1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
8 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997,
9 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
11 Permission is granted to copy, distribute and/or modify this document
12 under the terms of the GNU Free Documentation License, Version 1.2 or
13 any later version published by the Free Software Foundation; with the
14 Invariant Sections being ``GNU General Public License'' and ``Funding
15 Free Software'', the Front-Cover texts being (a) (see below), and with
16 the Back-Cover Texts being (b) (see below). A copy of the license is
17 included in the gfdl(7) man page.
19 (a) The FSF's Front-Cover Text is:
23 (b) The FSF's Back-Cover Text is:
25 You have freedom to copy and modify this GNU Manual, like GNU
26 software. Copies published by the Free Software Foundation raise
27 funds for GNU development.
29 @c Set file name and title for the man page.
31 @settitle GNU project C and C++ compiler
33 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
34 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
35 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
36 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
37 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
38 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
39 [@option{-o} @var{outfile}] @var{infile}@dots{}
41 Only the most useful options are listed here; see below for the
42 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
45 gpl(7), gfdl(7), fsf-funding(7),
46 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
47 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
48 @file{ld}, @file{binutils} and @file{gdb}.
51 For instructions on reporting bugs, see
52 @w{@uref{http://gcc.gnu.org/bugs.html}}.
55 See the Info entry for @command{gcc}, or
56 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
57 for contributors to GCC@.
62 @chapter GCC Command Options
63 @cindex GCC command options
64 @cindex command options
65 @cindex options, GCC command
67 @c man begin DESCRIPTION
68 When you invoke GCC, it normally does preprocessing, compilation,
69 assembly and linking. The ``overall options'' allow you to stop this
70 process at an intermediate stage. For example, the @option{-c} option
71 says not to run the linker. Then the output consists of object files
72 output by the assembler.
74 Other options are passed on to one stage of processing. Some options
75 control the preprocessor and others the compiler itself. Yet other
76 options control the assembler and linker; most of these are not
77 documented here, since you rarely need to use any of them.
79 @cindex C compilation options
80 Most of the command line options that you can use with GCC are useful
81 for C programs; when an option is only useful with another language
82 (usually C++), the explanation says so explicitly. If the description
83 for a particular option does not mention a source language, you can use
84 that option with all supported languages.
86 @cindex C++ compilation options
87 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
88 options for compiling C++ programs.
90 @cindex grouping options
91 @cindex options, grouping
92 The @command{gcc} program accepts options and file names as operands. Many
93 options have multi-letter names; therefore multiple single-letter options
94 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
97 @cindex order of options
98 @cindex options, order
99 You can mix options and other arguments. For the most part, the order
100 you use doesn't matter. Order does matter when you use several options
101 of the same kind; for example, if you specify @option{-L} more than once,
102 the directories are searched in the order specified.
104 Many options have long names starting with @samp{-f} or with
105 @samp{-W}---for example, @option{-fforce-mem},
106 @option{-fstrength-reduce}, @option{-Wformat} and so on. Most of
107 these have both positive and negative forms; the negative form of
108 @option{-ffoo} would be @option{-fno-foo}. This manual documents
109 only one of these two forms, whichever one is not the default.
113 @xref{Option Index}, for an index to GCC's options.
116 * Option Summary:: Brief list of all options, without explanations.
117 * Overall Options:: Controlling the kind of output:
118 an executable, object files, assembler files,
119 or preprocessed source.
120 * Invoking G++:: Compiling C++ programs.
121 * C Dialect Options:: Controlling the variant of C language compiled.
122 * C++ Dialect Options:: Variations on C++.
123 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
125 * Language Independent Options:: Controlling how diagnostics should be
127 * Warning Options:: How picky should the compiler be?
128 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
129 * Optimize Options:: How much optimization?
130 * Preprocessor Options:: Controlling header files and macro definitions.
131 Also, getting dependency information for Make.
132 * Assembler Options:: Passing options to the assembler.
133 * Link Options:: Specifying libraries and so on.
134 * Directory Options:: Where to find header files and libraries.
135 Where to find the compiler executable files.
136 * Spec Files:: How to pass switches to sub-processes.
137 * Target Options:: Running a cross-compiler, or an old version of GCC.
138 * Submodel Options:: Specifying minor hardware or convention variations,
139 such as 68010 vs 68020.
140 * Code Gen Options:: Specifying conventions for function calls, data layout
142 * Environment Variables:: Env vars that affect GCC.
143 * Precompiled Headers:: Compiling a header once, and using it many times.
144 * Running Protoize:: Automatically adding or removing function prototypes.
150 @section Option Summary
152 Here is a summary of all the options, grouped by type. Explanations are
153 in the following sections.
156 @item Overall Options
157 @xref{Overall Options,,Options Controlling the Kind of Output}.
158 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
159 -x @var{language} -v -### --help --target-help --version}
161 @item C Language Options
162 @xref{C Dialect Options,,Options Controlling C Dialect}.
163 @gccoptlist{-ansi -std=@var{standard} -aux-info @var{filename} @gol
164 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
165 -fhosted -ffreestanding -fms-extensions @gol
166 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
167 -fallow-single-precision -fcond-mismatch @gol
168 -fsigned-bitfields -fsigned-char @gol
169 -funsigned-bitfields -funsigned-char}
171 @item C++ Language Options
172 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
173 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
174 -fconserve-space -fno-const-strings @gol
175 -fno-elide-constructors @gol
176 -fno-enforce-eh-specs @gol
177 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
178 -fno-implicit-templates @gol
179 -fno-implicit-inline-templates @gol
180 -fno-implement-inlines -fms-extensions @gol
181 -fno-nonansi-builtins -fno-operator-names @gol
182 -fno-optional-diags -fpermissive @gol
183 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
184 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
185 -fno-default-inline -fvisibility-inlines-hidden @gol
186 -Wabi -Wctor-dtor-privacy @gol
187 -Wnon-virtual-dtor -Wreorder @gol
188 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel @gol
189 -Wno-non-template-friend -Wold-style-cast @gol
190 -Woverloaded-virtual -Wno-pmf-conversions @gol
193 @item Objective-C and Objective-C++ Language Options
194 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
195 Objective-C and Objective-C++ Dialects}.
197 -fconstant-string-class=@var{class-name} @gol
198 -fgnu-runtime -fnext-runtime @gol
199 -fno-nil-receivers @gol
200 -fobjc-call-cxx-cdtors @gol
201 -fobjc-direct-dispatch @gol
202 -fobjc-exceptions @gol
204 -freplace-objc-classes @gol
207 -Wassign-intercept @gol
208 -Wno-protocol -Wselector @gol
209 -Wstrict-selector-match @gol
210 -Wundeclared-selector}
212 @item Language Independent Options
213 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
214 @gccoptlist{-fmessage-length=@var{n} @gol
215 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}} @gol
216 -fdiagnostics-show-options
218 @item Warning Options
219 @xref{Warning Options,,Options to Request or Suppress Warnings}.
220 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
221 -w -Wextra -Wall -Waggregate-return -Wno-attributes @gol
222 -Wc++-compat -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
223 -Wconversion -Wno-deprecated-declarations @gol
224 -Wdisabled-optimization -Wno-div-by-zero -Wno-endif-labels @gol
225 -Werror -Werror-implicit-function-declaration @gol
226 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
227 -Wno-format-extra-args -Wformat-nonliteral @gol
228 -Wformat-security -Wformat-y2k @gol
229 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
230 -Wimport -Wno-import -Winit-self -Winline @gol
231 -Wno-int-to-pointer-cast @gol
232 -Wno-invalid-offsetof -Winvalid-pch @gol
233 -Wlarger-than-@var{len} -Wlong-long @gol
234 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
235 -Wmissing-format-attribute -Wmissing-include-dirs @gol
236 -Wmissing-noreturn @gol
237 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
238 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
239 -Wredundant-decls @gol
240 -Wreturn-type -Wsequence-point -Wshadow @gol
241 -Wsign-compare -Wstrict-aliasing -Wstrict-aliasing=2 @gol
242 -Wswitch -Wswitch-default -Wswitch-enum @gol
243 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
244 -Wunknown-pragmas -Wunreachable-code @gol
245 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
246 -Wunused-value -Wunused-variable -Wwrite-strings @gol
249 @item C-only Warning Options
250 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
251 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
252 -Wstrict-prototypes -Wtraditional @gol
253 -Wdeclaration-after-statement -Wno-pointer-sign}
255 @item Debugging Options
256 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
257 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
258 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
259 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
260 -fdump-ipa-all -fdump-ipa-cgraph @gol
262 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
263 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
264 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
267 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
268 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
269 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
270 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
271 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
272 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
273 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
274 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
275 -fdump-tree-nrv -fdump-tree-vect @gol
276 -fdump-tree-sink @gol
277 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
278 -fdump-tree-salias @gol
279 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
280 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
281 -ftree-vectorizer-verbose=@var{n} @gol
282 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
283 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
284 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs -ftree-based-profiling @gol
285 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
286 -ftest-coverage -ftime-report -fvar-tracking @gol
287 -g -g@var{level} -gcoff -gdwarf-2 @gol
288 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
289 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
290 -print-multi-directory -print-multi-lib @gol
291 -print-prog-name=@var{program} -print-search-dirs -Q @gol
294 @item Optimization Options
295 @xref{Optimize Options,,Options that Control Optimization}.
296 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
297 -falign-labels=@var{n} -falign-loops=@var{n} @gol
298 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
299 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
300 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
301 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
302 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
303 -fdelayed-branch -fdelete-null-pointer-checks @gol
304 -fexpensive-optimizations -ffast-math -ffloat-store @gol
305 -fforce-addr -fforce-mem -ffunction-sections @gol
306 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
307 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2 @gol
308 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
309 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
310 -fmodulo-sched -fno-branch-count-reg @gol
311 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants @gol
312 -fno-function-cse -fno-guess-branch-probability @gol
313 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
314 -funsafe-math-optimizations -ffinite-math-only @gol
315 -fno-trapping-math -fno-zero-initialized-in-bss @gol
316 -fomit-frame-pointer -foptimize-register-move @gol
317 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
318 -fprofile-generate -fprofile-use @gol
319 -fregmove -frename-registers @gol
320 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
321 -frerun-cse-after-loop -frerun-loop-opt @gol
322 -frounding-math -fschedule-insns -fschedule-insns2 @gol
323 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
324 -fsched-spec-load-dangerous @gol
325 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
326 -fsched2-use-superblocks @gol
327 -fsched2-use-traces -freschedule-modulo-scheduled-loops @gol
328 -fsignaling-nans -fsingle-precision-constant -fspeculative-prefetching @gol
329 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
330 -funroll-all-loops -funroll-loops -fpeel-loops @gol
331 -fsplit-ivs-in-unroller -funswitch-loops @gol
332 -fvariable-expansion-in-unroller @gol
333 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
334 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
335 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
336 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
337 -ftree-salias -fweb @gol
338 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
339 --param @var{name}=@var{value}
340 -O -O0 -O1 -O2 -O3 -Os}
342 @item Preprocessor Options
343 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
344 @gccoptlist{-A@var{question}=@var{answer} @gol
345 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
346 -C -dD -dI -dM -dN @gol
347 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
348 -idirafter @var{dir} @gol
349 -include @var{file} -imacros @var{file} @gol
350 -iprefix @var{file} -iwithprefix @var{dir} @gol
351 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
352 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
353 -P -fworking-directory -remap @gol
354 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
355 -Xpreprocessor @var{option}}
357 @item Assembler Option
358 @xref{Assembler Options,,Passing Options to the Assembler}.
359 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
362 @xref{Link Options,,Options for Linking}.
363 @gccoptlist{@var{object-file-name} -l@var{library} @gol
364 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
365 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
366 -Wl,@var{option} -Xlinker @var{option} @gol
369 @item Directory Options
370 @xref{Directory Options,,Options for Directory Search}.
371 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-}
374 @c I wrote this xref this way to avoid overfull hbox. -- rms
375 @xref{Target Options}.
376 @gccoptlist{-V @var{version} -b @var{machine}}
378 @item Machine Dependent Options
379 @xref{Submodel Options,,Hardware Models and Configurations}.
380 @c This list is ordered alphanumerically by subsection name.
381 @c Try and put the significant identifier (CPU or system) first,
382 @c so users have a clue at guessing where the ones they want will be.
