most common
6140 of these is when the application wishes to throw and catch exceptions
6141 across different shared libraries. In that case, each of the libraries
6142 as well as the application itself should use the shared @file{libgcc}.
6144 Therefore, the G++ and GCJ drivers automatically add
6145 @option{-shared-libgcc} whenever you build a shared library or a main
6146 executable, because C++ and Java programs typically use exceptions, so
6147 this is the right thing to do.
6149 If, instead, you use the GCC driver to create shared libraries, you may
6150 find that they will not always be linked with the shared @file{libgcc}.
6151 If GCC finds, at its configuration time, that you have a non-GNU linker
6152 or a GNU linker that does not support option @option{--eh-frame-hdr},
6153 it will link the shared version of @file{libgcc} into shared libraries
6154 by default. Otherwise, it will take advantage of the linker and optimize
6155 away the linking with the shared version of @file{libgcc}, linking with
6156 the static version of libgcc by default. This allows exceptions to
6157 propagate through such shared libraries, without incurring relocation
6158 costs at library load time.
6160 However, if a library or main executable is supposed to throw or catch
6161 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6162 for the languages used in the program, or using the option
6163 @option{-shared-libgcc}, such that it is linked with the shared
6168 Bind references to global symbols when building a shared object. Warn
6169 about any unresolved references (unless overridden by the link editor
6170 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6173 @item -Xlinker @var{option}
6175 Pass @var{option} as an option to the linker. You can use this to
6176 supply system-specific linker options which GCC does not know how to
6179 If you want to pass an option that takes an argument, you must use
6180 @option{-Xlinker} twice, once for the option and once for the argument.
6181 For example, to pass @option{-assert definitions}, you must write
6182 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6183 @option{-Xlinker "-assert definitions"}, because this passes the entire
6184 string as a single argument, which is not what the linker expects.
6186 @item -Wl,@var{option}
6188 Pass @var{option} as an option to the linker. If @var{option} contains
6189 commas, it is split into multiple options at the commas.
6191 @item -u @var{symbol}
6193 Pretend the symbol @var{symbol} is undefined, to force linking of
6194 library modules to define it. You can use @option{-u} multiple times with
6195 different symbols to force loading of additional library modules.
6198 @node Directory Options
6199 @section Options for Directory Search
6200 @cindex directory options
6201 @cindex options, directory search
6204 These options specify directories to search for header files, for
6205 libraries and for parts of the compiler:
6210 Add the directory @var{dir} to the head of the list of directories to be
6211 searched for header files. This can be used to override a system header
6212 file, substituting your own version, since these directories are
6213 searched before the system header file directories. However, you should
6214 not use this option to add directories that contain vendor-supplied
6215 system header files (use @option{-isystem} for that). If you use more than
6216 one @option{-I} option, the directories are scanned in left-to-right
6217 order; the standard system directories come after.
6219 If a standard system include directory, or a directory specified with
6220 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6221 option will be ignored. The directory will still be searched but as a
6222 system directory at its normal position in the system include chain.
6223 This is to ensure that GCC's procedure to fix buggy system headers and
6224 the ordering for the include_next directive are not inadvertently changed.
6225 If you really need to change the search order for system directories,
6226 use the @option{-nostdinc} and/or @option{-isystem} options.
6228 @item -iquote@var{dir}
6230 Add the directory @var{dir} to the head of the list of directories to
6231 be searched for header files only for the case of @samp{#include
6232 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6233 otherwise just like @option{-I}.
6237 Add directory @var{dir} to the list of directories to be searched
6240 @item -B@var{prefix}
6242 This option specifies where to find the executables, libraries,
6243 include files, and data files of the compiler itself.
6245 The compiler driver program runs one or more of the subprograms
6246 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6247 @var{prefix} as a prefix for each program it tries to run, both with and
6248 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6250 For each subprogram to be run, the compiler driver first tries the
6251 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6252 was not specified, the driver tries two standard prefixes, which are
6253 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6254 those results in a file name that is found, the unmodified program
6255 name is searched for using the directories specified in your
6256 @env{PATH} environment variable.
6258 The compiler will check to see if the path provided by the @option{-B}
6259 refers to a directory, and if necessary it will add a directory
6260 separator character at the end of the path.
6262 @option{-B} prefixes that effectively specify directory names also apply
6263 to libraries in the linker, because the compiler translates these
6264 options into @option{-L} options for the linker. They also apply to
6265 includes files in the preprocessor, because the compiler translates these
6266 options into @option{-isystem} options for the preprocessor. In this case,
6267 the compiler appends @samp{include} to the prefix.
6269 The run-time support file @file{libgcc.a} can also be searched for using
6270 the @option{-B} prefix, if needed. If it is not found there, the two
6271 standard prefixes above are tried, and that is all. The file is left
6272 out of the link if it is not found by those means.
6274 Another way to specify a prefix much like the @option{-B} prefix is to use
6275 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6278 As a special kludge, if the path provided by @option{-B} is
6279 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6280 9, then it will be replaced by @file{[dir/]include}. This is to help
6281 with boot-strapping the compiler.
6283 @item -specs=@var{file}
6285 Process @var{file} after the compiler reads in the standard @file{specs}
6286 file, in order to override the defaults that the @file{gcc} driver
6287 program uses when determining what switches to pass to @file{cc1},
6288 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6289 @option{-specs=@var{file}} can be specified on the command line, and they
6290 are processed in order, from left to right.
6294 This option has been deprecated. Please use @option{-iquote} instead for
6295 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6296 Any directories you specify with @option{-I} options before the @option{-I-}
6297 option are searched only for the case of @samp{#include "@var{file}"};
6298 they are not searched for @samp{#include <@var{file}>}.
6300 If additional directories are specified with @option{-I} options after
6301 the @option{-I-}, these directories are searched for all @samp{#include}
6302 directives. (Ordinarily @emph{all} @option{-I} directories are used
6305 In addition, the @option{-I-} option inhibits the use of the current
6306 directory (where the current input file came from) as the first search
6307 directory for @samp{#include "@var{file}"}. There is no way to
6308 override this effect of @option{-I-}. With @option{-I.} you can specify
6309 searching the directory which was current when the compiler was
6310 invoked. That is not exactly the same as what the preprocessor does
6311 by default, but it is often satisfactory.
6313 @option{-I-} does not inhibit the use of the standard system directories
6314 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6321 @section Specifying subprocesses and the switches to pass to them
6324 @command{gcc} is a driver program. It performs its job by invoking a
6325 sequence of other programs to do the work of compiling, assembling and
6326 linking. GCC interprets its command-line parameters and uses these to
6327 deduce which programs it should invoke, and which command-line options
6328 it ought to place on their command lines. This behavior is controlled
6329 by @dfn{spec strings}. In most cases there is one spec string for each
6330 program that GCC can invoke, but a few programs have multiple spec
6331 strings to control their behavior. The spec strings built into GCC can
6332 be overridden by using the @option{-specs=} command-line switch to specify
6335 @dfn{Spec files} are plaintext files that are used to construct spec
6336 strings. They consist of a sequence of directives separated by blank
6337 lines. The type of directive is determined by the first non-whitespace
6338 character on the line and it can be one of the following:
6341 @item %@var{command}
6342 Issues a @var{command} to the spec file processor. The commands that can
6346 @item %include <@var{file}>
6348 Search for @var{file} and insert its text at the current point in the
6351 @item %include_noerr <@var{file}>
6352 @cindex %include_noerr
6353 Just like @samp{%include}, but do not generate an error message if the include
6354 file cannot be found.
6356 @item %rename @var{old_name} @var{new_name}
6358 Rename the spec string @var{old_name} to @var{new_name}.
6362 @item *[@var{spec_name}]:
6363 This tells the compiler to create, override or delete the named spec
6364 string. All lines after this directive up to the next directive or
6365 blank line are considered to be the text for the spec string. If this
6366 results in an empty string then the spec will be deleted. (Or, if the
6367 spec did not exist, then nothing will happened.) Otherwise, if the spec
6368 does not currently exist a new spec will be created. If the spec does
6369 exist then its contents will be overridden by the text of this
6370 directive, unless the first character of that text is the @samp{+}
6371 character, in which case the text will be appended to the spec.
6373 @item [@var{suffix}]:
6374 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6375 and up to the next directive or blank line are considered to make up the
6376 spec string for the indicated suffix. When the compiler encounters an
6377 input file with the named suffix, it will processes the spec string in
6378 order to work out how to compile that file. For example:
6385 This says that any input file whose name ends in @samp{.ZZ} should be
6386 passed to the program @samp{z-compile}, which should be invoked with the
6387 command-line switch @option{-input} and with the result of performing the
6388 @samp{%i} substitution. (See below.)
6390 As an alternative to providing a spec string, the text that follows a
6391 suffix directive can be one of the following:
6394 @item @@@var{language}
6395 This says that the suffix is an alias for a known @var{language}. This is
6396 similar to using the @option{-x} command-line switch to GCC to specify a
6397 language explicitly. For example:
6404 Says that .ZZ files are, in fact, C++ source files.
6407 This causes an error messages saying:
6410 @var{name} compiler not installed on this system.
6414 GCC already has an extensive list of suffixes built into it.
6415 This directive will add an entry to the end of the list of suffixes, but
6416 since the list is searched from the end backwards, it is effectively
6417 possible to override earlier entries using this technique.
6421 GCC has the following spec strings built into it. Spec files can
6422 override these strings or create their own. Note that individual
6423 targets can also add their own spec strings to this list.
6426 asm Options to pass to the assembler
6427 asm_final Options to pass to the assembler post-processor
6428 cpp Options to pass to the C preprocessor
6429 cc1 Options to pass to the C compiler
6430 cc1plus Options to pass to the C++ compiler
6431 endfile Object files to include at the end of the link
6432 link Options to pass to the linker
6433 lib Libraries to include on the command line to the linker
6434 libgcc Decides which GCC support library to pass to the linker
6435 linker Sets the name of the linker
6436 predefines Defines to be passed to the C preprocessor
6437 signed_char Defines to pass to CPP to say whether @code{char} is signed
6439 startfile Object files to include at the start of the link
6442 Here is a small example of a spec file:
6448 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6451 This example renames the spec called @samp{lib} to @samp{old_lib} and
6452 then overrides the previous definition of @samp{lib} with a new one.
6453 The new definition adds in some extra command-line options before
6454 including the text of the old definition.
6456 @dfn{Spec strings} are a list of command-line options to be passed to their
6457 corresponding program. In addition, the spec strings can contain
6458 @samp{%}-prefixed sequences to substitute variable text or to
6459 conditionally insert text into the command line. Using these constructs
6460 it is possible to generate quite complex command lines.
6462 Here is a table of all defined @samp{%}-sequences for spec
6463 strings. Note that spaces are not generated automatically around the
6464 results of expanding these sequences. Therefore you can concatenate them
6465 together or combine them with constant text in a single argument.
6469 Substitute one @samp{%} into the program name or argument.
6472 Substitute the name of the input file being processed.
6475 Substitute the basename of the input file being processed.
6476 This is the substring up to (and not including) the last period
6477 and not including the directory.
6480 This is the same as @samp{%b}, but include the file suffix (text after
6484 Marks the argument containing or following the @samp{%d} as a
6485 temporary file name, so that that file will be deleted if GCC exits
6486 successfully. Unlike @samp{%g}, this contributes no text to the
6489 @item %g@var{suffix}
6490 Substitute a file name that has suffix @var{suffix} and is chosen
6491 once per compilation, and mark the argument in the same way as
6492 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6493 name is now chosen in a way that is hard to predict even when previously
6494 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6495 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6496 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6497 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6498 was simply substituted with a file name chosen once per compilation,
6499 without regard to any appended suffix (which was therefore treated
6500 just like ordinary text), making such attacks more likely to succeed.
6502 @item %u@var{suffix}
6503 Like @samp{%g}, but generates a new temporary file name even if
6504 @samp{%u@var{suffix}} was already seen.
6506 @item %U@var{suffix}
6507 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6508 new one if there is no such last file name. In the absence of any
6509 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6510 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6511 would involve the generation of two distinct file names, one
6512 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6513 simply substituted with a file name chosen for the previous @samp{%u},
6514 without regard to any appended suffix.
6516 @item %j@var{suffix}
6517 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6518 writable, and if save-temps is off; otherwise, substitute the name
6519 of a temporary file, just like @samp{%u}. This temporary file is not
6520 meant for communication between processes, but rather as a junk
6523 @item %|@var{suffix}
6524 @itemx %m@var{suffix}
6525 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6526 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6527 all. These are the two most common ways to instruct a program that it
6528 should read from standard input or write to standard output. If you
6529 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6530 construct: see for example @file{f/lang-specs.h}.
6532 @item %.@var{SUFFIX}
6533 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6534 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6535 terminated by the next space or %.
6538 Marks the argument containing or following the @samp{%w} as the
6539 designated output file of this compilation. This puts the argument
6540 into the sequence of arguments that @samp{%o} will substitute later.
6543 Substitutes the names of all the output files, with spaces
6544 automatically placed around them. You should write spaces
6545 around the @samp{%o} as well or the results are undefined.
6546 @samp{%o} is for use in the specs for running the linker.
6547 Input files whose names have no recognized suffix are not compiled
6548 at all, but they are included among the output files, so they will
6552 Substitutes the suffix for object files. Note that this is
6553 handled specially when it immediately follows @samp{%g, %u, or %U},
6554 because of the need for those to form complete file names. The
6555 handling is such that @samp{%O} is treated exactly as if it had already
6556 been substituted, except that @samp{%g, %u, and %U} do not currently
6557 support additional @var{suffix} characters following @samp{%O} as they would
6558 following, for example, @samp{.o}.
6561 Substitutes the standard macro predefinitions for the
6562 current target machine. Use this when running @code{cpp}.
6565 Like @samp{%p}, but puts @samp{__} before and after the name of each
6566 predefined macro, except for macros that start with @samp{__} or with
6567 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6571 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6572 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
6573 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6577 Current argument is the name of a library or startup file of some sort.
6578 Search for that file in a standard list of directories and substitute
6579 the full name found.
6582 Print @var{str} as an error message. @var{str} is terminated by a newline.
6583 Use this when inconsistent options are detected.
6586 Substitute the contents of spec string @var{name} at this point.
6589 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6591 @item %x@{@var{option}@}
6592 Accumulate an option for @samp{%X}.
6595 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6599 Output the accumulated assembler options specified by @option{-Wa}.
6602 Output the accumulated preprocessor options specified by @option{-Wp}.
6605 Process the @code{asm} spec. This is used to compute the
6606 switches to be passed to the assembler.
6609 Process the @code{asm_final} spec. This is a spec string for
6610 passing switches to an assembler post-processor, if such a program is
6614 Process the @code{link} spec. This is the spec for computing the
6615 command line passed to the linker. Typically it will make use of the
6616 @samp{%L %G %S %D and %E} sequences.
6619 Dump out a @option{-L} option for each directory that GCC believes might
6620 contain startup files. If the target supports multilibs then the
6621 current multilib directory will be prepended to each of these paths.
6624 Process the @code{lib} spec. This is a spec string for deciding which
6625 libraries should be included on the command line to the linker.
6628 Process the @code{libgcc} spec. This is a spec string for deciding
6629 which GCC support library should be included on the command line to the linker.
6632 Process the @code{startfile} spec. This is a spec for deciding which
6633 object files should be the first ones passed to the linker. Typically
6634 this might be a file named @file{crt0.o}.
6637 Process the @code{endfile} spec. This is a spec string that specifies
6638 the last object files that will be passed to the linker.
6641 Process the @code{cpp} spec. This is used to construct the arguments
6642 to be passed to the C preprocessor.
6645 Process the @code{cc1} spec. This is used to construct the options to be
6646 passed to the actual C compiler (@samp{cc1}).
6649 Process the @code{cc1plus} spec. This is used to construct the options to be
6650 passed to the actual C++ compiler (@samp{cc1plus}).
6653 Substitute the variable part of a matched option. See below.
6654 Note that each comma in the substituted string is replaced by
6658 Remove all occurrences of @code{-S} from the command line. Note---this
6659 command is position dependent. @samp{%} commands in the spec string
6660 before this one will see @code{-S}, @samp{%} commands in the spec string
6661 after this one will not.
6663 @item %:@var{function}(@var{args})
6664 Call the named function @var{function}, passing it @var{args}.
6665 @var{args} is first processed as a nested spec string, then split
6666 into an argument vector in the usual fashion. The function returns
6667 a string which is processed as if it had appeared literally as part
6668 of the current spec.
6670 The following built-in spec functions are provided:
6673 @item @code{if-exists}
6674 The @code{if-exists} spec function takes one argument, an absolute
6675 pathname to a file. If the file exists, @code{if-exists} returns the
6676 pathname. Here is a small example of its usage:
6680 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6683 @item @code{if-exists-else}
6684 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6685 spec function, except that it takes two arguments. The first argument is
6686 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6687 returns the pathname. If it does not exist, it returns the second argument.
6688 This way, @code{if-exists-else} can be used to select one file or another,
6689 based on the existence of the first. Here is a small example of its usage:
6693 crt0%O%s %:if-exists(crti%O%s) \
6694 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6697 @item @code{replace-outfile}
6698 The @code{replace-outfile} spec function takes two arguments. It looks for the
6699 first argument in the outfiles array and replaces it with the second argument. Here
6700 is a small example of its usage:
6703 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
6709 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6710 If that switch was not specified, this substitutes nothing. Note that
6711 the leading dash is omitted when specifying this option, and it is
6712 automatically inserted if the substitution is performed. Thus the spec
6713 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6714 and would output the command line option @option{-foo}.
6716 @item %W@{@code{S}@}
6717 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6720 @item %@{@code{S}*@}
6721 Substitutes all the switches specified to GCC whose names start
6722 with @code{-S}, but which also take an argument. This is used for
6723 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6724 GCC considers @option{-o foo} as being
6725 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6726 text, including the space. Thus two arguments would be generated.
6728 @item %@{@code{S}*&@code{T}*@}
6729 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6730 (the order of @code{S} and @code{T} in the spec is not significant).
6731 There can be any number of ampersand-separated variables; for each the
6732 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6734 @item %@{@code{S}:@code{X}@}
6735 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6737 @item %@{!@code{S}:@code{X}@}
6738 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6740 @item %@{@code{S}*:@code{X}@}
6741 Substitutes @code{X} if one or more switches whose names start with
6742 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6743 once, no matter how many such switches appeared. However, if @code{%*}
6744 appears somewhere in @code{X}, then @code{X} will be substituted once
6745 for each matching switch, with the @code{%*} replaced by the part of
6746 that switch that matched the @code{*}.
6748 @item %@{.@code{S}:@code{X}@}
6749 Substitutes @code{X}, if processing a file with suffix @code{S}.
6751 @item %@{!.@code{S}:@code{X}@}
6752 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6754 @item %@{@code{S}|@code{P}:@code{X}@}
6755 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6756 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
6757 although they have a stronger binding than the @samp{|}. If @code{%*}
6758 appears in @code{X}, all of the alternatives must be starred, and only
6759 the first matching alternative is substituted.
6761 For example, a spec string like this:
6764 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
6767 will output the following command-line options from the following input
6768 command-line options:
6773 -d fred.c -foo -baz -boggle
6774 -d jim.d -bar -baz -boggle
6777 @item %@{S:X; T:Y; :D@}
6779 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
6780 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
6781 be as many clauses as you need. This may be combined with @code{.},
6782 @code{!}, @code{|}, and @code{*} as needed.
6787 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
6788 construct may contain other nested @samp{%} constructs or spaces, or
6789 even newlines. They are processed as usual, as described above.
6790 Trailing white space in @code{X} is ignored. White space may also
6791 appear anywhere on the left side of the colon in these constructs,
6792 except between @code{.} or @code{*} and the corresponding word.
6794 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
6795 handled specifically in these constructs. If another value of
6796 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
6797 @option{-W} switch is found later in the command line, the earlier
6798 switch value is ignored, except with @{@code{S}*@} where @code{S} is
6799 just one letter, which passes all matching options.
6801 The character @samp{|} at the beginning of the predicate text is used to
6802 indicate that a command should be piped to the following command, but
6803 only if @option{-pipe} is specified.
6805 It is built into GCC which switches take arguments and which do not.
6806 (You might think it would be useful to generalize this to allow each
6807 compiler's spec to say which switches take arguments. But this cannot
6808 be done in a consistent fashion. GCC cannot even decide which input
6809 files have been specified without knowing which switches take arguments,
6810 and it must know which input files to compile in order to tell which
6813 GCC also knows implicitly that arguments starting in @option{-l} are to be
6814 treated as compiler output files, and passed to the linker in their
6815 proper position among the other output files.
6817 @c man begin OPTIONS
6819 @node Target Options
6820 @section Specifying Target Machine and Compiler Version
6821 @cindex target options
6822 @cindex cross compiling
6823 @cindex specifying machine version
6824 @cindex specifying compiler version and target machine
6825 @cindex compiler version, specifying
6826 @cindex target machine, specifying
6828 The usual way to run GCC is to run the executable called @file{gcc}, or
6829 @file{<machine>-gcc} when cross-compiling, or
6830 @file{<machine>-gcc-<version>} to run a version other than the one that
6831 was installed last. Sometimes this is inconvenient, so GCC provides
6832 options that will switch to another cross-compiler or version.
6835 @item -b @var{machine}
6837 The argument @var{machine} specifies the target machine for compilation.
6839 The value to use for @var{machine} is the same as was specified as the
6840 machine type when configuring GCC as a cross-compiler. For
6841 example, if a cross-compiler was configured with @samp{configure
6842 i386v}, meaning to compile for an 80386 running System V, then you
6843 would specify @option{-b i386v} to run that cross compiler.
6845 @item -V @var{version}
6847 The argument @var{version} specifies which version of GCC to run.
6848 This is useful when multiple versions are installed. For example,
6849 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
6852 The @option{-V} and @option{-b} options work by running the
6853 @file{<machine>-gcc-<version>} executable, so there's no real reason to
6854 use them if you can just run that directly.
6856 @node Submodel Options
6857 @section Hardware Models and Configurations
6858 @cindex submodel options
6859 @cindex specifying hardware config
6860 @cindex hardware models and configurations, specifying
6861 @cindex machine dependent options
6863 Earlier we discussed the standard option @option{-b} which chooses among
6864 different installed compilers for completely different target
6865 machines, such as VAX vs.@: 68000 vs.@: 80386.
6867 In addition, each of these target machine types can have its own
6868 special options, starting with @samp{-m}, to choose among various
6869 hardware models or configurations---for example, 68010 vs 68020,
6870 floating coprocessor or none. A single installed version of the
6871 compiler can compile for any model or configuration, according to the
6874 Some configurations of the compiler also support additional special
6875 options, usually for compatibility with other compilers on the same
6878 These options are defined by the macro @code{TARGET_SWITCHES} in the
6879 machine description. The default for the options is also defined by
6880 that macro, which enables you to change the defaults.
6882 @c This list is ordered alphanumerically by subsection name.
6883 @c It should be the same order and spelling as these options are listed
6884 @c in Machine Dependent Options
6890 * Blackfin Options::
6893 * DEC Alpha Options::
6894 * DEC Alpha/VMS Options::
6898 * i386 and x86-64 Options::
6910 * RS/6000 and PowerPC Options::
6911 * S/390 and zSeries Options::
6914 * System V Options::
6915 * TMS320C3x/C4x Options::
6919 * Xstormy16 Options::
6925 @subsection ARC Options
6928 These options are defined for ARC implementations:
6933 Compile code for little endian mode. This is the default.
6937 Compile code for big endian mode.
6940 @opindex mmangle-cpu
6941 Prepend the name of the cpu to all public symbol names.
6942 In multiple-processor systems, there are many ARC variants with different
6943 instruction and register set characteristics. This flag prevents code
6944 compiled for one cpu to be linked with code compiled for another.
6945 No facility exists for handling variants that are ``almost identical''.
6946 This is an all or nothing option.
6948 @item -mcpu=@var{cpu}
6950 Compile code for ARC variant @var{cpu}.
6951 Which variants are supported depend on the configuration.
6952 All variants support @option{-mcpu=base}, this is the default.
6954 @item -mtext=@var{text-section}
6955 @itemx -mdata=@var{data-section}
6956 @itemx -mrodata=@var{readonly-data-section}
6960 Put functions, data, and readonly data in @var{text-section},
6961 @var{data-section}, and @var{readonly-data-section} respectively
6962 by default. This can be overridden with the @code{section} attribute.
6963 @xref{Variable Attributes}.
6968 @subsection ARM Options
6971 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6975 @item -mabi=@var{name}
6977 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
6978 @samp{atpcs}, @samp{aapcs} and @samp{iwmmxt}.
6981 @opindex mapcs-frame
6982 Generate a stack frame that is compliant with the ARM Procedure Call
6983 Standard for all functions, even if this is not strictly necessary for
6984 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6985 with this option will cause the stack frames not to be generated for
6986 leaf functions. The default is @option{-mno-apcs-frame}.
6990 This is a synonym for @option{-mapcs-frame}.
6993 @c not currently implemented
6994 @item -mapcs-stack-check
6995 @opindex mapcs-stack-check
6996 Generate code to check the amount of stack space available upon entry to
6997 every function (that actually uses some stack space). If there is
6998 insufficient space available then either the function
6999 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7000 called, depending upon the amount of stack space required. The run time
7001 system is required to provide these functions. The default is
7002 @option{-mno-apcs-stack-check}, since this produces smaller code.
7004 @c not currently implemented
7006 @opindex mapcs-float
7007 Pass floating point arguments using the float point registers. This is
7008 one of the variants of the APCS@. This option is recommended if the
7009 target hardware has a floating point unit or if a lot of floating point
7010 arithmetic is going to be performed by the code. The default is
7011 @option{-mno-apcs-float}, since integer only code is slightly increased in
7012 size if @option{-mapcs-float} is used.
7014 @c not currently implemented
7015 @item -mapcs-reentrant
7016 @opindex mapcs-reentrant
7017 Generate reentrant, position independent code. The default is
7018 @option{-mno-apcs-reentrant}.
7021 @item -mthumb-interwork
7022 @opindex mthumb-interwork
7023 Generate code which supports calling between the ARM and Thumb
7024 instruction sets. Without this option the two instruction sets cannot
7025 be reliably used inside one program. The default is
7026 @option{-mno-thumb-interwork}, since slightly larger code is generated
7027 when @option{-mthumb-interwork} is specified.
