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-exceptions @gol
201 -freplace-objc-classes @gol
204 -Wno-protocol -Wselector -Wundeclared-selector}
206 @item Language Independent Options
207 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
208 @gccoptlist{-fmessage-length=@var{n} @gol
209 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}} @gol
210 -fdiagnostics-show-options
212 @item Warning Options
213 @xref{Warning Options,,Options to Request or Suppress Warnings}.
214 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
215 -w -Wextra -Wall -Waggregate-return @gol
216 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
217 -Wconversion -Wno-deprecated-declarations @gol
218 -Wdisabled-optimization -Wno-div-by-zero -Wno-endif-labels @gol
219 -Werror -Werror-implicit-function-declaration @gol
220 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
221 -Wno-format-extra-args -Wformat-nonliteral @gol
222 -Wformat-security -Wformat-y2k @gol
223 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
224 -Wimport -Wno-import -Winit-self -Winline @gol
225 -Wno-int-to-pointer-cast @gol
226 -Wno-invalid-offsetof -Winvalid-pch @gol
227 -Wlarger-than-@var{len} -Wlong-long @gol
228 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
229 -Wmissing-format-attribute -Wmissing-include-dirs @gol
230 -Wmissing-noreturn @gol
231 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
232 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
233 -Wredundant-decls @gol
234 -Wreturn-type -Wsequence-point -Wshadow @gol
235 -Wsign-compare -Wstrict-aliasing -Wstrict-aliasing=2 @gol
236 -Wswitch -Wswitch-default -Wswitch-enum @gol
237 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
238 -Wunknown-pragmas -Wunreachable-code @gol
239 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
240 -Wunused-value -Wunused-variable -Wwrite-strings @gol
243 @item C-only Warning Options
244 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
245 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
246 -Wstrict-prototypes -Wtraditional @gol
247 -Wdeclaration-after-statement -Wno-pointer-sign}
249 @item Debugging Options
250 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
251 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
252 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
253 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
254 -fdump-ipa-all -fdump-ipa-cgraph @gol
256 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
257 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
258 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
259 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
261 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
262 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
263 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
264 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
266 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
267 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
268 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
269 -fdump-tree-nrv -fdump-tree-vect @gol
270 -fdump-tree-sink @gol
271 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
272 -fdump-tree-salias @gol
273 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
274 -ftree-vectorizer-verbose=@var{n} @gol
275 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
276 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
277 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs -ftree-based-profiling @gol
278 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
279 -ftest-coverage -ftime-report -fvar-tracking @gol
280 -g -g@var{level} -gcoff -gdwarf-2 @gol
281 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
282 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
283 -print-multi-directory -print-multi-lib @gol
284 -print-prog-name=@var{program} -print-search-dirs -Q @gol
287 @item Optimization Options
288 @xref{Optimize Options,,Options that Control Optimization}.
289 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
290 -falign-labels=@var{n} -falign-loops=@var{n} @gol
291 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
292 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
293 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
294 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
295 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
296 -fdelayed-branch -fdelete-null-pointer-checks @gol
297 -fexpensive-optimizations -ffast-math -ffloat-store @gol
298 -fforce-addr -fforce-mem -ffunction-sections @gol
299 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
300 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2 @gol
301 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
302 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
303 -fmodulo-sched -fno-branch-count-reg @gol
304 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants @gol
305 -fno-function-cse -fno-guess-branch-probability @gol
306 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
307 -funsafe-math-optimizations -ffinite-math-only @gol
308 -fno-trapping-math -fno-zero-initialized-in-bss @gol
309 -fomit-frame-pointer -foptimize-register-move @gol
310 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
311 -fprofile-generate -fprofile-use @gol
312 -fregmove -frename-registers @gol
313 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
314 -frerun-cse-after-loop -frerun-loop-opt @gol
315 -frounding-math -fschedule-insns -fschedule-insns2 @gol
316 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
317 -fsched-spec-load-dangerous @gol
318 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
319 -fsched2-use-superblocks @gol
320 -fsched2-use-traces -freschedule-modulo-scheduled-loops @gol
321 -fsignaling-nans -fsingle-precision-constant -fspeculative-prefetching @gol
322 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
323 -funroll-all-loops -funroll-loops -fpeel-loops @gol
324 -fsplit-ivs-in-unroller -funswitch-loops @gol
325 -fvariable-expansion-in-unroller @gol
326 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
327 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
328 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
329 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
330 -ftree-salias -fweb @gol
331 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop @gol
332 --param @var{name}=@var{value}
333 -O -O0 -O1 -O2 -O3 -Os}
335 @item Preprocessor Options
336 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
337 @gccoptlist{-A@var{question}=@var{answer} @gol
338 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
339 -C -dD -dI -dM -dN @gol
340 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
341 -idirafter @var{dir} @gol
342 -include @var{file} -imacros @var{file} @gol
343 -iprefix @var{file} -iwithprefix @var{dir} @gol
344 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
345 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
346 -P -fworking-directory -remap @gol
347 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
348 -Xpreprocessor @var{option}}
350 @item Assembler Option
351 @xref{Assembler Options,,Passing Options to the Assembler}.
352 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
355 @xref{Link Options,,Options for Linking}.
356 @gccoptlist{@var{object-file-name} -l@var{library} @gol
357 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
358 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
359 -Wl,@var{option} -Xlinker @var{option} @gol
362 @item Directory Options
363 @xref{Directory Options,,Options for Directory Search}.
364 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-}
367 @c I wrote this xref this way to avoid overfull hbox. -- rms
368 @xref{Target Options}.
369 @gccoptlist{-V @var{version} -b @var{machine}}
371 @item Machine Dependent Options
372 @xref{Submodel Options,,Hardware Models and Configurations}.
373 @c This list is ordered alphanumerically by subsection name.
374 @c Try and put the significant identifier (CPU or system) first,
375 @c so users have a clue at guessing where the ones they want will be.
378 @gccoptlist{-EB -EL @gol
379 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
380 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
383 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
384 -mabi=@var{name} @gol
385 -mapcs-stack-check -mno-apcs-stack-check @gol
386 -mapcs-float -mno-apcs-float @gol
387 -mapcs-reentrant -mno-apcs-reentrant @gol
388 -msched-prolog -mno-sched-prolog @gol
389 -mlittle-endian -mbig-endian -mwords-little-endian @gol
390 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
391 -mthumb-interwork -mno-thumb-interwork @gol
392 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
393 -mstructure-size-boundary=@var{n} @gol
394 -mabort-on-noreturn @gol
395 -mlong-calls -mno-long-calls @gol
396 -msingle-pic-base -mno-single-pic-base @gol
397 -mpic-register=@var{reg} @gol
398 -mnop-fun-dllimport @gol
399 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
400 -mpoke-function-name @gol
402 -mtpcs-frame -mtpcs-leaf-frame @gol
403 -mcaller-super-interworking -mcallee-super-interworking}
406 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
407 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
409 @emph{Blackfin Options}
410 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer -mcsync @gol
411 -mno-csync -mlow-64k -mno-low64k -mid-shared-library @gol
412 -mno-id-shared-library -mshared-library-id=@var{n} @gol}
415 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
416 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
417 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
418 -mstack-align -mdata-align -mconst-align @gol
419 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
420 -melf -maout -melinux -mlinux -sim -sim2 @gol
421 -mmul-bug-workaround -mno-mul-bug-workaround}
423 @emph{Darwin Options}
424 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
425 -arch_only -bind_at_load -bundle -bundle_loader @gol
426 -client_name -compatibility_version -current_version @gol
428 -dependency-file -dylib_file -dylinker_install_name @gol
429 -dynamic -dynamiclib -exported_symbols_list @gol
430 -filelist -flat_namespace -force_cpusubtype_ALL @gol
431 -force_flat_namespace -headerpad_max_install_names @gol
432 -image_base -init -install_name -keep_private_externs @gol
433 -multi_module -multiply_defined -multiply_defined_unused @gol
434 -noall_load -no_dead_strip_inits_and_terms @gol
435 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
436 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
437 -private_bundle -read_only_relocs -sectalign @gol
438 -sectobjectsymbols -whyload -seg1addr @gol
439 -sectcreate -sectobjectsymbols -sectorder @gol
440 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
441 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
442 -segprot -segs_read_only_addr -segs_read_write_addr @gol
443 -single_module -static -sub_library -sub_umbrella @gol
444 -twolevel_namespace -umbrella -undefined @gol
445 -unexported_symbols_list -weak_reference_mismatches @gol
446 -whatsloaded -F -gused -gfull -mone-byte-bool}
448 @emph{DEC Alpha Options}
449 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
450 -mieee -mieee-with-inexact -mieee-conformant @gol
451 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
452 -mtrap-precision=@var{mode} -mbuild-constants @gol
453 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
454 -mbwx -mmax -mfix -mcix @gol
455 -mfloat-vax -mfloat-ieee @gol
456 -mexplicit-relocs -msmall-data -mlarge-data @gol
457 -msmall-text -mlarge-text @gol
458 -mmemory-latency=@var{time}}
460 @emph{DEC Alpha/VMS Options}
461 @gccoptlist{-mvms-return-codes}
464 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
465 -mhard-float -msoft-float @gol
466 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
467 -mdouble -mno-double @gol
468 -mmedia -mno-media -mmuladd -mno-muladd @gol
469 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
470 -mlinked-fp -mlong-calls -malign-labels @gol
471 -mlibrary-pic -macc-4 -macc-8 @gol
472 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
473 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
474 -mvliw-branch -mno-vliw-branch @gol
475 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
476 -mno-nested-cond-exec -mtomcat-stats @gol
480 @emph{H8/300 Options}
481 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
484 @gccoptlist{-march=@var{architecture-type} @gol
485 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
486 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
487 -mfixed-range=@var{register-range} @gol
488 -mjump-in-delay -mlinker-opt -mlong-calls @gol
489 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
490 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
491 -mno-jump-in-delay -mno-long-load-store @gol
492 -mno-portable-runtime -mno-soft-float @gol
493 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
494 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
495 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
496 -munix=@var{unix-std} -nolibdld -static -threads}
498 @emph{i386 and x86-64 Options}
499 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
500 -mfpmath=@var{unit} @gol
501 -masm=@var{dialect} -mno-fancy-math-387 @gol
502 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
503 -mno-wide-multiply -mrtd -malign-double @gol
504 -mpreferred-stack-boundary=@var{num} @gol
505 -mmmx -msse -msse2 -msse3 -m3dnow @gol
506 -mthreads -mno-align-stringops -minline-all-stringops @gol
507 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
508 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
509 -mno-red-zone -mno-tls-direct-seg-refs @gol
510 -mcmodel=@var{code-model} @gol
514 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
515 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
516 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
517 -minline-float-divide-max-throughput @gol
518 -minline-int-divide-min-latency @gol
519 -minline-int-divide-max-throughput @gol
520 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
521 -mno-dwarf2-asm -mearly-stop-bits @gol
522 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
523 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64}
525 @emph{M32R/D Options}
526 @gccoptlist{-m32r2 -m32rx -m32r @gol
528 -malign-loops -mno-align-loops @gol
529 -missue-rate=@var{number} @gol
530 -mbranch-cost=@var{number} @gol
531 -mmodel=@var{code-size-model-type} @gol
532 -msdata=@var{sdata-type} @gol
533 -mno-flush-func -mflush-func=@var{name} @gol
534 -mno-flush-trap -mflush-trap=@var{number} @gol
537 @emph{M680x0 Options}
538 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
539 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
540 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
541 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
542 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
544 @emph{M68hc1x Options}
545 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
546 -mauto-incdec -minmax -mlong-calls -mshort @gol
547 -msoft-reg-count=@var{count}}
550 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
551 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
552 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
553 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
554 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
557 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
558 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
559 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
560 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
561 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
562 -mpaired-single -mips3d @gol
563 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
564 -G@var{num} -membedded-data -mno-embedded-data @gol
565 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
566 -msplit-addresses -mno-split-addresses @gol
567 -mexplicit-relocs -mno-explicit-relocs @gol
568 -mcheck-zero-division -mno-check-zero-division @gol
569 -mdivide-traps -mdivide-breaks @gol
570 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
571 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
572 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
573 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
574 -mfix-sb1 -mno-fix-sb1 @gol
575 -mflush-func=@var{func} -mno-flush-func @gol
576 -mbranch-likely -mno-branch-likely @gol
577 -mfp-exceptions -mno-fp-exceptions @gol
578 -mvr4130-align -mno-vr4130-align}
581 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
582 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
583 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
584 -mno-base-addresses -msingle-exit -mno-single-exit}
586 @emph{MN10300 Options}
587 @gccoptlist{-mmult-bug -mno-mult-bug @gol
588 -mam33 -mno-am33 @gol
589 -mam33-2 -mno-am33-2 @gol
593 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
594 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
595 -mregparam -mnoregparam -msb -mnosb @gol
596 -mbitfield -mnobitfield -mhimem -mnohimem}
598 @emph{PDP-11 Options}
599 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
600 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
601 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
602 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
603 -mbranch-expensive -mbranch-cheap @gol
604 -msplit -mno-split -munix-asm -mdec-asm}
606 @emph{PowerPC Options}
607 See RS/6000 and PowerPC Options.
609 @emph{RS/6000 and PowerPC Options}
610 @gccoptlist{-mcpu=@var{cpu-type} @gol
611 -mtune=@var{cpu-type} @gol
612 -mpower -mno-power -mpower2 -mno-power2 @gol
613 -mpowerpc -mpowerpc64 -mno-powerpc @gol
614 -maltivec -mno-altivec @gol
615 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
616 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
617 -mnew-mnemonics -mold-mnemonics @gol
618 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
619 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
620 -malign-power -malign-natural @gol
621 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
622 -mstring -mno-string -mupdate -mno-update @gol
623 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
624 -mstrict-align -mno-strict-align -mrelocatable @gol
625 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
626 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
627 -mdynamic-no-pic @gol
628 -mprioritize-restricted-insns=@var{priority} @gol
629 -msched-costly-dep=@var{dependence_type} @gol
630 -minsert-sched-nops=@var{scheme} @gol
631 -mcall-sysv -mcall-netbsd @gol
632 -maix-struct-return -msvr4-struct-return @gol
633 -mabi=@var{abi-type} @gol
634 -misel -mno-isel @gol
635 -misel=yes -misel=no @gol
637 -mspe=yes -mspe=no @gol
638 -mvrsave -mno-vrsave @gol
639 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
640 -mprototype -mno-prototype @gol
641 -msim -mmvme -mads -myellowknife -memb -msdata @gol
642 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
644 @emph{S/390 and zSeries Options}
645 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
646 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
647 -mpacked-stack -mno-packed-stack @gol
648 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
649 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
650 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
651 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
654 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
655 -m4-nofpu -m4-single-only -m4-single -m4 @gol
656 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
657 -m5-64media -m5-64media-nofpu @gol
658 -m5-32media -m5-32media-nofpu @gol
659 -m5-compact -m5-compact-nofpu @gol
660 -mb -ml -mdalign -mrelax @gol
661 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
662 -mieee -misize -mpadstruct -mspace @gol
663 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
664 -mdivsi3_libfunc=@var{name} @gol
665 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
669 @gccoptlist{-mcpu=@var{cpu-type} @gol
670 -mtune=@var{cpu-type} @gol
671 -mcmodel=@var{code-model} @gol
672 -m32 -m64 -mapp-regs -mno-app-regs @gol
673 -mfaster-structs -mno-faster-structs @gol
674 -mfpu -mno-fpu -mhard-float -msoft-float @gol
675 -mhard-quad-float -msoft-quad-float @gol
676 -mimpure-text -mno-impure-text -mlittle-endian @gol
677 -mstack-bias -mno-stack-bias @gol
678 -munaligned-doubles -mno-unaligned-doubles @gol
679 -mv8plus -mno-v8plus -mvis -mno-vis
682 @emph{System V Options}
683 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
685 @emph{TMS320C3x/C4x Options}
686 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
687 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
688 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
689 -mparallel-insns -mparallel-mpy -mpreserve-float}
692 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
693 -mprolog-function -mno-prolog-function -mspace @gol
694 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
695 -mapp-regs -mno-app-regs @gol
696 -mdisable-callt -mno-disable-callt @gol
702 @gccoptlist{-mg -mgnu -munix}
704 @emph{x86-64 Options}
705 See i386 and x86-64 Options.
707 @emph{Xstormy16 Options}
710 @emph{Xtensa Options}
711 @gccoptlist{-mconst16 -mno-const16 @gol
712 -mfused-madd -mno-fused-madd @gol
713 -mtext-section-literals -mno-text-section-literals @gol
714 -mtarget-align -mno-target-align @gol
715 -mlongcalls -mno-longcalls}
717 @emph{zSeries Options}
718 See S/390 and zSeries Options.
720 @item Code Generation Options
721 @xref{Code Gen Options,,Options for Code Generation Conventions}.
722 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
723 -ffixed-@var{reg} -fexceptions @gol
724 -fnon-call-exceptions -funwind-tables @gol
725 -fasynchronous-unwind-tables @gol
726 -finhibit-size-directive -finstrument-functions @gol
727 -fno-common -fno-ident @gol
728 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
729 -freg-struct-return -fshared-data -fshort-enums @gol
730 -fshort-double -fshort-wchar @gol
731 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
732 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
733 -fargument-alias -fargument-noalias @gol
734 -fargument-noalias-global -fleading-underscore @gol
735 -ftls-model=@var{model} @gol
736 -ftrapv -fwrapv -fbounds-check @gol
741 * Overall Options:: Controlling the kind of output:
742 an executable, object files, assembler files,
743 or preprocessed source.
744 * C Dialect Options:: Controlling the variant of C language compiled.
745 * C++ Dialect Options:: Variations on C++.
746 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
748 * Language Independent Options:: Controlling how diagnostics should be
750 * Warning Options:: How picky should the compiler be?
751 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
752 * Optimize Options:: How much optimization?
753 * Preprocessor Options:: Controlling header files and macro definitions.
754 Also, getting dependency information for Make.
755 * Assembler Options:: Passing options to the assembler.
756 * Link Options:: Specifying libraries and so on.
757 * Directory Options:: Where to find header files and libraries.
758 Where to find the compiler executable files.
759 * Spec Files:: How to pass switches to sub-processes.
760 * Target Options:: Running a cross-compiler, or an old version of GCC.
763 @node Overall Options
764 @section Options Controlling the Kind of Output
766 Compilation can involve up to four stages: preprocessing, compilation
767 proper, assembly and linking, always in that order. GCC is capable of
768 preprocessing and compiling several files either into several
769 assembler input files, or into one assembler input file; then each
770 assembler input file produces an object file, and linking combines all
771 the object files (those newly compiled, and those specified as input)
772 into an executable file.
774 @cindex file name suffix
775 For any given input file, the file name suffix determines what kind of
780 C source code which must be preprocessed.
783 C source code which should not be preprocessed.
786 C++ source code which should not be preprocessed.
789 Objective-C source code. Note that you must link with the @file{libobjc}
790 library to make an Objective-C program work.
793 Objective-C source code which should not be preprocessed.
797 Objective-C++ source code. Note that you must link with the @file{libobjc}
798 library to make an Objective-C++ program work. Note that @samp{.M} refers
799 to a literal capital M@.
802 Objective-C++ source code which should not be preprocessed.
805 C, C++, Objective-C or Objective-C++ header file to be turned into a
810 @itemx @var{file}.cxx
811 @itemx @var{file}.cpp
812 @itemx @var{file}.CPP
813 @itemx @var{file}.c++
815 C++ source code which must be preprocessed. Note that in @samp{.cxx},
816 the last two letters must both be literally @samp{x}. Likewise,
817 @samp{.C} refers to a literal capital C@.
821 C++ header file to be turned into a precompiled header.
824 @itemx @var{file}.for
825 @itemx @var{file}.FOR
826 Fortran source code which should not be preprocessed.
829 @itemx @var{file}.fpp
830 @itemx @var{file}.FPP
831 Fortran source code which must be preprocessed (with the traditional
835 Fortran source code which must be preprocessed with a RATFOR
836 preprocessor (not included with GCC)@.
839 @itemx @var{file}.f95
840 Fortran 90/95 source code which should not be preprocessed.
842 @c FIXME: Descriptions of Java file types.
849 Ada source code file which contains a library unit declaration (a
850 declaration of a package, subprogram, or generic, or a generic
851 instantiation), or a library unit renaming declaration (a package,
852 generic, or subprogram renaming declaration). Such files are also
855 @itemx @var{file}.adb
856 Ada source code file containing a library unit body (a subprogram or
857 package body). Such files are also called @dfn{bodies}.
859 @c GCC also knows about some suffixes for languages not yet included:
868 Assembler code which must be preprocessed.
871 An object file to be fed straight into linking.
872 Any file name with no recognized suffix is treated this way.
876 You can specify the input language explicitly with the @option{-x} option:
879 @item -x @var{language}
880 Specify explicitly the @var{language} for the following input files
881 (rather than letting the compiler choose a default based on the file
882 name suffix). This option applies to all following input files until
883 the next @option{-x} option. Possible values for @var{language} are:
885 c c-header c-cpp-output
886 c++ c++-header c++-cpp-output
887 objective-c objective-c-header objective-c-cpp-output
888 objective-c++ objective-c++-header objective-c++-cpp-output
889 assembler assembler-with-cpp
891 f77 f77-cpp-input ratfor
898 Turn off any specification of a language, so that subsequent files are
899 handled according to their file name suffixes (as they are if @option{-x}
900 has not been used at all).
902 @item -pass-exit-codes
903 @opindex pass-exit-codes
904 Normally the @command{gcc} program will exit with the code of 1 if any
905 phase of the compiler returns a non-success return code. If you specify
906 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
907 numerically highest error produced by any phase that returned an error
911 If you only want some of the stages of compilation, you can use
912 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
913 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
914 @command{gcc} is to stop. Note that some combinations (for example,
915 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
920 Compile or assemble the source files, but do not link. The linking
921 stage simply is not done. The ultimate output is in the form of an
922 object file for each source file.
924 By default, the object file name for a source file is made by replacing
925 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
927 Unrecognized input files, not requiring compilation or assembly, are
932 Stop after the stage of compilation proper; do not assemble. The output
933 is in the form of an assembler code file for each non-assembler input
936 By default, the assembler file name for a source file is made by
937 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
939 Input files that don't require compilation are ignored.
943 Stop after the preprocessing stage; do not run the compiler proper. The
944 output is in the form of preprocessed source code, which is sent to the
947 Input files which don't require preprocessing are ignored.
949 @cindex output file option
952 Place output in file @var{file}. This applies regardless to whatever
953 sort of output is being produced, whether it be an executable file,
954 an object file, an assembler file or preprocessed C code.
956 If @option{-o} is not specified, the default is to put an executable
957 file in @file{a.out}, the object file for
958 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
959 assembler file in @file{@var{source}.s}, a precompiled header file in
960 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
965 Print (on standard error output) the commands executed to run the stages
966 of compilation. Also print the version number of the compiler driver
967 program and of the preprocessor and the compiler proper.
971 Like @option{-v} except the commands are not executed and all command
972 arguments are quoted. This is useful for shell scripts to capture the
973 driver-generated command lines.
977 Use pipes rather than temporary files for communication between the
978 various stages of compilation. This fails to work on some systems where
979 the assembler is unable to read from a pipe; but the GNU assembler has
984 If you are compiling multiple source files, this option tells the driver
985 to pass all the source files to the compiler at once (for those
986 languages for which the compiler can handle this). This will allow
987 intermodule analysis (IMA) to be performed by the compiler. Currently the only
988 language for which this is supported is C@. If you pass source files for
989 multiple languages to the driver, using this option, the driver will invoke
990 the compiler(s) that support IMA once each, passing each compiler all the
991 source files appropriate for it. For those languages that do not support
992 IMA this option will be ignored, and the compiler will be invoked once for
993 each source file in that language. If you use this option in conjunction
994 with @option{-save-temps}, the compiler will generate multiple
996 (one for each source file), but only one (combined) @file{.o} or
1001 Print (on the standard output) a description of the command line options
1002 understood by @command{gcc}. If the @option{-v} option is also specified
1003 then @option{--help} will also be passed on to the various processes
1004 invoked by @command{gcc}, so that they can display the command line options
1005 they accept. If the @option{-Wextra} option is also specified then command
1006 line options which have no documentation associated with them will also
1010 @opindex target-help
1011 Print (on the standard output) a description of target specific command
1012 line options for each tool.
1016 Display the version number and copyrights of the invoked GCC@.
1020 @section Compiling C++ Programs
1022 @cindex suffixes for C++ source
1023 @cindex C++ source file suffixes
1024 C++ source files conventionally use one of the suffixes @samp{.C},
1025 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1026 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1027 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1028 files with these names and compiles them as C++ programs even if you
1029 call the compiler the same way as for compiling C programs (usually
1030 with the name @command{gcc}).
1034 However, C++ programs often require class libraries as well as a
1035 compiler that understands the C++ language---and under some
1036 circumstances, you might want to compile programs or header files from
1037 standard input, or otherwise without a suffix that flags them as C++
1038 programs. You might also like to precompile a C header file with a
1039 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
1040 program that calls GCC with the default language set to C++, and
1041 automatically specifies linking against the C++ library. On many
1042 systems, @command{g++} is also installed with the name @command{c++}.
1044 @cindex invoking @command{g++}
1045 When you compile C++ programs, you may specify many of the same
1046 command-line options that you use for compiling programs in any
1047 language; or command-line options meaningful for C and related
1048 languages; or options that are meaningful only for C++ programs.
1049 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1050 explanations of options for languages related to C@.
1051 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1052 explanations of options that are meaningful only for C++ programs.
1054 @node C Dialect Options
1055 @section Options Controlling C Dialect
1056 @cindex dialect options
1057 @cindex language dialect options
1058 @cindex options, dialect
1060 The following options control the dialect of C (or languages derived
1061 from C, such as C++, Objective-C and Objective-C++) that the compiler
1065 @cindex ANSI support
1069 In C mode, support all ISO C90 programs. In C++ mode,
1070 remove GNU extensions that conflict with ISO C++.
1072 This turns off certain features of GCC that are incompatible with ISO
1073 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1074 such as the @code{asm} and @code{typeof} keywords, and
1075 predefined macros such as @code{unix} and @code{vax} that identify the
1076 type of system you are using. It also enables the undesirable and
1077 rarely used ISO trigraph feature. For the C compiler,
1078 it disables recognition of C++ style @samp{//} comments as well as
1079 the @code{inline} keyword.
1081 The alternate keywords @code{__asm__}, @code{__extension__},
1082 @code{__inline__} and @code{__typeof__} continue to work despite
1083 @option{-ansi}. You would not want to use them in an ISO C program, of
1084 course, but it is useful to put them in header files that might be included
1085 in compilations done with @option{-ansi}. Alternate predefined macros
1086 such as @code{__unix__} and @code{__vax__} are also available, with or
1087 without @option{-ansi}.
1089 The @option{-ansi} option does not cause non-ISO programs to be
1090 rejected gratuitously. For that, @option{-pedantic} is required in
1091 addition to @option{-ansi}. @xref{Warning Options}.
1093 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1094 option is used. Some header files may notice this macro and refrain
1095 from declaring certain functions or defining certain macros that the
1096 ISO standard doesn't call for; this is to avoid interfering with any
1097 programs that might use these names for other things.
1099 Functions which would normally be built in but do not have semantics
1100 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1101 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1102 built-in functions provided by GCC}, for details of the functions
1107 Determine the language standard. This option is currently only
1108 supported when compiling C or C++. A value for this option must be
1109 provided; possible values are
1114 ISO C90 (same as @option{-ansi}).
1116 @item iso9899:199409
1117 ISO C90 as modified in amendment 1.
1123 ISO C99. Note that this standard is not yet fully supported; see
1124 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1125 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1128 Default, ISO C90 plus GNU extensions (including some C99 features).
1132 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1133 this will become the default. The name @samp{gnu9x} is deprecated.
1136 The 1998 ISO C++ standard plus amendments.
1139 The same as @option{-std=c++98} plus GNU extensions. This is the
1140 default for C++ code.
1143 Even when this option is not specified, you can still use some of the
1144 features of newer standards in so far as they do not conflict with
1145 previous C standards. For example, you may use @code{__restrict__} even
1146 when @option{-std=c99} is not specified.
1148 The @option{-std} options specifying some version of ISO C have the same
1149 effects as @option{-ansi}, except that features that were not in ISO C90
1150 but are in the specified version (for example, @samp{//} comments and
1151 the @code{inline} keyword in ISO C99) are not disabled.
1153 @xref{Standards,,Language Standards Supported by GCC}, for details of
1154 these standard versions.
1156 @item -aux-info @var{filename}
1158 Output to the given filename prototyped declarations for all functions
1159 declared and/or defined in a translation unit, including those in header
1160 files. This option is silently ignored in any language other than C@.
1162 Besides declarations, the file indicates, in comments, the origin of
1163 each declaration (source file and line), whether the declaration was
1164 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1165 @samp{O} for old, respectively, in the first character after the line
1166 number and the colon), and whether it came from a declaration or a
1167 definition (@samp{C} or @samp{F}, respectively, in the following
1168 character). In the case of function definitions, a K&R-style list of
1169 arguments followed by their declarations is also provided, inside
1170 comments, after the declaration.
1174 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1175 keyword, so that code can use these words as identifiers. You can use
1176 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1177 instead. @option{-ansi} implies @option{-fno-asm}.
1179 In C++, this switch only affects the @code{typeof} keyword, since
1180 @code{asm} and @code{inline} are standard keywords. You may want to
1181 use the @option{-fno-gnu-keywords} flag instead, which has the same
1182 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1183 switch only affects the @code{asm} and @code{typeof} keywords, since
1184 @code{inline} is a standard keyword in ISO C99.
1187 @itemx -fno-builtin-@var{function}
1188 @opindex fno-builtin
1189 @cindex built-in functions
1190 Don't recognize built-in functions that do not begin with
1191 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1192 functions provided by GCC}, for details of the functions affected,
1193 including those which are not built-in functions when @option{-ansi} or
1194 @option{-std} options for strict ISO C conformance are used because they
1195 do not have an ISO standard meaning.
1197 GCC normally generates special code to handle certain built-in functions
1198 more efficiently; for instance, calls to @code{alloca} may become single
1199 instructions that adjust the stack directly, and calls to @code{memcpy}
1200 may become inline copy loops. The resulting code is often both smaller
1201 and faster, but since the function calls no longer appear as such, you
1202 cannot set a breakpoint on those calls, nor can you change the behavior
1203 of the functions by linking with a different library. In addition,
1204 when a function is recognized as a built-in function, GCC may use
1205 information about that function to warn about problems with calls to
1206 that function, or to generate more efficient code, even if the
1207 resulting code still contains calls to that function. For example,
1208 warnings are given with @option{-Wformat} for bad calls to
1209 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1210 known not to modify global memory.
1212 With the @option{-fno-builtin-@var{function}} option
1213 only the built-in function @var{function} is
1214 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1215 function is named this is not built-in in this version of GCC, this
1216 option is ignored. There is no corresponding
1217 @option{-fbuiltin-@var{function}} option; if you wish to enable
1218 built-in functions selectively when using @option{-fno-builtin} or
1219 @option{-ffreestanding}, you may define macros such as:
1222 #define abs(n) __builtin_abs ((n))
1223 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1228 @cindex hosted environment
1230 Assert that compilation takes place in a hosted environment. This implies
1231 @option{-fbuiltin}. A hosted environment is one in which the
1232 entire standard library is available, and in which @code{main} has a return
1233 type of @code{int}. Examples are nearly everything except a kernel.
1234 This is equivalent to @option{-fno-freestanding}.
1236 @item -ffreestanding
1237 @opindex ffreestanding
1238 @cindex hosted environment
1240 Assert that compilation takes place in a freestanding environment. This
1241 implies @option{-fno-builtin}. A freestanding environment
1242 is one in which the standard library may not exist, and program startup may
1243 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1244 This is equivalent to @option{-fno-hosted}.
1246 @xref{Standards,,Language Standards Supported by GCC}, for details of
1247 freestanding and hosted environments.
1249 @item -fms-extensions
1250 @opindex fms-extensions
1251 Accept some non-standard constructs used in Microsoft header files.
1253 Some cases of unnamed fields in structures and unions are only
1254 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1255 fields within structs/unions}, for details.
1259 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1260 options for strict ISO C conformance) implies @option{-trigraphs}.
1262 @item -no-integrated-cpp
1263 @opindex no-integrated-cpp
1264 Performs a compilation in two passes: preprocessing and compiling. This
1265 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1266 @option{-B} option. The user supplied compilation step can then add in
1267 an additional preprocessing step after normal preprocessing but before
1268 compiling. The default is to use the integrated cpp (internal cpp)
1270 The semantics of this option will change if "cc1", "cc1plus", and
1271 "cc1obj" are merged.
1273 @cindex traditional C language
1274 @cindex C language, traditional
1276 @itemx -traditional-cpp
1277 @opindex traditional-cpp
1278 @opindex traditional
1279 Formerly, these options caused GCC to attempt to emulate a pre-standard
1280 C compiler. They are now only supported with the @option{-E} switch.
1281 The preprocessor continues to support a pre-standard mode. See the GNU
1282 CPP manual for details.
1284 @item -fcond-mismatch
1285 @opindex fcond-mismatch
1286 Allow conditional expressions with mismatched types in the second and
1287 third arguments. The value of such an expression is void. This option
1288 is not supported for C++.
1290 @item -funsigned-char
1291 @opindex funsigned-char
1292 Let the type @code{char} be unsigned, like @code{unsigned char}.
1294 Each kind of machine has a default for what @code{char} should
1295 be. It is either like @code{unsigned char} by default or like
1296 @code{signed char} by default.
1298 Ideally, a portable program should always use @code{signed char} or
1299 @code{unsigned char} when it depends on the signedness of an object.
1300 But many programs have been written to use plain @code{char} and
1301 expect it to be signed, or expect it to be unsigned, depending on the
1302 machines they were written for. This option, and its inverse, let you
1303 make such a program work with the opposite default.
1305 The type @code{char} is always a distinct type from each of
1306 @code{signed char} or @code{unsigned char}, even though its behavior
1307 is always just like one of those two.
1310 @opindex fsigned-char
1311 Let the type @code{char} be signed, like @code{signed char}.
1313 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1314 the negative form of @option{-funsigned-char}. Likewise, the option
1315 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1317 @item -fsigned-bitfields
1318 @itemx -funsigned-bitfields
1319 @itemx -fno-signed-bitfields
1320 @itemx -fno-unsigned-bitfields
1321 @opindex fsigned-bitfields
1322 @opindex funsigned-bitfields
1323 @opindex fno-signed-bitfields
1324 @opindex fno-unsigned-bitfields
1325 These options control whether a bit-field is signed or unsigned, when the
1326 declaration does not use either @code{signed} or @code{unsigned}. By
1327 default, such a bit-field is signed, because this is consistent: the
1328 basic integer types such as @code{int} are signed types.
1331 @node C++ Dialect Options
1332 @section Options Controlling C++ Dialect
1334 @cindex compiler options, C++
1335 @cindex C++ options, command line
1336 @cindex options, C++
1337 This section describes the command-line options that are only meaningful
1338 for C++ programs; but you can also use most of the GNU compiler options
1339 regardless of what language your program is in. For example, you
1340 might compile a file @code{firstClass.C} like this:
1343 g++ -g -frepo -O -c firstClass.C
1347 In this example, only @option{-frepo} is an option meant
1348 only for C++ programs; you can use the other options with any
1349 language supported by GCC@.
1351 Here is a list of options that are @emph{only} for compiling C++ programs:
1355 @item -fabi-version=@var{n}
1356 @opindex fabi-version
1357 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1358 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1359 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1360 the version that conforms most closely to the C++ ABI specification.
1361 Therefore, the ABI obtained using version 0 will change as ABI bugs
1364 The default is version 2.
1366 @item -fno-access-control
1367 @opindex fno-access-control
1368 Turn off all access checking. This switch is mainly useful for working
1369 around bugs in the access control code.
1373 Check that the pointer returned by @code{operator new} is non-null
1374 before attempting to modify the storage allocated. This check is
1375 normally unnecessary because the C++ standard specifies that
1376 @code{operator new} will only return @code{0} if it is declared
1377 @samp{throw()}, in which case the compiler will always check the
1378 return value even without this option. In all other cases, when
1379 @code{operator new} has a non-empty exception specification, memory
1380 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1381 @samp{new (nothrow)}.
1383 @item -fconserve-space
1384 @opindex fconserve-space
1385 Put uninitialized or runtime-initialized global variables into the
1386 common segment, as C does. This saves space in the executable at the
1387 cost of not diagnosing duplicate definitions. If you compile with this
1388 flag and your program mysteriously crashes after @code{main()} has
1389 completed, you may have an object that is being destroyed twice because
1390 two definitions were merged.
1392 This option is no longer useful on most targets, now that support has
1393 been added for putting variables into BSS without making them common.
1395 @item -fno-const-strings
1396 @opindex fno-const-strings
1397 Give string constants type @code{char *} instead of type @code{const
1398 char *}. By default, G++ uses type @code{const char *} as required by
1399 the standard. Even if you use @option{-fno-const-strings}, you cannot
1400 actually modify the value of a string constant.
1402 This option might be removed in a future release of G++. For maximum
1403 portability, you should structure your code so that it works with
1404 string constants that have type @code{const char *}.
1406 @item -fno-elide-constructors
1407 @opindex fno-elide-constructors
1408 The C++ standard allows an implementation to omit creating a temporary
1409 which is only used to initialize another object of the same type.
1410 Specifying this option disables that optimization, and forces G++ to
1411 call the copy constructor in all cases.
1413 @item -fno-enforce-eh-specs
1414 @opindex fno-enforce-eh-specs
1415 Don't check for violation of exception specifications at runtime. This
1416 option violates the C++ standard, but may be useful for reducing code
1417 size in production builds, much like defining @samp{NDEBUG}. The compiler
1418 will still optimize based on the exception specifications.
1421 @itemx -fno-for-scope
1423 @opindex fno-for-scope
1424 If @option{-ffor-scope} is specified, the scope of variables declared in
1425 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1426 as specified by the C++ standard.
1427 If @option{-fno-for-scope} is specified, the scope of variables declared in
1428 a @i{for-init-statement} extends to the end of the enclosing scope,
1429 as was the case in old versions of G++, and other (traditional)
1430 implementations of C++.
1432 The default if neither flag is given to follow the standard,
1433 but to allow and give a warning for old-style code that would
1434 otherwise be invalid, or have different behavior.
1436 @item -fno-gnu-keywords
1437 @opindex fno-gnu-keywords
1438 Do not recognize @code{typeof} as a keyword, so that code can use this
1439 word as an identifier. You can use the keyword @code{__typeof__} instead.
1440 @option{-ansi} implies @option{-fno-gnu-keywords}.
1442 @item -fno-implicit-templates
1443 @opindex fno-implicit-templates
1444 Never emit code for non-inline templates which are instantiated
1445 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1446 @xref{Template Instantiation}, for more information.
1448 @item -fno-implicit-inline-templates
1449 @opindex fno-implicit-inline-templates
1450 Don't emit code for implicit instantiations of inline templates, either.
1451 The default is to handle inlines differently so that compiles with and
1452 without optimization will need the same set of explicit instantiations.
1454 @item -fno-implement-inlines
1455 @opindex fno-implement-inlines
1456 To save space, do not emit out-of-line copies of inline functions
1457 controlled by @samp{#pragma implementation}. This will cause linker
1458 errors if these functions are not inlined everywhere they are called.
1460 @item -fms-extensions
1461 @opindex fms-extensions
1462 Disable pedantic warnings about constructs used in MFC, such as implicit
1463 int and getting a pointer to member function via non-standard syntax.
1465 @item -fno-nonansi-builtins
1466 @opindex fno-nonansi-builtins
1467 Disable built-in declarations of functions that are not mandated by
1468 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1469 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1471 @item -fno-operator-names
1472 @opindex fno-operator-names
1473 Do not treat the operator name keywords @code{and}, @code{bitand},
1474 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1475 synonyms as keywords.
1477 @item -fno-optional-diags
1478 @opindex fno-optional-diags
1479 Disable diagnostics that the standard says a compiler does not need to
1480 issue. Currently, the only such diagnostic issued by G++ is the one for
1481 a name having multiple meanings within a class.
1484 @opindex fpermissive
1485 Downgrade some diagnostics about nonconformant code from errors to
1486 warnings. Thus, using @option{-fpermissive} will allow some
1487 nonconforming code to compile.
1491 Enable automatic template instantiation at link time. This option also
1492 implies @option{-fno-implicit-templates}. @xref{Template
1493 Instantiation}, for more information.
1497 Disable generation of information about every class with virtual
1498 functions for use by the C++ runtime type identification features
1499 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1500 of the language, you can save some space by using this flag. Note that
1501 exception handling uses the same information, but it will generate it as
1506 Emit statistics about front-end processing at the end of the compilation.
1507 This information is generally only useful to the G++ development team.
1509 @item -ftemplate-depth-@var{n}
1510 @opindex ftemplate-depth
1511 Set the maximum instantiation depth for template classes to @var{n}.
1512 A limit on the template instantiation depth is needed to detect
1513 endless recursions during template class instantiation. ANSI/ISO C++
1514 conforming programs must not rely on a maximum depth greater than 17.
1516 @item -fno-threadsafe-statics
1517 @opindex fno-threadsafe-statics
1518 Do not emit the extra code to use the routines specified in the C++
1519 ABI for thread-safe initialization of local statics. You can use this
1520 option to reduce code size slightly in code that doesn't need to be
1523 @item -fuse-cxa-atexit
1524 @opindex fuse-cxa-atexit
1525 Register destructors for objects with static storage duration with the
1526 @code{__cxa_atexit} function rather than the @code{atexit} function.
1527 This option is required for fully standards-compliant handling of static
1528 destructors, but will only work if your C library supports
1529 @code{__cxa_atexit}.
1531 @item -fvisibility-inlines-hidden
1532 @opindex fvisibility-inlines-hidden
1533 Causes all inlined methods to be marked with
1534 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1535 appear in the export table of a DSO and do not require a PLT indirection
1536 when used within the DSO@. Enabling this option can have a dramatic effect
1537 on load and link times of a DSO as it massively reduces the size of the
1538 dynamic export table when the library makes heavy use of templates. While
1539 it can cause bloating through duplication of code within each DSO where
1540 it is used, often the wastage is less than the considerable space occupied
1541 by a long symbol name in the export table which is typical when using
1542 templates and namespaces. For even more savings, combine with the
1543 @option{-fvisibility=hidden} switch.
1547 Do not use weak symbol support, even if it is provided by the linker.
1548 By default, G++ will use weak symbols if they are available. This
1549 option exists only for testing, and should not be used by end-users;
1550 it will result in inferior code and has no benefits. This option may
1551 be removed in a future release of G++.
1555 Do not search for header files in the standard directories specific to
1556 C++, but do still search the other standard directories. (This option
1557 is used when building the C++ library.)
1560 In addition, these optimization, warning, and code generation options
1561 have meanings only for C++ programs:
1564 @item -fno-default-inline
1565 @opindex fno-default-inline
1566 Do not assume @samp{inline} for functions defined inside a class scope.
1567 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1568 functions will have linkage like inline functions; they just won't be
1571 @item -Wabi @r{(C++ only)}
1573 Warn when G++ generates code that is probably not compatible with the
1574 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1575 all such cases, there are probably some cases that are not warned about,
1576 even though G++ is generating incompatible code. There may also be
1577 cases where warnings are emitted even though the code that is generated
1580 You should rewrite your code to avoid these warnings if you are
1581 concerned about the fact that code generated by G++ may not be binary
1582 compatible with code generated by other compilers.
1584 The known incompatibilities at this point include:
1589 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1590 pack data into the same byte as a base class. For example:
1593 struct A @{ virtual void f(); int f1 : 1; @};
1594 struct B : public A @{ int f2 : 1; @};
1598 In this case, G++ will place @code{B::f2} into the same byte
1599 as@code{A::f1}; other compilers will not. You can avoid this problem
1600 by explicitly padding @code{A} so that its size is a multiple of the
1601 byte size on your platform; that will cause G++ and other compilers to
1602 layout @code{B} identically.
1605 Incorrect handling of tail-padding for virtual bases. G++ does not use
1606 tail padding when laying out virtual bases. For example:
1609 struct A @{ virtual void f(); char c1; @};
1610 struct B @{ B(); char c2; @};
1611 struct C : public A, public virtual B @{@};
1615 In this case, G++ will not place @code{B} into the tail-padding for
1616 @code{A}; other compilers will. You can avoid this problem by
1617 explicitly padding @code{A} so that its size is a multiple of its
1618 alignment (ignoring virtual base classes); that will cause G++ and other
1619 compilers to layout @code{C} identically.
1622 Incorrect handling of bit-fields with declared widths greater than that
1623 of their underlying types, when the bit-fields appear in a union. For
1627 union U @{ int i : 4096; @};
1631 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1632 union too small by the number of bits in an @code{int}.
1635 Empty classes can be placed at incorrect offsets. For example:
1645 struct C : public B, public A @{@};
1649 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1650 it should be placed at offset zero. G++ mistakenly believes that the
1651 @code{A} data member of @code{B} is already at offset zero.
1654 Names of template functions whose types involve @code{typename} or
1655 template template parameters can be mangled incorrectly.
1658 template <typename Q>
1659 void f(typename Q::X) @{@}
1661 template <template <typename> class Q>
1662 void f(typename Q<int>::X) @{@}
1666 Instantiations of these templates may be mangled incorrectly.
1670 @item -Wctor-dtor-privacy @r{(C++ only)}
1671 @opindex Wctor-dtor-privacy
1672 Warn when a class seems unusable because all the constructors or
1673 destructors in that class are private, and it has neither friends nor
1674 public static member functions.
1676 @item -Wnon-virtual-dtor @r{(C++ only)}
1677 @opindex Wnon-virtual-dtor
1678 Warn when a class appears to be polymorphic, thereby requiring a virtual
1679 destructor, yet it declares a non-virtual one.
1680 This warning is enabled by @option{-Wall}.
1682 @item -Wreorder @r{(C++ only)}
1684 @cindex reordering, warning
1685 @cindex warning for reordering of member initializers
1686 Warn when the order of member initializers given in the code does not
1687 match the order in which they must be executed. For instance:
1693 A(): j (0), i (1) @{ @}
1697 The compiler will rearrange the member initializers for @samp{i}
1698 and @samp{j} to match the declaration order of the members, emitting
1699 a warning to that effect. This warning is enabled by @option{-Wall}.
1702 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1705 @item -Weffc++ @r{(C++ only)}
1707 Warn about violations of the following style guidelines from Scott Meyers'
1708 @cite{Effective C++} book:
1712 Item 11: Define a copy constructor and an assignment operator for classes
1713 with dynamically allocated memory.
1716 Item 12: Prefer initialization to assignment in constructors.
1719 Item 14: Make destructors virtual in base classes.
1722 Item 15: Have @code{operator=} return a reference to @code{*this}.
1725 Item 23: Don't try to return a reference when you must return an object.
1729 Also warn about violations of the following style guidelines from
1730 Scott Meyers' @cite{More Effective C++} book:
1734 Item 6: Distinguish between prefix and postfix forms of increment and
1735 decrement operators.
1738 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1742 When selecting this option, be aware that the standard library
1743 headers do not obey all of these guidelines; use @samp{grep -v}
1744 to filter out those warnings.
1746 @item -Wno-deprecated @r{(C++ only)}
1747 @opindex Wno-deprecated
1748 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1750 @item -Wstrict-null-sentinel @r{(C++ only)}
1751 @opindex Wstrict-null-sentinel
1752 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1753 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1754 to @code{__null}. Although it is a null pointer constant not a null pointer,
1755 it is guaranteed to of the same size as a pointer. But this use is
1756 not portable across different compilers.
1758 @item -Wno-non-template-friend @r{(C++ only)}
1759 @opindex Wno-non-template-friend
1760 Disable warnings when non-templatized friend functions are declared
1761 within a template. Since the advent of explicit template specification
1762 support in G++, if the name of the friend is an unqualified-id (i.e.,
1763 @samp{friend foo(int)}), the C++ language specification demands that the
1764 friend declare or define an ordinary, nontemplate function. (Section
1765 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1766 could be interpreted as a particular specialization of a templatized
1767 function. Because this non-conforming behavior is no longer the default
1768 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1769 check existing code for potential trouble spots and is on by default.
1770 This new compiler behavior can be turned off with
1771 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1772 but disables the helpful warning.
1774 @item -Wold-style-cast @r{(C++ only)}
1775 @opindex Wold-style-cast
1776 Warn if an old-style (C-style) cast to a non-void type is used within
1777 a C++ program. The new-style casts (@samp{static_cast},
1778 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1779 unintended effects and much easier to search for.
1781 @item -Woverloaded-virtual @r{(C++ only)}
1782 @opindex Woverloaded-virtual
1783 @cindex overloaded virtual fn, warning
1784 @cindex warning for overloaded virtual fn
1785 Warn when a function declaration hides virtual functions from a
1786 base class. For example, in:
1793 struct B: public A @{
1798 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1806 will fail to compile.
1808 @item -Wno-pmf-conversions @r{(C++ only)}
1809 @opindex Wno-pmf-conversions
1810 Disable the diagnostic for converting a bound pointer to member function
1813 @item -Wsign-promo @r{(C++ only)}
1814 @opindex Wsign-promo
1815 Warn when overload resolution chooses a promotion from unsigned or
1816 enumerated type to a signed type, over a conversion to an unsigned type of
1817 the same size. Previous versions of G++ would try to preserve
1818 unsignedness, but the standard mandates the current behavior.