385 @gccoptlist{-EB -EL @gol
386 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
387 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
390 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
391 -mabi=@var{name} @gol
392 -mapcs-stack-check -mno-apcs-stack-check @gol
393 -mapcs-float -mno-apcs-float @gol
394 -mapcs-reentrant -mno-apcs-reentrant @gol
395 -msched-prolog -mno-sched-prolog @gol
396 -mlittle-endian -mbig-endian -mwords-little-endian @gol
397 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
398 -mthumb-interwork -mno-thumb-interwork @gol
399 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
400 -mstructure-size-boundary=@var{n} @gol
401 -mabort-on-noreturn @gol
402 -mlong-calls -mno-long-calls @gol
403 -msingle-pic-base -mno-single-pic-base @gol
404 -mpic-register=@var{reg} @gol
405 -mnop-fun-dllimport @gol
406 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
407 -mpoke-function-name @gol
409 -mtpcs-frame -mtpcs-leaf-frame @gol
410 -mcaller-super-interworking -mcallee-super-interworking}
413 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
414 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
416 @emph{Blackfin Options}
417 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer -mcsync @gol
418 -mno-csync -mlow-64k -mno-low64k -mid-shared-library @gol
419 -mno-id-shared-library -mshared-library-id=@var{n} @gol
420 -mlong-calls -mno-long-calls}
423 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
424 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
425 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
426 -mstack-align -mdata-align -mconst-align @gol
427 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
428 -melf -maout -melinux -mlinux -sim -sim2 @gol
429 -mmul-bug-workaround -mno-mul-bug-workaround}
431 @emph{Darwin Options}
432 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
433 -arch_only -bind_at_load -bundle -bundle_loader @gol
434 -client_name -compatibility_version -current_version @gol
436 -dependency-file -dylib_file -dylinker_install_name @gol
437 -dynamic -dynamiclib -exported_symbols_list @gol
438 -filelist -flat_namespace -force_cpusubtype_ALL @gol
439 -force_flat_namespace -headerpad_max_install_names @gol
440 -image_base -init -install_name -keep_private_externs @gol
441 -multi_module -multiply_defined -multiply_defined_unused @gol
442 -noall_load -no_dead_strip_inits_and_terms @gol
443 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
444 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
445 -private_bundle -read_only_relocs -sectalign @gol
446 -sectobjectsymbols -whyload -seg1addr @gol
447 -sectcreate -sectobjectsymbols -sectorder @gol
448 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
449 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
450 -segprot -segs_read_only_addr -segs_read_write_addr @gol
451 -single_module -static -sub_library -sub_umbrella @gol
452 -twolevel_namespace -umbrella -undefined @gol
453 -unexported_symbols_list -weak_reference_mismatches @gol
454 -whatsloaded -F -gused -gfull -mone-byte-bool}
456 @emph{DEC Alpha Options}
457 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
458 -mieee -mieee-with-inexact -mieee-conformant @gol
459 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
460 -mtrap-precision=@var{mode} -mbuild-constants @gol
461 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
462 -mbwx -mmax -mfix -mcix @gol
463 -mfloat-vax -mfloat-ieee @gol
464 -mexplicit-relocs -msmall-data -mlarge-data @gol
465 -msmall-text -mlarge-text @gol
466 -mmemory-latency=@var{time}}
468 @emph{DEC Alpha/VMS Options}
469 @gccoptlist{-mvms-return-codes}
472 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
473 -mhard-float -msoft-float @gol
474 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
475 -mdouble -mno-double @gol
476 -mmedia -mno-media -mmuladd -mno-muladd @gol
477 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
478 -mlinked-fp -mlong-calls -malign-labels @gol
479 -mlibrary-pic -macc-4 -macc-8 @gol
480 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
481 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
482 -mvliw-branch -mno-vliw-branch @gol
483 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
484 -mno-nested-cond-exec -mtomcat-stats @gol
488 @emph{H8/300 Options}
489 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
492 @gccoptlist{-march=@var{architecture-type} @gol
493 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
494 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
495 -mfixed-range=@var{register-range} @gol
496 -mjump-in-delay -mlinker-opt -mlong-calls @gol
497 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
498 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
499 -mno-jump-in-delay -mno-long-load-store @gol
500 -mno-portable-runtime -mno-soft-float @gol
501 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
502 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
503 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
504 -munix=@var{unix-std} -nolibdld -static -threads}
506 @emph{i386 and x86-64 Options}
507 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
508 -mfpmath=@var{unit} @gol
509 -masm=@var{dialect} -mno-fancy-math-387 @gol
510 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
511 -mno-wide-multiply -mrtd -malign-double @gol
512 -mpreferred-stack-boundary=@var{num} @gol
513 -mmmx -msse -msse2 -msse3 -m3dnow @gol
514 -mthreads -mno-align-stringops -minline-all-stringops @gol
515 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
516 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
517 -mno-red-zone -mno-tls-direct-seg-refs @gol
518 -mcmodel=@var{code-model} @gol
522 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
523 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
524 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
525 -minline-float-divide-max-throughput @gol
526 -minline-int-divide-min-latency @gol
527 -minline-int-divide-max-throughput @gol
528 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
529 -mno-dwarf2-asm -mearly-stop-bits @gol
530 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
531 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64}
533 @emph{M32R/D Options}
534 @gccoptlist{-m32r2 -m32rx -m32r @gol
536 -malign-loops -mno-align-loops @gol
537 -missue-rate=@var{number} @gol
538 -mbranch-cost=@var{number} @gol
539 -mmodel=@var{code-size-model-type} @gol
540 -msdata=@var{sdata-type} @gol
541 -mno-flush-func -mflush-func=@var{name} @gol
542 -mno-flush-trap -mflush-trap=@var{number} @gol
545 @emph{M680x0 Options}
546 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
547 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
548 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
549 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
550 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
552 @emph{M68hc1x Options}
553 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
554 -mauto-incdec -minmax -mlong-calls -mshort @gol
555 -msoft-reg-count=@var{count}}
558 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
559 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
560 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
561 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
562 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
565 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
566 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
567 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
568 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
569 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
570 -mpaired-single -mips3d @gol
571 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
572 -G@var{num} -membedded-data -mno-embedded-data @gol
573 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
574 -msplit-addresses -mno-split-addresses @gol
575 -mexplicit-relocs -mno-explicit-relocs @gol
576 -mcheck-zero-division -mno-check-zero-division @gol
577 -mdivide-traps -mdivide-breaks @gol
578 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
579 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
580 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
581 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
582 -mfix-sb1 -mno-fix-sb1 @gol
583 -mflush-func=@var{func} -mno-flush-func @gol
584 -mbranch-likely -mno-branch-likely @gol
585 -mfp-exceptions -mno-fp-exceptions @gol
586 -mvr4130-align -mno-vr4130-align}
589 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
590 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
591 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
592 -mno-base-addresses -msingle-exit -mno-single-exit}
594 @emph{MN10300 Options}
595 @gccoptlist{-mmult-bug -mno-mult-bug @gol
596 -mam33 -mno-am33 @gol
597 -mam33-2 -mno-am33-2 @gol
601 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
602 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
603 -mregparam -mnoregparam -msb -mnosb @gol
604 -mbitfield -mnobitfield -mhimem -mnohimem}
606 @emph{PDP-11 Options}
607 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
608 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
609 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
610 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
611 -mbranch-expensive -mbranch-cheap @gol
612 -msplit -mno-split -munix-asm -mdec-asm}
614 @emph{PowerPC Options}
615 See RS/6000 and PowerPC Options.
617 @emph{RS/6000 and PowerPC Options}
618 @gccoptlist{-mcpu=@var{cpu-type} @gol
619 -mtune=@var{cpu-type} @gol
620 -mpower -mno-power -mpower2 -mno-power2 @gol
621 -mpowerpc -mpowerpc64 -mno-powerpc @gol
622 -maltivec -mno-altivec @gol
623 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
624 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
625 -mnew-mnemonics -mold-mnemonics @gol
626 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
627 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
628 -malign-power -malign-natural @gol
629 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
630 -mstring -mno-string -mupdate -mno-update @gol
631 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
632 -mstrict-align -mno-strict-align -mrelocatable @gol
633 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
634 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
635 -mdynamic-no-pic -maltivec -mswdiv @gol
636 -mprioritize-restricted-insns=@var{priority} @gol
637 -msched-costly-dep=@var{dependence_type} @gol
638 -minsert-sched-nops=@var{scheme} @gol
639 -mcall-sysv -mcall-netbsd @gol
640 -maix-struct-return -msvr4-struct-return @gol
641 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
642 -misel -mno-isel @gol
643 -misel=yes -misel=no @gol
645 -mspe=yes -mspe=no @gol
646 -mvrsave -mno-vrsave @gol
647 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
648 -mprototype -mno-prototype @gol
649 -msim -mmvme -mads -myellowknife -memb -msdata @gol
650 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
652 @emph{S/390 and zSeries Options}
653 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
654 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
655 -mpacked-stack -mno-packed-stack @gol
656 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
657 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
658 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
659 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
662 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
663 -m4-nofpu -m4-single-only -m4-single -m4 @gol
664 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
665 -m5-64media -m5-64media-nofpu @gol
666 -m5-32media -m5-32media-nofpu @gol
667 -m5-compact -m5-compact-nofpu @gol
668 -mb -ml -mdalign -mrelax @gol
669 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
670 -mieee -misize -mpadstruct -mspace @gol
671 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
672 -mdivsi3_libfunc=@var{name} @gol
673 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
677 @gccoptlist{-mcpu=@var{cpu-type} @gol
678 -mtune=@var{cpu-type} @gol
679 -mcmodel=@var{code-model} @gol
680 -m32 -m64 -mapp-regs -mno-app-regs @gol
681 -mfaster-structs -mno-faster-structs @gol
682 -mfpu -mno-fpu -mhard-float -msoft-float @gol
683 -mhard-quad-float -msoft-quad-float @gol
684 -mimpure-text -mno-impure-text -mlittle-endian @gol
685 -mstack-bias -mno-stack-bias @gol
686 -munaligned-doubles -mno-unaligned-doubles @gol
687 -mv8plus -mno-v8plus -mvis -mno-vis
690 @emph{System V Options}
691 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
693 @emph{TMS320C3x/C4x Options}
694 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
695 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
696 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
697 -mparallel-insns -mparallel-mpy -mpreserve-float}
700 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
701 -mprolog-function -mno-prolog-function -mspace @gol
702 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
703 -mapp-regs -mno-app-regs @gol
704 -mdisable-callt -mno-disable-callt @gol
710 @gccoptlist{-mg -mgnu -munix}
712 @emph{x86-64 Options}
713 See i386 and x86-64 Options.
715 @emph{Xstormy16 Options}
718 @emph{Xtensa Options}
719 @gccoptlist{-mconst16 -mno-const16 @gol
720 -mfused-madd -mno-fused-madd @gol
721 -mtext-section-literals -mno-text-section-literals @gol
722 -mtarget-align -mno-target-align @gol
723 -mlongcalls -mno-longcalls}
725 @emph{zSeries Options}
726 See S/390 and zSeries Options.
728 @item Code Generation Options
729 @xref{Code Gen Options,,Options for Code Generation Conventions}.
730 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
731 -ffixed-@var{reg} -fexceptions @gol
732 -fnon-call-exceptions -funwind-tables @gol
733 -fasynchronous-unwind-tables @gol
734 -finhibit-size-directive -finstrument-functions @gol
735 -fno-common -fno-ident @gol
736 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
737 -fno-jump-tables @gol
738 -freg-struct-return -fshared-data -fshort-enums @gol
739 -fshort-double -fshort-wchar @gol
740 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
741 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
742 -fargument-alias -fargument-noalias @gol
743 -fargument-noalias-global -fleading-underscore @gol
744 -ftls-model=@var{model} @gol
745 -ftrapv -fwrapv -fbounds-check @gol
750 * Overall Options:: Controlling the kind of output:
751 an executable, object files, assembler files,
752 or preprocessed source.
753 * C Dialect Options:: Controlling the variant of C language compiled.
754 * C++ Dialect Options:: Variations on C++.
755 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
757 * Language Independent Options:: Controlling how diagnostics should be
759 * Warning Options:: How picky should the compiler be?
760 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
761 * Optimize Options:: How much optimization?
762 * Preprocessor Options:: Controlling header files and macro definitions.
763 Also, getting dependency information for Make.
764 * Assembler Options:: Passing options to the assembler.
765 * Link Options:: Specifying libraries and so on.
766 * Directory Options:: Where to find header files and libraries.
767 Where to find the compiler executable files.
768 * Spec Files:: How to pass switches to sub-processes.
769 * Target Options:: Running a cross-compiler, or an old version of GCC.
772 @node Overall Options
773 @section Options Controlling the Kind of Output
775 Compilation can involve up to four stages: preprocessing, compilation
776 proper, assembly and linking, always in that order. GCC is capable of
777 preprocessing and compiling several files either into several
778 assembler input files, or into one assembler input file; then each
779 assembler input file produces an object file, and linking combines all
780 the object files (those newly compiled, and those specified as input)
781 into an executable file.
783 @cindex file name suffix
784 For any given input file, the file name suffix determines what kind of
789 C source code which must be preprocessed.
792 C source code which should not be preprocessed.
795 C++ source code which should not be preprocessed.
798 Objective-C source code. Note that you must link with the @file{libobjc}
799 library to make an Objective-C program work.
802 Objective-C source code which should not be preprocessed.
806 Objective-C++ source code. Note that you must link with the @file{libobjc}
807 library to make an Objective-C++ program work. Note that @samp{.M} refers
808 to a literal capital M@.
811 Objective-C++ source code which should not be preprocessed.
814 C, C++, Objective-C or Objective-C++ header file to be turned into a
819 @itemx @var{file}.cxx
820 @itemx @var{file}.cpp
821 @itemx @var{file}.CPP
822 @itemx @var{file}.c++
824 C++ source code which must be preprocessed. Note that in @samp{.cxx},
825 the last two letters must both be literally @samp{x}. Likewise,
826 @samp{.C} refers to a literal capital C@.
830 Objective-C++ source code which must be preprocessed.
833 Objective-C++ source code which should not be preprocessed.
837 C++ header file to be turned into a precompiled header.
840 @itemx @var{file}.for
841 @itemx @var{file}.FOR
842 Fortran source code which should not be preprocessed.
845 @itemx @var{file}.fpp
846 @itemx @var{file}.FPP
847 Fortran source code which must be preprocessed (with the traditional
851 Fortran source code which must be preprocessed with a RATFOR
852 preprocessor (not included with GCC)@.
855 @itemx @var{file}.f95
856 Fortran 90/95 source code which should not be preprocessed.
858 @c FIXME: Descriptions of Java file types.
865 Ada source code file which contains a library unit declaration (a
866 declaration of a package, subprogram, or generic, or a generic
867 instantiation), or a library unit renaming declaration (a package,
868 generic, or subprogram renaming declaration). Such files are also
871 @itemx @var{file}.adb
872 Ada source code file containing a library unit body (a subprogram or
873 package body). Such files are also called @dfn{bodies}.
875 @c GCC also knows about some suffixes for languages not yet included:
884 Assembler code which must be preprocessed.
887 An object file to be fed straight into linking.
888 Any file name with no recognized suffix is treated this way.
892 You can specify the input language explicitly with the @option{-x} option:
895 @item -x @var{language}
896 Specify explicitly the @var{language} for the following input files
897 (rather than letting the compiler choose a default based on the file
898 name suffix). This option applies to all following input files until
899 the next @option{-x} option. Possible values for @var{language} are:
901 c c-header c-cpp-output
902 c++ c++-header c++-cpp-output
903 objective-c objective-c-header objective-c-cpp-output
904 objective-c++ objective-c++-header objective-c++-cpp-output
905 assembler assembler-with-cpp
907 f77 f77-cpp-input ratfor
914 Turn off any specification of a language, so that subsequent files are
915 handled according to their file name suffixes (as they are if @option{-x}
916 has not been used at all).
918 @item -pass-exit-codes
919 @opindex pass-exit-codes
920 Normally the @command{gcc} program will exit with the code of 1 if any
921 phase of the compiler returns a non-success return code. If you specify
922 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
923 numerically highest error produced by any phase that returned an error
927 If you only want some of the stages of compilation, you can use
928 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
929 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
930 @command{gcc} is to stop. Note that some combinations (for example,
931 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
936 Compile or assemble the source files, but do not link. The linking
937 stage simply is not done. The ultimate output is in the form of an
938 object file for each source file.
940 By default, the object file name for a source file is made by replacing
941 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
943 Unrecognized input files, not requiring compilation or assembly, are
948 Stop after the stage of compilation proper; do not assemble. The output
949 is in the form of an assembler code file for each non-assembler input
952 By default, the assembler file name for a source file is made by
953 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
955 Input files that don't require compilation are ignored.
959 Stop after the preprocessing stage; do not run the compiler proper. The
960 output is in the form of preprocessed source code, which is sent to the
963 Input files which don't require preprocessing are ignored.
965 @cindex output file option
968 Place output in file @var{file}. This applies regardless to whatever
969 sort of output is being produced, whether it be an executable file,
970 an object file, an assembler file or preprocessed C code.
972 If @option{-o} is not specified, the default is to put an executable
973 file in @file{a.out}, the object file for
974 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
975 assembler file in @file{@var{source}.s}, a precompiled header file in
976 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
981 Print (on standard error output) the commands executed to run the stages
982 of compilation. Also print the version number of the compiler driver
983 program and of the preprocessor and the compiler proper.
987 Like @option{-v} except the commands are not executed and all command
988 arguments are quoted. This is useful for shell scripts to capture the
989 driver-generated command lines.
993 Use pipes rather than temporary files for communication between the
994 various stages of compilation. This fails to work on some systems where
995 the assembler is unable to read from a pipe; but the GNU assembler has
1000 If you are compiling multiple source files, this option tells the driver
1001 to pass all the source files to the compiler at once (for those
1002 languages for which the compiler can handle this). This will allow
1003 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1004 language for which this is supported is C@. If you pass source files for
1005 multiple languages to the driver, using this option, the driver will invoke
1006 the compiler(s) that support IMA once each, passing each compiler all the
1007 source files appropriate for it. For those languages that do not support
1008 IMA this option will be ignored, and the compiler will be invoked once for
1009 each source file in that language. If you use this option in conjunction
1010 with @option{-save-temps}, the compiler will generate multiple
1012 (one for each source file), but only one (combined) @file{.o} or
1017 Print (on the standard output) a description of the command line options
1018 understood by @command{gcc}. If the @option{-v} option is also specified
1019 then @option{--help} will also be passed on to the various processes
1020 invoked by @command{gcc}, so that they can display the command line options
1021 they accept. If the @option{-Wextra} option is also specified then command
1022 line options which have no documentation associated with them will also
1026 @opindex target-help
1027 Print (on the standard output) a description of target specific command
1028 line options for each tool.
1032 Display the version number and copyrights of the invoked GCC@.
1036 @section Compiling C++ Programs
1038 @cindex suffixes for C++ source
1039 @cindex C++ source file suffixes
1040 C++ source files conventionally use one of the suffixes @samp{.C},
1041 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1042 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1043 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1044 files with these names and compiles them as C++ programs even if you
1045 call the compiler the same way as for compiling C programs (usually
1046 with the name @command{gcc}).
1050 However, C++ programs often require class libraries as well as a
1051 compiler that understands the C++ language---and under some
1052 circumstances, you might want to compile programs or header files from
1053 standard input, or otherwise without a suffix that flags them as C++
1054 programs. You might also like to precompile a C header file with a
1055 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
1056 program that calls GCC with the default language set to C++, and
1057 automatically specifies linking against the C++ library. On many
1058 systems, @command{g++} is also installed with the name @command{c++}.
1060 @cindex invoking @command{g++}
1061 When you compile C++ programs, you may specify many of the same
1062 command-line options that you use for compiling programs in any
1063 language; or command-line options meaningful for C and related
1064 languages; or options that are meaningful only for C++ programs.
1065 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1066 explanations of options for languages related to C@.
1067 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1068 explanations of options that are meaningful only for C++ programs.
1070 @node C Dialect Options
1071 @section Options Controlling C Dialect
1072 @cindex dialect options
1073 @cindex language dialect options
1074 @cindex options, dialect
1076 The following options control the dialect of C (or languages derived
1077 from C, such as C++, Objective-C and Objective-C++) that the compiler
1081 @cindex ANSI support
1085 In C mode, support all ISO C90 programs. In C++ mode,
1086 remove GNU extensions that conflict with ISO C++.
1088 This turns off certain features of GCC that are incompatible with ISO
1089 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1090 such as the @code{asm} and @code{typeof} keywords, and
1091 predefined macros such as @code{unix} and @code{vax} that identify the
1092 type of system you are using. It also enables the undesirable and
1093 rarely used ISO trigraph feature. For the C compiler,
1094 it disables recognition of C++ style @samp{//} comments as well as
1095 the @code{inline} keyword.
1097 The alternate keywords @code{__asm__}, @code{__extension__},
1098 @code{__inline__} and @code{__typeof__} continue to work despite
1099 @option{-ansi}. You would not want to use them in an ISO C program, of
1100 course, but it is useful to put them in header files that might be included
1101 in compilations done with @option{-ansi}. Alternate predefined macros
1102 such as @code{__unix__} and @code{__vax__} are also available, with or
1103 without @option{-ansi}.
1105 The @option{-ansi} option does not cause non-ISO programs to be
1106 rejected gratuitously. For that, @option{-pedantic} is required in
1107 addition to @option{-ansi}. @xref{Warning Options}.
1109 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1110 option is used. Some header files may notice this macro and refrain
1111 from declaring certain functions or defining certain macros that the
1112 ISO standard doesn't call for; this is to avoid interfering with any
1113 programs that might use these names for other things.
1115 Functions which would normally be built in but do not have semantics
1116 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1117 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1118 built-in functions provided by GCC}, for details of the functions
1123 Determine the language standard. This option is currently only
1124 supported when compiling C or C++. A value for this option must be
1125 provided; possible values are
1130 ISO C90 (same as @option{-ansi}).
1132 @item iso9899:199409
1133 ISO C90 as modified in amendment 1.
1139 ISO C99. Note that this standard is not yet fully supported; see
1140 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1141 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1144 Default, ISO C90 plus GNU extensions (including some C99 features).
1148 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1149 this will become the default. The name @samp{gnu9x} is deprecated.
1152 The 1998 ISO C++ standard plus amendments.
1155 The same as @option{-std=c++98} plus GNU extensions. This is the
1156 default for C++ code.
1159 Even when this option is not specified, you can still use some of the
1160 features of newer standards in so far as they do not conflict with
1161 previous C standards. For example, you may use @code{__restrict__} even
1162 when @option{-std=c99} is not specified.
1164 The @option{-std} options specifying some version of ISO C have the same
1165 effects as @option{-ansi}, except that features that were not in ISO C90
1166 but are in the specified version (for example, @samp{//} comments and
1167 the @code{inline} keyword in ISO C99) are not disabled.
1169 @xref{Standards,,Language Standards Supported by GCC}, for details of
1170 these standard versions.
1172 @item -aux-info @var{filename}
1174 Output to the given filename prototyped declarations for all functions
1175 declared and/or defined in a translation unit, including those in header
1176 files. This option is silently ignored in any language other than C@.
1178 Besides declarations, the file indicates, in comments, the origin of
1179 each declaration (source file and line), whether the declaration was
1180 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1181 @samp{O} for old, respectively, in the first character after the line
1182 number and the colon), and whether it came from a declaration or a
1183 definition (@samp{C} or @samp{F}, respectively, in the following
1184 character). In the case of function definitions, a K&R-style list of
1185 arguments followed by their declarations is also provided, inside
1186 comments, after the declaration.
1190 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1191 keyword, so that code can use these words as identifiers. You can use
1192 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1193 instead. @option{-ansi} implies @option{-fno-asm}.
1195 In C++, this switch only affects the @code{typeof} keyword, since
1196 @code{asm} and @code{inline} are standard keywords. You may want to
1197 use the @option{-fno-gnu-keywords} flag instead, which has the same
1198 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1199 switch only affects the @code{asm} and @code{typeof} keywords, since
1200 @code{inline} is a standard keyword in ISO C99.
1203 @itemx -fno-builtin-@var{function}
1204 @opindex fno-builtin
1205 @cindex built-in functions
1206 Don't recognize built-in functions that do not begin with
1207 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1208 functions provided by GCC}, for details of the functions affected,
1209 including those which are not built-in functions when @option{-ansi} or
1210 @option{-std} options for strict ISO C conformance are used because they
1211 do not have an ISO standard meaning.