7029 @item -mno-sched-prolog
7030 @opindex mno-sched-prolog
7031 Prevent the reordering of instructions in the function prolog, or the
7032 merging of those instruction with the instructions in the function's
7033 body. This means that all functions will start with a recognizable set
7034 of instructions (or in fact one of a choice from a small set of
7035 different function prologues), and this information can be used to
7036 locate the start if functions inside an executable piece of code. The
7037 default is @option{-msched-prolog}.
7040 @opindex mhard-float
7041 Generate output containing floating point instructions. This is the
7045 @opindex msoft-float
7046 Generate output containing library calls for floating point.
7047 @strong{Warning:} the requisite libraries are not available for all ARM
7048 targets. Normally the facilities of the machine's usual C compiler are
7049 used, but this cannot be done directly in cross-compilation. You must make
7050 your own arrangements to provide suitable library functions for
7053 @option{-msoft-float} changes the calling convention in the output file;
7054 therefore, it is only useful if you compile @emph{all} of a program with
7055 this option. In particular, you need to compile @file{libgcc.a}, the
7056 library that comes with GCC, with @option{-msoft-float} in order for
7059 @item -mfloat-abi=@var{name}
7061 Specifies which ABI to use for floating point values. Permissible values
7062 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7064 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7065 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7066 of floating point instructions, but still uses the soft-float calling
7069 @item -mlittle-endian
7070 @opindex mlittle-endian
7071 Generate code for a processor running in little-endian mode. This is
7072 the default for all standard configurations.
7075 @opindex mbig-endian
7076 Generate code for a processor running in big-endian mode; the default is
7077 to compile code for a little-endian processor.
7079 @item -mwords-little-endian
7080 @opindex mwords-little-endian
7081 This option only applies when generating code for big-endian processors.
7082 Generate code for a little-endian word order but a big-endian byte
7083 order. That is, a byte order of the form @samp{32107654}. Note: this
7084 option should only be used if you require compatibility with code for
7085 big-endian ARM processors generated by versions of the compiler prior to
7088 @item -mcpu=@var{name}
7090 This specifies the name of the target ARM processor. GCC uses this name
7091 to determine what kind of instructions it can emit when generating
7092 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7093 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7094 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7095 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7096 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7097 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7098 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7099 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7100 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7101 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7102 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7103 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7104 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7105 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
7108 @itemx -mtune=@var{name}
7110 This option is very similar to the @option{-mcpu=} option, except that
7111 instead of specifying the actual target processor type, and hence
7112 restricting which instructions can be used, it specifies that GCC should
7113 tune the performance of the code as if the target were of the type
7114 specified in this option, but still choosing the instructions that it
7115 will generate based on the cpu specified by a @option{-mcpu=} option.
7116 For some ARM implementations better performance can be obtained by using
7119 @item -march=@var{name}
7121 This specifies the name of the target ARM architecture. GCC uses this
7122 name to determine what kind of instructions it can emit when generating
7123 assembly code. This option can be used in conjunction with or instead
7124 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7125 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7126 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7127 @samp{iwmmxt}, @samp{ep9312}.
7129 @item -mfpu=@var{name}
7130 @itemx -mfpe=@var{number}
7131 @itemx -mfp=@var{number}
7135 This specifies what floating point hardware (or hardware emulation) is
7136 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7137 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7138 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7139 with older versions of GCC@.
7141 If @option{-msoft-float} is specified this specifies the format of
7142 floating point values.
7144 @item -mstructure-size-boundary=@var{n}
7145 @opindex mstructure-size-boundary
7146 The size of all structures and unions will be rounded up to a multiple
7147 of the number of bits set by this option. Permissible values are 8, 32
7148 and 64. The default value varies for different toolchains. For the COFF
7149 targeted toolchain the default value is 8. A value of 64 is only allowed
7150 if the underlying ABI supports it.
7152 Specifying the larger number can produce faster, more efficient code, but
7153 can also increase the size of the program. Different values are potentially
7154 incompatible. Code compiled with one value cannot necessarily expect to
7155 work with code or libraries compiled with another value, if they exchange
7156 information using structures or unions.
7158 @item -mabort-on-noreturn
7159 @opindex mabort-on-noreturn
7160 Generate a call to the function @code{abort} at the end of a
7161 @code{noreturn} function. It will be executed if the function tries to
7165 @itemx -mno-long-calls
7166 @opindex mlong-calls
7167 @opindex mno-long-calls
7168 Tells the compiler to perform function calls by first loading the
7169 address of the function into a register and then performing a subroutine
7170 call on this register. This switch is needed if the target function
7171 will lie outside of the 64 megabyte addressing range of the offset based
7172 version of subroutine call instruction.
7174 Even if this switch is enabled, not all function calls will be turned
7175 into long calls. The heuristic is that static functions, functions
7176 which have the @samp{short-call} attribute, functions that are inside
7177 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7178 definitions have already been compiled within the current compilation
7179 unit, will not be turned into long calls. The exception to this rule is
7180 that weak function definitions, functions with the @samp{long-call}
7181 attribute or the @samp{section} attribute, and functions that are within
7182 the scope of a @samp{#pragma long_calls} directive, will always be
7183 turned into long calls.
7185 This feature is not enabled by default. Specifying
7186 @option{-mno-long-calls} will restore the default behavior, as will
7187 placing the function calls within the scope of a @samp{#pragma
7188 long_calls_off} directive. Note these switches have no effect on how
7189 the compiler generates code to handle function calls via function
7192 @item -mnop-fun-dllimport
7193 @opindex mnop-fun-dllimport
7194 Disable support for the @code{dllimport} attribute.
7196 @item -msingle-pic-base
7197 @opindex msingle-pic-base
7198 Treat the register used for PIC addressing as read-only, rather than
7199 loading it in the prologue for each function. The run-time system is
7200 responsible for initializing this register with an appropriate value
7201 before execution begins.
7203 @item -mpic-register=@var{reg}
7204 @opindex mpic-register
7205 Specify the register to be used for PIC addressing. The default is R10
7206 unless stack-checking is enabled, when R9 is used.
7208 @item -mcirrus-fix-invalid-insns
7209 @opindex mcirrus-fix-invalid-insns
7210 @opindex mno-cirrus-fix-invalid-insns
7211 Insert NOPs into the instruction stream to in order to work around
7212 problems with invalid Maverick instruction combinations. This option
7213 is only valid if the @option{-mcpu=ep9312} option has been used to
7214 enable generation of instructions for the Cirrus Maverick floating
7215 point co-processor. This option is not enabled by default, since the
7216 problem is only present in older Maverick implementations. The default
7217 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7220 @item -mpoke-function-name
7221 @opindex mpoke-function-name
7222 Write the name of each function into the text section, directly
7223 preceding the function prologue. The generated code is similar to this:
7227 .ascii "arm_poke_function_name", 0
7230 .word 0xff000000 + (t1 - t0)
7231 arm_poke_function_name
7233 stmfd sp!, @{fp, ip, lr, pc@}
7237 When performing a stack backtrace, code can inspect the value of
7238 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7239 location @code{pc - 12} and the top 8 bits are set, then we know that
7240 there is a function name embedded immediately preceding this location
7241 and has length @code{((pc[-3]) & 0xff000000)}.
7245 Generate code for the 16-bit Thumb instruction set. The default is to
7246 use the 32-bit ARM instruction set.
7249 @opindex mtpcs-frame
7250 Generate a stack frame that is compliant with the Thumb Procedure Call
7251 Standard for all non-leaf functions. (A leaf function is one that does
7252 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7254 @item -mtpcs-leaf-frame
7255 @opindex mtpcs-leaf-frame
7256 Generate a stack frame that is compliant with the Thumb Procedure Call
7257 Standard for all leaf functions. (A leaf function is one that does
7258 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7260 @item -mcallee-super-interworking
7261 @opindex mcallee-super-interworking
7262 Gives all externally visible functions in the file being compiled an ARM
7263 instruction set header which switches to Thumb mode before executing the
7264 rest of the function. This allows these functions to be called from
7265 non-interworking code.
7267 @item -mcaller-super-interworking
7268 @opindex mcaller-super-interworking
7269 Allows calls via function pointers (including virtual functions) to
7270 execute correctly regardless of whether the target code has been
7271 compiled for interworking or not. There is a small overhead in the cost
7272 of executing a function pointer if this option is enabled.
7277 @subsection AVR Options
7280 These options are defined for AVR implementations:
7283 @item -mmcu=@var{mcu}
7285 Specify ATMEL AVR instruction set or MCU type.
7287 Instruction set avr1 is for the minimal AVR core, not supported by the C
7288 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7289 attiny11, attiny12, attiny15, attiny28).
7291 Instruction set avr2 (default) is for the classic AVR core with up to
7292 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7293 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7294 at90c8534, at90s8535).
7296 Instruction set avr3 is for the classic AVR core with up to 128K program
7297 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7299 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7300 memory space (MCU types: atmega8, atmega83, atmega85).
7302 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7303 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7304 atmega64, atmega128, at43usb355, at94k).
7308 Output instruction sizes to the asm file.
7310 @item -minit-stack=@var{N}
7311 @opindex minit-stack
7312 Specify the initial stack address, which may be a symbol or numeric value,
7313 @samp{__stack} is the default.
7315 @item -mno-interrupts
7316 @opindex mno-interrupts
7317 Generated code is not compatible with hardware interrupts.
7318 Code size will be smaller.
7320 @item -mcall-prologues
7321 @opindex mcall-prologues
7322 Functions prologues/epilogues expanded as call to appropriate
7323 subroutines. Code size will be smaller.
7325 @item -mno-tablejump
7326 @opindex mno-tablejump
7327 Do not generate tablejump insns which sometimes increase code size.
7330 @opindex mtiny-stack
7331 Change only the low 8 bits of the stack pointer.
7335 Assume int to be 8 bit integer. This affects the sizes of all types: A
7336 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7337 and long long will be 4 bytes. Please note that this option does not
7338 comply to the C standards, but it will provide you with smaller code
7342 @node Blackfin Options
7343 @subsection Blackfin Options
7344 @cindex Blackfin Options
7347 @item -momit-leaf-frame-pointer
7348 @opindex momit-leaf-frame-pointer
7349 Don't keep the frame pointer in a register for leaf functions. This
7350 avoids the instructions to save, set up and restore frame pointers and
7351 makes an extra register available in leaf functions. The option
7352 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7353 which might make debugging harder.
7357 When enabled, the compiler will ensure that the generated code does not
7358 contain speculative loads after jump instructions. This option is enabled
7363 Don't generate extra code to prevent speculative loads from occurring.
7367 When enabled, the compiler is free to take advantage of the knowledge that
7368 the entire program fits into the low 64k of memory.
7371 @opindex mno-low-64k
7372 Assume that the program is arbitrarily large. This is the default.
7374 @item -mid-shared-library
7375 @opindex mid-shared-library
7376 Generate code that supports shared libraries via the library ID method.
7377 This allows for execute in place and shared libraries in an environment
7378 without virtual memory management. This option implies @option{-fPIC}.
7380 @item -mno-id-shared-library
7381 @opindex mno-id-shared-library
7382 Generate code that doesn't assume ID based shared libraries are being used.
7383 This is the default.
7385 @item -mshared-library-id=n
7386 @opindex mshared-library-id
7387 Specified the identification number of the ID based shared library being
7388 compiled. Specifying a value of 0 will generate more compact code, specifying
7389 other values will force the allocation of that number to the current
7390 library but is no more space or time efficient than omitting this option.
7394 @subsection CRIS Options
7395 @cindex CRIS Options
7397 These options are defined specifically for the CRIS ports.
7400 @item -march=@var{architecture-type}
7401 @itemx -mcpu=@var{architecture-type}
7404 Generate code for the specified architecture. The choices for
7405 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7406 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7407 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7410 @item -mtune=@var{architecture-type}
7412 Tune to @var{architecture-type} everything applicable about the generated
7413 code, except for the ABI and the set of available instructions. The
7414 choices for @var{architecture-type} are the same as for
7415 @option{-march=@var{architecture-type}}.
7417 @item -mmax-stack-frame=@var{n}
7418 @opindex mmax-stack-frame
7419 Warn when the stack frame of a function exceeds @var{n} bytes.
7421 @item -melinux-stacksize=@var{n}
7422 @opindex melinux-stacksize
7423 Only available with the @samp{cris-axis-aout} target. Arranges for
7424 indications in the program to the kernel loader that the stack of the
7425 program should be set to @var{n} bytes.
7431 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7432 @option{-march=v3} and @option{-march=v8} respectively.
7434 @item -mmul-bug-workaround
7435 @itemx -mno-mul-bug-workaround
7436 @opindex mmul-bug-workaround
7437 @opindex mno-mul-bug-workaround
7438 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7439 models where it applies. This option is active by default.
7443 Enable CRIS-specific verbose debug-related information in the assembly
7444 code. This option also has the effect to turn off the @samp{#NO_APP}
7445 formatted-code indicator to the assembler at the beginning of the
7450 Do not use condition-code results from previous instruction; always emit
7451 compare and test instructions before use of condition codes.
7453 @item -mno-side-effects
7454 @opindex mno-side-effects
7455 Do not emit instructions with side-effects in addressing modes other than
7459 @itemx -mno-stack-align
7461 @itemx -mno-data-align
7462 @itemx -mconst-align
7463 @itemx -mno-const-align
7464 @opindex mstack-align
7465 @opindex mno-stack-align
7466 @opindex mdata-align
7467 @opindex mno-data-align
7468 @opindex mconst-align
7469 @opindex mno-const-align
7470 These options (no-options) arranges (eliminate arrangements) for the
7471 stack-frame, individual data and constants to be aligned for the maximum
7472 single data access size for the chosen CPU model. The default is to
7473 arrange for 32-bit alignment. ABI details such as structure layout are
7474 not affected by these options.
7482 Similar to the stack- data- and const-align options above, these options
7483 arrange for stack-frame, writable data and constants to all be 32-bit,
7484 16-bit or 8-bit aligned. The default is 32-bit alignment.
7486 @item -mno-prologue-epilogue
7487 @itemx -mprologue-epilogue
7488 @opindex mno-prologue-epilogue
7489 @opindex mprologue-epilogue
7490 With @option{-mno-prologue-epilogue}, the normal function prologue and
7491 epilogue that sets up the stack-frame are omitted and no return
7492 instructions or return sequences are generated in the code. Use this
7493 option only together with visual inspection of the compiled code: no
7494 warnings or errors are generated when call-saved registers must be saved,
7495 or storage for local variable needs to be allocated.
7501 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7502 instruction sequences that load addresses for functions from the PLT part
7503 of the GOT rather than (traditional on other architectures) calls to the
7504 PLT@. The default is @option{-mgotplt}.
7508 Legacy no-op option only recognized with the cris-axis-aout target.
7512 Legacy no-op option only recognized with the cris-axis-elf and
7513 cris-axis-linux-gnu targets.
7517 Only recognized with the cris-axis-aout target, where it selects a
7518 GNU/linux-like multilib, include files and instruction set for
7523 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7527 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7528 to link with input-output functions from a simulator library. Code,
7529 initialized data and zero-initialized data are allocated consecutively.
7533 Like @option{-sim}, but pass linker options to locate initialized data at
7534 0x40000000 and zero-initialized data at 0x80000000.
7537 @node Darwin Options
7538 @subsection Darwin Options
7539 @cindex Darwin options
7541 These options are defined for all architectures running the Darwin operating
7544 FSF GCC on Darwin does not create ``fat'' object files; it will create
7545 an object file for the single architecture that it was built to
7546 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7547 @option{-arch} options are used; it does so by running the compiler or
7548 linker multiple times and joining the results together with
7551 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7552 @samp{i686}) is determined by the flags that specify the ISA
7553 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7554 @option{-force_cpusubtype_ALL} option can be used to override this.
7556 The Darwin tools vary in their behavior when presented with an ISA
7557 mismatch. The assembler, @file{as}, will only permit instructions to
7558 be used that are valid for the subtype of the file it is generating,
7559 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7560 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
7561 and print an error if asked to create a shared library with a less
7562 restrictive subtype than its input files (for instance, trying to put
7563 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
7564 for executables, @file{ld}, will quietly give the executable the most
7565 restrictive subtype of any of its input files.
7570 Add the framework directory @var{dir} to the head of the list of
7571 directories to be searched for header files. These directories are
7572 interleaved with those specified by @option{-I} options and are
7573 scanned in a left-to-right order.
7575 A framework directory is a directory with frameworks in it. A
7576 framework is a directory with a @samp{"Headers"} and/or
7577 @samp{"PrivateHeaders"} directory contained directly in it that ends
7578 in @samp{".framework"}. The name of a framework is the name of this
7579 directory excluding the @samp{".framework"}. Headers associated with
7580 the framework are found in one of those two directories, with
7581 @samp{"Headers"} being searched first. A subframework is a framework
7582 directory that is in a framework's @samp{"Frameworks"} directory.
7583 Includes of subframework headers can only appear in a header of a
7584 framework that contains the subframework, or in a sibling subframework
7585 header. Two subframeworks are siblings if they occur in the same
7586 framework. A subframework should not have the same name as a
7587 framework, a warning will be issued if this is violated. Currently a
7588 subframework cannot have subframeworks, in the future, the mechanism
7589 may be extended to support this. The standard frameworks can be found
7590 in @samp{"/System/Library/Frameworks"} and
7591 @samp{"/Library/Frameworks"}. An example include looks like
7592 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
7593 the name of the framework and header.h is found in the
7594 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
7598 Emit debugging information for symbols that are used. For STABS
7599 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
7600 This is by default ON@.
7604 Emit debugging information for all symbols and types.
7606 @item -mone-byte-bool
7607 @opindex -mone-byte-bool
7608 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7609 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7610 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7611 option has no effect on x86.
7613 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
7614 to generate code that is not binary compatible with code generated
7615 without that switch. Using this switch may require recompiling all
7616 other modules in a program, including system libraries. Use this
7617 switch to conform to a non-default data model.
7619 @item -mfix-and-continue
7620 @itemx -ffix-and-continue
7621 @itemx -findirect-data
7622 @opindex mfix-and-continue
7623 @opindex ffix-and-continue
7624 @opindex findirect-data
7625 Generate code suitable for fast turn around development. Needed to
7626 enable gdb to dynamically load @code{.o} files into already running
7627 programs. @option{-findirect-data} and @option{-ffix-and-continue}
7628 are provided for backwards compatibility.
7632 Loads all members of static archive libraries.
7633 See man ld(1) for more information.
7635 @item -arch_errors_fatal
7636 @opindex arch_errors_fatal
7637 Cause the errors having to do with files that have the wrong architecture
7641 @opindex bind_at_load
7642 Causes the output file to be marked such that the dynamic linker will
7643 bind all undefined references when the file is loaded or launched.
7647 Produce a Mach-o bundle format file.
7648 See man ld(1) for more information.
7650 @item -bundle_loader @var{executable}
7651 @opindex bundle_loader
7652 This option specifies the @var{executable} that will be loading the build
7653 output file being linked. See man ld(1) for more information.
7656 @opindex -dynamiclib
7657 When passed this option, GCC will produce a dynamic library instead of
7658 an executable when linking, using the Darwin @file{libtool} command.
7660 @item -force_cpusubtype_ALL
7661 @opindex -force_cpusubtype_ALL
7662 This causes GCC's output file to have the @var{ALL} subtype, instead of
7663 one controlled by the @option{-mcpu} or @option{-march} option.
7665 @item -allowable_client @var{client_name}
7667 @itemx -compatibility_version
7668 @itemx -current_version
7670 @itemx -dependency-file
7672 @itemx -dylinker_install_name
7674 @itemx -exported_symbols_list
7676 @itemx -flat_namespace
7677 @itemx -force_flat_namespace
7678 @itemx -headerpad_max_install_names
7681 @itemx -install_name
7682 @itemx -keep_private_externs
7683 @itemx -multi_module
7684 @itemx -multiply_defined
7685 @itemx -multiply_defined_unused
7687 @itemx -no_dead_strip_inits_and_terms
7688 @itemx -nofixprebinding
7691 @itemx -noseglinkedit
7692 @itemx -pagezero_size
7694 @itemx -prebind_all_twolevel_modules
7695 @itemx -private_bundle
7696 @itemx -read_only_relocs
7698 @itemx -sectobjectsymbols
7702 @itemx -sectobjectsymbols
7705 @itemx -segs_read_only_addr
7706 @itemx -segs_read_write_addr
7707 @itemx -seg_addr_table
7708 @itemx -seg_addr_table_filename
7711 @itemx -segs_read_only_addr
7712 @itemx -segs_read_write_addr
7713 @itemx -single_module
7716 @itemx -sub_umbrella
7717 @itemx -twolevel_namespace
7720 @itemx -unexported_symbols_list
7721 @itemx -weak_reference_mismatches
7724 @opindex allowable_client
7725 @opindex client_name
7726 @opindex compatibility_version
7727 @opindex current_version
7729 @opindex dependency-file
7731 @opindex dylinker_install_name
7733 @opindex exported_symbols_list
7735 @opindex flat_namespace
7736 @opindex force_flat_namespace
7737 @opindex headerpad_max_install_names
7740 @opindex install_name
7741 @opindex keep_private_externs
7742 @opindex multi_module
7743 @opindex multiply_defined
7744 @opindex multiply_defined_unused
7746 @opindex no_dead_strip_inits_and_terms
7747 @opindex nofixprebinding
7748 @opindex nomultidefs
7750 @opindex noseglinkedit
7751 @opindex pagezero_size
7753 @opindex prebind_all_twolevel_modules
7754 @opindex private_bundle
7755 @opindex read_only_relocs
7757 @opindex sectobjectsymbols
7761 @opindex sectobjectsymbols
7764 @opindex segs_read_only_addr
7765 @opindex segs_read_write_addr
7766 @opindex seg_addr_table
7767 @opindex seg_addr_table_filename
7768 @opindex seglinkedit
7770 @opindex segs_read_only_addr
7771 @opindex segs_read_write_addr
7772 @opindex single_module
7774 @opindex sub_library
7775 @opindex sub_umbrella
7776 @opindex twolevel_namespace
7779 @opindex unexported_symbols_list
7780 @opindex weak_reference_mismatches
7781 @opindex whatsloaded
7783 These options are passed to the Darwin linker. The Darwin linker man page
7784 describes them in detail.
7787 @node DEC Alpha Options
7788 @subsection DEC Alpha Options
7790 These @samp{-m} options are defined for the DEC Alpha implementations:
7793 @item -mno-soft-float
7795 @opindex mno-soft-float
7796 @opindex msoft-float
7797 Use (do not use) the hardware floating-point instructions for
7798 floating-point operations. When @option{-msoft-float} is specified,
7799 functions in @file{libgcc.a} will be used to perform floating-point
7800 operations. Unless they are replaced by routines that emulate the
7801 floating-point operations, or compiled in such a way as to call such
7802 emulations routines, these routines will issue floating-point
7803 operations. If you are compiling for an Alpha without floating-point
7804 operations, you must ensure that the library is built so as not to call
7807 Note that Alpha implementations without floating-point operations are
7808 required to have floating-point registers.
7813 @opindex mno-fp-regs
7814 Generate code that uses (does not use) the floating-point register set.
7815 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
7816 register set is not used, floating point operands are passed in integer
7817 registers as if they were integers and floating-point results are passed
7818 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
7819 so any function with a floating-point argument or return value called by code
7820 compiled with @option{-mno-fp-regs} must also be compiled with that
7823 A typical use of this option is building a kernel that does not use,
7824 and hence need not save and restore, any floating-point registers.
7828 The Alpha architecture implements floating-point hardware optimized for
7829 maximum performance. It is mostly compliant with the IEEE floating
7830 point standard. However, for full compliance, software assistance is
7831 required. This option generates code fully IEEE compliant code
7832 @emph{except} that the @var{inexact-flag} is not maintained (see below).
7833 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
7834 defined during compilation. The resulting code is less efficient but is
7835 able to correctly support denormalized numbers and exceptional IEEE
7836 values such as not-a-number and plus/minus infinity. Other Alpha
7837 compilers call this option @option{-ieee_with_no_inexact}.
7839 @item -mieee-with-inexact
7840 @opindex mieee-with-inexact
7841 This is like @option{-mieee} except the generated code also maintains
7842 the IEEE @var{inexact-flag}. Turning on this option causes the
7843 generated code to implement fully-compliant IEEE math. In addition to
7844 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
7845 macro. On some Alpha implementations the resulting code may execute
7846 significantly slower than the code generated by default. Since there is
7847 very little code that depends on the @var{inexact-flag}, you should
7848 normally not specify this option. Other Alpha compilers call this
7849 option @option{-ieee_with_inexact}.
7851 @item -mfp-trap-mode=@var{trap-mode}
7852 @opindex mfp-trap-mode
7853 This option controls what floating-point related traps are enabled.
7854 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
7855 The trap mode can be set to one of four values:
7859 This is the default (normal) setting. The only traps that are enabled
7860 are the ones that cannot be disabled in software (e.g., division by zero
7864 In addition to the traps enabled by @samp{n}, underflow traps are enabled
7868 Like @samp{su}, but the instructions are marked to be safe for software
7869 completion (see Alpha architecture manual for details).
7872 Like @samp{su}, but inexact traps are enabled as well.
7875 @item -mfp-rounding-mode=@var{rounding-mode}
7876 @opindex mfp-rounding-mode
7877 Selects the IEEE rounding mode. Other Alpha compilers call this option
7878 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
7883 Normal IEEE rounding mode. Floating point numbers are rounded towards
7884 the nearest machine number or towards the even machine number in case
7888 Round towards minus infinity.
7891 Chopped rounding mode. Floating point numbers are rounded towards zero.
7894 Dynamic rounding mode. A field in the floating point control register
7895 (@var{fpcr}, see Alpha architecture reference manual) controls the
7896 rounding mode in effect. The C library initializes this register for
7897 rounding towards plus infinity. Thus, unless your program modifies the
7898 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
7901 @item -mtrap-precision=@var{trap-precision}
7902 @opindex mtrap-precision
7903 In the Alpha architecture, floating point traps are imprecise. This
7904 means without software assistance it is impossible to recover from a
7905 floating trap and program execution normally needs to be terminated.
7906 GCC can generate code that can assist operating system trap handlers
7907 in determining the exact location that caused a floating point trap.