1823 A& operator = (int);
1833 In this example, G++ will synthesize a default @samp{A& operator =
1834 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1837 @node Objective-C and Objective-C++ Dialect Options
1838 @section Options Controlling Objective-C and Objective-C++ Dialects
1840 @cindex compiler options, Objective-C and Objective-C++
1841 @cindex Objective-C and Objective-C++ options, command line
1842 @cindex options, Objective-C and Objective-C++
1843 (NOTE: This manual does not describe the Objective-C and Objective-C++
1844 languages themselves. See @xref{Standards,,Language Standards
1845 Supported by GCC}, for references.)
1847 This section describes the command-line options that are only meaningful
1848 for Objective-C and Objective-C++ programs, but you can also use most of
1849 the language-independent GNU compiler options.
1850 For example, you might compile a file @code{some_class.m} like this:
1853 gcc -g -fgnu-runtime -O -c some_class.m
1857 In this example, @option{-fgnu-runtime} is an option meant only for
1858 Objective-C and Objective-C++ programs; you can use the other options with
1859 any language supported by GCC@.
1861 Note that since Objective-C is an extension of the C language, Objective-C
1862 compilations may also use options specific to the C front-end (e.g.,
1863 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1864 C++-specific options (e.g., @option{-Wabi}).
1866 Here is a list of options that are @emph{only} for compiling Objective-C
1867 and Objective-C++ programs:
1870 @item -fconstant-string-class=@var{class-name}
1871 @opindex fconstant-string-class
1872 Use @var{class-name} as the name of the class to instantiate for each
1873 literal string specified with the syntax @code{@@"@dots{}"}. The default
1874 class name is @code{NXConstantString} if the GNU runtime is being used, and
1875 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1876 @option{-fconstant-cfstrings} option, if also present, will override the
1877 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1878 to be laid out as constant CoreFoundation strings.
1881 @opindex fgnu-runtime
1882 Generate object code compatible with the standard GNU Objective-C
1883 runtime. This is the default for most types of systems.
1885 @item -fnext-runtime
1886 @opindex fnext-runtime
1887 Generate output compatible with the NeXT runtime. This is the default
1888 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1889 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1892 @item -fno-nil-receivers
1893 @opindex fno-nil-receivers
1894 Assume that all Objective-C message dispatches (e.g.,
1895 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1896 is not @code{nil}. This allows for more efficient entry points in the runtime
1897 to be used. Currently, this option is only available in conjunction with
1898 the NeXT runtime on Mac OS X 10.3 and later.
1900 @item -fobjc-exceptions
1901 @opindex fobjc-exceptions
1902 Enable syntactic support for structured exception handling in Objective-C,
1903 similar to what is offered by C++ and Java. Currently, this option is only
1904 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1912 @@catch (AnObjCClass *exc) @{
1919 @@catch (AnotherClass *exc) @{
1922 @@catch (id allOthers) @{
1932 The @code{@@throw} statement may appear anywhere in an Objective-C or
1933 Objective-C++ program; when used inside of a @code{@@catch} block, the
1934 @code{@@throw} may appear without an argument (as shown above), in which case
1935 the object caught by the @code{@@catch} will be rethrown.
1937 Note that only (pointers to) Objective-C objects may be thrown and
1938 caught using this scheme. When an object is thrown, it will be caught
1939 by the nearest @code{@@catch} clause capable of handling objects of that type,
1940 analogously to how @code{catch} blocks work in C++ and Java. A
1941 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1942 any and all Objective-C exceptions not caught by previous @code{@@catch}
1945 The @code{@@finally} clause, if present, will be executed upon exit from the
1946 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1947 regardless of whether any exceptions are thrown, caught or rethrown
1948 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1949 of the @code{finally} clause in Java.
1951 There are several caveats to using the new exception mechanism:
1955 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
1956 idioms provided by the @code{NSException} class, the new
1957 exceptions can only be used on Mac OS X 10.3 (Panther) and later
1958 systems, due to additional functionality needed in the (NeXT) Objective-C
1962 As mentioned above, the new exceptions do not support handling
1963 types other than Objective-C objects. Furthermore, when used from
1964 Objective-C++, the Objective-C exception model does not interoperate with C++
1965 exceptions at this time. This means you cannot @code{@@throw} an exception
1966 from Objective-C and @code{catch} it in C++, or vice versa
1967 (i.e., @code{throw @dots{} @@catch}).
1970 The @option{-fobjc-exceptions} switch also enables the use of synchronization
1971 blocks for thread-safe execution:
1974 @@synchronized (ObjCClass *guard) @{
1979 Upon entering the @code{@@synchronized} block, a thread of execution shall
1980 first check whether a lock has been placed on the corresponding @code{guard}
1981 object by another thread. If it has, the current thread shall wait until
1982 the other thread relinquishes its lock. Once @code{guard} becomes available,
1983 the current thread will place its own lock on it, execute the code contained in
1984 the @code{@@synchronized} block, and finally relinquish the lock (thereby
1985 making @code{guard} available to other threads).
1987 Unlike Java, Objective-C does not allow for entire methods to be marked
1988 @code{@@synchronized}. Note that throwing exceptions out of
1989 @code{@@synchronized} blocks is allowed, and will cause the guarding object
1990 to be unlocked properly.
1992 @item -freplace-objc-classes
1993 @opindex freplace-objc-classes
1994 Emit a special marker instructing @command{ld(1)} not to statically link in
1995 the resulting object file, and allow @command{dyld(1)} to load it in at
1996 run time instead. This is used in conjunction with the Fix-and-Continue
1997 debugging mode, where the object file in question may be recompiled and
1998 dynamically reloaded in the course of program execution, without the need
1999 to restart the program itself. Currently, Fix-and-Continue functionality
2000 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2005 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2006 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2007 compile time) with static class references that get initialized at load time,
2008 which improves run-time performance. Specifying the @option{-fzero-link} flag
2009 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2010 to be retained. This is useful in Zero-Link debugging mode, since it allows
2011 for individual class implementations to be modified during program execution.
2015 Dump interface declarations for all classes seen in the source file to a
2016 file named @file{@var{sourcename}.decl}.
2019 @opindex Wno-protocol
2020 If a class is declared to implement a protocol, a warning is issued for
2021 every method in the protocol that is not implemented by the class. The
2022 default behavior is to issue a warning for every method not explicitly
2023 implemented in the class, even if a method implementation is inherited
2024 from the superclass. If you use the @option{-Wno-protocol} option, then
2025 methods inherited from the superclass are considered to be implemented,
2026 and no warning is issued for them.
2030 Warn if multiple methods of different types for the same selector are
2031 found during compilation. The check is performed on the list of methods
2032 in the final stage of compilation. Additionally, a check is performed
2033 for each selector appearing in a @code{@@selector(@dots{})}
2034 expression, and a corresponding method for that selector has been found
2035 during compilation. Because these checks scan the method table only at
2036 the end of compilation, these warnings are not produced if the final
2037 stage of compilation is not reached, for example because an error is
2038 found during compilation, or because the @option{-fsyntax-only} option is
2041 @item -Wundeclared-selector
2042 @opindex Wundeclared-selector
2043 Warn if a @code{@@selector(@dots{})} expression referring to an
2044 undeclared selector is found. A selector is considered undeclared if no
2045 method with that name has been declared before the
2046 @code{@@selector(@dots{})} expression, either explicitly in an
2047 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2048 an @code{@@implementation} section. This option always performs its
2049 checks as soon as a @code{@@selector(@dots{})} expression is found,
2050 while @option{-Wselector} only performs its checks in the final stage of
2051 compilation. This also enforces the coding style convention
2052 that methods and selectors must be declared before being used.
2054 @item -print-objc-runtime-info
2055 @opindex print-objc-runtime-info
2056 Generate C header describing the largest structure that is passed by
2061 @node Language Independent Options
2062 @section Options to Control Diagnostic Messages Formatting
2063 @cindex options to control diagnostics formatting
2064 @cindex diagnostic messages
2065 @cindex message formatting
2067 Traditionally, diagnostic messages have been formatted irrespective of
2068 the output device's aspect (e.g.@: its width, @dots{}). The options described
2069 below can be used to control the diagnostic messages formatting
2070 algorithm, e.g.@: how many characters per line, how often source location
2071 information should be reported. Right now, only the C++ front end can
2072 honor these options. However it is expected, in the near future, that
2073 the remaining front ends would be able to digest them correctly.
2076 @item -fmessage-length=@var{n}
2077 @opindex fmessage-length
2078 Try to format error messages so that they fit on lines of about @var{n}
2079 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2080 the front ends supported by GCC@. If @var{n} is zero, then no
2081 line-wrapping will be done; each error message will appear on a single
2084 @opindex fdiagnostics-show-location
2085 @item -fdiagnostics-show-location=once
2086 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2087 reporter to emit @emph{once} source location information; that is, in
2088 case the message is too long to fit on a single physical line and has to
2089 be wrapped, the source location won't be emitted (as prefix) again,
2090 over and over, in subsequent continuation lines. This is the default
2093 @item -fdiagnostics-show-location=every-line
2094 Only meaningful in line-wrapping mode. Instructs the diagnostic
2095 messages reporter to emit the same source location information (as
2096 prefix) for physical lines that result from the process of breaking
2097 a message which is too long to fit on a single line.
2099 @item -fdiagnostics-show-options
2100 @opindex fdiagnostics-show-options
2101 This option instructs the diagnostic machinery to add text to each
2102 diagnostic emitted, which indicates which command line option directly
2103 controls that diagnostic, when such an option is known to the
2104 diagnostic machinery.
2108 @node Warning Options
2109 @section Options to Request or Suppress Warnings
2110 @cindex options to control warnings
2111 @cindex warning messages
2112 @cindex messages, warning
2113 @cindex suppressing warnings
2115 Warnings are diagnostic messages that report constructions which
2116 are not inherently erroneous but which are risky or suggest there
2117 may have been an error.
2119 You can request many specific warnings with options beginning @samp{-W},
2120 for example @option{-Wimplicit} to request warnings on implicit
2121 declarations. Each of these specific warning options also has a
2122 negative form beginning @samp{-Wno-} to turn off warnings;
2123 for example, @option{-Wno-implicit}. This manual lists only one of the
2124 two forms, whichever is not the default.
2126 The following options control the amount and kinds of warnings produced
2127 by GCC; for further, language-specific options also refer to
2128 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2132 @cindex syntax checking
2134 @opindex fsyntax-only
2135 Check the code for syntax errors, but don't do anything beyond that.
2139 Issue all the warnings demanded by strict ISO C and ISO C++;
2140 reject all programs that use forbidden extensions, and some other
2141 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2142 version of the ISO C standard specified by any @option{-std} option used.
2144 Valid ISO C and ISO C++ programs should compile properly with or without
2145 this option (though a rare few will require @option{-ansi} or a
2146 @option{-std} option specifying the required version of ISO C)@. However,
2147 without this option, certain GNU extensions and traditional C and C++
2148 features are supported as well. With this option, they are rejected.
2150 @option{-pedantic} does not cause warning messages for use of the
2151 alternate keywords whose names begin and end with @samp{__}. Pedantic
2152 warnings are also disabled in the expression that follows
2153 @code{__extension__}. However, only system header files should use
2154 these escape routes; application programs should avoid them.
2155 @xref{Alternate Keywords}.
2157 Some users try to use @option{-pedantic} to check programs for strict ISO
2158 C conformance. They soon find that it does not do quite what they want:
2159 it finds some non-ISO practices, but not all---only those for which
2160 ISO C @emph{requires} a diagnostic, and some others for which
2161 diagnostics have been added.
2163 A feature to report any failure to conform to ISO C might be useful in
2164 some instances, but would require considerable additional work and would
2165 be quite different from @option{-pedantic}. We don't have plans to
2166 support such a feature in the near future.
2168 Where the standard specified with @option{-std} represents a GNU
2169 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2170 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2171 extended dialect is based. Warnings from @option{-pedantic} are given
2172 where they are required by the base standard. (It would not make sense
2173 for such warnings to be given only for features not in the specified GNU
2174 C dialect, since by definition the GNU dialects of C include all
2175 features the compiler supports with the given option, and there would be
2176 nothing to warn about.)
2178 @item -pedantic-errors
2179 @opindex pedantic-errors
2180 Like @option{-pedantic}, except that errors are produced rather than
2185 Inhibit all warning messages.
2189 Inhibit warning messages about the use of @samp{#import}.
2191 @item -Wchar-subscripts
2192 @opindex Wchar-subscripts
2193 Warn if an array subscript has type @code{char}. This is a common cause
2194 of error, as programmers often forget that this type is signed on some
2196 This warning is enabled by @option{-Wall}.
2200 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2201 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2202 This warning is enabled by @option{-Wall}.
2204 @item -Wfatal-errors
2205 @opindex Wfatal-errors
2206 This option causes the compiler to abort compilation on the first error
2207 occurred rather than trying to keep going and printing further error
2212 @opindex ffreestanding
2213 @opindex fno-builtin
2214 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2215 the arguments supplied have types appropriate to the format string
2216 specified, and that the conversions specified in the format string make
2217 sense. This includes standard functions, and others specified by format
2218 attributes (@pxref{Function Attributes}), in the @code{printf},
2219 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2220 not in the C standard) families (or other target-specific families).
2221 Which functions are checked without format attributes having been
2222 specified depends on the standard version selected, and such checks of
2223 functions without the attribute specified are disabled by
2224 @option{-ffreestanding} or @option{-fno-builtin}.
2226 The formats are checked against the format features supported by GNU
2227 libc version 2.2. These include all ISO C90 and C99 features, as well
2228 as features from the Single Unix Specification and some BSD and GNU
2229 extensions. Other library implementations may not support all these
2230 features; GCC does not support warning about features that go beyond a
2231 particular library's limitations. However, if @option{-pedantic} is used
2232 with @option{-Wformat}, warnings will be given about format features not
2233 in the selected standard version (but not for @code{strfmon} formats,
2234 since those are not in any version of the C standard). @xref{C Dialect
2235 Options,,Options Controlling C Dialect}.
2237 Since @option{-Wformat} also checks for null format arguments for
2238 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2240 @option{-Wformat} is included in @option{-Wall}. For more control over some
2241 aspects of format checking, the options @option{-Wformat-y2k},
2242 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2243 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2244 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2247 @opindex Wformat-y2k
2248 If @option{-Wformat} is specified, also warn about @code{strftime}
2249 formats which may yield only a two-digit year.
2251 @item -Wno-format-extra-args
2252 @opindex Wno-format-extra-args
2253 If @option{-Wformat} is specified, do not warn about excess arguments to a
2254 @code{printf} or @code{scanf} format function. The C standard specifies
2255 that such arguments are ignored.
2257 Where the unused arguments lie between used arguments that are
2258 specified with @samp{$} operand number specifications, normally
2259 warnings are still given, since the implementation could not know what
2260 type to pass to @code{va_arg} to skip the unused arguments. However,
2261 in the case of @code{scanf} formats, this option will suppress the
2262 warning if the unused arguments are all pointers, since the Single
2263 Unix Specification says that such unused arguments are allowed.
2265 @item -Wno-format-zero-length
2266 @opindex Wno-format-zero-length
2267 If @option{-Wformat} is specified, do not warn about zero-length formats.
2268 The C standard specifies that zero-length formats are allowed.
2270 @item -Wformat-nonliteral
2271 @opindex Wformat-nonliteral
2272 If @option{-Wformat} is specified, also warn if the format string is not a
2273 string literal and so cannot be checked, unless the format function
2274 takes its format arguments as a @code{va_list}.
2276 @item -Wformat-security
2277 @opindex Wformat-security
2278 If @option{-Wformat} is specified, also warn about uses of format
2279 functions that represent possible security problems. At present, this
2280 warns about calls to @code{printf} and @code{scanf} functions where the
2281 format string is not a string literal and there are no format arguments,
2282 as in @code{printf (foo);}. This may be a security hole if the format
2283 string came from untrusted input and contains @samp{%n}. (This is
2284 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2285 in future warnings may be added to @option{-Wformat-security} that are not
2286 included in @option{-Wformat-nonliteral}.)
2290 Enable @option{-Wformat} plus format checks not included in
2291 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2292 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2296 Warn about passing a null pointer for arguments marked as
2297 requiring a non-null value by the @code{nonnull} function attribute.
2299 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2300 can be disabled with the @option{-Wno-nonnull} option.
2302 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2304 Warn about uninitialized variables which are initialized with themselves.
2305 Note this option can only be used with the @option{-Wuninitialized} option,
2306 which in turn only works with @option{-O1} and above.
2308 For example, GCC will warn about @code{i} being uninitialized in the
2309 following snippet only when @option{-Winit-self} has been specified:
2320 @item -Wimplicit-int
2321 @opindex Wimplicit-int
2322 Warn when a declaration does not specify a type.
2323 This warning is enabled by @option{-Wall}.
2325 @item -Wimplicit-function-declaration
2326 @itemx -Werror-implicit-function-declaration
2327 @opindex Wimplicit-function-declaration
2328 @opindex Werror-implicit-function-declaration
2329 Give a warning (or error) whenever a function is used before being
2330 declared. The form @option{-Wno-error-implicit-function-declaration}
2332 This warning is enabled by @option{-Wall} (as a warning, not an error).
2336 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2337 This warning is enabled by @option{-Wall}.
2341 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2342 function with external linkage, returning int, taking either zero
2343 arguments, two, or three arguments of appropriate types.
2344 This warning is enabled by @option{-Wall}.
2346 @item -Wmissing-braces
2347 @opindex Wmissing-braces
2348 Warn if an aggregate or union initializer is not fully bracketed. In
2349 the following example, the initializer for @samp{a} is not fully
2350 bracketed, but that for @samp{b} is fully bracketed.
2353 int a[2][2] = @{ 0, 1, 2, 3 @};
2354 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2357 This warning is enabled by @option{-Wall}.
2359 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2360 @opindex Wmissing-include-dirs
2361 Warn if a user-supplied include directory does not exist.
2364 @opindex Wparentheses
2365 Warn if parentheses are omitted in certain contexts, such
2366 as when there is an assignment in a context where a truth value
2367 is expected, or when operators are nested whose precedence people
2368 often get confused about. Only the warning for an assignment used as
2369 a truth value is supported when compiling C++; the other warnings are
2370 only supported when compiling C@.
2372 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2373 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2374 interpretation from that of ordinary mathematical notation.
2376 Also warn about constructions where there may be confusion to which
2377 @code{if} statement an @code{else} branch belongs. Here is an example of
2392 In C, every @code{else} branch belongs to the innermost possible @code{if}
2393 statement, which in this example is @code{if (b)}. This is often not
2394 what the programmer expected, as illustrated in the above example by
2395 indentation the programmer chose. When there is the potential for this
2396 confusion, GCC will issue a warning when this flag is specified.
2397 To eliminate the warning, add explicit braces around the innermost
2398 @code{if} statement so there is no way the @code{else} could belong to
2399 the enclosing @code{if}. The resulting code would look like this:
2415 This warning is enabled by @option{-Wall}.
2417 @item -Wsequence-point
2418 @opindex Wsequence-point
2419 Warn about code that may have undefined semantics because of violations
2420 of sequence point rules in the C standard.
2422 The C standard defines the order in which expressions in a C program are
2423 evaluated in terms of @dfn{sequence points}, which represent a partial
2424 ordering between the execution of parts of the program: those executed
2425 before the sequence point, and those executed after it. These occur
2426 after the evaluation of a full expression (one which is not part of a
2427 larger expression), after the evaluation of the first operand of a
2428 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2429 function is called (but after the evaluation of its arguments and the
2430 expression denoting the called function), and in certain other places.
2431 Other than as expressed by the sequence point rules, the order of
2432 evaluation of subexpressions of an expression is not specified. All
2433 these rules describe only a partial order rather than a total order,
2434 since, for example, if two functions are called within one expression
2435 with no sequence point between them, the order in which the functions
2436 are called is not specified. However, the standards committee have
2437 ruled that function calls do not overlap.
2439 It is not specified when between sequence points modifications to the
2440 values of objects take effect. Programs whose behavior depends on this
2441 have undefined behavior; the C standard specifies that ``Between the
2442 previous and next sequence point an object shall have its stored value
2443 modified at most once by the evaluation of an expression. Furthermore,
2444 the prior value shall be read only to determine the value to be
2445 stored.''. If a program breaks these rules, the results on any
2446 particular implementation are entirely unpredictable.
2448 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2449 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2450 diagnosed by this option, and it may give an occasional false positive
2451 result, but in general it has been found fairly effective at detecting
2452 this sort of problem in programs.
2454 The present implementation of this option only works for C programs. A
2455 future implementation may also work for C++ programs.
2457 The C standard is worded confusingly, therefore there is some debate
2458 over the precise meaning of the sequence point rules in subtle cases.
2459 Links to discussions of the problem, including proposed formal
2460 definitions, may be found on the GCC readings page, at
2461 @w{@uref{http://gcc.gnu.org/readings.html}}.
2463 This warning is enabled by @option{-Wall}.
2466 @opindex Wreturn-type
2467 Warn whenever a function is defined with a return-type that defaults to
2468 @code{int}. Also warn about any @code{return} statement with no
2469 return-value in a function whose return-type is not @code{void}.
2471 For C, also warn if the return type of a function has a type qualifier
2472 such as @code{const}. Such a type qualifier has no effect, since the
2473 value returned by a function is not an lvalue. ISO C prohibits
2474 qualified @code{void} return types on function definitions, so such
2475 return types always receive a warning even without this option.
2477 For C++, a function without return type always produces a diagnostic
2478 message, even when @option{-Wno-return-type} is specified. The only
2479 exceptions are @samp{main} and functions defined in system headers.
2481 This warning is enabled by @option{-Wall}.
2485 Warn whenever a @code{switch} statement has an index of enumerated type
2486 and lacks a @code{case} for one or more of the named codes of that
2487 enumeration. (The presence of a @code{default} label prevents this
2488 warning.) @code{case} labels outside the enumeration range also
2489 provoke warnings when this option is used.
2490 This warning is enabled by @option{-Wall}.
2492 @item -Wswitch-default
2493 @opindex Wswitch-switch
2494 Warn whenever a @code{switch} statement does not have a @code{default}
2498 @opindex Wswitch-enum
2499 Warn whenever a @code{switch} statement has an index of enumerated type
2500 and lacks a @code{case} for one or more of the named codes of that
2501 enumeration. @code{case} labels outside the enumeration range also
2502 provoke warnings when this option is used.
2506 Warn if any trigraphs are encountered that might change the meaning of
2507 the program (trigraphs within comments are not warned about).
2508 This warning is enabled by @option{-Wall}.
2510 @item -Wunused-function
2511 @opindex Wunused-function
2512 Warn whenever a static function is declared but not defined or a
2513 non\-inline static function is unused.
2514 This warning is enabled by @option{-Wall}.
2516 @item -Wunused-label
2517 @opindex Wunused-label
2518 Warn whenever a label is declared but not used.
2519 This warning is enabled by @option{-Wall}.
2521 To suppress this warning use the @samp{unused} attribute
2522 (@pxref{Variable Attributes}).
2524 @item -Wunused-parameter
2525 @opindex Wunused-parameter
2526 Warn whenever a function parameter is unused aside from its declaration.
2528 To suppress this warning use the @samp{unused} attribute
2529 (@pxref{Variable Attributes}).
2531 @item -Wunused-variable
2532 @opindex Wunused-variable
2533 Warn whenever a local variable or non-constant static variable is unused
2534 aside from its declaration
2535 This warning is enabled by @option{-Wall}.
2537 To suppress this warning use the @samp{unused} attribute
2538 (@pxref{Variable Attributes}).
2540 @item -Wunused-value
2541 @opindex Wunused-value
2542 Warn whenever a statement computes a result that is explicitly not used.
2543 This warning is enabled by @option{-Wall}.
2545 To suppress this warning cast the expression to @samp{void}.
2549 All the above @option{-Wunused} options combined.
2551 In order to get a warning about an unused function parameter, you must
2552 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2553 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2555 @item -Wuninitialized
2556 @opindex Wuninitialized
2557 Warn if an automatic variable is used without first being initialized or
2558 if a variable may be clobbered by a @code{setjmp} call.
2560 These warnings are possible only in optimizing compilation,
2561 because they require data flow information that is computed only
2562 when optimizing. If you don't specify @option{-O}, you simply won't
2565 If you want to warn about code which uses the uninitialized value of the
2566 variable in its own initializer, use the @option{-Winit-self} option.
2568 These warnings occur for individual uninitialized or clobbered
2569 elements of structure, union or array variables as well as for
2570 variables which are uninitialized or clobbered as a whole. They do
2571 not occur for variables or elements declared @code{volatile}. Because
2572 these warnings depend on optimization, the exact variables or elements
2573 for which there are warnings will depend on the precise optimization
2574 options and version of GCC used.
2576 Note that there may be no warning about a variable that is used only
2577 to compute a value that itself is never used, because such
2578 computations may be deleted by data flow analysis before the warnings
2581 These warnings are made optional because GCC is not smart
2582 enough to see all the reasons why the code might be correct
2583 despite appearing to have an error. Here is one example of how
2604 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2605 always initialized, but GCC doesn't know this. Here is
2606 another common case:
2611 if (change_y) save_y = y, y = new_y;
2613 if (change_y) y = save_y;
2618 This has no bug because @code{save_y} is used only if it is set.
2620 @cindex @code{longjmp} warnings
2621 This option also warns when a non-volatile automatic variable might be
2622 changed by a call to @code{longjmp}. These warnings as well are possible
2623 only in optimizing compilation.
2625 The compiler sees only the calls to @code{setjmp}. It cannot know
2626 where @code{longjmp} will be called; in fact, a signal handler could
2627 call it at any point in the code. As a result, you may get a warning
2628 even when there is in fact no problem because @code{longjmp} cannot
2629 in fact be called at the place which would cause a problem.
2631 Some spurious warnings can be avoided if you declare all the functions
2632 you use that never return as @code{noreturn}. @xref{Function
2635 This warning is enabled by @option{-Wall}.
2637 @item -Wunknown-pragmas
2638 @opindex Wunknown-pragmas
2639 @cindex warning for unknown pragmas
2640 @cindex unknown pragmas, warning
2641 @cindex pragmas, warning of unknown
2642 Warn when a #pragma directive is encountered which is not understood by
2643 GCC@. If this command line option is used, warnings will even be issued
2644 for unknown pragmas in system header files. This is not the case if
2645 the warnings were only enabled by the @option{-Wall} command line option.
2647 @item -Wstrict-aliasing
2648 @opindex Wstrict-aliasing
2649 This option is only active when @option{-fstrict-aliasing} is active.
2650 It warns about code which might break the strict aliasing rules that the
2651 compiler is using for optimization. The warning does not catch all
2652 cases, but does attempt to catch the more common pitfalls. It is
2653 included in @option{-Wall}.
2655 @item -Wstrict-aliasing=2
2656 @opindex Wstrict-aliasing=2
2657 This option is only active when @option{-fstrict-aliasing} is active.
2658 It warns about all code which might break the strict aliasing rules that the
2659 compiler is using for optimization. This warning catches all cases, but
2660 it will also give a warning for some ambiguous cases that are safe.
2664 All of the above @samp{-W} options combined. This enables all the
2665 warnings about constructions that some users consider questionable, and
2666 that are easy to avoid (or modify to prevent the warning), even in
2667 conjunction with macros. This also enables some language-specific
2668 warnings described in @ref{C++ Dialect Options} and
2669 @ref{Objective-C and Objective-C++ Dialect Options}.
2672 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2673 Some of them warn about constructions that users generally do not
2674 consider questionable, but which occasionally you might wish to check
2675 for; others warn about constructions that are necessary or hard to avoid
2676 in some cases, and there is no simple way to modify the code to suppress
2683 (This option used to be called @option{-W}. The older name is still
2684 supported, but the newer name is more descriptive.) Print extra warning
2685 messages for these events:
2689 A function can return either with or without a value. (Falling
2690 off the end of the function body is considered returning without
2691 a value.) For example, this function would evoke such a
2705 An expression-statement or the left-hand side of a comma expression
2706 contains no side effects.
2707 To suppress the warning, cast the unused expression to void.
2708 For example, an expression such as @samp{x[i,j]} will cause a warning,
2709 but @samp{x[(void)i,j]} will not.
2712 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2715 Storage-class specifiers like @code{static} are not the first things in
2716 a declaration. According to the C Standard, this usage is obsolescent.
2719 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2723 A comparison between signed and unsigned values could produce an
2724 incorrect result when the signed value is converted to unsigned.
2725 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2728 An aggregate has an initializer which does not initialize all members.
2729 This warning can be independently controlled by
2730 @option{-Wmissing-field-initializers}.
2733 A function parameter is declared without a type specifier in K&R-style
2741 An empty body occurs in an @samp{if} or @samp{else} statement.
2744 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2745 @samp{>}, or @samp{>=}.
2748 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2751 Any of several floating-point events that often indicate errors, such as
2752 overflow, underflow, loss of precision, etc.
2754 @item @r{(C++ only)}
2755 An enumerator and a non-enumerator both appear in a conditional expression.
2757 @item @r{(C++ only)}
2758 A non-static reference or non-static @samp{const} member appears in a
2759 class without constructors.
2761 @item @r{(C++ only)}
2762 Ambiguous virtual bases.
2764 @item @r{(C++ only)}
2765 Subscripting an array which has been declared @samp{register}.
2767 @item @r{(C++ only)}
2768 Taking the address of a variable which has been declared @samp{register}.
2770 @item @r{(C++ only)}
2771 A base class is not initialized in a derived class' copy constructor.
2774 @item -Wno-div-by-zero
2775 @opindex Wno-div-by-zero
2776 @opindex Wdiv-by-zero
2777 Do not warn about compile-time integer division by zero. Floating point
2778 division by zero is not warned about, as it can be a legitimate way of
2779 obtaining infinities and NaNs.
2781 @item -Wsystem-headers
2782 @opindex Wsystem-headers
2783 @cindex warnings from system headers
2784 @cindex system headers, warnings from
2785 Print warning messages for constructs found in system header files.
2786 Warnings from system headers are normally suppressed, on the assumption
2787 that they usually do not indicate real problems and would only make the
2788 compiler output harder to read. Using this command line option tells
2789 GCC to emit warnings from system headers as if they occurred in user
2790 code. However, note that using @option{-Wall} in conjunction with this
2791 option will @emph{not} warn about unknown pragmas in system
2792 headers---for that, @option{-Wunknown-pragmas} must also be used.
2795 @opindex Wfloat-equal
2796 Warn if floating point values are used in equality comparisons.
2798 The idea behind this is that sometimes it is convenient (for the
2799 programmer) to consider floating-point values as approximations to
2800 infinitely precise real numbers. If you are doing this, then you need
2801 to compute (by analyzing the code, or in some other way) the maximum or
2802 likely maximum error that the computation introduces, and allow for it
2803 when performing comparisons (and when producing output, but that's a
2804 different problem). In particular, instead of testing for equality, you
2805 would check to see whether the two values have ranges that overlap; and
2806 this is done with the relational operators, so equality comparisons are
2809 @item -Wtraditional @r{(C only)}
2810 @opindex Wtraditional
2811 Warn about certain constructs that behave differently in traditional and
2812 ISO C@. Also warn about ISO C constructs that have no traditional C
2813 equivalent, and/or problematic constructs which should be avoided.
2817 Macro parameters that appear within string literals in the macro body.
2818 In traditional C macro replacement takes place within string literals,
2819 but does not in ISO C@.
2822 In traditional C, some preprocessor directives did not exist.
2823 Traditional preprocessors would only consider a line to be a directive
2824 if the @samp{#} appeared in column 1 on the line. Therefore
2825 @option{-Wtraditional} warns about directives that traditional C
2826 understands but would ignore because the @samp{#} does not appear as the
2827 first character on the line. It also suggests you hide directives like
2828 @samp{#pragma} not understood by traditional C by indenting them. Some
2829 traditional implementations would not recognize @samp{#elif}, so it
2830 suggests avoiding it altogether.
2833 A function-like macro that appears without arguments.
2836 The unary plus operator.
2839 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2840 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2841 constants.) Note, these suffixes appear in macros defined in the system
2842 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2843 Use of these macros in user code might normally lead to spurious
2844 warnings, however GCC's integrated preprocessor has enough context to
2845 avoid warning in these cases.
2848 A function declared external in one block and then used after the end of
2852 A @code{switch} statement has an operand of type @code{long}.
2855 A non-@code{static} function declaration follows a @code{static} one.
2856 This construct is not accepted by some traditional C compilers.
2859 The ISO type of an integer constant has a different width or
2860 signedness from its traditional type. This warning is only issued if
2861 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2862 typically represent bit patterns, are not warned about.
2865 Usage of ISO string concatenation is detected.
2868 Initialization of automatic aggregates.
2871 Identifier conflicts with labels. Traditional C lacks a separate
2872 namespace for labels.
2875 Initialization of unions. If the initializer is zero, the warning is
2876 omitted. This is done under the assumption that the zero initializer in
2877 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2878 initializer warnings and relies on default initialization to zero in the
2882 Conversions by prototypes between fixed/floating point values and vice
2883 versa. The absence of these prototypes when compiling with traditional
2884 C would cause serious problems. This is a subset of the possible
2885 conversion warnings, for the full set use @option{-Wconversion}.
2888 Use of ISO C style function definitions. This warning intentionally is
2889 @emph{not} issued for prototype declarations or variadic functions
2890 because these ISO C features will appear in your code when using
2891 libiberty's traditional C compatibility macros, @code{PARAMS} and
2892 @code{VPARAMS}. This warning is also bypassed for nested functions
2893 because that feature is already a GCC extension and thus not relevant to
2894 traditional C compatibility.
2897 @item -Wdeclaration-after-statement @r{(C only)}
2898 @opindex Wdeclaration-after-statement
2899 Warn when a declaration is found after a statement in a block. This
2900 construct, known from C++, was introduced with ISO C99 and is by default
2901 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2902 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2906 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2908 @item -Wno-endif-labels
2909 @opindex Wno-endif-labels
2910 @opindex Wendif-labels
2911 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2915 Warn whenever a local variable shadows another local variable, parameter or
2916 global variable or whenever a built-in function is shadowed.
2918 @item -Wlarger-than-@var{len}
2919 @opindex Wlarger-than
2920 Warn whenever an object of larger than @var{len} bytes is defined.
2922 @item -Wpointer-arith
2923 @opindex Wpointer-arith
2924 Warn about anything that depends on the ``size of'' a function type or
2925 of @code{void}. GNU C assigns these types a size of 1, for
2926 convenience in calculations with @code{void *} pointers and pointers
2929 @item -Wbad-function-cast @r{(C only)}
2930 @opindex Wbad-function-cast
2931 Warn whenever a function call is cast to a non-matching type.
2932 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2936 Warn whenever a pointer is cast so as to remove a type qualifier from
2937 the target type. For example, warn if a @code{const char *} is cast
2938 to an ordinary @code{char *}.
2941 @opindex Wcast-align
2942 Warn whenever a pointer is cast such that the required alignment of the
2943 target is increased. For example, warn if a @code{char *} is cast to
2944 an @code{int *} on machines where integers can only be accessed at
2945 two- or four-byte boundaries.
2947 @item -Wwrite-strings
2948 @opindex Wwrite-strings
2949 When compiling C, give string constants the type @code{const
2950 char[@var{length}]} so that
2951 copying the address of one into a non-@code{const} @code{char *}
2952 pointer will get a warning; when compiling C++, warn about the
2953 deprecated conversion from string constants to @code{char *}.
2954 These warnings will help you find at
2955 compile time code that can try to write into a string constant, but
2956 only if you have been very careful about using @code{const} in
2957 declarations and prototypes. Otherwise, it will just be a nuisance;
2958 this is why we did not make @option{-Wall} request these warnings.
2961 @opindex Wconversion
2962 Warn if a prototype causes a type conversion that is different from what
2963 would happen to the same argument in the absence of a prototype. This
2964 includes conversions of fixed point to floating and vice versa, and
2965 conversions changing the width or signedness of a fixed point argument
2966 except when the same as the default promotion.
2968 Also, warn if a negative integer constant expression is implicitly
2969 converted to an unsigned type. For example, warn about the assignment
2970 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2971 casts like @code{(unsigned) -1}.
2973 @item -Wsign-compare
2974 @opindex Wsign-compare
2975 @cindex warning for comparison of signed and unsigned values
2976 @cindex comparison of signed and unsigned values, warning
2977 @cindex signed and unsigned values, comparison warning
2978 Warn when a comparison between signed and unsigned values could produce
2979 an incorrect result when the signed value is converted to unsigned.
2980 This warning is also enabled by @option{-Wextra}; to get the other warnings
2981 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2983 @item -Waggregate-return
2984 @opindex Waggregate-return
2985 Warn if any functions that return structures or unions are defined or
2986 called. (In languages where you can return an array, this also elicits
2989 @item -Wstrict-prototypes @r{(C only)}
2990 @opindex Wstrict-prototypes
2991 Warn if a function is declared or defined without specifying the
2992 argument types. (An old-style function definition is permitted without
2993 a warning if preceded by a declaration which specifies the argument
2996 @item -Wold-style-definition @r{(C only)}
2997 @opindex Wold-style-definition
2998 Warn if an old-style function definition is used. A warning is given
2999 even if there is a previous prototype.
3001 @item -Wmissing-prototypes @r{(C only)}
3002 @opindex Wmissing-prototypes
3003 Warn if a global function is defined without a previous prototype
3004 declaration. This warning is issued even if the definition itself
3005 provides a prototype. The aim is to detect global functions that fail
3006 to be declared in header files.
3008 @item -Wmissing-declarations @r{(C only)}
3009 @opindex Wmissing-declarations
3010 Warn if a global function is defined without a previous declaration.
3011 Do so even if the definition itself provides a prototype.
3012 Use this option to detect global functions that are not declared in
3015 @item -Wmissing-field-initializers
3016 @opindex Wmissing-field-initializers
3019 Warn if a structure's initializer has some fields missing. For
3020 example, the following code would cause such a warning, because
3021 @code{x.h} is implicitly zero:
3024 struct s @{ int f, g, h; @};
3025 struct s x = @{ 3, 4 @};
3028 This option does not warn about designated initializers, so the following
3029 modification would not trigger a warning:
3032 struct s @{ int f, g, h; @};
3033 struct s x = @{ .f = 3, .g = 4 @};
3036 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3037 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3039 @item -Wmissing-noreturn
3040 @opindex Wmissing-noreturn
3041 Warn about functions which might be candidates for attribute @code{noreturn}.
3042 Note these are only possible candidates, not absolute ones. Care should
3043 be taken to manually verify functions actually do not ever return before
3044 adding the @code{noreturn} attribute, otherwise subtle code generation
3045 bugs could be introduced. You will not get a warning for @code{main} in
3046 hosted C environments.
3048 @item -Wmissing-format-attribute
3049 @opindex Wmissing-format-attribute
3051 If @option{-Wformat} is enabled, also warn about functions which might be
3052 candidates for @code{format} attributes. Note these are only possible
3053 candidates, not absolute ones. GCC will guess that @code{format}
3054 attributes might be appropriate for any function that calls a function
3055 like @code{vprintf} or @code{vscanf}, but this might not always be the
3056 case, and some functions for which @code{format} attributes are
3057 appropriate may not be detected. This option has no effect unless
3058 @option{-Wformat} is enabled (possibly by @option{-Wall}).
3060 @item -Wno-multichar
3061 @opindex Wno-multichar
3063 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3064 Usually they indicate a typo in the user's code, as they have
3065 implementation-defined values, and should not be used in portable code.
3067 @item -Wnormalized=<none|id|nfc|nfkc>
3068 @opindex Wnormalized
3071 @cindex character set, input normalization
3072 In ISO C and ISO C++, two identifiers are different if they are
3073 different sequences of characters. However, sometimes when characters
3074 outside the basic ASCII character set are used, you can have two
3075 different character sequences that look the same. To avoid confusion,
3076 the ISO 10646 standard sets out some @dfn{normalization rules} which
3077 when applied ensure that two sequences that look the same are turned into
3078 the same sequence. GCC can warn you if you are using identifiers which
3079 have not been normalized; this option controls that warning.
3081 There are four levels of warning that GCC supports. The default is
3082 @option{-Wnormalized=nfc}, which warns about any identifier which is
3083 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3084 recommended form for most uses.
3086 Unfortunately, there are some characters which ISO C and ISO C++ allow
3087 in identifiers that when turned into NFC aren't allowable as
3088 identifiers. That is, there's no way to use these symbols in portable
3089 ISO C or C++ and have all your identifiers in NFC.
3090 @option{-Wnormalized=id} suppresses the warning for these characters.
3091 It is hoped that future versions of the standards involved will correct
3092 this, which is why this option is not the default.
3094 You can switch the warning off for all characters by writing
3095 @option{-Wnormalized=none}. You would only want to do this if you
3096 were using some other normalization scheme (like ``D''), because
3097 otherwise you can easily create bugs that are literally impossible to see.
3099 Some characters in ISO 10646 have distinct meanings but look identical
3100 in some fonts or display methodologies, especially once formatting has
3101 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3102 LETTER N'', will display just like a regular @code{n} which has been
3103 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3104 normalisation scheme to convert all these into a standard form as
3105 well, and GCC will warn if your code is not in NFKC if you use
3106 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3107 about every identifier that contains the letter O because it might be
3108 confused with the digit 0, and so is not the default, but may be
3109 useful as a local coding convention if the programming environment is
3110 unable to be fixed to display these characters distinctly.
3112 @item -Wno-deprecated-declarations
3113 @opindex Wno-deprecated-declarations
3114 Do not warn about uses of functions, variables, and types marked as
3115 deprecated by using the @code{deprecated} attribute.
3116 (@pxref{Function Attributes}, @pxref{Variable Attributes},
3117 @pxref{Type Attributes}.)
3121 Warn if a structure is given the packed attribute, but the packed
3122 attribute has no effect on the layout or size of the structure.
3123 Such structures may be mis-aligned for little benefit. For
3124 instance, in this code, the variable @code{f.x} in @code{struct bar}
3125 will be misaligned even though @code{struct bar} does not itself
3126 have the packed attribute:
3133 @} __attribute__((packed));
3143 Warn if padding is included in a structure, either to align an element
3144 of the structure or to align the whole structure. Sometimes when this
3145 happens it is possible to rearrange the fields of the structure to
3146 reduce the padding and so make the structure smaller.
3148 @item -Wredundant-decls
3149 @opindex Wredundant-decls
3150 Warn if anything is declared more than once in the same scope, even in
3151 cases where multiple declaration is valid and changes nothing.
3153 @item -Wnested-externs @r{(C only)}
3154 @opindex Wnested-externs
3155 Warn if an @code{extern} declaration is encountered within a function.
3157 @item -Wunreachable-code
3158 @opindex Wunreachable-code
3159 Warn if the compiler detects that code will never be executed.
3161 This option is intended to warn when the compiler detects that at
3162 least a whole line of source code will never be executed, because
3163 some condition is never satisfied or because it is after a
3164 procedure that never returns.
3166 It is possible for this option to produce a warning even though there
3167 are circumstances under which part of the affected line can be executed,
3168 so care should be taken when removing apparently-unreachable code.
3170 For instance, when a function is inlined, a warning may mean that the
3171 line is unreachable in only one inlined copy of the function.
3173 This option is not made part of @option{-Wall} because in a debugging
3174 version of a program there is often substantial code which checks
3175 correct functioning of the program and is, hopefully, unreachable
3176 because the program does work. Another common use of unreachable
3177 code is to provide behavior which is selectable at compile-time.
3181 Warn if a function can not be inlined and it was declared as inline.
3182 Even with this option, the compiler will not warn about failures to
3183 inline functions declared in system headers.
3185 The compiler uses a variety of heuristics to determine whether or not
3186 to inline a function. For example, the compiler takes into account
3187 the size of the function being inlined and the amount of inlining
3188 that has already been done in the current function. Therefore,
3189 seemingly insignificant changes in the source program can cause the
3190 warnings produced by @option{-Winline} to appear or disappear.
3192 @item -Wno-invalid-offsetof @r{(C++ only)}
3193 @opindex Wno-invalid-offsetof
3194 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3195 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3196 to a non-POD type is undefined. In existing C++ implementations,
3197 however, @samp{offsetof} typically gives meaningful results even when
3198 applied to certain kinds of non-POD types. (Such as a simple
3199 @samp{struct} that fails to be a POD type only by virtue of having a
3200 constructor.) This flag is for users who are aware that they are
3201 writing nonportable code and who have deliberately chosen to ignore the
3204 The restrictions on @samp{offsetof} may be relaxed in a future version
3205 of the C++ standard.
3207 @item -Wno-int-to-pointer-cast @r{(C only)}
3208 @opindex Wno-int-to-pointer-cast
3209 Suppress warnings from casts to pointer type of an integer of a
3212 @item -Wno-pointer-to-int-cast @r{(C only)}
3213 @opindex Wno-pointer-to-int-cast
3214 Suppress warnings from casts from a pointer to an integer type of a
3218 @opindex Winvalid-pch
3219 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3220 the search path but can't be used.
3224 @opindex Wno-long-long
3225 Warn if @samp{long long} type is used. This is default. To inhibit
3226 the warning messages, use @option{-Wno-long-long}. Flags
3227 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3228 only when @option{-pedantic} flag is used.
3230 @item -Wvariadic-macros
3231 @opindex Wvariadic-macros
3232 @opindex Wno-variadic-macros
3233 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3234 alternate syntax when in pedantic ISO C99 mode. This is default.
3235 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3237 @item -Wdisabled-optimization
3238 @opindex Wdisabled-optimization
3239 Warn if a requested optimization pass is disabled. This warning does
3240 not generally indicate that there is anything wrong with your code; it
3241 merely indicates that GCC's optimizers were unable to handle the code
3242 effectively. Often, the problem is that your code is too big or too
3243 complex; GCC will refuse to optimize programs when the optimization
3244 itself is likely to take inordinate amounts of time.
3246 @item -Wno-pointer-sign
3247 @opindex Wno-pointer-sign
3248 Don't warn for pointer argument passing or assignment with different signedness.
3249 Only useful in the negative form since this warning is enabled by default.
3250 This option is only supported for C and Objective-C@.
3254 Make all warnings into errors.
3257 @node Debugging Options
3258 @section Options for Debugging Your Program or GCC
3259 @cindex options, debugging
3260 @cindex debugging information options
3262 GCC has various special options that are used for debugging
3263 either your program or GCC:
3268 Produce debugging information in the operating system's native format
3269 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3272 On most systems that use stabs format, @option{-g} enables use of extra
3273 debugging information that only GDB can use; this extra information
3274 makes debugging work better in GDB but will probably make other debuggers
3276 refuse to read the program. If you want to control for certain whether
3277 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3278 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3280 GCC allows you to use @option{-g} with
3281 @option{-O}. The shortcuts taken by optimized code may occasionally
3282 produce surprising results: some variables you declared may not exist
3283 at all; flow of control may briefly move where you did not expect it;
3284 some statements may not be executed because they compute constant
3285 results or their values were already at hand; some statements may
3286 execute in different places because they were moved out of loops.
3288 Nevertheless it proves possible to debug optimized output. This makes
3289 it reasonable to use the optimizer for programs that might have bugs.
3291 The following options are useful when GCC is generated with the
3292 capability for more than one debugging format.
3296 Produce debugging information for use by GDB@. This means to use the
3297 most expressive format available (DWARF 2, stabs, or the native format
3298 if neither of those are supported), including GDB extensions if at all
3303 Produce debugging information in stabs format (if that is supported),
3304 without GDB extensions. This is the format used by DBX on most BSD
3305 systems. On MIPS, Alpha and System V Release 4 systems this option
3306 produces stabs debugging output which is not understood by DBX or SDB@.
3307 On System V Release 4 systems this option requires the GNU assembler.
3309 @item -feliminate-unused-debug-symbols
3310 @opindex feliminate-unused-debug-symbols
3311 Produce debugging information in stabs format (if that is supported),
3312 for only symbols that are actually used.
3316 Produce debugging information in stabs format (if that is supported),
3317 using GNU extensions understood only by the GNU debugger (GDB)@. The
3318 use of these extensions is likely to make other debuggers crash or
3319 refuse to read the program.
3323 Produce debugging information in COFF format (if that is supported).
3324 This is the format used by SDB on most System V systems prior to
3329 Produce debugging information in XCOFF format (if that is supported).
3330 This is the format used by the DBX debugger on IBM RS/6000 systems.
3334 Produce debugging information in XCOFF format (if that is supported),
3335 using GNU extensions understood only by the GNU debugger (GDB)@. The
3336 use of these extensions is likely to make other debuggers crash or
3337 refuse to read the program, and may cause assemblers other than the GNU
3338 assembler (GAS) to fail with an error.
3342 Produce debugging information in DWARF version 2 format (if that is
3343 supported). This is the format used by DBX on IRIX 6. With this
3344 option, GCC uses features of DWARF version 3 when they are useful;
3345 version 3 is upward compatible with version 2, but may still cause
3346 problems for older debuggers.
3350 Produce debugging information in VMS debug format (if that is
3351 supported). This is the format used by DEBUG on VMS systems.
3354 @itemx -ggdb@var{level}
3355 @itemx -gstabs@var{level}
3356 @itemx -gcoff@var{level}
3357 @itemx -gxcoff@var{level}
3358 @itemx -gvms@var{level}
3359 Request debugging information and also use @var{level} to specify how
3360 much information. The default level is 2.
3362 Level 1 produces minimal information, enough for making backtraces in
3363 parts of the program that you don't plan to debug. This includes
3364 descriptions of functions and external variables, but no information
3365 about local variables and no line numbers.
3367 Level 3 includes extra information, such as all the macro definitions
3368 present in the program. Some debuggers support macro expansion when
3369 you use @option{-g3}.