1213 GCC normally generates special code to handle certain built-in functions
1214 more efficiently; for instance, calls to @code{alloca} may become single
1215 instructions that adjust the stack directly, and calls to @code{memcpy}
1216 may become inline copy loops. The resulting code is often both smaller
1217 and faster, but since the function calls no longer appear as such, you
1218 cannot set a breakpoint on those calls, nor can you change the behavior
1219 of the functions by linking with a different library. In addition,
1220 when a function is recognized as a built-in function, GCC may use
1221 information about that function to warn about problems with calls to
1222 that function, or to generate more efficient code, even if the
1223 resulting code still contains calls to that function. For example,
1224 warnings are given with @option{-Wformat} for bad calls to
1225 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1226 known not to modify global memory.
1228 With the @option{-fno-builtin-@var{function}} option
1229 only the built-in function @var{function} is
1230 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1231 function is named this is not built-in in this version of GCC, this
1232 option is ignored. There is no corresponding
1233 @option{-fbuiltin-@var{function}} option; if you wish to enable
1234 built-in functions selectively when using @option{-fno-builtin} or
1235 @option{-ffreestanding}, you may define macros such as:
1238 #define abs(n) __builtin_abs ((n))
1239 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1244 @cindex hosted environment
1246 Assert that compilation takes place in a hosted environment. This implies
1247 @option{-fbuiltin}. A hosted environment is one in which the
1248 entire standard library is available, and in which @code{main} has a return
1249 type of @code{int}. Examples are nearly everything except a kernel.
1250 This is equivalent to @option{-fno-freestanding}.
1252 @item -ffreestanding
1253 @opindex ffreestanding
1254 @cindex hosted environment
1256 Assert that compilation takes place in a freestanding environment. This
1257 implies @option{-fno-builtin}. A freestanding environment
1258 is one in which the standard library may not exist, and program startup may
1259 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1260 This is equivalent to @option{-fno-hosted}.
1262 @xref{Standards,,Language Standards Supported by GCC}, for details of
1263 freestanding and hosted environments.
1265 @item -fms-extensions
1266 @opindex fms-extensions
1267 Accept some non-standard constructs used in Microsoft header files.
1269 Some cases of unnamed fields in structures and unions are only
1270 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1271 fields within structs/unions}, for details.
1275 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1276 options for strict ISO C conformance) implies @option{-trigraphs}.
1278 @item -no-integrated-cpp
1279 @opindex no-integrated-cpp
1280 Performs a compilation in two passes: preprocessing and compiling. This
1281 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1282 @option{-B} option. The user supplied compilation step can then add in
1283 an additional preprocessing step after normal preprocessing but before
1284 compiling. The default is to use the integrated cpp (internal cpp)
1286 The semantics of this option will change if "cc1", "cc1plus", and
1287 "cc1obj" are merged.
1289 @cindex traditional C language
1290 @cindex C language, traditional
1292 @itemx -traditional-cpp
1293 @opindex traditional-cpp
1294 @opindex traditional
1295 Formerly, these options caused GCC to attempt to emulate a pre-standard
1296 C compiler. They are now only supported with the @option{-E} switch.
1297 The preprocessor continues to support a pre-standard mode. See the GNU
1298 CPP manual for details.
1300 @item -fcond-mismatch
1301 @opindex fcond-mismatch
1302 Allow conditional expressions with mismatched types in the second and
1303 third arguments. The value of such an expression is void. This option
1304 is not supported for C++.
1306 @item -funsigned-char
1307 @opindex funsigned-char
1308 Let the type @code{char} be unsigned, like @code{unsigned char}.
1310 Each kind of machine has a default for what @code{char} should
1311 be. It is either like @code{unsigned char} by default or like
1312 @code{signed char} by default.
1314 Ideally, a portable program should always use @code{signed char} or
1315 @code{unsigned char} when it depends on the signedness of an object.
1316 But many programs have been written to use plain @code{char} and
1317 expect it to be signed, or expect it to be unsigned, depending on the
1318 machines they were written for. This option, and its inverse, let you
1319 make such a program work with the opposite default.
1321 The type @code{char} is always a distinct type from each of
1322 @code{signed char} or @code{unsigned char}, even though its behavior
1323 is always just like one of those two.
1326 @opindex fsigned-char
1327 Let the type @code{char} be signed, like @code{signed char}.
1329 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1330 the negative form of @option{-funsigned-char}. Likewise, the option
1331 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1333 @item -fsigned-bitfields
1334 @itemx -funsigned-bitfields
1335 @itemx -fno-signed-bitfields
1336 @itemx -fno-unsigned-bitfields
1337 @opindex fsigned-bitfields
1338 @opindex funsigned-bitfields
1339 @opindex fno-signed-bitfields
1340 @opindex fno-unsigned-bitfields
1341 These options control whether a bit-field is signed or unsigned, when the
1342 declaration does not use either @code{signed} or @code{unsigned}. By
1343 default, such a bit-field is signed, because this is consistent: the
1344 basic integer types such as @code{int} are signed types.
1347 @node C++ Dialect Options
1348 @section Options Controlling C++ Dialect
1350 @cindex compiler options, C++
1351 @cindex C++ options, command line
1352 @cindex options, C++
1353 This section describes the command-line options that are only meaningful
1354 for C++ programs; but you can also use most of the GNU compiler options
1355 regardless of what language your program is in. For example, you
1356 might compile a file @code{firstClass.C} like this:
1359 g++ -g -frepo -O -c firstClass.C
1363 In this example, only @option{-frepo} is an option meant
1364 only for C++ programs; you can use the other options with any
1365 language supported by GCC@.
1367 Here is a list of options that are @emph{only} for compiling C++ programs:
1371 @item -fabi-version=@var{n}
1372 @opindex fabi-version
1373 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1374 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1375 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1376 the version that conforms most closely to the C++ ABI specification.
1377 Therefore, the ABI obtained using version 0 will change as ABI bugs
1380 The default is version 2.
1382 @item -fno-access-control
1383 @opindex fno-access-control
1384 Turn off all access checking. This switch is mainly useful for working
1385 around bugs in the access control code.
1389 Check that the pointer returned by @code{operator new} is non-null
1390 before attempting to modify the storage allocated. This check is
1391 normally unnecessary because the C++ standard specifies that
1392 @code{operator new} will only return @code{0} if it is declared
1393 @samp{throw()}, in which case the compiler will always check the
1394 return value even without this option. In all other cases, when
1395 @code{operator new} has a non-empty exception specification, memory
1396 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1397 @samp{new (nothrow)}.
1399 @item -fconserve-space
1400 @opindex fconserve-space
1401 Put uninitialized or runtime-initialized global variables into the
1402 common segment, as C does. This saves space in the executable at the
1403 cost of not diagnosing duplicate definitions. If you compile with this
1404 flag and your program mysteriously crashes after @code{main()} has
1405 completed, you may have an object that is being destroyed twice because
1406 two definitions were merged.
1408 This option is no longer useful on most targets, now that support has
1409 been added for putting variables into BSS without making them common.
1411 @item -fno-const-strings
1412 @opindex fno-const-strings
1413 Give string constants type @code{char *} instead of type @code{const
1414 char *}. By default, G++ uses type @code{const char *} as required by
1415 the standard. Even if you use @option{-fno-const-strings}, you cannot
1416 actually modify the value of a string constant.
1418 This option might be removed in a future release of G++. For maximum
1419 portability, you should structure your code so that it works with
1420 string constants that have type @code{const char *}.
1422 @item -fno-elide-constructors
1423 @opindex fno-elide-constructors
1424 The C++ standard allows an implementation to omit creating a temporary
1425 which is only used to initialize another object of the same type.
1426 Specifying this option disables that optimization, and forces G++ to
1427 call the copy constructor in all cases.
1429 @item -fno-enforce-eh-specs
1430 @opindex fno-enforce-eh-specs
1431 Don't check for violation of exception specifications at runtime. This
1432 option violates the C++ standard, but may be useful for reducing code
1433 size in production builds, much like defining @samp{NDEBUG}. The compiler
1434 will still optimize based on the exception specifications.
1437 @itemx -fno-for-scope
1439 @opindex fno-for-scope
1440 If @option{-ffor-scope} is specified, the scope of variables declared in
1441 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1442 as specified by the C++ standard.
1443 If @option{-fno-for-scope} is specified, the scope of variables declared in
1444 a @i{for-init-statement} extends to the end of the enclosing scope,
1445 as was the case in old versions of G++, and other (traditional)
1446 implementations of C++.
1448 The default if neither flag is given to follow the standard,
1449 but to allow and give a warning for old-style code that would
1450 otherwise be invalid, or have different behavior.
1452 @item -fno-gnu-keywords
1453 @opindex fno-gnu-keywords
1454 Do not recognize @code{typeof} as a keyword, so that code can use this
1455 word as an identifier. You can use the keyword @code{__typeof__} instead.
1456 @option{-ansi} implies @option{-fno-gnu-keywords}.
1458 @item -fno-implicit-templates
1459 @opindex fno-implicit-templates
1460 Never emit code for non-inline templates which are instantiated
1461 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1462 @xref{Template Instantiation}, for more information.
1464 @item -fno-implicit-inline-templates
1465 @opindex fno-implicit-inline-templates
1466 Don't emit code for implicit instantiations of inline templates, either.
1467 The default is to handle inlines differently so that compiles with and
1468 without optimization will need the same set of explicit instantiations.
1470 @item -fno-implement-inlines
1471 @opindex fno-implement-inlines
1472 To save space, do not emit out-of-line copies of inline functions
1473 controlled by @samp{#pragma implementation}. This will cause linker
1474 errors if these functions are not inlined everywhere they are called.
1476 @item -fms-extensions
1477 @opindex fms-extensions
1478 Disable pedantic warnings about constructs used in MFC, such as implicit
1479 int and getting a pointer to member function via non-standard syntax.
1481 @item -fno-nonansi-builtins
1482 @opindex fno-nonansi-builtins
1483 Disable built-in declarations of functions that are not mandated by
1484 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1485 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1487 @item -fno-operator-names
1488 @opindex fno-operator-names
1489 Do not treat the operator name keywords @code{and}, @code{bitand},
1490 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1491 synonyms as keywords.
1493 @item -fno-optional-diags
1494 @opindex fno-optional-diags
1495 Disable diagnostics that the standard says a compiler does not need to
1496 issue. Currently, the only such diagnostic issued by G++ is the one for
1497 a name having multiple meanings within a class.
1500 @opindex fpermissive
1501 Downgrade some diagnostics about nonconformant code from errors to
1502 warnings. Thus, using @option{-fpermissive} will allow some
1503 nonconforming code to compile.
1507 Enable automatic template instantiation at link time. This option also
1508 implies @option{-fno-implicit-templates}. @xref{Template
1509 Instantiation}, for more information.
1513 Disable generation of information about every class with virtual
1514 functions for use by the C++ runtime type identification features
1515 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1516 of the language, you can save some space by using this flag. Note that
1517 exception handling uses the same information, but it will generate it as
1522 Emit statistics about front-end processing at the end of the compilation.
1523 This information is generally only useful to the G++ development team.
1525 @item -ftemplate-depth-@var{n}
1526 @opindex ftemplate-depth
1527 Set the maximum instantiation depth for template classes to @var{n}.
1528 A limit on the template instantiation depth is needed to detect
1529 endless recursions during template class instantiation. ANSI/ISO C++
1530 conforming programs must not rely on a maximum depth greater than 17.
1532 @item -fno-threadsafe-statics
1533 @opindex fno-threadsafe-statics
1534 Do not emit the extra code to use the routines specified in the C++
1535 ABI for thread-safe initialization of local statics. You can use this
1536 option to reduce code size slightly in code that doesn't need to be
1539 @item -fuse-cxa-atexit
1540 @opindex fuse-cxa-atexit
1541 Register destructors for objects with static storage duration with the
1542 @code{__cxa_atexit} function rather than the @code{atexit} function.
1543 This option is required for fully standards-compliant handling of static
1544 destructors, but will only work if your C library supports
1545 @code{__cxa_atexit}.
1547 @item -fvisibility-inlines-hidden
1548 @opindex fvisibility-inlines-hidden
1549 Causes all inlined methods to be marked with
1550 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1551 appear in the export table of a DSO and do not require a PLT indirection
1552 when used within the DSO@. Enabling this option can have a dramatic effect
1553 on load and link times of a DSO as it massively reduces the size of the
1554 dynamic export table when the library makes heavy use of templates. While
1555 it can cause bloating through duplication of code within each DSO where
1556 it is used, often the wastage is less than the considerable space occupied
1557 by a long symbol name in the export table which is typical when using
1558 templates and namespaces. For even more savings, combine with the
1559 @option{-fvisibility=hidden} switch.
1563 Do not use weak symbol support, even if it is provided by the linker.
1564 By default, G++ will use weak symbols if they are available. This
1565 option exists only for testing, and should not be used by end-users;
1566 it will result in inferior code and has no benefits. This option may
1567 be removed in a future release of G++.
1571 Do not search for header files in the standard directories specific to
1572 C++, but do still search the other standard directories. (This option
1573 is used when building the C++ library.)
1576 In addition, these optimization, warning, and code generation options
1577 have meanings only for C++ programs:
1580 @item -fno-default-inline
1581 @opindex fno-default-inline
1582 Do not assume @samp{inline} for functions defined inside a class scope.
1583 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1584 functions will have linkage like inline functions; they just won't be
1587 @item -Wabi @r{(C++ only)}
1589 Warn when G++ generates code that is probably not compatible with the
1590 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1591 all such cases, there are probably some cases that are not warned about,
1592 even though G++ is generating incompatible code. There may also be
1593 cases where warnings are emitted even though the code that is generated
1596 You should rewrite your code to avoid these warnings if you are
1597 concerned about the fact that code generated by G++ may not be binary
1598 compatible with code generated by other compilers.
1600 The known incompatibilities at this point include:
1605 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1606 pack data into the same byte as a base class. For example:
1609 struct A @{ virtual void f(); int f1 : 1; @};
1610 struct B : public A @{ int f2 : 1; @};
1614 In this case, G++ will place @code{B::f2} into the same byte
1615 as@code{A::f1}; other compilers will not. You can avoid this problem
1616 by explicitly padding @code{A} so that its size is a multiple of the
1617 byte size on your platform; that will cause G++ and other compilers to
1618 layout @code{B} identically.
1621 Incorrect handling of tail-padding for virtual bases. G++ does not use
1622 tail padding when laying out virtual bases. For example:
1625 struct A @{ virtual void f(); char c1; @};
1626 struct B @{ B(); char c2; @};
1627 struct C : public A, public virtual B @{@};
1631 In this case, G++ will not place @code{B} into the tail-padding for
1632 @code{A}; other compilers will. You can avoid this problem by
1633 explicitly padding @code{A} so that its size is a multiple of its
1634 alignment (ignoring virtual base classes); that will cause G++ and other
1635 compilers to layout @code{C} identically.
1638 Incorrect handling of bit-fields with declared widths greater than that
1639 of their underlying types, when the bit-fields appear in a union. For
1643 union U @{ int i : 4096; @};
1647 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1648 union too small by the number of bits in an @code{int}.
1651 Empty classes can be placed at incorrect offsets. For example:
1661 struct C : public B, public A @{@};
1665 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1666 it should be placed at offset zero. G++ mistakenly believes that the
1667 @code{A} data member of @code{B} is already at offset zero.
1670 Names of template functions whose types involve @code{typename} or
1671 template template parameters can be mangled incorrectly.
1674 template <typename Q>
1675 void f(typename Q::X) @{@}
1677 template <template <typename> class Q>
1678 void f(typename Q<int>::X) @{@}
1682 Instantiations of these templates may be mangled incorrectly.
1686 @item -Wctor-dtor-privacy @r{(C++ only)}
1687 @opindex Wctor-dtor-privacy
1688 Warn when a class seems unusable because all the constructors or
1689 destructors in that class are private, and it has neither friends nor
1690 public static member functions.
1692 @item -Wnon-virtual-dtor @r{(C++ only)}
1693 @opindex Wnon-virtual-dtor
1694 Warn when a class appears to be polymorphic, thereby requiring a virtual
1695 destructor, yet it declares a non-virtual one.
1696 This warning is enabled by @option{-Wall}.
1698 @item -Wreorder @r{(C++ only)}
1700 @cindex reordering, warning
1701 @cindex warning for reordering of member initializers
1702 Warn when the order of member initializers given in the code does not
1703 match the order in which they must be executed. For instance:
1709 A(): j (0), i (1) @{ @}
1713 The compiler will rearrange the member initializers for @samp{i}
1714 and @samp{j} to match the declaration order of the members, emitting
1715 a warning to that effect. This warning is enabled by @option{-Wall}.
1718 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1721 @item -Weffc++ @r{(C++ only)}
1723 Warn about violations of the following style guidelines from Scott Meyers'
1724 @cite{Effective C++} book:
1728 Item 11: Define a copy constructor and an assignment operator for classes
1729 with dynamically allocated memory.
1732 Item 12: Prefer initialization to assignment in constructors.
1735 Item 14: Make destructors virtual in base classes.
1738 Item 15: Have @code{operator=} return a reference to @code{*this}.
1741 Item 23: Don't try to return a reference when you must return an object.
1745 Also warn about violations of the following style guidelines from
1746 Scott Meyers' @cite{More Effective C++} book:
1750 Item 6: Distinguish between prefix and postfix forms of increment and
1751 decrement operators.
1754 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1758 When selecting this option, be aware that the standard library
1759 headers do not obey all of these guidelines; use @samp{grep -v}
1760 to filter out those warnings.
1762 @item -Wno-deprecated @r{(C++ only)}
1763 @opindex Wno-deprecated
1764 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1766 @item -Wstrict-null-sentinel @r{(C++ only)}
1767 @opindex Wstrict-null-sentinel
1768 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1769 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1770 to @code{__null}. Although it is a null pointer constant not a null pointer,
1771 it is guaranteed to of the same size as a pointer. But this use is
1772 not portable across different compilers.
1774 @item -Wno-non-template-friend @r{(C++ only)}
1775 @opindex Wno-non-template-friend
1776 Disable warnings when non-templatized friend functions are declared
1777 within a template. Since the advent of explicit template specification
1778 support in G++, if the name of the friend is an unqualified-id (i.e.,
1779 @samp{friend foo(int)}), the C++ language specification demands that the
1780 friend declare or define an ordinary, nontemplate function. (Section
1781 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1782 could be interpreted as a particular specialization of a templatized
1783 function. Because this non-conforming behavior is no longer the default
1784 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1785 check existing code for potential trouble spots and is on by default.