7908 Depending on the requirements of an application, different levels of
7909 precisions can be selected:
7913 Program precision. This option is the default and means a trap handler
7914 can only identify which program caused a floating point exception.
7917 Function precision. The trap handler can determine the function that
7918 caused a floating point exception.
7921 Instruction precision. The trap handler can determine the exact
7922 instruction that caused a floating point exception.
7925 Other Alpha compilers provide the equivalent options called
7926 @option{-scope_safe} and @option{-resumption_safe}.
7928 @item -mieee-conformant
7929 @opindex mieee-conformant
7930 This option marks the generated code as IEEE conformant. You must not
7931 use this option unless you also specify @option{-mtrap-precision=i} and either
7932 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
7933 is to emit the line @samp{.eflag 48} in the function prologue of the
7934 generated assembly file. Under DEC Unix, this has the effect that
7935 IEEE-conformant math library routines will be linked in.
7937 @item -mbuild-constants
7938 @opindex mbuild-constants
7939 Normally GCC examines a 32- or 64-bit integer constant to
7940 see if it can construct it from smaller constants in two or three
7941 instructions. If it cannot, it will output the constant as a literal and
7942 generate code to load it from the data segment at runtime.
7944 Use this option to require GCC to construct @emph{all} integer constants
7945 using code, even if it takes more instructions (the maximum is six).
7947 You would typically use this option to build a shared library dynamic
7948 loader. Itself a shared library, it must relocate itself in memory
7949 before it can find the variables and constants in its own data segment.
7955 Select whether to generate code to be assembled by the vendor-supplied
7956 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
7974 Indicate whether GCC should generate code to use the optional BWX,
7975 CIX, FIX and MAX instruction sets. The default is to use the instruction
7976 sets supported by the CPU type specified via @option{-mcpu=} option or that
7977 of the CPU on which GCC was built if none was specified.
7982 @opindex mfloat-ieee
7983 Generate code that uses (does not use) VAX F and G floating point
7984 arithmetic instead of IEEE single and double precision.
7986 @item -mexplicit-relocs
7987 @itemx -mno-explicit-relocs
7988 @opindex mexplicit-relocs
7989 @opindex mno-explicit-relocs
7990 Older Alpha assemblers provided no way to generate symbol relocations
7991 except via assembler macros. Use of these macros does not allow
7992 optimal instruction scheduling. GNU binutils as of version 2.12
7993 supports a new syntax that allows the compiler to explicitly mark
7994 which relocations should apply to which instructions. This option
7995 is mostly useful for debugging, as GCC detects the capabilities of
7996 the assembler when it is built and sets the default accordingly.
8000 @opindex msmall-data
8001 @opindex mlarge-data
8002 When @option{-mexplicit-relocs} is in effect, static data is
8003 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8004 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8005 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8006 16-bit relocations off of the @code{$gp} register. This limits the
8007 size of the small data area to 64KB, but allows the variables to be
8008 directly accessed via a single instruction.
8010 The default is @option{-mlarge-data}. With this option the data area
8011 is limited to just below 2GB@. Programs that require more than 2GB of
8012 data must use @code{malloc} or @code{mmap} to allocate the data in the
8013 heap instead of in the program's data segment.
8015 When generating code for shared libraries, @option{-fpic} implies
8016 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8020 @opindex msmall-text
8021 @opindex mlarge-text
8022 When @option{-msmall-text} is used, the compiler assumes that the
8023 code of the entire program (or shared library) fits in 4MB, and is
8024 thus reachable with a branch instruction. When @option{-msmall-data}
8025 is used, the compiler can assume that all local symbols share the
8026 same @code{$gp} value, and thus reduce the number of instructions
8027 required for a function call from 4 to 1.
8029 The default is @option{-mlarge-text}.
8031 @item -mcpu=@var{cpu_type}
8033 Set the instruction set and instruction scheduling parameters for
8034 machine type @var{cpu_type}. You can specify either the @samp{EV}
8035 style name or the corresponding chip number. GCC supports scheduling
8036 parameters for the EV4, EV5 and EV6 family of processors and will
8037 choose the default values for the instruction set from the processor
8038 you specify. If you do not specify a processor type, GCC will default
8039 to the processor on which the compiler was built.
8041 Supported values for @var{cpu_type} are
8047 Schedules as an EV4 and has no instruction set extensions.
8051 Schedules as an EV5 and has no instruction set extensions.
8055 Schedules as an EV5 and supports the BWX extension.
8060 Schedules as an EV5 and supports the BWX and MAX extensions.
8064 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8068 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8071 @item -mtune=@var{cpu_type}
8073 Set only the instruction scheduling parameters for machine type
8074 @var{cpu_type}. The instruction set is not changed.
8076 @item -mmemory-latency=@var{time}
8077 @opindex mmemory-latency
8078 Sets the latency the scheduler should assume for typical memory
8079 references as seen by the application. This number is highly
8080 dependent on the memory access patterns used by the application
8081 and the size of the external cache on the machine.
8083 Valid options for @var{time} are
8087 A decimal number representing clock cycles.
8093 The compiler contains estimates of the number of clock cycles for
8094 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8095 (also called Dcache, Scache, and Bcache), as well as to main memory.
8096 Note that L3 is only valid for EV5.
8101 @node DEC Alpha/VMS Options
8102 @subsection DEC Alpha/VMS Options
8104 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8107 @item -mvms-return-codes
8108 @opindex mvms-return-codes
8109 Return VMS condition codes from main. The default is to return POSIX
8110 style condition (e.g.@ error) codes.
8114 @subsection FRV Options
8121 Only use the first 32 general purpose registers.
8126 Use all 64 general purpose registers.
8131 Use only the first 32 floating point registers.
8136 Use all 64 floating point registers
8139 @opindex mhard-float
8141 Use hardware instructions for floating point operations.
8144 @opindex msoft-float
8146 Use library routines for floating point operations.
8151 Dynamically allocate condition code registers.
8156 Do not try to dynamically allocate condition code registers, only
8157 use @code{icc0} and @code{fcc0}.
8162 Change ABI to use double word insns.
8167 Do not use double word instructions.
8172 Use floating point double instructions.
8177 Do not use floating point double instructions.
8182 Use media instructions.
8187 Do not use media instructions.
8192 Use multiply and add/subtract instructions.
8197 Do not use multiply and add/subtract instructions.
8202 Select the FDPIC ABI, that uses function descriptors to represent
8203 pointers to functions. Without any PIC/PIE-related options, it
8204 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8205 assumes GOT entries and small data are within a 12-bit range from the
8206 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8207 are computed with 32 bits.
8210 @opindex minline-plt
8212 Enable inlining of PLT entries in function calls to functions that are
8213 not known to bind locally. It has no effect without @option{-mfdpic}.
8214 It's enabled by default if optimizing for speed and compiling for
8215 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8216 optimization option such as @option{-O3} or above is present in the
8222 Assume a large TLS segment when generating thread-local code.
8227 Do not assume a large TLS segment when generating thread-local code.
8232 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8233 that is known to be in read-only sections. It's enabled by default,
8234 except for @option{-fpic} or @option{-fpie}: even though it may help
8235 make the global offset table smaller, it trades 1 instruction for 4.
8236 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8237 one of which may be shared by multiple symbols, and it avoids the need
8238 for a GOT entry for the referenced symbol, so it's more likely to be a
8239 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8241 @item -multilib-library-pic
8242 @opindex multilib-library-pic
8244 Link with the (library, not FD) pic libraries. It's implied by
8245 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8246 @option{-fpic} without @option{-mfdpic}. You should never have to use
8252 Follow the EABI requirement of always creating a frame pointer whenever
8253 a stack frame is allocated. This option is enabled by default and can
8254 be disabled with @option{-mno-linked-fp}.
8257 @opindex mlong-calls
8259 Use indirect addressing to call functions outside the current
8260 compilation unit. This allows the functions to be placed anywhere
8261 within the 32-bit address space.
8263 @item -malign-labels
8264 @opindex malign-labels
8266 Try to align labels to an 8-byte boundary by inserting nops into the
8267 previous packet. This option only has an effect when VLIW packing
8268 is enabled. It doesn't create new packets; it merely adds nops to
8272 @opindex mlibrary-pic
8274 Generate position-independent EABI code.
8279 Use only the first four media accumulator registers.
8284 Use all eight media accumulator registers.
8289 Pack VLIW instructions.
8294 Do not pack VLIW instructions.
8299 Do not mark ABI switches in e_flags.
8304 Enable the use of conditional-move instructions (default).
8306 This switch is mainly for debugging the compiler and will likely be removed
8307 in a future version.
8309 @item -mno-cond-move
8310 @opindex mno-cond-move
8312 Disable the use of conditional-move instructions.
8314 This switch is mainly for debugging the compiler and will likely be removed
8315 in a future version.
8320 Enable the use of conditional set instructions (default).
8322 This switch is mainly for debugging the compiler and will likely be removed
8323 in a future version.
8328 Disable the use of conditional set instructions.
8330 This switch is mainly for debugging the compiler and will likely be removed
8331 in a future version.
8336 Enable the use of conditional execution (default).
8338 This switch is mainly for debugging the compiler and will likely be removed
8339 in a future version.
8341 @item -mno-cond-exec
8342 @opindex mno-cond-exec
8344 Disable the use of conditional execution.
8346 This switch is mainly for debugging the compiler and will likely be removed
8347 in a future version.
8350 @opindex mvliw-branch
8352 Run a pass to pack branches into VLIW instructions (default).
8354 This switch is mainly for debugging the compiler and will likely be removed
8355 in a future version.
8357 @item -mno-vliw-branch
8358 @opindex mno-vliw-branch
8360 Do not run a pass to pack branches into VLIW instructions.
8362 This switch is mainly for debugging the compiler and will likely be removed
8363 in a future version.
8365 @item -mmulti-cond-exec
8366 @opindex mmulti-cond-exec
8368 Enable optimization of @code{&&} and @code{||} in conditional execution
8371 This switch is mainly for debugging the compiler and will likely be removed
8372 in a future version.
8374 @item -mno-multi-cond-exec
8375 @opindex mno-multi-cond-exec
8377 Disable optimization of @code{&&} and @code{||} in conditional execution.
8379 This switch is mainly for debugging the compiler and will likely be removed
8380 in a future version.
8382 @item -mnested-cond-exec
8383 @opindex mnested-cond-exec
8385 Enable nested conditional execution optimizations (default).
8387 This switch is mainly for debugging the compiler and will likely be removed
8388 in a future version.
8390 @item -mno-nested-cond-exec
8391 @opindex mno-nested-cond-exec
8393 Disable nested conditional execution optimizations.
8395 This switch is mainly for debugging the compiler and will likely be removed
8396 in a future version.
8398 @item -mtomcat-stats
8399 @opindex mtomcat-stats
8401 Cause gas to print out tomcat statistics.
8403 @item -mcpu=@var{cpu}
8406 Select the processor type for which to generate code. Possible values are
8407 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8408 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8412 @node H8/300 Options
8413 @subsection H8/300 Options
8415 These @samp{-m} options are defined for the H8/300 implementations:
8420 Shorten some address references at link time, when possible; uses the
8421 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8422 ld, Using ld}, for a fuller description.
8426 Generate code for the H8/300H@.
8430 Generate code for the H8S@.
8434 Generate code for the H8S and H8/300H in the normal mode. This switch
8435 must be used either with @option{-mh} or @option{-ms}.
8439 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8443 Make @code{int} data 32 bits by default.
8447 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8448 The default for the H8/300H and H8S is to align longs and floats on 4
8450 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8451 This option has no effect on the H8/300.
8455 @subsection HPPA Options
8456 @cindex HPPA Options
8458 These @samp{-m} options are defined for the HPPA family of computers:
8461 @item -march=@var{architecture-type}
8463 Generate code for the specified architecture. The choices for
8464 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8465 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8466 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8467 architecture option for your machine. Code compiled for lower numbered
8468 architectures will run on higher numbered architectures, but not the
8472 @itemx -mpa-risc-1-1
8473 @itemx -mpa-risc-2-0
8474 @opindex mpa-risc-1-0
8475 @opindex mpa-risc-1-1
8476 @opindex mpa-risc-2-0
8477 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8480 @opindex mbig-switch
8481 Generate code suitable for big switch tables. Use this option only if
8482 the assembler/linker complain about out of range branches within a switch
8485 @item -mjump-in-delay
8486 @opindex mjump-in-delay
8487 Fill delay slots of function calls with unconditional jump instructions
8488 by modifying the return pointer for the function call to be the target
8489 of the conditional jump.
8491 @item -mdisable-fpregs
8492 @opindex mdisable-fpregs
8493 Prevent floating point registers from being used in any manner. This is
8494 necessary for compiling kernels which perform lazy context switching of
8495 floating point registers. If you use this option and attempt to perform
8496 floating point operations, the compiler will abort.
8498 @item -mdisable-indexing
8499 @opindex mdisable-indexing
8500 Prevent the compiler from using indexing address modes. This avoids some
8501 rather obscure problems when compiling MIG generated code under MACH@.
8503 @item -mno-space-regs
8504 @opindex mno-space-regs
8505 Generate code that assumes the target has no space registers. This allows
8506 GCC to generate faster indirect calls and use unscaled index address modes.
8508 Such code is suitable for level 0 PA systems and kernels.
8510 @item -mfast-indirect-calls
8511 @opindex mfast-indirect-calls
8512 Generate code that assumes calls never cross space boundaries. This
8513 allows GCC to emit code which performs faster indirect calls.
8515 This option will not work in the presence of shared libraries or nested
8518 @item -mfixed-range=@var{register-range}
8519 @opindex mfixed-range
8520 Generate code treating the given register range as fixed registers.
8521 A fixed register is one that the register allocator can not use. This is
8522 useful when compiling kernel code. A register range is specified as
8523 two registers separated by a dash. Multiple register ranges can be
8524 specified separated by a comma.
8526 @item -mlong-load-store
8527 @opindex mlong-load-store
8528 Generate 3-instruction load and store sequences as sometimes required by
8529 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8532 @item -mportable-runtime
8533 @opindex mportable-runtime
8534 Use the portable calling conventions proposed by HP for ELF systems.
8538 Enable the use of assembler directives only GAS understands.
8540 @item -mschedule=@var{cpu-type}
8542 Schedule code according to the constraints for the machine type
8543 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8544 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8545 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8546 proper scheduling option for your machine. The default scheduling is
8550 @opindex mlinker-opt
8551 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8552 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8553 linkers in which they give bogus error messages when linking some programs.
8556 @opindex msoft-float
8557 Generate output containing library calls for floating point.
8558 @strong{Warning:} the requisite libraries are not available for all HPPA
8559 targets. Normally the facilities of the machine's usual C compiler are
8560 used, but this cannot be done directly in cross-compilation. You must make
8561 your own arrangements to provide suitable library functions for
8562 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8563 does provide software floating point support.
8565 @option{-msoft-float} changes the calling convention in the output file;
8566 therefore, it is only useful if you compile @emph{all} of a program with
8567 this option. In particular, you need to compile @file{libgcc.a}, the
8568 library that comes with GCC, with @option{-msoft-float} in order for
8573 Generate the predefine, @code{_SIO}, for server IO@. The default is
8574 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8575 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
8576 options are available under HP-UX and HI-UX@.
8580 Use GNU ld specific options. This passes @option{-shared} to ld when
8581 building a shared library. It is the default when GCC is configured,
8582 explicitly or implicitly, with the GNU linker. This option does not
8583 have any affect on which ld is called, it only changes what parameters
8584 are passed to that ld. The ld that is called is determined by the
8585 @option{--with-ld} configure option, GCC's program search path, and
8586 finally by the user's @env{PATH}. The linker used by GCC can be printed
8587 using @samp{which `gcc -print-prog-name=ld`}.
8591 Use HP ld specific options. This passes @option{-b} to ld when building
8592 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8593 links. It is the default when GCC is configured, explicitly or
8594 implicitly, with the HP linker. This option does not have any affect on
8595 which ld is called, it only changes what parameters are passed to that
8596 ld. The ld that is called is determined by the @option{--with-ld}
8597 configure option, GCC's program search path, and finally by the user's
8598 @env{PATH}. The linker used by GCC can be printed using @samp{which
8599 `gcc -print-prog-name=ld`}.
8602 @opindex mno-long-calls
8603 Generate code that uses long call sequences. This ensures that a call
8604 is always able to reach linker generated stubs. The default is to generate
8605 long calls only when the distance from the call site to the beginning
8606 of the function or translation unit, as the case may be, exceeds a
8607 predefined limit set by the branch type being used. The limits for
8608 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8609 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8612 Distances are measured from the beginning of functions when using the
8613 @option{-ffunction-sections} option, or when using the @option{-mgas}
8614 and @option{-mno-portable-runtime} options together under HP-UX with
8617 It is normally not desirable to use this option as it will degrade
8618 performance. However, it may be useful in large applications,
8619 particularly when partial linking is used to build the application.
8621 The types of long calls used depends on the capabilities of the
8622 assembler and linker, and the type of code being generated. The
8623 impact on systems that support long absolute calls, and long pic
8624 symbol-difference or pc-relative calls should be relatively small.
8625 However, an indirect call is used on 32-bit ELF systems in pic code
8626 and it is quite long.
8628 @item -munix=@var{unix-std}
8630 Generate compiler predefines and select a startfile for the specified
8631 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
8632 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
8633 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
8634 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
8635 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
8638 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
8639 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
8640 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
8641 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
8642 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
8643 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
8645 It is @emph{important} to note that this option changes the interfaces
8646 for various library routines. It also affects the operational behavior
8647 of the C library. Thus, @emph{extreme} care is needed in using this
8650 Library code that is intended to operate with more than one UNIX
8651 standard must test, set and restore the variable @var{__xpg4_extended_mask}
8652 as appropriate. Most GNU software doesn't provide this capability.
8656 Suppress the generation of link options to search libdld.sl when the
8657 @option{-static} option is specified on HP-UX 10 and later.
8661 The HP-UX implementation of setlocale in libc has a dependency on
8662 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8663 when the @option{-static} option is specified, special link options
8664 are needed to resolve this dependency.
8666 On HP-UX 10 and later, the GCC driver adds the necessary options to
8667 link with libdld.sl when the @option{-static} option is specified.
8668 This causes the resulting binary to be dynamic. On the 64-bit port,
8669 the linkers generate dynamic binaries by default in any case. The
8670 @option{-nolibdld} option can be used to prevent the GCC driver from
8671 adding these link options.
8675 Add support for multithreading with the @dfn{dce thread} library
8676 under HP-UX@. This option sets flags for both the preprocessor and
8680 @node i386 and x86-64 Options
8681 @subsection Intel 386 and AMD x86-64 Options
8682 @cindex i386 Options
8683 @cindex x86-64 Options
8684 @cindex Intel 386 Options
8685 @cindex AMD x86-64 Options
8687 These @samp{-m} options are defined for the i386 and x86-64 family of
8691 @item -mtune=@var{cpu-type}
8693 Tune to @var{cpu-type} everything applicable about the generated code, except
8694 for the ABI and the set of available instructions. The choices for
8698 Original Intel's i386 CPU@.
8700 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
8702 Intel Pentium CPU with no MMX support.
8704 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
8705 @item i686, pentiumpro
8706 Intel PentiumPro CPU@.
8708 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
8709 @item pentium3, pentium3m
8710 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
8713 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
8714 support. Used by Centrino notebooks.
8715 @item pentium4, pentium4m
8716 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
8718 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
8721 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
8722 SSE2 and SSE3 instruction set support.
8724 AMD K6 CPU with MMX instruction set support.
8726 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
8727 @item athlon, athlon-tbird
8728 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
8730 @item athlon-4, athlon-xp, athlon-mp
8731 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
8732 instruction set support.
8733 @item k8, opteron, athlon64, athlon-fx
8734 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
8735 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
8737 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
8740 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
8741 instruction set support.
8743 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
8744 implemented for this chip.)
8746 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
8747 implemented for this chip.)
8750 While picking a specific @var{cpu-type} will schedule things appropriately
8751 for that particular chip, the compiler will not generate any code that
8752 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8755 @item -march=@var{cpu-type}
8757 Generate instructions for the machine type @var{cpu-type}. The choices
8758 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8759 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8761 @item -mcpu=@var{cpu-type}
8763 A deprecated synonym for @option{-mtune}.
8772 @opindex mpentiumpro
8773 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8774 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8775 These synonyms are deprecated.
8777 @item -mfpmath=@var{unit}
8779 Generate floating point arithmetics for selected unit @var{unit}. The choices
8784 Use the standard 387 floating point coprocessor present majority of chips and
8785 emulated otherwise. Code compiled with this option will run almost everywhere.
8786 The temporary results are computed in 80bit precision instead of precision
8787 specified by the type resulting in slightly different results compared to most
8788 of other chips. See @option{-ffloat-store} for more detailed description.
8790 This is the default choice for i386 compiler.
8793 Use scalar floating point instructions present in the SSE instruction set.
8794 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8795 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8796 instruction set supports only single precision arithmetics, thus the double and
8797 extended precision arithmetics is still done using 387. Later version, present
8798 only in Pentium4 and the future AMD x86-64 chips supports double precision
8801 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
8802 or @option{-msse2} switches to enable SSE extensions and make this option
8803 effective. For the x86-64 compiler, these extensions are enabled by default.
8805 The resulting code should be considerably faster in the majority of cases and avoid
8806 the numerical instability problems of 387 code, but may break some existing
8807 code that expects temporaries to be 80bit.
8809 This is the default choice for the x86-64 compiler.
8812 Attempt to utilize both instruction sets at once. This effectively double the
8813 amount of available registers and on chips with separate execution units for
8814 387 and SSE the execution resources too. Use this option with care, as it is
8815 still experimental, because the GCC register allocator does not model separate
8816 functional units well resulting in instable performance.
8819 @item -masm=@var{dialect}
8820 @opindex masm=@var{dialect}
8821 Output asm instructions using selected @var{dialect}. Supported choices are
8822 @samp{intel} or @samp{att} (the default one).
8827 @opindex mno-ieee-fp
8828 Control whether or not the compiler uses IEEE floating point
8829 comparisons. These handle correctly the case where the result of a
8830 comparison is unordered.
8833 @opindex msoft-float
8834 Generate output containing library calls for floating point.
8835 @strong{Warning:} the requisite libraries are not part of GCC@.
8836 Normally the facilities of the machine's usual C compiler are used, but
8837 this can't be done directly in cross-compilation. You must make your
8838 own arrangements to provide suitable library functions for
8841 On machines where a function returns floating point results in the 80387
8842 register stack, some floating point opcodes may be emitted even if
8843 @option{-msoft-float} is used.
8845 @item -mno-fp-ret-in-387
8846 @opindex mno-fp-ret-in-387
8847 Do not use the FPU registers for return values of functions.
8849 The usual calling convention has functions return values of types
8850 @code{float} and @code{double} in an FPU register, even if there
8851 is no FPU@. The idea is that the operating system should emulate
8854 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8855 in ordinary CPU registers instead.
8857 @item -mno-fancy-math-387
8858 @opindex mno-fancy-math-387
8859 Some 387 emulators do not support the @code{sin}, @code{cos} and
8860 @code{sqrt} instructions for the 387. Specify this option to avoid
8861 generating those instructions. This option is the default on FreeBSD,
8862 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8863 indicates that the target cpu will always have an FPU and so the
8864 instruction will not need emulation. As of revision 2.6.1, these
8865 instructions are not generated unless you also use the
8866 @option{-funsafe-math-optimizations} switch.
8868 @item -malign-double
8869 @itemx -mno-align-double
8870 @opindex malign-double
8871 @opindex mno-align-double
8872 Control whether GCC aligns @code{double}, @code{long double}, and
8873 @code{long long} variables on a two word boundary or a one word
8874 boundary. Aligning @code{double} variables on a two word boundary will
8875 produce code that runs somewhat faster on a @samp{Pentium} at the
8876 expense of more memory.
8878 @strong{Warning:} if you use the @option{-malign-double} switch,
8879 structures containing the above types will be aligned differently than
8880 the published application binary interface specifications for the 386
8881 and will not be binary compatible with structures in code compiled
8882 without that switch.
8884 @item -m96bit-long-double
8885 @itemx -m128bit-long-double
8886 @opindex m96bit-long-double
8887 @opindex m128bit-long-double
8888 These switches control the size of @code{long double} type. The i386
8889 application binary interface specifies the size to be 96 bits,
8890 so @option{-m96bit-long-double} is the default in 32 bit mode.
8892 Modern architectures (Pentium and newer) would prefer @code{long double}
8893 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8894 conforming to the ABI, this would not be possible. So specifying a
8895 @option{-m128bit-long-double} will align @code{long double}
8896 to a 16 byte boundary by padding the @code{long double} with an additional
8899 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8900 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8902 Notice that neither of these options enable any extra precision over the x87
8903 standard of 80 bits for a @code{long double}.
8905 @strong{Warning:} if you override the default value for your target ABI, the
8906 structures and arrays containing @code{long double} variables will change
8907 their size as well as function calling convention for function taking
8908 @code{long double} will be modified. Hence they will not be binary
8909 compatible with arrays or structures in code compiled without that switch.
8913 @itemx -mno-svr3-shlib
8914 @opindex msvr3-shlib
8915 @opindex mno-svr3-shlib
8916 Control whether GCC places uninitialized local variables into the
8917 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8918 into @code{bss}. These options are meaningful only on System V Release 3.
8922 Use a different function-calling convention, in which functions that
8923 take a fixed number of arguments return with the @code{ret} @var{num}
8924 instruction, which pops their arguments while returning. This saves one
8925 instruction in the caller since there is no need to pop the arguments
8928 You can specify that an individual function is called with this calling
8929 sequence with the function attribute @samp{stdcall}. You can also
8930 override the @option{-mrtd} option by using the function attribute
8931 @samp{cdecl}. @xref{Function Attributes}.
8933 @strong{Warning:} this calling convention is incompatible with the one
8934 normally used on Unix, so you cannot use it if you need to call
8935 libraries compiled with the Unix compiler.
8937 Also, you must provide function prototypes for all functions that
8938 take variable numbers of arguments (including @code{printf});
8939 otherwise incorrect code will be generated for calls to those
8942 In addition, seriously incorrect code will result if you call a
8943 function with too many arguments. (Normally, extra arguments are
8944 harmlessly ignored.)