3371 @option{-gdwarf-2} does not accept a concatenated debug level, because
3372 GCC used to support an option @option{-gdwarf} that meant to generate
3373 debug information in version 1 of the DWARF format (which is very
3374 different from version 2), and it would have been too confusing. That
3375 debug format is long obsolete, but the option cannot be changed now.
3376 Instead use an additional @option{-g@var{level}} option to change the
3377 debug level for DWARF2.
3379 @item -feliminate-dwarf2-dups
3380 @opindex feliminate-dwarf2-dups
3381 Compress DWARF2 debugging information by eliminating duplicated
3382 information about each symbol. This option only makes sense when
3383 generating DWARF2 debugging information with @option{-gdwarf-2}.
3385 @cindex @command{prof}
3388 Generate extra code to write profile information suitable for the
3389 analysis program @command{prof}. You must use this option when compiling
3390 the source files you want data about, and you must also use it when
3393 @cindex @command{gprof}
3396 Generate extra code to write profile information suitable for the
3397 analysis program @command{gprof}. You must use this option when compiling
3398 the source files you want data about, and you must also use it when
3403 Makes the compiler print out each function name as it is compiled, and
3404 print some statistics about each pass when it finishes.
3407 @opindex ftime-report
3408 Makes the compiler print some statistics about the time consumed by each
3409 pass when it finishes.
3412 @opindex fmem-report
3413 Makes the compiler print some statistics about permanent memory
3414 allocation when it finishes.
3416 @item -fprofile-arcs
3417 @opindex fprofile-arcs
3418 Add code so that program flow @dfn{arcs} are instrumented. During
3419 execution the program records how many times each branch and call is
3420 executed and how many times it is taken or returns. When the compiled
3421 program exits it saves this data to a file called
3422 @file{@var{auxname}.gcda} for each source file. The data may be used for
3423 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3424 test coverage analysis (@option{-ftest-coverage}). Each object file's
3425 @var{auxname} is generated from the name of the output file, if
3426 explicitly specified and it is not the final executable, otherwise it is
3427 the basename of the source file. In both cases any suffix is removed
3428 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3429 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3431 @cindex @command{gcov}
3435 This option is used to compile and link code instrumented for coverage
3436 analysis. The option is a synonym for @option{-fprofile-arcs}
3437 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3438 linking). See the documentation for those options for more details.
3443 Compile the source files with @option{-fprofile-arcs} plus optimization
3444 and code generation options. For test coverage analysis, use the
3445 additional @option{-ftest-coverage} option. You do not need to profile
3446 every source file in a program.
3449 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3450 (the latter implies the former).
3453 Run the program on a representative workload to generate the arc profile
3454 information. This may be repeated any number of times. You can run
3455 concurrent instances of your program, and provided that the file system
3456 supports locking, the data files will be correctly updated. Also
3457 @code{fork} calls are detected and correctly handled (double counting
3461 For profile-directed optimizations, compile the source files again with
3462 the same optimization and code generation options plus
3463 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3464 Control Optimization}).
3467 For test coverage analysis, use @command{gcov} to produce human readable
3468 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3469 @command{gcov} documentation for further information.
3473 With @option{-fprofile-arcs}, for each function of your program GCC
3474 creates a program flow graph, then finds a spanning tree for the graph.
3475 Only arcs that are not on the spanning tree have to be instrumented: the
3476 compiler adds code to count the number of times that these arcs are
3477 executed. When an arc is the only exit or only entrance to a block, the
3478 instrumentation code can be added to the block; otherwise, a new basic
3479 block must be created to hold the instrumentation code.
3481 @item -ftree-based-profiling
3482 @opindex ftree-based-profiling
3483 This option is used in addition to @option{-fprofile-arcs} or
3484 @option{-fbranch-probabilities} to control whether those optimizations
3485 are performed on a tree-based or rtl-based internal representation.
3486 If you use this option when compiling with @option{-fprofile-arcs},
3487 you must also use it when compiling later with @option{-fbranch-probabilities}.
3488 Currently the tree-based optimization is in an early stage of
3489 development, and this option is recommended only for those people
3490 working on improving it.
3493 @item -ftest-coverage
3494 @opindex ftest-coverage
3495 Produce a notes file that the @command{gcov} code-coverage utility
3496 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3497 show program coverage. Each source file's note file is called
3498 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3499 above for a description of @var{auxname} and instructions on how to
3500 generate test coverage data. Coverage data will match the source files
3501 more closely, if you do not optimize.
3503 @item -d@var{letters}
3504 @item -fdump-rtl-@var{pass}
3506 Says to make debugging dumps during compilation at times specified by
3507 @var{letters}. This is used for debugging the RTL-based passes of the
3508 compiler. The file names for most of the dumps are made by appending a
3509 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3510 from the name of the output file, if explicitly specified and it is not
3511 an executable, otherwise it is the basename of the source file.
3513 Most debug dumps can be enabled either passing a letter to the @option{-d}
3514 option, or with a long @option{-fdump-rtl} switch; here are the possible
3515 letters for use in @var{letters} and @var{pass}, and their meanings:
3520 Annotate the assembler output with miscellaneous debugging information.
3523 @itemx -fdump-rtl-bp
3525 @opindex fdump-rtl-bp
3526 Dump after computing branch probabilities, to @file{@var{file}.09.bp}.
3529 @itemx -fdump-rtl-bbro
3531 @opindex fdump-rtl-bbro
3532 Dump after block reordering, to @file{@var{file}.30.bbro}.
3535 @itemx -fdump-rtl-combine
3537 @opindex fdump-rtl-combine
3538 Dump after instruction combination, to the file @file{@var{file}.17.combine}.
3541 @itemx -fdump-rtl-ce1
3542 @itemx -fdump-rtl-ce2
3544 @opindex fdump-rtl-ce1
3545 @opindex fdump-rtl-ce2
3546 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3547 first if conversion, to the file @file{@var{file}.11.ce1}. @option{-dC}
3548 and @option{-fdump-rtl-ce2} enable dumping after the second if
3549 conversion, to the file @file{@var{file}.18.ce2}.
3552 @itemx -fdump-rtl-btl
3553 @itemx -fdump-rtl-dbr
3555 @opindex fdump-rtl-btl
3556 @opindex fdump-rtl-dbr
3557 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3558 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3559 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3560 scheduling, to @file{@var{file}.36.dbr}.
3564 Dump all macro definitions, at the end of preprocessing, in addition to
3568 @itemx -fdump-rtl-ce3
3570 @opindex fdump-rtl-ce3
3571 Dump after the third if conversion, to @file{@var{file}.28.ce3}.
3574 @itemx -fdump-rtl-cfg
3575 @itemx -fdump-rtl-life
3577 @opindex fdump-rtl-cfg
3578 @opindex fdump-rtl-life
3579 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3580 and data flow analysis, to @file{@var{file}.08.cfg}. @option{-df}
3581 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3582 to @file{@var{file}.16.life}.
3585 @itemx -fdump-rtl-greg
3587 @opindex fdump-rtl-greg
3588 Dump after global register allocation, to @file{@var{file}.23.greg}.
3591 @itemx -fdump-rtl-gcse
3592 @itemx -fdump-rtl-bypass
3594 @opindex fdump-rtl-gcse
3595 @opindex fdump-rtl-bypass
3596 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3597 @file{@var{file}.05.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3598 enable dumping after jump bypassing and control flow optimizations, to
3599 @file{@var{file}.07.bypass}.
3602 @itemx -fdump-rtl-eh
3604 @opindex fdump-rtl-eh
3605 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3608 @itemx -fdump-rtl-sibling
3610 @opindex fdump-rtl-sibling
3611 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
3614 @itemx -fdump-rtl-jump
3616 @opindex fdump-rtl-jump
3617 Dump after the first jump optimization, to @file{@var{file}.03.jump}.
3620 @itemx -fdump-rtl-stack
3622 @opindex fdump-rtl-stack
3623 Dump after conversion from registers to stack, to @file{@var{file}.33.stack}.
3626 @itemx -fdump-rtl-lreg
3628 @opindex fdump-rtl-lreg
3629 Dump after local register allocation, to @file{@var{file}.22.lreg}.
3632 @itemx -fdump-rtl-loop
3633 @itemx -fdump-rtl-loop2
3635 @opindex fdump-rtl-loop
3636 @opindex fdump-rtl-loop2
3637 @option{-dL} and @option{-fdump-rtl-loop} enable dumping after the first
3638 loop optimization pass, to @file{@var{file}.06.loop}. @option{-dL} and
3639 @option{-fdump-rtl-loop2} enable dumping after the second pass, to
3640 @file{@var{file}.13.loop2}.
3643 @itemx -fdump-rtl-sms
3645 @opindex fdump-rtl-sms
3646 Dump after modulo scheduling, to @file{@var{file}.20.sms}.
3649 @itemx -fdump-rtl-mach
3651 @opindex fdump-rtl-mach
3652 Dump after performing the machine dependent reorganization pass, to
3653 @file{@var{file}.35.mach}.
3656 @itemx -fdump-rtl-rnreg
3658 @opindex fdump-rtl-rnreg
3659 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3662 @itemx -fdump-rtl-regmove
3664 @opindex fdump-rtl-regmove
3665 Dump after the register move pass, to @file{@var{file}.19.regmove}.
3668 @itemx -fdump-rtl-postreload
3670 @opindex fdump-rtl-postreload
3671 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3674 @itemx -fdump-rtl-expand
3676 @opindex fdump-rtl-expand
3677 Dump after RTL generation, to @file{@var{file}.00.expand}.
3680 @itemx -fdump-rtl-sched2
3682 @opindex fdump-rtl-sched2
3683 Dump after the second scheduling pass, to @file{@var{file}.32.sched2}.
3686 @itemx -fdump-rtl-cse
3688 @opindex fdump-rtl-cse
3689 Dump after CSE (including the jump optimization that sometimes follows
3690 CSE), to @file{@var{file}.04.cse}.
3693 @itemx -fdump-rtl-sched
3695 @opindex fdump-rtl-sched
3696 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3699 @itemx -fdump-rtl-cse2
3701 @opindex fdump-rtl-cse2
3702 Dump after the second CSE pass (including the jump optimization that
3703 sometimes follows CSE), to @file{@var{file}.15.cse2}.
3706 @itemx -fdump-rtl-tracer
3708 @opindex fdump-rtl-tracer
3709 Dump after running tracer, to @file{@var{file}.12.tracer}.
3712 @itemx -fdump-rtl-vpt
3713 @itemx -fdump-rtl-vartrack
3715 @opindex fdump-rtl-vpt
3716 @opindex fdump-rtl-vartrack
3717 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3718 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3719 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3720 to @file{@var{file}.34.vartrack}.
3723 @itemx -fdump-rtl-flow2
3725 @opindex fdump-rtl-flow2
3726 Dump after the second flow pass, to @file{@var{file}.26.flow2}.
3729 @itemx -fdump-rtl-peephole2
3731 @opindex fdump-rtl-peephole2
3732 Dump after the peephole pass, to @file{@var{file}.27.peephole2}.
3735 @itemx -fdump-rtl-web
3737 @opindex fdump-rtl-web
3738 Dump after live range splitting, to @file{@var{file}.14.web}.
3741 @itemx -fdump-rtl-all
3743 @opindex fdump-rtl-all
3744 Produce all the dumps listed above.
3748 Produce a core dump whenever an error occurs.
3752 Print statistics on memory usage, at the end of the run, to
3757 Annotate the assembler output with a comment indicating which
3758 pattern and alternative was used. The length of each instruction is
3763 Dump the RTL in the assembler output as a comment before each instruction.
3764 Also turns on @option{-dp} annotation.
3768 For each of the other indicated dump files (either with @option{-d} or
3769 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3770 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3774 Just generate RTL for a function instead of compiling it. Usually used
3775 with @samp{r} (@option{-fdump-rtl-expand}).
3779 Dump debugging information during parsing, to standard error.
3782 @item -fdump-unnumbered
3783 @opindex fdump-unnumbered
3784 When doing debugging dumps (see @option{-d} option above), suppress instruction
3785 numbers and line number note output. This makes it more feasible to
3786 use diff on debugging dumps for compiler invocations with different
3787 options, in particular with and without @option{-g}.
3789 @item -fdump-translation-unit @r{(C and C++ only)}
3790 @itemx -fdump-translation-unit-@var{options} @r{(C and C++ only)}
3791 @opindex fdump-translation-unit
3792 Dump a representation of the tree structure for the entire translation
3793 unit to a file. The file name is made by appending @file{.tu} to the
3794 source file name. If the @samp{-@var{options}} form is used, @var{options}
3795 controls the details of the dump as described for the
3796 @option{-fdump-tree} options.
3798 @item -fdump-class-hierarchy @r{(C++ only)}
3799 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3800 @opindex fdump-class-hierarchy
3801 Dump a representation of each class's hierarchy and virtual function
3802 table layout to a file. The file name is made by appending @file{.class}
3803 to the source file name. If the @samp{-@var{options}} form is used,
3804 @var{options} controls the details of the dump as described for the
3805 @option{-fdump-tree} options.
3807 @item -fdump-ipa-@var{switch}
3809 Control the dumping at various stages of inter-procedural analysis
3810 language tree to a file. The file name is generated by appending a switch
3811 specific suffix to the source file name. The following dumps are possible:
3815 Enables all inter-procedural analysis dumps; currently the only produced
3816 dump is the @samp{cgraph} dump.
3819 Dumps information about call-graph optimization, unused function removal,
3820 and inlining decisions.
3823 @item -fdump-tree-@var{switch} @r{(C and C++ only)}
3824 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C and C++ only)}
3826 Control the dumping at various stages of processing the intermediate
3827 language tree to a file. The file name is generated by appending a switch
3828 specific suffix to the source file name. If the @samp{-@var{options}}
3829 form is used, @var{options} is a list of @samp{-} separated options that
3830 control the details of the dump. Not all options are applicable to all
3831 dumps, those which are not meaningful will be ignored. The following
3832 options are available
3836 Print the address of each node. Usually this is not meaningful as it
3837 changes according to the environment and source file. Its primary use
3838 is for tying up a dump file with a debug environment.
3840 Inhibit dumping of members of a scope or body of a function merely
3841 because that scope has been reached. Only dump such items when they
3842 are directly reachable by some other path. When dumping pretty-printed
3843 trees, this option inhibits dumping the bodies of control structures.
3845 Print a raw representation of the tree. By default, trees are
3846 pretty-printed into a C-like representation.
3848 Enable more detailed dumps (not honored by every dump option).
3850 Enable dumping various statistics about the pass (not honored by every dump
3853 Enable showing basic block boundaries (disabled in raw dumps).
3855 Enable showing virtual operands for every statement.
3857 Enable showing line numbers for statements.
3859 Enable showing the unique ID (@code{DECL_UID}) for each variable.
3861 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
3864 The following tree dumps are possible:
3868 Dump before any tree based optimization, to @file{@var{file}.original}.
3871 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3874 Dump after function inlining, to @file{@var{file}.inlined}.
3877 @opindex fdump-tree-gimple
3878 Dump each function before and after the gimplification pass to a file. The
3879 file name is made by appending @file{.gimple} to the source file name.
3882 @opindex fdump-tree-cfg
3883 Dump the control flow graph of each function to a file. The file name is
3884 made by appending @file{.cfg} to the source file name.
3887 @opindex fdump-tree-vcg
3888 Dump the control flow graph of each function to a file in VCG format. The
3889 file name is made by appending @file{.vcg} to the source file name. Note
3890 that if the file contains more than one function, the generated file cannot
3891 be used directly by VCG@. You will need to cut and paste each function's
3892 graph into its own separate file first.
3895 @opindex fdump-tree-ch
3896 Dump each function after copying loop headers. The file name is made by
3897 appending @file{.ch} to the source file name.
3900 @opindex fdump-tree-ssa
3901 Dump SSA related information to a file. The file name is made by appending
3902 @file{.ssa} to the source file name.
3905 @opindex fdump-tree-salias
3906 Dump structure aliasing variable information to a file. This file name
3907 is made by appending @file{.salias} to the source file name.
3910 @opindex fdump-tree-alias
3911 Dump aliasing information for each function. The file name is made by
3912 appending @file{.alias} to the source file name.
3915 @opindex fdump-tree-ccp
3916 Dump each function after CCP@. The file name is made by appending
3917 @file{.ccp} to the source file name.
3920 @opindex fdump-tree-storeccp
3921 Dump each function after STORE-CCP. The file name is made by appending
3922 @file{.storeccp} to the source file name.
3925 @opindex fdump-tree-pre
3926 Dump trees after partial redundancy elimination. The file name is made
3927 by appending @file{.pre} to the source file name.
3930 @opindex fdump-tree-fre
3931 Dump trees after full redundancy elimination. The file name is made
3932 by appending @file{.fre} to the source file name.
3935 @opindex fdump-tree-copyprop
3936 Dump trees after copy propagation. The file name is made
3937 by appending @file{.copyprop} to the source file name.
3939 @item store_copyprop
3940 @opindex fdump-tree-store_copyprop
3941 Dump trees after store copy-propagation. The file name is made
3942 by appending @file{.store_copyprop} to the source file name.
3945 @opindex fdump-tree-dce
3946 Dump each function after dead code elimination. The file name is made by
3947 appending @file{.dce} to the source file name.
3950 @opindex fdump-tree-mudflap
3951 Dump each function after adding mudflap instrumentation. The file name is
3952 made by appending @file{.mudflap} to the source file name.
3955 @opindex fdump-tree-sra
3956 Dump each function after performing scalar replacement of aggregates. The
3957 file name is made by appending @file{.sra} to the source file name.
3960 @opindex fdump-tree-sink
3961 Dump each function after performing code sinking. The file name is made
3962 by appending @file{.sink} to the source file name.
3965 @opindex fdump-tree-dom
3966 Dump each function after applying dominator tree optimizations. The file
3967 name is made by appending @file{.dom} to the source file name.
3970 @opindex fdump-tree-dse
3971 Dump each function after applying dead store elimination. The file
3972 name is made by appending @file{.dse} to the source file name.
3975 @opindex fdump-tree-phiopt
3976 Dump each function after optimizing PHI nodes into straightline code. The file
3977 name is made by appending @file{.phiopt} to the source file name.
3980 @opindex fdump-tree-forwprop
3981 Dump each function after forward propagating single use variables. The file
3982 name is made by appending @file{.forwprop} to the source file name.
3985 @opindex fdump-tree-copyrename
3986 Dump each function after applying the copy rename optimization. The file
3987 name is made by appending @file{.copyrename} to the source file name.
3990 @opindex fdump-tree-nrv
3991 Dump each function after applying the named return value optimization on
3992 generic trees. The file name is made by appending @file{.nrv} to the source
3996 @opindex fdump-tree-vect
3997 Dump each function after applying vectorization of loops. The file name is
3998 made by appending @file{.vect} to the source file name.
4001 @opindex fdump-tree-all
4002 Enable all the available tree dumps with the flags provided in this option.
4005 @item -ftree-vectorizer-verbose=@var{n}
4006 @opindex ftree-vectorizer-verbose
4007 This option controls the amount of debugging output the vectorizer prints.
4008 This information is written to standard error, unless @option{-fdump-tree-all}
4009 or @option{-fdump-tree-vect} is specified, in which case it is output to the
4010 usual dump listing file, @file{.vect}.
4012 @item -frandom-seed=@var{string}
4013 @opindex frandom-string
4014 This option provides a seed that GCC uses when it would otherwise use
4015 random numbers. It is used to generate certain symbol names
4016 that have to be different in every compiled file. It is also used to
4017 place unique stamps in coverage data files and the object files that
4018 produce them. You can use the @option{-frandom-seed} option to produce
4019 reproducibly identical object files.
4021 The @var{string} should be different for every file you compile.
4023 @item -fsched-verbose=@var{n}
4024 @opindex fsched-verbose
4025 On targets that use instruction scheduling, this option controls the
4026 amount of debugging output the scheduler prints. This information is
4027 written to standard error, unless @option{-dS} or @option{-dR} is
4028 specified, in which case it is output to the usual dump
4029 listing file, @file{.sched} or @file{.sched2} respectively. However
4030 for @var{n} greater than nine, the output is always printed to standard
4033 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4034 same information as @option{-dRS}. For @var{n} greater than one, it
4035 also output basic block probabilities, detailed ready list information
4036 and unit/insn info. For @var{n} greater than two, it includes RTL
4037 at abort point, control-flow and regions info. And for @var{n} over
4038 four, @option{-fsched-verbose} also includes dependence info.
4042 Store the usual ``temporary'' intermediate files permanently; place them
4043 in the current directory and name them based on the source file. Thus,
4044 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4045 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4046 preprocessed @file{foo.i} output file even though the compiler now
4047 normally uses an integrated preprocessor.
4049 When used in combination with the @option{-x} command line option,
4050 @option{-save-temps} is sensible enough to avoid over writing an
4051 input source file with the same extension as an intermediate file.
4052 The corresponding intermediate file may be obtained by renaming the
4053 source file before using @option{-save-temps}.
4057 Report the CPU time taken by each subprocess in the compilation
4058 sequence. For C source files, this is the compiler proper and assembler
4059 (plus the linker if linking is done). The output looks like this:
4066 The first number on each line is the ``user time'', that is time spent
4067 executing the program itself. The second number is ``system time'',
4068 time spent executing operating system routines on behalf of the program.
4069 Both numbers are in seconds.
4071 @item -fvar-tracking
4072 @opindex fvar-tracking
4073 Run variable tracking pass. It computes where variables are stored at each
4074 position in code. Better debugging information is then generated
4075 (if the debugging information format supports this information).
4077 It is enabled by default when compiling with optimization (@option{-Os},
4078 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4079 the debug info format supports it.
4081 @item -print-file-name=@var{library}
4082 @opindex print-file-name
4083 Print the full absolute name of the library file @var{library} that
4084 would be used when linking---and don't do anything else. With this
4085 option, GCC does not compile or link anything; it just prints the
4088 @item -print-multi-directory
4089 @opindex print-multi-directory
4090 Print the directory name corresponding to the multilib selected by any
4091 other switches present in the command line. This directory is supposed
4092 to exist in @env{GCC_EXEC_PREFIX}.
4094 @item -print-multi-lib
4095 @opindex print-multi-lib
4096 Print the mapping from multilib directory names to compiler switches
4097 that enable them. The directory name is separated from the switches by
4098 @samp{;}, and each switch starts with an @samp{@@} instead of the
4099 @samp{-}, without spaces between multiple switches. This is supposed to
4100 ease shell-processing.
4102 @item -print-prog-name=@var{program}
4103 @opindex print-prog-name
4104 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4106 @item -print-libgcc-file-name
4107 @opindex print-libgcc-file-name
4108 Same as @option{-print-file-name=libgcc.a}.
4110 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4111 but you do want to link with @file{libgcc.a}. You can do
4114 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4117 @item -print-search-dirs
4118 @opindex print-search-dirs
4119 Print the name of the configured installation directory and a list of
4120 program and library directories @command{gcc} will search---and don't do anything else.
4122 This is useful when @command{gcc} prints the error message
4123 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4124 To resolve this you either need to put @file{cpp0} and the other compiler
4125 components where @command{gcc} expects to find them, or you can set the environment
4126 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4127 Don't forget the trailing @samp{/}.
4128 @xref{Environment Variables}.
4131 @opindex dumpmachine
4132 Print the compiler's target machine (for example,
4133 @samp{i686-pc-linux-gnu})---and don't do anything else.
4136 @opindex dumpversion
4137 Print the compiler version (for example, @samp{3.0})---and don't do
4142 Print the compiler's built-in specs---and don't do anything else. (This
4143 is used when GCC itself is being built.) @xref{Spec Files}.
4145 @item -feliminate-unused-debug-types
4146 @opindex feliminate-unused-debug-types
4147 Normally, when producing DWARF2 output, GCC will emit debugging
4148 information for all types declared in a compilation
4149 unit, regardless of whether or not they are actually used
4150 in that compilation unit. Sometimes this is useful, such as
4151 if, in the debugger, you want to cast a value to a type that is
4152 not actually used in your program (but is declared). More often,
4153 however, this results in a significant amount of wasted space.
4154 With this option, GCC will avoid producing debug symbol output
4155 for types that are nowhere used in the source file being compiled.
4158 @node Optimize Options
4159 @section Options That Control Optimization
4160 @cindex optimize options
4161 @cindex options, optimization
4163 These options control various sorts of optimizations.
4165 Without any optimization option, the compiler's goal is to reduce the
4166 cost of compilation and to make debugging produce the expected
4167 results. Statements are independent: if you stop the program with a
4168 breakpoint between statements, you can then assign a new value to any
4169 variable or change the program counter to any other statement in the
4170 function and get exactly the results you would expect from the source
4173 Turning on optimization flags makes the compiler attempt to improve
4174 the performance and/or code size at the expense of compilation time
4175 and possibly the ability to debug the program.
4177 The compiler performs optimization based on the knowledge it has of
4178 the program. Optimization levels @option{-O2} and above, in
4179 particular, enable @emph{unit-at-a-time} mode, which allows the
4180 compiler to consider information gained from later functions in
4181 the file when compiling a function. Compiling multiple files at
4182 once to a single output file in @emph{unit-at-a-time} mode allows
4183 the compiler to use information gained from all of the files when
4184 compiling each of them.
4186 Not all optimizations are controlled directly by a flag. Only
4187 optimizations that have a flag are listed.
4194 Optimize. Optimizing compilation takes somewhat more time, and a lot
4195 more memory for a large function.
4197 With @option{-O}, the compiler tries to reduce code size and execution
4198 time, without performing any optimizations that take a great deal of
4201 @option{-O} turns on the following optimization flags:
4202 @gccoptlist{-fdefer-pop @gol
4203 -fdelayed-branch @gol
4204 -fguess-branch-probability @gol
4205 -fcprop-registers @gol
4206 -floop-optimize @gol
4207 -fif-conversion @gol
4208 -fif-conversion2 @gol
4211 -ftree-dominator-opts @gol
4216 -ftree-copyrename @gol
4221 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4222 where doing so does not interfere with debugging.
4226 Optimize even more. GCC performs nearly all supported optimizations
4227 that do not involve a space-speed tradeoff. The compiler does not
4228 perform loop unrolling or function inlining when you specify @option{-O2}.
4229 As compared to @option{-O}, this option increases both compilation time
4230 and the performance of the generated code.
4232 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4233 also turns on the following optimization flags:
4234 @gccoptlist{-fthread-jumps @gol
4236 -foptimize-sibling-calls @gol
4237 -fcse-follow-jumps -fcse-skip-blocks @gol
4238 -fgcse -fgcse-lm @gol
4239 -fexpensive-optimizations @gol
4240 -fstrength-reduce @gol
4241 -frerun-cse-after-loop -frerun-loop-opt @gol
4245 -fschedule-insns -fschedule-insns2 @gol
4246 -fsched-interblock -fsched-spec @gol
4248 -fstrict-aliasing @gol
4249 -fdelete-null-pointer-checks @gol
4250 -freorder-blocks -freorder-functions @gol
4251 -funit-at-a-time @gol
4252 -falign-functions -falign-jumps @gol
4253 -falign-loops -falign-labels @gol
4256 Please note the warning under @option{-fgcse} about
4257 invoking @option{-O2} on programs that use computed gotos.
4261 Optimize yet more. @option{-O3} turns on all optimizations specified by
4262 @option{-O2} and also turns on the @option{-finline-functions},
4263 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4267 Do not optimize. This is the default.
4271 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4272 do not typically increase code size. It also performs further
4273 optimizations designed to reduce code size.
4275 @option{-Os} disables the following optimization flags:
4276 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4277 -falign-labels -freorder-blocks -freorder-blocks-and-partition -fprefetch-loop-arrays}
4279 If you use multiple @option{-O} options, with or without level numbers,
4280 the last such option is the one that is effective.
4283 Options of the form @option{-f@var{flag}} specify machine-independent
4284 flags. Most flags have both positive and negative forms; the negative
4285 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4286 below, only one of the forms is listed---the one you typically will
4287 use. You can figure out the other form by either removing @samp{no-}
4290 The following options control specific optimizations. They are either
4291 activated by @option{-O} options or are related to ones that are. You
4292 can use the following flags in the rare cases when ``fine-tuning'' of
4293 optimizations to be performed is desired.
4296 @item -fno-default-inline
4297 @opindex fno-default-inline
4298 Do not make member functions inline by default merely because they are
4299 defined inside the class scope (C++ only). Otherwise, when you specify
4300 @w{@option{-O}}, member functions defined inside class scope are compiled
4301 inline by default; i.e., you don't need to add @samp{inline} in front of
4302 the member function name.
4304 @item -fno-defer-pop
4305 @opindex fno-defer-pop
4306 Always pop the arguments to each function call as soon as that function
4307 returns. For machines which must pop arguments after a function call,
4308 the compiler normally lets arguments accumulate on the stack for several
4309 function calls and pops them all at once.
4311 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4315 Force memory operands to be copied into registers before doing
4316 arithmetic on them. This produces better code by making all memory
4317 references potential common subexpressions. When they are not common
4318 subexpressions, instruction combination should eliminate the separate
4321 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4324 @opindex fforce-addr
4325 Force memory address constants to be copied into registers before
4326 doing arithmetic on them. This may produce better code just as
4327 @option{-fforce-mem} may.
4329 @item -fomit-frame-pointer
4330 @opindex fomit-frame-pointer
4331 Don't keep the frame pointer in a register for functions that
4332 don't need one. This avoids the instructions to save, set up and
4333 restore frame pointers; it also makes an extra register available
4334 in many functions. @strong{It also makes debugging impossible on
4337 On some machines, such as the VAX, this flag has no effect, because
4338 the standard calling sequence automatically handles the frame pointer
4339 and nothing is saved by pretending it doesn't exist. The
4340 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4341 whether a target machine supports this flag. @xref{Registers,,Register
4342 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4344 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4346 @item -foptimize-sibling-calls
4347 @opindex foptimize-sibling-calls
4348 Optimize sibling and tail recursive calls.
4350 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4354 Don't pay attention to the @code{inline} keyword. Normally this option
4355 is used to keep the compiler from expanding any functions inline.
4356 Note that if you are not optimizing, no functions can be expanded inline.
4358 @item -finline-functions
4359 @opindex finline-functions
4360 Integrate all simple functions into their callers. The compiler
4361 heuristically decides which functions are simple enough to be worth
4362 integrating in this way.
4364 If all calls to a given function are integrated, and the function is
4365 declared @code{static}, then the function is normally not output as
4366 assembler code in its own right.
4368 Enabled at level @option{-O3}.
4370 @item -finline-limit=@var{n}
4371 @opindex finline-limit
4372 By default, GCC limits the size of functions that can be inlined. This flag
4373 allows the control of this limit for functions that are explicitly marked as
4374 inline (i.e., marked with the inline keyword or defined within the class
4375 definition in c++). @var{n} is the size of functions that can be inlined in
4376 number of pseudo instructions (not counting parameter handling). The default
4377 value of @var{n} is 600.
4378 Increasing this value can result in more inlined code at
4379 the cost of compilation time and memory consumption. Decreasing usually makes
4380 the compilation faster and less code will be inlined (which presumably
4381 means slower programs). This option is particularly useful for programs that
4382 use inlining heavily such as those based on recursive templates with C++.
4384 Inlining is actually controlled by a number of parameters, which may be
4385 specified individually by using @option{--param @var{name}=@var{value}}.
4386 The @option{-finline-limit=@var{n}} option sets some of these parameters
4390 @item max-inline-insns-single
4391 is set to @var{n}/2.
4392 @item max-inline-insns-auto
4393 is set to @var{n}/2.
4394 @item min-inline-insns
4395 is set to 130 or @var{n}/4, whichever is smaller.
4396 @item max-inline-insns-rtl
4400 See below for a documentation of the individual
4401 parameters controlling inlining.
4403 @emph{Note:} pseudo instruction represents, in this particular context, an
4404 abstract measurement of function's size. In no way, it represents a count
4405 of assembly instructions and as such its exact meaning might change from one
4406 release to an another.
4408 @item -fkeep-inline-functions
4409 @opindex fkeep-inline-functions
4410 In C, emit @code{static} functions that are declared @code{inline}
4411 into the object file, even if the function has been inlined into all
4412 of its callers. This switch does not affect functions using the
4413 @code{extern inline} extension in GNU C@. In C++, emit any and all
4414 inline functions into the object file.
4416 @item -fkeep-static-consts
4417 @opindex fkeep-static-consts
4418 Emit variables declared @code{static const} when optimization isn't turned
4419 on, even if the variables aren't referenced.
4421 GCC enables this option by default. If you want to force the compiler to
4422 check if the variable was referenced, regardless of whether or not
4423 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4425 @item -fmerge-constants
4426 Attempt to merge identical constants (string constants and floating point
4427 constants) across compilation units.
4429 This option is the default for optimized compilation if the assembler and
4430 linker support it. Use @option{-fno-merge-constants} to inhibit this
4433 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4435 @item -fmerge-all-constants
4436 Attempt to merge identical constants and identical variables.
4438 This option implies @option{-fmerge-constants}. In addition to
4439 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4440 arrays or initialized constant variables with integral or floating point
4441 types. Languages like C or C++ require each non-automatic variable to
4442 have distinct location, so using this option will result in non-conforming
4445 @item -fmodulo-sched
4446 @opindex fmodulo-sched
4447 Perform swing modulo scheduling immediately before the first scheduling
4448 pass. This pass looks at innermost loops and reorders their
4449 instructions by overlapping different iterations.
4451 @item -fno-branch-count-reg
4452 @opindex fno-branch-count-reg
4453 Do not use ``decrement and branch'' instructions on a count register,
4454 but instead generate a sequence of instructions that decrement a
4455 register, compare it against zero, then branch based upon the result.
4456 This option is only meaningful on architectures that support such
4457 instructions, which include x86, PowerPC, IA-64 and S/390.
4459 The default is @option{-fbranch-count-reg}, enabled when
4460 @option{-fstrength-reduce} is enabled.
4462 @item -fno-function-cse
4463 @opindex fno-function-cse
4464 Do not put function addresses in registers; make each instruction that
4465 calls a constant function contain the function's address explicitly.
4467 This option results in less efficient code, but some strange hacks
4468 that alter the assembler output may be confused by the optimizations
4469 performed when this option is not used.
4471 The default is @option{-ffunction-cse}
4473 @item -fno-zero-initialized-in-bss
4474 @opindex fno-zero-initialized-in-bss
4475 If the target supports a BSS section, GCC by default puts variables that
4476 are initialized to zero into BSS@. This can save space in the resulting
4479 This option turns off this behavior because some programs explicitly
4480 rely on variables going to the data section. E.g., so that the
4481 resulting executable can find the beginning of that section and/or make
4482 assumptions based on that.
4484 The default is @option{-fzero-initialized-in-bss}.
4486 @item -fbounds-check
4487 @opindex fbounds-check
4488 For front-ends that support it, generate additional code to check that
4489 indices used to access arrays are within the declared range. This is
4490 currently only supported by the Java and Fortran front-ends, where
4491 this option defaults to true and false respectively.
4493 @item -fmudflap -fmudflapth -fmudflapir
4497 @cindex bounds checking
4499 For front-ends that support it (C and C++), instrument all risky
4500 pointer/array dereferencing operations, some standard library
4501 string/heap functions, and some other associated constructs with
4502 range/validity tests. Modules so instrumented should be immune to
4503 buffer overflows, invalid heap use, and some other classes of C/C++
4504 programming errors. The instrumentation relies on a separate runtime
4505 library (@file{libmudflap}), which will be linked into a program if
4506 @option{-fmudflap} is given at link time. Run-time behavior of the
4507 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4508 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4511 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4512 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4513 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4514 instrumentation should ignore pointer reads. This produces less
4515 instrumentation (and therefore faster execution) and still provides
4516 some protection against outright memory corrupting writes, but allows
4517 erroneously read data to propagate within a program.
4519 @item -fstrength-reduce
4520 @opindex fstrength-reduce
4521 Perform the optimizations of loop strength reduction and
4522 elimination of iteration variables.
4524 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4526 @item -fthread-jumps
4527 @opindex fthread-jumps
4528 Perform optimizations where we check to see if a jump branches to a
4529 location where another comparison subsumed by the first is found. If
4530 so, the first branch is redirected to either the destination of the
4531 second branch or a point immediately following it, depending on whether
4532 the condition is known to be true or false.
4534 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4536 @item -fcse-follow-jumps
4537 @opindex fcse-follow-jumps
4538 In common subexpression elimination, scan through jump instructions
4539 when the target of the jump is not reached by any other path. For
4540 example, when CSE encounters an @code{if} statement with an
4541 @code{else} clause, CSE will follow the jump when the condition
4544 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4546 @item -fcse-skip-blocks
4547 @opindex fcse-skip-blocks
4548 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4549 follow jumps which conditionally skip over blocks. When CSE
4550 encounters a simple @code{if} statement with no else clause,
4551 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4552 body of the @code{if}.
4554 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4556 @item -frerun-cse-after-loop
4557 @opindex frerun-cse-after-loop
4558 Re-run common subexpression elimination after loop optimizations has been
4561 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4563 @item -frerun-loop-opt
4564 @opindex frerun-loop-opt
4565 Run the loop optimizer twice.
4567 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4571 Perform a global common subexpression elimination pass.
4572 This pass also performs global constant and copy propagation.
4574 @emph{Note:} When compiling a program using computed gotos, a GCC
4575 extension, you may get better runtime performance if you disable
4576 the global common subexpression elimination pass by adding
4577 @option{-fno-gcse} to the command line.
4579 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4583 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4584 attempt to move loads which are only killed by stores into themselves. This
4585 allows a loop containing a load/store sequence to be changed to a load outside
4586 the loop, and a copy/store within the loop.
4588 Enabled by default when gcse is enabled.
4592 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4593 global common subexpression elimination. This pass will attempt to move
4594 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4595 loops containing a load/store sequence can be changed to a load before
4596 the loop and a store after the loop.
4598 Not enabled at any optimization level.
4602 When @option{-fgcse-las} is enabled, the global common subexpression
4603 elimination pass eliminates redundant loads that come after stores to the
4604 same memory location (both partial and full redundancies).
4606 Not enabled at any optimization level.
4608 @item -fgcse-after-reload
4609 @opindex fgcse-after-reload
4610 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4611 pass is performed after reload. The purpose of this pass is to cleanup
4614 @item -floop-optimize
4615 @opindex floop-optimize
4616 Perform loop optimizations: move constant expressions out of loops, simplify
4617 exit test conditions and optionally do strength-reduction as well.
4619 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4621 @item -floop-optimize2
4622 @opindex floop-optimize2
4623 Perform loop optimizations using the new loop optimizer. The optimizations
4624 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4627 @item -fcrossjumping
4628 @opindex crossjumping
4629 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4630 resulting code may or may not perform better than without cross-jumping.
4632 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4634 @item -fif-conversion
4635 @opindex if-conversion
4636 Attempt to transform conditional jumps into branch-less equivalents. This
4637 include use of conditional moves, min, max, set flags and abs instructions, and
4638 some tricks doable by standard arithmetics. The use of conditional execution
4639 on chips where it is available is controlled by @code{if-conversion2}.
4641 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4643 @item -fif-conversion2
4644 @opindex if-conversion2
4645 Use conditional execution (where available) to transform conditional jumps into
4646 branch-less equivalents.
4648 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4650 @item -fdelete-null-pointer-checks
4651 @opindex fdelete-null-pointer-checks
4652 Use global dataflow analysis to identify and eliminate useless checks
4653 for null pointers. The compiler assumes that dereferencing a null
4654 pointer would have halted the program. If a pointer is checked after
4655 it has already been dereferenced, it cannot be null.
4657 In some environments, this assumption is not true, and programs can
4658 safely dereference null pointers. Use
4659 @option{-fno-delete-null-pointer-checks} to disable this optimization
4660 for programs which depend on that behavior.
4662 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4664 @item -fexpensive-optimizations
4665 @opindex fexpensive-optimizations
4666 Perform a number of minor optimizations that are relatively expensive.
4668 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4670 @item -foptimize-register-move
4672 @opindex foptimize-register-move
4674 Attempt to reassign register numbers in move instructions and as
4675 operands of other simple instructions in order to maximize the amount of
4676 register tying. This is especially helpful on machines with two-operand
4679 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4682 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4684 @item -fdelayed-branch
4685 @opindex fdelayed-branch
4686 If supported for the target machine, attempt to reorder instructions
4687 to exploit instruction slots available after delayed branch
4690 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4692 @item -fschedule-insns
4693 @opindex fschedule-insns
4694 If supported for the target machine, attempt to reorder instructions to
4695 eliminate execution stalls due to required data being unavailable. This
4696 helps machines that have slow floating point or memory load instructions
4697 by allowing other instructions to be issued until the result of the load
4698 or floating point instruction is required.
4700 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4702 @item -fschedule-insns2
4703 @opindex fschedule-insns2
4704 Similar to @option{-fschedule-insns}, but requests an additional pass of
4705 instruction scheduling after register allocation has been done. This is
4706 especially useful on machines with a relatively small number of
4707 registers and where memory load instructions take more than one cycle.
4709 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4711 @item -fno-sched-interblock
4712 @opindex fno-sched-interblock
4713 Don't schedule instructions across basic blocks. This is normally
4714 enabled by default when scheduling before register allocation, i.e.@:
4715 with @option{-fschedule-insns} or at @option{-O2} or higher.
4717 @item -fno-sched-spec
4718 @opindex fno-sched-spec
4719 Don't allow speculative motion of non-load instructions. This is normally
4720 enabled by default when scheduling before register allocation, i.e.@:
4721 with @option{-fschedule-insns} or at @option{-O2} or higher.
4723 @item -fsched-spec-load
4724 @opindex fsched-spec-load
4725 Allow speculative motion of some load instructions. This only makes
4726 sense when scheduling before register allocation, i.e.@: with
4727 @option{-fschedule-insns} or at @option{-O2} or higher.
4729 @item -fsched-spec-load-dangerous
4730 @opindex fsched-spec-load-dangerous
4731 Allow speculative motion of more load instructions. This only makes
4732 sense when scheduling before register allocation, i.e.@: with
4733 @option{-fschedule-insns} or at @option{-O2} or higher.
4735 @item -fsched-stalled-insns=@var{n}
4736 @opindex fsched-stalled-insns
4737 Define how many insns (if any) can be moved prematurely from the queue
4738 of stalled insns into the ready list, during the second scheduling pass.
4740 @item -fsched-stalled-insns-dep=@var{n}
4741 @opindex fsched-stalled-insns-dep
4742 Define how many insn groups (cycles) will be examined for a dependency
4743 on a stalled insn that is candidate for premature removal from the queue
4744 of stalled insns. Has an effect only during the second scheduling pass,
4745 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4747 @item -fsched2-use-superblocks
4748 @opindex fsched2-use-superblocks
4749 When scheduling after register allocation, do use superblock scheduling
4750 algorithm. Superblock scheduling allows motion across basic block boundaries
4751 resulting on faster schedules. This option is experimental, as not all machine
4752 descriptions used by GCC model the CPU closely enough to avoid unreliable
4753 results from the algorithm.
4755 This only makes sense when scheduling after register allocation, i.e.@: with
4756 @option{-fschedule-insns2} or at @option{-O2} or higher.
4758 @item -fsched2-use-traces
4759 @opindex fsched2-use-traces
4760 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4761 allocation and additionally perform code duplication in order to increase the
4762 size of superblocks using tracer pass. See @option{-ftracer} for details on
4765 This mode should produce faster but significantly longer programs. Also
4766 without @option{-fbranch-probabilities} the traces constructed may not
4767 match the reality and hurt the performance. This only makes
4768 sense when scheduling after register allocation, i.e.@: with
4769 @option{-fschedule-insns2} or at @option{-O2} or higher.
4771 @item -freschedule-modulo-scheduled-loops
4772 @opindex fscheduling-in-modulo-scheduled-loops
4773 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4774 we may want to prevent the later scheduling passes from changing its schedule, we use this
4775 option to control that.
4777 @item -fcaller-saves
4778 @opindex fcaller-saves
4779 Enable values to be allocated in registers that will be clobbered by
4780 function calls, by emitting extra instructions to save and restore the
4781 registers around such calls. Such allocation is done only when it
4782 seems to result in better code than would otherwise be produced.
4784 This option is always enabled by default on certain machines, usually
4785 those which have no call-preserved registers to use instead.
4787 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4790 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4791 enabled by default at @option{-O2} and @option{-O3}.
4794 Perform Full Redundancy Elimination (FRE) on trees. The difference
4795 between FRE and PRE is that FRE only considers expressions
4796 that are computed on all paths leading to the redundant computation.
4797 This analysis faster than PRE, though it exposes fewer redundancies.
4798 This flag is enabled by default at @option{-O} and higher.
4800 @item -ftree-copy-prop
4801 Perform copy propagation on trees. This pass eliminates unnecessary
4802 copy operations. This flag is enabled by default at @option{-O} and
4805 @item -ftree-store-copy-prop
4806 Perform copy propagation of memory loads and stores. This pass
4807 eliminates unnecessary copy operations in memory references
4808 (structures, global variables, arrays, etc). This flag is enabled by
4809 default at @option{-O2} and higher.
4812 Perform structural alias analysis on trees. This flag
4813 is enabled by default at @option{-O} and higher.
4816 Perform forward store motion on trees. This flag is
4817 enabled by default at @option{-O} and higher.
4820 Perform sparse conditional constant propagation (CCP) on trees. This
4821 pass only operates on local scalar variables and is enabled by default
4822 at @option{-O} and higher.
4824 @item -ftree-store-ccp
4825 Perform sparse conditional constant propagation (CCP) on trees. This
4826 pass operates on both local scalar variables and memory stores and
4827 loads (global variables, structures, arrays, etc). This flag is
4828 enabled by default at @option{-O2} and higher.
4831 Perform dead code elimination (DCE) on trees. This flag is enabled by
4832 default at @option{-O} and higher.
4834 @item -ftree-dominator-opts
4835 Perform dead code elimination (DCE) on trees. This flag is enabled by
4836 default at @option{-O} and higher.
4839 Perform loop header copying on trees. This is beneficial since it increases
4840 effectiveness of code motion optimizations. It also saves one jump. This flag
4841 is enabled by default at @option{-O} and higher. It is not enabled
4842 for @option{-Os}, since it usually increases code size.
4844 @item -ftree-loop-optimize
4845 Perform loop optimizations on trees. This flag is enabled by default
4846 at @option{-O} and higher.
4848 @item -ftree-loop-linear
4849 Perform linear loop transformations on tree. This flag can improve cache
4850 performance and allow further loop optimizations to take place.
4852 @item -ftree-loop-im
4853 Perform loop invariant motion on trees. This pass moves only invariants that
4854 would be hard to handle at RTL level (function calls, operations that expand to
4855 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
4856 operands of conditions that are invariant out of the loop, so that we can use
4857 just trivial invariantness analysis in loop unswitching. The pass also includes
4860 @item -ftree-loop-ivcanon
4861 Create a canonical counter for number of iterations in the loop for that
4862 determining number of iterations requires complicated analysis. Later
4863 optimizations then may determine the number easily. Useful especially
4864 in connection with unrolling.
4867 Perform induction variable optimizations (strength reduction, induction
4868 variable merging and induction variable elimination) on trees.
4871 Perform scalar replacement of aggregates. This pass replaces structure
4872 references with scalars to prevent committing structures to memory too
4873 early. This flag is enabled by default at @option{-O} and higher.
4875 @item -ftree-copyrename
4876 Perform copy renaming on trees. This pass attempts to rename compiler
4877 temporaries to other variables at copy locations, usually resulting in
4878 variable names which more closely resemble the original variables. This flag
4879 is enabled by default at @option{-O} and higher.
4882 Perform temporary expression replacement during the SSA->normal phase. Single
4883 use/single def temporaries are replaced at their use location with their
4884 defining expression. This results in non-GIMPLE code, but gives the expanders
4885 much more complex trees to work on resulting in better RTL generation. This is
4886 enabled by default at @option{-O} and higher.
4889 Perform live range splitting during the SSA->normal phase. Distinct live
4890 ranges of a variable are split into unique variables, allowing for better
4891 optimization later. This is enabled by default at @option{-O} and higher.
4893 @item -ftree-vectorize
4894 Perform loop vectorization on trees.
4898 Perform tail duplication to enlarge superblock size. This transformation
4899 simplifies the control flow of the function allowing other optimizations to do
4902 @item -funroll-loops
4903 @opindex funroll-loops
4904 Unroll loops whose number of iterations can be determined at compile
4905 time or upon entry to the loop. @option{-funroll-loops} implies both
4906 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4907 option makes code larger, and may or may not make it run faster.
4909 @item -funroll-all-loops
4910 @opindex funroll-all-loops
4911 Unroll all loops, even if their number of iterations is uncertain when
4912 the loop is entered. This usually makes programs run more slowly.
4913 @option{-funroll-all-loops} implies the same options as
4914 @option{-funroll-loops},
4916 @item -fsplit-ivs-in-unroller
4917 @opindex -fsplit-ivs-in-unroller
4918 Enables expressing of values of induction variables in later iterations
4919 of the unrolled loop using the value in the first iteration. This breaks
4920 long dependency chains, thus improving efficiency of the scheduling passes
4921 (for best results, @option{-fweb} should be used as well).
4923 Combination of @option{-fweb} and CSE is often sufficient to obtain the
4924 same effect. However in cases the loop body is more complicated than
4925 a single basic block, this is not reliable. It also does not work at all
4926 on some of the architectures due to restrictions in the CSE pass.
4928 This optimization is enabled by default.
4930 @item -fvariable-expansion-in-unroller
4931 @opindex -fvariable-expansion-in-unroller
4932 With this option, the compiler will create multiple copies of some
4933 local variables when unrolling a loop which can result in superior code.