1786 This new compiler behavior can be turned off with
1787 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1788 but disables the helpful warning.
1790 @item -Wold-style-cast @r{(C++ only)}
1791 @opindex Wold-style-cast
1792 Warn if an old-style (C-style) cast to a non-void type is used within
1793 a C++ program. The new-style casts (@samp{static_cast},
1794 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1795 unintended effects and much easier to search for.
1797 @item -Woverloaded-virtual @r{(C++ only)}
1798 @opindex Woverloaded-virtual
1799 @cindex overloaded virtual fn, warning
1800 @cindex warning for overloaded virtual fn
1801 Warn when a function declaration hides virtual functions from a
1802 base class. For example, in:
1809 struct B: public A @{
1814 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1822 will fail to compile.
1824 @item -Wno-pmf-conversions @r{(C++ only)}
1825 @opindex Wno-pmf-conversions
1826 Disable the diagnostic for converting a bound pointer to member function
1829 @item -Wsign-promo @r{(C++ only)}
1830 @opindex Wsign-promo
1831 Warn when overload resolution chooses a promotion from unsigned or
1832 enumerated type to a signed type, over a conversion to an unsigned type of
1833 the same size. Previous versions of G++ would try to preserve
1834 unsignedness, but the standard mandates the current behavior.
1839 A& operator = (int);
1849 In this example, G++ will synthesize a default @samp{A& operator =
1850 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1853 @node Objective-C and Objective-C++ Dialect Options
1854 @section Options Controlling Objective-C and Objective-C++ Dialects
1856 @cindex compiler options, Objective-C and Objective-C++
1857 @cindex Objective-C and Objective-C++ options, command line
1858 @cindex options, Objective-C and Objective-C++
1859 (NOTE: This manual does not describe the Objective-C and Objective-C++
1860 languages themselves. See @xref{Standards,,Language Standards
1861 Supported by GCC}, for references.)
1863 This section describes the command-line options that are only meaningful
1864 for Objective-C and Objective-C++ programs, but you can also use most of
1865 the language-independent GNU compiler options.
1866 For example, you might compile a file @code{some_class.m} like this:
1869 gcc -g -fgnu-runtime -O -c some_class.m
1873 In this example, @option{-fgnu-runtime} is an option meant only for
1874 Objective-C and Objective-C++ programs; you can use the other options with
1875 any language supported by GCC@.
1877 Note that since Objective-C is an extension of the C language, Objective-C
1878 compilations may also use options specific to the C front-end (e.g.,
1879 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1880 C++-specific options (e.g., @option{-Wabi}).
1882 Here is a list of options that are @emph{only} for compiling Objective-C
1883 and Objective-C++ programs:
1886 @item -fconstant-string-class=@var{class-name}
1887 @opindex fconstant-string-class
1888 Use @var{class-name} as the name of the class to instantiate for each
1889 literal string specified with the syntax @code{@@"@dots{}"}. The default
1890 class name is @code{NXConstantString} if the GNU runtime is being used, and
1891 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1892 @option{-fconstant-cfstrings} option, if also present, will override the
1893 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1894 to be laid out as constant CoreFoundation strings.
1897 @opindex fgnu-runtime
1898 Generate object code compatible with the standard GNU Objective-C
1899 runtime. This is the default for most types of systems.
1901 @item -fnext-runtime
1902 @opindex fnext-runtime
1903 Generate output compatible with the NeXT runtime. This is the default
1904 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1905 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1908 @item -fno-nil-receivers
1909 @opindex fno-nil-receivers
1910 Assume that all Objective-C message dispatches (e.g.,
1911 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1912 is not @code{nil}. This allows for more efficient entry points in the runtime
1913 to be used. Currently, this option is only available in conjunction with
1914 the NeXT runtime on Mac OS X 10.3 and later.
1916 @item -fobjc-call-cxx-cdtors
1917 @opindex fobjc-call-cxx-cdtors
1918 For each Objective-C class, check if any of its instance variables is a
1919 C++ object with a non-trivial default constructor. If so, synthesize a
1920 special @code{- (id) .cxx_construct} instance method that will run
1921 non-trivial default constructors on any such instance variables, in order,
1922 and then return @code{self}. Similarly, check if any instance variable
1923 is a C++ object with a non-trivial destructor, and if so, synthesize a
1924 special @code{- (void) .cxx_destruct} method that will run
1925 all such default destructors, in reverse order.
1927 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
1928 thusly generated will only operate on instance variables declared in the
1929 current Objective-C class, and not those inherited from superclasses. It
1930 is the responsibility of the Objective-C runtime to invoke all such methods
1931 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
1932 will be invoked by the runtime immediately after a new object
1933 instance is allocated; the @code{- (void) .cxx_destruct} methods will
1934 be invoked immediately before the runtime deallocates an object instance.
1936 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
1937 support for invoking the @code{- (id) .cxx_construct} and
1938 @code{- (void) .cxx_destruct} methods.
1940 @item -fobjc-direct-dispatch
1941 @opindex fobjc-direct-dispatch
1942 Allow fast jumps to the message dispatcher. On Darwin this is
1943 accomplished via the comm page.
1945 @item -fobjc-exceptions
1946 @opindex fobjc-exceptions
1947 Enable syntactic support for structured exception handling in Objective-C,
1948 similar to what is offered by C++ and Java. Currently, this option is only
1949 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1957 @@catch (AnObjCClass *exc) @{
1964 @@catch (AnotherClass *exc) @{
1967 @@catch (id allOthers) @{
1977 The @code{@@throw} statement may appear anywhere in an Objective-C or
1978 Objective-C++ program; when used inside of a @code{@@catch} block, the
1979 @code{@@throw} may appear without an argument (as shown above), in which case
1980 the object caught by the @code{@@catch} will be rethrown.
1982 Note that only (pointers to) Objective-C objects may be thrown and
1983 caught using this scheme. When an object is thrown, it will be caught
1984 by the nearest @code{@@catch} clause capable of handling objects of that type,
1985 analogously to how @code{catch} blocks work in C++ and Java. A
1986 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1987 any and all Objective-C exceptions not caught by previous @code{@@catch}
1990 The @code{@@finally} clause, if present, will be executed upon exit from the
1991 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1992 regardless of whether any exceptions are thrown, caught or rethrown
1993 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1994 of the @code{finally} clause in Java.
1996 There are several caveats to using the new exception mechanism:
2000 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2001 idioms provided by the @code{NSException} class, the new
2002 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2003 systems, due to additional functionality needed in the (NeXT) Objective-C
2007 As mentioned above, the new exceptions do not support handling
2008 types other than Objective-C objects. Furthermore, when used from
2009 Objective-C++, the Objective-C exception model does not interoperate with C++
2010 exceptions at this time. This means you cannot @code{@@throw} an exception
2011 from Objective-C and @code{catch} it in C++, or vice versa
2012 (i.e., @code{throw @dots{} @@catch}).
2015 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2016 blocks for thread-safe execution:
2019 @@synchronized (ObjCClass *guard) @{
2024 Upon entering the @code{@@synchronized} block, a thread of execution shall
2025 first check whether a lock has been placed on the corresponding @code{guard}
2026 object by another thread. If it has, the current thread shall wait until
2027 the other thread relinquishes its lock. Once @code{guard} becomes available,
2028 the current thread will place its own lock on it, execute the code contained in
2029 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2030 making @code{guard} available to other threads).
2032 Unlike Java, Objective-C does not allow for entire methods to be marked
2033 @code{@@synchronized}. Note that throwing exceptions out of
2034 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2035 to be unlocked properly.
2039 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2041 @item -freplace-objc-classes
2042 @opindex freplace-objc-classes
2043 Emit a special marker instructing @command{ld(1)} not to statically link in
2044 the resulting object file, and allow @command{dyld(1)} to load it in at
2045 run time instead. This is used in conjunction with the Fix-and-Continue
2046 debugging mode, where the object file in question may be recompiled and
2047 dynamically reloaded in the course of program execution, without the need
2048 to restart the program itself. Currently, Fix-and-Continue functionality
2049 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2054 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2055 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2056 compile time) with static class references that get initialized at load time,
2057 which improves run-time performance. Specifying the @option{-fzero-link} flag
2058 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2059 to be retained. This is useful in Zero-Link debugging mode, since it allows
2060 for individual class implementations to be modified during program execution.
2064 Dump interface declarations for all classes seen in the source file to a
2065 file named @file{@var{sourcename}.decl}.
2067 @item -Wassign-intercept
2068 @opindex Wassign-intercept
2069 Warn whenever an Objective-C assignment is being intercepted by the
2073 @opindex Wno-protocol
2074 If a class is declared to implement a protocol, a warning is issued for
2075 every method in the protocol that is not implemented by the class. The
2076 default behavior is to issue a warning for every method not explicitly
2077 implemented in the class, even if a method implementation is inherited
2078 from the superclass. If you use the @option{-Wno-protocol} option, then
2079 methods inherited from the superclass are considered to be implemented,
2080 and no warning is issued for them.
2084 Warn if multiple methods of different types for the same selector are
2085 found during compilation. The check is performed on the list of methods
2086 in the final stage of compilation. Additionally, a check is performed
2087 for each selector appearing in a @code{@@selector(@dots{})}
2088 expression, and a corresponding method for that selector has been found
2089 during compilation. Because these checks scan the method table only at
2090 the end of compilation, these warnings are not produced if the final
2091 stage of compilation is not reached, for example because an error is
2092 found during compilation, or because the @option{-fsyntax-only} option is
2095 @item -Wstrict-selector-match
2096 @opindex Wstrict-selector-match
2097 Warn if multiple methods with differing argument and/or return types are
2098 found for a given selector when attempting to send a message using this
2099 selector to a receiver of type @code{id} or @code{Class}. When this flag
2100 is off (which is the default behavior), the compiler will omit such warnings
2101 if any differences found are confined to types which share the same size
2104 @item -Wundeclared-selector
2105 @opindex Wundeclared-selector
2106 Warn if a @code{@@selector(@dots{})} expression referring to an
2107 undeclared selector is found. A selector is considered undeclared if no
2108 method with that name has been declared before the
2109 @code{@@selector(@dots{})} expression, either explicitly in an
2110 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2111 an @code{@@implementation} section. This option always performs its
2112 checks as soon as a @code{@@selector(@dots{})} expression is found,
2113 while @option{-Wselector} only performs its checks in the final stage of
2114 compilation. This also enforces the coding style convention
2115 that methods and selectors must be declared before being used.
2117 @item -print-objc-runtime-info
2118 @opindex print-objc-runtime-info
2119 Generate C header describing the largest structure that is passed by
2124 @node Language Independent Options
2125 @section Options to Control Diagnostic Messages Formatting
2126 @cindex options to control diagnostics formatting
2127 @cindex diagnostic messages
2128 @cindex message formatting
2130 Traditionally, diagnostic messages have been formatted irrespective of
2131 the output device's aspect (e.g.@: its width, @dots{}). The options described
2132 below can be used to control the diagnostic messages formatting
2133 algorithm, e.g.@: how many characters per line, how often source location
2134 information should be reported. Right now, only the C++ front end can
2135 honor these options. However it is expected, in the near future, that
2136 the remaining front ends would be able to digest them correctly.
2139 @item -fmessage-length=@var{n}
2140 @opindex fmessage-length
2141 Try to format error messages so that they fit on lines of about @var{n}
2142 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2143 the front ends supported by GCC@. If @var{n} is zero, then no
2144 line-wrapping will be done; each error message will appear on a single
2147 @opindex fdiagnostics-show-location
2148 @item -fdiagnostics-show-location=once
2149 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2150 reporter to emit @emph{once} source location information; that is, in
2151 case the message is too long to fit on a single physical line and has to
2152 be wrapped, the source location won't be emitted (as prefix) again,
2153 over and over, in subsequent continuation lines. This is the default
2156 @item -fdiagnostics-show-location=every-line
2157 Only meaningful in line-wrapping mode. Instructs the diagnostic
2158 messages reporter to emit the same source location information (as
2159 prefix) for physical lines that result from the process of breaking
2160 a message which is too long to fit on a single line.
2162 @item -fdiagnostics-show-options
2163 @opindex fdiagnostics-show-options
2164 This option instructs the diagnostic machinery to add text to each
2165 diagnostic emitted, which indicates which command line option directly
2166 controls that diagnostic, when such an option is known to the
2167 diagnostic machinery.
2171 @node Warning Options
2172 @section Options to Request or Suppress Warnings
2173 @cindex options to control warnings
2174 @cindex warning messages
2175 @cindex messages, warning
2176 @cindex suppressing warnings
2178 Warnings are diagnostic messages that report constructions which
2179 are not inherently erroneous but which are risky or suggest there
2180 may have been an error.
2182 You can request many specific warnings with options beginning @samp{-W},
2183 for example @option{-Wimplicit} to request warnings on implicit
2184 declarations. Each of these specific warning options also has a
2185 negative form beginning @samp{-Wno-} to turn off warnings;
2186 for example, @option{-Wno-implicit}. This manual lists only one of the
2187 two forms, whichever is not the default.
2189 The following options control the amount and kinds of warnings produced
2190 by GCC; for further, language-specific options also refer to
2191 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2195 @cindex syntax checking
2197 @opindex fsyntax-only
2198 Check the code for syntax errors, but don't do anything beyond that.
2202 Issue all the warnings demanded by strict ISO C and ISO C++;
2203 reject all programs that use forbidden extensions, and some other
2204 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2205 version of the ISO C standard specified by any @option{-std} option used.
2207 Valid ISO C and ISO C++ programs should compile properly with or without
2208 this option (though a rare few will require @option{-ansi} or a
2209 @option{-std} option specifying the required version of ISO C)@. However,
2210 without this option, certain GNU extensions and traditional C and C++
2211 features are supported as well. With this option, they are rejected.
2213 @option{-pedantic} does not cause warning messages for use of the
2214 alternate keywords whose names begin and end with @samp{__}. Pedantic
2215 warnings are also disabled in the expression that follows
2216 @code{__extension__}. However, only system header files should use
2217 these escape routes; application programs should avoid them.
2218 @xref{Alternate Keywords}.
2220 Some users try to use @option{-pedantic} to check programs for strict ISO
2221 C conformance. They soon find that it does not do quite what they want:
2222 it finds some non-ISO practices, but not all---only those for which
2223 ISO C @emph{requires} a diagnostic, and some others for which
2224 diagnostics have been added.
2226 A feature to report any failure to conform to ISO C might be useful in
2227 some instances, but would require considerable additional work and would
2228 be quite different from @option{-pedantic}. We don't have plans to
2229 support such a feature in the near future.
2231 Where the standard specified with @option{-std} represents a GNU
2232 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2233 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2234 extended dialect is based. Warnings from @option{-pedantic} are given
2235 where they are required by the base standard. (It would not make sense
2236 for such warnings to be given only for features not in the specified GNU
2237 C dialect, since by definition the GNU dialects of C include all
2238 features the compiler supports with the given option, and there would be
2239 nothing to warn about.)
2241 @item -pedantic-errors
2242 @opindex pedantic-errors
2243 Like @option{-pedantic}, except that errors are produced rather than
2248 Inhibit all warning messages.
2252 Inhibit warning messages about the use of @samp{#import}.
2254 @item -Wchar-subscripts
2255 @opindex Wchar-subscripts
2256 Warn if an array subscript has type @code{char}. This is a common cause
2257 of error, as programmers often forget that this type is signed on some
2259 This warning is enabled by @option{-Wall}.
2263 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2264 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2265 This warning is enabled by @option{-Wall}.
2267 @item -Wfatal-errors
2268 @opindex Wfatal-errors
2269 This option causes the compiler to abort compilation on the first error
2270 occurred rather than trying to keep going and printing further error
2275 @opindex ffreestanding
2276 @opindex fno-builtin
2277 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2278 the arguments supplied have types appropriate to the format string
2279 specified, and that the conversions specified in the format string make
2280 sense. This includes standard functions, and others specified by format
2281 attributes (@pxref{Function Attributes}), in the @code{printf},
2282 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2283 not in the C standard) families (or other target-specific families).
2284 Which functions are checked without format attributes having been
2285 specified depends on the standard version selected, and such checks of
2286 functions without the attribute specified are disabled by
2287 @option{-ffreestanding} or @option{-fno-builtin}.
2289 The formats are checked against the format features supported by GNU
2290 libc version 2.2. These include all ISO C90 and C99 features, as well
2291 as features from the Single Unix Specification and some BSD and GNU
2292 extensions. Other library implementations may not support all these
2293 features; GCC does not support warning about features that go beyond a
2294 particular library's limitations. However, if @option{-pedantic} is used
2295 with @option{-Wformat}, warnings will be given about format features not
2296 in the selected standard version (but not for @code{strfmon} formats,
2297 since those are not in any version of the C standard). @xref{C Dialect
2298 Options,,Options Controlling C Dialect}.
2300 Since @option{-Wformat} also checks for null format arguments for
2301 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2303 @option{-Wformat} is included in @option{-Wall}. For more control over some
2304 aspects of format checking, the options @option{-Wformat-y2k},
2305 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2306 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2307 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2310 @opindex Wformat-y2k
2311 If @option{-Wformat} is specified, also warn about @code{strftime}
2312 formats which may yield only a two-digit year.
2314 @item -Wno-format-extra-args
2315 @opindex Wno-format-extra-args
2316 If @option{-Wformat} is specified, do not warn about excess arguments to a
2317 @code{printf} or @code{scanf} format function. The C standard specifies
2318 that such arguments are ignored.
2320 Where the unused arguments lie between used arguments that are
2321 specified with @samp{$} operand number specifications, normally
2322 warnings are still given, since the implementation could not know what
2323 type to pass to @code{va_arg} to skip the unused arguments. However,
2324 in the case of @code{scanf} formats, this option will suppress the
2325 warning if the unused arguments are all pointers, since the Single
2326 Unix Specification says that such unused arguments are allowed.
2328 @item -Wno-format-zero-length
2329 @opindex Wno-format-zero-length
2330 If @option{-Wformat} is specified, do not warn about zero-length formats.
2331 The C standard specifies that zero-length formats are allowed.
2333 @item -Wformat-nonliteral
2334 @opindex Wformat-nonliteral
2335 If @option{-Wformat} is specified, also warn if the format string is not a
2336 string literal and so cannot be checked, unless the format function
2337 takes its format arguments as a @code{va_list}.