8946 @item -mregparm=@var{num}
8948 Control how many registers are used to pass integer arguments. By
8949 default, no registers are used to pass arguments, and at most 3
8950 registers can be used. You can control this behavior for a specific
8951 function by using the function attribute @samp{regparm}.
8952 @xref{Function Attributes}.
8954 @strong{Warning:} if you use this switch, and
8955 @var{num} is nonzero, then you must build all modules with the same
8956 value, including any libraries. This includes the system libraries and
8959 @item -mpreferred-stack-boundary=@var{num}
8960 @opindex mpreferred-stack-boundary
8961 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8962 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8963 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8964 size (@option{-Os}), in which case the default is the minimum correct
8965 alignment (4 bytes for x86, and 8 bytes for x86-64).
8967 On Pentium and PentiumPro, @code{double} and @code{long double} values
8968 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8969 suffer significant run time performance penalties. On Pentium III, the
8970 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8971 penalties if it is not 16 byte aligned.
8973 To ensure proper alignment of this values on the stack, the stack boundary
8974 must be as aligned as that required by any value stored on the stack.
8975 Further, every function must be generated such that it keeps the stack
8976 aligned. Thus calling a function compiled with a higher preferred
8977 stack boundary from a function compiled with a lower preferred stack
8978 boundary will most likely misalign the stack. It is recommended that
8979 libraries that use callbacks always use the default setting.
8981 This extra alignment does consume extra stack space, and generally
8982 increases code size. Code that is sensitive to stack space usage, such
8983 as embedded systems and operating system kernels, may want to reduce the
8984 preferred alignment to @option{-mpreferred-stack-boundary=2}.
9002 These switches enable or disable the use of built-in functions that allow
9003 direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow extensions of the
9006 @xref{X86 Built-in Functions}, for details of the functions enabled
9007 and disabled by these switches.
9009 To have SSE/SSE2 instructions generated automatically from floating-point
9010 code, see @option{-mfpmath=sse}.
9013 @itemx -mno-push-args
9015 @opindex mno-push-args
9016 Use PUSH operations to store outgoing parameters. This method is shorter
9017 and usually equally fast as method using SUB/MOV operations and is enabled
9018 by default. In some cases disabling it may improve performance because of
9019 improved scheduling and reduced dependencies.
9021 @item -maccumulate-outgoing-args
9022 @opindex maccumulate-outgoing-args
9023 If enabled, the maximum amount of space required for outgoing arguments will be
9024 computed in the function prologue. This is faster on most modern CPUs
9025 because of reduced dependencies, improved scheduling and reduced stack usage
9026 when preferred stack boundary is not equal to 2. The drawback is a notable
9027 increase in code size. This switch implies @option{-mno-push-args}.
9031 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
9032 on thread-safe exception handling must compile and link all code with the
9033 @option{-mthreads} option. When compiling, @option{-mthreads} defines
9034 @option{-D_MT}; when linking, it links in a special thread helper library
9035 @option{-lmingwthrd} which cleans up per thread exception handling data.
9037 @item -mno-align-stringops
9038 @opindex mno-align-stringops
9039 Do not align destination of inlined string operations. This switch reduces
9040 code size and improves performance in case the destination is already aligned,
9041 but GCC doesn't know about it.
9043 @item -minline-all-stringops
9044 @opindex minline-all-stringops
9045 By default GCC inlines string operations only when destination is known to be
9046 aligned at least to 4 byte boundary. This enables more inlining, increase code
9047 size, but may improve performance of code that depends on fast memcpy, strlen
9048 and memset for short lengths.
9050 @item -momit-leaf-frame-pointer
9051 @opindex momit-leaf-frame-pointer
9052 Don't keep the frame pointer in a register for leaf functions. This
9053 avoids the instructions to save, set up and restore frame pointers and
9054 makes an extra register available in leaf functions. The option
9055 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9056 which might make debugging harder.
9058 @item -mtls-direct-seg-refs
9059 @itemx -mno-tls-direct-seg-refs
9060 @opindex mtls-direct-seg-refs
9061 Controls whether TLS variables may be accessed with offsets from the
9062 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
9063 or whether the thread base pointer must be added. Whether or not this
9064 is legal depends on the operating system, and whether it maps the
9065 segment to cover the entire TLS area.
9067 For systems that use GNU libc, the default is on.
9070 These @samp{-m} switches are supported in addition to the above
9071 on AMD x86-64 processors in 64-bit environments.
9078 Generate code for a 32-bit or 64-bit environment.
9079 The 32-bit environment sets int, long and pointer to 32 bits and
9080 generates code that runs on any i386 system.
9081 The 64-bit environment sets int to 32 bits and long and pointer
9082 to 64 bits and generates code for AMD's x86-64 architecture.
9085 @opindex no-red-zone
9086 Do not use a so called red zone for x86-64 code. The red zone is mandated
9087 by the x86-64 ABI, it is a 128-byte area beyond the location of the
9088 stack pointer that will not be modified by signal or interrupt handlers
9089 and therefore can be used for temporary data without adjusting the stack
9090 pointer. The flag @option{-mno-red-zone} disables this red zone.
9092 @item -mcmodel=small
9093 @opindex mcmodel=small
9094 Generate code for the small code model: the program and its symbols must
9095 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
9096 Programs can be statically or dynamically linked. This is the default
9099 @item -mcmodel=kernel
9100 @opindex mcmodel=kernel
9101 Generate code for the kernel code model. The kernel runs in the
9102 negative 2 GB of the address space.
9103 This model has to be used for Linux kernel code.
9105 @item -mcmodel=medium
9106 @opindex mcmodel=medium
9107 Generate code for the medium model: The program is linked in the lower 2
9108 GB of the address space but symbols can be located anywhere in the
9109 address space. Programs can be statically or dynamically linked, but
9110 building of shared libraries are not supported with the medium model.
9112 @item -mcmodel=large
9113 @opindex mcmodel=large
9114 Generate code for the large model: This model makes no assumptions
9115 about addresses and sizes of sections. Currently GCC does not implement
9120 @subsection IA-64 Options
9121 @cindex IA-64 Options
9123 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9127 @opindex mbig-endian
9128 Generate code for a big endian target. This is the default for HP-UX@.
9130 @item -mlittle-endian
9131 @opindex mlittle-endian
9132 Generate code for a little endian target. This is the default for AIX5
9139 Generate (or don't) code for the GNU assembler. This is the default.
9140 @c Also, this is the default if the configure option @option{--with-gnu-as}
9147 Generate (or don't) code for the GNU linker. This is the default.
9148 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9153 Generate code that does not use a global pointer register. The result
9154 is not position independent code, and violates the IA-64 ABI@.
9156 @item -mvolatile-asm-stop
9157 @itemx -mno-volatile-asm-stop
9158 @opindex mvolatile-asm-stop
9159 @opindex mno-volatile-asm-stop
9160 Generate (or don't) a stop bit immediately before and after volatile asm
9163 @item -mregister-names
9164 @itemx -mno-register-names
9165 @opindex mregister-names
9166 @opindex mno-register-names
9167 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9168 the stacked registers. This may make assembler output more readable.
9174 Disable (or enable) optimizations that use the small data section. This may
9175 be useful for working around optimizer bugs.
9178 @opindex mconstant-gp
9179 Generate code that uses a single constant global pointer value. This is
9180 useful when compiling kernel code.
9184 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9185 This is useful when compiling firmware code.
9187 @item -minline-float-divide-min-latency
9188 @opindex minline-float-divide-min-latency
9189 Generate code for inline divides of floating point values
9190 using the minimum latency algorithm.
9192 @item -minline-float-divide-max-throughput
9193 @opindex minline-float-divide-max-throughput
9194 Generate code for inline divides of floating point values
9195 using the maximum throughput algorithm.
9197 @item -minline-int-divide-min-latency
9198 @opindex minline-int-divide-min-latency
9199 Generate code for inline divides of integer values
9200 using the minimum latency algorithm.
9202 @item -minline-int-divide-max-throughput
9203 @opindex minline-int-divide-max-throughput
9204 Generate code for inline divides of integer values
9205 using the maximum throughput algorithm.
9207 @item -minline-sqrt-min-latency
9208 @opindex minline-sqrt-min-latency
9209 Generate code for inline square roots
9210 using the minimum latency algorithm.
9212 @item -minline-sqrt-max-throughput
9213 @opindex minline-sqrt-max-throughput
9214 Generate code for inline square roots
9215 using the maximum throughput algorithm.
9217 @item -mno-dwarf2-asm
9219 @opindex mno-dwarf2-asm
9220 @opindex mdwarf2-asm
9221 Don't (or do) generate assembler code for the DWARF2 line number debugging
9222 info. This may be useful when not using the GNU assembler.
9224 @item -mearly-stop-bits
9225 @itemx -mno-early-stop-bits
9226 @opindex mearly-stop-bits
9227 @opindex mno-early-stop-bits
9228 Allow stop bits to be placed earlier than immediately preceding the
9229 instruction that triggered the stop bit. This can improve instruction
9230 scheduling, but does not always do so.
9232 @item -mfixed-range=@var{register-range}
9233 @opindex mfixed-range
9234 Generate code treating the given register range as fixed registers.
9235 A fixed register is one that the register allocator can not use. This is
9236 useful when compiling kernel code. A register range is specified as
9237 two registers separated by a dash. Multiple register ranges can be
9238 specified separated by a comma.
9240 @item -mtls-size=@var{tls-size}
9242 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9245 @item -mtune-arch=@var{cpu-type}
9247 Tune the instruction scheduling for a particular CPU, Valid values are
9248 itanium, itanium1, merced, itanium2, and mckinley.
9254 Add support for multithreading using the POSIX threads library. This
9255 option sets flags for both the preprocessor and linker. It does
9256 not affect the thread safety of object code produced by the compiler or
9257 that of libraries supplied with it. These are HP-UX specific flags.
9263 Generate code for a 32-bit or 64-bit environment.
9264 The 32-bit environment sets int, long and pointer to 32 bits.
9265 The 64-bit environment sets int to 32 bits and long and pointer
9266 to 64 bits. These are HP-UX specific flags.
9270 @node M32R/D Options
9271 @subsection M32R/D Options
9272 @cindex M32R/D options
9274 These @option{-m} options are defined for Renesas M32R/D architectures:
9279 Generate code for the M32R/2@.
9283 Generate code for the M32R/X@.
9287 Generate code for the M32R@. This is the default.
9290 @opindex mmodel=small
9291 Assume all objects live in the lower 16MB of memory (so that their addresses
9292 can be loaded with the @code{ld24} instruction), and assume all subroutines
9293 are reachable with the @code{bl} instruction.
9294 This is the default.
9296 The addressability of a particular object can be set with the
9297 @code{model} attribute.
9299 @item -mmodel=medium
9300 @opindex mmodel=medium
9301 Assume objects may be anywhere in the 32-bit address space (the compiler
9302 will generate @code{seth/add3} instructions to load their addresses), and
9303 assume all subroutines are reachable with the @code{bl} instruction.
9306 @opindex mmodel=large
9307 Assume objects may be anywhere in the 32-bit address space (the compiler
9308 will generate @code{seth/add3} instructions to load their addresses), and
9309 assume subroutines may not be reachable with the @code{bl} instruction
9310 (the compiler will generate the much slower @code{seth/add3/jl}
9311 instruction sequence).
9314 @opindex msdata=none
9315 Disable use of the small data area. Variables will be put into
9316 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
9317 @code{section} attribute has been specified).
9318 This is the default.
9320 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
9321 Objects may be explicitly put in the small data area with the
9322 @code{section} attribute using one of these sections.
9325 @opindex msdata=sdata
9326 Put small global and static data in the small data area, but do not
9327 generate special code to reference them.
9331 Put small global and static data in the small data area, and generate
9332 special instructions to reference them.
9336 @cindex smaller data references
9337 Put global and static objects less than or equal to @var{num} bytes
9338 into the small data or bss sections instead of the normal data or bss
9339 sections. The default value of @var{num} is 8.
9340 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
9341 for this option to have any effect.
9343 All modules should be compiled with the same @option{-G @var{num}} value.
9344 Compiling with different values of @var{num} may or may not work; if it
9345 doesn't the linker will give an error message---incorrect code will not be
9350 Makes the M32R specific code in the compiler display some statistics
9351 that might help in debugging programs.
9354 @opindex malign-loops
9355 Align all loops to a 32-byte boundary.
9357 @item -mno-align-loops
9358 @opindex mno-align-loops
9359 Do not enforce a 32-byte alignment for loops. This is the default.
9361 @item -missue-rate=@var{number}
9362 @opindex missue-rate=@var{number}
9363 Issue @var{number} instructions per cycle. @var{number} can only be 1
9366 @item -mbranch-cost=@var{number}
9367 @opindex mbranch-cost=@var{number}
9368 @var{number} can only be 1 or 2. If it is 1 then branches will be
9369 preferred over conditional code, if it is 2, then the opposite will
9372 @item -mflush-trap=@var{number}
9373 @opindex mflush-trap=@var{number}
9374 Specifies the trap number to use to flush the cache. The default is
9375 12. Valid numbers are between 0 and 15 inclusive.
9377 @item -mno-flush-trap
9378 @opindex mno-flush-trap
9379 Specifies that the cache cannot be flushed by using a trap.
9381 @item -mflush-func=@var{name}
9382 @opindex mflush-func=@var{name}
9383 Specifies the name of the operating system function to call to flush
9384 the cache. The default is @emph{_flush_cache}, but a function call
9385 will only be used if a trap is not available.
9387 @item -mno-flush-func
9388 @opindex mno-flush-func
9389 Indicates that there is no OS function for flushing the cache.
9393 @node M680x0 Options
9394 @subsection M680x0 Options
9395 @cindex M680x0 options
9397 These are the @samp{-m} options defined for the 68000 series. The default
9398 values for these options depends on which style of 68000 was selected when
9399 the compiler was configured; the defaults for the most common choices are
9407 Generate output for a 68000. This is the default
9408 when the compiler is configured for 68000-based systems.
9410 Use this option for microcontrollers with a 68000 or EC000 core,
9411 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
9417 Generate output for a 68020. This is the default
9418 when the compiler is configured for 68020-based systems.
9422 Generate output containing 68881 instructions for floating point.
9423 This is the default for most 68020 systems unless @option{--nfp} was
9424 specified when the compiler was configured.
9428 Generate output for a 68030. This is the default when the compiler is
9429 configured for 68030-based systems.
9433 Generate output for a 68040. This is the default when the compiler is
9434 configured for 68040-based systems.
9436 This option inhibits the use of 68881/68882 instructions that have to be
9437 emulated by software on the 68040. Use this option if your 68040 does not
9438 have code to emulate those instructions.
9442 Generate output for a 68060. This is the default when the compiler is
9443 configured for 68060-based systems.
9445 This option inhibits the use of 68020 and 68881/68882 instructions that
9446 have to be emulated by software on the 68060. Use this option if your 68060
9447 does not have code to emulate those instructions.
9451 Generate output for a CPU32. This is the default
9452 when the compiler is configured for CPU32-based systems.
9454 Use this option for microcontrollers with a
9455 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
9456 68336, 68340, 68341, 68349 and 68360.
9460 Generate output for a 520X ``coldfire'' family cpu. This is the default
9461 when the compiler is configured for 520X-based systems.
9463 Use this option for microcontroller with a 5200 core, including
9464 the MCF5202, MCF5203, MCF5204 and MCF5202.
9469 Generate output for a 68040, without using any of the new instructions.
9470 This results in code which can run relatively efficiently on either a
9471 68020/68881 or a 68030 or a 68040. The generated code does use the
9472 68881 instructions that are emulated on the 68040.
9476 Generate output for a 68060, without using any of the new instructions.
9477 This results in code which can run relatively efficiently on either a
9478 68020/68881 or a 68030 or a 68040. The generated code does use the
9479 68881 instructions that are emulated on the 68060.
9482 @opindex msoft-float
9483 Generate output containing library calls for floating point.
9484 @strong{Warning:} the requisite libraries are not available for all m68k
9485 targets. Normally the facilities of the machine's usual C compiler are
9486 used, but this can't be done directly in cross-compilation. You must
9487 make your own arrangements to provide suitable library functions for
9488 cross-compilation. The embedded targets @samp{m68k-*-aout} and
9489 @samp{m68k-*-coff} do provide software floating point support.
9493 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9494 Additionally, parameters passed on the stack are also aligned to a
9495 16-bit boundary even on targets whose API mandates promotion to 32-bit.
9498 @opindex mnobitfield
9499 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
9500 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
9504 Do use the bit-field instructions. The @option{-m68020} option implies
9505 @option{-mbitfield}. This is the default if you use a configuration
9506 designed for a 68020.
9510 Use a different function-calling convention, in which functions
9511 that take a fixed number of arguments return with the @code{rtd}
9512 instruction, which pops their arguments while returning. This
9513 saves one instruction in the caller since there is no need to pop
9514 the arguments there.
9516 This calling convention is incompatible with the one normally
9517 used on Unix, so you cannot use it if you need to call libraries
9518 compiled with the Unix compiler.
9520 Also, you must provide function prototypes for all functions that
9521 take variable numbers of arguments (including @code{printf});
9522 otherwise incorrect code will be generated for calls to those
9525 In addition, seriously incorrect code will result if you call a
9526 function with too many arguments. (Normally, extra arguments are
9527 harmlessly ignored.)
9529 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
9530 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
9533 @itemx -mno-align-int
9535 @opindex mno-align-int
9536 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
9537 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
9538 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
9539 Aligning variables on 32-bit boundaries produces code that runs somewhat
9540 faster on processors with 32-bit busses at the expense of more memory.
9542 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
9543 align structures containing the above types differently than
9544 most published application binary interface specifications for the m68k.
9548 Use the pc-relative addressing mode of the 68000 directly, instead of
9549 using a global offset table. At present, this option implies @option{-fpic},
9550 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
9551 not presently supported with @option{-mpcrel}, though this could be supported for
9552 68020 and higher processors.
9554 @item -mno-strict-align
9555 @itemx -mstrict-align
9556 @opindex mno-strict-align
9557 @opindex mstrict-align
9558 Do not (do) assume that unaligned memory references will be handled by
9562 Generate code that allows the data segment to be located in a different
9563 area of memory from the text segment. This allows for execute in place in
9564 an environment without virtual memory management. This option implies
9568 Generate code that assumes that the data segment follows the text segment.
9569 This is the default.
9571 @item -mid-shared-library
9572 Generate code that supports shared libraries via the library ID method.
9573 This allows for execute in place and shared libraries in an environment
9574 without virtual memory management. This option implies @option{-fPIC}.
9576 @item -mno-id-shared-library
9577 Generate code that doesn't assume ID based shared libraries are being used.
9578 This is the default.
9580 @item -mshared-library-id=n
9581 Specified the identification number of the ID based shared library being
9582 compiled. Specifying a value of 0 will generate more compact code, specifying
9583 other values will force the allocation of that number to the current
9584 library but is no more space or time efficient than omitting this option.
9588 @node M68hc1x Options
9589 @subsection M68hc1x Options
9590 @cindex M68hc1x options
9592 These are the @samp{-m} options defined for the 68hc11 and 68hc12
9593 microcontrollers. The default values for these options depends on
9594 which style of microcontroller was selected when the compiler was configured;
9595 the defaults for the most common choices are given below.
9602 Generate output for a 68HC11. This is the default
9603 when the compiler is configured for 68HC11-based systems.
9609 Generate output for a 68HC12. This is the default
9610 when the compiler is configured for 68HC12-based systems.
9616 Generate output for a 68HCS12.
9619 @opindex mauto-incdec
9620 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
9627 Enable the use of 68HC12 min and max instructions.
9630 @itemx -mno-long-calls
9631 @opindex mlong-calls
9632 @opindex mno-long-calls
9633 Treat all calls as being far away (near). If calls are assumed to be
9634 far away, the compiler will use the @code{call} instruction to
9635 call a function and the @code{rtc} instruction for returning.
9639 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9641 @item -msoft-reg-count=@var{count}
9642 @opindex msoft-reg-count
9643 Specify the number of pseudo-soft registers which are used for the
9644 code generation. The maximum number is 32. Using more pseudo-soft
9645 register may or may not result in better code depending on the program.
9646 The default is 4 for 68HC11 and 2 for 68HC12.
9651 @subsection MCore Options
9652 @cindex MCore options
9654 These are the @samp{-m} options defined for the Motorola M*Core
9662 @opindex mno-hardlit
9663 Inline constants into the code stream if it can be done in two
9664 instructions or less.
9670 Use the divide instruction. (Enabled by default).
9672 @item -mrelax-immediate
9673 @itemx -mno-relax-immediate
9674 @opindex mrelax-immediate
9675 @opindex mno-relax-immediate
9676 Allow arbitrary sized immediates in bit operations.
9678 @item -mwide-bitfields
9679 @itemx -mno-wide-bitfields
9680 @opindex mwide-bitfields
9681 @opindex mno-wide-bitfields
9682 Always treat bit-fields as int-sized.
9684 @item -m4byte-functions
9685 @itemx -mno-4byte-functions
9686 @opindex m4byte-functions
9687 @opindex mno-4byte-functions
9688 Force all functions to be aligned to a four byte boundary.
9690 @item -mcallgraph-data
9691 @itemx -mno-callgraph-data
9692 @opindex mcallgraph-data
9693 @opindex mno-callgraph-data
9694 Emit callgraph information.
9697 @itemx -mno-slow-bytes
9698 @opindex mslow-bytes
9699 @opindex mno-slow-bytes
9700 Prefer word access when reading byte quantities.
9702 @item -mlittle-endian
9704 @opindex mlittle-endian
9705 @opindex mbig-endian
9706 Generate code for a little endian target.
9712 Generate code for the 210 processor.
9716 @subsection MIPS Options
9717 @cindex MIPS options
9723 Generate big-endian code.
9727 Generate little-endian code. This is the default for @samp{mips*el-*-*}
9730 @item -march=@var{arch}
9732 Generate code that will run on @var{arch}, which can be the name of a
9733 generic MIPS ISA, or the name of a particular processor.
9735 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
9736 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
9737 The processor names are:
9738 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
9740 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
9741 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
9745 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
9746 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
9747 The special value @samp{from-abi} selects the
9748 most compatible architecture for the selected ABI (that is,
9749 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
9751 In processor names, a final @samp{000} can be abbreviated as @samp{k}
9752 (for example, @samp{-march=r2k}). Prefixes are optional, and
9753 @samp{vr} may be written @samp{r}.
9755 GCC defines two macros based on the value of this option. The first
9756 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
9757 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
9758 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
9759 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
9760 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
9762 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
9763 above. In other words, it will have the full prefix and will not
9764 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
9765 the macro names the resolved architecture (either @samp{"mips1"} or
9766 @samp{"mips3"}). It names the default architecture when no
9767 @option{-march} option is given.
9769 @item -mtune=@var{arch}
9771 Optimize for @var{arch}. Among other things, this option controls
9772 the way instructions are scheduled, and the perceived cost of arithmetic
9773 operations. The list of @var{arch} values is the same as for
9776 When this option is not used, GCC will optimize for the processor
9777 specified by @option{-march}. By using @option{-march} and
9778 @option{-mtune} together, it is possible to generate code that will
9779 run on a family of processors, but optimize the code for one
9780 particular member of that family.
9782 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
9783 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
9784 @samp{-march} ones described above.
9788 Equivalent to @samp{-march=mips1}.
9792 Equivalent to @samp{-march=mips2}.
9796 Equivalent to @samp{-march=mips3}.
9800 Equivalent to @samp{-march=mips4}.
9804 Equivalent to @samp{-march=mips32}.
9808 Equivalent to @samp{-march=mips32r2}.
9812 Equivalent to @samp{-march=mips64}.
9818 Use (do not use) the MIPS16 ISA@.
9830 Generate code for the given ABI@.
9832 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
9833 generates 64-bit code when you select a 64-bit architecture, but you
9834 can use @option{-mgp32} to get 32-bit code instead.
9836 For information about the O64 ABI, see
9837 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
9840 @itemx -mno-abicalls
9842 @opindex mno-abicalls
9843 Generate (do not generate) SVR4-style position-independent code.
9844 @option{-mabicalls} is the default for SVR4-based systems.
9850 Lift (do not lift) the usual restrictions on the size of the global
9853 GCC normally uses a single instruction to load values from the GOT@.
9854 While this is relatively efficient, it will only work if the GOT
9855 is smaller than about 64k. Anything larger will cause the linker
9856 to report an error such as:
9858 @cindex relocation truncated to fit (MIPS)
9860 relocation truncated to fit: R_MIPS_GOT16 foobar
9863 If this happens, you should recompile your code with @option{-mxgot}.
9864 It should then work with very large GOTs, although it will also be
9865 less efficient, since it will take three instructions to fetch the
9866 value of a global symbol.
9868 Note that some linkers can create multiple GOTs. If you have such a
9869 linker, you should only need to use @option{-mxgot} when a single object
9870 file accesses more than 64k's worth of GOT entries. Very few do.
9872 These options have no effect unless GCC is generating position
9877 Assume that general-purpose registers are 32 bits wide.
9881 Assume that general-purpose registers are 64 bits wide.
9885 Assume that floating-point registers are 32 bits wide.
9889 Assume that floating-point registers are 64 bits wide.
9892 @opindex mhard-float
9893 Use floating-point coprocessor instructions.
9896 @opindex msoft-float
9897 Do not use floating-point coprocessor instructions. Implement
9898 floating-point calculations using library calls instead.
9900 @item -msingle-float
9901 @opindex msingle-float
9902 Assume that the floating-point coprocessor only supports single-precision
9905 @itemx -mdouble-float
9906 @opindex mdouble-float
9907 Assume that the floating-point coprocessor supports double-precision
9908 operations. This is the default.
9910 @itemx -mpaired-single
9911 @itemx -mno-paired-single
9912 @opindex mpaired-single
9913 @opindex mno-paired-single
9914 Use (do not use) paired-single floating-point instructions.
9915 @xref{MIPS Paired-Single Support}. This option can only be used
9916 when generating 64-bit code and requires hardware floating-point
9917 support to be enabled.
9923 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
9924 The option @option{-mips3d} implies @option{-mpaired-single}.
9928 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
9929 an explanation of the default and the way that the pointer size is
9934 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
9936 The default size of @code{int}s, @code{long}s and pointers depends on
9937 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
9938 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
9939 32-bit @code{long}s. Pointers are the same size as @code{long}s,
9940 or the same size as integer registers, whichever is smaller.