4935 @item -fprefetch-loop-arrays
4936 @opindex fprefetch-loop-arrays
4937 If supported by the target machine, generate instructions to prefetch
4938 memory to improve the performance of loops that access large arrays.
4940 These options may generate better or worse code; results are highly
4941 dependent on the structure of loops within the source code.
4944 @itemx -fno-peephole2
4945 @opindex fno-peephole
4946 @opindex fno-peephole2
4947 Disable any machine-specific peephole optimizations. The difference
4948 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
4949 are implemented in the compiler; some targets use one, some use the
4950 other, a few use both.
4952 @option{-fpeephole} is enabled by default.
4953 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4955 @item -fno-guess-branch-probability
4956 @opindex fno-guess-branch-probability
4957 Do not guess branch probabilities using heuristics.
4959 GCC will use heuristics to guess branch probabilities if they are
4960 not provided by profiling feedback (@option{-fprofile-arcs}). These
4961 heuristics are based on the control flow graph. If some branch probabilities
4962 are specified by @samp{__builtin_expect}, then the heuristics will be
4963 used to guess branch probabilities for the rest of the control flow graph,
4964 taking the @samp{__builtin_expect} info into account. The interactions
4965 between the heuristics and @samp{__builtin_expect} can be complex, and in
4966 some cases, it may be useful to disable the heuristics so that the effects
4967 of @samp{__builtin_expect} are easier to understand.
4969 The default is @option{-fguess-branch-probability} at levels
4970 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4972 @item -freorder-blocks
4973 @opindex freorder-blocks
4974 Reorder basic blocks in the compiled function in order to reduce number of
4975 taken branches and improve code locality.
4977 Enabled at levels @option{-O2}, @option{-O3}.
4979 @item -freorder-blocks-and-partition
4980 @opindex freorder-blocks-and-partition
4981 In addition to reordering basic blocks in the compiled function, in order
4982 to reduce number of taken branches, partitions hot and cold basic blocks
4983 into separate sections of the assembly and .o files, to improve
4984 paging and cache locality performance.
4986 This optimization is automatically turned off in the presence of
4987 exception handling, for linkonce sections, for functions with a user-defined
4988 section attribute and on any architecture that does not support named
4991 @item -freorder-functions
4992 @opindex freorder-functions
4993 Reorder functions in the object file in order to
4994 improve code locality. This is implemented by using special
4995 subsections @code{.text.hot} for most frequently executed functions and
4996 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
4997 the linker so object file format must support named sections and linker must
4998 place them in a reasonable way.
5000 Also profile feedback must be available in to make this option effective. See
5001 @option{-fprofile-arcs} for details.
5003 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5005 @item -fstrict-aliasing
5006 @opindex fstrict-aliasing
5007 Allows the compiler to assume the strictest aliasing rules applicable to
5008 the language being compiled. For C (and C++), this activates
5009 optimizations based on the type of expressions. In particular, an
5010 object of one type is assumed never to reside at the same address as an
5011 object of a different type, unless the types are almost the same. For
5012 example, an @code{unsigned int} can alias an @code{int}, but not a
5013 @code{void*} or a @code{double}. A character type may alias any other
5016 Pay special attention to code like this:
5029 The practice of reading from a different union member than the one most
5030 recently written to (called ``type-punning'') is common. Even with
5031 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5032 is accessed through the union type. So, the code above will work as
5033 expected. However, this code might not:
5044 Every language that wishes to perform language-specific alias analysis
5045 should define a function that computes, given an @code{tree}
5046 node, an alias set for the node. Nodes in different alias sets are not
5047 allowed to alias. For an example, see the C front-end function
5048 @code{c_get_alias_set}.
5050 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5052 @item -falign-functions
5053 @itemx -falign-functions=@var{n}
5054 @opindex falign-functions
5055 Align the start of functions to the next power-of-two greater than
5056 @var{n}, skipping up to @var{n} bytes. For instance,
5057 @option{-falign-functions=32} aligns functions to the next 32-byte
5058 boundary, but @option{-falign-functions=24} would align to the next
5059 32-byte boundary only if this can be done by skipping 23 bytes or less.
5061 @option{-fno-align-functions} and @option{-falign-functions=1} are
5062 equivalent and mean that functions will not be aligned.
5064 Some assemblers only support this flag when @var{n} is a power of two;
5065 in that case, it is rounded up.
5067 If @var{n} is not specified or is zero, use a machine-dependent default.
5069 Enabled at levels @option{-O2}, @option{-O3}.
5071 @item -falign-labels
5072 @itemx -falign-labels=@var{n}
5073 @opindex falign-labels
5074 Align all branch targets to a power-of-two boundary, skipping up to
5075 @var{n} bytes like @option{-falign-functions}. This option can easily
5076 make code slower, because it must insert dummy operations for when the
5077 branch target is reached in the usual flow of the code.
5079 @option{-fno-align-labels} and @option{-falign-labels=1} are
5080 equivalent and mean that labels will not be aligned.
5082 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5083 are greater than this value, then their values are used instead.
5085 If @var{n} is not specified or is zero, use a machine-dependent default
5086 which is very likely to be @samp{1}, meaning no alignment.
5088 Enabled at levels @option{-O2}, @option{-O3}.
5091 @itemx -falign-loops=@var{n}
5092 @opindex falign-loops
5093 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5094 like @option{-falign-functions}. The hope is that the loop will be
5095 executed many times, which will make up for any execution of the dummy
5098 @option{-fno-align-loops} and @option{-falign-loops=1} are
5099 equivalent and mean that loops will not be aligned.
5101 If @var{n} is not specified or is zero, use a machine-dependent default.
5103 Enabled at levels @option{-O2}, @option{-O3}.
5106 @itemx -falign-jumps=@var{n}
5107 @opindex falign-jumps
5108 Align branch targets to a power-of-two boundary, for branch targets
5109 where the targets can only be reached by jumping, skipping up to @var{n}
5110 bytes like @option{-falign-functions}. In this case, no dummy operations
5113 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5114 equivalent and mean that loops will not be aligned.
5116 If @var{n} is not specified or is zero, use a machine-dependent default.
5118 Enabled at levels @option{-O2}, @option{-O3}.
5120 @item -funit-at-a-time
5121 @opindex funit-at-a-time
5122 Parse the whole compilation unit before starting to produce code.
5123 This allows some extra optimizations to take place but consumes
5124 more memory (in general). There are some compatibility issues
5125 with @emph{unit-at-at-time} mode:
5128 enabling @emph{unit-at-a-time} mode may change the order
5129 in which functions, variables, and top-level @code{asm} statements
5130 are emitted, and will likely break code relying on some particular
5131 ordering. The majority of such top-level @code{asm} statements,
5132 though, can be replaced by @code{section} attributes.
5135 @emph{unit-at-a-time} mode removes unreferenced static variables
5136 and functions are removed. This may result in undefined references
5137 when an @code{asm} statement refers directly to variables or functions
5138 that are otherwise unused. In that case either the variable/function
5139 shall be listed as an operand of the @code{asm} statement operand or,
5140 in the case of top-level @code{asm} statements the attribute @code{used}
5141 shall be used on the declaration.
5144 Static functions now can use non-standard passing conventions that
5145 may break @code{asm} statements calling functions directly. Again,
5146 attribute @code{used} will prevent this behavior.
5149 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5150 but this scheme may not be supported by future releases of GCC@.
5152 Enabled at levels @option{-O2}, @option{-O3}.
5156 Constructs webs as commonly used for register allocation purposes and assign
5157 each web individual pseudo register. This allows the register allocation pass
5158 to operate on pseudos directly, but also strengthens several other optimization
5159 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5160 however, make debugging impossible, since variables will no longer stay in a
5163 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os},
5164 on targets where the default format for debugging information supports
5167 @item -fno-cprop-registers
5168 @opindex fno-cprop-registers
5169 After register allocation and post-register allocation instruction splitting,
5170 we perform a copy-propagation pass to try to reduce scheduling dependencies
5171 and occasionally eliminate the copy.
5173 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5175 @item -fprofile-generate
5176 @opindex fprofile-generate
5178 Enable options usually used for instrumenting application to produce
5179 profile useful for later recompilation with profile feedback based
5180 optimization. You must use @option{-fprofile-generate} both when
5181 compiling and when linking your program.
5183 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5186 @opindex fprofile-use
5187 Enable profile feedback directed optimizations, and optimizations
5188 generally profitable only with profile feedback available.
5190 The following options are enabled: @code{-fbranch-probabilities},
5191 @code{-fvpt}, @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}.
5195 The following options control compiler behavior regarding floating
5196 point arithmetic. These options trade off between speed and
5197 correctness. All must be specifically enabled.
5201 @opindex ffloat-store
5202 Do not store floating point variables in registers, and inhibit other
5203 options that might change whether a floating point value is taken from a
5206 @cindex floating point precision
5207 This option prevents undesirable excess precision on machines such as
5208 the 68000 where the floating registers (of the 68881) keep more
5209 precision than a @code{double} is supposed to have. Similarly for the
5210 x86 architecture. For most programs, the excess precision does only
5211 good, but a few programs rely on the precise definition of IEEE floating
5212 point. Use @option{-ffloat-store} for such programs, after modifying
5213 them to store all pertinent intermediate computations into variables.
5217 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5218 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5219 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5220 and @option{fcx-limited-range}.
5222 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5224 This option should never be turned on by any @option{-O} option since
5225 it can result in incorrect output for programs which depend on
5226 an exact implementation of IEEE or ISO rules/specifications for
5229 @item -fno-math-errno
5230 @opindex fno-math-errno
5231 Do not set ERRNO after calling math functions that are executed
5232 with a single instruction, e.g., sqrt. A program that relies on
5233 IEEE exceptions for math error handling may want to use this flag
5234 for speed while maintaining IEEE arithmetic compatibility.
5236 This option should never be turned on by any @option{-O} option since
5237 it can result in incorrect output for programs which depend on
5238 an exact implementation of IEEE or ISO rules/specifications for
5241 The default is @option{-fmath-errno}.
5243 @item -funsafe-math-optimizations
5244 @opindex funsafe-math-optimizations
5245 Allow optimizations for floating-point arithmetic that (a) assume
5246 that arguments and results are valid and (b) may violate IEEE or
5247 ANSI standards. When used at link-time, it may include libraries
5248 or startup files that change the default FPU control word or other
5249 similar optimizations.
5251 This option should never be turned on by any @option{-O} option since
5252 it can result in incorrect output for programs which depend on
5253 an exact implementation of IEEE or ISO rules/specifications for
5256 The default is @option{-fno-unsafe-math-optimizations}.
5258 @item -ffinite-math-only
5259 @opindex ffinite-math-only
5260 Allow optimizations for floating-point arithmetic that assume
5261 that arguments and results are not NaNs or +-Infs.
5263 This option should never be turned on by any @option{-O} option since
5264 it can result in incorrect output for programs which depend on
5265 an exact implementation of IEEE or ISO rules/specifications.
5267 The default is @option{-fno-finite-math-only}.
5269 @item -fno-trapping-math
5270 @opindex fno-trapping-math
5271 Compile code assuming that floating-point operations cannot generate
5272 user-visible traps. These traps include division by zero, overflow,
5273 underflow, inexact result and invalid operation. This option implies
5274 @option{-fno-signaling-nans}. Setting this option may allow faster
5275 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5277 This option should never be turned on by any @option{-O} option since
5278 it can result in incorrect output for programs which depend on
5279 an exact implementation of IEEE or ISO rules/specifications for
5282 The default is @option{-ftrapping-math}.
5284 @item -frounding-math
5285 @opindex frounding-math
5286 Disable transformations and optimizations that assume default floating
5287 point rounding behavior. This is round-to-zero for all floating point
5288 to integer conversions, and round-to-nearest for all other arithmetic
5289 truncations. This option should be specified for programs that change
5290 the FP rounding mode dynamically, or that may be executed with a
5291 non-default rounding mode. This option disables constant folding of
5292 floating point expressions at compile-time (which may be affected by
5293 rounding mode) and arithmetic transformations that are unsafe in the
5294 presence of sign-dependent rounding modes.
5296 The default is @option{-fno-rounding-math}.
5298 This option is experimental and does not currently guarantee to
5299 disable all GCC optimizations that are affected by rounding mode.
5300 Future versions of GCC may provide finer control of this setting
5301 using C99's @code{FENV_ACCESS} pragma. This command line option
5302 will be used to specify the default state for @code{FENV_ACCESS}.
5304 @item -fsignaling-nans
5305 @opindex fsignaling-nans
5306 Compile code assuming that IEEE signaling NaNs may generate user-visible
5307 traps during floating-point operations. Setting this option disables
5308 optimizations that may change the number of exceptions visible with
5309 signaling NaNs. This option implies @option{-ftrapping-math}.
5311 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5314 The default is @option{-fno-signaling-nans}.
5316 This option is experimental and does not currently guarantee to
5317 disable all GCC optimizations that affect signaling NaN behavior.
5319 @item -fsingle-precision-constant
5320 @opindex fsingle-precision-constant
5321 Treat floating point constant as single precision constant instead of
5322 implicitly converting it to double precision constant.
5324 @item -fcx-limited-range
5325 @itemx -fno-cx-limited-range
5326 @opindex fcx-limited-range
5327 @opindex fno-cx-limited-range
5328 When enabled, this option states that a range reduction step is not
5329 needed when performing complex division. The default is
5330 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5332 This option controls the default setting of the ISO C99
5333 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5338 The following options control optimizations that may improve
5339 performance, but are not enabled by any @option{-O} options. This
5340 section includes experimental options that may produce broken code.
5343 @item -fbranch-probabilities
5344 @opindex fbranch-probabilities
5345 After running a program compiled with @option{-fprofile-arcs}
5346 (@pxref{Debugging Options,, Options for Debugging Your Program or
5347 @command{gcc}}), you can compile it a second time using
5348 @option{-fbranch-probabilities}, to improve optimizations based on
5349 the number of times each branch was taken. When the program
5350 compiled with @option{-fprofile-arcs} exits it saves arc execution
5351 counts to a file called @file{@var{sourcename}.gcda} for each source
5352 file The information in this data file is very dependent on the
5353 structure of the generated code, so you must use the same source code
5354 and the same optimization options for both compilations.
5356 With @option{-fbranch-probabilities}, GCC puts a
5357 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5358 These can be used to improve optimization. Currently, they are only
5359 used in one place: in @file{reorg.c}, instead of guessing which path a
5360 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5361 exactly determine which path is taken more often.
5363 @item -fprofile-values
5364 @opindex fprofile-values
5365 If combined with @option{-fprofile-arcs}, it adds code so that some
5366 data about values of expressions in the program is gathered.
5368 With @option{-fbranch-probabilities}, it reads back the data gathered
5369 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5370 notes to instructions for their later usage in optimizations.
5372 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5376 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5377 a code to gather information about values of expressions.
5379 With @option{-fbranch-probabilities}, it reads back the data gathered
5380 and actually performs the optimizations based on them.
5381 Currently the optimizations include specialization of division operation
5382 using the knowledge about the value of the denominator.
5384 @item -fspeculative-prefetching
5385 @opindex fspeculative-prefetching
5386 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5387 a code to gather information about addresses of memory references in the
5390 With @option{-fbranch-probabilities}, it reads back the data gathered
5391 and issues prefetch instructions according to them. In addition to the opportunities
5392 noticed by @option{-fprefetch-loop-arrays}, it also notices more complicated
5393 memory access patterns---for example accesses to the data stored in linked
5394 list whose elements are usually allocated sequentially.
5396 In order to prevent issuing double prefetches, usage of
5397 @option{-fspeculative-prefetching} implies @option{-fno-prefetch-loop-arrays}.
5399 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5401 @item -frename-registers
5402 @opindex frename-registers
5403 Attempt to avoid false dependencies in scheduled code by making use
5404 of registers left over after register allocation. This optimization
5405 will most benefit processors with lots of registers. Depending on the
5406 debug information format adopted by the target, however, it can
5407 make debugging impossible, since variables will no longer stay in
5408 a ``home register''.
5410 Not enabled by default at any level because it has known bugs.
5414 Perform tail duplication to enlarge superblock size. This transformation
5415 simplifies the control flow of the function allowing other optimizations to do
5418 Enabled with @option{-fprofile-use}.
5420 @item -funroll-loops
5421 @opindex funroll-loops
5422 Unroll loops whose number of iterations can be determined at compile time or
5423 upon entry to the loop. @option{-funroll-loops} implies
5424 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
5425 (i.e.@: complete removal of loops with small constant number of iterations).
5426 This option makes code larger, and may or may not make it run faster.
5428 Enabled with @option{-fprofile-use}.
5430 @item -funroll-all-loops
5431 @opindex funroll-all-loops
5432 Unroll all loops, even if their number of iterations is uncertain when
5433 the loop is entered. This usually makes programs run more slowly.
5434 @option{-funroll-all-loops} implies the same options as
5435 @option{-funroll-loops}.
5438 @opindex fpeel-loops
5439 Peels the loops for that there is enough information that they do not
5440 roll much (from profile feedback). It also turns on complete loop peeling
5441 (i.e.@: complete removal of loops with small constant number of iterations).
5443 Enabled with @option{-fprofile-use}.
5445 @item -fmove-loop-invariants
5446 @opindex fmove-loop-invariants
5447 Enables the loop invariant motion pass in the new loop optimizer. Enabled
5448 at level @option{-O1}
5450 @item -funswitch-loops
5451 @opindex funswitch-loops
5452 Move branches with loop invariant conditions out of the loop, with duplicates
5453 of the loop on both branches (modified according to result of the condition).
5455 @item -fprefetch-loop-arrays
5456 @opindex fprefetch-loop-arrays
5457 If supported by the target machine, generate instructions to prefetch
5458 memory to improve the performance of loops that access large arrays.
5460 Disabled at level @option{-Os}.
5462 @item -ffunction-sections
5463 @itemx -fdata-sections
5464 @opindex ffunction-sections
5465 @opindex fdata-sections
5466 Place each function or data item into its own section in the output
5467 file if the target supports arbitrary sections. The name of the
5468 function or the name of the data item determines the section's name
5471 Use these options on systems where the linker can perform optimizations
5472 to improve locality of reference in the instruction space. Most systems
5473 using the ELF object format and SPARC processors running Solaris 2 have
5474 linkers with such optimizations. AIX may have these optimizations in
5477 Only use these options when there are significant benefits from doing
5478 so. When you specify these options, the assembler and linker will
5479 create larger object and executable files and will also be slower.
5480 You will not be able to use @code{gprof} on all systems if you
5481 specify this option and you may have problems with debugging if
5482 you specify both this option and @option{-g}.
5484 @item -fbranch-target-load-optimize
5485 @opindex fbranch-target-load-optimize
5486 Perform branch target register load optimization before prologue / epilogue
5488 The use of target registers can typically be exposed only during reload,
5489 thus hoisting loads out of loops and doing inter-block scheduling needs
5490 a separate optimization pass.
5492 @item -fbranch-target-load-optimize2
5493 @opindex fbranch-target-load-optimize2
5494 Perform branch target register load optimization after prologue / epilogue
5497 @item -fbtr-bb-exclusive
5498 @opindex fbtr-bb-exclusive
5499 When performing branch target register load optimization, don't reuse
5500 branch target registers in within any basic block.
5502 @item --param @var{name}=@var{value}
5504 In some places, GCC uses various constants to control the amount of
5505 optimization that is done. For example, GCC will not inline functions
5506 that contain more that a certain number of instructions. You can
5507 control some of these constants on the command-line using the
5508 @option{--param} option.
5510 The names of specific parameters, and the meaning of the values, are
5511 tied to the internals of the compiler, and are subject to change
5512 without notice in future releases.
5514 In each case, the @var{value} is an integer. The allowable choices for
5515 @var{name} are given in the following table:
5518 @item salias-max-implicit-fields
5519 The maximum number of fields in a variable without direct
5520 structure accesses for which structure aliasing will consider trying
5521 to track each field. The default is 5
5523 @item sra-max-structure-size
5524 The maximum structure size, in bytes, at which the scalar replacement
5525 of aggregates (SRA) optimization will perform block copies. The
5526 default value, 0, implies that GCC will select the most appropriate
5529 @item sra-field-structure-ratio
5530 The threshold ratio (as a percentage) between instantiated fields and
5531 the complete structure size. We say that if the ratio of the number
5532 of bytes in instantiated fields to the number of bytes in the complete
5533 structure exceeds this parameter, then block copies are not used. The
5536 @item max-crossjump-edges
5537 The maximum number of incoming edges to consider for crossjumping.
5538 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5539 the number of edges incoming to each block. Increasing values mean
5540 more aggressive optimization, making the compile time increase with
5541 probably small improvement in executable size.
5543 @item min-crossjump-insns
5544 The minimum number of instructions which must be matched at the end
5545 of two blocks before crossjumping will be performed on them. This
5546 value is ignored in the case where all instructions in the block being
5547 crossjumped from are matched. The default value is 5.
5549 @item max-goto-duplication-insns
5550 The maximum number of instructions to duplicate to a block that jumps
5551 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5552 passes, GCC factors computed gotos early in the compilation process,
5553 and unfactors them as late as possible. Only computed jumps at the
5554 end of a basic blocks with no more than max-goto-duplication-insns are
5555 unfactored. The default value is 8.
5557 @item max-delay-slot-insn-search
5558 The maximum number of instructions to consider when looking for an
5559 instruction to fill a delay slot. If more than this arbitrary number of
5560 instructions is searched, the time savings from filling the delay slot
5561 will be minimal so stop searching. Increasing values mean more
5562 aggressive optimization, making the compile time increase with probably
5563 small improvement in executable run time.
5565 @item max-delay-slot-live-search
5566 When trying to fill delay slots, the maximum number of instructions to
5567 consider when searching for a block with valid live register
5568 information. Increasing this arbitrarily chosen value means more
5569 aggressive optimization, increasing the compile time. This parameter
5570 should be removed when the delay slot code is rewritten to maintain the
5573 @item max-gcse-memory
5574 The approximate maximum amount of memory that will be allocated in
5575 order to perform the global common subexpression elimination
5576 optimization. If more memory than specified is required, the
5577 optimization will not be done.
5579 @item max-gcse-passes
5580 The maximum number of passes of GCSE to run. The default is 1.
5582 @item max-pending-list-length
5583 The maximum number of pending dependencies scheduling will allow
5584 before flushing the current state and starting over. Large functions
5585 with few branches or calls can create excessively large lists which
5586 needlessly consume memory and resources.
5588 @item max-inline-insns-single
5589 Several parameters control the tree inliner used in gcc.
5590 This number sets the maximum number of instructions (counted in GCC's
5591 internal representation) in a single function that the tree inliner
5592 will consider for inlining. This only affects functions declared
5593 inline and methods implemented in a class declaration (C++).
5594 The default value is 450.
5596 @item max-inline-insns-auto
5597 When you use @option{-finline-functions} (included in @option{-O3}),
5598 a lot of functions that would otherwise not be considered for inlining
5599 by the compiler will be investigated. To those functions, a different
5600 (more restrictive) limit compared to functions declared inline can
5602 The default value is 90.
5604 @item large-function-insns
5605 The limit specifying really large functions. For functions larger than this
5606 limit after inlining inlining is constrained by
5607 @option{--param large-function-growth}. This parameter is useful primarily
5608 to avoid extreme compilation time caused by non-linear algorithms used by the
5610 This parameter is ignored when @option{-funit-at-a-time} is not used.
5611 The default value is 2700.
5613 @item large-function-growth
5614 Specifies maximal growth of large function caused by inlining in percents.
5615 This parameter is ignored when @option{-funit-at-a-time} is not used.
5616 The default value is 100 which limits large function growth to 2.0 times
5619 @item inline-unit-growth
5620 Specifies maximal overall growth of the compilation unit caused by inlining.
5621 This parameter is ignored when @option{-funit-at-a-time} is not used.
5622 The default value is 50 which limits unit growth to 1.5 times the original
5625 @item max-inline-insns-recursive
5626 @itemx max-inline-insns-recursive-auto
5627 Specifies maximum number of instructions out-of-line copy of self recursive inline
5628 function can grow into by performing recursive inlining.
5630 For functions declared inline @option{--param max-inline-insns-recursive} is
5631 taken into acount. For function not declared inline, recursive inlining
5632 happens only when @option{-finline-functions} (included in @option{-O3}) is
5633 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5634 default value is 450.
5636 @item max-inline-recursive-depth
5637 @itemx max-inline-recursive-depth-auto
5638 Specifies maximum recursion depth used by the recursive inlining.
5640 For functions declared inline @option{--param max-inline-recursive-depth} is
5641 taken into acount. For function not declared inline, recursive inlining
5642 happens only when @option{-finline-functions} (included in @option{-O3}) is
5643 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5644 default value is 450.
5646 @item inline-call-cost
5647 Specify cost of call instruction relative to simple arithmetics operations
5648 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
5649 functions and at the same time increases size of leaf function that is believed to
5650 reduce function size by being inlined. In effect it increases amount of
5651 inlining for code having large abstraction penalty (many functions that just
5652 pass the arguments to other functions) and decrease inlining for code with low
5653 abstraction penalty. The default value is 16.
5655 @item max-unrolled-insns
5656 The maximum number of instructions that a loop should have if that loop
5657 is unrolled, and if the loop is unrolled, it determines how many times
5658 the loop code is unrolled.
5660 @item max-average-unrolled-insns
5661 The maximum number of instructions biased by probabilities of their execution
5662 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5663 it determines how many times the loop code is unrolled.
5665 @item max-unroll-times
5666 The maximum number of unrollings of a single loop.
5668 @item max-peeled-insns
5669 The maximum number of instructions that a loop should have if that loop
5670 is peeled, and if the loop is peeled, it determines how many times
5671 the loop code is peeled.
5673 @item max-peel-times
5674 The maximum number of peelings of a single loop.
5676 @item max-completely-peeled-insns
5677 The maximum number of insns of a completely peeled loop.
5679 @item max-completely-peel-times
5680 The maximum number of iterations of a loop to be suitable for complete peeling.
5682 @item max-unswitch-insns
5683 The maximum number of insns of an unswitched loop.
5685 @item max-unswitch-level
5686 The maximum number of branches unswitched in a single loop.
5689 The minimum cost of an expensive expression in the loop invariant motion.
5691 @item iv-consider-all-candidates-bound
5692 Bound on number of candidates for induction variables below that
5693 all candidates are considered for each use in induction variable
5694 optimizations. Only the most relevant candidates are considered
5695 if there are more candidates, to avoid quadratic time complexity.
5697 @item iv-max-considered-uses
5698 The induction variable optimizations give up on loops that contain more
5699 induction variable uses.
5701 @item iv-always-prune-cand-set-bound
5702 If number of candidates in the set is smaller than this value,
5703 we always try to remove unnecessary ivs from the set during its
5704 optimization when a new iv is added to the set.
5706 @item scev-max-expr-size
5707 Bound on size of expressions used in the scalar evolutions analyzer.
5708 Large expressions slow the analyzer.
5710 @item max-iterations-to-track
5712 The maximum number of iterations of a loop the brute force algorithm
5713 for analysis of # of iterations of the loop tries to evaluate.
5715 @item hot-bb-count-fraction
5716 Select fraction of the maximal count of repetitions of basic block in program
5717 given basic block needs to have to be considered hot.
5719 @item hot-bb-frequency-fraction
5720 Select fraction of the maximal frequency of executions of basic block in
5721 function given basic block needs to have to be considered hot
5723 @item tracer-dynamic-coverage
5724 @itemx tracer-dynamic-coverage-feedback
5726 This value is used to limit superblock formation once the given percentage of
5727 executed instructions is covered. This limits unnecessary code size
5730 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5731 feedback is available. The real profiles (as opposed to statically estimated
5732 ones) are much less balanced allowing the threshold to be larger value.
5734 @item tracer-max-code-growth
5735 Stop tail duplication once code growth has reached given percentage. This is
5736 rather hokey argument, as most of the duplicates will be eliminated later in
5737 cross jumping, so it may be set to much higher values than is the desired code
5740 @item tracer-min-branch-ratio
5742 Stop reverse growth when the reverse probability of best edge is less than this
5743 threshold (in percent).
5745 @item tracer-min-branch-ratio
5746 @itemx tracer-min-branch-ratio-feedback
5748 Stop forward growth if the best edge do have probability lower than this
5751 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
5752 compilation for profile feedback and one for compilation without. The value
5753 for compilation with profile feedback needs to be more conservative (higher) in
5754 order to make tracer effective.
5756 @item max-cse-path-length
5758 Maximum number of basic blocks on path that cse considers. The default is 10.
5760 @item global-var-threshold
5762 Counts the number of function calls (@var{n}) and the number of
5763 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
5764 single artificial variable will be created to represent all the
5765 call-clobbered variables at function call sites. This artificial
5766 variable will then be made to alias every call-clobbered variable.
5767 (done as @code{int * size_t} on the host machine; beware overflow).
5769 @item max-aliased-vops
5771 Maximum number of virtual operands allowed to represent aliases
5772 before triggering the alias grouping heuristic. Alias grouping
5773 reduces compile times and memory consumption needed for aliasing at
5774 the expense of precision loss in alias information.
5776 @item ggc-min-expand
5778 GCC uses a garbage collector to manage its own memory allocation. This
5779 parameter specifies the minimum percentage by which the garbage
5780 collector's heap should be allowed to expand between collections.
5781 Tuning this may improve compilation speed; it has no effect on code
5784 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
5785 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
5786 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
5787 GCC is not able to calculate RAM on a particular platform, the lower
5788 bound of 30% is used. Setting this parameter and
5789 @option{ggc-min-heapsize} to zero causes a full collection to occur at
5790 every opportunity. This is extremely slow, but can be useful for
5793 @item ggc-min-heapsize
5795 Minimum size of the garbage collector's heap before it begins bothering
5796 to collect garbage. The first collection occurs after the heap expands
5797 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
5798 tuning this may improve compilation speed, and has no effect on code
5801 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
5802 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
5803 with a lower bound of 4096 (four megabytes) and an upper bound of
5804 131072 (128 megabytes). If GCC is not able to calculate RAM on a
5805 particular platform, the lower bound is used. Setting this parameter
5806 very large effectively disables garbage collection. Setting this
5807 parameter and @option{ggc-min-expand} to zero causes a full collection
5808 to occur at every opportunity.
5810 @item max-reload-search-insns
5811 The maximum number of instruction reload should look backward for equivalent
5812 register. Increasing values mean more aggressive optimization, making the
5813 compile time increase with probably slightly better performance. The default
5816 @item max-cselib-memory-location
5817 The maximum number of memory locations cselib should take into acount.
5818 Increasing values mean more aggressive optimization, making the compile time
5819 increase with probably slightly better performance. The default value is 500.
5821 @item reorder-blocks-duplicate
5822 @itemx reorder-blocks-duplicate-feedback
5824 Used by basic block reordering pass to decide whether to use unconditional
5825 branch or duplicate the code on its destination. Code is duplicated when its
5826 estimated size is smaller than this value multiplied by the estimated size of
5827 unconditional jump in the hot spots of the program.
5829 The @option{reorder-block-duplicate-feedback} is used only when profile
5830 feedback is available and may be set to higher values than
5831 @option{reorder-block-duplicate} since information about the hot spots is more
5834 @item max-sched-region-blocks
5835 The maximum number of blocks in a region to be considered for
5836 interblock scheduling. The default value is 10.
5838 @item max-sched-region-insns
5839 The maximum number of insns in a region to be considered for
5840 interblock scheduling. The default value is 100.
5842 @item max-last-value-rtl
5844 The maximum size measured as number of RTLs that can be recorded in an expression
5845 in combiner for a pseudo register as last known value of that register. The default
5848 @item integer-share-limit
5849 Small integer constants can use a shared data structure, reducing the
5850 compiler's memory usage and increasing its speed. This sets the maximum
5851 value of a shared integer constant's. The default value is 256.
5853 @item min-virtual-mappings
5854 Specifies the minimum number of virtual mappings in the incremental
5855 SSA updater that should be registered to trigger the virtual mappings
5856 heuristic defined by virtual-mappings-ratio. The default value is
5859 @item virtual-mappings-ratio
5860 If the number of virtual mappings is virtual-mappings-ratio bigger
5861 than the number of virtual symbols to be updated, then the incremental
5862 SSA updater switches to a full update for those symbols. The default
5868 @node Preprocessor Options
5869 @section Options Controlling the Preprocessor
5870 @cindex preprocessor options
5871 @cindex options, preprocessor
5873 These options control the C preprocessor, which is run on each C source
5874 file before actual compilation.
5876 If you use the @option{-E} option, nothing is done except preprocessing.
5877 Some of these options make sense only together with @option{-E} because
5878 they cause the preprocessor output to be unsuitable for actual
5883 You can use @option{-Wp,@var{option}} to bypass the compiler driver
5884 and pass @var{option} directly through to the preprocessor. If
5885 @var{option} contains commas, it is split into multiple options at the
5886 commas. However, many options are modified, translated or interpreted
5887 by the compiler driver before being passed to the preprocessor, and
5888 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
5889 interface is undocumented and subject to change, so whenever possible
5890 you should avoid using @option{-Wp} and let the driver handle the
5893 @item -Xpreprocessor @var{option}
5894 @opindex preprocessor
5895 Pass @var{option} as an option to the preprocessor. You can use this to
5896 supply system-specific preprocessor options which GCC does not know how to
5899 If you want to pass an option that takes an argument, you must use
5900 @option{-Xpreprocessor} twice, once for the option and once for the argument.
5903 @include cppopts.texi
5905 @node Assembler Options
5906 @section Passing Options to the Assembler
5908 @c prevent bad page break with this line
5909 You can pass options to the assembler.
5912 @item -Wa,@var{option}
5914 Pass @var{option} as an option to the assembler. If @var{option}
5915 contains commas, it is split into multiple options at the commas.
5917 @item -Xassembler @var{option}
5919 Pass @var{option} as an option to the assembler. You can use this to
5920 supply system-specific assembler options which GCC does not know how to
5923 If you want to pass an option that takes an argument, you must use
5924 @option{-Xassembler} twice, once for the option and once for the argument.
5929 @section Options for Linking
5930 @cindex link options
5931 @cindex options, linking
5933 These options come into play when the compiler links object files into
5934 an executable output file. They are meaningless if the compiler is
5935 not doing a link step.
5939 @item @var{object-file-name}
5940 A file name that does not end in a special recognized suffix is
5941 considered to name an object file or library. (Object files are
5942 distinguished from libraries by the linker according to the file
5943 contents.) If linking is done, these object files are used as input
5952 If any of these options is used, then the linker is not run, and
5953 object file names should not be used as arguments. @xref{Overall
5957 @item -l@var{library}
5958 @itemx -l @var{library}
5960 Search the library named @var{library} when linking. (The second
5961 alternative with the library as a separate argument is only for
5962 POSIX compliance and is not recommended.)
5964 It makes a difference where in the command you write this option; the
5965 linker searches and processes libraries and object files in the order they
5966 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
5967 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
5968 to functions in @samp{z}, those functions may not be loaded.
5970 The linker searches a standard list of directories for the library,
5971 which is actually a file named @file{lib@var{library}.a}. The linker
5972 then uses this file as if it had been specified precisely by name.
5974 The directories searched include several standard system directories
5975 plus any that you specify with @option{-L}.
5977 Normally the files found this way are library files---archive files
5978 whose members are object files. The linker handles an archive file by
5979 scanning through it for members which define symbols that have so far
5980 been referenced but not defined. But if the file that is found is an
5981 ordinary object file, it is linked in the usual fashion. The only
5982 difference between using an @option{-l} option and specifying a file name
5983 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
5984 and searches several directories.
5988 You need this special case of the @option{-l} option in order to
5989 link an Objective-C or Objective-C++ program.
5992 @opindex nostartfiles
5993 Do not use the standard system startup files when linking.
5994 The standard system libraries are used normally, unless @option{-nostdlib}
5995 or @option{-nodefaultlibs} is used.
5997 @item -nodefaultlibs
5998 @opindex nodefaultlibs
5999 Do not use the standard system libraries when linking.
6000 Only the libraries you specify will be passed to the linker.
6001 The standard startup files are used normally, unless @option{-nostartfiles}
6002 is used. The compiler may generate calls to @code{memcmp},
6003 @code{memset}, @code{memcpy} and @code{memmove}.
6004 These entries are usually resolved by entries in
6005 libc. These entry points should be supplied through some other
6006 mechanism when this option is specified.
6010 Do not use the standard system startup files or libraries when linking.
6011 No startup files and only the libraries you specify will be passed to
6012 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6013 @code{memcpy} and @code{memmove}.
6014 These entries are usually resolved by entries in
6015 libc. These entry points should be supplied through some other
6016 mechanism when this option is specified.
6018 @cindex @option{-lgcc}, use with @option{-nostdlib}
6019 @cindex @option{-nostdlib} and unresolved references
6020 @cindex unresolved references and @option{-nostdlib}
6021 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6022 @cindex @option{-nodefaultlibs} and unresolved references
6023 @cindex unresolved references and @option{-nodefaultlibs}
6024 One of the standard libraries bypassed by @option{-nostdlib} and
6025 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6026 that GCC uses to overcome shortcomings of particular machines, or special
6027 needs for some languages.
6028 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6029 Collection (GCC) Internals},
6030 for more discussion of @file{libgcc.a}.)
6031 In most cases, you need @file{libgcc.a} even when you want to avoid
6032 other standard libraries. In other words, when you specify @option{-nostdlib}
6033 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6034 This ensures that you have no unresolved references to internal GCC
6035 library subroutines. (For example, @samp{__main}, used to ensure C++
6036 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6037 GNU Compiler Collection (GCC) Internals}.)
6041 Produce a position independent executable on targets which support it.
6042 For predictable results, you must also specify the same set of options
6043 that were used to generate code (@option{-fpie}, @option{-fPIE},
6044 or model suboptions) when you specify this option.
6048 Remove all symbol table and relocation information from the executable.
6052 On systems that support dynamic linking, this prevents linking with the shared
6053 libraries. On other systems, this option has no effect.
6057 Produce a shared object which can then be linked with other objects to
6058 form an executable. Not all systems support this option. For predictable
6059 results, you must also specify the same set of options that were used to
6060 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6061 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6062 needs to build supplementary stub code for constructors to work. On
6063 multi-libbed systems, @samp{gcc -shared} must select the correct support
6064 libraries to link against. Failing to supply the correct flags may lead
6065 to subtle defects. Supplying them in cases where they are not necessary
6068 @item -shared-libgcc
6069 @itemx -static-libgcc
6070 @opindex shared-libgcc
6071 @opindex static-libgcc
6072 On systems that provide @file{libgcc} as a shared library, these options
6073 force the use of either the shared or static version respectively.
6074 If no shared version of @file{libgcc} was built when the compiler was
6075 configured, these options have no effect.
6077 There are several situations in which an application should use the
6078 shared @file{libgcc} instead of the static version. The most common
6079 of these is when the application wishes to throw and catch exceptions
6080 across different shared libraries. In that case, each of the libraries
6081 as well as the application itself should use the shared @file{libgcc}.
6083 Therefore, the G++ and GCJ drivers automatically add
6084 @option{-shared-libgcc} whenever you build a shared library or a main
6085 executable, because C++ and Java programs typically use exceptions, so
6086 this is the right thing to do.
6088 If, instead, you use the GCC driver to create shared libraries, you may
6089 find that they will not always be linked with the shared @file{libgcc}.
6090 If GCC finds, at its configuration time, that you have a non-GNU linker
6091 or a GNU linker that does not support option @option{--eh-frame-hdr},
6092 it will link the shared version of @file{libgcc} into shared libraries
6093 by default. Otherwise, it will take advantage of the linker and optimize
6094 away the linking with the shared version of @file{libgcc}, linking with
6095 the static version of libgcc by default. This allows exceptions to
6096 propagate through such shared libraries, without incurring relocation
6097 costs at library load time.
6099 However, if a library or main executable is supposed to throw or catch
6100 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6101 for the languages used in the program, or using the option
6102 @option{-shared-libgcc}, such that it is linked with the shared
6107 Bind references to global symbols when building a shared object. Warn
6108 about any unresolved references (unless overridden by the link editor
6109 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6112 @item -Xlinker @var{option}
6114 Pass @var{option} as an option to the linker. You can use this to
6115 supply system-specific linker options which GCC does not know how to
6118 If you want to pass an option that takes an argument, you must use
6119 @option{-Xlinker} twice, once for the option and once for the argument.
6120 For example, to pass @option{-assert definitions}, you must write
6121 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6122 @option{-Xlinker "-assert definitions"}, because this passes the entire
6123 string as a single argument, which is not what the linker expects.
6125 @item -Wl,@var{option}
6127 Pass @var{option} as an option to the linker. If @var{option} contains
6128 commas, it is split into multiple options at the commas.
6130 @item -u @var{symbol}
6132 Pretend the symbol @var{symbol} is undefined, to force linking of
6133 library modules to define it. You can use @option{-u} multiple times with
6134 different symbols to force loading of additional library modules.
6137 @node Directory Options
6138 @section Options for Directory Search
6139 @cindex directory options
6140 @cindex options, directory search
6143 These options specify directories to search for header files, for
6144 libraries and for parts of the compiler:
6149 Add the directory @var{dir} to the head of the list of directories to be
6150 searched for header files. This can be used to override a system header
6151 file, substituting your own version, since these directories are
6152 searched before the system header file directories. However, you should
6153 not use this option to add directories that contain vendor-supplied
6154 system header files (use @option{-isystem} for that). If you use more than
6155 one @option{-I} option, the directories are scanned in left-to-right
6156 order; the standard system directories come after.
6158 If a standard system include directory, or a directory specified with
6159 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6160 option will be ignored. The directory will still be searched but as a
6161 system directory at its normal position in the system include chain.
6162 This is to ensure that GCC's procedure to fix buggy system headers and
6163 the ordering for the include_next directive are not inadvertently changed.
6164 If you really need to change the search order for system directories,
6165 use the @option{-nostdinc} and/or @option{-isystem} options.
6167 @item -iquote@var{dir}
6169 Add the directory @var{dir} to the head of the list of directories to
6170 be searched for header files only for the case of @samp{#include
6171 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6172 otherwise just like @option{-I}.
6176 Add directory @var{dir} to the list of directories to be searched
6179 @item -B@var{prefix}
6181 This option specifies where to find the executables, libraries,
6182 include files, and data files of the compiler itself.
6184 The compiler driver program runs one or more of the subprograms
6185 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6186 @var{prefix} as a prefix for each program it tries to run, both with and
6187 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6189 For each subprogram to be run, the compiler driver first tries the
6190 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6191 was not specified, the driver tries two standard prefixes, which are
6192 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6193 those results in a file name that is found, the unmodified program
6194 name is searched for using the directories specified in your
6195 @env{PATH} environment variable.
6197 The compiler will check to see if the path provided by the @option{-B}
6198 refers to a directory, and if necessary it will add a directory
6199 separator character at the end of the path.
6201 @option{-B} prefixes that effectively specify directory names also apply
6202 to libraries in the linker, because the compiler translates these
6203 options into @option{-L} options for the linker. They also apply to
6204 includes files in the preprocessor, because the compiler translates these
6205 options into @option{-isystem} options for the preprocessor. In this case,
6206 the compiler appends @samp{include} to the prefix.
6208 The run-time support file @file{libgcc.a} can also be searched for using
6209 the @option{-B} prefix, if needed. If it is not found there, the two
6210 standard prefixes above are tried, and that is all. The file is left
6211 out of the link if it is not found by those means.
6213 Another way to specify a prefix much like the @option{-B} prefix is to use
6214 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6217 As a special kludge, if the path provided by @option{-B} is
6218 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6219 9, then it will be replaced by @file{[dir/]include}. This is to help
6220 with boot-strapping the compiler.
6222 @item -specs=@var{file}
6224 Process @var{file} after the compiler reads in the standard @file{specs}
6225 file, in order to override the defaults that the @file{gcc} driver
6226 program uses when determining what switches to pass to @file{cc1},
6227 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6228 @option{-specs=@var{file}} can be specified on the command line, and they
6229 are processed in order, from left to right.
6233 This option has been deprecated. Please use @option{-iquote} instead for
6234 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6235 Any directories you specify with @option{-I} options before the @option{-I-}
6236 option are searched only for the case of @samp{#include "@var{file}"};
6237 they are not searched for @samp{#include <@var{file}>}.
6239 If additional directories are specified with @option{-I} options after
6240 the @option{-I-}, these directories are searched for all @samp{#include}
6241 directives. (Ordinarily @emph{all} @option{-I} directories are used
6244 In addition, the @option{-I-} option inhibits the use of the current
6245 directory (where the current input file came from) as the first search
6246 directory for @samp{#include "@var{file}"}. There is no way to
6247 override this effect of @option{-I-}. With @option{-I.} you can specify
6248 searching the directory which was current when the compiler was
6249 invoked. That is not exactly the same as what the preprocessor does
6250 by default, but it is often satisfactory.
6252 @option{-I-} does not inhibit the use of the standard system directories
6253 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6260 @section Specifying subprocesses and the switches to pass to them
6263 @command{gcc} is a driver program. It performs its job by invoking a
6264 sequence of other programs to do the work of compiling, assembling and
6265 linking. GCC interprets its command-line parameters and uses these to
6266 deduce which programs it should invoke, and which command-line options
6267 it ought to place on their command lines. This behavior is controlled
6268 by @dfn{spec strings}. In most cases there is one spec string for each
6269 program that GCC can invoke, but a few programs have multiple spec
6270 strings to control their behavior. The spec strings built into GCC can
6271 be overridden by using the @option{-specs=} command-line switch to specify
6274 @dfn{Spec files} are plaintext files that are used to construct spec
6275 strings. They consist of a sequence of directives separated by blank
6276 lines. The type of directive is determined by the first non-whitespace
6277 character on the line and it can be one of the following:
6280 @item %@var{command}
6281 Issues a @var{command} to the spec file processor. The commands that can
6285 @item %include <@var{file}>
6287 Search for @var{file} and insert its text at the current point in the
6290 @item %include_noerr <@var{file}>
6291 @cindex %include_noerr
6292 Just like @samp{%include}, but do not generate an error message if the include
6293 file cannot be found.
6295 @item %rename @var{old_name} @var{new_name}
6297 Rename the spec string @var{old_name} to @var{new_name}.
6301 @item *[@var{spec_name}]:
6302 This tells the compiler to create, override or delete the named spec
6303 string. All lines after this directive up to the next directive or
6304 blank line are considered to be the text for the spec string. If this
6305 results in an empty string then the spec will be deleted. (Or, if the
6306 spec did not exist, then nothing will happened.) Otherwise, if the spec
6307 does not currently exist a new spec will be created. If the spec does
6308 exist then its contents will be overridden by the text of this
6309 directive, unless the first character of that text is the @samp{+}
6310 character, in which case the text will be appended to the spec.
6312 @item [@var{suffix}]:
6313 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6314 and up to the next directive or blank line are considered to make up the
6315 spec string for the indicated suffix. When the compiler encounters an
6316 input file with the named suffix, it will processes the spec string in
6317 order to work out how to compile that file. For example:
6324 This says that any input file whose name ends in @samp{.ZZ} should be
6325 passed to the program @samp{z-compile}, which should be invoked with the
6326 command-line switch @option{-input} and with the result of performing the
6327 @samp{%i} substitution. (See below.)
6329 As an alternative to providing a spec string, the text that follows a
6330 suffix directive can be one of the following:
6333 @item @@@var{language}
6334 This says that the suffix is an alias for a known @var{language}. This is
6335 similar to using the @option{-x} command-line switch to GCC to specify a
6336 language explicitly. For example:
6343 Says that .ZZ files are, in fact, C++ source files.
6346 This causes an error messages saying:
6349 @var{name} compiler not installed on this system.
6353 GCC already has an extensive list of suffixes built into it.
6354 This directive will add an entry to the end of the list of suffixes, but
6355 since the list is searched from the end backwards, it is effectively
6356 possible to override earlier entries using this technique.
6360 GCC has the following spec strings built into it. Spec files can
6361 override these strings or create their own. Note that individual
6362 targets can also add their own spec strings to this list.
6365 asm Options to pass to the assembler
6366 asm_final Options to pass to the assembler post-processor
6367 cpp Options to pass to the C preprocessor
6368 cc1 Options to pass to the C compiler
6369 cc1plus Options to pass to the C++ compiler
6370 endfile Object files to include at the end of the link
6371 link Options to pass to the linker
6372 lib Libraries to include on the command line to the linker
6373 libgcc Decides which GCC support library to pass to the linker
6374 linker Sets the name of the linker
6375 predefines Defines to be passed to the C preprocessor
6376 signed_char Defines to pass to CPP to say whether @code{char} is signed
6378 startfile Object files to include at the start of the link
6381 Here is a small example of a spec file:
6387 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6390 This example renames the spec called @samp{lib} to @samp{old_lib} and
6391 then overrides the previous definition of @samp{lib} with a new one.
6392 The new definition adds in some extra command-line options before
6393 including the text of the old definition.
6395 @dfn{Spec strings} are a list of command-line options to be passed to their
6396 corresponding program. In addition, the spec strings can contain
6397 @samp{%}-prefixed sequences to substitute variable text or to
6398 conditionally insert text into the command line. Using these constructs
6399 it is possible to generate quite complex command lines.
6401 Here is a table of all defined @samp{%}-sequences for spec
6402 strings. Note that spaces are not generated automatically around the
6403 results of expanding these sequences. Therefore you can concatenate them
6404 together or combine them with constant text in a single argument.
6408 Substitute one @samp{%} into the program name or argument.
6411 Substitute the name of the input file being processed.
6414 Substitute the basename of the input file being processed.
6415 This is the substring up to (and not including) the last period
6416 and not including the directory.