2339 @item -Wformat-security
2340 @opindex Wformat-security
2341 If @option{-Wformat} is specified, also warn about uses of format
2342 functions that represent possible security problems. At present, this
2343 warns about calls to @code{printf} and @code{scanf} functions where the
2344 format string is not a string literal and there are no format arguments,
2345 as in @code{printf (foo);}. This may be a security hole if the format
2346 string came from untrusted input and contains @samp{%n}. (This is
2347 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2348 in future warnings may be added to @option{-Wformat-security} that are not
2349 included in @option{-Wformat-nonliteral}.)
2353 Enable @option{-Wformat} plus format checks not included in
2354 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2355 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2359 Warn about passing a null pointer for arguments marked as
2360 requiring a non-null value by the @code{nonnull} function attribute.
2362 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2363 can be disabled with the @option{-Wno-nonnull} option.
2365 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2367 Warn about uninitialized variables which are initialized with themselves.
2368 Note this option can only be used with the @option{-Wuninitialized} option,
2369 which in turn only works with @option{-O1} and above.
2371 For example, GCC will warn about @code{i} being uninitialized in the
2372 following snippet only when @option{-Winit-self} has been specified:
2383 @item -Wimplicit-int
2384 @opindex Wimplicit-int
2385 Warn when a declaration does not specify a type.
2386 This warning is enabled by @option{-Wall}.
2388 @item -Wimplicit-function-declaration
2389 @itemx -Werror-implicit-function-declaration
2390 @opindex Wimplicit-function-declaration
2391 @opindex Werror-implicit-function-declaration
2392 Give a warning (or error) whenever a function is used before being
2393 declared. The form @option{-Wno-error-implicit-function-declaration}
2395 This warning is enabled by @option{-Wall} (as a warning, not an error).
2399 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2400 This warning is enabled by @option{-Wall}.
2404 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2405 function with external linkage, returning int, taking either zero
2406 arguments, two, or three arguments of appropriate types.
2407 This warning is enabled by @option{-Wall}.
2409 @item -Wmissing-braces
2410 @opindex Wmissing-braces
2411 Warn if an aggregate or union initializer is not fully bracketed. In
2412 the following example, the initializer for @samp{a} is not fully
2413 bracketed, but that for @samp{b} is fully bracketed.
2416 int a[2][2] = @{ 0, 1, 2, 3 @};
2417 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2420 This warning is enabled by @option{-Wall}.
2422 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2423 @opindex Wmissing-include-dirs
2424 Warn if a user-supplied include directory does not exist.
2427 @opindex Wparentheses
2428 Warn if parentheses are omitted in certain contexts, such
2429 as when there is an assignment in a context where a truth value
2430 is expected, or when operators are nested whose precedence people
2431 often get confused about. Only the warning for an assignment used as
2432 a truth value is supported when compiling C++; the other warnings are
2433 only supported when compiling C@.
2435 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2436 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2437 interpretation from that of ordinary mathematical notation.
2439 Also warn about constructions where there may be confusion to which
2440 @code{if} statement an @code{else} branch belongs. Here is an example of
2455 In C, every @code{else} branch belongs to the innermost possible @code{if}
2456 statement, which in this example is @code{if (b)}. This is often not
2457 what the programmer expected, as illustrated in the above example by
2458 indentation the programmer chose. When there is the potential for this
2459 confusion, GCC will issue a warning when this flag is specified.
2460 To eliminate the warning, add explicit braces around the innermost
2461 @code{if} statement so there is no way the @code{else} could belong to
2462 the enclosing @code{if}. The resulting code would look like this:
2478 This warning is enabled by @option{-Wall}.
2480 @item -Wsequence-point
2481 @opindex Wsequence-point
2482 Warn about code that may have undefined semantics because of violations
2483 of sequence point rules in the C standard.
2485 The C standard defines the order in which expressions in a C program are
2486 evaluated in terms of @dfn{sequence points}, which represent a partial
2487 ordering between the execution of parts of the program: those executed
2488 before the sequence point, and those executed after it. These occur
2489 after the evaluation of a full expression (one which is not part of a
2490 larger expression), after the evaluation of the first operand of a
2491 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2492 function is called (but after the evaluation of its arguments and the
2493 expression denoting the called function), and in certain other places.
2494 Other than as expressed by the sequence point rules, the order of
2495 evaluation of subexpressions of an expression is not specified. All
2496 these rules describe only a partial order rather than a total order,
2497 since, for example, if two functions are called within one expression
2498 with no sequence point between them, the order in which the functions
2499 are called is not specified. However, the standards committee have
2500 ruled that function calls do not overlap.
2502 It is not specified when between sequence points modifications to the
2503 values of objects take effect. Programs whose behavior depends on this
2504 have undefined behavior; the C standard specifies that ``Between the
2505 previous and next sequence point an object shall have its stored value
2506 modified at most once by the evaluation of an expression. Furthermore,
2507 the prior value shall be read only to determine the value to be
2508 stored.''. If a program breaks these rules, the results on any
2509 particular implementation are entirely unpredictable.
2511 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2512 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2513 diagnosed by this option, and it may give an occasional false positive
2514 result, but in general it has been found fairly effective at detecting
2515 this sort of problem in programs.
2517 The present implementation of this option only works for C programs. A
2518 future implementation may also work for C++ programs.
2520 The C standard is worded confusingly, therefore there is some debate
2521 over the precise meaning of the sequence point rules in subtle cases.
2522 Links to discussions of the problem, including proposed formal
2523 definitions, may be found on the GCC readings page, at
2524 @w{@uref{http://gcc.gnu.org/readings.html}}.
2526 This warning is enabled by @option{-Wall}.
2529 @opindex Wreturn-type
2530 Warn whenever a function is defined with a return-type that defaults to
2531 @code{int}. Also warn about any @code{return} statement with no
2532 return-value in a function whose return-type is not @code{void}.
2534 For C, also warn if the return type of a function has a type qualifier
2535 such as @code{const}. Such a type qualifier has no effect, since the
2536 value returned by a function is not an lvalue. ISO C prohibits
2537 qualified @code{void} return types on function definitions, so such
2538 return types always receive a warning even without this option.
2540 For C++, a function without return type always produces a diagnostic
2541 message, even when @option{-Wno-return-type} is specified. The only
2542 exceptions are @samp{main} and functions defined in system headers.
2544 This warning is enabled by @option{-Wall}.
2548 Warn whenever a @code{switch} statement has an index of enumerated type
2549 and lacks a @code{case} for one or more of the named codes of that
2550 enumeration. (The presence of a @code{default} label prevents this
2551 warning.) @code{case} labels outside the enumeration range also
2552 provoke warnings when this option is used.
2553 This warning is enabled by @option{-Wall}.
2555 @item -Wswitch-default
2556 @opindex Wswitch-switch
2557 Warn whenever a @code{switch} statement does not have a @code{default}
2561 @opindex Wswitch-enum
2562 Warn whenever a @code{switch} statement has an index of enumerated type
2563 and lacks a @code{case} for one or more of the named codes of that
2564 enumeration. @code{case} labels outside the enumeration range also
2565 provoke warnings when this option is used.
2569 Warn if any trigraphs are encountered that might change the meaning of
2570 the program (trigraphs within comments are not warned about).
2571 This warning is enabled by @option{-Wall}.
2573 @item -Wunused-function
2574 @opindex Wunused-function
2575 Warn whenever a static function is declared but not defined or a
2576 non\-inline static function is unused.
2577 This warning is enabled by @option{-Wall}.
2579 @item -Wunused-label
2580 @opindex Wunused-label
2581 Warn whenever a label is declared but not used.
2582 This warning is enabled by @option{-Wall}.
2584 To suppress this warning use the @samp{unused} attribute
2585 (@pxref{Variable Attributes}).
2587 @item -Wunused-parameter
2588 @opindex Wunused-parameter
2589 Warn whenever a function parameter is unused aside from its declaration.
2591 To suppress this warning use the @samp{unused} attribute
2592 (@pxref{Variable Attributes}).
2594 @item -Wunused-variable
2595 @opindex Wunused-variable
2596 Warn whenever a local variable or non-constant static variable is unused
2597 aside from its declaration
2598 This warning is enabled by @option{-Wall}.
2600 To suppress this warning use the @samp{unused} attribute
2601 (@pxref{Variable Attributes}).
2603 @item -Wunused-value
2604 @opindex Wunused-value
2605 Warn whenever a statement computes a result that is explicitly not used.
2606 This warning is enabled by @option{-Wall}.
2608 To suppress this warning cast the expression to @samp{void}.
2612 All the above @option{-Wunused} options combined.
2614 In order to get a warning about an unused function parameter, you must
2615 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2616 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2618 @item -Wuninitialized
2619 @opindex Wuninitialized
2620 Warn if an automatic variable is used without first being initialized or
2621 if a variable may be clobbered by a @code{setjmp} call.
2623 These warnings are possible only in optimizing compilation,
2624 because they require data flow information that is computed only
2625 when optimizing. If you don't specify @option{-O}, you simply won't
2628 If you want to warn about code which uses the uninitialized value of the
2629 variable in its own initializer, use the @option{-Winit-self} option.
2631 These warnings occur for individual uninitialized or clobbered
2632 elements of structure, union or array variables as well as for
2633 variables which are uninitialized or clobbered as a whole. They do
2634 not occur for variables or elements declared @code{volatile}. Because
2635 these warnings depend on optimization, the exact variables or elements
2636 for which there are warnings will depend on the precise optimization
2637 options and version of GCC used.
2639 Note that there may be no warning about a variable that is used only
2640 to compute a value that itself is never used, because such
2641 computations may be deleted by data flow analysis before the warnings
2644 These warnings are made optional because GCC is not smart
2645 enough to see all the reasons why the code might be correct
2646 despite appearing to have an error. Here is one example of how
2667 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2668 always initialized, but GCC doesn't know this. Here is
2669 another common case:
2674 if (change_y) save_y = y, y = new_y;
2676 if (change_y) y = save_y;
2681 This has no bug because @code{save_y} is used only if it is set.
2683 @cindex @code{longjmp} warnings
2684 This option also warns when a non-volatile automatic variable might be
2685 changed by a call to @code{longjmp}. These warnings as well are possible
2686 only in optimizing compilation.
2688 The compiler sees only the calls to @code{setjmp}. It cannot know
2689 where @code{longjmp} will be called; in fact, a signal handler could
2690 call it at any point in the code. As a result, you may get a warning
2691 even when there is in fact no problem because @code{longjmp} cannot
2692 in fact be called at the place which would cause a problem.
2694 Some spurious warnings can be avoided if you declare all the functions
2695 you use that never return as @code{noreturn}. @xref{Function
2698 This warning is enabled by @option{-Wall}.
2700 @item -Wunknown-pragmas
2701 @opindex Wunknown-pragmas
2702 @cindex warning for unknown pragmas
2703 @cindex unknown pragmas, warning
2704 @cindex pragmas, warning of unknown
2705 Warn when a #pragma directive is encountered which is not understood by
2706 GCC@. If this command line option is used, warnings will even be issued
2707 for unknown pragmas in system header files. This is not the case if
2708 the warnings were only enabled by the @option{-Wall} command line option.
2710 @item -Wstrict-aliasing
2711 @opindex Wstrict-aliasing
2712 This option is only active when @option{-fstrict-aliasing} is active.
2713 It warns about code which might break the strict aliasing rules that the
2714 compiler is using for optimization. The warning does not catch all
2715 cases, but does attempt to catch the more common pitfalls. It is
2716 included in @option{-Wall}.
2718 @item -Wstrict-aliasing=2
2719 @opindex Wstrict-aliasing=2
2720 This option is only active when @option{-fstrict-aliasing} is active.
2721 It warns about all code which might break the strict aliasing rules that the
2722 compiler is using for optimization. This warning catches all cases, but
2723 it will also give a warning for some ambiguous cases that are safe.
2727 All of the above @samp{-W} options combined. This enables all the
2728 warnings about constructions that some users consider questionable, and
2729 that are easy to avoid (or modify to prevent the warning), even in
2730 conjunction with macros. This also enables some language-specific
2731 warnings described in @ref{C++ Dialect Options} and
2732 @ref{Objective-C and Objective-C++ Dialect Options}.
2735 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2736 Some of them warn about constructions that users generally do not
2737 consider questionable, but which occasionally you might wish to check
2738 for; others warn about constructions that are necessary or hard to avoid
2739 in some cases, and there is no simple way to modify the code to suppress
2746 (This option used to be called @option{-W}. The older name is still
2747 supported, but the newer name is more descriptive.) Print extra warning
2748 messages for these events:
2752 A function can return either with or without a value. (Falling
2753 off the end of the function body is considered returning without
2754 a value.) For example, this function would evoke such a
2768 An expression-statement or the left-hand side of a comma expression
2769 contains no side effects.
2770 To suppress the warning, cast the unused expression to void.
2771 For example, an expression such as @samp{x[i,j]} will cause a warning,
2772 but @samp{x[(void)i,j]} will not.
2775 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2778 Storage-class specifiers like @code{static} are not the first things in
2779 a declaration. According to the C Standard, this usage is obsolescent.
2782 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2786 A comparison between signed and unsigned values could produce an
2787 incorrect result when the signed value is converted to unsigned.
2788 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2791 An aggregate has an initializer which does not initialize all members.
2792 This warning can be independently controlled by
2793 @option{-Wmissing-field-initializers}.
2796 A function parameter is declared without a type specifier in K&R-style
2804 An empty body occurs in an @samp{if} or @samp{else} statement.
2807 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2808 @samp{>}, or @samp{>=}.
2811 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2814 Any of several floating-point events that often indicate errors, such as
2815 overflow, underflow, loss of precision, etc.
2817 @item @r{(C++ only)}
2818 An enumerator and a non-enumerator both appear in a conditional expression.
2820 @item @r{(C++ only)}
2821 A non-static reference or non-static @samp{const} member appears in a
2822 class without constructors.
2824 @item @r{(C++ only)}
2825 Ambiguous virtual bases.
2827 @item @r{(C++ only)}
2828 Subscripting an array which has been declared @samp{register}.
2830 @item @r{(C++ only)}
2831 Taking the address of a variable which has been declared @samp{register}.
2833 @item @r{(C++ only)}
2834 A base class is not initialized in a derived class' copy constructor.
2837 @item -Wno-div-by-zero
2838 @opindex Wno-div-by-zero
2839 @opindex Wdiv-by-zero
2840 Do not warn about compile-time integer division by zero. Floating point
2841 division by zero is not warned about, as it can be a legitimate way of
2842 obtaining infinities and NaNs.
2844 @item -Wsystem-headers
2845 @opindex Wsystem-headers
2846 @cindex warnings from system headers
2847 @cindex system headers, warnings from
2848 Print warning messages for constructs found in system header files.
2849 Warnings from system headers are normally suppressed, on the assumption
2850 that they usually do not indicate real problems and would only make the
2851 compiler output harder to read. Using this command line option tells
2852 GCC to emit warnings from system headers as if they occurred in user
2853 code. However, note that using @option{-Wall} in conjunction with this
2854 option will @emph{not} warn about unknown pragmas in system
2855 headers---for that, @option{-Wunknown-pragmas} must also be used.
2858 @opindex Wfloat-equal
2859 Warn if floating point values are used in equality comparisons.
2861 The idea behind this is that sometimes it is convenient (for the
2862 programmer) to consider floating-point values as approximations to
2863 infinitely precise real numbers. If you are doing this, then you need
2864 to compute (by analyzing the code, or in some other way) the maximum or
2865 likely maximum error that the computation introduces, and allow for it
2866 when performing comparisons (and when producing output, but that's a
2867 different problem). In particular, instead of testing for equality, you
2868 would check to see whether the two values have ranges that overlap; and
2869 this is done with the relational operators, so equality comparisons are
2872 @item -Wtraditional @r{(C only)}
2873 @opindex Wtraditional
2874 Warn about certain constructs that behave differently in traditional and
2875 ISO C@. Also warn about ISO C constructs that have no traditional C
2876 equivalent, and/or problematic constructs which should be avoided.
2880 Macro parameters that appear within string literals in the macro body.
2881 In traditional C macro replacement takes place within string literals,
2882 but does not in ISO C@.
2885 In traditional C, some preprocessor directives did not exist.
2886 Traditional preprocessors would only consider a line to be a directive
2887 if the @samp{#} appeared in column 1 on the line. Therefore
2888 @option{-Wtraditional} warns about directives that traditional C
2889 understands but would ignore because the @samp{#} does not appear as the
2890 first character on the line. It also suggests you hide directives like
2891 @samp{#pragma} not understood by traditional C by indenting them. Some
2892 traditional implementations would not recognize @samp{#elif}, so it
2893 suggests avoiding it altogether.
2896 A function-like macro that appears without arguments.
2899 The unary plus operator.
2902 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2903 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2904 constants.) Note, these suffixes appear in macros defined in the system
2905 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2906 Use of these macros in user code might normally lead to spurious
2907 warnings, however GCC's integrated preprocessor has enough context to
2908 avoid warning in these cases.
2911 A function declared external in one block and then used after the end of
2915 A @code{switch} statement has an operand of type @code{long}.
2918 A non-@code{static} function declaration follows a @code{static} one.
2919 This construct is not accepted by some traditional C compilers.
2922 The ISO type of an integer constant has a different width or
2923 signedness from its traditional type. This warning is only issued if
2924 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2925 typically represent bit patterns, are not warned about.
2928 Usage of ISO string concatenation is detected.
2931 Initialization of automatic aggregates.
2934 Identifier conflicts with labels. Traditional C lacks a separate
2935 namespace for labels.
2938 Initialization of unions. If the initializer is zero, the warning is
2939 omitted. This is done under the assumption that the zero initializer in
2940 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2941 initializer warnings and relies on default initialization to zero in the
2945 Conversions by prototypes between fixed/floating point values and vice
2946 versa. The absence of these prototypes when compiling with traditional
2947 C would cause serious problems. This is a subset of the possible
2948 conversion warnings, for the full set use @option{-Wconversion}.
2951 Use of ISO C style function definitions. This warning intentionally is
2952 @emph{not} issued for prototype declarations or variadic functions
2953 because these ISO C features will appear in your code when using
2954 libiberty's traditional C compatibility macros, @code{PARAMS} and
2955 @code{VPARAMS}. This warning is also bypassed for nested functions
2956 because that feature is already a GCC extension and thus not relevant to
2957 traditional C compatibility.