9946 Assume (do not assume) that all symbols have 32-bit values, regardless
9947 of the selected ABI@. This option is useful in combination with
9948 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
9949 to generate shorter and faster references to symbolic addresses.
9953 @cindex smaller data references (MIPS)
9954 @cindex gp-relative references (MIPS)
9955 Put global and static items less than or equal to @var{num} bytes into
9956 the small data or bss section instead of the normal data or bss section.
9957 This allows the data to be accessed using a single instruction.
9959 All modules should be compiled with the same @option{-G @var{num}}
9962 @item -membedded-data
9963 @itemx -mno-embedded-data
9964 @opindex membedded-data
9965 @opindex mno-embedded-data
9966 Allocate variables to the read-only data section first if possible, then
9967 next in the small data section if possible, otherwise in data. This gives
9968 slightly slower code than the default, but reduces the amount of RAM required
9969 when executing, and thus may be preferred for some embedded systems.
9971 @item -muninit-const-in-rodata
9972 @itemx -mno-uninit-const-in-rodata
9973 @opindex muninit-const-in-rodata
9974 @opindex mno-uninit-const-in-rodata
9975 Put uninitialized @code{const} variables in the read-only data section.
9976 This option is only meaningful in conjunction with @option{-membedded-data}.
9978 @item -msplit-addresses
9979 @itemx -mno-split-addresses
9980 @opindex msplit-addresses
9981 @opindex mno-split-addresses
9982 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
9983 relocation operators. This option has been superseded by
9984 @option{-mexplicit-relocs} but is retained for backwards compatibility.
9986 @item -mexplicit-relocs
9987 @itemx -mno-explicit-relocs
9988 @opindex mexplicit-relocs
9989 @opindex mno-explicit-relocs
9990 Use (do not use) assembler relocation operators when dealing with symbolic
9991 addresses. The alternative, selected by @option{-mno-explicit-relocs},
9992 is to use assembler macros instead.
9994 @option{-mexplicit-relocs} is the default if GCC was configured
9995 to use an assembler that supports relocation operators.
9997 @item -mcheck-zero-division
9998 @itemx -mno-check-zero-division
9999 @opindex mcheck-zero-division
10000 @opindex mno-check-zero-division
10001 Trap (do not trap) on integer division by zero. The default is
10002 @option{-mcheck-zero-division}.
10004 @item -mdivide-traps
10005 @itemx -mdivide-breaks
10006 @opindex mdivide-traps
10007 @opindex mdivide-breaks
10008 MIPS systems check for division by zero by generating either a
10009 conditional trap or a break instruction. Using traps results in
10010 smaller code, but is only supported on MIPS II and later. Also, some
10011 versions of the Linux kernel have a bug that prevents trap from
10012 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
10013 allow conditional traps on architectures that support them and
10014 @option{-mdivide-breaks} to force the use of breaks.
10016 The default is usually @option{-mdivide-traps}, but this can be
10017 overridden at configure time using @option{--with-divide=breaks}.
10018 Divide-by-zero checks can be completely disabled using
10019 @option{-mno-check-zero-division}.
10024 @opindex mno-memcpy
10025 Force (do not force) the use of @code{memcpy()} for non-trivial block
10026 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
10027 most constant-sized copies.
10030 @itemx -mno-long-calls
10031 @opindex mlong-calls
10032 @opindex mno-long-calls
10033 Disable (do not disable) use of the @code{jal} instruction. Calling
10034 functions using @code{jal} is more efficient but requires the caller
10035 and callee to be in the same 256 megabyte segment.
10037 This option has no effect on abicalls code. The default is
10038 @option{-mno-long-calls}.
10044 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
10045 instructions, as provided by the R4650 ISA@.
10048 @itemx -mno-fused-madd
10049 @opindex mfused-madd
10050 @opindex mno-fused-madd
10051 Enable (disable) use of the floating point multiply-accumulate
10052 instructions, when they are available. The default is
10053 @option{-mfused-madd}.
10055 When multiply-accumulate instructions are used, the intermediate
10056 product is calculated to infinite precision and is not subject to
10057 the FCSR Flush to Zero bit. This may be undesirable in some
10062 Tell the MIPS assembler to not run its preprocessor over user
10063 assembler files (with a @samp{.s} suffix) when assembling them.
10066 @itemx -mno-fix-r4000
10067 @opindex mfix-r4000
10068 @opindex mno-fix-r4000
10069 Work around certain R4000 CPU errata:
10072 A double-word or a variable shift may give an incorrect result if executed
10073 immediately after starting an integer division.
10075 A double-word or a variable shift may give an incorrect result if executed
10076 while an integer multiplication is in progress.
10078 An integer division may give an incorrect result if started in a delay slot
10079 of a taken branch or a jump.
10083 @itemx -mno-fix-r4400
10084 @opindex mfix-r4400
10085 @opindex mno-fix-r4400
10086 Work around certain R4400 CPU errata:
10089 A double-word or a variable shift may give an incorrect result if executed
10090 immediately after starting an integer division.
10094 @itemx -mno-fix-vr4120
10095 @opindex mfix-vr4120
10096 Work around certain VR4120 errata:
10099 @code{dmultu} does not always produce the correct result.
10101 @code{div} and @code{ddiv} do not always produce the correct result if one
10102 of the operands is negative.
10104 The workarounds for the division errata rely on special functions in
10105 @file{libgcc.a}. At present, these functions are only provided by
10106 the @code{mips64vr*-elf} configurations.
10108 Other VR4120 errata require a nop to be inserted between certain pairs of
10109 instructions. These errata are handled by the assembler, not by GCC itself.
10112 @opindex mfix-vr4130
10113 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
10114 workarounds are implemented by the assembler rather than by GCC,
10115 although GCC will avoid using @code{mflo} and @code{mfhi} if the
10116 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
10117 instructions are available instead.
10120 @itemx -mno-fix-sb1
10122 Work around certain SB-1 CPU core errata.
10123 (This flag currently works around the SB-1 revision 2
10124 ``F1'' and ``F2'' floating point errata.)
10126 @item -mflush-func=@var{func}
10127 @itemx -mno-flush-func
10128 @opindex mflush-func
10129 Specifies the function to call to flush the I and D caches, or to not
10130 call any such function. If called, the function must take the same
10131 arguments as the common @code{_flush_func()}, that is, the address of the
10132 memory range for which the cache is being flushed, the size of the
10133 memory range, and the number 3 (to flush both caches). The default
10134 depends on the target GCC was configured for, but commonly is either
10135 @samp{_flush_func} or @samp{__cpu_flush}.
10137 @item -mbranch-likely
10138 @itemx -mno-branch-likely
10139 @opindex mbranch-likely
10140 @opindex mno-branch-likely
10141 Enable or disable use of Branch Likely instructions, regardless of the
10142 default for the selected architecture. By default, Branch Likely
10143 instructions may be generated if they are supported by the selected
10144 architecture. An exception is for the MIPS32 and MIPS64 architectures
10145 and processors which implement those architectures; for those, Branch
10146 Likely instructions will not be generated by default because the MIPS32
10147 and MIPS64 architectures specifically deprecate their use.
10149 @item -mfp-exceptions
10150 @itemx -mno-fp-exceptions
10151 @opindex mfp-exceptions
10152 Specifies whether FP exceptions are enabled. This affects how we schedule
10153 FP instructions for some processors. The default is that FP exceptions are
10156 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
10157 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
10160 @item -mvr4130-align
10161 @itemx -mno-vr4130-align
10162 @opindex mvr4130-align
10163 The VR4130 pipeline is two-way superscalar, but can only issue two
10164 instructions together if the first one is 8-byte aligned. When this
10165 option is enabled, GCC will align pairs of instructions that it
10166 thinks should execute in parallel.
10168 This option only has an effect when optimizing for the VR4130.
10169 It normally makes code faster, but at the expense of making it bigger.
10170 It is enabled by default at optimization level @option{-O3}.
10174 @subsection MMIX Options
10175 @cindex MMIX Options
10177 These options are defined for the MMIX:
10181 @itemx -mno-libfuncs
10183 @opindex mno-libfuncs
10184 Specify that intrinsic library functions are being compiled, passing all
10185 values in registers, no matter the size.
10188 @itemx -mno-epsilon
10190 @opindex mno-epsilon
10191 Generate floating-point comparison instructions that compare with respect
10192 to the @code{rE} epsilon register.
10194 @item -mabi=mmixware
10196 @opindex mabi-mmixware
10198 Generate code that passes function parameters and return values that (in
10199 the called function) are seen as registers @code{$0} and up, as opposed to
10200 the GNU ABI which uses global registers @code{$231} and up.
10202 @item -mzero-extend
10203 @itemx -mno-zero-extend
10204 @opindex mzero-extend
10205 @opindex mno-zero-extend
10206 When reading data from memory in sizes shorter than 64 bits, use (do not
10207 use) zero-extending load instructions by default, rather than
10208 sign-extending ones.
10211 @itemx -mno-knuthdiv
10213 @opindex mno-knuthdiv
10214 Make the result of a division yielding a remainder have the same sign as
10215 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10216 remainder follows the sign of the dividend. Both methods are
10217 arithmetically valid, the latter being almost exclusively used.
10219 @item -mtoplevel-symbols
10220 @itemx -mno-toplevel-symbols
10221 @opindex mtoplevel-symbols
10222 @opindex mno-toplevel-symbols
10223 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10224 code can be used with the @code{PREFIX} assembly directive.
10228 Generate an executable in the ELF format, rather than the default
10229 @samp{mmo} format used by the @command{mmix} simulator.
10231 @item -mbranch-predict
10232 @itemx -mno-branch-predict
10233 @opindex mbranch-predict
10234 @opindex mno-branch-predict
10235 Use (do not use) the probable-branch instructions, when static branch
10236 prediction indicates a probable branch.
10238 @item -mbase-addresses
10239 @itemx -mno-base-addresses
10240 @opindex mbase-addresses
10241 @opindex mno-base-addresses
10242 Generate (do not generate) code that uses @emph{base addresses}. Using a
10243 base address automatically generates a request (handled by the assembler
10244 and the linker) for a constant to be set up in a global register. The
10245 register is used for one or more base address requests within the range 0
10246 to 255 from the value held in the register. The generally leads to short
10247 and fast code, but the number of different data items that can be
10248 addressed is limited. This means that a program that uses lots of static
10249 data may require @option{-mno-base-addresses}.
10251 @item -msingle-exit
10252 @itemx -mno-single-exit
10253 @opindex msingle-exit
10254 @opindex mno-single-exit
10255 Force (do not force) generated code to have a single exit point in each
10259 @node MN10300 Options
10260 @subsection MN10300 Options
10261 @cindex MN10300 options
10263 These @option{-m} options are defined for Matsushita MN10300 architectures:
10268 Generate code to avoid bugs in the multiply instructions for the MN10300
10269 processors. This is the default.
10271 @item -mno-mult-bug
10272 @opindex mno-mult-bug
10273 Do not generate code to avoid bugs in the multiply instructions for the
10274 MN10300 processors.
10278 Generate code which uses features specific to the AM33 processor.
10282 Do not generate code which uses features specific to the AM33 processor. This
10287 Do not link in the C run-time initialization object file.
10291 Indicate to the linker that it should perform a relaxation optimization pass
10292 to shorten branches, calls and absolute memory addresses. This option only
10293 has an effect when used on the command line for the final link step.
10295 This option makes symbolic debugging impossible.
10298 @node NS32K Options
10299 @subsection NS32K Options
10300 @cindex NS32K options
10302 These are the @samp{-m} options defined for the 32000 series. The default
10303 values for these options depends on which style of 32000 was selected when
10304 the compiler was configured; the defaults for the most common choices are
10312 Generate output for a 32032. This is the default
10313 when the compiler is configured for 32032 and 32016 based systems.
10319 Generate output for a 32332. This is the default
10320 when the compiler is configured for 32332-based systems.
10326 Generate output for a 32532. This is the default
10327 when the compiler is configured for 32532-based systems.
10331 Generate output containing 32081 instructions for floating point.
10332 This is the default for all systems.
10336 Generate output containing 32381 instructions for floating point. This
10337 also implies @option{-m32081}. The 32381 is only compatible with the 32332
10338 and 32532 cpus. This is the default for the pc532-netbsd configuration.
10341 @opindex mmulti-add
10342 Try and generate multiply-add floating point instructions @code{polyF}
10343 and @code{dotF}. This option is only available if the @option{-m32381}
10344 option is in effect. Using these instructions requires changes to
10345 register allocation which generally has a negative impact on
10346 performance. This option should only be enabled when compiling code
10347 particularly likely to make heavy use of multiply-add instructions.
10349 @item -mnomulti-add
10350 @opindex mnomulti-add
10351 Do not try and generate multiply-add floating point instructions
10352 @code{polyF} and @code{dotF}. This is the default on all platforms.
10355 @opindex msoft-float
10356 Generate output containing library calls for floating point.
10357 @strong{Warning:} the requisite libraries may not be available.
10359 @item -mieee-compare
10360 @itemx -mno-ieee-compare
10361 @opindex mieee-compare
10362 @opindex mno-ieee-compare
10363 Control whether or not the compiler uses IEEE floating point
10364 comparisons. These handle correctly the case where the result of a
10365 comparison is unordered.
10366 @strong{Warning:} the requisite kernel support may not be available.
10369 @opindex mnobitfield
10370 Do not use the bit-field instructions. On some machines it is faster to
10371 use shifting and masking operations. This is the default for the pc532.
10375 Do use the bit-field instructions. This is the default for all platforms
10380 Use a different function-calling convention, in which functions
10381 that take a fixed number of arguments return pop their
10382 arguments on return with the @code{ret} instruction.
10384 This calling convention is incompatible with the one normally
10385 used on Unix, so you cannot use it if you need to call libraries
10386 compiled with the Unix compiler.
10388 Also, you must provide function prototypes for all functions that
10389 take variable numbers of arguments (including @code{printf});
10390 otherwise incorrect code will be generated for calls to those
10393 In addition, seriously incorrect code will result if you call a
10394 function with too many arguments. (Normally, extra arguments are
10395 harmlessly ignored.)
10397 This option takes its name from the 680x0 @code{rtd} instruction.
10402 Use a different function-calling convention where the first two arguments
10403 are passed in registers.
10405 This calling convention is incompatible with the one normally
10406 used on Unix, so you cannot use it if you need to call libraries
10407 compiled with the Unix compiler.
10410 @opindex mnoregparam
10411 Do not pass any arguments in registers. This is the default for all
10416 It is OK to use the sb as an index register which is always loaded with
10417 zero. This is the default for the pc532-netbsd target.
10421 The sb register is not available for use or has not been initialized to
10422 zero by the run time system. This is the default for all targets except
10423 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
10424 @option{-fpic} is set.
10428 Many ns32000 series addressing modes use displacements of up to 512MB@.
10429 If an address is above 512MB then displacements from zero can not be used.
10430 This option causes code to be generated which can be loaded above 512MB@.
10431 This may be useful for operating systems or ROM code.
10435 Assume code will be loaded in the first 512MB of virtual address space.
10436 This is the default for all platforms.
10440 @node PDP-11 Options
10441 @subsection PDP-11 Options
10442 @cindex PDP-11 Options
10444 These options are defined for the PDP-11:
10449 Use hardware FPP floating point. This is the default. (FIS floating
10450 point on the PDP-11/40 is not supported.)
10453 @opindex msoft-float
10454 Do not use hardware floating point.
10458 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10462 Return floating-point results in memory. This is the default.
10466 Generate code for a PDP-11/40.
10470 Generate code for a PDP-11/45. This is the default.
10474 Generate code for a PDP-11/10.
10476 @item -mbcopy-builtin
10477 @opindex bcopy-builtin
10478 Use inline @code{movmemhi} patterns for copying memory. This is the
10483 Do not use inline @code{movmemhi} patterns for copying memory.
10489 Use 16-bit @code{int}. This is the default.
10495 Use 32-bit @code{int}.
10498 @itemx -mno-float32
10500 @opindex mno-float32
10501 Use 64-bit @code{float}. This is the default.
10504 @itemx -mno-float64
10506 @opindex mno-float64
10507 Use 32-bit @code{float}.
10511 Use @code{abshi2} pattern. This is the default.
10515 Do not use @code{abshi2} pattern.
10517 @item -mbranch-expensive
10518 @opindex mbranch-expensive
10519 Pretend that branches are expensive. This is for experimenting with
10520 code generation only.
10522 @item -mbranch-cheap
10523 @opindex mbranch-cheap
10524 Do not pretend that branches are expensive. This is the default.
10528 Generate code for a system with split I&D@.
10532 Generate code for a system without split I&D@. This is the default.
10536 Use Unix assembler syntax. This is the default when configured for
10537 @samp{pdp11-*-bsd}.
10541 Use DEC assembler syntax. This is the default when configured for any
10542 PDP-11 target other than @samp{pdp11-*-bsd}.
10545 @node PowerPC Options
10546 @subsection PowerPC Options
10547 @cindex PowerPC options
10549 These are listed under @xref{RS/6000 and PowerPC Options}.
10551 @node RS/6000 and PowerPC Options
10552 @subsection IBM RS/6000 and PowerPC Options
10553 @cindex RS/6000 and PowerPC Options
10554 @cindex IBM RS/6000 and PowerPC Options
10556 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
10563 @itemx -mno-powerpc
10564 @itemx -mpowerpc-gpopt
10565 @itemx -mno-powerpc-gpopt
10566 @itemx -mpowerpc-gfxopt
10567 @itemx -mno-powerpc-gfxopt
10569 @itemx -mno-powerpc64
10573 @opindex mno-power2
10575 @opindex mno-powerpc
10576 @opindex mpowerpc-gpopt
10577 @opindex mno-powerpc-gpopt
10578 @opindex mpowerpc-gfxopt
10579 @opindex mno-powerpc-gfxopt
10580 @opindex mpowerpc64
10581 @opindex mno-powerpc64
10582 GCC supports two related instruction set architectures for the
10583 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
10584 instructions supported by the @samp{rios} chip set used in the original
10585 RS/6000 systems and the @dfn{PowerPC} instruction set is the
10586 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
10587 the IBM 4xx microprocessors.
10589 Neither architecture is a subset of the other. However there is a
10590 large common subset of instructions supported by both. An MQ
10591 register is included in processors supporting the POWER architecture.
10593 You use these options to specify which instructions are available on the
10594 processor you are using. The default value of these options is
10595 determined when configuring GCC@. Specifying the
10596 @option{-mcpu=@var{cpu_type}} overrides the specification of these
10597 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
10598 rather than the options listed above.
10600 The @option{-mpower} option allows GCC to generate instructions that
10601 are found only in the POWER architecture and to use the MQ register.
10602 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
10603 to generate instructions that are present in the POWER2 architecture but
10604 not the original POWER architecture.
10606 The @option{-mpowerpc} option allows GCC to generate instructions that
10607 are found only in the 32-bit subset of the PowerPC architecture.
10608 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
10609 GCC to use the optional PowerPC architecture instructions in the
10610 General Purpose group, including floating-point square root. Specifying
10611 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
10612 use the optional PowerPC architecture instructions in the Graphics
10613 group, including floating-point select.
10615 The @option{-mpowerpc64} option allows GCC to generate the additional
10616 64-bit instructions that are found in the full PowerPC64 architecture
10617 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
10618 @option{-mno-powerpc64}.
10620 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
10621 will use only the instructions in the common subset of both
10622 architectures plus some special AIX common-mode calls, and will not use
10623 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
10624 permits GCC to use any instruction from either architecture and to
10625 allow use of the MQ register; specify this for the Motorola MPC601.
10627 @item -mnew-mnemonics
10628 @itemx -mold-mnemonics
10629 @opindex mnew-mnemonics
10630 @opindex mold-mnemonics
10631 Select which mnemonics to use in the generated assembler code. With
10632 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
10633 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
10634 assembler mnemonics defined for the POWER architecture. Instructions
10635 defined in only one architecture have only one mnemonic; GCC uses that
10636 mnemonic irrespective of which of these options is specified.
10638 GCC defaults to the mnemonics appropriate for the architecture in
10639 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
10640 value of these option. Unless you are building a cross-compiler, you
10641 should normally not specify either @option{-mnew-mnemonics} or
10642 @option{-mold-mnemonics}, but should instead accept the default.
10644 @item -mcpu=@var{cpu_type}
10646 Set architecture type, register usage, choice of mnemonics, and
10647 instruction scheduling parameters for machine type @var{cpu_type}.
10648 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
10649 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
10650 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
10651 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
10652 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
10653 @samp{860}, @samp{970}, @samp{8540}, @samp{common}, @samp{ec603e}, @samp{G3},
10654 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
10655 @samp{power4}, @samp{power5}, @samp{powerpc}, @samp{powerpc64},
10656 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
10658 @option{-mcpu=common} selects a completely generic processor. Code
10659 generated under this option will run on any POWER or PowerPC processor.
10660 GCC will use only the instructions in the common subset of both
10661 architectures, and will not use the MQ register. GCC assumes a generic
10662 processor model for scheduling purposes.
10664 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
10665 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
10666 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
10667 types, with an appropriate, generic processor model assumed for
10668 scheduling purposes.
10670 The other options specify a specific processor. Code generated under
10671 those options will run best on that processor, and may not run at all on
10674 The @option{-mcpu} options automatically enable or disable the
10675 following options: @option{-maltivec}, @option{-mhard-float},
10676 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
10677 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
10678 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
10679 @option{-mstring}. The particular options set for any particular CPU
10680 will vary between compiler versions, depending on what setting seems
10681 to produce optimal code for that CPU; it doesn't necessarily reflect
10682 the actual hardware's capabilities. If you wish to set an individual
10683 option to a particular value, you may specify it after the
10684 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
10686 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
10687 not enabled or disabled by the @option{-mcpu} option at present, since
10688 AIX does not have full support for these options. You may still
10689 enable or disable them individually if you're sure it'll work in your
10692 @item -mtune=@var{cpu_type}
10694 Set the instruction scheduling parameters for machine type
10695 @var{cpu_type}, but do not set the architecture type, register usage, or
10696 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
10697 values for @var{cpu_type} are used for @option{-mtune} as for
10698 @option{-mcpu}. If both are specified, the code generated will use the
10699 architecture, registers, and mnemonics set by @option{-mcpu}, but the
10700 scheduling parameters set by @option{-mtune}.
10703 @itemx -mno-altivec
10705 @opindex mno-altivec
10706 Generate code that uses (does not use) AltiVec instructions, and also
10707 enable the use of built-in functions that allow more direct access to
10708 the AltiVec instruction set. You may also need to set
10709 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
10715 @opindex mno-vrsave
10716 Generate VRSAVE instructions when generating AltiVec code.
10720 Extend the current ABI with SPE ABI extensions. This does not change
10721 the default ABI, instead it adds the SPE ABI extensions to the current
10725 @opindex mabi=no-spe
10726 Disable Booke SPE ABI extensions for the current ABI@.
10732 This switch enables or disables the generation of ISEL instructions.
10734 @item -misel=@var{yes/no}
10735 This switch has been deprecated. Use @option{-misel} and
10736 @option{-mno-isel} instead.
10742 This switch enables or disables the generation of SPE simd
10745 @item -mspe=@var{yes/no}
10746 This option has been deprecated. Use @option{-mspe} and
10747 @option{-mno-spe} instead.
10749 @item -mfloat-gprs=@var{yes/single/double/no}
10750 @itemx -mfloat-gprs
10751 @opindex mfloat-gprs
10752 This switch enables or disables the generation of floating point
10753 operations on the general purpose registers for architectures that
10756 The argument @var{yes} or @var{single} enables the use of
10757 single-precision floating point operations.
10759 The argument @var{double} enables the use of single and
10760 double-precision floating point operations.
10762 The argument @var{no} disables floating point operations on the
10763 general purpose registers.
10765 This option is currently only available on the MPC854x.
10771 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
10772 targets (including GNU/Linux). The 32-bit environment sets int, long
10773 and pointer to 32 bits and generates code that runs on any PowerPC
10774 variant. The 64-bit environment sets int to 32 bits and long and
10775 pointer to 64 bits, and generates code for PowerPC64, as for
10776 @option{-mpowerpc64}.
10779 @itemx -mno-fp-in-toc
10780 @itemx -mno-sum-in-toc
10781 @itemx -mminimal-toc
10783 @opindex mno-fp-in-toc
10784 @opindex mno-sum-in-toc
10785 @opindex mminimal-toc
10786 Modify generation of the TOC (Table Of Contents), which is created for
10787 every executable file. The @option{-mfull-toc} option is selected by
10788 default. In that case, GCC will allocate at least one TOC entry for
10789 each unique non-automatic variable reference in your program. GCC
10790 will also place floating-point constants in the TOC@. However, only
10791 16,384 entries are available in the TOC@.
10793 If you receive a linker error message that saying you have overflowed
10794 the available TOC space, you can reduce the amount of TOC space used
10795 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
10796 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
10797 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
10798 generate code to calculate the sum of an address and a constant at
10799 run-time instead of putting that sum into the TOC@. You may specify one
10800 or both of these options. Each causes GCC to produce very slightly
10801 slower and larger code at the expense of conserving TOC space.
10803 If you still run out of space in the TOC even when you specify both of
10804 these options, specify @option{-mminimal-toc} instead. This option causes
10805 GCC to make only one TOC entry for every file. When you specify this
10806 option, GCC will produce code that is slower and larger but which
10807 uses extremely little TOC space. You may wish to use this option
10808 only on files that contain less frequently executed code.
10814 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
10815 @code{long} type, and the infrastructure needed to support them.
10816 Specifying @option{-maix64} implies @option{-mpowerpc64} and
10817 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
10818 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
10821 @itemx -mno-xl-compat
10822 @opindex mxl-compat
10823 @opindex mno-xl-compat
10824 Produce code that conforms more closely to IBM XLC semantics when using
10825 AIX-compatible ABI. Pass floating-point arguments to prototyped
10826 functions beyond the register save area (RSA) on the stack in addition
10827 to argument FPRs. Do not assume that most significant double in 128
10828 bit long double value is properly rounded when comparing values.