6419 This is the same as @samp{%b}, but include the file suffix (text after
6423 Marks the argument containing or following the @samp{%d} as a
6424 temporary file name, so that that file will be deleted if GCC exits
6425 successfully. Unlike @samp{%g}, this contributes no text to the
6428 @item %g@var{suffix}
6429 Substitute a file name that has suffix @var{suffix} and is chosen
6430 once per compilation, and mark the argument in the same way as
6431 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6432 name is now chosen in a way that is hard to predict even when previously
6433 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6434 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6435 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6436 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6437 was simply substituted with a file name chosen once per compilation,
6438 without regard to any appended suffix (which was therefore treated
6439 just like ordinary text), making such attacks more likely to succeed.
6441 @item %u@var{suffix}
6442 Like @samp{%g}, but generates a new temporary file name even if
6443 @samp{%u@var{suffix}} was already seen.
6445 @item %U@var{suffix}
6446 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6447 new one if there is no such last file name. In the absence of any
6448 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6449 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6450 would involve the generation of two distinct file names, one
6451 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6452 simply substituted with a file name chosen for the previous @samp{%u},
6453 without regard to any appended suffix.
6455 @item %j@var{suffix}
6456 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6457 writable, and if save-temps is off; otherwise, substitute the name
6458 of a temporary file, just like @samp{%u}. This temporary file is not
6459 meant for communication between processes, but rather as a junk
6462 @item %|@var{suffix}
6463 @itemx %m@var{suffix}
6464 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6465 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6466 all. These are the two most common ways to instruct a program that it
6467 should read from standard input or write to standard output. If you
6468 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6469 construct: see for example @file{f/lang-specs.h}.
6471 @item %.@var{SUFFIX}
6472 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6473 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6474 terminated by the next space or %.
6477 Marks the argument containing or following the @samp{%w} as the
6478 designated output file of this compilation. This puts the argument
6479 into the sequence of arguments that @samp{%o} will substitute later.
6482 Substitutes the names of all the output files, with spaces
6483 automatically placed around them. You should write spaces
6484 around the @samp{%o} as well or the results are undefined.
6485 @samp{%o} is for use in the specs for running the linker.
6486 Input files whose names have no recognized suffix are not compiled
6487 at all, but they are included among the output files, so they will
6491 Substitutes the suffix for object files. Note that this is
6492 handled specially when it immediately follows @samp{%g, %u, or %U},
6493 because of the need for those to form complete file names. The
6494 handling is such that @samp{%O} is treated exactly as if it had already
6495 been substituted, except that @samp{%g, %u, and %U} do not currently
6496 support additional @var{suffix} characters following @samp{%O} as they would
6497 following, for example, @samp{.o}.
6500 Substitutes the standard macro predefinitions for the
6501 current target machine. Use this when running @code{cpp}.
6504 Like @samp{%p}, but puts @samp{__} before and after the name of each
6505 predefined macro, except for macros that start with @samp{__} or with
6506 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6510 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6511 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
6512 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6516 Current argument is the name of a library or startup file of some sort.
6517 Search for that file in a standard list of directories and substitute
6518 the full name found.
6521 Print @var{str} as an error message. @var{str} is terminated by a newline.
6522 Use this when inconsistent options are detected.
6525 Substitute the contents of spec string @var{name} at this point.
6528 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6530 @item %x@{@var{option}@}
6531 Accumulate an option for @samp{%X}.
6534 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6538 Output the accumulated assembler options specified by @option{-Wa}.
6541 Output the accumulated preprocessor options specified by @option{-Wp}.
6544 Process the @code{asm} spec. This is used to compute the
6545 switches to be passed to the assembler.
6548 Process the @code{asm_final} spec. This is a spec string for
6549 passing switches to an assembler post-processor, if such a program is
6553 Process the @code{link} spec. This is the spec for computing the
6554 command line passed to the linker. Typically it will make use of the
6555 @samp{%L %G %S %D and %E} sequences.
6558 Dump out a @option{-L} option for each directory that GCC believes might
6559 contain startup files. If the target supports multilibs then the
6560 current multilib directory will be prepended to each of these paths.
6563 Process the @code{lib} spec. This is a spec string for deciding which
6564 libraries should be included on the command line to the linker.
6567 Process the @code{libgcc} spec. This is a spec string for deciding
6568 which GCC support library should be included on the command line to the linker.
6571 Process the @code{startfile} spec. This is a spec for deciding which
6572 object files should be the first ones passed to the linker. Typically
6573 this might be a file named @file{crt0.o}.
6576 Process the @code{endfile} spec. This is a spec string that specifies
6577 the last object files that will be passed to the linker.
6580 Process the @code{cpp} spec. This is used to construct the arguments
6581 to be passed to the C preprocessor.
6584 Process the @code{cc1} spec. This is used to construct the options to be
6585 passed to the actual C compiler (@samp{cc1}).
6588 Process the @code{cc1plus} spec. This is used to construct the options to be
6589 passed to the actual C++ compiler (@samp{cc1plus}).
6592 Substitute the variable part of a matched option. See below.
6593 Note that each comma in the substituted string is replaced by
6597 Remove all occurrences of @code{-S} from the command line. Note---this
6598 command is position dependent. @samp{%} commands in the spec string
6599 before this one will see @code{-S}, @samp{%} commands in the spec string
6600 after this one will not.
6602 @item %:@var{function}(@var{args})
6603 Call the named function @var{function}, passing it @var{args}.
6604 @var{args} is first processed as a nested spec string, then split
6605 into an argument vector in the usual fashion. The function returns
6606 a string which is processed as if it had appeared literally as part
6607 of the current spec.
6609 The following built-in spec functions are provided:
6612 @item @code{if-exists}
6613 The @code{if-exists} spec function takes one argument, an absolute
6614 pathname to a file. If the file exists, @code{if-exists} returns the
6615 pathname. Here is a small example of its usage:
6619 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6622 @item @code{if-exists-else}
6623 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6624 spec function, except that it takes two arguments. The first argument is
6625 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6626 returns the pathname. If it does not exist, it returns the second argument.
6627 This way, @code{if-exists-else} can be used to select one file or another,
6628 based on the existence of the first. Here is a small example of its usage:
6632 crt0%O%s %:if-exists(crti%O%s) \
6633 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6636 @item @code{replace-outfile}
6637 The @code{replace-outfile} spec function takes two arguments. It looks for the
6638 first argument in the outfiles array and replaces it with the second argument. Here
6639 is a small example of its usage:
6642 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
6648 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6649 If that switch was not specified, this substitutes nothing. Note that
6650 the leading dash is omitted when specifying this option, and it is
6651 automatically inserted if the substitution is performed. Thus the spec
6652 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6653 and would output the command line option @option{-foo}.
6655 @item %W@{@code{S}@}
6656 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6659 @item %@{@code{S}*@}
6660 Substitutes all the switches specified to GCC whose names start
6661 with @code{-S}, but which also take an argument. This is used for
6662 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6663 GCC considers @option{-o foo} as being
6664 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6665 text, including the space. Thus two arguments would be generated.
6667 @item %@{@code{S}*&@code{T}*@}
6668 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6669 (the order of @code{S} and @code{T} in the spec is not significant).
6670 There can be any number of ampersand-separated variables; for each the
6671 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6673 @item %@{@code{S}:@code{X}@}
6674 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6676 @item %@{!@code{S}:@code{X}@}
6677 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6679 @item %@{@code{S}*:@code{X}@}
6680 Substitutes @code{X} if one or more switches whose names start with
6681 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6682 once, no matter how many such switches appeared. However, if @code{%*}
6683 appears somewhere in @code{X}, then @code{X} will be substituted once
6684 for each matching switch, with the @code{%*} replaced by the part of
6685 that switch that matched the @code{*}.
6687 @item %@{.@code{S}:@code{X}@}
6688 Substitutes @code{X}, if processing a file with suffix @code{S}.
6690 @item %@{!.@code{S}:@code{X}@}
6691 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6693 @item %@{@code{S}|@code{P}:@code{X}@}
6694 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6695 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
6696 although they have a stronger binding than the @samp{|}. If @code{%*}
6697 appears in @code{X}, all of the alternatives must be starred, and only
6698 the first matching alternative is substituted.
6700 For example, a spec string like this:
6703 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
6706 will output the following command-line options from the following input
6707 command-line options:
6712 -d fred.c -foo -baz -boggle
6713 -d jim.d -bar -baz -boggle
6716 @item %@{S:X; T:Y; :D@}
6718 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
6719 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
6720 be as many clauses as you need. This may be combined with @code{.},
6721 @code{!}, @code{|}, and @code{*} as needed.
6726 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
6727 construct may contain other nested @samp{%} constructs or spaces, or
6728 even newlines. They are processed as usual, as described above.
6729 Trailing white space in @code{X} is ignored. White space may also
6730 appear anywhere on the left side of the colon in these constructs,
6731 except between @code{.} or @code{*} and the corresponding word.
6733 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
6734 handled specifically in these constructs. If another value of
6735 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
6736 @option{-W} switch is found later in the command line, the earlier
6737 switch value is ignored, except with @{@code{S}*@} where @code{S} is
6738 just one letter, which passes all matching options.
6740 The character @samp{|} at the beginning of the predicate text is used to
6741 indicate that a command should be piped to the following command, but
6742 only if @option{-pipe} is specified.
6744 It is built into GCC which switches take arguments and which do not.
6745 (You might think it would be useful to generalize this to allow each
6746 compiler's spec to say which switches take arguments. But this cannot
6747 be done in a consistent fashion. GCC cannot even decide which input
6748 files have been specified without knowing which switches take arguments,
6749 and it must know which input files to compile in order to tell which
6752 GCC also knows implicitly that arguments starting in @option{-l} are to be
6753 treated as compiler output files, and passed to the linker in their
6754 proper position among the other output files.
6756 @c man begin OPTIONS
6758 @node Target Options
6759 @section Specifying Target Machine and Compiler Version
6760 @cindex target options
6761 @cindex cross compiling
6762 @cindex specifying machine version
6763 @cindex specifying compiler version and target machine
6764 @cindex compiler version, specifying
6765 @cindex target machine, specifying
6767 The usual way to run GCC is to run the executable called @file{gcc}, or
6768 @file{<machine>-gcc} when cross-compiling, or
6769 @file{<machine>-gcc-<version>} to run a version other than the one that
6770 was installed last. Sometimes this is inconvenient, so GCC provides
6771 options that will switch to another cross-compiler or version.
6774 @item -b @var{machine}
6776 The argument @var{machine} specifies the target machine for compilation.
6778 The value to use for @var{machine} is the same as was specified as the
6779 machine type when configuring GCC as a cross-compiler. For
6780 example, if a cross-compiler was configured with @samp{configure
6781 i386v}, meaning to compile for an 80386 running System V, then you
6782 would specify @option{-b i386v} to run that cross compiler.
6784 @item -V @var{version}
6786 The argument @var{version} specifies which version of GCC to run.
6787 This is useful when multiple versions are installed. For example,
6788 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
6791 The @option{-V} and @option{-b} options work by running the
6792 @file{<machine>-gcc-<version>} executable, so there's no real reason to
6793 use them if you can just run that directly.
6795 @node Submodel Options
6796 @section Hardware Models and Configurations
6797 @cindex submodel options
6798 @cindex specifying hardware config
6799 @cindex hardware models and configurations, specifying
6800 @cindex machine dependent options
6802 Earlier we discussed the standard option @option{-b} which chooses among
6803 different installed compilers for completely different target
6804 machines, such as VAX vs.@: 68000 vs.@: 80386.
6806 In addition, each of these target machine types can have its own
6807 special options, starting with @samp{-m}, to choose among various
6808 hardware models or configurations---for example, 68010 vs 68020,
6809 floating coprocessor or none. A single installed version of the
6810 compiler can compile for any model or configuration, according to the
6813 Some configurations of the compiler also support additional special
6814 options, usually for compatibility with other compilers on the same
6817 These options are defined by the macro @code{TARGET_SWITCHES} in the
6818 machine description. The default for the options is also defined by
6819 that macro, which enables you to change the defaults.
6821 @c This list is ordered alphanumerically by subsection name.
6822 @c It should be the same order and spelling as these options are listed
6823 @c in Machine Dependent Options
6829 * Blackfin Options::
6832 * DEC Alpha Options::
6833 * DEC Alpha/VMS Options::
6837 * i386 and x86-64 Options::
6849 * RS/6000 and PowerPC Options::
6850 * S/390 and zSeries Options::
6853 * System V Options::
6854 * TMS320C3x/C4x Options::
6858 * Xstormy16 Options::
6864 @subsection ARC Options
6867 These options are defined for ARC implementations:
6872 Compile code for little endian mode. This is the default.
6876 Compile code for big endian mode.
6879 @opindex mmangle-cpu
6880 Prepend the name of the cpu to all public symbol names.
6881 In multiple-processor systems, there are many ARC variants with different
6882 instruction and register set characteristics. This flag prevents code
6883 compiled for one cpu to be linked with code compiled for another.
6884 No facility exists for handling variants that are ``almost identical''.
6885 This is an all or nothing option.
6887 @item -mcpu=@var{cpu}
6889 Compile code for ARC variant @var{cpu}.
6890 Which variants are supported depend on the configuration.
6891 All variants support @option{-mcpu=base}, this is the default.
6893 @item -mtext=@var{text-section}
6894 @itemx -mdata=@var{data-section}
6895 @itemx -mrodata=@var{readonly-data-section}
6899 Put functions, data, and readonly data in @var{text-section},
6900 @var{data-section}, and @var{readonly-data-section} respectively
6901 by default. This can be overridden with the @code{section} attribute.
6902 @xref{Variable Attributes}.
6907 @subsection ARM Options
6910 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6914 @item -mabi=@var{name}
6916 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
6917 @samp{atpcs}, @samp{aapcs} and @samp{iwmmxt}.
6920 @opindex mapcs-frame
6921 Generate a stack frame that is compliant with the ARM Procedure Call
6922 Standard for all functions, even if this is not strictly necessary for
6923 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6924 with this option will cause the stack frames not to be generated for
6925 leaf functions. The default is @option{-mno-apcs-frame}.
6929 This is a synonym for @option{-mapcs-frame}.
6932 @c not currently implemented
6933 @item -mapcs-stack-check
6934 @opindex mapcs-stack-check
6935 Generate code to check the amount of stack space available upon entry to
6936 every function (that actually uses some stack space). If there is
6937 insufficient space available then either the function
6938 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
6939 called, depending upon the amount of stack space required. The run time
6940 system is required to provide these functions. The default is
6941 @option{-mno-apcs-stack-check}, since this produces smaller code.
6943 @c not currently implemented
6945 @opindex mapcs-float
6946 Pass floating point arguments using the float point registers. This is
6947 one of the variants of the APCS@. This option is recommended if the
6948 target hardware has a floating point unit or if a lot of floating point
6949 arithmetic is going to be performed by the code. The default is
6950 @option{-mno-apcs-float}, since integer only code is slightly increased in
6951 size if @option{-mapcs-float} is used.
6953 @c not currently implemented
6954 @item -mapcs-reentrant
6955 @opindex mapcs-reentrant
6956 Generate reentrant, position independent code. The default is
6957 @option{-mno-apcs-reentrant}.
6960 @item -mthumb-interwork
6961 @opindex mthumb-interwork
6962 Generate code which supports calling between the ARM and Thumb
6963 instruction sets. Without this option the two instruction sets cannot
6964 be reliably used inside one program. The default is
6965 @option{-mno-thumb-interwork}, since slightly larger code is generated
6966 when @option{-mthumb-interwork} is specified.
6968 @item -mno-sched-prolog
6969 @opindex mno-sched-prolog
6970 Prevent the reordering of instructions in the function prolog, or the
6971 merging of those instruction with the instructions in the function's
6972 body. This means that all functions will start with a recognizable set
6973 of instructions (or in fact one of a choice from a small set of
6974 different function prologues), and this information can be used to
6975 locate the start if functions inside an executable piece of code. The
6976 default is @option{-msched-prolog}.
6979 @opindex mhard-float
6980 Generate output containing floating point instructions. This is the
6984 @opindex msoft-float
6985 Generate output containing library calls for floating point.
6986 @strong{Warning:} the requisite libraries are not available for all ARM
6987 targets. Normally the facilities of the machine's usual C compiler are
6988 used, but this cannot be done directly in cross-compilation. You must make
6989 your own arrangements to provide suitable library functions for
6992 @option{-msoft-float} changes the calling convention in the output file;
6993 therefore, it is only useful if you compile @emph{all} of a program with
6994 this option. In particular, you need to compile @file{libgcc.a}, the
6995 library that comes with GCC, with @option{-msoft-float} in order for
6998 @item -mfloat-abi=@var{name}
7000 Specifies which ABI to use for floating point values. Permissible values
7001 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7003 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7004 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7005 of floating point instructions, but still uses the soft-float calling
7008 @item -mlittle-endian
7009 @opindex mlittle-endian
7010 Generate code for a processor running in little-endian mode. This is
7011 the default for all standard configurations.
7014 @opindex mbig-endian
7015 Generate code for a processor running in big-endian mode; the default is
7016 to compile code for a little-endian processor.
7018 @item -mwords-little-endian
7019 @opindex mwords-little-endian
7020 This option only applies when generating code for big-endian processors.
7021 Generate code for a little-endian word order but a big-endian byte
7022 order. That is, a byte order of the form @samp{32107654}. Note: this
7023 option should only be used if you require compatibility with code for
7024 big-endian ARM processors generated by versions of the compiler prior to
7027 @item -mcpu=@var{name}
7029 This specifies the name of the target ARM processor. GCC uses this name
7030 to determine what kind of instructions it can emit when generating
7031 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7032 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7033 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7034 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7035 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7036 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7037 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7038 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7039 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7040 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7041 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7042 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7043 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7044 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
7047 @itemx -mtune=@var{name}
7049 This option is very similar to the @option{-mcpu=} option, except that
7050 instead of specifying the actual target processor type, and hence
7051 restricting which instructions can be used, it specifies that GCC should
7052 tune the performance of the code as if the target were of the type
7053 specified in this option, but still choosing the instructions that it
7054 will generate based on the cpu specified by a @option{-mcpu=} option.
7055 For some ARM implementations better performance can be obtained by using
7058 @item -march=@var{name}
7060 This specifies the name of the target ARM architecture. GCC uses this
7061 name to determine what kind of instructions it can emit when generating
7062 assembly code. This option can be used in conjunction with or instead
7063 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7064 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7065 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7066 @samp{iwmmxt}, @samp{ep9312}.
7068 @item -mfpu=@var{name}
7069 @itemx -mfpe=@var{number}
7070 @itemx -mfp=@var{number}
7074 This specifies what floating point hardware (or hardware emulation) is
7075 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7076 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7077 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7078 with older versions of GCC@.
7080 If @option{-msoft-float} is specified this specifies the format of
7081 floating point values.
7083 @item -mstructure-size-boundary=@var{n}
7084 @opindex mstructure-size-boundary
7085 The size of all structures and unions will be rounded up to a multiple
7086 of the number of bits set by this option. Permissible values are 8, 32
7087 and 64. The default value varies for different toolchains. For the COFF
7088 targeted toolchain the default value is 8. A value of 64 is only allowed
7089 if the underlying ABI supports it.
7091 Specifying the larger number can produce faster, more efficient code, but
7092 can also increase the size of the program. Different values are potentially
7093 incompatible. Code compiled with one value cannot necessarily expect to
7094 work with code or libraries compiled with another value, if they exchange
7095 information using structures or unions.
7097 @item -mabort-on-noreturn
7098 @opindex mabort-on-noreturn
7099 Generate a call to the function @code{abort} at the end of a
7100 @code{noreturn} function. It will be executed if the function tries to
7104 @itemx -mno-long-calls
7105 @opindex mlong-calls
7106 @opindex mno-long-calls
7107 Tells the compiler to perform function calls by first loading the
7108 address of the function into a register and then performing a subroutine
7109 call on this register. This switch is needed if the target function
7110 will lie outside of the 64 megabyte addressing range of the offset based
7111 version of subroutine call instruction.
7113 Even if this switch is enabled, not all function calls will be turned
7114 into long calls. The heuristic is that static functions, functions
7115 which have the @samp{short-call} attribute, functions that are inside
7116 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7117 definitions have already been compiled within the current compilation
7118 unit, will not be turned into long calls. The exception to this rule is
7119 that weak function definitions, functions with the @samp{long-call}
7120 attribute or the @samp{section} attribute, and functions that are within
7121 the scope of a @samp{#pragma long_calls} directive, will always be
7122 turned into long calls.
7124 This feature is not enabled by default. Specifying
7125 @option{-mno-long-calls} will restore the default behavior, as will
7126 placing the function calls within the scope of a @samp{#pragma
7127 long_calls_off} directive. Note these switches have no effect on how
7128 the compiler generates code to handle function calls via function
7131 @item -mnop-fun-dllimport
7132 @opindex mnop-fun-dllimport
7133 Disable support for the @code{dllimport} attribute.
7135 @item -msingle-pic-base
7136 @opindex msingle-pic-base
7137 Treat the register used for PIC addressing as read-only, rather than
7138 loading it in the prologue for each function. The run-time system is
7139 responsible for initializing this register with an appropriate value
7140 before execution begins.
7142 @item -mpic-register=@var{reg}
7143 @opindex mpic-register
7144 Specify the register to be used for PIC addressing. The default is R10
7145 unless stack-checking is enabled, when R9 is used.
7147 @item -mcirrus-fix-invalid-insns
7148 @opindex mcirrus-fix-invalid-insns
7149 @opindex mno-cirrus-fix-invalid-insns
7150 Insert NOPs into the instruction stream to in order to work around
7151 problems with invalid Maverick instruction combinations. This option
7152 is only valid if the @option{-mcpu=ep9312} option has been used to
7153 enable generation of instructions for the Cirrus Maverick floating
7154 point co-processor. This option is not enabled by default, since the
7155 problem is only present in older Maverick implementations. The default
7156 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7159 @item -mpoke-function-name
7160 @opindex mpoke-function-name
7161 Write the name of each function into the text section, directly
7162 preceding the function prologue. The generated code is similar to this:
7166 .ascii "arm_poke_function_name", 0
7169 .word 0xff000000 + (t1 - t0)
7170 arm_poke_function_name
7172 stmfd sp!, @{fp, ip, lr, pc@}
7176 When performing a stack backtrace, code can inspect the value of
7177 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7178 location @code{pc - 12} and the top 8 bits are set, then we know that
7179 there is a function name embedded immediately preceding this location
7180 and has length @code{((pc[-3]) & 0xff000000)}.
7184 Generate code for the 16-bit Thumb instruction set. The default is to
7185 use the 32-bit ARM instruction set.
7188 @opindex mtpcs-frame
7189 Generate a stack frame that is compliant with the Thumb Procedure Call
7190 Standard for all non-leaf functions. (A leaf function is one that does
7191 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7193 @item -mtpcs-leaf-frame
7194 @opindex mtpcs-leaf-frame
7195 Generate a stack frame that is compliant with the Thumb Procedure Call
7196 Standard for all leaf functions. (A leaf function is one that does
7197 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7199 @item -mcallee-super-interworking
7200 @opindex mcallee-super-interworking
7201 Gives all externally visible functions in the file being compiled an ARM
7202 instruction set header which switches to Thumb mode before executing the
7203 rest of the function. This allows these functions to be called from
7204 non-interworking code.
7206 @item -mcaller-super-interworking
7207 @opindex mcaller-super-interworking
7208 Allows calls via function pointers (including virtual functions) to
7209 execute correctly regardless of whether the target code has been
7210 compiled for interworking or not. There is a small overhead in the cost
7211 of executing a function pointer if this option is enabled.
7216 @subsection AVR Options
7219 These options are defined for AVR implementations:
7222 @item -mmcu=@var{mcu}
7224 Specify ATMEL AVR instruction set or MCU type.
7226 Instruction set avr1 is for the minimal AVR core, not supported by the C
7227 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7228 attiny11, attiny12, attiny15, attiny28).
7230 Instruction set avr2 (default) is for the classic AVR core with up to
7231 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7232 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7233 at90c8534, at90s8535).
7235 Instruction set avr3 is for the classic AVR core with up to 128K program
7236 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7238 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7239 memory space (MCU types: atmega8, atmega83, atmega85).
7241 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7242 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7243 atmega64, atmega128, at43usb355, at94k).
7247 Output instruction sizes to the asm file.
7249 @item -minit-stack=@var{N}
7250 @opindex minit-stack
7251 Specify the initial stack address, which may be a symbol or numeric value,
7252 @samp{__stack} is the default.
7254 @item -mno-interrupts
7255 @opindex mno-interrupts
7256 Generated code is not compatible with hardware interrupts.
7257 Code size will be smaller.
7259 @item -mcall-prologues
7260 @opindex mcall-prologues
7261 Functions prologues/epilogues expanded as call to appropriate
7262 subroutines. Code size will be smaller.
7264 @item -mno-tablejump
7265 @opindex mno-tablejump
7266 Do not generate tablejump insns which sometimes increase code size.
7269 @opindex mtiny-stack
7270 Change only the low 8 bits of the stack pointer.
7274 Assume int to be 8 bit integer. This affects the sizes of all types: A
7275 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7276 and long long will be 4 bytes. Please note that this option does not
7277 comply to the C standards, but it will provide you with smaller code
7281 @node Blackfin Options
7282 @subsection Blackfin Options
7283 @cindex Blackfin Options
7286 @item -momit-leaf-frame-pointer
7287 @opindex momit-leaf-frame-pointer
7288 Don't keep the frame pointer in a register for leaf functions. This
7289 avoids the instructions to save, set up and restore frame pointers and
7290 makes an extra register available in leaf functions. The option
7291 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7292 which might make debugging harder.
7296 When enabled, the compiler will ensure that the generated code does not
7297 contain speculative loads after jump instructions. This option is enabled
7302 Don't generate extra code to prevent speculative loads from occurring.
7306 When enabled, the compiler is free to take advantage of the knowledge that
7307 the entire program fits into the low 64k of memory.
7310 @opindex mno-low-64k
7311 Assume that the program is arbitrarily large. This is the default.
7313 @item -mid-shared-library
7314 @opindex mid-shared-library
7315 Generate code that supports shared libraries via the library ID method.
7316 This allows for execute in place and shared libraries in an environment
7317 without virtual memory management. This option implies @option{-fPIC}.
7319 @item -mno-id-shared-library
7320 @opindex mno-id-shared-library
7321 Generate code that doesn't assume ID based shared libraries are being used.
7322 This is the default.
7324 @item -mshared-library-id=n
7325 @opindex mshared-library-id
7326 Specified the identification number of the ID based shared library being
7327 compiled. Specifying a value of 0 will generate more compact code, specifying
7328 other values will force the allocation of that number to the current
7329 library but is no more space or time efficient than omitting this option.
7333 @subsection CRIS Options
7334 @cindex CRIS Options
7336 These options are defined specifically for the CRIS ports.
7339 @item -march=@var{architecture-type}
7340 @itemx -mcpu=@var{architecture-type}
7343 Generate code for the specified architecture. The choices for
7344 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7345 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7346 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7349 @item -mtune=@var{architecture-type}
7351 Tune to @var{architecture-type} everything applicable about the generated
7352 code, except for the ABI and the set of available instructions. The
7353 choices for @var{architecture-type} are the same as for
7354 @option{-march=@var{architecture-type}}.
7356 @item -mmax-stack-frame=@var{n}
7357 @opindex mmax-stack-frame
7358 Warn when the stack frame of a function exceeds @var{n} bytes.
7360 @item -melinux-stacksize=@var{n}
7361 @opindex melinux-stacksize
7362 Only available with the @samp{cris-axis-aout} target. Arranges for
7363 indications in the program to the kernel loader that the stack of the
7364 program should be set to @var{n} bytes.
7370 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7371 @option{-march=v3} and @option{-march=v8} respectively.
7373 @item -mmul-bug-workaround
7374 @itemx -mno-mul-bug-workaround
7375 @opindex mmul-bug-workaround
7376 @opindex mno-mul-bug-workaround
7377 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7378 models where it applies. This option is active by default.
7382 Enable CRIS-specific verbose debug-related information in the assembly
7383 code. This option also has the effect to turn off the @samp{#NO_APP}
7384 formatted-code indicator to the assembler at the beginning of the
7389 Do not use condition-code results from previous instruction; always emit
7390 compare and test instructions before use of condition codes.
7392 @item -mno-side-effects
7393 @opindex mno-side-effects
7394 Do not emit instructions with side-effects in addressing modes other than
7398 @itemx -mno-stack-align
7400 @itemx -mno-data-align
7401 @itemx -mconst-align
7402 @itemx -mno-const-align
7403 @opindex mstack-align
7404 @opindex mno-stack-align
7405 @opindex mdata-align
7406 @opindex mno-data-align
7407 @opindex mconst-align
7408 @opindex mno-const-align
7409 These options (no-options) arranges (eliminate arrangements) for the
7410 stack-frame, individual data and constants to be aligned for the maximum
7411 single data access size for the chosen CPU model. The default is to
7412 arrange for 32-bit alignment. ABI details such as structure layout are
7413 not affected by these options.
7421 Similar to the stack- data- and const-align options above, these options
7422 arrange for stack-frame, writable data and constants to all be 32-bit,
7423 16-bit or 8-bit aligned. The default is 32-bit alignment.
7425 @item -mno-prologue-epilogue
7426 @itemx -mprologue-epilogue
7427 @opindex mno-prologue-epilogue
7428 @opindex mprologue-epilogue
7429 With @option{-mno-prologue-epilogue}, the normal function prologue and
7430 epilogue that sets up the stack-frame are omitted and no return
7431 instructions or return sequences are generated in the code. Use this
7432 option only together with visual inspection of the compiled code: no
7433 warnings or errors are generated when call-saved registers must be saved,
7434 or storage for local variable needs to be allocated.
7440 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7441 instruction sequences that load addresses for functions from the PLT part
7442 of the GOT rather than (traditional on other architectures) calls to the
7443 PLT@. The default is @option{-mgotplt}.
7447 Legacy no-op option only recognized with the cris-axis-aout target.
7451 Legacy no-op option only recognized with the cris-axis-elf and
7452 cris-axis-linux-gnu targets.
7456 Only recognized with the cris-axis-aout target, where it selects a
7457 GNU/linux-like multilib, include files and instruction set for
7462 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7466 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7467 to link with input-output functions from a simulator library. Code,
7468 initialized data and zero-initialized data are allocated consecutively.
7472 Like @option{-sim}, but pass linker options to locate initialized data at
7473 0x40000000 and zero-initialized data at 0x80000000.
7476 @node Darwin Options
7477 @subsection Darwin Options
7478 @cindex Darwin options
7480 These options are defined for all architectures running the Darwin operating
7483 FSF GCC on Darwin does not create ``fat'' object files; it will create
7484 an object file for the single architecture that it was built to
7485 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7486 @option{-arch} options are used; it does so by running the compiler or
7487 linker multiple times and joining the results together with
7490 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7491 @samp{i686}) is determined by the flags that specify the ISA
7492 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7493 @option{-force_cpusubtype_ALL} option can be used to override this.
7495 The Darwin tools vary in their behavior when presented with an ISA
7496 mismatch. The assembler, @file{as}, will only permit instructions to
7497 be used that are valid for the subtype of the file it is generating,
7498 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7499 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
7500 and print an error if asked to create a shared library with a less
7501 restrictive subtype than its input files (for instance, trying to put
7502 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
7503 for executables, @file{ld}, will quietly give the executable the most
7504 restrictive subtype of any of its input files.
7509 Add the framework directory @var{dir} to the head of the list of
7510 directories to be searched for header files. These directories are
7511 interleaved with those specified by @option{-I} options and are
7512 scanned in a left-to-right order.
7514 A framework directory is a directory with frameworks in it. A
7515 framework is a directory with a @samp{"Headers"} and/or
7516 @samp{"PrivateHeaders"} directory contained directly in it that ends
7517 in @samp{".framework"}. The name of a framework is the name of this
7518 directory excluding the @samp{".framework"}. Headers associated with
7519 the framework are found in one of those two directories, with
7520 @samp{"Headers"} being searched first. A subframework is a framework
7521 directory that is in a framework's @samp{"Frameworks"} directory.
7522 Includes of subframework headers can only appear in a header of a
7523 framework that contains the subframework, or in a sibling subframework
7524 header. Two subframeworks are siblings if they occur in the same
7525 framework. A subframework should not have the same name as a
7526 framework, a warning will be issued if this is violated. Currently a
7527 subframework cannot have subframeworks, in the future, the mechanism
7528 may be extended to support this. The standard frameworks can be found
7529 in @samp{"/System/Library/Frameworks"} and
7530 @samp{"/Library/Frameworks"}. An example include looks like
7531 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
7532 the name of the framework and header.h is found in the
7533 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
7537 Emit debugging information for symbols that are used. For STABS
7538 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
7539 This is by default ON@.
7543 Emit debugging information for all symbols and types.
7545 @item -mone-byte-bool
7546 @opindex -mone-byte-bool
7547 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7548 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7549 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7550 option has no effect on x86.
7552 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
7553 to generate code that is not binary compatible with code generated
7554 without that switch. Using this switch may require recompiling all
7555 other modules in a program, including system libraries. Use this
7556 switch to conform to a non-default data model.
7558 @item -mfix-and-continue
7559 @itemx -ffix-and-continue
7560 @itemx -findirect-data
7561 @opindex mfix-and-continue
7562 @opindex ffix-and-continue
7563 @opindex findirect-data
7564 Generate code suitable for fast turn around development. Needed to
7565 enable gdb to dynamically load @code{.o} files into already running
7566 programs. @option{-findirect-data} and @option{-ffix-and-continue}
7567 are provided for backwards compatibility.
7571 Loads all members of static archive libraries.
7572 See man ld(1) for more information.
7574 @item -arch_errors_fatal
7575 @opindex arch_errors_fatal
7576 Cause the errors having to do with files that have the wrong architecture
7580 @opindex bind_at_load
7581 Causes the output file to be marked such that the dynamic linker will
7582 bind all undefined references when the file is loaded or launched.
7586 Produce a Mach-o bundle format file.
7587 See man ld(1) for more information.
7589 @item -bundle_loader @var{executable}
7590 @opindex bundle_loader
7591 This option specifies the @var{executable} that will be loading the build
7592 output file being linked. See man ld(1) for more information.
7595 @opindex -dynamiclib
7596 When passed this option, GCC will produce a dynamic library instead of
7597 an executable when linking, using the Darwin @file{libtool} command.
7599 @item -force_cpusubtype_ALL
7600 @opindex -force_cpusubtype_ALL
7601 This causes GCC's output file to have the @var{ALL} subtype, instead of
7602 one controlled by the @option{-mcpu} or @option{-march} option.
7604 @item -allowable_client @var{client_name}
7606 @itemx -compatibility_version
7607 @itemx -current_version
7609 @itemx -dependency-file
7611 @itemx -dylinker_install_name
7613 @itemx -exported_symbols_list
7615 @itemx -flat_namespace
7616 @itemx -force_flat_namespace
7617 @itemx -headerpad_max_install_names
7620 @itemx -install_name
7621 @itemx -keep_private_externs
7622 @itemx -multi_module
7623 @itemx -multiply_defined
7624 @itemx -multiply_defined_unused
7626 @itemx -no_dead_strip_inits_and_terms
7627 @itemx -nofixprebinding
7630 @itemx -noseglinkedit
7631 @itemx -pagezero_size
7633 @itemx -prebind_all_twolevel_modules
7634 @itemx -private_bundle
7635 @itemx -read_only_relocs
7637 @itemx -sectobjectsymbols
7641 @itemx -sectobjectsymbols
7644 @itemx -segs_read_only_addr
7645 @itemx -segs_read_write_addr
7646 @itemx -seg_addr_table
7647 @itemx -seg_addr_table_filename
7650 @itemx -segs_read_only_addr
7651 @itemx -segs_read_write_addr
7652 @itemx -single_module
7655 @itemx -sub_umbrella
7656 @itemx -twolevel_namespace
7659 @itemx -unexported_symbols_list
7660 @itemx -weak_reference_mismatches
7663 @opindex allowable_client
7664 @opindex client_name
7665 @opindex compatibility_version
7666 @opindex current_version
7668 @opindex dependency-file
7670 @opindex dylinker_install_name
7672 @opindex exported_symbols_list
7674 @opindex flat_namespace
7675 @opindex force_flat_namespace
7676 @opindex headerpad_max_install_names
7679 @opindex install_name
7680 @opindex keep_private_externs
7681 @opindex multi_module
7682 @opindex multiply_defined
7683 @opindex multiply_defined_unused
7685 @opindex no_dead_strip_inits_and_terms
7686 @opindex nofixprebinding
7687 @opindex nomultidefs
7689 @opindex noseglinkedit
7690 @opindex pagezero_size
7692 @opindex prebind_all_twolevel_modules
7693 @opindex private_bundle
7694 @opindex read_only_relocs
7696 @opindex sectobjectsymbols
7700 @opindex sectobjectsymbols
7703 @opindex segs_read_only_addr
7704 @opindex segs_read_write_addr
7705 @opindex seg_addr_table
7706 @opindex seg_addr_table_filename
7707 @opindex seglinkedit
7709 @opindex segs_read_only_addr
7710 @opindex segs_read_write_addr
7711 @opindex single_module
7713 @opindex sub_library
7714 @opindex sub_umbrella
7715 @opindex twolevel_namespace
7718 @opindex unexported_symbols_list
7719 @opindex weak_reference_mismatches
7720 @opindex whatsloaded
7722 These options are passed to the Darwin linker. The Darwin linker man page
7723 describes them in detail.
7726 @node DEC Alpha Options
7727 @subsection DEC Alpha Options
7729 These @samp{-m} options are defined for the DEC Alpha implementations:
7732 @item -mno-soft-float
7734 @opindex mno-soft-float
7735 @opindex msoft-float
7736 Use (do not use) the hardware floating-point instructions for
7737 floating-point operations. When @option{-msoft-float} is specified,
7738 functions in @file{libgcc.a} will be used to perform floating-point
7739 operations. Unless they are replaced by routines that emulate the
7740 floating-point operations, or compiled in such a way as to call such
7741 emulations routines, these routines will issue floating-point
7742 operations. If you are compiling for an Alpha without floating-point
7743 operations, you must ensure that the library is built so as not to call
7746 Note that Alpha implementations without floating-point operations are
7747 required to have floating-point registers.
7752 @opindex mno-fp-regs
7753 Generate code that uses (does not use) the floating-point register set.
7754 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
7755 register set is not used, floating point operands are passed in integer
7756 registers as if they were integers and floating-point results are passed
7757 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
7758 so any function with a floating-point argument or return value called by code
7759 compiled with @option{-mno-fp-regs} must also be compiled with that
7762 A typical use of this option is building a kernel that does not use,
7763 and hence need not save and restore, any floating-point registers.
7767 The Alpha architecture implements floating-point hardware optimized for
7768 maximum performance. It is mostly compliant with the IEEE floating
7769 point standard. However, for full compliance, software assistance is
7770 required. This option generates code fully IEEE compliant code
7771 @emph{except} that the @var{inexact-flag} is not maintained (see below).
7772 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
7773 defined during compilation. The resulting code is less efficient but is
7774 able to correctly support denormalized numbers and exceptional IEEE
7775 values such as not-a-number and plus/minus infinity. Other Alpha
7776 compilers call this option @option{-ieee_with_no_inexact}.
7778 @item -mieee-with-inexact
7779 @opindex mieee-with-inexact
7780 This is like @option{-mieee} except the generated code also maintains
7781 the IEEE @var{inexact-flag}. Turning on this option causes the
7782 generated code to implement fully-compliant IEEE math. In addition to
7783 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
7784 macro. On some Alpha implementations the resulting code may execute
7785 significantly slower than the code generated by default. Since there is
7786 very little code that depends on the @var{inexact-flag}, you should
7787 normally not specify this option. Other Alpha compilers call this
7788 option @option{-ieee_with_inexact}.
7790 @item -mfp-trap-mode=@var{trap-mode}
7791 @opindex mfp-trap-mode
7792 This option controls what floating-point related traps are enabled.
7793 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
7794 The trap mode can be set to one of four values:
7798 This is the default (normal) setting. The only traps that are enabled
7799 are the ones that cannot be disabled in software (e.g., division by zero
7803 In addition to the traps enabled by @samp{n}, underflow traps are enabled
7807 Like @samp{su}, but the instructions are marked to be safe for software
7808 completion (see Alpha architecture manual for details).
7811 Like @samp{su}, but inexact traps are enabled as well.
7814 @item -mfp-rounding-mode=@var{rounding-mode}
7815 @opindex mfp-rounding-mode
7816 Selects the IEEE rounding mode. Other Alpha compilers call this option
7817 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
7822 Normal IEEE rounding mode. Floating point numbers are rounded towards
7823 the nearest machine number or towards the even machine number in case
7827 Round towards minus infinity.
7830 Chopped rounding mode. Floating point numbers are rounded towards zero.
7833 Dynamic rounding mode. A field in the floating point control register
7834 (@var{fpcr}, see Alpha architecture reference manual) controls the
7835 rounding mode in effect. The C library initializes this register for
7836 rounding towards plus infinity. Thus, unless your program modifies the
7837 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
7840 @item -mtrap-precision=@var{trap-precision}
7841 @opindex mtrap-precision
7842 In the Alpha architecture, floating point traps are imprecise. This
7843 means without software assistance it is impossible to recover from a
7844 floating trap and program execution normally needs to be terminated.
7845 GCC can generate code that can assist operating system trap handlers
7846 in determining the exact location that caused a floating point trap.
7847 Depending on the requirements of an application, different levels of
7848 precisions can be selected:
7852 Program precision. This option is the default and means a trap handler
7853 can only identify which program caused a floating point exception.
7856 Function precision. The trap handler can determine the function that
7857 caused a floating point exception.
7860 Instruction precision. The trap handler can determine the exact
7861 instruction that caused a floating point exception.
7864 Other Alpha compilers provide the equivalent options called
7865 @option{-scope_safe} and @option{-resumption_safe}.
7867 @item -mieee-conformant
7868 @opindex mieee-conformant
7869 This option marks the generated code as IEEE conformant. You must not
7870 use this option unless you also specify @option{-mtrap-precision=i} and either
7871 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
7872 is to emit the line @samp{.eflag 48} in the function prologue of the
7873 generated assembly file. Under DEC Unix, this has the effect that
7874 IEEE-conformant math library routines will be linked in.
7876 @item -mbuild-constants
7877 @opindex mbuild-constants
7878 Normally GCC examines a 32- or 64-bit integer constant to
7879 see if it can construct it from smaller constants in two or three
7880 instructions. If it cannot, it will output the constant as a literal and
7881 generate code to load it from the data segment at runtime.
7883 Use this option to require GCC to construct @emph{all} integer constants
7884 using code, even if it takes more instructions (the maximum is six).
7886 You would typically use this option to build a shared library dynamic
7887 loader. Itself a shared library, it must relocate itself in memory
7888 before it can find the variables and constants in its own data segment.
7894 Select whether to generate code to be assembled by the vendor-supplied
7895 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
7913 Indicate whether GCC should generate code to use the optional BWX,
7914 CIX, FIX and MAX instruction sets. The default is to use the instruction
7915 sets supported by the CPU type specified via @option{-mcpu=} option or that
7916 of the CPU on which GCC was built if none was specified.
7921 @opindex mfloat-ieee
7922 Generate code that uses (does not use) VAX F and G floating point
7923 arithmetic instead of IEEE single and double precision.
7925 @item -mexplicit-relocs
7926 @itemx -mno-explicit-relocs
7927 @opindex mexplicit-relocs
7928 @opindex mno-explicit-relocs
7929 Older Alpha assemblers provided no way to generate symbol relocations
7930 except via assembler macros. Use of these macros does not allow
7931 optimal instruction scheduling. GNU binutils as of version 2.12
7932 supports a new syntax that allows the compiler to explicitly mark
7933 which relocations should apply to which instructions. This option
7934 is mostly useful for debugging, as GCC detects the capabilities of
7935 the assembler when it is built and sets the default accordingly.
7939 @opindex msmall-data
7940 @opindex mlarge-data
7941 When @option{-mexplicit-relocs} is in effect, static data is
7942 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
7943 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
7944 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
7945 16-bit relocations off of the @code{$gp} register. This limits the
7946 size of the small data area to 64KB, but allows the variables to be
7947 directly accessed via a single instruction.
7949 The default is @option{-mlarge-data}. With this option the data area
7950 is limited to just below 2GB@. Programs that require more than 2GB of
7951 data must use @code{malloc} or @code{mmap} to allocate the data in the
7952 heap instead of in the program's data segment.
7954 When generating code for shared libraries, @option{-fpic} implies
7955 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
7959 @opindex msmall-text
7960 @opindex mlarge-text
7961 When @option{-msmall-text} is used, the compiler assumes that the
7962 code of the entire program (or shared library) fits in 4MB, and is
7963 thus reachable with a branch instruction. When @option{-msmall-data}
7964 is used, the compiler can assume that all local symbols share the
7965 same @code{$gp} value, and thus reduce the number of instructions
7966 required for a function call from 4 to 1.
7968 The default is @option{-mlarge-text}.
7970 @item -mcpu=@var{cpu_type}
7972 Set the instruction set and instruction scheduling parameters for
7973 machine type @var{cpu_type}. You can specify either the @samp{EV}
7974 style name or the corresponding chip number. GCC supports scheduling
7975 parameters for the EV4, EV5 and EV6 family of processors and will
7976 choose the default values for the instruction set from the processor
7977 you specify. If you do not specify a processor type, GCC will default
7978 to the processor on which the compiler was built.
7980 Supported values for @var{cpu_type} are
7986 Schedules as an EV4 and has no instruction set extensions.
7990 Schedules as an EV5 and has no instruction set extensions.
7994 Schedules as an EV5 and supports the BWX extension.
7999 Schedules as an EV5 and supports the BWX and MAX extensions.
8003 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8007 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8010 @item -mtune=@var{cpu_type}
8012 Set only the instruction scheduling parameters for machine type
8013 @var{cpu_type}. The instruction set is not changed.
8015 @item -mmemory-latency=@var{time}
8016 @opindex mmemory-latency
8017 Sets the latency the scheduler should assume for typical memory
8018 references as seen by the application. This number is highly
8019 dependent on the memory access patterns used by the application
8020 and the size of the external cache on the machine.
8022 Valid options for @var{time} are
8026 A decimal number representing clock cycles.
8032 The compiler contains estimates of the number of clock cycles for
8033 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8034 (also called Dcache, Scache, and Bcache), as well as to main memory.
8035 Note that L3 is only valid for EV5.
8040 @node DEC Alpha/VMS Options
8041 @subsection DEC Alpha/VMS Options
8043 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8046 @item -mvms-return-codes
8047 @opindex mvms-return-codes
8048 Return VMS condition codes from main. The default is to return POSIX
8049 style condition (e.g.@ error) codes.
8053 @subsection FRV Options
8060 Only use the first 32 general purpose registers.
8065 Use all 64 general purpose registers.
8070 Use only the first 32 floating point registers.
8075 Use all 64 floating point registers
8078 @opindex mhard-float
8080 Use hardware instructions for floating point operations.
8083 @opindex msoft-float
8085 Use library routines for floating point operations.
8090 Dynamically allocate condition code registers.
8095 Do not try to dynamically allocate condition code registers, only
8096 use @code{icc0} and @code{fcc0}.
8101 Change ABI to use double word insns.
8106 Do not use double word instructions.
8111 Use floating point double instructions.
8116 Do not use floating point double instructions.
8121 Use media instructions.
8126 Do not use media instructions.
8131 Use multiply and add/subtract instructions.
8136 Do not use multiply and add/subtract instructions.
8141 Select the FDPIC ABI, that uses function descriptors to represent
8142 pointers to functions. Without any PIC/PIE-related options, it
8143 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8144 assumes GOT entries and small data are within a 12-bit range from the
8145 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8146 are computed with 32 bits.
8149 @opindex minline-plt
8151 Enable inlining of PLT entries in function calls to functions that are
8152 not known to bind locally. It has no effect without @option{-mfdpic}.
8153 It's enabled by default if optimizing for speed and compiling for
8154 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8155 optimization option such as @option{-O3} or above is present in the
8161 Assume a large TLS segment when generating thread-local code.
8166 Do not assume a large TLS segment when generating thread-local code.
8171 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8172 that is known to be in read-only sections. It's enabled by default,
8173 except for @option{-fpic} or @option{-fpie}: even though it may help
8174 make the global offset table smaller, it trades 1 instruction for 4.
8175 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8176 one of which may be shared by multiple symbols, and it avoids the need
8177 for a GOT entry for the referenced symbol, so it's more likely to be a
8178 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8180 @item -multilib-library-pic
8181 @opindex multilib-library-pic
8183 Link with the (library, not FD) pic libraries. It's implied by
8184 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8185 @option{-fpic} without @option{-mfdpic}. You should never have to use
8191 Follow the EABI requirement of always creating a frame pointer whenever
8192 a stack frame is allocated. This option is enabled by default and can
8193 be disabled with @option{-mno-linked-fp}.
8196 @opindex mlong-calls
8198 Use indirect addressing to call functions outside the current
8199 compilation unit. This allows the functions to be placed anywhere
8200 within the 32-bit address space.
8202 @item -malign-labels
8203 @opindex malign-labels
8205 Try to align labels to an 8-byte boundary by inserting nops into the
8206 previous packet. This option only has an effect when VLIW packing
8207 is enabled. It doesn't create new packets; it merely adds nops to
8211 @opindex mlibrary-pic
8213 Generate position-independent EABI code.
8218 Use only the first four media accumulator registers.
8223 Use all eight media accumulator registers.
8228 Pack VLIW instructions.
8233 Do not pack VLIW instructions.
8238 Do not mark ABI switches in e_flags.
8243 Enable the use of conditional-move instructions (default).
8245 This switch is mainly for debugging the compiler and will likely be removed
8246 in a future version.