2960 @item -Wdeclaration-after-statement @r{(C only)}
2961 @opindex Wdeclaration-after-statement
2962 Warn when a declaration is found after a statement in a block. This
2963 construct, known from C++, was introduced with ISO C99 and is by default
2964 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2965 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2969 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2971 @item -Wno-endif-labels
2972 @opindex Wno-endif-labels
2973 @opindex Wendif-labels
2974 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2978 Warn whenever a local variable shadows another local variable, parameter or
2979 global variable or whenever a built-in function is shadowed.
2981 @item -Wlarger-than-@var{len}
2982 @opindex Wlarger-than
2983 Warn whenever an object of larger than @var{len} bytes is defined.
2985 @item -Wpointer-arith
2986 @opindex Wpointer-arith
2987 Warn about anything that depends on the ``size of'' a function type or
2988 of @code{void}. GNU C assigns these types a size of 1, for
2989 convenience in calculations with @code{void *} pointers and pointers
2992 @item -Wbad-function-cast @r{(C only)}
2993 @opindex Wbad-function-cast
2994 Warn whenever a function call is cast to a non-matching type.
2995 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2998 Warn about ISO C constructs that are outside of the common subset of
2999 ISO C and ISO C++, e.g.@: request for implicit conversion from
3000 @code{void *} to a pointer to non-@code{void} type.
3004 Warn whenever a pointer is cast so as to remove a type qualifier from
3005 the target type. For example, warn if a @code{const char *} is cast
3006 to an ordinary @code{char *}.
3009 @opindex Wcast-align
3010 Warn whenever a pointer is cast such that the required alignment of the
3011 target is increased. For example, warn if a @code{char *} is cast to
3012 an @code{int *} on machines where integers can only be accessed at
3013 two- or four-byte boundaries.
3015 @item -Wwrite-strings
3016 @opindex Wwrite-strings
3017 When compiling C, give string constants the type @code{const
3018 char[@var{length}]} so that
3019 copying the address of one into a non-@code{const} @code{char *}
3020 pointer will get a warning; when compiling C++, warn about the
3021 deprecated conversion from string constants to @code{char *}.
3022 These warnings will help you find at
3023 compile time code that can try to write into a string constant, but
3024 only if you have been very careful about using @code{const} in
3025 declarations and prototypes. Otherwise, it will just be a nuisance;
3026 this is why we did not make @option{-Wall} request these warnings.
3029 @opindex Wconversion
3030 Warn if a prototype causes a type conversion that is different from what
3031 would happen to the same argument in the absence of a prototype. This
3032 includes conversions of fixed point to floating and vice versa, and
3033 conversions changing the width or signedness of a fixed point argument
3034 except when the same as the default promotion.
3036 Also, warn if a negative integer constant expression is implicitly
3037 converted to an unsigned type. For example, warn about the assignment
3038 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
3039 casts like @code{(unsigned) -1}.
3041 @item -Wsign-compare
3042 @opindex Wsign-compare
3043 @cindex warning for comparison of signed and unsigned values
3044 @cindex comparison of signed and unsigned values, warning
3045 @cindex signed and unsigned values, comparison warning
3046 Warn when a comparison between signed and unsigned values could produce
3047 an incorrect result when the signed value is converted to unsigned.
3048 This warning is also enabled by @option{-Wextra}; to get the other warnings
3049 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3051 @item -Waggregate-return
3052 @opindex Waggregate-return
3053 Warn if any functions that return structures or unions are defined or
3054 called. (In languages where you can return an array, this also elicits
3057 @item -Wno-attributes
3058 @opindex Wno-attributes
3059 @opindex Wattributes
3060 Do not warn if an unexpected @code{__attribute__} is used, such as
3061 unrecognized attributes, function attributes applied to variables,
3062 etc. This will not stop errors for incorrect use of supported
3065 @item -Wstrict-prototypes @r{(C only)}
3066 @opindex Wstrict-prototypes
3067 Warn if a function is declared or defined without specifying the
3068 argument types. (An old-style function definition is permitted without
3069 a warning if preceded by a declaration which specifies the argument
3072 @item -Wold-style-definition @r{(C only)}
3073 @opindex Wold-style-definition
3074 Warn if an old-style function definition is used. A warning is given
3075 even if there is a previous prototype.
3077 @item -Wmissing-prototypes @r{(C only)}
3078 @opindex Wmissing-prototypes
3079 Warn if a global function is defined without a previous prototype
3080 declaration. This warning is issued even if the definition itself
3081 provides a prototype. The aim is to detect global functions that fail
3082 to be declared in header files.
3084 @item -Wmissing-declarations @r{(C only)}
3085 @opindex Wmissing-declarations
3086 Warn if a global function is defined without a previous declaration.
3087 Do so even if the definition itself provides a prototype.
3088 Use this option to detect global functions that are not declared in
3091 @item -Wmissing-field-initializers
3092 @opindex Wmissing-field-initializers
3095 Warn if a structure's initializer has some fields missing. For
3096 example, the following code would cause such a warning, because
3097 @code{x.h} is implicitly zero:
3100 struct s @{ int f, g, h; @};
3101 struct s x = @{ 3, 4 @};
3104 This option does not warn about designated initializers, so the following
3105 modification would not trigger a warning:
3108 struct s @{ int f, g, h; @};
3109 struct s x = @{ .f = 3, .g = 4 @};
3112 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3113 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3115 @item -Wmissing-noreturn
3116 @opindex Wmissing-noreturn
3117 Warn about functions which might be candidates for attribute @code{noreturn}.
3118 Note these are only possible candidates, not absolute ones. Care should
3119 be taken to manually verify functions actually do not ever return before
3120 adding the @code{noreturn} attribute, otherwise subtle code generation
3121 bugs could be introduced. You will not get a warning for @code{main} in
3122 hosted C environments.
3124 @item -Wmissing-format-attribute
3125 @opindex Wmissing-format-attribute
3127 If @option{-Wformat} is enabled, also warn about functions which might be
3128 candidates for @code{format} attributes. Note these are only possible
3129 candidates, not absolute ones. GCC will guess that @code{format}
3130 attributes might be appropriate for any function that calls a function
3131 like @code{vprintf} or @code{vscanf}, but this might not always be the
3132 case, and some functions for which @code{format} attributes are
3133 appropriate may not be detected. This option has no effect unless
3134 @option{-Wformat} is enabled (possibly by @option{-Wall}).
3136 @item -Wno-multichar
3137 @opindex Wno-multichar
3139 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3140 Usually they indicate a typo in the user's code, as they have
3141 implementation-defined values, and should not be used in portable code.
3143 @item -Wnormalized=<none|id|nfc|nfkc>
3144 @opindex Wnormalized
3147 @cindex character set, input normalization
3148 In ISO C and ISO C++, two identifiers are different if they are
3149 different sequences of characters. However, sometimes when characters
3150 outside the basic ASCII character set are used, you can have two
3151 different character sequences that look the same. To avoid confusion,
3152 the ISO 10646 standard sets out some @dfn{normalization rules} which
3153 when applied ensure that two sequences that look the same are turned into
3154 the same sequence. GCC can warn you if you are using identifiers which
3155 have not been normalized; this option controls that warning.
3157 There are four levels of warning that GCC supports. The default is
3158 @option{-Wnormalized=nfc}, which warns about any identifier which is
3159 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3160 recommended form for most uses.
3162 Unfortunately, there are some characters which ISO C and ISO C++ allow
3163 in identifiers that when turned into NFC aren't allowable as
3164 identifiers. That is, there's no way to use these symbols in portable
3165 ISO C or C++ and have all your identifiers in NFC.
3166 @option{-Wnormalized=id} suppresses the warning for these characters.
3167 It is hoped that future versions of the standards involved will correct
3168 this, which is why this option is not the default.
3170 You can switch the warning off for all characters by writing
3171 @option{-Wnormalized=none}. You would only want to do this if you
3172 were using some other normalization scheme (like ``D''), because
3173 otherwise you can easily create bugs that are literally impossible to see.
3175 Some characters in ISO 10646 have distinct meanings but look identical
3176 in some fonts or display methodologies, especially once formatting has
3177 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3178 LETTER N'', will display just like a regular @code{n} which has been
3179 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3180 normalisation scheme to convert all these into a standard form as
3181 well, and GCC will warn if your code is not in NFKC if you use
3182 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3183 about every identifier that contains the letter O because it might be
3184 confused with the digit 0, and so is not the default, but may be
3185 useful as a local coding convention if the programming environment is
3186 unable to be fixed to display these characters distinctly.
3188 @item -Wno-deprecated-declarations
3189 @opindex Wno-deprecated-declarations
3190 Do not warn about uses of functions, variables, and types marked as
3191 deprecated by using the @code{deprecated} attribute.
3192 (@pxref{Function Attributes}, @pxref{Variable Attributes},
3193 @pxref{Type Attributes}.)
3197 Warn if a structure is given the packed attribute, but the packed
3198 attribute has no effect on the layout or size of the structure.
3199 Such structures may be mis-aligned for little benefit. For
3200 instance, in this code, the variable @code{f.x} in @code{struct bar}
3201 will be misaligned even though @code{struct bar} does not itself
3202 have the packed attribute:
3209 @} __attribute__((packed));
3219 Warn if padding is included in a structure, either to align an element
3220 of the structure or to align the whole structure. Sometimes when this
3221 happens it is possible to rearrange the fields of the structure to
3222 reduce the padding and so make the structure smaller.
3224 @item -Wredundant-decls
3225 @opindex Wredundant-decls
3226 Warn if anything is declared more than once in the same scope, even in
3227 cases where multiple declaration is valid and changes nothing.
3229 @item -Wnested-externs @r{(C only)}
3230 @opindex Wnested-externs
3231 Warn if an @code{extern} declaration is encountered within a function.
3233 @item -Wunreachable-code
3234 @opindex Wunreachable-code
3235 Warn if the compiler detects that code will never be executed.
3237 This option is intended to warn when the compiler detects that at
3238 least a whole line of source code will never be executed, because
3239 some condition is never satisfied or because it is after a
3240 procedure that never returns.
3242 It is possible for this option to produce a warning even though there
3243 are circumstances under which part of the affected line can be executed,
3244 so care should be taken when removing apparently-unreachable code.
3246 For instance, when a function is inlined, a warning may mean that the
3247 line is unreachable in only one inlined copy of the function.
3249 This option is not made part of @option{-Wall} because in a debugging
3250 version of a program there is often substantial code which checks
3251 correct functioning of the program and is, hopefully, unreachable
3252 because the program does work. Another common use of unreachable
3253 code is to provide behavior which is selectable at compile-time.
3257 Warn if a function can not be inlined and it was declared as inline.
3258 Even with this option, the compiler will not warn about failures to
3259 inline functions declared in system headers.
3261 The compiler uses a variety of heuristics to determine whether or not
3262 to inline a function. For example, the compiler takes into account
3263 the size of the function being inlined and the amount of inlining
3264 that has already been done in the current function. Therefore,
3265 seemingly insignificant changes in the source program can cause the
3266 warnings produced by @option{-Winline} to appear or disappear.
3268 @item -Wno-invalid-offsetof @r{(C++ only)}
3269 @opindex Wno-invalid-offsetof
3270 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3271 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3272 to a non-POD type is undefined. In existing C++ implementations,
3273 however, @samp{offsetof} typically gives meaningful results even when
3274 applied to certain kinds of non-POD types. (Such as a simple
3275 @samp{struct} that fails to be a POD type only by virtue of having a
3276 constructor.) This flag is for users who are aware that they are
3277 writing nonportable code and who have deliberately chosen to ignore the
3280 The restrictions on @samp{offsetof} may be relaxed in a future version
3281 of the C++ standard.
3283 @item -Wno-int-to-pointer-cast @r{(C only)}
3284 @opindex Wno-int-to-pointer-cast
3285 Suppress warnings from casts to pointer type of an integer of a
3288 @item -Wno-pointer-to-int-cast @r{(C only)}
3289 @opindex Wno-pointer-to-int-cast
3290 Suppress warnings from casts from a pointer to an integer type of a
3294 @opindex Winvalid-pch
3295 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3296 the search path but can't be used.
3300 @opindex Wno-long-long
3301 Warn if @samp{long long} type is used. This is default. To inhibit
3302 the warning messages, use @option{-Wno-long-long}. Flags
3303 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3304 only when @option{-pedantic} flag is used.
3306 @item -Wvariadic-macros
3307 @opindex Wvariadic-macros
3308 @opindex Wno-variadic-macros
3309 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3310 alternate syntax when in pedantic ISO C99 mode. This is default.
3311 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3313 @item -Wdisabled-optimization
3314 @opindex Wdisabled-optimization
3315 Warn if a requested optimization pass is disabled. This warning does
3316 not generally indicate that there is anything wrong with your code; it
3317 merely indicates that GCC's optimizers were unable to handle the code
3318 effectively. Often, the problem is that your code is too big or too
3319 complex; GCC will refuse to optimize programs when the optimization
3320 itself is likely to take inordinate amounts of time.
3322 @item -Wno-pointer-sign
3323 @opindex Wno-pointer-sign
3324 Don't warn for pointer argument passing or assignment with different signedness.
3325 Only useful in the negative form since this warning is enabled by default.
3326 This option is only supported for C and Objective-C@.
3330 Make all warnings into errors.
3333 @node Debugging Options
3334 @section Options for Debugging Your Program or GCC
3335 @cindex options, debugging
3336 @cindex debugging information options
3338 GCC has various special options that are used for debugging
3339 either your program or GCC:
3344 Produce debugging information in the operating system's native format
3345 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3348 On most systems that use stabs format, @option{-g} enables use of extra
3349 debugging information that only GDB can use; this extra information
3350 makes debugging work better in GDB but will probably make other debuggers
3352 refuse to read the program. If you want to control for certain whether
3353 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3354 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3356 GCC allows you to use @option{-g} with
3357 @option{-O}. The shortcuts taken by optimized code may occasionally
3358 produce surprising results: some variables you declared may not exist
3359 at all; flow of control may briefly move where you did not expect it;
3360 some statements may not be executed because they compute constant
3361 results or their values were already at hand; some statements may
3362 execute in different places because they were moved out of loops.
3364 Nevertheless it proves possible to debug optimized output. This makes
3365 it reasonable to use the optimizer for programs that might have bugs.
3367 The following options are useful when GCC is generated with the
3368 capability for more than one debugging format.
3372 Produce debugging information for use by GDB@. This means to use the
3373 most expressive format available (DWARF 2, stabs, or the native format
3374 if neither of those are supported), including GDB extensions if at all
3379 Produce debugging information in stabs format (if that is supported),
3380 without GDB extensions. This is the format used by DBX on most BSD
3381 systems. On MIPS, Alpha and System V Release 4 systems this option
3382 produces stabs debugging output which is not understood by DBX or SDB@.
3383 On System V Release 4 systems this option requires the GNU assembler.
3385 @item -feliminate-unused-debug-symbols
3386 @opindex feliminate-unused-debug-symbols
3387 Produce debugging information in stabs format (if that is supported),
3388 for only symbols that are actually used.
3392 Produce debugging information in stabs format (if that is supported),
3393 using GNU extensions understood only by the GNU debugger (GDB)@. The
3394 use of these extensions is likely to make other debuggers crash or
3395 refuse to read the program.
3399 Produce debugging information in COFF format (if that is supported).
3400 This is the format used by SDB on most System V systems prior to
3405 Produce debugging information in XCOFF format (if that is supported).
3406 This is the format used by the DBX debugger on IBM RS/6000 systems.
3410 Produce debugging information in XCOFF format (if that is supported),
3411 using GNU extensions understood only by the GNU debugger (GDB)@. The
3412 use of these extensions is likely to make other debuggers crash or
3413 refuse to read the program, and may cause assemblers other than the GNU
3414 assembler (GAS) to fail with an error.
3418 Produce debugging information in DWARF version 2 format (if that is
3419 supported). This is the format used by DBX on IRIX 6. With this
3420 option, GCC uses features of DWARF version 3 when they are useful;
3421 version 3 is upward compatible with version 2, but may still cause
3422 problems for older debuggers.
3426 Produce debugging information in VMS debug format (if that is
3427 supported). This is the format used by DEBUG on VMS systems.
3430 @itemx -ggdb@var{level}
3431 @itemx -gstabs@var{level}
3432 @itemx -gcoff@var{level}
3433 @itemx -gxcoff@var{level}
3434 @itemx -gvms@var{level}
3435 Request debugging information and also use @var{level} to specify how
3436 much information. The default level is 2.
3438 Level 1 produces minimal information, enough for making backtraces in
3439 parts of the program that you don't plan to debug. This includes
3440 descriptions of functions and external variables, but no information
3441 about local variables and no line numbers.
3443 Level 3 includes extra information, such as all the macro definitions
3444 present in the program. Some debuggers support macro expansion when
3445 you use @option{-g3}.
3447 @option{-gdwarf-2} does not accept a concatenated debug level, because
3448 GCC used to support an option @option{-gdwarf} that meant to generate
3449 debug information in version 1 of the DWARF format (which is very
3450 different from version 2), and it would have been too confusing. That
3451 debug format is long obsolete, but the option cannot be changed now.
3452 Instead use an additional @option{-g@var{level}} option to change the
3453 debug level for DWARF2.
3455 @item -feliminate-dwarf2-dups
3456 @opindex feliminate-dwarf2-dups
3457 Compress DWARF2 debugging information by eliminating duplicated
3458 information about each symbol. This option only makes sense when
3459 generating DWARF2 debugging information with @option{-gdwarf-2}.
3461 @cindex @command{prof}
3464 Generate extra code to write profile information suitable for the
3465 analysis program @command{prof}. You must use this option when compiling
3466 the source files you want data about, and you must also use it when
3469 @cindex @command{gprof}
3472 Generate extra code to write profile information suitable for the
3473 analysis program @command{gprof}. You must use this option when compiling
3474 the source files you want data about, and you must also use it when
3479 Makes the compiler print out each function name as it is compiled, and
3480 print some statistics about each pass when it finishes.
3483 @opindex ftime-report
3484 Makes the compiler print some statistics about the time consumed by each
3485 pass when it finishes.
3488 @opindex fmem-report
3489 Makes the compiler print some statistics about permanent memory
3490 allocation when it finishes.