10830 The AIX calling convention was extended but not initially documented to
10831 handle an obscure K&R C case of calling a function that takes the
10832 address of its arguments with fewer arguments than declared. AIX XL
10833 compilers access floating point arguments which do not fit in the
10834 RSA from the stack when a subroutine is compiled without
10835 optimization. Because always storing floating-point arguments on the
10836 stack is inefficient and rarely needed, this option is not enabled by
10837 default and only is necessary when calling subroutines compiled by AIX
10838 XL compilers without optimization.
10842 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
10843 application written to use message passing with special startup code to
10844 enable the application to run. The system must have PE installed in the
10845 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
10846 must be overridden with the @option{-specs=} option to specify the
10847 appropriate directory location. The Parallel Environment does not
10848 support threads, so the @option{-mpe} option and the @option{-pthread}
10849 option are incompatible.
10851 @item -malign-natural
10852 @itemx -malign-power
10853 @opindex malign-natural
10854 @opindex malign-power
10855 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
10856 @option{-malign-natural} overrides the ABI-defined alignment of larger
10857 types, such as floating-point doubles, on their natural size-based boundary.
10858 The option @option{-malign-power} instructs GCC to follow the ABI-specified
10859 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
10861 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
10865 @itemx -mhard-float
10866 @opindex msoft-float
10867 @opindex mhard-float
10868 Generate code that does not use (uses) the floating-point register set.
10869 Software floating point emulation is provided if you use the
10870 @option{-msoft-float} option, and pass the option to GCC when linking.
10873 @itemx -mno-multiple
10875 @opindex mno-multiple
10876 Generate code that uses (does not use) the load multiple word
10877 instructions and the store multiple word instructions. These
10878 instructions are generated by default on POWER systems, and not
10879 generated on PowerPC systems. Do not use @option{-mmultiple} on little
10880 endian PowerPC systems, since those instructions do not work when the
10881 processor is in little endian mode. The exceptions are PPC740 and
10882 PPC750 which permit the instructions usage in little endian mode.
10887 @opindex mno-string
10888 Generate code that uses (does not use) the load string instructions
10889 and the store string word instructions to save multiple registers and
10890 do small block moves. These instructions are generated by default on
10891 POWER systems, and not generated on PowerPC systems. Do not use
10892 @option{-mstring} on little endian PowerPC systems, since those
10893 instructions do not work when the processor is in little endian mode.
10894 The exceptions are PPC740 and PPC750 which permit the instructions
10895 usage in little endian mode.
10900 @opindex mno-update
10901 Generate code that uses (does not use) the load or store instructions
10902 that update the base register to the address of the calculated memory
10903 location. These instructions are generated by default. If you use
10904 @option{-mno-update}, there is a small window between the time that the
10905 stack pointer is updated and the address of the previous frame is
10906 stored, which means code that walks the stack frame across interrupts or
10907 signals may get corrupted data.
10910 @itemx -mno-fused-madd
10911 @opindex mfused-madd
10912 @opindex mno-fused-madd
10913 Generate code that uses (does not use) the floating point multiply and
10914 accumulate instructions. These instructions are generated by default if
10915 hardware floating is used.
10917 @item -mno-bit-align
10919 @opindex mno-bit-align
10920 @opindex mbit-align
10921 On System V.4 and embedded PowerPC systems do not (do) force structures
10922 and unions that contain bit-fields to be aligned to the base type of the
10925 For example, by default a structure containing nothing but 8
10926 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
10927 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
10928 the structure would be aligned to a 1 byte boundary and be one byte in
10931 @item -mno-strict-align
10932 @itemx -mstrict-align
10933 @opindex mno-strict-align
10934 @opindex mstrict-align
10935 On System V.4 and embedded PowerPC systems do not (do) assume that
10936 unaligned memory references will be handled by the system.
10938 @item -mrelocatable
10939 @itemx -mno-relocatable
10940 @opindex mrelocatable
10941 @opindex mno-relocatable
10942 On embedded PowerPC systems generate code that allows (does not allow)
10943 the program to be relocated to a different address at runtime. If you
10944 use @option{-mrelocatable} on any module, all objects linked together must
10945 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
10947 @item -mrelocatable-lib
10948 @itemx -mno-relocatable-lib
10949 @opindex mrelocatable-lib
10950 @opindex mno-relocatable-lib
10951 On embedded PowerPC systems generate code that allows (does not allow)
10952 the program to be relocated to a different address at runtime. Modules
10953 compiled with @option{-mrelocatable-lib} can be linked with either modules
10954 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
10955 with modules compiled with the @option{-mrelocatable} options.
10961 On System V.4 and embedded PowerPC systems do not (do) assume that
10962 register 2 contains a pointer to a global area pointing to the addresses
10963 used in the program.
10966 @itemx -mlittle-endian
10968 @opindex mlittle-endian
10969 On System V.4 and embedded PowerPC systems compile code for the
10970 processor in little endian mode. The @option{-mlittle-endian} option is
10971 the same as @option{-mlittle}.
10974 @itemx -mbig-endian
10976 @opindex mbig-endian
10977 On System V.4 and embedded PowerPC systems compile code for the
10978 processor in big endian mode. The @option{-mbig-endian} option is
10979 the same as @option{-mbig}.
10981 @item -mdynamic-no-pic
10982 @opindex mdynamic-no-pic
10983 On Darwin and Mac OS X systems, compile code so that it is not
10984 relocatable, but that its external references are relocatable. The
10985 resulting code is suitable for applications, but not shared
10988 @item -mprioritize-restricted-insns=@var{priority}
10989 @opindex mprioritize-restricted-insns
10990 This option controls the priority that is assigned to
10991 dispatch-slot restricted instructions during the second scheduling
10992 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
10993 @var{no/highest/second-highest} priority to dispatch slot restricted
10996 @item -msched-costly-dep=@var{dependence_type}
10997 @opindex msched-costly-dep
10998 This option controls which dependences are considered costly
10999 by the target during instruction scheduling. The argument
11000 @var{dependence_type} takes one of the following values:
11001 @var{no}: no dependence is costly,
11002 @var{all}: all dependences are costly,
11003 @var{true_store_to_load}: a true dependence from store to load is costly,
11004 @var{store_to_load}: any dependence from store to load is costly,
11005 @var{number}: any dependence which latency >= @var{number} is costly.
11007 @item -minsert-sched-nops=@var{scheme}
11008 @opindex minsert-sched-nops
11009 This option controls which nop insertion scheme will be used during
11010 the second scheduling pass. The argument @var{scheme} takes one of the
11012 @var{no}: Don't insert nops.
11013 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
11014 according to the scheduler's grouping.
11015 @var{regroup_exact}: Insert nops to force costly dependent insns into
11016 separate groups. Insert exactly as many nops as needed to force an insn
11017 to a new group, according to the estimated processor grouping.
11018 @var{number}: Insert nops to force costly dependent insns into
11019 separate groups. Insert @var{number} nops to force an insn to a new group.
11022 @opindex mcall-sysv
11023 On System V.4 and embedded PowerPC systems compile code using calling
11024 conventions that adheres to the March 1995 draft of the System V
11025 Application Binary Interface, PowerPC processor supplement. This is the
11026 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
11028 @item -mcall-sysv-eabi
11029 @opindex mcall-sysv-eabi
11030 Specify both @option{-mcall-sysv} and @option{-meabi} options.
11032 @item -mcall-sysv-noeabi
11033 @opindex mcall-sysv-noeabi
11034 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
11036 @item -mcall-solaris
11037 @opindex mcall-solaris
11038 On System V.4 and embedded PowerPC systems compile code for the Solaris
11042 @opindex mcall-linux
11043 On System V.4 and embedded PowerPC systems compile code for the
11044 Linux-based GNU system.
11048 On System V.4 and embedded PowerPC systems compile code for the
11049 Hurd-based GNU system.
11051 @item -mcall-netbsd
11052 @opindex mcall-netbsd
11053 On System V.4 and embedded PowerPC systems compile code for the
11054 NetBSD operating system.
11056 @item -maix-struct-return
11057 @opindex maix-struct-return
11058 Return all structures in memory (as specified by the AIX ABI)@.
11060 @item -msvr4-struct-return
11061 @opindex msvr4-struct-return
11062 Return structures smaller than 8 bytes in registers (as specified by the
11065 @item -mabi=@var{abi-type}
11067 Extend the current ABI with a particular extension, or remove such extension.
11068 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
11072 @itemx -mno-prototype
11073 @opindex mprototype
11074 @opindex mno-prototype
11075 On System V.4 and embedded PowerPC systems assume that all calls to
11076 variable argument functions are properly prototyped. Otherwise, the
11077 compiler must insert an instruction before every non prototyped call to
11078 set or clear bit 6 of the condition code register (@var{CR}) to
11079 indicate whether floating point values were passed in the floating point
11080 registers in case the function takes a variable arguments. With
11081 @option{-mprototype}, only calls to prototyped variable argument functions
11082 will set or clear the bit.
11086 On embedded PowerPC systems, assume that the startup module is called
11087 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
11088 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
11093 On embedded PowerPC systems, assume that the startup module is called
11094 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
11099 On embedded PowerPC systems, assume that the startup module is called
11100 @file{crt0.o} and the standard C libraries are @file{libads.a} and
11103 @item -myellowknife
11104 @opindex myellowknife
11105 On embedded PowerPC systems, assume that the startup module is called
11106 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
11111 On System V.4 and embedded PowerPC systems, specify that you are
11112 compiling for a VxWorks system.
11116 Specify that you are compiling for the WindISS simulation environment.
11120 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
11121 header to indicate that @samp{eabi} extended relocations are used.
11127 On System V.4 and embedded PowerPC systems do (do not) adhere to the
11128 Embedded Applications Binary Interface (eabi) which is a set of
11129 modifications to the System V.4 specifications. Selecting @option{-meabi}
11130 means that the stack is aligned to an 8 byte boundary, a function
11131 @code{__eabi} is called to from @code{main} to set up the eabi
11132 environment, and the @option{-msdata} option can use both @code{r2} and
11133 @code{r13} to point to two separate small data areas. Selecting
11134 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
11135 do not call an initialization function from @code{main}, and the
11136 @option{-msdata} option will only use @code{r13} to point to a single
11137 small data area. The @option{-meabi} option is on by default if you
11138 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
11141 @opindex msdata=eabi
11142 On System V.4 and embedded PowerPC systems, put small initialized
11143 @code{const} global and static data in the @samp{.sdata2} section, which
11144 is pointed to by register @code{r2}. Put small initialized
11145 non-@code{const} global and static data in the @samp{.sdata} section,
11146 which is pointed to by register @code{r13}. Put small uninitialized
11147 global and static data in the @samp{.sbss} section, which is adjacent to
11148 the @samp{.sdata} section. The @option{-msdata=eabi} option is
11149 incompatible with the @option{-mrelocatable} option. The
11150 @option{-msdata=eabi} option also sets the @option{-memb} option.
11153 @opindex msdata=sysv
11154 On System V.4 and embedded PowerPC systems, put small global and static
11155 data in the @samp{.sdata} section, which is pointed to by register
11156 @code{r13}. Put small uninitialized global and static data in the
11157 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11158 The @option{-msdata=sysv} option is incompatible with the
11159 @option{-mrelocatable} option.
11161 @item -msdata=default
11163 @opindex msdata=default
11165 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
11166 compile code the same as @option{-msdata=eabi}, otherwise compile code the
11167 same as @option{-msdata=sysv}.
11170 @opindex msdata-data
11171 On System V.4 and embedded PowerPC systems, put small global and static
11172 data in the @samp{.sdata} section. Put small uninitialized global and
11173 static data in the @samp{.sbss} section. Do not use register @code{r13}
11174 to address small data however. This is the default behavior unless
11175 other @option{-msdata} options are used.
11179 @opindex msdata=none
11181 On embedded PowerPC systems, put all initialized global and static data
11182 in the @samp{.data} section, and all uninitialized data in the
11183 @samp{.bss} section.
11187 @cindex smaller data references (PowerPC)
11188 @cindex .sdata/.sdata2 references (PowerPC)
11189 On embedded PowerPC systems, put global and static items less than or
11190 equal to @var{num} bytes into the small data or bss sections instead of
11191 the normal data or bss section. By default, @var{num} is 8. The
11192 @option{-G @var{num}} switch is also passed to the linker.
11193 All modules should be compiled with the same @option{-G @var{num}} value.
11196 @itemx -mno-regnames
11198 @opindex mno-regnames
11199 On System V.4 and embedded PowerPC systems do (do not) emit register
11200 names in the assembly language output using symbolic forms.
11203 @itemx -mno-longcall
11205 @opindex mno-longcall
11206 Default to making all function calls indirectly, using a register, so
11207 that functions which reside further than 32 megabytes (33,554,432
11208 bytes) from the current location can be called. This setting can be
11209 overridden by the @code{shortcall} function attribute, or by
11210 @code{#pragma longcall(0)}.
11212 Some linkers are capable of detecting out-of-range calls and generating
11213 glue code on the fly. On these systems, long calls are unnecessary and
11214 generate slower code. As of this writing, the AIX linker can do this,
11215 as can the GNU linker for PowerPC/64. It is planned to add this feature
11216 to the GNU linker for 32-bit PowerPC systems as well.
11218 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11219 callee, L42'', plus a ``branch island'' (glue code). The two target
11220 addresses represent the callee and the ``branch island''. The
11221 Darwin/PPC linker will prefer the first address and generate a ``bl
11222 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11223 otherwise, the linker will generate ``bl L42'' to call the ``branch
11224 island''. The ``branch island'' is appended to the body of the
11225 calling function; it computes the full 32-bit address of the callee
11228 On Mach-O (Darwin) systems, this option directs the compiler emit to
11229 the glue for every direct call, and the Darwin linker decides whether
11230 to use or discard it.
11232 In the future, we may cause GCC to ignore all longcall specifications
11233 when the linker is known to generate glue.
11237 Adds support for multithreading with the @dfn{pthreads} library.
11238 This option sets flags for both the preprocessor and linker.
11242 @node S/390 and zSeries Options
11243 @subsection S/390 and zSeries Options
11244 @cindex S/390 and zSeries Options
11246 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11250 @itemx -msoft-float
11251 @opindex mhard-float
11252 @opindex msoft-float
11253 Use (do not use) the hardware floating-point instructions and registers
11254 for floating-point operations. When @option{-msoft-float} is specified,
11255 functions in @file{libgcc.a} will be used to perform floating-point
11256 operations. When @option{-mhard-float} is specified, the compiler
11257 generates IEEE floating-point instructions. This is the default.
11260 @itemx -mno-backchain
11261 @opindex mbackchain
11262 @opindex mno-backchain
11263 Store (do not store) the address of the caller's frame as backchain pointer
11264 into the callee's stack frame.
11265 A backchain may be needed to allow debugging using tools that do not understand
11266 DWARF-2 call frame information.
11267 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
11268 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
11269 the backchain is placed into the topmost word of the 96/160 byte register
11272 In general, code compiled with @option{-mbackchain} is call-compatible with
11273 code compiled with @option{-mmo-backchain}; however, use of the backchain
11274 for debugging purposes usually requires that the whole binary is built with
11275 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
11276 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11277 to build a linux kernel use @option{-msoft-float}.
11279 The default is to not maintain the backchain.
11281 @item -mpacked-stack
11282 @item -mno-packed-stack
11283 @opindex mpacked-stack
11284 @opindex mno-packed-stack
11285 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
11286 specified, the compiler uses the all fields of the 96/160 byte register save
11287 area only for their default purpose; unused fields still take up stack space.
11288 When @option{-mpacked-stack} is specified, register save slots are densely
11289 packed at the top of the register save area; unused space is reused for other
11290 purposes, allowing for more efficient use of the available stack space.
11291 However, when @option{-mbackchain} is also in effect, the topmost word of
11292 the save area is always used to store the backchain, and the return address
11293 register is always saved two words below the backchain.
11295 As long as the stack frame backchain is not used, code generated with
11296 @option{-mpacked-stack} is call-compatible with code generated with
11297 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
11298 S/390 or zSeries generated code that uses the stack frame backchain at run
11299 time, not just for debugging purposes. Such code is not call-compatible
11300 with code compiled with @option{-mpacked-stack}. Also, note that the
11301 combination of @option{-mbackchain},
11302 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11303 to build a linux kernel use @option{-msoft-float}.
11305 The default is to not use the packed stack layout.
11308 @itemx -mno-small-exec
11309 @opindex msmall-exec
11310 @opindex mno-small-exec
11311 Generate (or do not generate) code using the @code{bras} instruction
11312 to do subroutine calls.
11313 This only works reliably if the total executable size does not
11314 exceed 64k. The default is to use the @code{basr} instruction instead,
11315 which does not have this limitation.
11321 When @option{-m31} is specified, generate code compliant to the
11322 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
11323 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
11324 particular to generate 64-bit instructions. For the @samp{s390}
11325 targets, the default is @option{-m31}, while the @samp{s390x}
11326 targets default to @option{-m64}.
11332 When @option{-mzarch} is specified, generate code using the
11333 instructions available on z/Architecture.
11334 When @option{-mesa} is specified, generate code using the
11335 instructions available on ESA/390. Note that @option{-mesa} is
11336 not possible with @option{-m64}.
11337 When generating code compliant to the GNU/Linux for S/390 ABI,
11338 the default is @option{-mesa}. When generating code compliant
11339 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
11345 Generate (or do not generate) code using the @code{mvcle} instruction
11346 to perform block moves. When @option{-mno-mvcle} is specified,
11347 use a @code{mvc} loop instead. This is the default.
11353 Print (or do not print) additional debug information when compiling.
11354 The default is to not print debug information.
11356 @item -march=@var{cpu-type}
11358 Generate code that will run on @var{cpu-type}, which is the name of a system
11359 representing a certain processor type. Possible values for
11360 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
11361 When generating code using the instructions available on z/Architecture,
11362 the default is @option{-march=z900}. Otherwise, the default is
11363 @option{-march=g5}.
11365 @item -mtune=@var{cpu-type}
11367 Tune to @var{cpu-type} everything applicable about the generated code,
11368 except for the ABI and the set of available instructions.
11369 The list of @var{cpu-type} values is the same as for @option{-march}.
11370 The default is the value used for @option{-march}.
11373 @itemx -mno-tpf-trace
11374 @opindex mtpf-trace
11375 @opindex mno-tpf-trace
11376 Generate code that adds (does not add) in TPF OS specific branches to trace
11377 routines in the operating system. This option is off by default, even
11378 when compiling for the TPF OS@.
11381 @itemx -mno-fused-madd
11382 @opindex mfused-madd
11383 @opindex mno-fused-madd
11384 Generate code that uses (does not use) the floating point multiply and
11385 accumulate instructions. These instructions are generated by default if
11386 hardware floating point is used.
11388 @item -mwarn-framesize=@var{framesize}
11389 @opindex mwarn-framesize
11390 Emit a warning if the current function exceeds the given frame size. Because
11391 this is a compile time check it doesn't need to be a real problem when the program
11392 runs. It is intended to identify functions which most probably cause
11393 a stack overflow. It is useful to be used in an environment with limited stack
11394 size e.g.@: the linux kernel.
11396 @item -mwarn-dynamicstack
11397 @opindex mwarn-dynamicstack
11398 Emit a warning if the function calls alloca or uses dynamically
11399 sized arrays. This is generally a bad idea with a limited stack size.
11401 @item -mstack-guard=@var{stack-guard}
11402 @item -mstack-size=@var{stack-size}
11403 @opindex mstack-guard
11404 @opindex mstack-size
11405 These arguments always have to be used in conjunction. If they are present the s390
11406 back end emits additional instructions in the function prologue which trigger a trap
11407 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
11408 (remember that the stack on s390 grows downward). These options are intended to
11409 be used to help debugging stack overflow problems. The additionally emitted code
11410 cause only little overhead and hence can also be used in production like systems
11411 without greater performance degradation. The given values have to be exact
11412 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard}.
11413 In order to be efficient the extra code makes the assumption that the stack starts
11414 at an address aligned to the value given by @var{stack-size}.
11418 @subsection SH Options
11420 These @samp{-m} options are defined for the SH implementations:
11425 Generate code for the SH1.
11429 Generate code for the SH2.
11432 Generate code for the SH2e.
11436 Generate code for the SH3.
11440 Generate code for the SH3e.
11444 Generate code for the SH4 without a floating-point unit.
11446 @item -m4-single-only
11447 @opindex m4-single-only
11448 Generate code for the SH4 with a floating-point unit that only
11449 supports single-precision arithmetic.
11453 Generate code for the SH4 assuming the floating-point unit is in
11454 single-precision mode by default.
11458 Generate code for the SH4.
11462 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
11463 floating-point unit is not used.
11465 @item -m4a-single-only
11466 @opindex m4a-single-only
11467 Generate code for the SH4a, in such a way that no double-precision
11468 floating point operations are used.
11471 @opindex m4a-single
11472 Generate code for the SH4a assuming the floating-point unit is in
11473 single-precision mode by default.
11477 Generate code for the SH4a.
11481 Same as @option{-m4a-nofpu}, except that it implicitly passes
11482 @option{-dsp} to the assembler. GCC doesn't generate any DSP
11483 instructions at the moment.
11487 Compile code for the processor in big endian mode.
11491 Compile code for the processor in little endian mode.
11495 Align doubles at 64-bit boundaries. Note that this changes the calling
11496 conventions, and thus some functions from the standard C library will
11497 not work unless you recompile it first with @option{-mdalign}.
11501 Shorten some address references at link time, when possible; uses the
11502 linker option @option{-relax}.
11506 Use 32-bit offsets in @code{switch} tables. The default is to use
11511 Enable the use of the instruction @code{fmovd}.
11515 Comply with the calling conventions defined by Renesas.
11519 Comply with the calling conventions defined by Renesas.
11523 Comply with the calling conventions defined for GCC before the Renesas
11524 conventions were available. This option is the default for all
11525 targets of the SH toolchain except for @samp{sh-symbianelf}.
11528 @opindex mnomacsave
11529 Mark the @code{MAC} register as call-clobbered, even if
11530 @option{-mhitachi} is given.
11534 Increase IEEE-compliance of floating-point code.
11535 At the moment, this is equivalent to @option{-fno-finite-math-only}.
11536 When generating 16 bit SH opcodes, getting IEEE-conforming results for
11537 comparisons of NANs / infinities incurs extra overhead in every
11538 floating point comparison, therefore the default is set to
11539 @option{-ffinite-math-only}.
11543 Dump instruction size and location in the assembly code.
11546 @opindex mpadstruct
11547 This option is deprecated. It pads structures to multiple of 4 bytes,
11548 which is incompatible with the SH ABI@.
11552 Optimize for space instead of speed. Implied by @option{-Os}.
11555 @opindex mprefergot
11556 When generating position-independent code, emit function calls using
11557 the Global Offset Table instead of the Procedure Linkage Table.
11561 Generate a library function call to invalidate instruction cache
11562 entries, after fixing up a trampoline. This library function call
11563 doesn't assume it can write to the whole memory address space. This
11564 is the default when the target is @code{sh-*-linux*}.
11566 @item -multcost=@var{number}
11567 @opindex multcost=@var{number}
11568 Set the cost to assume for a multiply insn.
11570 @item -mdiv=@var{strategy}
11571 @opindex mdiv=@var{strategy}
11572 Set the division strategy to use for SHmedia code. @var{strategy} must be
11573 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
11574 inv:call2, inv:fp .
11575 "fp" performs the operation in floating point. This has a very high latency,
11576 but needs only a few instructions, so it might be a good choice if
11577 your code has enough easily exploitable ILP to allow the compiler to
11578 schedule the floating point instructions together with other instructions.
11579 Division by zero causes a floating point exception.
11580 "inv" uses integer operations to calculate the inverse of the divisor,
11581 and then multiplies the dividend with the inverse. This strategy allows
11582 cse and hoisting of the inverse calculation. Division by zero calculates
11583 an unspecified result, but does not trap.
11584 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
11585 have been found, or if the entire operation has been hoisted to the same
11586 place, the last stages of the inverse calculation are intertwined with the
11587 final multiply to reduce the overall latency, at the expense of using a few
11588 more instructions, and thus offering fewer scheduling opportunities with
11590 "call" calls a library function that usually implements the inv:minlat
11592 This gives high code density for m5-*media-nofpu compilations.
11593 "call2" uses a different entry point of the same library function, where it
11594 assumes that a pointer to a lookup table has already been set up, which
11595 exposes the pointer load to cse / code hoisting optimizations.
11596 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
11597 code generation, but if the code stays unoptimized, revert to the "call",
11598 "call2", or "fp" strategies, respectively. Note that the
11599 potentially-trapping side effect of division by zero is carried by a
11600 separate instruction, so it is possible that all the integer instructions
11601 are hoisted out, but the marker for the side effect stays where it is.
11602 A recombination to fp operations or a call is not possible in that case.
11603 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
11604 that the inverse calculation was nor separated from the multiply, they speed
11605 up division where the dividend fits into 20 bits (plus sign where applicable),
11606 by inserting a test to skip a number of operations in this case; this test
11607 slows down the case of larger dividends. inv20u assumes the case of a such
11608 a small dividend to be unlikely, and inv20l assumes it to be likely.
11610 @item -mdivsi3_libfunc=@var{name}
11611 @opindex mdivsi3_libfunc=@var{name}
11612 Set the name of the library function used for 32 bit signed division to
11613 @var{name}. This only affect the name used in the call and inv:call
11614 division strategies, and the compiler will still expect the same
11615 sets of input/output/clobbered registers as if this option was not present.
11617 @item -madjust-unroll
11618 @opindex madjust-unroll
11619 Throttle unrolling to avoid thrashing target registers.
11620 This option only has an effect if the gcc code base supports the
11621 TARGET_ADJUST_UNROLL_MAX target hook.
11623 @item -mindexed-addressing
11624 @opindex mindexed-addressing
11625 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
11626 This is only safe if the hardware and/or OS implement 32 bit wrap-around
11627 semantics for the indexed addressing mode. The architecture allows the
11628 implementation of processors with 64 bit MMU, which the OS could use to
11629 get 32 bit addressing, but since no current hardware implementation supports
11630 this or any other way to make the indexed addressing mode safe to use in
11631 the 32 bit ABI, the default is -mno-indexed-addressing.