8248 @item -mno-cond-move
8249 @opindex mno-cond-move
8251 Disable the use of conditional-move instructions.
8253 This switch is mainly for debugging the compiler and will likely be removed
8254 in a future version.
8259 Enable the use of conditional set instructions (default).
8261 This switch is mainly for debugging the compiler and will likely be removed
8262 in a future version.
8267 Disable the use of conditional set instructions.
8269 This switch is mainly for debugging the compiler and will likely be removed
8270 in a future version.
8275 Enable the use of conditional execution (default).
8277 This switch is mainly for debugging the compiler and will likely be removed
8278 in a future version.
8280 @item -mno-cond-exec
8281 @opindex mno-cond-exec
8283 Disable the use of conditional execution.
8285 This switch is mainly for debugging the compiler and will likely be removed
8286 in a future version.
8289 @opindex mvliw-branch
8291 Run a pass to pack branches into VLIW instructions (default).
8293 This switch is mainly for debugging the compiler and will likely be removed
8294 in a future version.
8296 @item -mno-vliw-branch
8297 @opindex mno-vliw-branch
8299 Do not run a pass to pack branches into VLIW instructions.
8301 This switch is mainly for debugging the compiler and will likely be removed
8302 in a future version.
8304 @item -mmulti-cond-exec
8305 @opindex mmulti-cond-exec
8307 Enable optimization of @code{&&} and @code{||} in conditional execution
8310 This switch is mainly for debugging the compiler and will likely be removed
8311 in a future version.
8313 @item -mno-multi-cond-exec
8314 @opindex mno-multi-cond-exec
8316 Disable optimization of @code{&&} and @code{||} in conditional execution.
8318 This switch is mainly for debugging the compiler and will likely be removed
8319 in a future version.
8321 @item -mnested-cond-exec
8322 @opindex mnested-cond-exec
8324 Enable nested conditional execution optimizations (default).
8326 This switch is mainly for debugging the compiler and will likely be removed
8327 in a future version.
8329 @item -mno-nested-cond-exec
8330 @opindex mno-nested-cond-exec
8332 Disable nested conditional execution optimizations.
8334 This switch is mainly for debugging the compiler and will likely be removed
8335 in a future version.
8337 @item -mtomcat-stats
8338 @opindex mtomcat-stats
8340 Cause gas to print out tomcat statistics.
8342 @item -mcpu=@var{cpu}
8345 Select the processor type for which to generate code. Possible values are
8346 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8347 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8351 @node H8/300 Options
8352 @subsection H8/300 Options
8354 These @samp{-m} options are defined for the H8/300 implementations:
8359 Shorten some address references at link time, when possible; uses the
8360 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8361 ld, Using ld}, for a fuller description.
8365 Generate code for the H8/300H@.
8369 Generate code for the H8S@.
8373 Generate code for the H8S and H8/300H in the normal mode. This switch
8374 must be used either with @option{-mh} or @option{-ms}.
8378 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8382 Make @code{int} data 32 bits by default.
8386 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8387 The default for the H8/300H and H8S is to align longs and floats on 4
8389 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8390 This option has no effect on the H8/300.
8394 @subsection HPPA Options
8395 @cindex HPPA Options
8397 These @samp{-m} options are defined for the HPPA family of computers:
8400 @item -march=@var{architecture-type}
8402 Generate code for the specified architecture. The choices for
8403 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8404 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8405 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8406 architecture option for your machine. Code compiled for lower numbered
8407 architectures will run on higher numbered architectures, but not the
8411 @itemx -mpa-risc-1-1
8412 @itemx -mpa-risc-2-0
8413 @opindex mpa-risc-1-0
8414 @opindex mpa-risc-1-1
8415 @opindex mpa-risc-2-0
8416 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8419 @opindex mbig-switch
8420 Generate code suitable for big switch tables. Use this option only if
8421 the assembler/linker complain about out of range branches within a switch
8424 @item -mjump-in-delay
8425 @opindex mjump-in-delay
8426 Fill delay slots of function calls with unconditional jump instructions
8427 by modifying the return pointer for the function call to be the target
8428 of the conditional jump.
8430 @item -mdisable-fpregs
8431 @opindex mdisable-fpregs
8432 Prevent floating point registers from being used in any manner. This is
8433 necessary for compiling kernels which perform lazy context switching of
8434 floating point registers. If you use this option and attempt to perform
8435 floating point operations, the compiler will abort.
8437 @item -mdisable-indexing
8438 @opindex mdisable-indexing
8439 Prevent the compiler from using indexing address modes. This avoids some
8440 rather obscure problems when compiling MIG generated code under MACH@.
8442 @item -mno-space-regs
8443 @opindex mno-space-regs
8444 Generate code that assumes the target has no space registers. This allows
8445 GCC to generate faster indirect calls and use unscaled index address modes.
8447 Such code is suitable for level 0 PA systems and kernels.
8449 @item -mfast-indirect-calls
8450 @opindex mfast-indirect-calls
8451 Generate code that assumes calls never cross space boundaries. This
8452 allows GCC to emit code which performs faster indirect calls.
8454 This option will not work in the presence of shared libraries or nested
8457 @item -mfixed-range=@var{register-range}
8458 @opindex mfixed-range
8459 Generate code treating the given register range as fixed registers.
8460 A fixed register is one that the register allocator can not use. This is
8461 useful when compiling kernel code. A register range is specified as
8462 two registers separated by a dash. Multiple register ranges can be
8463 specified separated by a comma.
8465 @item -mlong-load-store
8466 @opindex mlong-load-store
8467 Generate 3-instruction load and store sequences as sometimes required by
8468 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8471 @item -mportable-runtime
8472 @opindex mportable-runtime
8473 Use the portable calling conventions proposed by HP for ELF systems.
8477 Enable the use of assembler directives only GAS understands.
8479 @item -mschedule=@var{cpu-type}
8481 Schedule code according to the constraints for the machine type
8482 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8483 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8484 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8485 proper scheduling option for your machine. The default scheduling is
8489 @opindex mlinker-opt
8490 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8491 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8492 linkers in which they give bogus error messages when linking some programs.
8495 @opindex msoft-float
8496 Generate output containing library calls for floating point.
8497 @strong{Warning:} the requisite libraries are not available for all HPPA
8498 targets. Normally the facilities of the machine's usual C compiler are
8499 used, but this cannot be done directly in cross-compilation. You must make
8500 your own arrangements to provide suitable library functions for
8501 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8502 does provide software floating point support.
8504 @option{-msoft-float} changes the calling convention in the output file;
8505 therefore, it is only useful if you compile @emph{all} of a program with
8506 this option. In particular, you need to compile @file{libgcc.a}, the
8507 library that comes with GCC, with @option{-msoft-float} in order for
8512 Generate the predefine, @code{_SIO}, for server IO@. The default is
8513 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8514 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
8515 options are available under HP-UX and HI-UX@.
8519 Use GNU ld specific options. This passes @option{-shared} to ld when
8520 building a shared library. It is the default when GCC is configured,
8521 explicitly or implicitly, with the GNU linker. This option does not
8522 have any affect on which ld is called, it only changes what parameters
8523 are passed to that ld. The ld that is called is determined by the
8524 @option{--with-ld} configure option, GCC's program search path, and
8525 finally by the user's @env{PATH}. The linker used by GCC can be printed
8526 using @samp{which `gcc -print-prog-name=ld`}.
8530 Use HP ld specific options. This passes @option{-b} to ld when building
8531 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8532 links. It is the default when GCC is configured, explicitly or
8533 implicitly, with the HP linker. This option does not have any affect on
8534 which ld is called, it only changes what parameters are passed to that
8535 ld. The ld that is called is determined by the @option{--with-ld}
8536 configure option, GCC's program search path, and finally by the user's
8537 @env{PATH}. The linker used by GCC can be printed using @samp{which
8538 `gcc -print-prog-name=ld`}.
8541 @opindex mno-long-calls
8542 Generate code that uses long call sequences. This ensures that a call
8543 is always able to reach linker generated stubs. The default is to generate
8544 long calls only when the distance from the call site to the beginning
8545 of the function or translation unit, as the case may be, exceeds a
8546 predefined limit set by the branch type being used. The limits for
8547 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8548 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8551 Distances are measured from the beginning of functions when using the
8552 @option{-ffunction-sections} option, or when using the @option{-mgas}
8553 and @option{-mno-portable-runtime} options together under HP-UX with
8556 It is normally not desirable to use this option as it will degrade
8557 performance. However, it may be useful in large applications,
8558 particularly when partial linking is used to build the application.
8560 The types of long calls used depends on the capabilities of the
8561 assembler and linker, and the type of code being generated. The
8562 impact on systems that support long absolute calls, and long pic
8563 symbol-difference or pc-relative calls should be relatively small.
8564 However, an indirect call is used on 32-bit ELF systems in pic code
8565 and it is quite long.
8567 @item -munix=@var{unix-std}
8569 Generate compiler predefines and select a startfile for the specified
8570 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
8571 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
8572 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
8573 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
8574 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
8577 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
8578 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
8579 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
8580 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
8581 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
8582 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
8584 It is @emph{important} to note that this option changes the interfaces
8585 for various library routines. It also affects the operational behavior
8586 of the C library. Thus, @emph{extreme} care is needed in using this
8589 Library code that is intended to operate with more than one UNIX
8590 standard must test, set and restore the variable @var{__xpg4_extended_mask}
8591 as appropriate. Most GNU software doesn't provide this capability.
8595 Suppress the generation of link options to search libdld.sl when the
8596 @option{-static} option is specified on HP-UX 10 and later.
8600 The HP-UX implementation of setlocale in libc has a dependency on
8601 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8602 when the @option{-static} option is specified, special link options
8603 are needed to resolve this dependency.
8605 On HP-UX 10 and later, the GCC driver adds the necessary options to
8606 link with libdld.sl when the @option{-static} option is specified.
8607 This causes the resulting binary to be dynamic. On the 64-bit port,
8608 the linkers generate dynamic binaries by default in any case. The
8609 @option{-nolibdld} option can be used to prevent the GCC driver from
8610 adding these link options.
8614 Add support for multithreading with the @dfn{dce thread} library
8615 under HP-UX@. This option sets flags for both the preprocessor and
8619 @node i386 and x86-64 Options
8620 @subsection Intel 386 and AMD x86-64 Options
8621 @cindex i386 Options
8622 @cindex x86-64 Options
8623 @cindex Intel 386 Options
8624 @cindex AMD x86-64 Options
8626 These @samp{-m} options are defined for the i386 and x86-64 family of
8630 @item -mtune=@var{cpu-type}
8632 Tune to @var{cpu-type} everything applicable about the generated code, except
8633 for the ABI and the set of available instructions. The choices for
8637 Original Intel's i386 CPU@.
8639 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
8641 Intel Pentium CPU with no MMX support.
8643 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
8644 @item i686, pentiumpro
8645 Intel PentiumPro CPU@.
8647 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
8648 @item pentium3, pentium3m
8649 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
8652 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
8653 support. Used by Centrino notebooks.
8654 @item pentium4, pentium4m
8655 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
8657 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
8660 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
8661 SSE2 and SSE3 instruction set support.
8663 AMD K6 CPU with MMX instruction set support.
8665 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
8666 @item athlon, athlon-tbird
8667 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
8669 @item athlon-4, athlon-xp, athlon-mp
8670 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
8671 instruction set support.
8672 @item k8, opteron, athlon64, athlon-fx
8673 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
8674 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
8676 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
8679 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
8680 instruction set support.
8682 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
8683 implemented for this chip.)
8685 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
8686 implemented for this chip.)
8689 While picking a specific @var{cpu-type} will schedule things appropriately
8690 for that particular chip, the compiler will not generate any code that
8691 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8694 @item -march=@var{cpu-type}
8696 Generate instructions for the machine type @var{cpu-type}. The choices
8697 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8698 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8700 @item -mcpu=@var{cpu-type}
8702 A deprecated synonym for @option{-mtune}.
8711 @opindex mpentiumpro
8712 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8713 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8714 These synonyms are deprecated.
8716 @item -mfpmath=@var{unit}
8718 Generate floating point arithmetics for selected unit @var{unit}. The choices
8723 Use the standard 387 floating point coprocessor present majority of chips and
8724 emulated otherwise. Code compiled with this option will run almost everywhere.
8725 The temporary results are computed in 80bit precision instead of precision
8726 specified by the type resulting in slightly different results compared to most
8727 of other chips. See @option{-ffloat-store} for more detailed description.
8729 This is the default choice for i386 compiler.
8732 Use scalar floating point instructions present in the SSE instruction set.
8733 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8734 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8735 instruction set supports only single precision arithmetics, thus the double and
8736 extended precision arithmetics is still done using 387. Later version, present
8737 only in Pentium4 and the future AMD x86-64 chips supports double precision
8740 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
8741 or @option{-msse2} switches to enable SSE extensions and make this option
8742 effective. For the x86-64 compiler, these extensions are enabled by default.
8744 The resulting code should be considerably faster in the majority of cases and avoid
8745 the numerical instability problems of 387 code, but may break some existing
8746 code that expects temporaries to be 80bit.
8748 This is the default choice for the x86-64 compiler.
8751 Attempt to utilize both instruction sets at once. This effectively double the
8752 amount of available registers and on chips with separate execution units for
8753 387 and SSE the execution resources too. Use this option with care, as it is
8754 still experimental, because the GCC register allocator does not model separate
8755 functional units well resulting in instable performance.
8758 @item -masm=@var{dialect}
8759 @opindex masm=@var{dialect}
8760 Output asm instructions using selected @var{dialect}. Supported choices are
8761 @samp{intel} or @samp{att} (the default one).
8766 @opindex mno-ieee-fp
8767 Control whether or not the compiler uses IEEE floating point
8768 comparisons. These handle correctly the case where the result of a
8769 comparison is unordered.
8772 @opindex msoft-float
8773 Generate output containing library calls for floating point.
8774 @strong{Warning:} the requisite libraries are not part of GCC@.
8775 Normally the facilities of the machine's usual C compiler are used, but
8776 this can't be done directly in cross-compilation. You must make your
8777 own arrangements to provide suitable library functions for
8780 On machines where a function returns floating point results in the 80387
8781 register stack, some floating point opcodes may be emitted even if
8782 @option{-msoft-float} is used.
8784 @item -mno-fp-ret-in-387
8785 @opindex mno-fp-ret-in-387
8786 Do not use the FPU registers for return values of functions.
8788 The usual calling convention has functions return values of types
8789 @code{float} and @code{double} in an FPU register, even if there
8790 is no FPU@. The idea is that the operating system should emulate
8793 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8794 in ordinary CPU registers instead.
8796 @item -mno-fancy-math-387
8797 @opindex mno-fancy-math-387
8798 Some 387 emulators do not support the @code{sin}, @code{cos} and
8799 @code{sqrt} instructions for the 387. Specify this option to avoid
8800 generating those instructions. This option is the default on FreeBSD,
8801 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8802 indicates that the target cpu will always have an FPU and so the
8803 instruction will not need emulation. As of revision 2.6.1, these
8804 instructions are not generated unless you also use the
8805 @option{-funsafe-math-optimizations} switch.
8807 @item -malign-double
8808 @itemx -mno-align-double
8809 @opindex malign-double
8810 @opindex mno-align-double
8811 Control whether GCC aligns @code{double}, @code{long double}, and
8812 @code{long long} variables on a two word boundary or a one word
8813 boundary. Aligning @code{double} variables on a two word boundary will
8814 produce code that runs somewhat faster on a @samp{Pentium} at the
8815 expense of more memory.
8817 @strong{Warning:} if you use the @option{-malign-double} switch,
8818 structures containing the above types will be aligned differently than
8819 the published application binary interface specifications for the 386
8820 and will not be binary compatible with structures in code compiled
8821 without that switch.
8823 @item -m96bit-long-double
8824 @itemx -m128bit-long-double
8825 @opindex m96bit-long-double
8826 @opindex m128bit-long-double
8827 These switches control the size of @code{long double} type. The i386
8828 application binary interface specifies the size to be 96 bits,
8829 so @option{-m96bit-long-double} is the default in 32 bit mode.
8831 Modern architectures (Pentium and newer) would prefer @code{long double}
8832 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8833 conforming to the ABI, this would not be possible. So specifying a
8834 @option{-m128bit-long-double} will align @code{long double}
8835 to a 16 byte boundary by padding the @code{long double} with an additional
8838 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8839 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8841 Notice that neither of these options enable any extra precision over the x87
8842 standard of 80 bits for a @code{long double}.
8844 @strong{Warning:} if you override the default value for your target ABI, the
8845 structures and arrays containing @code{long double} variables will change
8846 their size as well as function calling convention for function taking
8847 @code{long double} will be modified. Hence they will not be binary
8848 compatible with arrays or structures in code compiled without that switch.
8852 @itemx -mno-svr3-shlib
8853 @opindex msvr3-shlib
8854 @opindex mno-svr3-shlib
8855 Control whether GCC places uninitialized local variables into the
8856 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8857 into @code{bss}. These options are meaningful only on System V Release 3.
8861 Use a different function-calling convention, in which functions that
8862 take a fixed number of arguments return with the @code{ret} @var{num}
8863 instruction, which pops their arguments while returning. This saves one
8864 instruction in the caller since there is no need to pop the arguments
8867 You can specify that an individual function is called with this calling
8868 sequence with the function attribute @samp{stdcall}. You can also
8869 override the @option{-mrtd} option by using the function attribute
8870 @samp{cdecl}. @xref{Function Attributes}.
8872 @strong{Warning:} this calling convention is incompatible with the one
8873 normally used on Unix, so you cannot use it if you need to call
8874 libraries compiled with the Unix compiler.
8876 Also, you must provide function prototypes for all functions that
8877 take variable numbers of arguments (including @code{printf});
8878 otherwise incorrect code will be generated for calls to those
8881 In addition, seriously incorrect code will result if you call a
8882 function with too many arguments. (Normally, extra arguments are
8883 harmlessly ignored.)
8885 @item -mregparm=@var{num}
8887 Control how many registers are used to pass integer arguments. By
8888 default, no registers are used to pass arguments, and at most 3
8889 registers can be used. You can control this behavior for a specific
8890 function by using the function attribute @samp{regparm}.
8891 @xref{Function Attributes}.
8893 @strong{Warning:} if you use this switch, and
8894 @var{num} is nonzero, then you must build all modules with the same
8895 value, including any libraries. This includes the system libraries and
8898 @item -mpreferred-stack-boundary=@var{num}
8899 @opindex mpreferred-stack-boundary
8900 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8901 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8902 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8903 size (@option{-Os}), in which case the default is the minimum correct
8904 alignment (4 bytes for x86, and 8 bytes for x86-64).
8906 On Pentium and PentiumPro, @code{double} and @code{long double} values
8907 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8908 suffer significant run time performance penalties. On Pentium III, the
8909 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8910 penalties if it is not 16 byte aligned.
8912 To ensure proper alignment of this values on the stack, the stack boundary
8913 must be as aligned as that required by any value stored on the stack.
8914 Further, every function must be generated such that it keeps the stack
8915 aligned. Thus calling a function compiled with a higher preferred
8916 stack boundary from a function compiled with a lower preferred stack
8917 boundary will most likely misalign the stack. It is recommended that
8918 libraries that use callbacks always use the default setting.
8920 This extra alignment does consume extra stack space, and generally
8921 increases code size. Code that is sensitive to stack space usage, such
8922 as embedded systems and operating system kernels, may want to reduce the
8923 preferred alignment to @option{-mpreferred-stack-boundary=2}.
8941 These switches enable or disable the use of built-in functions that allow
8942 direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow extensions of the
8945 @xref{X86 Built-in Functions}, for details of the functions enabled
8946 and disabled by these switches.
8948 To have SSE/SSE2 instructions generated automatically from floating-point
8949 code, see @option{-mfpmath=sse}.
8952 @itemx -mno-push-args
8954 @opindex mno-push-args
8955 Use PUSH operations to store outgoing parameters. This method is shorter
8956 and usually equally fast as method using SUB/MOV operations and is enabled
8957 by default. In some cases disabling it may improve performance because of
8958 improved scheduling and reduced dependencies.
8960 @item -maccumulate-outgoing-args
8961 @opindex maccumulate-outgoing-args
8962 If enabled, the maximum amount of space required for outgoing arguments will be
8963 computed in the function prologue. This is faster on most modern CPUs
8964 because of reduced dependencies, improved scheduling and reduced stack usage
8965 when preferred stack boundary is not equal to 2. The drawback is a notable
8966 increase in code size. This switch implies @option{-mno-push-args}.
8970 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8971 on thread-safe exception handling must compile and link all code with the
8972 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8973 @option{-D_MT}; when linking, it links in a special thread helper library
8974 @option{-lmingwthrd} which cleans up per thread exception handling data.
8976 @item -mno-align-stringops
8977 @opindex mno-align-stringops
8978 Do not align destination of inlined string operations. This switch reduces
8979 code size and improves performance in case the destination is already aligned,
8980 but GCC doesn't know about it.
8982 @item -minline-all-stringops
8983 @opindex minline-all-stringops
8984 By default GCC inlines string operations only when destination is known to be
8985 aligned at least to 4 byte boundary. This enables more inlining, increase code
8986 size, but may improve performance of code that depends on fast memcpy, strlen
8987 and memset for short lengths.
8989 @item -momit-leaf-frame-pointer
8990 @opindex momit-leaf-frame-pointer
8991 Don't keep the frame pointer in a register for leaf functions. This
8992 avoids the instructions to save, set up and restore frame pointers and
8993 makes an extra register available in leaf functions. The option
8994 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8995 which might make debugging harder.
8997 @item -mtls-direct-seg-refs
8998 @itemx -mno-tls-direct-seg-refs
8999 @opindex mtls-direct-seg-refs
9000 Controls whether TLS variables may be accessed with offsets from the
9001 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
9002 or whether the thread base pointer must be added. Whether or not this
9003 is legal depends on the operating system, and whether it maps the
9004 segment to cover the entire TLS area.
9006 For systems that use GNU libc, the default is on.
9009 These @samp{-m} switches are supported in addition to the above
9010 on AMD x86-64 processors in 64-bit environments.
9017 Generate code for a 32-bit or 64-bit environment.
9018 The 32-bit environment sets int, long and pointer to 32 bits and
9019 generates code that runs on any i386 system.
9020 The 64-bit environment sets int to 32 bits and long and pointer
9021 to 64 bits and generates code for AMD's x86-64 architecture.
9024 @opindex no-red-zone
9025 Do not use a so called red zone for x86-64 code. The red zone is mandated
9026 by the x86-64 ABI, it is a 128-byte area beyond the location of the
9027 stack pointer that will not be modified by signal or interrupt handlers
9028 and therefore can be used for temporary data without adjusting the stack
9029 pointer. The flag @option{-mno-red-zone} disables this red zone.
9031 @item -mcmodel=small
9032 @opindex mcmodel=small
9033 Generate code for the small code model: the program and its symbols must
9034 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
9035 Programs can be statically or dynamically linked. This is the default
9038 @item -mcmodel=kernel
9039 @opindex mcmodel=kernel
9040 Generate code for the kernel code model. The kernel runs in the
9041 negative 2 GB of the address space.
9042 This model has to be used for Linux kernel code.
9044 @item -mcmodel=medium
9045 @opindex mcmodel=medium
9046 Generate code for the medium model: The program is linked in the lower 2
9047 GB of the address space but symbols can be located anywhere in the
9048 address space. Programs can be statically or dynamically linked, but
9049 building of shared libraries are not supported with the medium model.
9051 @item -mcmodel=large
9052 @opindex mcmodel=large
9053 Generate code for the large model: This model makes no assumptions
9054 about addresses and sizes of sections. Currently GCC does not implement
9059 @subsection IA-64 Options
9060 @cindex IA-64 Options
9062 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9066 @opindex mbig-endian
9067 Generate code for a big endian target. This is the default for HP-UX@.
9069 @item -mlittle-endian
9070 @opindex mlittle-endian
9071 Generate code for a little endian target. This is the default for AIX5
9078 Generate (or don't) code for the GNU assembler. This is the default.
9079 @c Also, this is the default if the configure option @option{--with-gnu-as}
9086 Generate (or don't) code for the GNU linker. This is the default.
9087 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9092 Generate code that does not use a global pointer register. The result
9093 is not position independent code, and violates the IA-64 ABI@.
9095 @item -mvolatile-asm-stop
9096 @itemx -mno-volatile-asm-stop
9097 @opindex mvolatile-asm-stop
9098 @opindex mno-volatile-asm-stop
9099 Generate (or don't) a stop bit immediately before and after volatile asm
9102 @item -mregister-names
9103 @itemx -mno-register-names
9104 @opindex mregister-names
9105 @opindex mno-register-names
9106 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9107 the stacked registers. This may make assembler output more readable.
9113 Disable (or enable) optimizations that use the small data section. This may
9114 be useful for working around optimizer bugs.
9117 @opindex mconstant-gp
9118 Generate code that uses a single constant global pointer value. This is
9119 useful when compiling kernel code.
9123 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9124 This is useful when compiling firmware code.
9126 @item -minline-float-divide-min-latency
9127 @opindex minline-float-divide-min-latency
9128 Generate code for inline divides of floating point values
9129 using the minimum latency algorithm.
9131 @item -minline-float-divide-max-throughput
9132 @opindex minline-float-divide-max-throughput
9133 Generate code for inline divides of floating point values
9134 using the maximum throughput algorithm.
9136 @item -minline-int-divide-min-latency
9137 @opindex minline-int-divide-min-latency
9138 Generate code for inline divides of integer values
9139 using the minimum latency algorithm.
9141 @item -minline-int-divide-max-throughput
9142 @opindex minline-int-divide-max-throughput
9143 Generate code for inline divides of integer values
9144 using the maximum throughput algorithm.
9146 @item -minline-sqrt-min-latency
9147 @opindex minline-sqrt-min-latency
9148 Generate code for inline square roots
9149 using the minimum latency algorithm.
9151 @item -minline-sqrt-max-throughput
9152 @opindex minline-sqrt-max-throughput
9153 Generate code for inline square roots
9154 using the maximum throughput algorithm.
9156 @item -mno-dwarf2-asm
9158 @opindex mno-dwarf2-asm
9159 @opindex mdwarf2-asm
9160 Don't (or do) generate assembler code for the DWARF2 line number debugging
9161 info. This may be useful when not using the GNU assembler.
9163 @item -mearly-stop-bits
9164 @itemx -mno-early-stop-bits
9165 @opindex mearly-stop-bits
9166 @opindex mno-early-stop-bits
9167 Allow stop bits to be placed earlier than immediately preceding the
9168 instruction that triggered the stop bit. This can improve instruction
9169 scheduling, but does not always do so.
9171 @item -mfixed-range=@var{register-range}
9172 @opindex mfixed-range
9173 Generate code treating the given register range as fixed registers.
9174 A fixed register is one that the register allocator can not use. This is
9175 useful when compiling kernel code. A register range is specified as
9176 two registers separated by a dash. Multiple register ranges can be
9177 specified separated by a comma.
9179 @item -mtls-size=@var{tls-size}
9181 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9184 @item -mtune-arch=@var{cpu-type}
9186 Tune the instruction scheduling for a particular CPU, Valid values are
9187 itanium, itanium1, merced, itanium2, and mckinley.
9193 Add support for multithreading using the POSIX threads library. This
9194 option sets flags for both the preprocessor and linker. It does
9195 not affect the thread safety of object code produced by the compiler or
9196 that of libraries supplied with it. These are HP-UX specific flags.
9202 Generate code for a 32-bit or 64-bit environment.
9203 The 32-bit environment sets int, long and pointer to 32 bits.
9204 The 64-bit environment sets int to 32 bits and long and pointer
9205 to 64 bits. These are HP-UX specific flags.
9209 @node M32R/D Options
9210 @subsection M32R/D Options
9211 @cindex M32R/D options
9213 These @option{-m} options are defined for Renesas M32R/D architectures:
9218 Generate code for the M32R/2@.
9222 Generate code for the M32R/X@.
9226 Generate code for the M32R@. This is the default.
9229 @opindex mmodel=small
9230 Assume all objects live in the lower 16MB of memory (so that their addresses
9231 can be loaded with the @code{ld24} instruction), and assume all subroutines
9232 are reachable with the @code{bl} instruction.
9233 This is the default.
9235 The addressability of a particular object can be set with the
9236 @code{model} attribute.
9238 @item -mmodel=medium
9239 @opindex mmodel=medium
9240 Assume objects may be anywhere in the 32-bit address space (the compiler
9241 will generate @code{seth/add3} instructions to load their addresses), and
9242 assume all subroutines are reachable with the @code{bl} instruction.
9245 @opindex mmodel=large
9246 Assume objects may be anywhere in the 32-bit address space (the compiler
9247 will generate @code{seth/add3} instructions to load their addresses), and
9248 assume subroutines may not be reachable with the @code{bl} instruction
9249 (the compiler will generate the much slower @code{seth/add3/jl}
9250 instruction sequence).
9253 @opindex msdata=none
9254 Disable use of the small data area. Variables will be put into
9255 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
9256 @code{section} attribute has been specified).
9257 This is the default.
9259 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
9260 Objects may be explicitly put in the small data area with the
9261 @code{section} attribute using one of these sections.
9264 @opindex msdata=sdata
9265 Put small global and static data in the small data area, but do not
9266 generate special code to reference them.
9270 Put small global and static data in the small data area, and generate
9271 special instructions to reference them.
9275 @cindex smaller data references
9276 Put global and static objects less than or equal to @var{num} bytes
9277 into the small data or bss sections instead of the normal data or bss
9278 sections. The default value of @var{num} is 8.
9279 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
9280 for this option to have any effect.
9282 All modules should be compiled with the same @option{-G @var{num}} value.
9283 Compiling with different values of @var{num} may or may not work; if it
9284 doesn't the linker will give an error message---incorrect code will not be
9289 Makes the M32R specific code in the compiler display some statistics
9290 that might help in debugging programs.
9293 @opindex malign-loops
9294 Align all loops to a 32-byte boundary.
9296 @item -mno-align-loops
9297 @opindex mno-align-loops
9298 Do not enforce a 32-byte alignment for loops. This is the default.
9300 @item -missue-rate=@var{number}
9301 @opindex missue-rate=@var{number}
9302 Issue @var{number} instructions per cycle. @var{number} can only be 1
9305 @item -mbranch-cost=@var{number}
9306 @opindex mbranch-cost=@var{number}
9307 @var{number} can only be 1 or 2. If it is 1 then branches will be
9308 preferred over conditional code, if it is 2, then the opposite will
9311 @item -mflush-trap=@var{number}
9312 @opindex mflush-trap=@var{number}
9313 Specifies the trap number to use to flush the cache. The default is
9314 12. Valid numbers are between 0 and 15 inclusive.
9316 @item -mno-flush-trap
9317 @opindex mno-flush-trap
9318 Specifies that the cache cannot be flushed by using a trap.
9320 @item -mflush-func=@var{name}
9321 @opindex mflush-func=@var{name}
9322 Specifies the name of the operating system function to call to flush
9323 the cache. The default is @emph{_flush_cache}, but a function call
9324 will only be used if a trap is not available.
9326 @item -mno-flush-func
9327 @opindex mno-flush-func
9328 Indicates that there is no OS function for flushing the cache.
9332 @node M680x0 Options
9333 @subsection M680x0 Options
9334 @cindex M680x0 options
9336 These are the @samp{-m} options defined for the 68000 series. The default
9337 values for these options depends on which style of 68000 was selected when
9338 the compiler was configured; the defaults for the most common choices are
9346 Generate output for a 68000. This is the default
9347 when the compiler is configured for 68000-based systems.
9349 Use this option for microcontrollers with a 68000 or EC000 core,
9350 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
9356 Generate output for a 68020. This is the default
9357 when the compiler is configured for 68020-based systems.
9361 Generate output containing 68881 instructions for floating point.
9362 This is the default for most 68020 systems unless @option{--nfp} was
9363 specified when the compiler was configured.
9367 Generate output for a 68030. This is the default when the compiler is
9368 configured for 68030-based systems.
9372 Generate output for a 68040. This is the default when the compiler is
9373 configured for 68040-based systems.
9375 This option inhibits the use of 68881/68882 instructions that have to be
9376 emulated by software on the 68040. Use this option if your 68040 does not
9377 have code to emulate those instructions.
9381 Generate output for a 68060. This is the default when the compiler is
9382 configured for 68060-based systems.
9384 This option inhibits the use of 68020 and 68881/68882 instructions that
9385 have to be emulated by software on the 68060. Use this option if your 68060
9386 does not have code to emulate those instructions.
9390 Generate output for a CPU32. This is the default
9391 when the compiler is configured for CPU32-based systems.
9393 Use this option for microcontrollers with a
9394 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
9395 68336, 68340, 68341, 68349 and 68360.
9399 Generate output for a 520X ``coldfire'' family cpu. This is the default
9400 when the compiler is configured for 520X-based systems.
9402 Use this option for microcontroller with a 5200 core, including
9403 the MCF5202, MCF5203, MCF5204 and MCF5202.
9408 Generate output for a 68040, without using any of the new instructions.
9409 This results in code which can run relatively efficiently on either a
9410 68020/68881 or a 68030 or a 68040. The generated code does use the
9411 68881 instructions that are emulated on the 68040.
9415 Generate output for a 68060, without using any of the new instructions.
9416 This results in code which can run relatively efficiently on either a
9417 68020/68881 or a 68030 or a 68040. The generated code does use the
9418 68881 instructions that are emulated on the 68060.
9421 @opindex msoft-float
9422 Generate output containing library calls for floating point.
9423 @strong{Warning:} the requisite libraries are not available for all m68k
9424 targets. Normally the facilities of the machine's usual C compiler are
9425 used, but this can't be done directly in cross-compilation. You must
9426 make your own arrangements to provide suitable library functions for
9427 cross-compilation. The embedded targets @samp{m68k-*-aout} and
9428 @samp{m68k-*-coff} do provide software floating point support.
9432 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9433 Additionally, parameters passed on the stack are also aligned to a
9434 16-bit boundary even on targets whose API mandates promotion to 32-bit.
9437 @opindex mnobitfield
9438 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
9439 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
9443 Do use the bit-field instructions. The @option{-m68020} option implies
9444 @option{-mbitfield}. This is the default if you use a configuration
9445 designed for a 68020.
9449 Use a different function-calling convention, in which functions
9450 that take a fixed number of arguments return with the @code{rtd}
9451 instruction, which pops their arguments while returning. This
9452 saves one instruction in the caller since there is no need to pop
9453 the arguments there.
9455 This calling convention is incompatible with the one normally
9456 used on Unix, so you cannot use it if you need to call libraries
9457 compiled with the Unix compiler.
9459 Also, you must provide function prototypes for all functions that
9460 take variable numbers of arguments (including @code{printf});
9461 otherwise incorrect code will be generated for calls to those
9464 In addition, seriously incorrect code will result if you call a
9465 function with too many arguments. (Normally, extra arguments are
9466 harmlessly ignored.)
9468 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
9469 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
9472 @itemx -mno-align-int
9474 @opindex mno-align-int
9475 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
9476 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
9477 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
9478 Aligning variables on 32-bit boundaries produces code that runs somewhat
9479 faster on processors with 32-bit busses at the expense of more memory.
9481 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
9482 align structures containing the above types differently than
9483 most published application binary interface specifications for the m68k.
9487 Use the pc-relative addressing mode of the 68000 directly, instead of
9488 using a global offset table. At present, this option implies @option{-fpic},
9489 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
9490 not presently supported with @option{-mpcrel}, though this could be supported for
9491 68020 and higher processors.
9493 @item -mno-strict-align
9494 @itemx -mstrict-align
9495 @opindex mno-strict-align
9496 @opindex mstrict-align
9497 Do not (do) assume that unaligned memory references will be handled by
9501 Generate code that allows the data segment to be located in a different
9502 area of memory from the text segment. This allows for execute in place in
9503 an environment without virtual memory management. This option implies
9507 Generate code that assumes that the data segment follows the text segment.
9508 This is the default.
9510 @item -mid-shared-library
9511 Generate code that supports shared libraries via the library ID method.
9512 This allows for execute in place and shared libraries in an environment
9513 without virtual memory management. This option implies @option{-fPIC}.
9515 @item -mno-id-shared-library
9516 Generate code that doesn't assume ID based shared libraries are being used.
9517 This is the default.
9519 @item -mshared-library-id=n
9520 Specified the identification number of the ID based shared library being
9521 compiled. Specifying a value of 0 will generate more compact code, specifying
9522 other values will force the allocation of that number to the current
9523 library but is no more space or time efficient than omitting this option.
9527 @node M68hc1x Options
9528 @subsection M68hc1x Options
9529 @cindex M68hc1x options
9531 These are the @samp{-m} options defined for the 68hc11 and 68hc12
9532 microcontrollers. The default values for these options depends on
9533 which style of microcontroller was selected when the compiler was configured;
9534 the defaults for the most common choices are given below.
9541 Generate output for a 68HC11. This is the default
9542 when the compiler is configured for 68HC11-based systems.
9548 Generate output for a 68HC12. This is the default
9549 when the compiler is configured for 68HC12-based systems.
9555 Generate output for a 68HCS12.
9558 @opindex mauto-incdec
9559 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
9566 Enable the use of 68HC12 min and max instructions.
9569 @itemx -mno-long-calls
9570 @opindex mlong-calls
9571 @opindex mno-long-calls
9572 Treat all calls as being far away (near). If calls are assumed to be
9573 far away, the compiler will use the @code{call} instruction to
9574 call a function and the @code{rtc} instruction for returning.
9578 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9580 @item -msoft-reg-count=@var{count}
9581 @opindex msoft-reg-count
9582 Specify the number of pseudo-soft registers which are used for the
9583 code generation. The maximum number is 32. Using more pseudo-soft
9584 register may or may not result in better code depending on the program.
9585 The default is 4 for 68HC11 and 2 for 68HC12.
9590 @subsection MCore Options
9591 @cindex MCore options
9593 These are the @samp{-m} options defined for the Motorola M*Core
9601 @opindex mno-hardlit
9602 Inline constants into the code stream if it can be done in two
9603 instructions or less.
9609 Use the divide instruction. (Enabled by default).
9611 @item -mrelax-immediate
9612 @itemx -mno-relax-immediate
9613 @opindex mrelax-immediate
9614 @opindex mno-relax-immediate
9615 Allow arbitrary sized immediates in bit operations.
9617 @item -mwide-bitfields
9618 @itemx -mno-wide-bitfields
9619 @opindex mwide-bitfields
9620 @opindex mno-wide-bitfields
9621 Always treat bit-fields as int-sized.
9623 @item -m4byte-functions
9624 @itemx -mno-4byte-functions
9625 @opindex m4byte-functions
9626 @opindex mno-4byte-functions
9627 Force all functions to be aligned to a four byte boundary.
9629 @item -mcallgraph-data
9630 @itemx -mno-callgraph-data
9631 @opindex mcallgraph-data
9632 @opindex mno-callgraph-data
9633 Emit callgraph information.
9636 @itemx -mno-slow-bytes
9637 @opindex mslow-bytes
9638 @opindex mno-slow-bytes
9639 Prefer word access when reading byte quantities.
9641 @item -mlittle-endian
9643 @opindex mlittle-endian
9644 @opindex mbig-endian
9645 Generate code for a little endian target.
9651 Generate code for the 210 processor.
9655 @subsection MIPS Options
9656 @cindex MIPS options
9662 Generate big-endian code.
9666 Generate little-endian code. This is the default for @samp{mips*el-*-*}
9669 @item -march=@var{arch}
9671 Generate code that will run on @var{arch}, which can be the name of a
9672 generic MIPS ISA, or the name of a particular processor.
9674 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
9675 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
9676 The processor names are:
9677 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
9679 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
9680 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
9684 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
9685 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
9686 The special value @samp{from-abi} selects the
9687 most compatible architecture for the selected ABI (that is,
9688 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
9690 In processor names, a final @samp{000} can be abbreviated as @samp{k}
9691 (for example, @samp{-march=r2k}). Prefixes are optional, and
9692 @samp{vr} may be written @samp{r}.
9694 GCC defines two macros based on the value of this option. The first
9695 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
9696 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
9697 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
9698 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
9699 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
9701 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
9702 above. In other words, it will have the full prefix and will not
9703 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
9704 the macro names the resolved architecture (either @samp{"mips1"} or
9705 @samp{"mips3"}). It names the default architecture when no
9706 @option{-march} option is given.
9708 @item -mtune=@var{arch}
9710 Optimize for @var{arch}. Among other things, this option controls
9711 the way instructions are scheduled, and the perceived cost of arithmetic
9712 operations. The list of @var{arch} values is the same as for
9715 When this option is not used, GCC will optimize for the processor
9716 specified by @option{-march}. By using @option{-march} and
9717 @option{-mtune} together, it is possible to generate code that will
9718 run on a family of processors, but optimize the code for one
9719 particular member of that family.
9721 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
9722 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
9723 @samp{-march} ones described above.
9727 Equivalent to @samp{-march=mips1}.
9731 Equivalent to @samp{-march=mips2}.
9735 Equivalent to @samp{-march=mips3}.
9739 Equivalent to @samp{-march=mips4}.
9743 Equivalent to @samp{-march=mips32}.
9747 Equivalent to @samp{-march=mips32r2}.
9751 Equivalent to @samp{-march=mips64}.
9757 Use (do not use) the MIPS16 ISA@.
9769 Generate code for the given ABI@.
9771 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
9772 generates 64-bit code when you select a 64-bit architecture, but you
9773 can use @option{-mgp32} to get 32-bit code instead.
9775 For information about the O64 ABI, see
9776 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
9779 @itemx -mno-abicalls
9781 @opindex mno-abicalls
9782 Generate (do not generate) SVR4-style position-independent code.
9783 @option{-mabicalls} is the default for SVR4-based systems.
9789 Lift (do not lift) the usual restrictions on the size of the global
9792 GCC normally uses a single instruction to load values from the GOT@.
9793 While this is relatively efficient, it will only work if the GOT
9794 is smaller than about 64k. Anything larger will cause the linker
9795 to report an error such as:
9797 @cindex relocation truncated to fit (MIPS)
9799 relocation truncated to fit: R_MIPS_GOT16 foobar
9802 If this happens, you should recompile your code with @option{-mxgot}.
9803 It should then work with very large GOTs, although it will also be
9804 less efficient, since it will take three instructions to fetch the
9805 value of a global symbol.
9807 Note that some linkers can create multiple GOTs. If you have such a
9808 linker, you should only need to use @option{-mxgot} when a single object
9809 file accesses more than 64k's worth of GOT entries. Very few do.
9811 These options have no effect unless GCC is generating position
9816 Assume that general-purpose registers are 32 bits wide.
9820 Assume that general-purpose registers are 64 bits wide.
9824 Assume that floating-point registers are 32 bits wide.
9828 Assume that floating-point registers are 64 bits wide.
9831 @opindex mhard-float
9832 Use floating-point coprocessor instructions.
9835 @opindex msoft-float
9836 Do not use floating-point coprocessor instructions. Implement
9837 floating-point calculations using library calls instead.
9839 @item -msingle-float
9840 @opindex msingle-float
9841 Assume that the floating-point coprocessor only supports single-precision
9844 @itemx -mdouble-float
9845 @opindex mdouble-float
9846 Assume that the floating-point coprocessor supports double-precision
9847 operations. This is the default.
9849 @itemx -mpaired-single
9850 @itemx -mno-paired-single
9851 @opindex mpaired-single
9852 @opindex mno-paired-single
9853 Use (do not use) paired-single floating-point instructions.
9854 @xref{MIPS Paired-Single Support}. This option can only be used
9855 when generating 64-bit code and requires hardware floating-point
9856 support to be enabled.
9862 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
9863 The option @option{-mips3d} implies @option{-mpaired-single}.
9867 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
9868 an explanation of the default and the way that the pointer size is
9873 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
9875 The default size of @code{int}s, @code{long}s and pointers depends on
9876 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
9877 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
9878 32-bit @code{long}s. Pointers are the same size as @code{long}s,
9879 or the same size as integer registers, whichever is smaller.
9885 Assume (do not assume) that all symbols have 32-bit values, regardless
9886 of the selected ABI@. This option is useful in combination with
9887 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
9888 to generate shorter and faster references to symbolic addresses.
9892 @cindex smaller data references (MIPS)
9893 @cindex gp-relative references (MIPS)
9894 Put global and static items less than or equal to @var{num} bytes into
9895 the small data or bss section instead of the normal data or bss section.
9896 This allows the data to be accessed using a single instruction.
9898 All modules should be compiled with the same @option{-G @var{num}}
9901 @item -membedded-data
9902 @itemx -mno-embedded-data
9903 @opindex membedded-data
9904 @opindex mno-embedded-data
9905 Allocate variables to the read-only data section first if possible, then
9906 next in the small data section if possible, otherwise in data. This gives
9907 slightly slower code than the default, but reduces the amount of RAM required
9908 when executing, and thus may be preferred for some embedded systems.
9910 @item -muninit-const-in-rodata
9911 @itemx -mno-uninit-const-in-rodata
9912 @opindex muninit-const-in-rodata
9913 @opindex mno-uninit-const-in-rodata
9914 Put uninitialized @code{const} variables in the read-only data section.
9915 This option is only meaningful in conjunction with @option{-membedded-data}.
9917 @item -msplit-addresses
9918 @itemx -mno-split-addresses
9919 @opindex msplit-addresses
9920 @opindex mno-split-addresses
9921 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
9922 relocation operators. This option has been superseded by
9923 @option{-mexplicit-relocs} but is retained for backwards compatibility.
9925 @item -mexplicit-relocs
9926 @itemx -mno-explicit-relocs
9927 @opindex mexplicit-relocs
9928 @opindex mno-explicit-relocs
9929 Use (do not use) assembler relocation operators when dealing with symbolic
9930 addresses. The alternative, selected by @option{-mno-explicit-relocs},
9931 is to use assembler macros instead.
9933 @option{-mexplicit-relocs} is the default if GCC was configured
9934 to use an assembler that supports relocation operators.
9936 @item -mcheck-zero-division
9937 @itemx -mno-check-zero-division
9938 @opindex mcheck-zero-division
9939 @opindex mno-check-zero-division
9940 Trap (do not trap) on integer division by zero. The default is
9941 @option{-mcheck-zero-division}.
9943 @item -mdivide-traps
9944 @itemx -mdivide-breaks
9945 @opindex mdivide-traps
9946 @opindex mdivide-breaks
9947 MIPS systems check for division by zero by generating either a
9948 conditional trap or a break instruction. Using traps results in
9949 smaller code, but is only supported on MIPS II and later. Also, some
9950 versions of the Linux kernel have a bug that prevents trap from
9951 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
9952 allow conditional traps on architectures that support them and
9953 @option{-mdivide-breaks} to force the use of breaks.
9955 The default is usually @option{-mdivide-traps}, but this can be
9956 overridden at configure time using @option{--with-divide=breaks}.
9957 Divide-by-zero checks can be completely disabled using
9958 @option{-mno-check-zero-division}.
9964 Force (do not force) the use of @code{memcpy()} for non-trivial block
9965 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
9966 most constant-sized copies.
9969 @itemx -mno-long-calls
9970 @opindex mlong-calls
9971 @opindex mno-long-calls
9972 Disable (do not disable) use of the @code{jal} instruction. Calling
9973 functions using @code{jal} is more efficient but requires the caller
9974 and callee to be in the same 256 megabyte segment.
9976 This option has no effect on abicalls code. The default is
9977 @option{-mno-long-calls}.
9983 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
9984 instructions, as provided by the R4650 ISA@.
9987 @itemx -mno-fused-madd
9988 @opindex mfused-madd
9989 @opindex mno-fused-madd
9990 Enable (disable) use of the floating point multiply-accumulate
9991 instructions, when they are available. The default is
9992 @option{-mfused-madd}.
9994 When multiply-accumulate instructions are used, the intermediate
9995 product is calculated to infinite precision and is not subject to
9996 the FCSR Flush to Zero bit. This may be undesirable in some
10001 Tell the MIPS assembler to not run its preprocessor over user
10002 assembler files (with a @samp{.s} suffix) when assembling them.
10005 @itemx -mno-fix-r4000
10006 @opindex mfix-r4000
10007 @opindex mno-fix-r4000
10008 Work around certain R4000 CPU errata:
10011 A double-word or a variable shift may give an incorrect result if executed
10012 immediately after starting an integer division.
10014 A double-word or a variable shift may give an incorrect result if executed
10015 while an integer multiplication is in progress.
10017 An integer division may give an incorrect result if started in a delay slot
10018 of a taken branch or a jump.
10022 @itemx -mno-fix-r4400
10023 @opindex mfix-r4400
10024 @opindex mno-fix-r4400
10025 Work around certain R4400 CPU errata:
10028 A double-word or a variable shift may give an incorrect result if executed
10029 immediately after starting an integer division.
10033 @itemx -mno-fix-vr4120
10034 @opindex mfix-vr4120
10035 Work around certain VR4120 errata:
10038 @code{dmultu} does not always produce the correct result.
10040 @code{div} and @code{ddiv} do not always produce the correct result if one
10041 of the operands is negative.
10043 The workarounds for the division errata rely on special functions in
10044 @file{libgcc.a}. At present, these functions are only provided by
10045 the @code{mips64vr*-elf} configurations.
10047 Other VR4120 errata require a nop to be inserted between certain pairs of
10048 instructions. These errata are handled by the assembler, not by GCC itself.
10051 @opindex mfix-vr4130
10052 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
10053 workarounds are implemented by the assembler rather than by GCC,
10054 although GCC will avoid using @code{mflo} and @code{mfhi} if the
10055 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
10056 instructions are available instead.