3492 @item -fprofile-arcs
3493 @opindex fprofile-arcs
3494 Add code so that program flow @dfn{arcs} are instrumented. During
3495 execution the program records how many times each branch and call is
3496 executed and how many times it is taken or returns. When the compiled
3497 program exits it saves this data to a file called
3498 @file{@var{auxname}.gcda} for each source file. The data may be used for
3499 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3500 test coverage analysis (@option{-ftest-coverage}). Each object file's
3501 @var{auxname} is generated from the name of the output file, if
3502 explicitly specified and it is not the final executable, otherwise it is
3503 the basename of the source file. In both cases any suffix is removed
3504 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3505 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3506 @xref{Cross-profiling}.
3508 @cindex @command{gcov}
3512 This option is used to compile and link code instrumented for coverage
3513 analysis. The option is a synonym for @option{-fprofile-arcs}
3514 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3515 linking). See the documentation for those options for more details.
3520 Compile the source files with @option{-fprofile-arcs} plus optimization
3521 and code generation options. For test coverage analysis, use the
3522 additional @option{-ftest-coverage} option. You do not need to profile
3523 every source file in a program.
3526 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3527 (the latter implies the former).
3530 Run the program on a representative workload to generate the arc profile
3531 information. This may be repeated any number of times. You can run
3532 concurrent instances of your program, and provided that the file system
3533 supports locking, the data files will be correctly updated. Also
3534 @code{fork} calls are detected and correctly handled (double counting
3538 For profile-directed optimizations, compile the source files again with
3539 the same optimization and code generation options plus
3540 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3541 Control Optimization}).
3544 For test coverage analysis, use @command{gcov} to produce human readable
3545 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3546 @command{gcov} documentation for further information.
3550 With @option{-fprofile-arcs}, for each function of your program GCC
3551 creates a program flow graph, then finds a spanning tree for the graph.
3552 Only arcs that are not on the spanning tree have to be instrumented: the
3553 compiler adds code to count the number of times that these arcs are
3554 executed. When an arc is the only exit or only entrance to a block, the
3555 instrumentation code can be added to the block; otherwise, a new basic
3556 block must be created to hold the instrumentation code.
3558 @item -ftree-based-profiling
3559 @opindex ftree-based-profiling
3560 This option is used in addition to @option{-fprofile-arcs} or
3561 @option{-fbranch-probabilities} to control whether those optimizations
3562 are performed on a tree-based or rtl-based internal representation.
3563 If you use this option when compiling with @option{-fprofile-arcs},
3564 you must also use it when compiling later with @option{-fbranch-probabilities}.
3565 Currently the tree-based optimization is in an early stage of
3566 development, and this option is recommended only for those people
3567 working on improving it.
3570 @item -ftest-coverage
3571 @opindex ftest-coverage
3572 Produce a notes file that the @command{gcov} code-coverage utility
3573 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3574 show program coverage. Each source file's note file is called
3575 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3576 above for a description of @var{auxname} and instructions on how to
3577 generate test coverage data. Coverage data will match the source files
3578 more closely, if you do not optimize.
3580 @item -d@var{letters}
3581 @item -fdump-rtl-@var{pass}
3583 Says to make debugging dumps during compilation at times specified by
3584 @var{letters}. This is used for debugging the RTL-based passes of the
3585 compiler. The file names for most of the dumps are made by appending a
3586 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3587 from the name of the output file, if explicitly specified and it is not
3588 an executable, otherwise it is the basename of the source file.
3590 Most debug dumps can be enabled either passing a letter to the @option{-d}
3591 option, or with a long @option{-fdump-rtl} switch; here are the possible
3592 letters for use in @var{letters} and @var{pass}, and their meanings:
3597 Annotate the assembler output with miscellaneous debugging information.
3600 @itemx -fdump-rtl-bp
3602 @opindex fdump-rtl-bp
3603 Dump after computing branch probabilities, to @file{@var{file}.09.bp}.
3606 @itemx -fdump-rtl-bbro
3608 @opindex fdump-rtl-bbro
3609 Dump after block reordering, to @file{@var{file}.30.bbro}.
3612 @itemx -fdump-rtl-combine
3614 @opindex fdump-rtl-combine
3615 Dump after instruction combination, to the file @file{@var{file}.17.combine}.
3618 @itemx -fdump-rtl-ce1
3619 @itemx -fdump-rtl-ce2
3621 @opindex fdump-rtl-ce1
3622 @opindex fdump-rtl-ce2
3623 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3624 first if conversion, to the file @file{@var{file}.11.ce1}. @option{-dC}
3625 and @option{-fdump-rtl-ce2} enable dumping after the second if
3626 conversion, to the file @file{@var{file}.18.ce2}.
3629 @itemx -fdump-rtl-btl
3630 @itemx -fdump-rtl-dbr
3632 @opindex fdump-rtl-btl
3633 @opindex fdump-rtl-dbr
3634 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3635 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3636 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3637 scheduling, to @file{@var{file}.36.dbr}.
3641 Dump all macro definitions, at the end of preprocessing, in addition to
3645 @itemx -fdump-rtl-ce3
3647 @opindex fdump-rtl-ce3
3648 Dump after the third if conversion, to @file{@var{file}.28.ce3}.
3651 @itemx -fdump-rtl-cfg
3652 @itemx -fdump-rtl-life
3654 @opindex fdump-rtl-cfg
3655 @opindex fdump-rtl-life
3656 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3657 and data flow analysis, to @file{@var{file}.08.cfg}. @option{-df}
3658 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3659 to @file{@var{file}.16.life}.
3662 @itemx -fdump-rtl-greg
3664 @opindex fdump-rtl-greg
3665 Dump after global register allocation, to @file{@var{file}.23.greg}.
3668 @itemx -fdump-rtl-gcse
3669 @itemx -fdump-rtl-bypass
3671 @opindex fdump-rtl-gcse
3672 @opindex fdump-rtl-bypass
3673 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3674 @file{@var{file}.05.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3675 enable dumping after jump bypassing and control flow optimizations, to
3676 @file{@var{file}.07.bypass}.
3679 @itemx -fdump-rtl-eh
3681 @opindex fdump-rtl-eh
3682 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3685 @itemx -fdump-rtl-sibling
3687 @opindex fdump-rtl-sibling
3688 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
3691 @itemx -fdump-rtl-jump
3693 @opindex fdump-rtl-jump
3694 Dump after the first jump optimization, to @file{@var{file}.03.jump}.
3697 @itemx -fdump-rtl-stack
3699 @opindex fdump-rtl-stack
3700 Dump after conversion from registers to stack, to @file{@var{file}.33.stack}.
3703 @itemx -fdump-rtl-lreg
3705 @opindex fdump-rtl-lreg
3706 Dump after local register allocation, to @file{@var{file}.22.lreg}.
3709 @itemx -fdump-rtl-loop
3710 @itemx -fdump-rtl-loop2
3712 @opindex fdump-rtl-loop
3713 @opindex fdump-rtl-loop2
3714 @option{-dL} and @option{-fdump-rtl-loop} enable dumping after the first
3715 loop optimization pass, to @file{@var{file}.06.loop}. @option{-dL} and
3716 @option{-fdump-rtl-loop2} enable dumping after the second pass, to
3717 @file{@var{file}.13.loop2}.
3720 @itemx -fdump-rtl-sms
3722 @opindex fdump-rtl-sms
3723 Dump after modulo scheduling, to @file{@var{file}.20.sms}.
3726 @itemx -fdump-rtl-mach
3728 @opindex fdump-rtl-mach
3729 Dump after performing the machine dependent reorganization pass, to
3730 @file{@var{file}.35.mach}.
3733 @itemx -fdump-rtl-rnreg
3735 @opindex fdump-rtl-rnreg
3736 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3739 @itemx -fdump-rtl-regmove
3741 @opindex fdump-rtl-regmove
3742 Dump after the register move pass, to @file{@var{file}.19.regmove}.
3745 @itemx -fdump-rtl-postreload
3747 @opindex fdump-rtl-postreload
3748 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3751 @itemx -fdump-rtl-expand
3753 @opindex fdump-rtl-expand
3754 Dump after RTL generation, to @file{@var{file}.00.expand}.
3757 @itemx -fdump-rtl-sched2
3759 @opindex fdump-rtl-sched2
3760 Dump after the second scheduling pass, to @file{@var{file}.32.sched2}.
3763 @itemx -fdump-rtl-cse
3765 @opindex fdump-rtl-cse
3766 Dump after CSE (including the jump optimization that sometimes follows
3767 CSE), to @file{@var{file}.04.cse}.
3770 @itemx -fdump-rtl-sched
3772 @opindex fdump-rtl-sched
3773 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3776 @itemx -fdump-rtl-cse2
3778 @opindex fdump-rtl-cse2
3779 Dump after the second CSE pass (including the jump optimization that
3780 sometimes follows CSE), to @file{@var{file}.15.cse2}.
3783 @itemx -fdump-rtl-tracer
3785 @opindex fdump-rtl-tracer
3786 Dump after running tracer, to @file{@var{file}.12.tracer}.
3789 @itemx -fdump-rtl-vpt
3790 @itemx -fdump-rtl-vartrack
3792 @opindex fdump-rtl-vpt
3793 @opindex fdump-rtl-vartrack
3794 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3795 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3796 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3797 to @file{@var{file}.34.vartrack}.
3800 @itemx -fdump-rtl-flow2
3802 @opindex fdump-rtl-flow2
3803 Dump after the second flow pass, to @file{@var{file}.26.flow2}.
3806 @itemx -fdump-rtl-peephole2
3808 @opindex fdump-rtl-peephole2
3809 Dump after the peephole pass, to @file{@var{file}.27.peephole2}.
3812 @itemx -fdump-rtl-web
3814 @opindex fdump-rtl-web
3815 Dump after live range splitting, to @file{@var{file}.14.web}.
3818 @itemx -fdump-rtl-all
3820 @opindex fdump-rtl-all
3821 Produce all the dumps listed above.
3825 Produce a core dump whenever an error occurs.
3829 Print statistics on memory usage, at the end of the run, to
3834 Annotate the assembler output with a comment indicating which
3835 pattern and alternative was used. The length of each instruction is
3840 Dump the RTL in the assembler output as a comment before each instruction.
3841 Also turns on @option{-dp} annotation.
3845 For each of the other indicated dump files (either with @option{-d} or
3846 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3847 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3851 Just generate RTL for a function instead of compiling it. Usually used
3852 with @samp{r} (@option{-fdump-rtl-expand}).
3856 Dump debugging information during parsing, to standard error.
3859 @item -fdump-unnumbered
3860 @opindex fdump-unnumbered
3861 When doing debugging dumps (see @option{-d} option above), suppress instruction
3862 numbers and line number note output. This makes it more feasible to
3863 use diff on debugging dumps for compiler invocations with different
3864 options, in particular with and without @option{-g}.
3866 @item -fdump-translation-unit @r{(C++ only)}
3867 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
3868 @opindex fdump-translation-unit
3869 Dump a representation of the tree structure for the entire translation
3870 unit to a file. The file name is made by appending @file{.tu} to the
3871 source file name. If the @samp{-@var{options}} form is used, @var{options}
3872 controls the details of the dump as described for the
3873 @option{-fdump-tree} options.
3875 @item -fdump-class-hierarchy @r{(C++ only)}
3876 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3877 @opindex fdump-class-hierarchy
3878 Dump a representation of each class's hierarchy and virtual function
3879 table layout to a file. The file name is made by appending @file{.class}
3880 to the source file name. If the @samp{-@var{options}} form is used,
3881 @var{options} controls the details of the dump as described for the
3882 @option{-fdump-tree} options.
3884 @item -fdump-ipa-@var{switch}
3886 Control the dumping at various stages of inter-procedural analysis
3887 language tree to a file. The file name is generated by appending a switch
3888 specific suffix to the source file name. The following dumps are possible:
3892 Enables all inter-procedural analysis dumps; currently the only produced
3893 dump is the @samp{cgraph} dump.
3896 Dumps information about call-graph optimization, unused function removal,
3897 and inlining decisions.
3900 @item -fdump-tree-@var{switch} @r{(C and C++ only)}
3901 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C and C++ only)}
3903 Control the dumping at various stages of processing the intermediate
3904 language tree to a file. The file name is generated by appending a switch
3905 specific suffix to the source file name. If the @samp{-@var{options}}
3906 form is used, @var{options} is a list of @samp{-} separated options that
3907 control the details of the dump. Not all options are applicable to all
3908 dumps, those which are not meaningful will be ignored. The following
3909 options are available
3913 Print the address of each node. Usually this is not meaningful as it
3914 changes according to the environment and source file. Its primary use
3915 is for tying up a dump file with a debug environment.
3917 Inhibit dumping of members of a scope or body of a function merely
3918 because that scope has been reached. Only dump such items when they
3919 are directly reachable by some other path. When dumping pretty-printed
3920 trees, this option inhibits dumping the bodies of control structures.
3922 Print a raw representation of the tree. By default, trees are
3923 pretty-printed into a C-like representation.
3925 Enable more detailed dumps (not honored by every dump option).
3927 Enable dumping various statistics about the pass (not honored by every dump
3930 Enable showing basic block boundaries (disabled in raw dumps).
3932 Enable showing virtual operands for every statement.
3934 Enable showing line numbers for statements.
3936 Enable showing the unique ID (@code{DECL_UID}) for each variable.
3938 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
3941 The following tree dumps are possible:
3945 Dump before any tree based optimization, to @file{@var{file}.original}.
3948 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3951 Dump after function inlining, to @file{@var{file}.inlined}.
3954 @opindex fdump-tree-gimple
3955 Dump each function before and after the gimplification pass to a file. The
3956 file name is made by appending @file{.gimple} to the source file name.
3959 @opindex fdump-tree-cfg
3960 Dump the control flow graph of each function to a file. The file name is
3961 made by appending @file{.cfg} to the source file name.
3964 @opindex fdump-tree-vcg
3965 Dump the control flow graph of each function to a file in VCG format. The
3966 file name is made by appending @file{.vcg} to the source file name. Note
3967 that if the file contains more than one function, the generated file cannot
3968 be used directly by VCG@. You will need to cut and paste each function's
3969 graph into its own separate file first.
3972 @opindex fdump-tree-ch
3973 Dump each function after copying loop headers. The file name is made by
3974 appending @file{.ch} to the source file name.
3977 @opindex fdump-tree-ssa
3978 Dump SSA related information to a file. The file name is made by appending
3979 @file{.ssa} to the source file name.
3982 @opindex fdump-tree-salias
3983 Dump structure aliasing variable information to a file. This file name
3984 is made by appending @file{.salias} to the source file name.
3987 @opindex fdump-tree-alias
3988 Dump aliasing information for each function. The file name is made by
3989 appending @file{.alias} to the source file name.
3992 @opindex fdump-tree-ccp
3993 Dump each function after CCP@. The file name is made by appending
3994 @file{.ccp} to the source file name.
3997 @opindex fdump-tree-storeccp
3998 Dump each function after STORE-CCP. The file name is made by appending
3999 @file{.storeccp} to the source file name.
4002 @opindex fdump-tree-pre
4003 Dump trees after partial redundancy elimination. The file name is made
4004 by appending @file{.pre} to the source file name.
4007 @opindex fdump-tree-fre
4008 Dump trees after full redundancy elimination. The file name is made
4009 by appending @file{.fre} to the source file name.
4012 @opindex fdump-tree-copyprop
4013 Dump trees after copy propagation. The file name is made
4014 by appending @file{.copyprop} to the source file name.
4016 @item store_copyprop
4017 @opindex fdump-tree-store_copyprop
4018 Dump trees after store copy-propagation. The file name is made
4019 by appending @file{.store_copyprop} to the source file name.
4022 @opindex fdump-tree-dce
4023 Dump each function after dead code elimination. The file name is made by
4024 appending @file{.dce} to the source file name.
4027 @opindex fdump-tree-mudflap
4028 Dump each function after adding mudflap instrumentation. The file name is
4029 made by appending @file{.mudflap} to the source file name.
4032 @opindex fdump-tree-sra
4033 Dump each function after performing scalar replacement of aggregates. The
4034 file name is made by appending @file{.sra} to the source file name.
4037 @opindex fdump-tree-sink
4038 Dump each function after performing code sinking. The file name is made
4039 by appending @file{.sink} to the source file name.
4042 @opindex fdump-tree-dom
4043 Dump each function after applying dominator tree optimizations. The file
4044 name is made by appending @file{.dom} to the source file name.
4047 @opindex fdump-tree-dse
4048 Dump each function after applying dead store elimination. The file
4049 name is made by appending @file{.dse} to the source file name.
4052 @opindex fdump-tree-phiopt
4053 Dump each function after optimizing PHI nodes into straightline code. The file
4054 name is made by appending @file{.phiopt} to the source file name.
4057 @opindex fdump-tree-forwprop
4058 Dump each function after forward propagating single use variables. The file
4059 name is made by appending @file{.forwprop} to the source file name.
4062 @opindex fdump-tree-copyrename
4063 Dump each function after applying the copy rename optimization. The file
4064 name is made by appending @file{.copyrename} to the source file name.
4067 @opindex fdump-tree-nrv
4068 Dump each function after applying the named return value optimization on
4069 generic trees. The file name is made by appending @file{.nrv} to the source
4073 @opindex fdump-tree-vect
4074 Dump each function after applying vectorization of loops. The file name is
4075 made by appending @file{.vect} to the source file name.
4078 @opindex fdump-tree-vrp
4079 Dump each function after Value Range Propagation (VRP). The file name
4080 is made by appending @file{.vrp} to the source file name.
4083 @opindex fdump-tree-all
4084 Enable all the available tree dumps with the flags provided in this option.
4087 @item -ftree-vectorizer-verbose=@var{n}
4088 @opindex ftree-vectorizer-verbose
4089 This option controls the amount of debugging output the vectorizer prints.
4090 This information is written to standard error, unless @option{-fdump-tree-all}
4091 or @option{-fdump-tree-vect} is specified, in which case it is output to the
4092 usual dump listing file, @file{.vect}.
4094 @item -frandom-seed=@var{string}
4095 @opindex frandom-string
4096 This option provides a seed that GCC uses when it would otherwise use
4097 random numbers. It is used to generate certain symbol names
4098 that have to be different in every compiled file. It is also used to
4099 place unique stamps in coverage data files and the object files that
4100 produce them. You can use the @option{-frandom-seed} option to produce
4101 reproducibly identical object files.