11633 @item -mgettrcost=@var{number}
11634 @opindex mgettrcost=@var{number}
11635 Set the cost assumed for the gettr instruction to @var{number}.
11636 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
11640 Assume pt* instructions won't trap. This will generally generate better
11641 scheduled code, but is unsafe on current hardware. The current architecture
11642 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
11643 This has the unintentional effect of making it unsafe to schedule ptabs /
11644 ptrel before a branch, or hoist it out of a loop. For example,
11645 __do_global_ctors, a part of libgcc that runs constructors at program
11646 startup, calls functions in a list which is delimited by -1. With the
11647 -mpt-fixed option, the ptabs will be done before testing against -1.
11648 That means that all the constructors will be run a bit quicker, but when
11649 the loop comes to the end of the list, the program crashes because ptabs
11650 loads -1 into a target register. Since this option is unsafe for any
11651 hardware implementing the current architecture specification, the default
11652 is -mno-pt-fixed. Unless the user specifies a specific cost with
11653 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
11654 this deters register allocation using target registers for storing
11657 @item -minvalid-symbols
11658 @opindex minvalid-symbols
11659 Assume symbols might be invalid. Ordinary function symbols generated by
11660 the compiler will always be valid to load with movi/shori/ptabs or
11661 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
11662 to generate symbols that will cause ptabs / ptrel to trap.
11663 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
11664 It will then prevent cross-basic-block cse, hoisting and most scheduling
11665 of symbol loads. The default is @option{-mno-invalid-symbols}.
11668 @node SPARC Options
11669 @subsection SPARC Options
11670 @cindex SPARC options
11672 These @samp{-m} options are supported on the SPARC:
11675 @item -mno-app-regs
11677 @opindex mno-app-regs
11679 Specify @option{-mapp-regs} to generate output using the global registers
11680 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
11683 To be fully SVR4 ABI compliant at the cost of some performance loss,
11684 specify @option{-mno-app-regs}. You should compile libraries and system
11685 software with this option.
11688 @itemx -mhard-float
11690 @opindex mhard-float
11691 Generate output containing floating point instructions. This is the
11695 @itemx -msoft-float
11697 @opindex msoft-float
11698 Generate output containing library calls for floating point.
11699 @strong{Warning:} the requisite libraries are not available for all SPARC
11700 targets. Normally the facilities of the machine's usual C compiler are
11701 used, but this cannot be done directly in cross-compilation. You must make
11702 your own arrangements to provide suitable library functions for
11703 cross-compilation. The embedded targets @samp{sparc-*-aout} and
11704 @samp{sparclite-*-*} do provide software floating point support.
11706 @option{-msoft-float} changes the calling convention in the output file;
11707 therefore, it is only useful if you compile @emph{all} of a program with
11708 this option. In particular, you need to compile @file{libgcc.a}, the
11709 library that comes with GCC, with @option{-msoft-float} in order for
11712 @item -mhard-quad-float
11713 @opindex mhard-quad-float
11714 Generate output containing quad-word (long double) floating point
11717 @item -msoft-quad-float
11718 @opindex msoft-quad-float
11719 Generate output containing library calls for quad-word (long double)
11720 floating point instructions. The functions called are those specified
11721 in the SPARC ABI@. This is the default.
11723 As of this writing, there are no SPARC implementations that have hardware
11724 support for the quad-word floating point instructions. They all invoke
11725 a trap handler for one of these instructions, and then the trap handler
11726 emulates the effect of the instruction. Because of the trap handler overhead,
11727 this is much slower than calling the ABI library routines. Thus the
11728 @option{-msoft-quad-float} option is the default.
11730 @item -mno-unaligned-doubles
11731 @itemx -munaligned-doubles
11732 @opindex mno-unaligned-doubles
11733 @opindex munaligned-doubles
11734 Assume that doubles have 8 byte alignment. This is the default.
11736 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
11737 alignment only if they are contained in another type, or if they have an
11738 absolute address. Otherwise, it assumes they have 4 byte alignment.
11739 Specifying this option avoids some rare compatibility problems with code
11740 generated by other compilers. It is not the default because it results
11741 in a performance loss, especially for floating point code.
11743 @item -mno-faster-structs
11744 @itemx -mfaster-structs
11745 @opindex mno-faster-structs
11746 @opindex mfaster-structs
11747 With @option{-mfaster-structs}, the compiler assumes that structures
11748 should have 8 byte alignment. This enables the use of pairs of
11749 @code{ldd} and @code{std} instructions for copies in structure
11750 assignment, in place of twice as many @code{ld} and @code{st} pairs.
11751 However, the use of this changed alignment directly violates the SPARC
11752 ABI@. Thus, it's intended only for use on targets where the developer
11753 acknowledges that their resulting code will not be directly in line with
11754 the rules of the ABI@.
11756 @item -mimpure-text
11757 @opindex mimpure-text
11758 @option{-mimpure-text}, used in addition to @option{-shared}, tells
11759 the compiler to not pass @option{-z text} to the linker when linking a
11760 shared object. Using this option, you can link position-dependent
11761 code into a shared object.
11763 @option{-mimpure-text} suppresses the ``relocations remain against
11764 allocatable but non-writable sections'' linker error message.
11765 However, the necessary relocations will trigger copy-on-write, and the
11766 shared object is not actually shared across processes. Instead of
11767 using @option{-mimpure-text}, you should compile all source code with
11768 @option{-fpic} or @option{-fPIC}.
11770 This option is only available on SunOS and Solaris.
11772 @item -mcpu=@var{cpu_type}
11774 Set the instruction set, register set, and instruction scheduling parameters
11775 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
11776 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
11777 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
11778 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
11779 @samp{ultrasparc3}.
11781 Default instruction scheduling parameters are used for values that select
11782 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
11783 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
11785 Here is a list of each supported architecture and their supported
11790 v8: supersparc, hypersparc
11791 sparclite: f930, f934, sparclite86x
11793 v9: ultrasparc, ultrasparc3
11796 By default (unless configured otherwise), GCC generates code for the V7
11797 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
11798 additionally optimizes it for the Cypress CY7C602 chip, as used in the
11799 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
11800 SPARCStation 1, 2, IPX etc.
11802 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
11803 architecture. The only difference from V7 code is that the compiler emits
11804 the integer multiply and integer divide instructions which exist in SPARC-V8
11805 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
11806 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
11809 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
11810 the SPARC architecture. This adds the integer multiply, integer divide step
11811 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
11812 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
11813 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
11814 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
11815 MB86934 chip, which is the more recent SPARClite with FPU@.
11817 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
11818 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
11819 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
11820 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
11821 optimizes it for the TEMIC SPARClet chip.
11823 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
11824 architecture. This adds 64-bit integer and floating-point move instructions,
11825 3 additional floating-point condition code registers and conditional move
11826 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
11827 optimizes it for the Sun UltraSPARC I/II chips. With
11828 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
11829 Sun UltraSPARC III chip.
11831 @item -mtune=@var{cpu_type}
11833 Set the instruction scheduling parameters for machine type
11834 @var{cpu_type}, but do not set the instruction set or register set that the
11835 option @option{-mcpu=@var{cpu_type}} would.
11837 The same values for @option{-mcpu=@var{cpu_type}} can be used for
11838 @option{-mtune=@var{cpu_type}}, but the only useful values are those
11839 that select a particular cpu implementation. Those are @samp{cypress},
11840 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
11841 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
11842 @samp{ultrasparc3}.
11847 @opindex mno-v8plus
11848 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
11849 difference from the V8 ABI is that the global and out registers are
11850 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
11851 mode for all SPARC-V9 processors.
11857 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
11858 Visual Instruction Set extensions. The default is @option{-mno-vis}.
11861 These @samp{-m} options are supported in addition to the above
11862 on SPARC-V9 processors in 64-bit environments:
11865 @item -mlittle-endian
11866 @opindex mlittle-endian
11867 Generate code for a processor running in little-endian mode. It is only
11868 available for a few configurations and most notably not on Solaris and Linux.
11874 Generate code for a 32-bit or 64-bit environment.
11875 The 32-bit environment sets int, long and pointer to 32 bits.
11876 The 64-bit environment sets int to 32 bits and long and pointer
11879 @item -mcmodel=medlow
11880 @opindex mcmodel=medlow
11881 Generate code for the Medium/Low code model: 64-bit addresses, programs
11882 must be linked in the low 32 bits of memory. Programs can be statically
11883 or dynamically linked.
11885 @item -mcmodel=medmid
11886 @opindex mcmodel=medmid
11887 Generate code for the Medium/Middle code model: 64-bit addresses, programs
11888 must be linked in the low 44 bits of memory, the text and data segments must
11889 be less than 2GB in size and the data segment must be located within 2GB of
11892 @item -mcmodel=medany
11893 @opindex mcmodel=medany
11894 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
11895 may be linked anywhere in memory, the text and data segments must be less
11896 than 2GB in size and the data segment must be located within 2GB of the
11899 @item -mcmodel=embmedany
11900 @opindex mcmodel=embmedany
11901 Generate code for the Medium/Anywhere code model for embedded systems:
11902 64-bit addresses, the text and data segments must be less than 2GB in
11903 size, both starting anywhere in memory (determined at link time). The
11904 global register %g4 points to the base of the data segment. Programs
11905 are statically linked and PIC is not supported.
11908 @itemx -mno-stack-bias
11909 @opindex mstack-bias
11910 @opindex mno-stack-bias
11911 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
11912 frame pointer if present, are offset by @minus{}2047 which must be added back
11913 when making stack frame references. This is the default in 64-bit mode.
11914 Otherwise, assume no such offset is present.
11917 These switches are supported in addition to the above on Solaris:
11922 Add support for multithreading using the Solaris threads library. This
11923 option sets flags for both the preprocessor and linker. This option does
11924 not affect the thread safety of object code produced by the compiler or
11925 that of libraries supplied with it.
11929 Add support for multithreading using the POSIX threads library. This
11930 option sets flags for both the preprocessor and linker. This option does
11931 not affect the thread safety of object code produced by the compiler or
11932 that of libraries supplied with it.
11935 @node System V Options
11936 @subsection Options for System V
11938 These additional options are available on System V Release 4 for
11939 compatibility with other compilers on those systems:
11944 Create a shared object.
11945 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
11949 Identify the versions of each tool used by the compiler, in a
11950 @code{.ident} assembler directive in the output.
11954 Refrain from adding @code{.ident} directives to the output file (this is
11957 @item -YP,@var{dirs}
11959 Search the directories @var{dirs}, and no others, for libraries
11960 specified with @option{-l}.
11962 @item -Ym,@var{dir}
11964 Look in the directory @var{dir} to find the M4 preprocessor.
11965 The assembler uses this option.
11966 @c This is supposed to go with a -Yd for predefined M4 macro files, but
11967 @c the generic assembler that comes with Solaris takes just -Ym.
11970 @node TMS320C3x/C4x Options
11971 @subsection TMS320C3x/C4x Options
11972 @cindex TMS320C3x/C4x Options
11974 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
11978 @item -mcpu=@var{cpu_type}
11980 Set the instruction set, register set, and instruction scheduling
11981 parameters for machine type @var{cpu_type}. Supported values for
11982 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
11983 @samp{c44}. The default is @samp{c40} to generate code for the
11988 @itemx -msmall-memory
11990 @opindex mbig-memory
11992 @opindex msmall-memory
11994 Generates code for the big or small memory model. The small memory
11995 model assumed that all data fits into one 64K word page. At run-time
11996 the data page (DP) register must be set to point to the 64K page
11997 containing the .bss and .data program sections. The big memory model is
11998 the default and requires reloading of the DP register for every direct
12005 Allow (disallow) allocation of general integer operands into the block
12006 count register BK@.
12012 Enable (disable) generation of code using decrement and branch,
12013 DBcond(D), instructions. This is enabled by default for the C4x. To be
12014 on the safe side, this is disabled for the C3x, since the maximum
12015 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
12016 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
12017 that it can utilize the decrement and branch instruction, but will give
12018 up if there is more than one memory reference in the loop. Thus a loop
12019 where the loop counter is decremented can generate slightly more
12020 efficient code, in cases where the RPTB instruction cannot be utilized.
12022 @item -mdp-isr-reload
12024 @opindex mdp-isr-reload
12026 Force the DP register to be saved on entry to an interrupt service
12027 routine (ISR), reloaded to point to the data section, and restored on
12028 exit from the ISR@. This should not be required unless someone has
12029 violated the small memory model by modifying the DP register, say within
12036 For the C3x use the 24-bit MPYI instruction for integer multiplies
12037 instead of a library call to guarantee 32-bit results. Note that if one
12038 of the operands is a constant, then the multiplication will be performed
12039 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
12040 then squaring operations are performed inline instead of a library call.
12043 @itemx -mno-fast-fix
12045 @opindex mno-fast-fix
12046 The C3x/C4x FIX instruction to convert a floating point value to an
12047 integer value chooses the nearest integer less than or equal to the
12048 floating point value rather than to the nearest integer. Thus if the
12049 floating point number is negative, the result will be incorrectly
12050 truncated an additional code is necessary to detect and correct this
12051 case. This option can be used to disable generation of the additional
12052 code required to correct the result.
12058 Enable (disable) generation of repeat block sequences using the RPTB
12059 instruction for zero overhead looping. The RPTB construct is only used
12060 for innermost loops that do not call functions or jump across the loop
12061 boundaries. There is no advantage having nested RPTB loops due to the
12062 overhead required to save and restore the RC, RS, and RE registers.
12063 This is enabled by default with @option{-O2}.
12065 @item -mrpts=@var{count}
12069 Enable (disable) the use of the single instruction repeat instruction
12070 RPTS@. If a repeat block contains a single instruction, and the loop
12071 count can be guaranteed to be less than the value @var{count}, GCC will
12072 emit a RPTS instruction instead of a RPTB@. If no value is specified,
12073 then a RPTS will be emitted even if the loop count cannot be determined
12074 at compile time. Note that the repeated instruction following RPTS does
12075 not have to be reloaded from memory each iteration, thus freeing up the
12076 CPU buses for operands. However, since interrupts are blocked by this
12077 instruction, it is disabled by default.
12079 @item -mloop-unsigned
12080 @itemx -mno-loop-unsigned
12081 @opindex mloop-unsigned
12082 @opindex mno-loop-unsigned
12083 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
12084 is @math{2^{31} + 1} since these instructions test if the iteration count is
12085 negative to terminate the loop. If the iteration count is unsigned
12086 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
12087 exceeded. This switch allows an unsigned iteration count.
12091 Try to emit an assembler syntax that the TI assembler (asm30) is happy
12092 with. This also enforces compatibility with the API employed by the TI
12093 C3x C compiler. For example, long doubles are passed as structures
12094 rather than in floating point registers.
12100 Generate code that uses registers (stack) for passing arguments to functions.
12101 By default, arguments are passed in registers where possible rather
12102 than by pushing arguments on to the stack.
12104 @item -mparallel-insns
12105 @itemx -mno-parallel-insns
12106 @opindex mparallel-insns
12107 @opindex mno-parallel-insns
12108 Allow the generation of parallel instructions. This is enabled by
12109 default with @option{-O2}.
12111 @item -mparallel-mpy
12112 @itemx -mno-parallel-mpy
12113 @opindex mparallel-mpy
12114 @opindex mno-parallel-mpy
12115 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
12116 provided @option{-mparallel-insns} is also specified. These instructions have
12117 tight register constraints which can pessimize the code generation
12118 of large functions.
12123 @subsection V850 Options
12124 @cindex V850 Options
12126 These @samp{-m} options are defined for V850 implementations:
12130 @itemx -mno-long-calls
12131 @opindex mlong-calls
12132 @opindex mno-long-calls
12133 Treat all calls as being far away (near). If calls are assumed to be
12134 far away, the compiler will always load the functions address up into a
12135 register, and call indirect through the pointer.
12141 Do not optimize (do optimize) basic blocks that use the same index
12142 pointer 4 or more times to copy pointer into the @code{ep} register, and
12143 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
12144 option is on by default if you optimize.
12146 @item -mno-prolog-function
12147 @itemx -mprolog-function
12148 @opindex mno-prolog-function
12149 @opindex mprolog-function
12150 Do not use (do use) external functions to save and restore registers
12151 at the prologue and epilogue of a function. The external functions
12152 are slower, but use less code space if more than one function saves
12153 the same number of registers. The @option{-mprolog-function} option
12154 is on by default if you optimize.
12158 Try to make the code as small as possible. At present, this just turns
12159 on the @option{-mep} and @option{-mprolog-function} options.
12161 @item -mtda=@var{n}
12163 Put static or global variables whose size is @var{n} bytes or less into
12164 the tiny data area that register @code{ep} points to. The tiny data
12165 area can hold up to 256 bytes in total (128 bytes for byte references).
12167 @item -msda=@var{n}
12169 Put static or global variables whose size is @var{n} bytes or less into
12170 the small data area that register @code{gp} points to. The small data
12171 area can hold up to 64 kilobytes.
12173 @item -mzda=@var{n}
12175 Put static or global variables whose size is @var{n} bytes or less into
12176 the first 32 kilobytes of memory.
12180 Specify that the target processor is the V850.
12183 @opindex mbig-switch
12184 Generate code suitable for big switch tables. Use this option only if
12185 the assembler/linker complain about out of range branches within a switch
12190 This option will cause r2 and r5 to be used in the code generated by
12191 the compiler. This setting is the default.
12193 @item -mno-app-regs
12194 @opindex mno-app-regs
12195 This option will cause r2 and r5 to be treated as fixed registers.
12199 Specify that the target processor is the V850E1. The preprocessor
12200 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
12201 this option is used.
12205 Specify that the target processor is the V850E@. The preprocessor
12206 constant @samp{__v850e__} will be defined if this option is used.
12208 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
12209 are defined then a default target processor will be chosen and the
12210 relevant @samp{__v850*__} preprocessor constant will be defined.
12212 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
12213 defined, regardless of which processor variant is the target.
12215 @item -mdisable-callt
12216 @opindex mdisable-callt
12217 This option will suppress generation of the CALLT instruction for the
12218 v850e and v850e1 flavors of the v850 architecture. The default is
12219 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
12224 @subsection VAX Options
12225 @cindex VAX options
12227 These @samp{-m} options are defined for the VAX:
12232 Do not output certain jump instructions (@code{aobleq} and so on)
12233 that the Unix assembler for the VAX cannot handle across long
12238 Do output those jump instructions, on the assumption that you
12239 will assemble with the GNU assembler.
12243 Output code for g-format floating point numbers instead of d-format.
12246 @node x86-64 Options
12247 @subsection x86-64 Options
12248 @cindex x86-64 options
12250 These are listed under @xref{i386 and x86-64 Options}.
12252 @node Xstormy16 Options
12253 @subsection Xstormy16 Options
12254 @cindex Xstormy16 Options
12256 These options are defined for Xstormy16:
12261 Choose startup files and linker script suitable for the simulator.
12264 @node Xtensa Options
12265 @subsection Xtensa Options
12266 @cindex Xtensa Options
12268 These options are supported for Xtensa targets:
12272 @itemx -mno-const16
12274 @opindex mno-const16
12275 Enable or disable use of @code{CONST16} instructions for loading
12276 constant values. The @code{CONST16} instruction is currently not a
12277 standard option from Tensilica. When enabled, @code{CONST16}
12278 instructions are always used in place of the standard @code{L32R}
12279 instructions. The use of @code{CONST16} is enabled by default only if
12280 the @code{L32R} instruction is not available.
12283 @itemx -mno-fused-madd
12284 @opindex mfused-madd
12285 @opindex mno-fused-madd
12286 Enable or disable use of fused multiply/add and multiply/subtract
12287 instructions in the floating-point option. This has no effect if the
12288 floating-point option is not also enabled. Disabling fused multiply/add
12289 and multiply/subtract instructions forces the compiler to use separate
12290 instructions for the multiply and add/subtract operations. This may be
12291 desirable in some cases where strict IEEE 754-compliant results are
12292 required: the fused multiply add/subtract instructions do not round the
12293 intermediate result, thereby producing results with @emph{more} bits of
12294 precision than specified by the IEEE standard. Disabling fused multiply
12295 add/subtract instructions also ensures that the program output is not
12296 sensitive to the compiler's ability to combine multiply and add/subtract
12299 @item -mtext-section-literals
12300 @itemx -mno-text-section-literals
12301 @opindex mtext-section-literals
12302 @opindex mno-text-section-literals
12303 Control the treatment of literal pools. The default is
12304 @option{-mno-text-section-literals}, which places literals in a separate
12305 section in the output file. This allows the literal pool to be placed
12306 in a data RAM/ROM, and it also allows the linker to combine literal
12307 pools from separate object files to remove redundant literals and
12308 improve code size. With @option{-mtext-section-literals}, the literals
12309 are interspersed in the text section in order to keep them as close as
12310 possible to their references. This may be necessary for large assembly
12313 @item -mtarget-align
12314 @itemx -mno-target-align
12315 @opindex mtarget-align
12316 @opindex mno-target-align
12317 When this option is enabled, GCC instructs the assembler to
12318 automatically align instructions to reduce branch penalties at the
12319 expense of some code density. The assembler attempts to widen density
12320 instructions to align branch targets and the instructions following call
12321 instructions. If there are not enough preceding safe density
12322 instructions to align a target, no widening will be performed. The
12323 default is @option{-mtarget-align}. These options do not affect the
12324 treatment of auto-aligned instructions like @code{LOOP}, which the
12325 assembler will always align, either by widening density instructions or
12326 by inserting no-op instructions.
12329 @itemx -mno-longcalls
12330 @opindex mlongcalls
12331 @opindex mno-longcalls
12332 When this option is enabled, GCC instructs the assembler to translate
12333 direct calls to indirect calls unless it can determine that the target
12334 of a direct call is in the range allowed by the call instruction. This
12335 translation typically occurs for calls to functions in other source
12336 files. Specifically, the assembler translates a direct @code{CALL}
12337 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
12338 The default is @option{-mno-longcalls}. This option should be used in
12339 programs where the call target can potentially be out of range. This
12340 option is implemented in the assembler, not the compiler, so the
12341 assembly code generated by GCC will still show direct call
12342 instructions---look at the disassembled object code to see the actual
12343 instructions. Note that the assembler will use an indirect call for
12344 every cross-file call, not just those that really will be out of range.
12347 @node zSeries Options
12348 @subsection zSeries Options
12349 @cindex zSeries options
12351 These are listed under @xref{S/390 and zSeries Options}.
12353 @node Code Gen Options
12354 @section Options for Code Generation Conventions
12355 @cindex code generation conventions
12356 @cindex options, code generation
12357 @cindex run-time options
12359 These machine-independent options control the interface conventions
12360 used in code generation.
12362 Most of them have both positive and negative forms; the negative form
12363 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
12364 one of the forms is listed---the one which is not the default. You
12365 can figure out the other form by either removing @samp{no-} or adding
12369 @item -fbounds-check
12370 @opindex fbounds-check
12371 For front-ends that support it, generate additional code to check that
12372 indices used to access arrays are within the declared range. This is
12373 currently only supported by the Java and Fortran 77 front-ends, where
12374 this option defaults to true and false respectively.
12378 This option generates traps for signed overflow on addition, subtraction,
12379 multiplication operations.
12383 This option instructs the compiler to assume that signed arithmetic
12384 overflow of addition, subtraction and multiplication wraps around
12385 using twos-complement representation. This flag enables some optimizations
12386 and disables other. This option is enabled by default for the Java
12387 front-end, as required by the Java language specification.
12390 @opindex fexceptions
12391 Enable exception handling. Generates extra code needed to propagate
12392 exceptions. For some targets, this implies GCC will generate frame
12393 unwind information for all functions, which can produce significant data
12394 size overhead, although it does not affect execution. If you do not
12395 specify this option, GCC will enable it by default for languages like
12396 C++ which normally require exception handling, and disable it for
12397 languages like C that do not normally require it. However, you may need
12398 to enable this option when compiling C code that needs to interoperate
12399 properly with exception handlers written in C++. You may also wish to
12400 disable this option if you are compiling older C++ programs that don't
12401 use exception handling.
12403 @item -fnon-call-exceptions
12404 @opindex fnon-call-exceptions
12405 Generate code that allows trapping instructions to throw exceptions.
12406 Note that this requires platform-specific runtime support that does
12407 not exist everywhere. Moreover, it only allows @emph{trapping}
12408 instructions to throw exceptions, i.e.@: memory references or floating
12409 point instructions. It does not allow exceptions to be thrown from
12410 arbitrary signal handlers such as @code{SIGALRM}.
12412 @item -funwind-tables
12413 @opindex funwind-tables
12414 Similar to @option{-fexceptions}, except that it will just generate any needed
12415 static data, but will not affect the generated code in any other way.
12416 You will normally not enable this option; instead, a language processor
12417 that needs this handling would enable it on your behalf.
12419 @item -fasynchronous-unwind-tables
12420 @opindex fasynchronous-unwind-tables
12421 Generate unwind table in dwarf2 format, if supported by target machine. The
12422 table is exact at each instruction boundary, so it can be used for stack
12423 unwinding from asynchronous events (such as debugger or garbage collector).
12425 @item -fpcc-struct-return
12426 @opindex fpcc-struct-return
12427 Return ``short'' @code{struct} and @code{union} values in memory like
12428 longer ones, rather than in registers. This convention is less
12429 efficient, but it has the advantage of allowing intercallability between
12430 GCC-compiled files and files compiled with other compilers, particularly
12431 the Portable C Compiler (pcc).
12433 The precise convention for returning structures in memory depends
12434 on the target configuration macros.
12436 Short structures and unions are those whose size and alignment match
12437 that of some integer type.
12439 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12440 switch is not binary compatible with code compiled with the
12441 @option{-freg-struct-return} switch.
12442 Use it to conform to a non-default application binary interface.
12444 @item -freg-struct-return
12445 @opindex freg-struct-return
12446 Return @code{struct} and @code{union} values in registers when possible.
12447 This is more efficient for small structures than
12448 @option{-fpcc-struct-return}.
12450 If you specify neither @option{-fpcc-struct-return} nor
12451 @option{-freg-struct-return}, GCC defaults to whichever convention is
12452 standard for the target. If there is no standard convention, GCC
12453 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12454 the principal compiler. In those cases, we can choose the standard, and
12455 we chose the more efficient register return alternative.
12457 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12458 switch is not binary compatible with code compiled with the
12459 @option{-fpcc-struct-return} switch.