10059 @itemx -mno-fix-sb1
10061 Work around certain SB-1 CPU core errata.
10062 (This flag currently works around the SB-1 revision 2
10063 ``F1'' and ``F2'' floating point errata.)
10065 @item -mflush-func=@var{func}
10066 @itemx -mno-flush-func
10067 @opindex mflush-func
10068 Specifies the function to call to flush the I and D caches, or to not
10069 call any such function. If called, the function must take the same
10070 arguments as the common @code{_flush_func()}, that is, the address of the
10071 memory range for which the cache is being flushed, the size of the
10072 memory range, and the number 3 (to flush both caches). The default
10073 depends on the target GCC was configured for, but commonly is either
10074 @samp{_flush_func} or @samp{__cpu_flush}.
10076 @item -mbranch-likely
10077 @itemx -mno-branch-likely
10078 @opindex mbranch-likely
10079 @opindex mno-branch-likely
10080 Enable or disable use of Branch Likely instructions, regardless of the
10081 default for the selected architecture. By default, Branch Likely
10082 instructions may be generated if they are supported by the selected
10083 architecture. An exception is for the MIPS32 and MIPS64 architectures
10084 and processors which implement those architectures; for those, Branch
10085 Likely instructions will not be generated by default because the MIPS32
10086 and MIPS64 architectures specifically deprecate their use.
10088 @item -mfp-exceptions
10089 @itemx -mno-fp-exceptions
10090 @opindex mfp-exceptions
10091 Specifies whether FP exceptions are enabled. This affects how we schedule
10092 FP instructions for some processors. The default is that FP exceptions are
10095 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
10096 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
10099 @item -mvr4130-align
10100 @itemx -mno-vr4130-align
10101 @opindex mvr4130-align
10102 The VR4130 pipeline is two-way superscalar, but can only issue two
10103 instructions together if the first one is 8-byte aligned. When this
10104 option is enabled, GCC will align pairs of instructions that it
10105 thinks should execute in parallel.
10107 This option only has an effect when optimizing for the VR4130.
10108 It normally makes code faster, but at the expense of making it bigger.
10109 It is enabled by default at optimization level @option{-O3}.
10113 @subsection MMIX Options
10114 @cindex MMIX Options
10116 These options are defined for the MMIX:
10120 @itemx -mno-libfuncs
10122 @opindex mno-libfuncs
10123 Specify that intrinsic library functions are being compiled, passing all
10124 values in registers, no matter the size.
10127 @itemx -mno-epsilon
10129 @opindex mno-epsilon
10130 Generate floating-point comparison instructions that compare with respect
10131 to the @code{rE} epsilon register.
10133 @item -mabi=mmixware
10135 @opindex mabi-mmixware
10137 Generate code that passes function parameters and return values that (in
10138 the called function) are seen as registers @code{$0} and up, as opposed to
10139 the GNU ABI which uses global registers @code{$231} and up.
10141 @item -mzero-extend
10142 @itemx -mno-zero-extend
10143 @opindex mzero-extend
10144 @opindex mno-zero-extend
10145 When reading data from memory in sizes shorter than 64 bits, use (do not
10146 use) zero-extending load instructions by default, rather than
10147 sign-extending ones.
10150 @itemx -mno-knuthdiv
10152 @opindex mno-knuthdiv
10153 Make the result of a division yielding a remainder have the same sign as
10154 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10155 remainder follows the sign of the dividend. Both methods are
10156 arithmetically valid, the latter being almost exclusively used.
10158 @item -mtoplevel-symbols
10159 @itemx -mno-toplevel-symbols
10160 @opindex mtoplevel-symbols
10161 @opindex mno-toplevel-symbols
10162 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10163 code can be used with the @code{PREFIX} assembly directive.
10167 Generate an executable in the ELF format, rather than the default
10168 @samp{mmo} format used by the @command{mmix} simulator.
10170 @item -mbranch-predict
10171 @itemx -mno-branch-predict
10172 @opindex mbranch-predict
10173 @opindex mno-branch-predict
10174 Use (do not use) the probable-branch instructions, when static branch
10175 prediction indicates a probable branch.
10177 @item -mbase-addresses
10178 @itemx -mno-base-addresses
10179 @opindex mbase-addresses
10180 @opindex mno-base-addresses
10181 Generate (do not generate) code that uses @emph{base addresses}. Using a
10182 base address automatically generates a request (handled by the assembler
10183 and the linker) for a constant to be set up in a global register. The
10184 register is used for one or more base address requests within the range 0
10185 to 255 from the value held in the register. The generally leads to short
10186 and fast code, but the number of different data items that can be
10187 addressed is limited. This means that a program that uses lots of static
10188 data may require @option{-mno-base-addresses}.
10190 @item -msingle-exit
10191 @itemx -mno-single-exit
10192 @opindex msingle-exit
10193 @opindex mno-single-exit
10194 Force (do not force) generated code to have a single exit point in each
10198 @node MN10300 Options
10199 @subsection MN10300 Options
10200 @cindex MN10300 options
10202 These @option{-m} options are defined for Matsushita MN10300 architectures:
10207 Generate code to avoid bugs in the multiply instructions for the MN10300
10208 processors. This is the default.
10210 @item -mno-mult-bug
10211 @opindex mno-mult-bug
10212 Do not generate code to avoid bugs in the multiply instructions for the
10213 MN10300 processors.
10217 Generate code which uses features specific to the AM33 processor.
10221 Do not generate code which uses features specific to the AM33 processor. This
10226 Do not link in the C run-time initialization object file.
10230 Indicate to the linker that it should perform a relaxation optimization pass
10231 to shorten branches, calls and absolute memory addresses. This option only
10232 has an effect when used on the command line for the final link step.
10234 This option makes symbolic debugging impossible.
10237 @node NS32K Options
10238 @subsection NS32K Options
10239 @cindex NS32K options
10241 These are the @samp{-m} options defined for the 32000 series. The default
10242 values for these options depends on which style of 32000 was selected when
10243 the compiler was configured; the defaults for the most common choices are
10251 Generate output for a 32032. This is the default
10252 when the compiler is configured for 32032 and 32016 based systems.
10258 Generate output for a 32332. This is the default
10259 when the compiler is configured for 32332-based systems.
10265 Generate output for a 32532. This is the default
10266 when the compiler is configured for 32532-based systems.
10270 Generate output containing 32081 instructions for floating point.
10271 This is the default for all systems.
10275 Generate output containing 32381 instructions for floating point. This
10276 also implies @option{-m32081}. The 32381 is only compatible with the 32332
10277 and 32532 cpus. This is the default for the pc532-netbsd configuration.
10280 @opindex mmulti-add
10281 Try and generate multiply-add floating point instructions @code{polyF}
10282 and @code{dotF}. This option is only available if the @option{-m32381}
10283 option is in effect. Using these instructions requires changes to
10284 register allocation which generally has a negative impact on
10285 performance. This option should only be enabled when compiling code
10286 particularly likely to make heavy use of multiply-add instructions.
10288 @item -mnomulti-add
10289 @opindex mnomulti-add
10290 Do not try and generate multiply-add floating point instructions
10291 @code{polyF} and @code{dotF}. This is the default on all platforms.
10294 @opindex msoft-float
10295 Generate output containing library calls for floating point.
10296 @strong{Warning:} the requisite libraries may not be available.
10298 @item -mieee-compare
10299 @itemx -mno-ieee-compare
10300 @opindex mieee-compare
10301 @opindex mno-ieee-compare
10302 Control whether or not the compiler uses IEEE floating point
10303 comparisons. These handle correctly the case where the result of a
10304 comparison is unordered.
10305 @strong{Warning:} the requisite kernel support may not be available.
10308 @opindex mnobitfield
10309 Do not use the bit-field instructions. On some machines it is faster to
10310 use shifting and masking operations. This is the default for the pc532.
10314 Do use the bit-field instructions. This is the default for all platforms
10319 Use a different function-calling convention, in which functions
10320 that take a fixed number of arguments return pop their
10321 arguments on return with the @code{ret} instruction.
10323 This calling convention is incompatible with the one normally
10324 used on Unix, so you cannot use it if you need to call libraries
10325 compiled with the Unix compiler.
10327 Also, you must provide function prototypes for all functions that
10328 take variable numbers of arguments (including @code{printf});
10329 otherwise incorrect code will be generated for calls to those
10332 In addition, seriously incorrect code will result if you call a
10333 function with too many arguments. (Normally, extra arguments are
10334 harmlessly ignored.)
10336 This option takes its name from the 680x0 @code{rtd} instruction.
10341 Use a different function-calling convention where the first two arguments
10342 are passed in registers.
10344 This calling convention is incompatible with the one normally
10345 used on Unix, so you cannot use it if you need to call libraries
10346 compiled with the Unix compiler.
10349 @opindex mnoregparam
10350 Do not pass any arguments in registers. This is the default for all
10355 It is OK to use the sb as an index register which is always loaded with
10356 zero. This is the default for the pc532-netbsd target.
10360 The sb register is not available for use or has not been initialized to
10361 zero by the run time system. This is the default for all targets except
10362 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
10363 @option{-fpic} is set.
10367 Many ns32000 series addressing modes use displacements of up to 512MB@.
10368 If an address is above 512MB then displacements from zero can not be used.
10369 This option causes code to be generated which can be loaded above 512MB@.
10370 This may be useful for operating systems or ROM code.
10374 Assume code will be loaded in the first 512MB of virtual address space.
10375 This is the default for all platforms.
10379 @node PDP-11 Options
10380 @subsection PDP-11 Options
10381 @cindex PDP-11 Options
10383 These options are defined for the PDP-11:
10388 Use hardware FPP floating point. This is the default. (FIS floating
10389 point on the PDP-11/40 is not supported.)
10392 @opindex msoft-float
10393 Do not use hardware floating point.
10397 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10401 Return floating-point results in memory. This is the default.
10405 Generate code for a PDP-11/40.
10409 Generate code for a PDP-11/45. This is the default.
10413 Generate code for a PDP-11/10.
10415 @item -mbcopy-builtin
10416 @opindex bcopy-builtin
10417 Use inline @code{movmemhi} patterns for copying memory. This is the
10422 Do not use inline @code{movmemhi} patterns for copying memory.
10428 Use 16-bit @code{int}. This is the default.
10434 Use 32-bit @code{int}.
10437 @itemx -mno-float32
10439 @opindex mno-float32
10440 Use 64-bit @code{float}. This is the default.
10443 @itemx -mno-float64
10445 @opindex mno-float64
10446 Use 32-bit @code{float}.
10450 Use @code{abshi2} pattern. This is the default.
10454 Do not use @code{abshi2} pattern.
10456 @item -mbranch-expensive
10457 @opindex mbranch-expensive
10458 Pretend that branches are expensive. This is for experimenting with
10459 code generation only.
10461 @item -mbranch-cheap
10462 @opindex mbranch-cheap
10463 Do not pretend that branches are expensive. This is the default.
10467 Generate code for a system with split I&D@.
10471 Generate code for a system without split I&D@. This is the default.
10475 Use Unix assembler syntax. This is the default when configured for
10476 @samp{pdp11-*-bsd}.
10480 Use DEC assembler syntax. This is the default when configured for any
10481 PDP-11 target other than @samp{pdp11-*-bsd}.
10484 @node PowerPC Options
10485 @subsection PowerPC Options
10486 @cindex PowerPC options
10488 These are listed under @xref{RS/6000 and PowerPC Options}.
10490 @node RS/6000 and PowerPC Options
10491 @subsection IBM RS/6000 and PowerPC Options
10492 @cindex RS/6000 and PowerPC Options
10493 @cindex IBM RS/6000 and PowerPC Options
10495 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
10502 @itemx -mno-powerpc
10503 @itemx -mpowerpc-gpopt
10504 @itemx -mno-powerpc-gpopt
10505 @itemx -mpowerpc-gfxopt
10506 @itemx -mno-powerpc-gfxopt
10508 @itemx -mno-powerpc64
10512 @opindex mno-power2
10514 @opindex mno-powerpc
10515 @opindex mpowerpc-gpopt
10516 @opindex mno-powerpc-gpopt
10517 @opindex mpowerpc-gfxopt
10518 @opindex mno-powerpc-gfxopt
10519 @opindex mpowerpc64
10520 @opindex mno-powerpc64
10521 GCC supports two related instruction set architectures for the
10522 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
10523 instructions supported by the @samp{rios} chip set used in the original
10524 RS/6000 systems and the @dfn{PowerPC} instruction set is the
10525 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
10526 the IBM 4xx microprocessors.
10528 Neither architecture is a subset of the other. However there is a
10529 large common subset of instructions supported by both. An MQ
10530 register is included in processors supporting the POWER architecture.
10532 You use these options to specify which instructions are available on the
10533 processor you are using. The default value of these options is
10534 determined when configuring GCC@. Specifying the
10535 @option{-mcpu=@var{cpu_type}} overrides the specification of these
10536 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
10537 rather than the options listed above.
10539 The @option{-mpower} option allows GCC to generate instructions that
10540 are found only in the POWER architecture and to use the MQ register.
10541 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
10542 to generate instructions that are present in the POWER2 architecture but
10543 not the original POWER architecture.
10545 The @option{-mpowerpc} option allows GCC to generate instructions that
10546 are found only in the 32-bit subset of the PowerPC architecture.
10547 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
10548 GCC to use the optional PowerPC architecture instructions in the
10549 General Purpose group, including floating-point square root. Specifying
10550 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
10551 use the optional PowerPC architecture instructions in the Graphics
10552 group, including floating-point select.
10554 The @option{-mpowerpc64} option allows GCC to generate the additional
10555 64-bit instructions that are found in the full PowerPC64 architecture
10556 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
10557 @option{-mno-powerpc64}.
10559 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
10560 will use only the instructions in the common subset of both
10561 architectures plus some special AIX common-mode calls, and will not use
10562 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
10563 permits GCC to use any instruction from either architecture and to
10564 allow use of the MQ register; specify this for the Motorola MPC601.
10566 @item -mnew-mnemonics
10567 @itemx -mold-mnemonics
10568 @opindex mnew-mnemonics
10569 @opindex mold-mnemonics
10570 Select which mnemonics to use in the generated assembler code. With
10571 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
10572 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
10573 assembler mnemonics defined for the POWER architecture. Instructions
10574 defined in only one architecture have only one mnemonic; GCC uses that
10575 mnemonic irrespective of which of these options is specified.
10577 GCC defaults to the mnemonics appropriate for the architecture in
10578 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
10579 value of these option. Unless you are building a cross-compiler, you
10580 should normally not specify either @option{-mnew-mnemonics} or
10581 @option{-mold-mnemonics}, but should instead accept the default.
10583 @item -mcpu=@var{cpu_type}
10585 Set architecture type, register usage, choice of mnemonics, and
10586 instruction scheduling parameters for machine type @var{cpu_type}.
10587 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
10588 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
10589 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
10590 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
10591 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
10592 @samp{860}, @samp{970}, @samp{8540}, @samp{common}, @samp{ec603e}, @samp{G3},
10593 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
10594 @samp{power4}, @samp{power5}, @samp{powerpc}, @samp{powerpc64},
10595 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
10597 @option{-mcpu=common} selects a completely generic processor. Code
10598 generated under this option will run on any POWER or PowerPC processor.
10599 GCC will use only the instructions in the common subset of both
10600 architectures, and will not use the MQ register. GCC assumes a generic
10601 processor model for scheduling purposes.
10603 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
10604 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
10605 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
10606 types, with an appropriate, generic processor model assumed for
10607 scheduling purposes.
10609 The other options specify a specific processor. Code generated under
10610 those options will run best on that processor, and may not run at all on
10613 The @option{-mcpu} options automatically enable or disable the
10614 following options: @option{-maltivec}, @option{-mhard-float},
10615 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
10616 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
10617 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
10618 @option{-mstring}. The particular options set for any particular CPU
10619 will vary between compiler versions, depending on what setting seems
10620 to produce optimal code for that CPU; it doesn't necessarily reflect
10621 the actual hardware's capabilities. If you wish to set an individual
10622 option to a particular value, you may specify it after the
10623 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
10625 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
10626 not enabled or disabled by the @option{-mcpu} option at present, since
10627 AIX does not have full support for these options. You may still
10628 enable or disable them individually if you're sure it'll work in your
10631 @item -mtune=@var{cpu_type}
10633 Set the instruction scheduling parameters for machine type
10634 @var{cpu_type}, but do not set the architecture type, register usage, or
10635 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
10636 values for @var{cpu_type} are used for @option{-mtune} as for
10637 @option{-mcpu}. If both are specified, the code generated will use the
10638 architecture, registers, and mnemonics set by @option{-mcpu}, but the
10639 scheduling parameters set by @option{-mtune}.
10642 @itemx -mno-altivec
10644 @opindex mno-altivec
10645 Generate code that uses (does not use) AltiVec instructions, and also
10646 enable the use of built-in functions that allow more direct access to
10647 the AltiVec instruction set. You may also need to set
10648 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
10654 @opindex mno-vrsave
10655 Generate VRSAVE instructions when generating AltiVec code.
10659 Extend the current ABI with SPE ABI extensions. This does not change
10660 the default ABI, instead it adds the SPE ABI extensions to the current
10664 @opindex mabi=no-spe
10665 Disable Booke SPE ABI extensions for the current ABI@.
10671 This switch enables or disables the generation of ISEL instructions.
10673 @item -misel=@var{yes/no}
10674 This switch has been deprecated. Use @option{-misel} and
10675 @option{-mno-isel} instead.
10681 This switch enables or disables the generation of SPE simd
10684 @item -mspe=@var{yes/no}
10685 This option has been deprecated. Use @option{-mspe} and
10686 @option{-mno-spe} instead.
10688 @item -mfloat-gprs=@var{yes/single/double/no}
10689 @itemx -mfloat-gprs
10690 @opindex mfloat-gprs
10691 This switch enables or disables the generation of floating point
10692 operations on the general purpose registers for architectures that
10695 The argument @var{yes} or @var{single} enables the use of
10696 single-precision floating point operations.
10698 The argument @var{double} enables the use of single and
10699 double-precision floating point operations.
10701 The argument @var{no} disables floating point operations on the
10702 general purpose registers.
10704 This option is currently only available on the MPC854x.
10710 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
10711 targets (including GNU/Linux). The 32-bit environment sets int, long
10712 and pointer to 32 bits and generates code that runs on any PowerPC
10713 variant. The 64-bit environment sets int to 32 bits and long and
10714 pointer to 64 bits, and generates code for PowerPC64, as for
10715 @option{-mpowerpc64}.
10718 @itemx -mno-fp-in-toc
10719 @itemx -mno-sum-in-toc
10720 @itemx -mminimal-toc
10722 @opindex mno-fp-in-toc
10723 @opindex mno-sum-in-toc
10724 @opindex mminimal-toc
10725 Modify generation of the TOC (Table Of Contents), which is created for
10726 every executable file. The @option{-mfull-toc} option is selected by
10727 default. In that case, GCC will allocate at least one TOC entry for
10728 each unique non-automatic variable reference in your program. GCC
10729 will also place floating-point constants in the TOC@. However, only
10730 16,384 entries are available in the TOC@.
10732 If you receive a linker error message that saying you have overflowed
10733 the available TOC space, you can reduce the amount of TOC space used
10734 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
10735 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
10736 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
10737 generate code to calculate the sum of an address and a constant at
10738 run-time instead of putting that sum into the TOC@. You may specify one
10739 or both of these options. Each causes GCC to produce very slightly
10740 slower and larger code at the expense of conserving TOC space.
10742 If you still run out of space in the TOC even when you specify both of
10743 these options, specify @option{-mminimal-toc} instead. This option causes
10744 GCC to make only one TOC entry for every file. When you specify this
10745 option, GCC will produce code that is slower and larger but which
10746 uses extremely little TOC space. You may wish to use this option
10747 only on files that contain less frequently executed code.
10753 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
10754 @code{long} type, and the infrastructure needed to support them.
10755 Specifying @option{-maix64} implies @option{-mpowerpc64} and
10756 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
10757 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
10760 @itemx -mno-xl-compat
10761 @opindex mxl-compat
10762 @opindex mno-xl-compat
10763 Produce code that conforms more closely to IBM XLC semantics when using
10764 AIX-compatible ABI. Pass floating-point arguments to prototyped
10765 functions beyond the register save area (RSA) on the stack in addition
10766 to argument FPRs. Do not assume that most significant double in 128
10767 bit long double value is properly rounded when comparing values.
10769 The AIX calling convention was extended but not initially documented to
10770 handle an obscure K&R C case of calling a function that takes the
10771 address of its arguments with fewer arguments than declared. AIX XL
10772 compilers access floating point arguments which do not fit in the
10773 RSA from the stack when a subroutine is compiled without
10774 optimization. Because always storing floating-point arguments on the
10775 stack is inefficient and rarely needed, this option is not enabled by
10776 default and only is necessary when calling subroutines compiled by AIX
10777 XL compilers without optimization.
10781 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
10782 application written to use message passing with special startup code to
10783 enable the application to run. The system must have PE installed in the
10784 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
10785 must be overridden with the @option{-specs=} option to specify the
10786 appropriate directory location. The Parallel Environment does not
10787 support threads, so the @option{-mpe} option and the @option{-pthread}
10788 option are incompatible.
10790 @item -malign-natural
10791 @itemx -malign-power
10792 @opindex malign-natural
10793 @opindex malign-power
10794 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
10795 @option{-malign-natural} overrides the ABI-defined alignment of larger
10796 types, such as floating-point doubles, on their natural size-based boundary.
10797 The option @option{-malign-power} instructs GCC to follow the ABI-specified
10798 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
10800 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
10804 @itemx -mhard-float
10805 @opindex msoft-float
10806 @opindex mhard-float
10807 Generate code that does not use (uses) the floating-point register set.
10808 Software floating point emulation is provided if you use the
10809 @option{-msoft-float} option, and pass the option to GCC when linking.
10812 @itemx -mno-multiple
10814 @opindex mno-multiple
10815 Generate code that uses (does not use) the load multiple word
10816 instructions and the store multiple word instructions. These
10817 instructions are generated by default on POWER systems, and not
10818 generated on PowerPC systems. Do not use @option{-mmultiple} on little
10819 endian PowerPC systems, since those instructions do not work when the
10820 processor is in little endian mode. The exceptions are PPC740 and
10821 PPC750 which permit the instructions usage in little endian mode.
10826 @opindex mno-string
10827 Generate code that uses (does not use) the load string instructions
10828 and the store string word instructions to save multiple registers and
10829 do small block moves. These instructions are generated by default on
10830 POWER systems, and not generated on PowerPC systems. Do not use
10831 @option{-mstring} on little endian PowerPC systems, since those
10832 instructions do not work when the processor is in little endian mode.
10833 The exceptions are PPC740 and PPC750 which permit the instructions
10834 usage in little endian mode.
10839 @opindex mno-update
10840 Generate code that uses (does not use) the load or store instructions
10841 that update the base register to the address of the calculated memory
10842 location. These instructions are generated by default. If you use
10843 @option{-mno-update}, there is a small window between the time that the
10844 stack pointer is updated and the address of the previous frame is
10845 stored, which means code that walks the stack frame across interrupts or
10846 signals may get corrupted data.
10849 @itemx -mno-fused-madd
10850 @opindex mfused-madd
10851 @opindex mno-fused-madd
10852 Generate code that uses (does not use) the floating point multiply and
10853 accumulate instructions. These instructions are generated by default if
10854 hardware floating is used.
10856 @item -mno-bit-align
10858 @opindex mno-bit-align
10859 @opindex mbit-align
10860 On System V.4 and embedded PowerPC systems do not (do) force structures
10861 and unions that contain bit-fields to be aligned to the base type of the
10864 For example, by default a structure containing nothing but 8
10865 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
10866 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
10867 the structure would be aligned to a 1 byte boundary and be one byte in
10870 @item -mno-strict-align
10871 @itemx -mstrict-align
10872 @opindex mno-strict-align
10873 @opindex mstrict-align
10874 On System V.4 and embedded PowerPC systems do not (do) assume that
10875 unaligned memory references will be handled by the system.
10877 @item -mrelocatable
10878 @itemx -mno-relocatable
10879 @opindex mrelocatable
10880 @opindex mno-relocatable
10881 On embedded PowerPC systems generate code that allows (does not allow)
10882 the program to be relocated to a different address at runtime. If you
10883 use @option{-mrelocatable} on any module, all objects linked together must
10884 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
10886 @item -mrelocatable-lib
10887 @itemx -mno-relocatable-lib
10888 @opindex mrelocatable-lib
10889 @opindex mno-relocatable-lib
10890 On embedded PowerPC systems generate code that allows (does not allow)
10891 the program to be relocated to a different address at runtime. Modules
10892 compiled with @option{-mrelocatable-lib} can be linked with either modules
10893 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
10894 with modules compiled with the @option{-mrelocatable} options.
10900 On System V.4 and embedded PowerPC systems do not (do) assume that
10901 register 2 contains a pointer to a global area pointing to the addresses
10902 used in the program.
10905 @itemx -mlittle-endian
10907 @opindex mlittle-endian
10908 On System V.4 and embedded PowerPC systems compile code for the
10909 processor in little endian mode. The @option{-mlittle-endian} option is
10910 the same as @option{-mlittle}.
10913 @itemx -mbig-endian
10915 @opindex mbig-endian
10916 On System V.4 and embedded PowerPC systems compile code for the
10917 processor in big endian mode. The @option{-mbig-endian} option is
10918 the same as @option{-mbig}.
10920 @item -mdynamic-no-pic
10921 @opindex mdynamic-no-pic
10922 On Darwin and Mac OS X systems, compile code so that it is not
10923 relocatable, but that its external references are relocatable. The
10924 resulting code is suitable for applications, but not shared
10927 @item -mprioritize-restricted-insns=@var{priority}
10928 @opindex mprioritize-restricted-insns
10929 This option controls the priority that is assigned to
10930 dispatch-slot restricted instructions during the second scheduling
10931 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
10932 @var{no/highest/second-highest} priority to dispatch slot restricted
10935 @item -msched-costly-dep=@var{dependence_type}
10936 @opindex msched-costly-dep
10937 This option controls which dependences are considered costly
10938 by the target during instruction scheduling. The argument
10939 @var{dependence_type} takes one of the following values:
10940 @var{no}: no dependence is costly,
10941 @var{all}: all dependences are costly,
10942 @var{true_store_to_load}: a true dependence from store to load is costly,
10943 @var{store_to_load}: any dependence from store to load is costly,
10944 @var{number}: any dependence which latency >= @var{number} is costly.
10946 @item -minsert-sched-nops=@var{scheme}
10947 @opindex minsert-sched-nops
10948 This option controls which nop insertion scheme will be used during
10949 the second scheduling pass. The argument @var{scheme} takes one of the
10951 @var{no}: Don't insert nops.
10952 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
10953 according to the scheduler's grouping.
10954 @var{regroup_exact}: Insert nops to force costly dependent insns into
10955 separate groups. Insert exactly as many nops as needed to force an insn
10956 to a new group, according to the estimated processor grouping.
10957 @var{number}: Insert nops to force costly dependent insns into
10958 separate groups. Insert @var{number} nops to force an insn to a new group.
10961 @opindex mcall-sysv
10962 On System V.4 and embedded PowerPC systems compile code using calling
10963 conventions that adheres to the March 1995 draft of the System V
10964 Application Binary Interface, PowerPC processor supplement. This is the
10965 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
10967 @item -mcall-sysv-eabi
10968 @opindex mcall-sysv-eabi
10969 Specify both @option{-mcall-sysv} and @option{-meabi} options.
10971 @item -mcall-sysv-noeabi
10972 @opindex mcall-sysv-noeabi
10973 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
10975 @item -mcall-solaris
10976 @opindex mcall-solaris
10977 On System V.4 and embedded PowerPC systems compile code for the Solaris
10981 @opindex mcall-linux
10982 On System V.4 and embedded PowerPC systems compile code for the
10983 Linux-based GNU system.
10987 On System V.4 and embedded PowerPC systems compile code for the
10988 Hurd-based GNU system.
10990 @item -mcall-netbsd
10991 @opindex mcall-netbsd
10992 On System V.4 and embedded PowerPC systems compile code for the
10993 NetBSD operating system.
10995 @item -maix-struct-return
10996 @opindex maix-struct-return
10997 Return all structures in memory (as specified by the AIX ABI)@.
10999 @item -msvr4-struct-return
11000 @opindex msvr4-struct-return
11001 Return structures smaller than 8 bytes in registers (as specified by the
11004 @item -mabi=@var{abi-type}
11006 Extend the current ABI with a particular extension, or remove such extension.
11007 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
11011 @itemx -mno-prototype
11012 @opindex mprototype
11013 @opindex mno-prototype
11014 On System V.4 and embedded PowerPC systems assume that all calls to
11015 variable argument functions are properly prototyped. Otherwise, the
11016 compiler must insert an instruction before every non prototyped call to
11017 set or clear bit 6 of the condition code register (@var{CR}) to
11018 indicate whether floating point values were passed in the floating point
11019 registers in case the function takes a variable arguments. With
11020 @option{-mprototype}, only calls to prototyped variable argument functions
11021 will set or clear the bit.
11025 On embedded PowerPC systems, assume that the startup module is called
11026 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
11027 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
11032 On embedded PowerPC systems, assume that the startup module is called
11033 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
11038 On embedded PowerPC systems, assume that the startup module is called
11039 @file{crt0.o} and the standard C libraries are @file{libads.a} and
11042 @item -myellowknife
11043 @opindex myellowknife
11044 On embedded PowerPC systems, assume that the startup module is called
11045 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
11050 On System V.4 and embedded PowerPC systems, specify that you are
11051 compiling for a VxWorks system.
11055 Specify that you are compiling for the WindISS simulation environment.
11059 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
11060 header to indicate that @samp{eabi} extended relocations are used.
11066 On System V.4 and embedded PowerPC systems do (do not) adhere to the
11067 Embedded Applications Binary Interface (eabi) which is a set of
11068 modifications to the System V.4 specifications. Selecting @option{-meabi}
11069 means that the stack is aligned to an 8 byte boundary, a function
11070 @code{__eabi} is called to from @code{main} to set up the eabi
11071 environment, and the @option{-msdata} option can use both @code{r2} and
11072 @code{r13} to point to two separate small data areas. Selecting
11073 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
11074 do not call an initialization function from @code{main}, and the
11075 @option{-msdata} option will only use @code{r13} to point to a single
11076 small data area. The @option{-meabi} option is on by default if you
11077 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
11080 @opindex msdata=eabi
11081 On System V.4 and embedded PowerPC systems, put small initialized
11082 @code{const} global and static data in the @samp{.sdata2} section, which
11083 is pointed to by register @code{r2}. Put small initialized
11084 non-@code{const} global and static data in the @samp{.sdata} section,
11085 which is pointed to by register @code{r13}. Put small uninitialized
11086 global and static data in the @samp{.sbss} section, which is adjacent to
11087 the @samp{.sdata} section. The @option{-msdata=eabi} option is
11088 incompatible with the @option{-mrelocatable} option. The
11089 @option{-msdata=eabi} option also sets the @option{-memb} option.
11092 @opindex msdata=sysv
11093 On System V.4 and embedded PowerPC systems, put small global and static
11094 data in the @samp{.sdata} section, which is pointed to by register
11095 @code{r13}. Put small uninitialized global and static data in the
11096 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11097 The @option{-msdata=sysv} option is incompatible with the
11098 @option{-mrelocatable} option.
11100 @item -msdata=default
11102 @opindex msdata=default
11104 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
11105 compile code the same as @option{-msdata=eabi}, otherwise compile code the
11106 same as @option{-msdata=sysv}.
11109 @opindex msdata-data
11110 On System V.4 and embedded PowerPC systems, put small global and static
11111 data in the @samp{.sdata} section. Put small uninitialized global and
11112 static data in the @samp{.sbss} section. Do not use register @code{r13}
11113 to address small data however. This is the default behavior unless
11114 other @option{-msdata} options are used.
11118 @opindex msdata=none
11120 On embedded PowerPC systems, put all initialized global and static data
11121 in the @samp{.data} section, and all uninitialized data in the
11122 @samp{.bss} section.
11126 @cindex smaller data references (PowerPC)
11127 @cindex .sdata/.sdata2 references (PowerPC)
11128 On embedded PowerPC systems, put global and static items less than or
11129 equal to @var{num} bytes into the small data or bss sections instead of
11130 the normal data or bss section. By default, @var{num} is 8. The
11131 @option{-G @var{num}} switch is also passed to the linker.
11132 All modules should be compiled with the same @option{-G @var{num}} value.
11135 @itemx -mno-regnames
11137 @opindex mno-regnames
11138 On System V.4 and embedded PowerPC systems do (do not) emit register
11139 names in the assembly language output using symbolic forms.
11142 @itemx -mno-longcall
11144 @opindex mno-longcall
11145 Default to making all function calls indirectly, using a register, so
11146 that functions which reside further than 32 megabytes (33,554,432
11147 bytes) from the current location can be called. This setting can be
11148 overridden by the @code{shortcall} function attribute, or by
11149 @code{#pragma longcall(0)}.
11151 Some linkers are capable of detecting out-of-range calls and generating
11152 glue code on the fly. On these systems, long calls are unnecessary and
11153 generate slower code. As of this writing, the AIX linker can do this,
11154 as can the GNU linker for PowerPC/64. It is planned to add this feature
11155 to the GNU linker for 32-bit PowerPC systems as well.
11157 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11158 callee, L42'', plus a ``branch island'' (glue code). The two target
11159 addresses represent the callee and the ``branch island''. The
11160 Darwin/PPC linker will prefer the first address and generate a ``bl
11161 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11162 otherwise, the linker will generate ``bl L42'' to call the ``branch
11163 island''. The ``branch island'' is appended to the body of the
11164 calling function; it computes the full 32-bit address of the callee
11167 On Mach-O (Darwin) systems, this option directs the compiler emit to
11168 the glue for every direct call, and the Darwin linker decides whether
11169 to use or discard it.
11171 In the future, we may cause GCC to ignore all longcall specifications
11172 when the linker is known to generate glue.
11176 Adds support for multithreading with the @dfn{pthreads} library.
11177 This option sets flags for both the preprocessor and linker.
11181 @node S/390 and zSeries Options
11182 @subsection S/390 and zSeries Options
11183 @cindex S/390 and zSeries Options
11185 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11189 @itemx -msoft-float
11190 @opindex mhard-float
11191 @opindex msoft-float
11192 Use (do not use) the hardware floating-point instructions and registers
11193 for floating-point operations. When @option{-msoft-float} is specified,
11194 functions in @file{libgcc.a} will be used to perform floating-point
11195 operations. When @option{-mhard-float} is specified, the compiler
11196 generates IEEE floating-point instructions. This is the default.
11199 @itemx -mno-backchain
11200 @opindex mbackchain
11201 @opindex mno-backchain
11202 Store (do not store) the address of the caller's frame as backchain pointer
11203 into the callee's stack frame.
11204 A backchain may be needed to allow debugging using tools that do not understand
11205 DWARF-2 call frame information.
11206 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
11207 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
11208 the backchain is placed into the topmost word of the 96/160 byte register
11211 In general, code compiled with @option{-mbackchain} is call-compatible with
11212 code compiled with @option{-mmo-backchain}; however, use of the backchain
11213 for debugging purposes usually requires that the whole binary is built with
11214 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
11215 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11216 to build a linux kernel use @option{-msoft-float}.
11218 The default is to not maintain the backchain.
11220 @item -mpacked-stack
11221 @item -mno-packed-stack
11222 @opindex mpacked-stack
11223 @opindex mno-packed-stack
11224 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
11225 specified, the compiler uses the all fields of the 96/160 byte register save
11226 area only for their default purpose; unused fields still take up stack space.
11227 When @option{-mpacked-stack} is specified, register save slots are densely
11228 packed at the top of the register save area; unused space is reused for other
11229 purposes, allowing for more efficient use of the available stack space.
11230 However, when @option{-mbackchain} is also in effect, the topmost word of
11231 the save area is always used to store the backchain, and the return address
11232 register is always saved two words below the backchain.
11234 As long as the stack frame backchain is not used, code generated with
11235 @option{-mpacked-stack} is call-compatible with code generated with
11236 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
11237 S/390 or zSeries generated code that uses the stack frame backchain at run
11238 time, not just for debugging purposes. Such code is not call-compatible
11239 with code compiled with @option{-mpacked-stack}. Also, note that the
11240 combination of @option{-mbackchain},
11241 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11242 to build a linux kernel use @option{-msoft-float}.
11244 The default is to not use the packed stack layout.
11247 @itemx -mno-small-exec
11248 @opindex msmall-exec
11249 @opindex mno-small-exec
11250 Generate (or do not generate) code using the @code{bras} instruction
11251 to do subroutine calls.
11252 This only works reliably if the total executable size does not
11253 exceed 64k. The default is to use the @code{basr} instruction instead,
11254 which does not have this limitation.
11260 When @option{-m31} is specified, generate code compliant to the
11261 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
11262 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
11263 particular to generate 64-bit instructions. For the @samp{s390}
11264 targets, the default is @option{-m31}, while the @samp{s390x}
11265 targets default to @option{-m64}.
11271 When @option{-mzarch} is specified, generate code using the
11272 instructions available on z/Architecture.
11273 When @option{-mesa} is specified, generate code using the
11274 instructions available on ESA/390. Note that @option{-mesa} is
11275 not possible with @option{-m64}.
11276 When generating code compliant to the GNU/Linux for S/390 ABI,
11277 the default is @option{-mesa}. When generating code compliant
11278 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
11284 Generate (or do not generate) code using the @code{mvcle} instruction
11285 to perform block moves. When @option{-mno-mvcle} is specified,
11286 use a @code{mvc} loop instead. This is the default.
11292 Print (or do not print) additional debug information when compiling.
11293 The default is to not print debug information.
11295 @item -march=@var{cpu-type}
11297 Generate code that will run on @var{cpu-type}, which is the name of a system
11298 representing a certain processor type. Possible values for
11299 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
11300 When generating code using the instructions available on z/Architecture,
11301 the default is @option{-march=z900}. Otherwise, the default is
11302 @option{-march=g5}.
11304 @item -mtune=@var{cpu-type}
11306 Tune to @var{cpu-type} everything applicable about the generated code,
11307 except for the ABI and the set of available instructions.
11308 The list of @var{cpu-type} values is the same as for @option{-march}.
11309 The default is the value used for @option{-march}.
11312 @itemx -mno-tpf-trace
11313 @opindex mtpf-trace
11314 @opindex mno-tpf-trace
11315 Generate code that adds (does not add) in TPF OS specific branches to trace
11316 routines in the operating system. This option is off by default, even
11317 when compiling for the TPF OS@.
11320 @itemx -mno-fused-madd
11321 @opindex mfused-madd
11322 @opindex mno-fused-madd
11323 Generate code that uses (does not use) the floating point multiply and
11324 accumulate instructions. These instructions are generated by default if
11325 hardware floating point is used.
11327 @item -mwarn-framesize=@var{framesize}
11328 @opindex mwarn-framesize
11329 Emit a warning if the current function exceeds the given frame size. Because
11330 this is a compile time check it doesn't need to be a real problem when the program
11331 runs. It is intended to identify functions which most probably cause
11332 a stack overflow. It is useful to be used in an environment with limited stack
11333 size e.g.@: the linux kernel.
11335 @item -mwarn-dynamicstack
11336 @opindex mwarn-dynamicstack
11337 Emit a warning if the function calls alloca or uses dynamically
11338 sized arrays. This is generally a bad idea with a limited stack size.
11340 @item -mstack-guard=@var{stack-guard}
11341 @item -mstack-size=@var{stack-size}
11342 @opindex mstack-guard
11343 @opindex mstack-size
11344 These arguments always have to be used in conjunction. If they are present the s390
11345 back end emits additional instructions in the function prologue which trigger a trap
11346 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
11347 (remember that the stack on s390 grows downward). These options are intended to
11348 be used to help debugging stack overflow problems. The additionally emitted code
11349 cause only little overhead and hence can also be used in production like systems
11350 without greater performance degradation. The given values have to be exact
11351 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard}.
11352 In order to be efficient the extra code makes the assumption that the stack starts
11353 at an address aligned to the value given by @var{stack-size}.
11357 @subsection SH Options
11359 These @samp{-m} options are defined for the SH implementations:
11364 Generate code for the SH1.
11368 Generate code for the SH2.
11371 Generate code for the SH2e.
11375 Generate code for the SH3.
11379 Generate code for the SH3e.
11383 Generate code for the SH4 without a floating-point unit.
11385 @item -m4-single-only
11386 @opindex m4-single-only
11387 Generate code for the SH4 with a floating-point unit that only
11388 supports single-precision arithmetic.
11392 Generate code for the SH4 assuming the floating-point unit is in
11393 single-precision mode by default.
11397 Generate code for the SH4.
11401 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
11402 floating-point unit is not used.
11404 @item -m4a-single-only
11405 @opindex m4a-single-only
11406 Generate code for the SH4a, in such a way that no double-precision
11407 floating point operations are used.
11410 @opindex m4a-single
11411 Generate code for the SH4a assuming the floating-point unit is in
11412 single-precision mode by default.
11416 Generate code for the SH4a.
11420 Same as @option{-m4a-nofpu}, except that it implicitly passes
11421 @option{-dsp} to the assembler. GCC doesn't generate any DSP
11422 instructions at the moment.
11426 Compile code for the processor in big endian mode.
11430 Compile code for the processor in little endian mode.
11434 Align doubles at 64-bit boundaries. Note that this changes the calling
11435 conventions, and thus some functions from the standard C library will
11436 not work unless you recompile it first with @option{-mdalign}.
11440 Shorten some address references at link time, when possible; uses the
11441 linker option @option{-relax}.
11445 Use 32-bit offsets in @code{switch} tables. The default is to use
11450 Enable the use of the instruction @code{fmovd}.
11454 Comply with the calling conventions defined by Renesas.
11458 Comply with the calling conventions defined by Renesas.
11462 Comply with the calling conventions defined for GCC before the Renesas
11463 conventions were available. This option is the default for all
11464 targets of the SH toolchain except for @samp{sh-symbianelf}.
11467 @opindex mnomacsave
11468 Mark the @code{MAC} register as call-clobbered, even if
11469 @option{-mhitachi} is given.
11473 Increase IEEE-compliance of floating-point code.
11474 At the moment, this is equivalent to @option{-fno-finite-math-only}.
11475 When generating 16 bit SH opcodes, getting IEEE-conforming results for
11476 comparisons of NANs / infinities incurs extra overhead in every
11477 floating point comparison, therefore the default is set to
11478 @option{-ffinite-math-only}.
11482 Dump instruction size and location in the assembly code.
11485 @opindex mpadstruct
11486 This option is deprecated. It pads structures to multiple of 4 bytes,
11487 which is incompatible with the SH ABI@.
11491 Optimize for space instead of speed. Implied by @option{-Os}.
11494 @opindex mprefergot
11495 When generating position-independent code, emit function calls using
11496 the Global Offset Table instead of the Procedure Linkage Table.
11500 Generate a library function call to invalidate instruction cache
11501 entries, after fixing up a trampoline. This library function call
11502 doesn't assume it can write to the whole memory address space. This
11503 is the default when the target is @code{sh-*-linux*}.
11505 @item -multcost=@var{number}
11506 @opindex multcost=@var{number}
11507 Set the cost to assume for a multiply insn.
11509 @item -mdiv=@var{strategy}
11510 @opindex mdiv=@var{strategy}
11511 Set the division strategy to use for SHmedia code. @var{strategy} must be
11512 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
11513 inv:call2, inv:fp .
11514 "fp" performs the operation in floating point. This has a very high latency,
11515 but needs only a few instructions, so it might be a good choice if
11516 your code has enough easily esploitable ILP to allow the compiler to
11517 schedule the floating point instructions together with other instructions.
11518 Division by zero causes a floating point exception.
11519 "inv" uses integer operations to calculate the inverse of the divisor,
11520 and then multiplies the divident with the inverse. This strategy allows
11521 cse and hoisting of the inverse calculation. Division by zero calculates
11522 an unspecified result, but does not trap.
11523 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
11524 have been found, or if the entire operation has been hoisted to the same
11525 place, the last stages of the inverse calculation are intertwined with the
11526 final multiply to reduce the overall latency, at the expense of using a few
11527 more instructions, and thus offering fewer scheduling opportunities with
11529 "call" calls a library function that usually implements the inv:minlat
11531 This gives high code density for m5-*media-nofpu compilations.
11532 "call2" uses a different entry point of the same library function, where it
11533 assumes that a pointer to a lookup table has already been set up, which
11534 exposes the pointer load to cse / code hoisting optimizations.
11535 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
11536 code generation, but if the code stays unoptimized, revert to the "call",
11537 "call2", or "fp" strategies, resspectively. Note that the
11538 potentially-trapping side effect of division by zero is carried by a
11539 separate instruction, so it is possible that all the integer instructions
11540 are hoisted out, but the marker for the side effect stays where it is.
11541 A recombination to fp operations or a call is not possible in that case.
11542 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
11543 that the inverse calculation was nor separated from the multiply, they speed
11544 up division where the dividend fits into 20 bits (plus sign where applicable),
11545 by inserting a test to skip a number of operations in this case; this test
11546 slows down the case of larger divdends. inv20u assumes the case of a such
11547 a small dividend to be unlikely, and inv20l assumes it to be likely.
11549 @item -mdivsi3_libfunc=@var{name}
11550 @opindex mdivsi3_libfunc=@var{name}
11551 Set the name of the library function used for 32 bit signed division to
11552 @var{name}. This only affect the name used in the call and inv:call
11553 division strategies, and the compiler will still expect the same
11554 sets of input/output/clobbered registers as if this option was not present.
11556 @item -madjust-unroll
11557 @opindex madjust-unroll
11558 Throttle unrolling to avoid thrashing target registers.
11559 This option only has an effect if the gcc code base supports the
11560 TARGET_ADJUST_UNROLL_MAX target hook.
11562 @item -mindexed-addressing
11563 @opindex mindexed-addressing
11564 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
11565 This is only safe if the hardware and/or OS implement 32 bit wrap-around
11566 semantics for the indexed addressing mode. The architecture allows the
11567 implementation of processors with 64 bit MMU, which the OS could use to
11568 get 32 bit addressing, but since no current harware implementation supports
11569 this or any other way to make the indexed addressing mode safe to use in
11570 the 32 bit ABI, the default is -mno-indexed-addressing.
11572 @item -mgettrcost=@var{number}
11573 @opindex mgettrcost=@var{number}
11574 Set the cost assumed for the gettr instruction to @var{number}.
11575 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
11579 Assume pt* instructions won't trap. This will generally generate better
11580 scheduled code, but is unsafe on current hardware. The current architecture
11581 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
11582 This has the unintentional effect of making it unsafe to schedule ptabs /
11583 ptrel before a branch, or hoist it out of a loop. For example,
11584 __do_global_ctors, a part of libgcc that runs constructors at program
11585 startup, calls functions in a list which is delimited by -1. With the
11586 -mpt-fixed option, the ptabs will be done before testing against -1.
11587 That means that all the constructors will be run a bit quicker, but when
11588 the loop comes to the end of the list, the pprogram crashes because ptabs
11589 loads -1 into a target register. Since this option is unsafe for any
11590 hardware implementing the current architecture specification, the default
11591 is -mno-pt-fixed. Unless the user specifies a specific cost with
11592 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
11593 this deters register allocation using target registers for storing
11596 @item -minvalid-symbols
11597 @opindex minvalid-symbols
11598 Assume symbols might be invalid. Ordinary function symbols generated by
11599 the compiler will always be valid to load with movi/shori/ptabs or
11600 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
11601 to generate symbols that will cause ptabs / ptrel to trap.
11602 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
11603 It will then prevent cross-basic-block cse, hoisting and most scheduling
11604 of symbol loads. The default is @option{-mno-invalid-symbols}.
11607 @node SPARC Options
11608 @subsection SPARC Options
11609 @cindex SPARC options
11611 These @samp{-m} options are supported on the SPARC:
11614 @item -mno-app-regs
11616 @opindex mno-app-regs
11618 Specify @option{-mapp-regs} to generate output using the global registers
11619 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
11620 is the default, except on Solaris.
11622 To be fully SVR4 ABI compliant at the cost of some performance loss,
11623 specify @option{-mno-app-regs}. You should compile libraries and system
11624 software with this option.
11627 @itemx -mhard-float
11629 @opindex mhard-float
11630 Generate output containing floating point instructions. This is the
11634 @itemx -msoft-float
11636 @opindex msoft-float
11637 Generate output containing library calls for floating point.
11638 @strong{Warning:} the requisite libraries are not available for all SPARC
11639 targets. Normally the facilities of the machine's usual C compiler are
11640 used, but this cannot be done directly in cross-compilation. You must make
11641 your own arrangements to provide suitable library functions for
11642 cross-compilation. The embedded targets @samp{sparc-*-aout} and
11643 @samp{sparclite-*-*} do provide software floating point support.
11645 @option{-msoft-float} changes the calling convention in the output file;
11646 therefore, it is only useful if you compile @emph{all} of a program with
11647 this option. In particular, you need to compile @file{libgcc.a}, the
11648 library that comes with GCC, with @option{-msoft-float} in order for
11651 @item -mhard-quad-float
11652 @opindex mhard-quad-float
11653 Generate output containing quad-word (long double) floating point
11656 @item -msoft-quad-float
11657 @opindex msoft-quad-float
11658 Generate output containing library calls for quad-word (long double)
11659 floating point instructions. The functions called are those specified
11660 in the SPARC ABI@. This is the default.