12460 Use it to conform to a non-default application binary interface.
12462 @item -fshort-enums
12463 @opindex fshort-enums
12464 Allocate to an @code{enum} type only as many bytes as it needs for the
12465 declared range of possible values. Specifically, the @code{enum} type
12466 will be equivalent to the smallest integer type which has enough room.
12468 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12469 code that is not binary compatible with code generated without that switch.
12470 Use it to conform to a non-default application binary interface.
12472 @item -fshort-double
12473 @opindex fshort-double
12474 Use the same size for @code{double} as for @code{float}.
12476 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
12477 code that is not binary compatible with code generated without that switch.
12478 Use it to conform to a non-default application binary interface.
12480 @item -fshort-wchar
12481 @opindex fshort-wchar
12482 Override the underlying type for @samp{wchar_t} to be @samp{short
12483 unsigned int} instead of the default for the target. This option is
12484 useful for building programs to run under WINE@.
12486 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12487 code that is not binary compatible with code generated without that switch.
12488 Use it to conform to a non-default application binary interface.
12490 @item -fshared-data
12491 @opindex fshared-data
12492 Requests that the data and non-@code{const} variables of this
12493 compilation be shared data rather than private data. The distinction
12494 makes sense only on certain operating systems, where shared data is
12495 shared between processes running the same program, while private data
12496 exists in one copy per process.
12499 @opindex fno-common
12500 In C, allocate even uninitialized global variables in the data section of the
12501 object file, rather than generating them as common blocks. This has the
12502 effect that if the same variable is declared (without @code{extern}) in
12503 two different compilations, you will get an error when you link them.
12504 The only reason this might be useful is if you wish to verify that the
12505 program will work on other systems which always work this way.
12509 Ignore the @samp{#ident} directive.
12511 @item -finhibit-size-directive
12512 @opindex finhibit-size-directive
12513 Don't output a @code{.size} assembler directive, or anything else that
12514 would cause trouble if the function is split in the middle, and the
12515 two halves are placed at locations far apart in memory. This option is
12516 used when compiling @file{crtstuff.c}; you should not need to use it
12519 @item -fverbose-asm
12520 @opindex fverbose-asm
12521 Put extra commentary information in the generated assembly code to
12522 make it more readable. This option is generally only of use to those
12523 who actually need to read the generated assembly code (perhaps while
12524 debugging the compiler itself).
12526 @option{-fno-verbose-asm}, the default, causes the
12527 extra information to be omitted and is useful when comparing two assembler
12532 @cindex global offset table
12534 Generate position-independent code (PIC) suitable for use in a shared
12535 library, if supported for the target machine. Such code accesses all
12536 constant addresses through a global offset table (GOT)@. The dynamic
12537 loader resolves the GOT entries when the program starts (the dynamic
12538 loader is not part of GCC; it is part of the operating system). If
12539 the GOT size for the linked executable exceeds a machine-specific
12540 maximum size, you get an error message from the linker indicating that
12541 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12542 instead. (These maximums are 8k on the SPARC and 32k
12543 on the m68k and RS/6000. The 386 has no such limit.)
12545 Position-independent code requires special support, and therefore works
12546 only on certain machines. For the 386, GCC supports PIC for System V
12547 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12548 position-independent.
12552 If supported for the target machine, emit position-independent code,
12553 suitable for dynamic linking and avoiding any limit on the size of the
12554 global offset table. This option makes a difference on the m68k,
12555 PowerPC and SPARC@.
12557 Position-independent code requires special support, and therefore works
12558 only on certain machines.
12564 These options are similar to @option{-fpic} and @option{-fPIC}, but
12565 generated position independent code can be only linked into executables.
12566 Usually these options are used when @option{-pie} GCC option will be
12567 used during linking.
12569 @item -ffixed-@var{reg}
12571 Treat the register named @var{reg} as a fixed register; generated code
12572 should never refer to it (except perhaps as a stack pointer, frame
12573 pointer or in some other fixed role).
12575 @var{reg} must be the name of a register. The register names accepted
12576 are machine-specific and are defined in the @code{REGISTER_NAMES}
12577 macro in the machine description macro file.
12579 This flag does not have a negative form, because it specifies a
12582 @item -fcall-used-@var{reg}
12583 @opindex fcall-used
12584 Treat the register named @var{reg} as an allocable register that is
12585 clobbered by function calls. It may be allocated for temporaries or
12586 variables that do not live across a call. Functions compiled this way
12587 will not save and restore the register @var{reg}.
12589 It is an error to used this flag with the frame pointer or stack pointer.
12590 Use of this flag for other registers that have fixed pervasive roles in
12591 the machine's execution model will produce disastrous results.
12593 This flag does not have a negative form, because it specifies a
12596 @item -fcall-saved-@var{reg}
12597 @opindex fcall-saved
12598 Treat the register named @var{reg} as an allocable register saved by
12599 functions. It may be allocated even for temporaries or variables that
12600 live across a call. Functions compiled this way will save and restore
12601 the register @var{reg} if they use it.
12603 It is an error to used this flag with the frame pointer or stack pointer.
12604 Use of this flag for other registers that have fixed pervasive roles in
12605 the machine's execution model will produce disastrous results.
12607 A different sort of disaster will result from the use of this flag for
12608 a register in which function values may be returned.
12610 This flag does not have a negative form, because it specifies a
12613 @item -fpack-struct[=@var{n}]
12614 @opindex fpack-struct
12615 Without a value specified, pack all structure members together without
12616 holes. When a value is specified (which must be a small power of two), pack
12617 structure members according to this value, representing the maximum
12618 alignment (that is, objects with default alignment requirements larger than
12619 this will be output potentially unaligned at the next fitting location.
12621 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12622 code that is not binary compatible with code generated without that switch.
12623 Additionally, it makes the code suboptimal.
12624 Use it to conform to a non-default application binary interface.
12626 @item -finstrument-functions
12627 @opindex finstrument-functions
12628 Generate instrumentation calls for entry and exit to functions. Just
12629 after function entry and just before function exit, the following
12630 profiling functions will be called with the address of the current
12631 function and its call site. (On some platforms,
12632 @code{__builtin_return_address} does not work beyond the current
12633 function, so the call site information may not be available to the
12634 profiling functions otherwise.)
12637 void __cyg_profile_func_enter (void *this_fn,
12639 void __cyg_profile_func_exit (void *this_fn,
12643 The first argument is the address of the start of the current function,
12644 which may be looked up exactly in the symbol table.
12646 This instrumentation is also done for functions expanded inline in other
12647 functions. The profiling calls will indicate where, conceptually, the
12648 inline function is entered and exited. This means that addressable
12649 versions of such functions must be available. If all your uses of a
12650 function are expanded inline, this may mean an additional expansion of
12651 code size. If you use @samp{extern inline} in your C code, an
12652 addressable version of such functions must be provided. (This is
12653 normally the case anyways, but if you get lucky and the optimizer always
12654 expands the functions inline, you might have gotten away without
12655 providing static copies.)
12657 A function may be given the attribute @code{no_instrument_function}, in
12658 which case this instrumentation will not be done. This can be used, for
12659 example, for the profiling functions listed above, high-priority
12660 interrupt routines, and any functions from which the profiling functions
12661 cannot safely be called (perhaps signal handlers, if the profiling
12662 routines generate output or allocate memory).
12664 @item -fstack-check
12665 @opindex fstack-check
12666 Generate code to verify that you do not go beyond the boundary of the
12667 stack. You should specify this flag if you are running in an
12668 environment with multiple threads, but only rarely need to specify it in
12669 a single-threaded environment since stack overflow is automatically
12670 detected on nearly all systems if there is only one stack.
12672 Note that this switch does not actually cause checking to be done; the
12673 operating system must do that. The switch causes generation of code
12674 to ensure that the operating system sees the stack being extended.
12676 @item -fstack-limit-register=@var{reg}
12677 @itemx -fstack-limit-symbol=@var{sym}
12678 @itemx -fno-stack-limit
12679 @opindex fstack-limit-register
12680 @opindex fstack-limit-symbol
12681 @opindex fno-stack-limit
12682 Generate code to ensure that the stack does not grow beyond a certain value,
12683 either the value of a register or the address of a symbol. If the stack
12684 would grow beyond the value, a signal is raised. For most targets,
12685 the signal is raised before the stack overruns the boundary, so
12686 it is possible to catch the signal without taking special precautions.
12688 For instance, if the stack starts at absolute address @samp{0x80000000}
12689 and grows downwards, you can use the flags
12690 @option{-fstack-limit-symbol=__stack_limit} and
12691 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12692 of 128KB@. Note that this may only work with the GNU linker.
12694 @cindex aliasing of parameters
12695 @cindex parameters, aliased
12696 @item -fargument-alias
12697 @itemx -fargument-noalias
12698 @itemx -fargument-noalias-global
12699 @opindex fargument-alias
12700 @opindex fargument-noalias
12701 @opindex fargument-noalias-global
12702 Specify the possible relationships among parameters and between
12703 parameters and global data.
12705 @option{-fargument-alias} specifies that arguments (parameters) may
12706 alias each other and may alias global storage.@*
12707 @option{-fargument-noalias} specifies that arguments do not alias
12708 each other, but may alias global storage.@*
12709 @option{-fargument-noalias-global} specifies that arguments do not
12710 alias each other and do not alias global storage.
12712 Each language will automatically use whatever option is required by
12713 the language standard. You should not need to use these options yourself.
12715 @item -fleading-underscore
12716 @opindex fleading-underscore
12717 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12718 change the way C symbols are represented in the object file. One use
12719 is to help link with legacy assembly code.
12721 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12722 generate code that is not binary compatible with code generated without that
12723 switch. Use it to conform to a non-default application binary interface.
12724 Not all targets provide complete support for this switch.
12726 @item -ftls-model=@var{model}
12727 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12728 The @var{model} argument should be one of @code{global-dynamic},
12729 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
12731 The default without @option{-fpic} is @code{initial-exec}; with
12732 @option{-fpic} the default is @code{global-dynamic}.
12734 @item -fvisibility=@var{default|internal|hidden|protected}
12735 @opindex fvisibility
12736 Set the default ELF image symbol visibility to the specified option---all
12737 symbols will be marked with this unless overridden within the code.
12738 Using this feature can very substantially improve linking and
12739 load times of shared object libraries, produce more optimized
12740 code, provide near-perfect API export and prevent symbol clashes.
12741 It is @strong{strongly} recommended that you use this in any shared objects
12744 Despite the nomenclature, @code{default} always means public ie;
12745 available to be linked against from outside the shared object.
12746 @code{protected} and @code{internal} are pretty useless in real-world
12747 usage so the only other commonly used option will be @code{hidden}.
12748 The default if @option{-fvisibility} isn't specified is
12749 @code{default}, i.e., make every
12750 symbol public---this causes the same behavior as previous versions of
12753 A good explanation of the benefits offered by ensuring ELF
12754 symbols have the correct visibility is given by ``How To Write
12755 Shared Libraries'' by Ulrich Drepper (which can be found at
12756 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
12757 solution made possible by this option to marking things hidden when
12758 the default is public is to make the default hidden and mark things
12759 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
12760 and @code{__attribute__ ((visibility("default")))} instead of
12761 @code{__declspec(dllexport)} you get almost identical semantics with
12762 identical syntax. This is a great boon to those working with
12763 cross-platform projects.
12765 For those adding visibility support to existing code, you may find
12766 @samp{#pragma GCC visibility} of use. This works by you enclosing
12767 the declarations you wish to set visibility for with (for example)
12768 @samp{#pragma GCC visibility push(hidden)} and
12769 @samp{#pragma GCC visibility pop}. These can be nested up to sixteen
12770 times. Bear in mind that symbol visibility should be viewed @strong{as
12771 part of the API interface contract} and thus all new code should
12772 always specify visibility when it is not the default ie; declarations
12773 only for use within the local DSO should @strong{always} be marked explicitly
12774 as hidden as so to avoid PLT indirection overheads---making this
12775 abundantly clear also aids readability and self-documentation of the code.
12776 Note that due to ISO C++ specification requirements, operator new and
12777 operator delete must always be of default visibility.
12779 An overview of these techniques, their benefits and how to use them
12780 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
12786 @node Environment Variables
12787 @section Environment Variables Affecting GCC
12788 @cindex environment variables
12790 @c man begin ENVIRONMENT
12791 This section describes several environment variables that affect how GCC
12792 operates. Some of them work by specifying directories or prefixes to use
12793 when searching for various kinds of files. Some are used to specify other
12794 aspects of the compilation environment.
12796 Note that you can also specify places to search using options such as
12797 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
12798 take precedence over places specified using environment variables, which
12799 in turn take precedence over those specified by the configuration of GCC@.
12800 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
12801 GNU Compiler Collection (GCC) Internals}.
12806 @c @itemx LC_COLLATE
12808 @c @itemx LC_MONETARY
12809 @c @itemx LC_NUMERIC
12814 @c @findex LC_COLLATE
12815 @findex LC_MESSAGES
12816 @c @findex LC_MONETARY
12817 @c @findex LC_NUMERIC
12821 These environment variables control the way that GCC uses
12822 localization information that allow GCC to work with different
12823 national conventions. GCC inspects the locale categories
12824 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
12825 so. These locale categories can be set to any value supported by your
12826 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
12827 Kingdom encoded in UTF-8.
12829 The @env{LC_CTYPE} environment variable specifies character
12830 classification. GCC uses it to determine the character boundaries in
12831 a string; this is needed for some multibyte encodings that contain quote
12832 and escape characters that would otherwise be interpreted as a string
12835 The @env{LC_MESSAGES} environment variable specifies the language to
12836 use in diagnostic messages.
12838 If the @env{LC_ALL} environment variable is set, it overrides the value
12839 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
12840 and @env{LC_MESSAGES} default to the value of the @env{LANG}
12841 environment variable. If none of these variables are set, GCC
12842 defaults to traditional C English behavior.
12846 If @env{TMPDIR} is set, it specifies the directory to use for temporary
12847 files. GCC uses temporary files to hold the output of one stage of
12848 compilation which is to be used as input to the next stage: for example,
12849 the output of the preprocessor, which is the input to the compiler
12852 @item GCC_EXEC_PREFIX
12853 @findex GCC_EXEC_PREFIX
12854 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
12855 names of the subprograms executed by the compiler. No slash is added
12856 when this prefix is combined with the name of a subprogram, but you can
12857 specify a prefix that ends with a slash if you wish.
12859 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
12860 an appropriate prefix to use based on the pathname it was invoked with.
12862 If GCC cannot find the subprogram using the specified prefix, it
12863 tries looking in the usual places for the subprogram.
12865 The default value of @env{GCC_EXEC_PREFIX} is
12866 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
12867 of @code{prefix} when you ran the @file{configure} script.
12869 Other prefixes specified with @option{-B} take precedence over this prefix.
12871 This prefix is also used for finding files such as @file{crt0.o} that are
12874 In addition, the prefix is used in an unusual way in finding the
12875 directories to search for header files. For each of the standard
12876 directories whose name normally begins with @samp{/usr/local/lib/gcc}
12877 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
12878 replacing that beginning with the specified prefix to produce an
12879 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
12880 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
12881 These alternate directories are searched first; the standard directories
12884 @item COMPILER_PATH
12885 @findex COMPILER_PATH
12886 The value of @env{COMPILER_PATH} is a colon-separated list of
12887 directories, much like @env{PATH}. GCC tries the directories thus
12888 specified when searching for subprograms, if it can't find the
12889 subprograms using @env{GCC_EXEC_PREFIX}.
12892 @findex LIBRARY_PATH
12893 The value of @env{LIBRARY_PATH} is a colon-separated list of
12894 directories, much like @env{PATH}. When configured as a native compiler,
12895 GCC tries the directories thus specified when searching for special
12896 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
12897 using GCC also uses these directories when searching for ordinary
12898 libraries for the @option{-l} option (but directories specified with
12899 @option{-L} come first).
12903 @cindex locale definition
12904 This variable is used to pass locale information to the compiler. One way in
12905 which this information is used is to determine the character set to be used
12906 when character literals, string literals and comments are parsed in C and C++.
12907 When the compiler is configured to allow multibyte characters,
12908 the following values for @env{LANG} are recognized:
12912 Recognize JIS characters.
12914 Recognize SJIS characters.
12916 Recognize EUCJP characters.
12919 If @env{LANG} is not defined, or if it has some other value, then the
12920 compiler will use mblen and mbtowc as defined by the default locale to
12921 recognize and translate multibyte characters.
12925 Some additional environments variables affect the behavior of the
12928 @include cppenv.texi
12932 @node Precompiled Headers
12933 @section Using Precompiled Headers
12934 @cindex precompiled headers
12935 @cindex speed of compilation
12937 Often large projects have many header files that are included in every
12938 source file. The time the compiler takes to process these header files
12939 over and over again can account for nearly all of the time required to
12940 build the project. To make builds faster, GCC allows users to
12941 `precompile' a header file; then, if builds can use the precompiled
12942 header file they will be much faster.
12944 To create a precompiled header file, simply compile it as you would any
12945 other file, if necessary using the @option{-x} option to make the driver
12946 treat it as a C or C++ header file. You will probably want to use a
12947 tool like @command{make} to keep the precompiled header up-to-date when
12948 the headers it contains change.
12950 A precompiled header file will be searched for when @code{#include} is
12951 seen in the compilation. As it searches for the included file
12952 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
12953 compiler looks for a precompiled header in each directory just before it
12954 looks for the include file in that directory. The name searched for is
12955 the name specified in the @code{#include} with @samp{.gch} appended. If
12956 the precompiled header file can't be used, it is ignored.
12958 For instance, if you have @code{#include "all.h"}, and you have
12959 @file{all.h.gch} in the same directory as @file{all.h}, then the
12960 precompiled header file will be used if possible, and the original
12961 header will be used otherwise.
12963 Alternatively, you might decide to put the precompiled header file in a
12964 directory and use @option{-I} to ensure that directory is searched
12965 before (or instead of) the directory containing the original header.
12966 Then, if you want to check that the precompiled header file is always
12967 used, you can put a file of the same name as the original header in this
12968 directory containing an @code{#error} command.
12970 This also works with @option{-include}. So yet another way to use
12971 precompiled headers, good for projects not designed with precompiled
12972 header files in mind, is to simply take most of the header files used by
12973 a project, include them from another header file, precompile that header
12974 file, and @option{-include} the precompiled header. If the header files
12975 have guards against multiple inclusion, they will be skipped because
12976 they've already been included (in the precompiled header).
12978 If you need to precompile the same header file for different
12979 languages, targets, or compiler options, you can instead make a
12980 @emph{directory} named like @file{all.h.gch}, and put each precompiled
12981 header in the directory, perhaps using @option{-o}. It doesn't matter
12982 what you call the files in the directory, every precompiled header in
12983 the directory will be considered. The first precompiled header
12984 encountered in the directory that is valid for this compilation will
12985 be used; they're searched in no particular order.
12987 There are many other possibilities, limited only by your imagination,
12988 good sense, and the constraints of your build system.
12990 A precompiled header file can be used only when these conditions apply:
12994 Only one precompiled header can be used in a particular compilation.
12997 A precompiled header can't be used once the first C token is seen. You
12998 can have preprocessor directives before a precompiled header; you can
12999 even include a precompiled header from inside another header, so long as
13000 there are no C tokens before the @code{#include}.
13003 The precompiled header file must be produced for the same language as
13004 the current compilation. You can't use a C precompiled header for a C++
13008 The precompiled header file must have been produced by the same compiler
13009 binary as the current compilation is using.
13012 Any macros defined before the precompiled header is included must
13013 either be defined in the same way as when the precompiled header was
13014 generated, or must not affect the precompiled header, which usually
13015 means that they don't appear in the precompiled header at all.
13017 The @option{-D} option is one way to define a macro before a
13018 precompiled header is included; using a @code{#define} can also do it.
13019 There are also some options that define macros implicitly, like
13020 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
13023 @item If debugging information is output when using the precompiled
13024 header, using @option{-g} or similar, the same kind of debugging information
13025 must have been output when building the precompiled header. However,
13026 a precompiled header built using @option{-g} can be used in a compilation
13027 when no debugging information is being output.
13029 @item The same @option{-m} options must generally be used when building
13030 and using the precompiled header. @xref{Submodel Options},
13031 for any cases where this rule is relaxed.
13033 @item Each of the following options must be the same when building and using
13034 the precompiled header:
13036 @gccoptlist{-fexceptions -funit-at-a-time}
13039 Some other command-line options starting with @option{-f},
13040 @option{-p}, or @option{-O} must be defined in the same way as when
13041 the precompiled header was generated. At present, it's not clear
13042 which options are safe to change and which are not; the safest choice
13043 is to use exactly the same options when generating and using the
13044 precompiled header. The following are known to be safe:
13046 @gccoptlist{-fpreprocessed
13047 -fsched-interblock -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
13048 -fsched-verbose=<number> -fschedule-insns
13053 For all of these except the last, the compiler will automatically
13054 ignore the precompiled header if the conditions aren't met. If you
13055 find an option combination that doesn't work and doesn't cause the
13056 precompiled header to be ignored, please consider filing a bug report,
13059 If you do use differing options when generating and using the
13060 precompiled header, the actual behavior will be a mixture of the
13061 behavior for the options. For instance, if you use @option{-g} to
13062 generate the precompiled header but not when using it, you may or may
13063 not get debugging information for routines in the precompiled header.
13065 @node Running Protoize
13066 @section Running Protoize
13068 The program @code{protoize} is an optional part of GCC@. You can use
13069 it to add prototypes to a program, thus converting the program to ISO
13070 C in one respect. The companion program @code{unprotoize} does the
13071 reverse: it removes argument types from any prototypes that are found.
13073 When you run these programs, you must specify a set of source files as
13074 command line arguments. The conversion programs start out by compiling
13075 these files to see what functions they define. The information gathered
13076 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
13078 After scanning comes actual conversion. The specified files are all
13079 eligible to be converted; any files they include (whether sources or
13080 just headers) are eligible as well.
13082 But not all the eligible files are converted. By default,
13083 @code{protoize} and @code{unprotoize} convert only source and header
13084 files in the current directory. You can specify additional directories
13085 whose files should be converted with the @option{-d @var{directory}}
13086 option. You can also specify particular files to exclude with the
13087 @option{-x @var{file}} option. A file is converted if it is eligible, its
13088 directory name matches one of the specified directory names, and its
13089 name within the directory has not been excluded.
13091 Basic conversion with @code{protoize} consists of rewriting most
13092 function definitions and function declarations to specify the types of
13093 the arguments. The only ones not rewritten are those for varargs
13096 @code{protoize} optionally inserts prototype declarations at the
13097 beginning of the source file, to make them available for any calls that
13098 precede the function's definition. Or it can insert prototype
13099 declarations with block scope in the blocks where undeclared functions
13102 Basic conversion with @code{unprotoize} consists of rewriting most
13103 function declarations to remove any argument types, and rewriting
13104 function definitions to the old-style pre-ISO form.
13106 Both conversion programs print a warning for any function declaration or
13107 definition that they can't convert. You can suppress these warnings
13110 The output from @code{protoize} or @code{unprotoize} replaces the
13111 original source file. The original file is renamed to a name ending
13112 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
13113 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
13114 for DOS) file already exists, then the source file is simply discarded.
13116 @code{protoize} and @code{unprotoize} both depend on GCC itself to
13117 scan the program and collect information about the functions it uses.
13118 So neither of these programs will work until GCC is installed.
13120 Here is a table of the options you can use with @code{protoize} and
13121 @code{unprotoize}. Each option works with both programs unless
13125 @item -B @var{directory}
13126 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
13127 usual directory (normally @file{/usr/local/lib}). This file contains
13128 prototype information about standard system functions. This option
13129 applies only to @code{protoize}.
13131 @item -c @var{compilation-options}
13132 Use @var{compilation-options} as the options when running @command{gcc} to
13133 produce the @samp{.X} files. The special option @option{-aux-info} is
13134 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
13136 Note that the compilation options must be given as a single argument to
13137 @code{protoize} or @code{unprotoize}. If you want to specify several
13138 @command{gcc} options, you must quote the entire set of compilation options
13139 to make them a single word in the shell.
13141 There are certain @command{gcc} arguments that you cannot use, because they
13142 would produce the wrong kind of output. These include @option{-g},
13143 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
13144 the @var{compilation-options}, they are ignored.
13147 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
13148 systems) instead of @samp{.c}. This is convenient if you are converting
13149 a C program to C++. This option applies only to @code{protoize}.
13152 Add explicit global declarations. This means inserting explicit
13153 declarations at the beginning of each source file for each function
13154 that is called in the file and was not declared. These declarations
13155 precede the first function definition that contains a call to an
13156 undeclared function. This option applies only to @code{protoize}.
13158 @item -i @var{string}
13159 Indent old-style parameter declarations with the string @var{string}.
13160 This option applies only to @code{protoize}.
13162 @code{unprotoize} converts prototyped function definitions to old-style
13163 function definitions, where the arguments are declared between the
13164 argument list and the initial @samp{@{}. By default, @code{unprotoize}
13165 uses five spaces as the indentation. If you want to indent with just
13166 one space instead, use @option{-i " "}.
13169 Keep the @samp{.X} files. Normally, they are deleted after conversion
13173 Add explicit local declarations. @code{protoize} with @option{-l} inserts
13174 a prototype declaration for each function in each block which calls the
13175 function without any declaration. This option applies only to
13179 Make no real changes. This mode just prints information about the conversions
13180 that would have been done without @option{-n}.
13183 Make no @samp{.save} files. The original files are simply deleted.
13184 Use this option with caution.
13186 @item -p @var{program}
13187 Use the program @var{program} as the compiler. Normally, the name
13188 @file{gcc} is used.
13191 Work quietly. Most warnings are suppressed.
13194 Print the version number, just like @option{-v} for @command{gcc}.
13197 If you need special compiler options to compile one of your program's
13198 source files, then you should generate that file's @samp{.X} file
13199 specially, by running @command{gcc} on that source file with the
13200 appropriate options and the option @option{-aux-info}. Then run
13201 @code{protoize} on the entire set of files. @code{protoize} will use
13202 the existing @samp{.X} file because it is newer than the source file.
13206 gcc -Dfoo=bar file1.c -aux-info file1.X
13211 You need to include the special files along with the rest in the
13212 @code{protoize} command, even though their @samp{.X} files already
13213 exist, because otherwise they won't get converted.
13215 @xref{Protoize Caveats}, for more information on how to use
13216 @code{protoize} successfully.