11662 As of this writing, there are no SPARC implementations that have hardware
11663 support for the quad-word floating point instructions. They all invoke
11664 a trap handler for one of these instructions, and then the trap handler
11665 emulates the effect of the instruction. Because of the trap handler overhead,
11666 this is much slower than calling the ABI library routines. Thus the
11667 @option{-msoft-quad-float} option is the default.
11669 @item -mno-unaligned-doubles
11670 @itemx -munaligned-doubles
11671 @opindex mno-unaligned-doubles
11672 @opindex munaligned-doubles
11673 Assume that doubles have 8 byte alignment. This is the default.
11675 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
11676 alignment only if they are contained in another type, or if they have an
11677 absolute address. Otherwise, it assumes they have 4 byte alignment.
11678 Specifying this option avoids some rare compatibility problems with code
11679 generated by other compilers. It is not the default because it results
11680 in a performance loss, especially for floating point code.
11682 @item -mno-faster-structs
11683 @itemx -mfaster-structs
11684 @opindex mno-faster-structs
11685 @opindex mfaster-structs
11686 With @option{-mfaster-structs}, the compiler assumes that structures
11687 should have 8 byte alignment. This enables the use of pairs of
11688 @code{ldd} and @code{std} instructions for copies in structure
11689 assignment, in place of twice as many @code{ld} and @code{st} pairs.
11690 However, the use of this changed alignment directly violates the SPARC
11691 ABI@. Thus, it's intended only for use on targets where the developer
11692 acknowledges that their resulting code will not be directly in line with
11693 the rules of the ABI@.
11695 @item -mimpure-text
11696 @opindex mimpure-text
11697 @option{-mimpure-text}, used in addition to @option{-shared}, tells
11698 the compiler to not pass @option{-z text} to the linker when linking a
11699 shared object. Using this option, you can link position-dependent
11700 code into a shared object.
11702 @option{-mimpure-text} suppresses the ``relocations remain against
11703 allocatable but non-writable sections'' linker error message.
11704 However, the necessary relocations will trigger copy-on-write, and the
11705 shared object is not actually shared across processes. Instead of
11706 using @option{-mimpure-text}, you should compile all source code with
11707 @option{-fpic} or @option{-fPIC}.
11709 This option is only available on SunOS and Solaris.
11711 @item -mcpu=@var{cpu_type}
11713 Set the instruction set, register set, and instruction scheduling parameters
11714 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
11715 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
11716 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
11717 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
11718 @samp{ultrasparc3}.
11720 Default instruction scheduling parameters are used for values that select
11721 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
11722 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
11724 Here is a list of each supported architecture and their supported
11729 v8: supersparc, hypersparc
11730 sparclite: f930, f934, sparclite86x
11732 v9: ultrasparc, ultrasparc3
11735 By default (unless configured otherwise), GCC generates code for the V7
11736 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
11737 additionally optimizes it for the Cypress CY7C602 chip, as used in the
11738 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
11739 SPARCStation 1, 2, IPX etc.
11741 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
11742 architecture. The only difference from V7 code is that the compiler emits
11743 the integer multiply and integer divide instructions which exist in SPARC-V8
11744 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
11745 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
11748 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
11749 the SPARC architecture. This adds the integer multiply, integer divide step
11750 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
11751 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
11752 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
11753 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
11754 MB86934 chip, which is the more recent SPARClite with FPU@.
11756 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
11757 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
11758 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
11759 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
11760 optimizes it for the TEMIC SPARClet chip.
11762 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
11763 architecture. This adds 64-bit integer and floating-point move instructions,
11764 3 additional floating-point condition code registers and conditional move
11765 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
11766 optimizes it for the Sun UltraSPARC I/II chips. With
11767 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
11768 Sun UltraSPARC III chip.
11770 @item -mtune=@var{cpu_type}
11772 Set the instruction scheduling parameters for machine type
11773 @var{cpu_type}, but do not set the instruction set or register set that the
11774 option @option{-mcpu=@var{cpu_type}} would.
11776 The same values for @option{-mcpu=@var{cpu_type}} can be used for
11777 @option{-mtune=@var{cpu_type}}, but the only useful values are those
11778 that select a particular cpu implementation. Those are @samp{cypress},
11779 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
11780 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
11781 @samp{ultrasparc3}.
11786 @opindex mno-v8plus
11787 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
11788 difference from the V8 ABI is that the global and out registers are
11789 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
11790 mode for all SPARC-V9 processors.
11796 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
11797 Visual Instruction Set extensions. The default is @option{-mno-vis}.
11800 These @samp{-m} options are supported in addition to the above
11801 on SPARC-V9 processors in 64-bit environments:
11804 @item -mlittle-endian
11805 @opindex mlittle-endian
11806 Generate code for a processor running in little-endian mode. It is only
11807 available for a few configurations and most notably not on Solaris and Linux.
11813 Generate code for a 32-bit or 64-bit environment.
11814 The 32-bit environment sets int, long and pointer to 32 bits.
11815 The 64-bit environment sets int to 32 bits and long and pointer
11818 @item -mcmodel=medlow
11819 @opindex mcmodel=medlow
11820 Generate code for the Medium/Low code model: 64-bit addresses, programs
11821 must be linked in the low 32 bits of memory. Programs can be statically
11822 or dynamically linked.
11824 @item -mcmodel=medmid
11825 @opindex mcmodel=medmid
11826 Generate code for the Medium/Middle code model: 64-bit addresses, programs
11827 must be linked in the low 44 bits of memory, the text and data segments must
11828 be less than 2GB in size and the data segment must be located within 2GB of
11831 @item -mcmodel=medany
11832 @opindex mcmodel=medany
11833 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
11834 may be linked anywhere in memory, the text and data segments must be less
11835 than 2GB in size and the data segment must be located within 2GB of the
11838 @item -mcmodel=embmedany
11839 @opindex mcmodel=embmedany
11840 Generate code for the Medium/Anywhere code model for embedded systems:
11841 64-bit addresses, the text and data segments must be less than 2GB in
11842 size, both starting anywhere in memory (determined at link time). The
11843 global register %g4 points to the base of the data segment. Programs
11844 are statically linked and PIC is not supported.
11847 @itemx -mno-stack-bias
11848 @opindex mstack-bias
11849 @opindex mno-stack-bias
11850 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
11851 frame pointer if present, are offset by @minus{}2047 which must be added back
11852 when making stack frame references. This is the default in 64-bit mode.
11853 Otherwise, assume no such offset is present.
11856 These switches are supported in addition to the above on Solaris:
11861 Add support for multithreading using the Solaris threads library. This
11862 option sets flags for both the preprocessor and linker. This option does
11863 not affect the thread safety of object code produced by the compiler or
11864 that of libraries supplied with it.
11868 Add support for multithreading using the POSIX threads library. This
11869 option sets flags for both the preprocessor and linker. This option does
11870 not affect the thread safety of object code produced by the compiler or
11871 that of libraries supplied with it.
11874 @node System V Options
11875 @subsection Options for System V
11877 These additional options are available on System V Release 4 for
11878 compatibility with other compilers on those systems:
11883 Create a shared object.
11884 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
11888 Identify the versions of each tool used by the compiler, in a
11889 @code{.ident} assembler directive in the output.
11893 Refrain from adding @code{.ident} directives to the output file (this is
11896 @item -YP,@var{dirs}
11898 Search the directories @var{dirs}, and no others, for libraries
11899 specified with @option{-l}.
11901 @item -Ym,@var{dir}
11903 Look in the directory @var{dir} to find the M4 preprocessor.
11904 The assembler uses this option.
11905 @c This is supposed to go with a -Yd for predefined M4 macro files, but
11906 @c the generic assembler that comes with Solaris takes just -Ym.
11909 @node TMS320C3x/C4x Options
11910 @subsection TMS320C3x/C4x Options
11911 @cindex TMS320C3x/C4x Options
11913 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
11917 @item -mcpu=@var{cpu_type}
11919 Set the instruction set, register set, and instruction scheduling
11920 parameters for machine type @var{cpu_type}. Supported values for
11921 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
11922 @samp{c44}. The default is @samp{c40} to generate code for the
11927 @itemx -msmall-memory
11929 @opindex mbig-memory
11931 @opindex msmall-memory
11933 Generates code for the big or small memory model. The small memory
11934 model assumed that all data fits into one 64K word page. At run-time
11935 the data page (DP) register must be set to point to the 64K page
11936 containing the .bss and .data program sections. The big memory model is
11937 the default and requires reloading of the DP register for every direct
11944 Allow (disallow) allocation of general integer operands into the block
11945 count register BK@.
11951 Enable (disable) generation of code using decrement and branch,
11952 DBcond(D), instructions. This is enabled by default for the C4x. To be
11953 on the safe side, this is disabled for the C3x, since the maximum
11954 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
11955 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
11956 that it can utilize the decrement and branch instruction, but will give
11957 up if there is more than one memory reference in the loop. Thus a loop
11958 where the loop counter is decremented can generate slightly more
11959 efficient code, in cases where the RPTB instruction cannot be utilized.
11961 @item -mdp-isr-reload
11963 @opindex mdp-isr-reload
11965 Force the DP register to be saved on entry to an interrupt service
11966 routine (ISR), reloaded to point to the data section, and restored on
11967 exit from the ISR@. This should not be required unless someone has
11968 violated the small memory model by modifying the DP register, say within
11975 For the C3x use the 24-bit MPYI instruction for integer multiplies
11976 instead of a library call to guarantee 32-bit results. Note that if one
11977 of the operands is a constant, then the multiplication will be performed
11978 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
11979 then squaring operations are performed inline instead of a library call.
11982 @itemx -mno-fast-fix
11984 @opindex mno-fast-fix
11985 The C3x/C4x FIX instruction to convert a floating point value to an
11986 integer value chooses the nearest integer less than or equal to the
11987 floating point value rather than to the nearest integer. Thus if the
11988 floating point number is negative, the result will be incorrectly
11989 truncated an additional code is necessary to detect and correct this
11990 case. This option can be used to disable generation of the additional
11991 code required to correct the result.
11997 Enable (disable) generation of repeat block sequences using the RPTB
11998 instruction for zero overhead looping. The RPTB construct is only used
11999 for innermost loops that do not call functions or jump across the loop
12000 boundaries. There is no advantage having nested RPTB loops due to the
12001 overhead required to save and restore the RC, RS, and RE registers.
12002 This is enabled by default with @option{-O2}.
12004 @item -mrpts=@var{count}
12008 Enable (disable) the use of the single instruction repeat instruction
12009 RPTS@. If a repeat block contains a single instruction, and the loop
12010 count can be guaranteed to be less than the value @var{count}, GCC will
12011 emit a RPTS instruction instead of a RPTB@. If no value is specified,
12012 then a RPTS will be emitted even if the loop count cannot be determined
12013 at compile time. Note that the repeated instruction following RPTS does
12014 not have to be reloaded from memory each iteration, thus freeing up the
12015 CPU buses for operands. However, since interrupts are blocked by this
12016 instruction, it is disabled by default.
12018 @item -mloop-unsigned
12019 @itemx -mno-loop-unsigned
12020 @opindex mloop-unsigned
12021 @opindex mno-loop-unsigned
12022 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
12023 is @math{2^{31} + 1} since these instructions test if the iteration count is
12024 negative to terminate the loop. If the iteration count is unsigned
12025 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
12026 exceeded. This switch allows an unsigned iteration count.
12030 Try to emit an assembler syntax that the TI assembler (asm30) is happy
12031 with. This also enforces compatibility with the API employed by the TI
12032 C3x C compiler. For example, long doubles are passed as structures
12033 rather than in floating point registers.
12039 Generate code that uses registers (stack) for passing arguments to functions.
12040 By default, arguments are passed in registers where possible rather
12041 than by pushing arguments on to the stack.
12043 @item -mparallel-insns
12044 @itemx -mno-parallel-insns
12045 @opindex mparallel-insns
12046 @opindex mno-parallel-insns
12047 Allow the generation of parallel instructions. This is enabled by
12048 default with @option{-O2}.
12050 @item -mparallel-mpy
12051 @itemx -mno-parallel-mpy
12052 @opindex mparallel-mpy
12053 @opindex mno-parallel-mpy
12054 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
12055 provided @option{-mparallel-insns} is also specified. These instructions have
12056 tight register constraints which can pessimize the code generation
12057 of large functions.
12062 @subsection V850 Options
12063 @cindex V850 Options
12065 These @samp{-m} options are defined for V850 implementations:
12069 @itemx -mno-long-calls
12070 @opindex mlong-calls
12071 @opindex mno-long-calls
12072 Treat all calls as being far away (near). If calls are assumed to be
12073 far away, the compiler will always load the functions address up into a
12074 register, and call indirect through the pointer.
12080 Do not optimize (do optimize) basic blocks that use the same index
12081 pointer 4 or more times to copy pointer into the @code{ep} register, and
12082 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
12083 option is on by default if you optimize.
12085 @item -mno-prolog-function
12086 @itemx -mprolog-function
12087 @opindex mno-prolog-function
12088 @opindex mprolog-function
12089 Do not use (do use) external functions to save and restore registers
12090 at the prologue and epilogue of a function. The external functions
12091 are slower, but use less code space if more than one function saves
12092 the same number of registers. The @option{-mprolog-function} option
12093 is on by default if you optimize.
12097 Try to make the code as small as possible. At present, this just turns
12098 on the @option{-mep} and @option{-mprolog-function} options.
12100 @item -mtda=@var{n}
12102 Put static or global variables whose size is @var{n} bytes or less into
12103 the tiny data area that register @code{ep} points to. The tiny data
12104 area can hold up to 256 bytes in total (128 bytes for byte references).
12106 @item -msda=@var{n}
12108 Put static or global variables whose size is @var{n} bytes or less into
12109 the small data area that register @code{gp} points to. The small data
12110 area can hold up to 64 kilobytes.
12112 @item -mzda=@var{n}
12114 Put static or global variables whose size is @var{n} bytes or less into
12115 the first 32 kilobytes of memory.
12119 Specify that the target processor is the V850.
12122 @opindex mbig-switch
12123 Generate code suitable for big switch tables. Use this option only if
12124 the assembler/linker complain about out of range branches within a switch
12129 This option will cause r2 and r5 to be used in the code generated by
12130 the compiler. This setting is the default.
12132 @item -mno-app-regs
12133 @opindex mno-app-regs
12134 This option will cause r2 and r5 to be treated as fixed registers.
12138 Specify that the target processor is the V850E1. The preprocessor
12139 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
12140 this option is used.
12144 Specify that the target processor is the V850E@. The preprocessor
12145 constant @samp{__v850e__} will be defined if this option is used.
12147 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
12148 are defined then a default target processor will be chosen and the
12149 relevant @samp{__v850*__} preprocessor constant will be defined.
12151 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
12152 defined, regardless of which processor variant is the target.
12154 @item -mdisable-callt
12155 @opindex mdisable-callt
12156 This option will suppress generation of the CALLT instruction for the
12157 v850e and v850e1 flavors of the v850 architecture. The default is
12158 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
12163 @subsection VAX Options
12164 @cindex VAX options
12166 These @samp{-m} options are defined for the VAX:
12171 Do not output certain jump instructions (@code{aobleq} and so on)
12172 that the Unix assembler for the VAX cannot handle across long
12177 Do output those jump instructions, on the assumption that you
12178 will assemble with the GNU assembler.
12182 Output code for g-format floating point numbers instead of d-format.
12185 @node x86-64 Options
12186 @subsection x86-64 Options
12187 @cindex x86-64 options
12189 These are listed under @xref{i386 and x86-64 Options}.
12191 @node Xstormy16 Options
12192 @subsection Xstormy16 Options
12193 @cindex Xstormy16 Options
12195 These options are defined for Xstormy16:
12200 Choose startup files and linker script suitable for the simulator.
12203 @node Xtensa Options
12204 @subsection Xtensa Options
12205 @cindex Xtensa Options
12207 These options are supported for Xtensa targets:
12211 @itemx -mno-const16
12213 @opindex mno-const16
12214 Enable or disable use of @code{CONST16} instructions for loading
12215 constant values. The @code{CONST16} instruction is currently not a
12216 standard option from Tensilica. When enabled, @code{CONST16}
12217 instructions are always used in place of the standard @code{L32R}
12218 instructions. The use of @code{CONST16} is enabled by default only if
12219 the @code{L32R} instruction is not available.
12222 @itemx -mno-fused-madd
12223 @opindex mfused-madd
12224 @opindex mno-fused-madd
12225 Enable or disable use of fused multiply/add and multiply/subtract
12226 instructions in the floating-point option. This has no effect if the
12227 floating-point option is not also enabled. Disabling fused multiply/add
12228 and multiply/subtract instructions forces the compiler to use separate
12229 instructions for the multiply and add/subtract operations. This may be
12230 desirable in some cases where strict IEEE 754-compliant results are
12231 required: the fused multiply add/subtract instructions do not round the
12232 intermediate result, thereby producing results with @emph{more} bits of
12233 precision than specified by the IEEE standard. Disabling fused multiply
12234 add/subtract instructions also ensures that the program output is not
12235 sensitive to the compiler's ability to combine multiply and add/subtract
12238 @item -mtext-section-literals
12239 @itemx -mno-text-section-literals
12240 @opindex mtext-section-literals
12241 @opindex mno-text-section-literals
12242 Control the treatment of literal pools. The default is
12243 @option{-mno-text-section-literals}, which places literals in a separate
12244 section in the output file. This allows the literal pool to be placed
12245 in a data RAM/ROM, and it also allows the linker to combine literal
12246 pools from separate object files to remove redundant literals and
12247 improve code size. With @option{-mtext-section-literals}, the literals
12248 are interspersed in the text section in order to keep them as close as
12249 possible to their references. This may be necessary for large assembly
12252 @item -mtarget-align
12253 @itemx -mno-target-align
12254 @opindex mtarget-align
12255 @opindex mno-target-align
12256 When this option is enabled, GCC instructs the assembler to
12257 automatically align instructions to reduce branch penalties at the
12258 expense of some code density. The assembler attempts to widen density
12259 instructions to align branch targets and the instructions following call
12260 instructions. If there are not enough preceding safe density
12261 instructions to align a target, no widening will be performed. The
12262 default is @option{-mtarget-align}. These options do not affect the
12263 treatment of auto-aligned instructions like @code{LOOP}, which the
12264 assembler will always align, either by widening density instructions or
12265 by inserting no-op instructions.
12268 @itemx -mno-longcalls
12269 @opindex mlongcalls
12270 @opindex mno-longcalls
12271 When this option is enabled, GCC instructs the assembler to translate
12272 direct calls to indirect calls unless it can determine that the target
12273 of a direct call is in the range allowed by the call instruction. This
12274 translation typically occurs for calls to functions in other source
12275 files. Specifically, the assembler translates a direct @code{CALL}
12276 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
12277 The default is @option{-mno-longcalls}. This option should be used in
12278 programs where the call target can potentially be out of range. This
12279 option is implemented in the assembler, not the compiler, so the
12280 assembly code generated by GCC will still show direct call
12281 instructions---look at the disassembled object code to see the actual
12282 instructions. Note that the assembler will use an indirect call for
12283 every cross-file call, not just those that really will be out of range.
12286 @node zSeries Options
12287 @subsection zSeries Options
12288 @cindex zSeries options
12290 These are listed under @xref{S/390 and zSeries Options}.
12292 @node Code Gen Options
12293 @section Options for Code Generation Conventions
12294 @cindex code generation conventions
12295 @cindex options, code generation
12296 @cindex run-time options
12298 These machine-independent options control the interface conventions
12299 used in code generation.
12301 Most of them have both positive and negative forms; the negative form
12302 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
12303 one of the forms is listed---the one which is not the default. You
12304 can figure out the other form by either removing @samp{no-} or adding
12308 @item -fbounds-check
12309 @opindex fbounds-check
12310 For front-ends that support it, generate additional code to check that
12311 indices used to access arrays are within the declared range. This is
12312 currently only supported by the Java and Fortran 77 front-ends, where
12313 this option defaults to true and false respectively.
12317 This option generates traps for signed overflow on addition, subtraction,
12318 multiplication operations.
12322 This option instructs the compiler to assume that signed arithmetic
12323 overflow of addition, subtraction and multiplication wraps around
12324 using twos-complement representation. This flag enables some optimizations
12325 and disables other. This option is enabled by default for the Java
12326 front-end, as required by the Java language specification.
12329 @opindex fexceptions
12330 Enable exception handling. Generates extra code needed to propagate
12331 exceptions. For some targets, this implies GCC will generate frame
12332 unwind information for all functions, which can produce significant data
12333 size overhead, although it does not affect execution. If you do not
12334 specify this option, GCC will enable it by default for languages like
12335 C++ which normally require exception handling, and disable it for
12336 languages like C that do not normally require it. However, you may need
12337 to enable this option when compiling C code that needs to interoperate
12338 properly with exception handlers written in C++. You may also wish to
12339 disable this option if you are compiling older C++ programs that don't
12340 use exception handling.
12342 @item -fnon-call-exceptions
12343 @opindex fnon-call-exceptions
12344 Generate code that allows trapping instructions to throw exceptions.
12345 Note that this requires platform-specific runtime support that does
12346 not exist everywhere. Moreover, it only allows @emph{trapping}
12347 instructions to throw exceptions, i.e.@: memory references or floating
12348 point instructions. It does not allow exceptions to be thrown from
12349 arbitrary signal handlers such as @code{SIGALRM}.
12351 @item -funwind-tables
12352 @opindex funwind-tables
12353 Similar to @option{-fexceptions}, except that it will just generate any needed
12354 static data, but will not affect the generated code in any other way.
12355 You will normally not enable this option; instead, a language processor
12356 that needs this handling would enable it on your behalf.
12358 @item -fasynchronous-unwind-tables
12359 @opindex fasynchronous-unwind-tables
12360 Generate unwind table in dwarf2 format, if supported by target machine. The
12361 table is exact at each instruction boundary, so it can be used for stack
12362 unwinding from asynchronous events (such as debugger or garbage collector).
12364 @item -fpcc-struct-return
12365 @opindex fpcc-struct-return
12366 Return ``short'' @code{struct} and @code{union} values in memory like
12367 longer ones, rather than in registers. This convention is less
12368 efficient, but it has the advantage of allowing intercallability between
12369 GCC-compiled files and files compiled with other compilers, particularly
12370 the Portable C Compiler (pcc).
12372 The precise convention for returning structures in memory depends
12373 on the target configuration macros.
12375 Short structures and unions are those whose size and alignment match
12376 that of some integer type.
12378 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12379 switch is not binary compatible with code compiled with the
12380 @option{-freg-struct-return} switch.
12381 Use it to conform to a non-default application binary interface.
12383 @item -freg-struct-return
12384 @opindex freg-struct-return
12385 Return @code{struct} and @code{union} values in registers when possible.
12386 This is more efficient for small structures than
12387 @option{-fpcc-struct-return}.
12389 If you specify neither @option{-fpcc-struct-return} nor
12390 @option{-freg-struct-return}, GCC defaults to whichever convention is
12391 standard for the target. If there is no standard convention, GCC
12392 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12393 the principal compiler. In those cases, we can choose the standard, and
12394 we chose the more efficient register return alternative.
12396 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12397 switch is not binary compatible with code compiled with the
12398 @option{-fpcc-struct-return} switch.
12399 Use it to conform to a non-default application binary interface.
12401 @item -fshort-enums
12402 @opindex fshort-enums
12403 Allocate to an @code{enum} type only as many bytes as it needs for the
12404 declared range of possible values. Specifically, the @code{enum} type
12405 will be equivalent to the smallest integer type which has enough room.
12407 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12408 code that is not binary compatible with code generated without that switch.
12409 Use it to conform to a non-default application binary interface.
12411 @item -fshort-double
12412 @opindex fshort-double
12413 Use the same size for @code{double} as for @code{float}.
12415 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
12416 code that is not binary compatible with code generated without that switch.
12417 Use it to conform to a non-default application binary interface.
12419 @item -fshort-wchar
12420 @opindex fshort-wchar
12421 Override the underlying type for @samp{wchar_t} to be @samp{short
12422 unsigned int} instead of the default for the target. This option is
12423 useful for building programs to run under WINE@.
12425 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12426 code that is not binary compatible with code generated without that switch.
12427 Use it to conform to a non-default application binary interface.
12429 @item -fshared-data
12430 @opindex fshared-data
12431 Requests that the data and non-@code{const} variables of this
12432 compilation be shared data rather than private data. The distinction
12433 makes sense only on certain operating systems, where shared data is
12434 shared between processes running the same program, while private data
12435 exists in one copy per process.
12438 @opindex fno-common
12439 In C, allocate even uninitialized global variables in the data section of the
12440 object file, rather than generating them as common blocks. This has the
12441 effect that if the same variable is declared (without @code{extern}) in
12442 two different compilations, you will get an error when you link them.
12443 The only reason this might be useful is if you wish to verify that the
12444 program will work on other systems which always work this way.
12448 Ignore the @samp{#ident} directive.
12450 @item -finhibit-size-directive
12451 @opindex finhibit-size-directive
12452 Don't output a @code{.size} assembler directive, or anything else that
12453 would cause trouble if the function is split in the middle, and the
12454 two halves are placed at locations far apart in memory. This option is
12455 used when compiling @file{crtstuff.c}; you should not need to use it
12458 @item -fverbose-asm
12459 @opindex fverbose-asm
12460 Put extra commentary information in the generated assembly code to
12461 make it more readable. This option is generally only of use to those
12462 who actually need to read the generated assembly code (perhaps while
12463 debugging the compiler itself).
12465 @option{-fno-verbose-asm}, the default, causes the
12466 extra information to be omitted and is useful when comparing two assembler
12471 @cindex global offset table
12473 Generate position-independent code (PIC) suitable for use in a shared
12474 library, if supported for the target machine. Such code accesses all
12475 constant addresses through a global offset table (GOT)@. The dynamic
12476 loader resolves the GOT entries when the program starts (the dynamic
12477 loader is not part of GCC; it is part of the operating system). If
12478 the GOT size for the linked executable exceeds a machine-specific
12479 maximum size, you get an error message from the linker indicating that
12480 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12481 instead. (These maximums are 8k on the SPARC and 32k
12482 on the m68k and RS/6000. The 386 has no such limit.)
12484 Position-independent code requires special support, and therefore works
12485 only on certain machines. For the 386, GCC supports PIC for System V
12486 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12487 position-independent.
12491 If supported for the target machine, emit position-independent code,
12492 suitable for dynamic linking and avoiding any limit on the size of the
12493 global offset table. This option makes a difference on the m68k,
12494 PowerPC and SPARC@.
12496 Position-independent code requires special support, and therefore works
12497 only on certain machines.
12503 These options are similar to @option{-fpic} and @option{-fPIC}, but
12504 generated position independent code can be only linked into executables.
12505 Usually these options are used when @option{-pie} GCC option will be
12506 used during linking.
12508 @item -ffixed-@var{reg}
12510 Treat the register named @var{reg} as a fixed register; generated code
12511 should never refer to it (except perhaps as a stack pointer, frame
12512 pointer or in some other fixed role).
12514 @var{reg} must be the name of a register. The register names accepted
12515 are machine-specific and are defined in the @code{REGISTER_NAMES}
12516 macro in the machine description macro file.
12518 This flag does not have a negative form, because it specifies a
12521 @item -fcall-used-@var{reg}
12522 @opindex fcall-used
12523 Treat the register named @var{reg} as an allocable register that is
12524 clobbered by function calls. It may be allocated for temporaries or
12525 variables that do not live across a call. Functions compiled this way
12526 will not save and restore the register @var{reg}.
12528 It is an error to used this flag with the frame pointer or stack pointer.
12529 Use of this flag for other registers that have fixed pervasive roles in
12530 the machine's execution model will produce disastrous results.
12532 This flag does not have a negative form, because it specifies a
12535 @item -fcall-saved-@var{reg}
12536 @opindex fcall-saved
12537 Treat the register named @var{reg} as an allocable register saved by
12538 functions. It may be allocated even for temporaries or variables that
12539 live across a call. Functions compiled this way will save and restore
12540 the register @var{reg} if they use it.
12542 It is an error to used this flag with the frame pointer or stack pointer.
12543 Use of this flag for other registers that have fixed pervasive roles in
12544 the machine's execution model will produce disastrous results.
12546 A different sort of disaster will result from the use of this flag for
12547 a register in which function values may be returned.
12549 This flag does not have a negative form, because it specifies a
12552 @item -fpack-struct[=@var{n}]
12553 @opindex fpack-struct
12554 Without a value specified, pack all structure members together without
12555 holes. When a value is specified (which must be a small power of two), pack
12556 structure members according to this value, representing the maximum
12557 alignment (that is, objects with default alignment requirements larger than
12558 this will be output potentially unaligned at the next fitting location.
12560 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12561 code that is not binary compatible with code generated without that switch.
12562 Additionally, it makes the code suboptimal.
12563 Use it to conform to a non-default application binary interface.
12565 @item -finstrument-functions
12566 @opindex finstrument-functions
12567 Generate instrumentation calls for entry and exit to functions. Just
12568 after function entry and just before function exit, the following
12569 profiling functions will be called with the address of the current
12570 function and its call site. (On some platforms,
12571 @code{__builtin_return_address} does not work beyond the current
12572 function, so the call site information may not be available to the
12573 profiling functions otherwise.)
12576 void __cyg_profile_func_enter (void *this_fn,
12578 void __cyg_profile_func_exit (void *this_fn,
12582 The first argument is the address of the start of the current function,
12583 which may be looked up exactly in the symbol table.
12585 This instrumentation is also done for functions expanded inline in other
12586 functions. The profiling calls will indicate where, conceptually, the
12587 inline function is entered and exited. This means that addressable
12588 versions of such functions must be available. If all your uses of a
12589 function are expanded inline, this may mean an additional expansion of
12590 code size. If you use @samp{extern inline} in your C code, an
12591 addressable version of such functions must be provided. (This is
12592 normally the case anyways, but if you get lucky and the optimizer always
12593 expands the functions inline, you might have gotten away without
12594 providing static copies.)
12596 A function may be given the attribute @code{no_instrument_function}, in
12597 which case this instrumentation will not be done. This can be used, for
12598 example, for the profiling functions listed above, high-priority
12599 interrupt routines, and any functions from which the profiling functions
12600 cannot safely be called (perhaps signal handlers, if the profiling
12601 routines generate output or allocate memory).
12603 @item -fstack-check
12604 @opindex fstack-check
12605 Generate code to verify that you do not go beyond the boundary of the
12606 stack. You should specify this flag if you are running in an
12607 environment with multiple threads, but only rarely need to specify it in
12608 a single-threaded environment since stack overflow is automatically
12609 detected on nearly all systems if there is only one stack.
12611 Note that this switch does not actually cause checking to be done; the
12612 operating system must do that. The switch causes generation of code
12613 to ensure that the operating system sees the stack being extended.
12615 @item -fstack-limit-register=@var{reg}
12616 @itemx -fstack-limit-symbol=@var{sym}
12617 @itemx -fno-stack-limit
12618 @opindex fstack-limit-register
12619 @opindex fstack-limit-symbol
12620 @opindex fno-stack-limit
12621 Generate code to ensure that the stack does not grow beyond a certain value,
12622 either the value of a register or the address of a symbol. If the stack
12623 would grow beyond the value, a signal is raised. For most targets,
12624 the signal is raised before the stack overruns the boundary, so
12625 it is possible to catch the signal without taking special precautions.
12627 For instance, if the stack starts at absolute address @samp{0x80000000}
12628 and grows downwards, you can use the flags
12629 @option{-fstack-limit-symbol=__stack_limit} and
12630 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12631 of 128KB@. Note that this may only work with the GNU linker.
12633 @cindex aliasing of parameters
12634 @cindex parameters, aliased
12635 @item -fargument-alias
12636 @itemx -fargument-noalias
12637 @itemx -fargument-noalias-global
12638 @opindex fargument-alias
12639 @opindex fargument-noalias
12640 @opindex fargument-noalias-global
12641 Specify the possible relationships among parameters and between
12642 parameters and global data.
12644 @option{-fargument-alias} specifies that arguments (parameters) may
12645 alias each other and may alias global storage.@*
12646 @option{-fargument-noalias} specifies that arguments do not alias
12647 each other, but may alias global storage.@*
12648 @option{-fargument-noalias-global} specifies that arguments do not
12649 alias each other and do not alias global storage.
12651 Each language will automatically use whatever option is required by
12652 the language standard. You should not need to use these options yourself.
12654 @item -fleading-underscore
12655 @opindex fleading-underscore
12656 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12657 change the way C symbols are represented in the object file. One use
12658 is to help link with legacy assembly code.
12660 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12661 generate code that is not binary compatible with code generated without that
12662 switch. Use it to conform to a non-default application binary interface.
12663 Not all targets provide complete support for this switch.
12665 @item -ftls-model=@var{model}
12666 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12667 The @var{model} argument should be one of @code{global-dynamic},
12668 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
12670 The default without @option{-fpic} is @code{initial-exec}; with
12671 @option{-fpic} the default is @code{global-dynamic}.
12673 @item -fvisibility=@var{default|internal|hidden|protected}
12674 @opindex fvisibility
12675 Set the default ELF image symbol visibility to the specified option---all
12676 symbols will be marked with this unless overridden within the code.
12677 Using this feature can very substantially improve linking and
12678 load times of shared object libraries, produce more optimized
12679 code, provide near-perfect API export and prevent symbol clashes.
12680 It is @strong{strongly} recommended that you use this in any shared objects
12683 Despite the nomenclature, @code{default} always means public ie;
12684 available to be linked against from outside the shared object.
12685 @code{protected} and @code{internal} are pretty useless in real-world
12686 usage so the only other commonly used option will be @code{hidden}.
12687 The default if @option{-fvisibility} isn't specified is
12688 @code{default}, i.e., make every
12689 symbol public---this causes the same behavior as previous versions of
12692 A good explanation of the benefits offered by ensuring ELF
12693 symbols have the correct visibility is given by ``How To Write
12694 Shared Libraries'' by Ulrich Drepper (which can be found at
12695 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
12696 solution made possible by this option to marking things hidden when
12697 the default is public is to make the default hidden and mark things
12698 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
12699 and @code{__attribute__ ((visibility("default")))} instead of
12700 @code{__declspec(dllexport)} you get almost identical semantics with
12701 identical syntax. This is a great boon to those working with
12702 cross-platform projects.
12704 For those adding visibility support to existing code, you may find
12705 @samp{#pragma GCC visibility} of use. This works by you enclosing
12706 the declarations you wish to set visibility for with (for example)
12707 @samp{#pragma GCC visibility push(hidden)} and
12708 @samp{#pragma GCC visibility pop}. These can be nested up to sixteen
12709 times. Bear in mind that symbol visibility should be viewed @strong{as
12710 part of the API interface contract} and thus all new code should
12711 always specify visibility when it is not the default ie; declarations
12712 only for use within the local DSO should @strong{always} be marked explicitly
12713 as hidden as so to avoid PLT indirection overheads---making this
12714 abundantly clear also aids readability and self-documentation of the code.
12715 Note that due to ISO C++ specification requirements, operator new and
12716 operator delete must always be of default visibility.
12718 An overview of these techniques, their benefits and how to use them
12719 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
12725 @node Environment Variables
12726 @section Environment Variables Affecting GCC
12727 @cindex environment variables
12729 @c man begin ENVIRONMENT
12730 This section describes several environment variables that affect how GCC
12731 operates. Some of them work by specifying directories or prefixes to use
12732 when searching for various kinds of files. Some are used to specify other
12733 aspects of the compilation environment.
12735 Note that you can also specify places to search using options such as
12736 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
12737 take precedence over places specified using environment variables, which
12738 in turn take precedence over those specified by the configuration of GCC@.
12739 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
12740 GNU Compiler Collection (GCC) Internals}.
12745 @c @itemx LC_COLLATE
12747 @c @itemx LC_MONETARY
12748 @c @itemx LC_NUMERIC
12753 @c @findex LC_COLLATE
12754 @findex LC_MESSAGES
12755 @c @findex LC_MONETARY
12756 @c @findex LC_NUMERIC
12760 These environment variables control the way that GCC uses
12761 localization information that allow GCC to work with different
12762 national conventions. GCC inspects the locale categories
12763 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
12764 so. These locale categories can be set to any value supported by your
12765 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
12766 Kingdom encoded in UTF-8.
12768 The @env{LC_CTYPE} environment variable specifies character
12769 classification. GCC uses it to determine the character boundaries in
12770 a string; this is needed for some multibyte encodings that contain quote
12771 and escape characters that would otherwise be interpreted as a string
12774 The @env{LC_MESSAGES} environment variable specifies the language to
12775 use in diagnostic messages.
12777 If the @env{LC_ALL} environment variable is set, it overrides the value
12778 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
12779 and @env{LC_MESSAGES} default to the value of the @env{LANG}
12780 environment variable. If none of these variables are set, GCC
12781 defaults to traditional C English behavior.
12785 If @env{TMPDIR} is set, it specifies the directory to use for temporary
12786 files. GCC uses temporary files to hold the output of one stage of
12787 compilation which is to be used as input to the next stage: for example,
12788 the output of the preprocessor, which is the input to the compiler
12791 @item GCC_EXEC_PREFIX
12792 @findex GCC_EXEC_PREFIX
12793 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
12794 names of the subprograms executed by the compiler. No slash is added
12795 when this prefix is combined with the name of a subprogram, but you can
12796 specify a prefix that ends with a slash if you wish.
12798 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
12799 an appropriate prefix to use based on the pathname it was invoked with.
12801 If GCC cannot find the subprogram using the specified prefix, it
12802 tries looking in the usual places for the subprogram.
12804 The default value of @env{GCC_EXEC_PREFIX} is
12805 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
12806 of @code{prefix} when you ran the @file{configure} script.
12808 Other prefixes specified with @option{-B} take precedence over this prefix.
12810 This prefix is also used for finding files such as @file{crt0.o} that are
12813 In addition, the prefix is used in an unusual way in finding the
12814 directories to search for header files. For each of the standard
12815 directories whose name normally begins with @samp{/usr/local/lib/gcc}
12816 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
12817 replacing that beginning with the specified prefix to produce an
12818 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
12819 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
12820 These alternate directories are searched first; the standard directories
12823 @item COMPILER_PATH
12824 @findex COMPILER_PATH
12825 The value of @env{COMPILER_PATH} is a colon-separated list of
12826 directories, much like @env{PATH}. GCC tries the directories thus
12827 specified when searching for subprograms, if it can't find the
12828 subprograms using @env{GCC_EXEC_PREFIX}.
12831 @findex LIBRARY_PATH
12832 The value of @env{LIBRARY_PATH} is a colon-separated list of
12833 directories, much like @env{PATH}. When configured as a native compiler,
12834 GCC tries the directories thus specified when searching for special
12835 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
12836 using GCC also uses these directories when searching for ordinary
12837 libraries for the @option{-l} option (but directories specified with
12838 @option{-L} come first).
12842 @cindex locale definition
12843 This variable is used to pass locale information to the compiler. One way in
12844 which this information is used is to determine the character set to be used
12845 when character literals, string literals and comments are parsed in C and C++.
12846 When the compiler is configured to allow multibyte characters,
12847 the following values for @env{LANG} are recognized:
12851 Recognize JIS characters.
12853 Recognize SJIS characters.
12855 Recognize EUCJP characters.
12858 If @env{LANG} is not defined, or if it has some other value, then the
12859 compiler will use mblen and mbtowc as defined by the default locale to
12860 recognize and translate multibyte characters.
12864 Some additional environments variables affect the behavior of the
12867 @include cppenv.texi
12871 @node Precompiled Headers
12872 @section Using Precompiled Headers
12873 @cindex precompiled headers
12874 @cindex speed of compilation
12876 Often large projects have many header files that are included in every
12877 source file. The time the compiler takes to process these header files
12878 over and over again can account for nearly all of the time required to
12879 build the project. To make builds faster, GCC allows users to
12880 `precompile' a header file; then, if builds can use the precompiled
12881 header file they will be much faster.
12883 @strong{Caution:} There are a few known situations where GCC will
12884 crash when trying to use a precompiled header. If you have trouble
12885 with a precompiled header, you should remove the precompiled header
12886 and compile without it. In addition, please use GCC's on-line
12887 defect-tracking system to report any problems you encounter with
12888 precompiled headers. @xref{Bugs}.
12890 To create a precompiled header file, simply compile it as you would any
12891 other file, if necessary using the @option{-x} option to make the driver
12892 treat it as a C or C++ header file. You will probably want to use a
12893 tool like @command{make} to keep the precompiled header up-to-date when
12894 the headers it contains change.
12896 A precompiled header file will be searched for when @code{#include} is
12897 seen in the compilation. As it searches for the included file
12898 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
12899 compiler looks for a precompiled header in each directory just before it
12900 looks for the include file in that directory. The name searched for is
12901 the name specified in the @code{#include} with @samp{.gch} appended. If
12902 the precompiled header file can't be used, it is ignored.
12904 For instance, if you have @code{#include "all.h"}, and you have
12905 @file{all.h.gch} in the same directory as @file{all.h}, then the
12906 precompiled header file will be used if possible, and the original
12907 header will be used otherwise.
12909 Alternatively, you might decide to put the precompiled header file in a
12910 directory and use @option{-I} to ensure that directory is searched
12911 before (or instead of) the directory containing the original header.
12912 Then, if you want to check that the precompiled header file is always
12913 used, you can put a file of the same name as the original header in this
12914 directory containing an @code{#error} command.
12916 This also works with @option{-include}. So yet another way to use
12917 precompiled headers, good for projects not designed with precompiled
12918 header files in mind, is to simply take most of the header files used by
12919 a project, include them from another header file, precompile that header
12920 file, and @option{-include} the precompiled header. If the header files
12921 have guards against multiple inclusion, they will be skipped because
12922 they've already been included (in the precompiled header).
12924 If you need to precompile the same header file for different
12925 languages, targets, or compiler options, you can instead make a
12926 @emph{directory} named like @file{all.h.gch}, and put each precompiled
12927 header in the directory, perhaps using @option{-o}. It doesn't matter
12928 what you call the files in the directory, every precompiled header in
12929 the directory will be considered. The first precompiled header
12930 encountered in the directory that is valid for this compilation will
12931 be used; they're searched in no particular order.
12933 There are many other possibilities, limited only by your imagination,
12934 good sense, and the constraints of your build system.
12936 A precompiled header file can be used only when these conditions apply:
12940 Only one precompiled header can be used in a particular compilation.
12943 A precompiled header can't be used once the first C token is seen. You
12944 can have preprocessor directives before a precompiled header; you can
12945 even include a precompiled header from inside another header, so long as
12946 there are no C tokens before the @code{#include}.
12949 The precompiled header file must be produced for the same language as
12950 the current compilation. You can't use a C precompiled header for a C++
12954 The precompiled header file must be produced by the same compiler
12955 version and configuration as the current compilation is using.
12956 The easiest way to guarantee this is to use the same compiler binary
12957 for creating and using precompiled headers.
12960 Any macros defined before the precompiled header is included must
12961 either be defined in the same way as when the precompiled header was
12962 generated, or must not affect the precompiled header, which usually
12963 means that they don't appear in the precompiled header at all.
12965 The @option{-D} option is one way to define a macro before a
12966 precompiled header is included; using a @code{#define} can also do it.
12967 There are also some options that define macros implicitly, like
12968 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
12971 @item If debugging information is output when using the precompiled
12972 header, using @option{-g} or similar, the same kind of debugging information
12973 must have been output when building the precompiled header. However,
12974 a precompiled header built using @option{-g} can be used in a compilation
12975 when no debugging information is being output.
12977 @item The same @option{-m} options must generally be used when building
12978 and using the precompiled header. @xref{Submodel Options},
12979 for any cases where this rule is relaxed.
12981 @item Each of the following options must be the same when building and using
12982 the precompiled header:
12984 @gccoptlist{-fexceptions -funit-at-a-time}
12987 Some other command-line options starting with @option{-f},
12988 @option{-p}, or @option{-O} must be defined in the same way as when
12989 the precompiled header was generated. At present, it's not clear
12990 which options are safe to change and which are not; the safest choice
12991 is to use exactly the same options when generating and using the
12992 precompiled header. The following are known to be safe:
12994 @gccoptlist{-fpreprocessed -pedantic-errors}
12998 For all of these except the last, the compiler will automatically
12999 ignore the precompiled header if the conditions aren't met. If you
13000 find an option combination that doesn't work and doesn't cause the
13001 precompiled header to be ignored, please consider filing a bug report,
13004 If you do use differing options when generating and using the
13005 precompiled header, the actual behavior will be a mixture of the
13006 behavior for the options. For instance, if you use @option{-g} to
13007 generate the precompiled header but not when using it, you may or may
13008 not get debugging information for routines in the precompiled header.
13010 @node Running Protoize
13011 @section Running Protoize
13013 The program @code{protoize} is an optional part of GCC@. You can use
13014 it to add prototypes to a program, thus converting the program to ISO
13015 C in one respect. The companion program @code{unprotoize} does the
13016 reverse: it removes argument types from any prototypes that are found.
13018 When you run these programs, you must specify a set of source files as
13019 command line arguments. The conversion programs start out by compiling
13020 these files to see what functions they define. The information gathered
13021 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
13023 After scanning comes actual conversion. The specified files are all
13024 eligible to be converted; any files they include (whether sources or
13025 just headers) are eligible as well.
13027 But not all the eligible files are converted. By default,
13028 @code{protoize} and @code{unprotoize} convert only source and header
13029 files in the current directory. You can specify additional directories
13030 whose files should be converted with the @option{-d @var{directory}}
13031 option. You can also specify particular files to exclude with the
13032 @option{-x @var{file}} option. A file is converted if it is eligible, its
13033 directory name matches one of the specified directory names, and its
13034 name within the directory has not been excluded.
13036 Basic conversion with @code{protoize} consists of rewriting most
13037 function definitions and function declarations to specify the types of
13038 the arguments. The only ones not rewritten are those for varargs
13041 @code{protoize} optionally inserts prototype declarations at the
13042 beginning of the source file, to make them available for any calls that
13043 precede the function's definition. Or it can insert prototype
13044 declarations with block scope in the blocks where undeclared functions
13047 Basic conversion with @code{unprotoize} consists of rewriting most
13048 function declarations to remove any argument types, and rewriting
13049 function definitions to the old-style pre-ISO form.
13051 Both conversion programs print a warning for any function declaration or
13052 definition that they can't convert. You can suppress these warnings
13055 The output from @code{protoize} or @code{unprotoize} replaces the
13056 original source file. The original file is renamed to a name ending
13057 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
13058 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
13059 for DOS) file already exists, then the source file is simply discarded.
13061 @code{protoize} and @code{unprotoize} both depend on GCC itself to
13062 scan the program and collect information about the functions it uses.
13063 So neither of these programs will work until GCC is installed.
13065 Here is a table of the options you can use with @code{protoize} and
13066 @code{unprotoize}. Each option works with both programs unless
13070 @item -B @var{directory}
13071 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
13072 usual directory (normally @file{/usr/local/lib}). This file contains
13073 prototype information about standard system functions. This option
13074 applies only to @code{protoize}.
13076 @item -c @var{compilation-options}
13077 Use @var{compilation-options} as the options when running @command{gcc} to
13078 produce the @samp{.X} files. The special option @option{-aux-info} is
13079 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
13081 Note that the compilation options must be given as a single argument to
13082 @code{protoize} or @code{unprotoize}. If you want to specify several
13083 @command{gcc} options, you must quote the entire set of compilation options
13084 to make them a single word in the shell.
13086 There are certain @command{gcc} arguments that you cannot use, because they
13087 would produce the wrong kind of output. These include @option{-g},
13088 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
13089 the @var{compilation-options}, they are ignored.
13092 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
13093 systems) instead of @samp{.c}. This is convenient if you are converting
13094 a C program to C++. This option applies only to @code{protoize}.
13097 Add explicit global declarations. This means inserting explicit
13098 declarations at the beginning of each source file for each function
13099 that is called in the file and was not declared. These declarations
13100 precede the first function definition that contains a call to an
13101 undeclared function. This option applies only to @code{protoize}.
13103 @item -i @var{string}
13104 Indent old-style parameter declarations with the string @var{string}.
13105 This option applies only to @code{protoize}.
13107 @code{unprotoize} converts prototyped function definitions to old-style
13108 function definitions, where the arguments are declared between the
13109 argument list and the initial @samp{@{}. By default, @code{unprotoize}
13110 uses five spaces as the indentation. If you want to indent with just
13111 one space instead, use @option{-i " "}.
13114 Keep the @samp{.X} files. Normally, they are deleted after conversion
13118 Add explicit local declarations. @code{protoize} with @option{-l} inserts
13119 a prototype declaration for each function in each block which calls the
13120 function without any declaration. This option applies only to
13124 Make no real changes. This mode just prints information about the conversions
13125 that would have been done without @option{-n}.
13128 Make no @samp{.save} files. The original files are simply deleted.
13129 Use this option with caution.
13131 @item -p @var{program}
13132 Use the program @var{program} as the compiler. Normally, the name
13133 @file{gcc} is used.
13136 Work quietly. Most warnings are suppressed.
13139 Print the version number, just like @option{-v} for @command{gcc}.
13142 If you need special compiler options to compile one of your program's
13143 source files, then you should generate that file's @samp{.X} file
13144 specially, by running @command{gcc} on that source file with the
13145 appropriate options and the option @option{-aux-info}. Then run
13146 @code{protoize} on the entire set of files. @code{protoize} will use
13147 the existing @samp{.X} file because it is newer than the source file.
13151 gcc -Dfoo=bar file1.c -aux-info file1.X
13156 You need to include the special files along with the rest in the
13157 @code{protoize} command, even though their @samp{.X} files already
13158 exist, because otherwise they won't get converted.
13160 @xref{Protoize Caveats}, for more information on how to use
13161 @code{protoize} successfully.