4103 The @var{string} should be different for every file you compile.
4105 @item -fsched-verbose=@var{n}
4106 @opindex fsched-verbose
4107 On targets that use instruction scheduling, this option controls the
4108 amount of debugging output the scheduler prints. This information is
4109 written to standard error, unless @option{-dS} or @option{-dR} is
4110 specified, in which case it is output to the usual dump
4111 listing file, @file{.sched} or @file{.sched2} respectively. However
4112 for @var{n} greater than nine, the output is always printed to standard
4115 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4116 same information as @option{-dRS}. For @var{n} greater than one, it
4117 also output basic block probabilities, detailed ready list information
4118 and unit/insn info. For @var{n} greater than two, it includes RTL
4119 at abort point, control-flow and regions info. And for @var{n} over
4120 four, @option{-fsched-verbose} also includes dependence info.
4124 Store the usual ``temporary'' intermediate files permanently; place them
4125 in the current directory and name them based on the source file. Thus,
4126 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4127 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4128 preprocessed @file{foo.i} output file even though the compiler now
4129 normally uses an integrated preprocessor.
4131 When used in combination with the @option{-x} command line option,
4132 @option{-save-temps} is sensible enough to avoid over writing an
4133 input source file with the same extension as an intermediate file.
4134 The corresponding intermediate file may be obtained by renaming the
4135 source file before using @option{-save-temps}.
4139 Report the CPU time taken by each subprocess in the compilation
4140 sequence. For C source files, this is the compiler proper and assembler
4141 (plus the linker if linking is done). The output looks like this:
4148 The first number on each line is the ``user time'', that is time spent
4149 executing the program itself. The second number is ``system time'',
4150 time spent executing operating system routines on behalf of the program.
4151 Both numbers are in seconds.
4153 @item -fvar-tracking
4154 @opindex fvar-tracking
4155 Run variable tracking pass. It computes where variables are stored at each
4156 position in code. Better debugging information is then generated
4157 (if the debugging information format supports this information).
4159 It is enabled by default when compiling with optimization (@option{-Os},
4160 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4161 the debug info format supports it.
4163 @item -print-file-name=@var{library}
4164 @opindex print-file-name
4165 Print the full absolute name of the library file @var{library} that
4166 would be used when linking---and don't do anything else. With this
4167 option, GCC does not compile or link anything; it just prints the
4170 @item -print-multi-directory
4171 @opindex print-multi-directory
4172 Print the directory name corresponding to the multilib selected by any
4173 other switches present in the command line. This directory is supposed
4174 to exist in @env{GCC_EXEC_PREFIX}.
4176 @item -print-multi-lib
4177 @opindex print-multi-lib
4178 Print the mapping from multilib directory names to compiler switches
4179 that enable them. The directory name is separated from the switches by
4180 @samp{;}, and each switch starts with an @samp{@@} instead of the
4181 @samp{-}, without spaces between multiple switches. This is supposed to
4182 ease shell-processing.
4184 @item -print-prog-name=@var{program}
4185 @opindex print-prog-name
4186 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4188 @item -print-libgcc-file-name
4189 @opindex print-libgcc-file-name
4190 Same as @option{-print-file-name=libgcc.a}.
4192 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4193 but you do want to link with @file{libgcc.a}. You can do
4196 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4199 @item -print-search-dirs
4200 @opindex print-search-dirs
4201 Print the name of the configured installation directory and a list of
4202 program and library directories @command{gcc} will search---and don't do anything else.
4204 This is useful when @command{gcc} prints the error message
4205 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4206 To resolve this you either need to put @file{cpp0} and the other compiler
4207 components where @command{gcc} expects to find them, or you can set the environment
4208 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4209 Don't forget the trailing @samp{/}.
4210 @xref{Environment Variables}.
4213 @opindex dumpmachine
4214 Print the compiler's target machine (for example,
4215 @samp{i686-pc-linux-gnu})---and don't do anything else.
4218 @opindex dumpversion
4219 Print the compiler version (for example, @samp{3.0})---and don't do
4224 Print the compiler's built-in specs---and don't do anything else. (This
4225 is used when GCC itself is being built.) @xref{Spec Files}.
4227 @item -feliminate-unused-debug-types
4228 @opindex feliminate-unused-debug-types
4229 Normally, when producing DWARF2 output, GCC will emit debugging
4230 information for all types declared in a compilation
4231 unit, regardless of whether or not they are actually used
4232 in that compilation unit. Sometimes this is useful, such as
4233 if, in the debugger, you want to cast a value to a type that is
4234 not actually used in your program (but is declared). More often,
4235 however, this results in a significant amount of wasted space.
4236 With this option, GCC will avoid producing debug symbol output
4237 for types that are nowhere used in the source file being compiled.
4240 @node Optimize Options
4241 @section Options That Control Optimization
4242 @cindex optimize options
4243 @cindex options, optimization
4245 These options control various sorts of optimizations.
4247 Without any optimization option, the compiler's goal is to reduce the
4248 cost of compilation and to make debugging produce the expected
4249 results. Statements are independent: if you stop the program with a
4250 breakpoint between statements, you can then assign a new value to any
4251 variable or change the program counter to any other statement in the
4252 function and get exactly the results you would expect from the source
4255 Turning on optimization flags makes the compiler attempt to improve
4256 the performance and/or code size at the expense of compilation time
4257 and possibly the ability to debug the program.
4259 The compiler performs optimization based on the knowledge it has of
4260 the program. Optimization levels @option{-O2} and above, in
4261 particular, enable @emph{unit-at-a-time} mode, which allows the
4262 compiler to consider information gained from later functions in
4263 the file when compiling a function. Compiling multiple files at
4264 once to a single output file in @emph{unit-at-a-time} mode allows
4265 the compiler to use information gained from all of the files when
4266 compiling each of them.
4268 Not all optimizations are controlled directly by a flag. Only
4269 optimizations that have a flag are listed.
4276 Optimize. Optimizing compilation takes somewhat more time, and a lot
4277 more memory for a large function.
4279 With @option{-O}, the compiler tries to reduce code size and execution
4280 time, without performing any optimizations that take a great deal of
4283 @option{-O} turns on the following optimization flags:
4284 @gccoptlist{-fdefer-pop @gol
4285 -fdelayed-branch @gol
4286 -fguess-branch-probability @gol
4287 -fcprop-registers @gol
4288 -floop-optimize @gol
4289 -fif-conversion @gol
4290 -fif-conversion2 @gol
4293 -ftree-dominator-opts @gol
4298 -ftree-copyrename @gol
4303 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4304 where doing so does not interfere with debugging.
4308 Optimize even more. GCC performs nearly all supported optimizations
4309 that do not involve a space-speed tradeoff. The compiler does not
4310 perform loop unrolling or function inlining when you specify @option{-O2}.
4311 As compared to @option{-O}, this option increases both compilation time
4312 and the performance of the generated code.
4314 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4315 also turns on the following optimization flags:
4316 @gccoptlist{-fthread-jumps @gol
4318 -foptimize-sibling-calls @gol
4319 -fcse-follow-jumps -fcse-skip-blocks @gol
4320 -fgcse -fgcse-lm @gol
4321 -fexpensive-optimizations @gol
4322 -fstrength-reduce @gol
4323 -frerun-cse-after-loop -frerun-loop-opt @gol
4327 -fschedule-insns -fschedule-insns2 @gol
4328 -fsched-interblock -fsched-spec @gol
4330 -fstrict-aliasing @gol
4331 -fdelete-null-pointer-checks @gol
4332 -freorder-blocks -freorder-functions @gol
4333 -funit-at-a-time @gol
4334 -falign-functions -falign-jumps @gol
4335 -falign-loops -falign-labels @gol
4339 Please note the warning under @option{-fgcse} about
4340 invoking @option{-O2} on programs that use computed gotos.
4344 Optimize yet more. @option{-O3} turns on all optimizations specified by
4345 @option{-O2} and also turns on the @option{-finline-functions},
4346 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4350 Do not optimize. This is the default.
4354 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4355 do not typically increase code size. It also performs further
4356 optimizations designed to reduce code size.
4358 @option{-Os} disables the following optimization flags:
4359 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4360 -falign-labels -freorder-blocks -freorder-blocks-and-partition -fprefetch-loop-arrays}
4362 If you use multiple @option{-O} options, with or without level numbers,
4363 the last such option is the one that is effective.
4366 Options of the form @option{-f@var{flag}} specify machine-independent
4367 flags. Most flags have both positive and negative forms; the negative
4368 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4369 below, only one of the forms is listed---the one you typically will
4370 use. You can figure out the other form by either removing @samp{no-}
4373 The following options control specific optimizations. They are either
4374 activated by @option{-O} options or are related to ones that are. You
4375 can use the following flags in the rare cases when ``fine-tuning'' of
4376 optimizations to be performed is desired.
4379 @item -fno-default-inline
4380 @opindex fno-default-inline
4381 Do not make member functions inline by default merely because they are
4382 defined inside the class scope (C++ only). Otherwise, when you specify
4383 @w{@option{-O}}, member functions defined inside class scope are compiled
4384 inline by default; i.e., you don't need to add @samp{inline} in front of
4385 the member function name.
4387 @item -fno-defer-pop
4388 @opindex fno-defer-pop
4389 Always pop the arguments to each function call as soon as that function
4390 returns. For machines which must pop arguments after a function call,
4391 the compiler normally lets arguments accumulate on the stack for several
4392 function calls and pops them all at once.
4394 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4398 Force memory operands to be copied into registers before doing
4399 arithmetic on them. This produces better code by making all memory
4400 references potential common subexpressions. When they are not common
4401 subexpressions, instruction combination should eliminate the separate
4404 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4407 @opindex fforce-addr
4408 Force memory address constants to be copied into registers before
4409 doing arithmetic on them. This may produce better code just as
4410 @option{-fforce-mem} may.
4412 @item -fomit-frame-pointer
4413 @opindex fomit-frame-pointer
4414 Don't keep the frame pointer in a register for functions that
4415 don't need one. This avoids the instructions to save, set up and
4416 restore frame pointers; it also makes an extra register available
4417 in many functions. @strong{It also makes debugging impossible on
4420 On some machines, such as the VAX, this flag has no effect, because
4421 the standard calling sequence automatically handles the frame pointer
4422 and nothing is saved by pretending it doesn't exist. The
4423 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4424 whether a target machine supports this flag. @xref{Registers,,Register
4425 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4427 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4429 @item -foptimize-sibling-calls
4430 @opindex foptimize-sibling-calls
4431 Optimize sibling and tail recursive calls.
4433 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4437 Don't pay attention to the @code{inline} keyword. Normally this option
4438 is used to keep the compiler from expanding any functions inline.
4439 Note that if you are not optimizing, no functions can be expanded inline.
4441 @item -finline-functions
4442 @opindex finline-functions
4443 Integrate all simple functions into their callers. The compiler
4444 heuristically decides which functions are simple enough to be worth
4445 integrating in this way.
4447 If all calls to a given function are integrated, and the function is
4448 declared @code{static}, then the function is normally not output as
4449 assembler code in its own right.
4451 Enabled at level @option{-O3}.
4453 @item -finline-limit=@var{n}
4454 @opindex finline-limit
4455 By default, GCC limits the size of functions that can be inlined. This flag
4456 allows the control of this limit for functions that are explicitly marked as
4457 inline (i.e., marked with the inline keyword or defined within the class
4458 definition in c++). @var{n} is the size of functions that can be inlined in
4459 number of pseudo instructions (not counting parameter handling). The default
4460 value of @var{n} is 600.
4461 Increasing this value can result in more inlined code at
4462 the cost of compilation time and memory consumption. Decreasing usually makes
4463 the compilation faster and less code will be inlined (which presumably
4464 means slower programs). This option is particularly useful for programs that
4465 use inlining heavily such as those based on recursive templates with C++.
4467 Inlining is actually controlled by a number of parameters, which may be
4468 specified individually by using @option{--param @var{name}=@var{value}}.
4469 The @option{-finline-limit=@var{n}} option sets some of these parameters
4473 @item max-inline-insns-single
4474 is set to @var{n}/2.
4475 @item max-inline-insns-auto
4476 is set to @var{n}/2.
4477 @item min-inline-insns
4478 is set to 130 or @var{n}/4, whichever is smaller.
4479 @item max-inline-insns-rtl
4483 See below for a documentation of the individual
4484 parameters controlling inlining.
4486 @emph{Note:} pseudo instruction represents, in this particular context, an
4487 abstract measurement of function's size. In no way does it represent a count
4488 of assembly instructions and as such its exact meaning might change from one
4489 release to an another.
4491 @item -fkeep-inline-functions
4492 @opindex fkeep-inline-functions
4493 In C, emit @code{static} functions that are declared @code{inline}
4494 into the object file, even if the function has been inlined into all
4495 of its callers. This switch does not affect functions using the
4496 @code{extern inline} extension in GNU C@. In C++, emit any and all
4497 inline functions into the object file.
4499 @item -fkeep-static-consts
4500 @opindex fkeep-static-consts
4501 Emit variables declared @code{static const} when optimization isn't turned
4502 on, even if the variables aren't referenced.
4504 GCC enables this option by default. If you want to force the compiler to
4505 check if the variable was referenced, regardless of whether or not
4506 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4508 @item -fmerge-constants
4509 Attempt to merge identical constants (string constants and floating point
4510 constants) across compilation units.
4512 This option is the default for optimized compilation if the assembler and
4513 linker support it. Use @option{-fno-merge-constants} to inhibit this
4516 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4518 @item -fmerge-all-constants
4519 Attempt to merge identical constants and identical variables.
4521 This option implies @option{-fmerge-constants}. In addition to
4522 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4523 arrays or initialized constant variables with integral or floating point
4524 types. Languages like C or C++ require each non-automatic variable to
4525 have distinct location, so using this option will result in non-conforming
4528 @item -fmodulo-sched
4529 @opindex fmodulo-sched
4530 Perform swing modulo scheduling immediately before the first scheduling
4531 pass. This pass looks at innermost loops and reorders their
4532 instructions by overlapping different iterations.
4534 @item -fno-branch-count-reg
4535 @opindex fno-branch-count-reg
4536 Do not use ``decrement and branch'' instructions on a count register,
4537 but instead generate a sequence of instructions that decrement a
4538 register, compare it against zero, then branch based upon the result.
4539 This option is only meaningful on architectures that support such
4540 instructions, which include x86, PowerPC, IA-64 and S/390.
4542 The default is @option{-fbranch-count-reg}, enabled when
4543 @option{-fstrength-reduce} is enabled.
4545 @item -fno-function-cse
4546 @opindex fno-function-cse
4547 Do not put function addresses in registers; make each instruction that
4548 calls a constant function contain the function's address explicitly.
4550 This option results in less efficient code, but some strange hacks
4551 that alter the assembler output may be confused by the optimizations
4552 performed when this option is not used.
4554 The default is @option{-ffunction-cse}
4556 @item -fno-zero-initialized-in-bss
4557 @opindex fno-zero-initialized-in-bss
4558 If the target supports a BSS section, GCC by default puts variables that
4559 are initialized to zero into BSS@. This can save space in the resulting
4562 This option turns off this behavior because some programs explicitly
4563 rely on variables going to the data section. E.g., so that the
4564 resulting executable can find the beginning of that section and/or make
4565 assumptions based on that.
4567 The default is @option{-fzero-initialized-in-bss}.
4569 @item -fbounds-check
4570 @opindex fbounds-check
4571 For front-ends that support it, generate additional code to check that
4572 indices used to access arrays are within the declared range. This is
4573 currently only supported by the Java and Fortran front-ends, where
4574 this option defaults to true and false respectively.
4576 @item -fmudflap -fmudflapth -fmudflapir
4580 @cindex bounds checking
4582 For front-ends that support it (C and C++), instrument all risky
4583 pointer/array dereferencing operations, some standard library
4584 string/heap functions, and some other associated constructs with
4585 range/validity tests. Modules so instrumented should be immune to
4586 buffer overflows, invalid heap use, and some other classes of C/C++
4587 programming errors. The instrumentation relies on a separate runtime
4588 library (@file{libmudflap}), which will be linked into a program if
4589 @option{-fmudflap} is given at link time. Run-time behavior of the
4590 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4591 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4594 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4595 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4596 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4597 instrumentation should ignore pointer reads. This produces less
4598 instrumentation (and therefore faster execution) and still provides
4599 some protection against outright memory corrupting writes, but allows
4600 erroneously read data to propagate within a program.
4602 @item -fstrength-reduce
4603 @opindex fstrength-reduce
4604 Perform the optimizations of loop strength reduction and
4605 elimination of iteration variables.
4607 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4609 @item -fthread-jumps
4610 @opindex fthread-jumps
4611 Perform optimizations where we check to see if a jump branches to a
4612 location where another comparison subsumed by the first is found. If
4613 so, the first branch is redirected to either the destination of the
4614 second branch or a point immediately following it, depending on whether
4615 the condition is known to be true or false.
4617 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4619 @item -fcse-follow-jumps
4620 @opindex fcse-follow-jumps
4621 In common subexpression elimination, scan through jump instructions
4622 when the target of the jump is not reached by any other path. For
4623 example, when CSE encounters an @code{if} statement with an
4624 @code{else} clause, CSE will follow the jump when the condition
4627 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4629 @item -fcse-skip-blocks
4630 @opindex fcse-skip-blocks
4631 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4632 follow jumps which conditionally skip over blocks. When CSE
4633 encounters a simple @code{if} statement with no else clause,
4634 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4635 body of the @code{if}.
4637 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4639 @item -frerun-cse-after-loop
4640 @opindex frerun-cse-after-loop
4641 Re-run common subexpression elimination after loop optimizations has been
4644 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4646 @item -frerun-loop-opt
4647 @opindex frerun-loop-opt
4648 Run the loop optimizer twice.
4650 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4654 Perform a global common subexpression elimination pass.
4655 This pass also performs global constant and copy propagation.
4657 @emph{Note:} When compiling a program using computed gotos, a GCC
4658 extension, you may get better runtime performance if you disable
4659 the global common subexpression elimination pass by adding
4660 @option{-fno-gcse} to the command line.
4662 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4666 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4667 attempt to move loads which are only killed by stores into themselves. This
4668 allows a loop containing a load/store sequence to be changed to a load outside
4669 the loop, and a copy/store within the loop.
4671 Enabled by default when gcse is enabled.
4675 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4676 global common subexpression elimination. This pass will attempt to move
4677 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4678 loops containing a load/store sequence can be changed to a load before
4679 the loop and a store after the loop.
4681 Not enabled at any optimization level.
4685 When @option{-fgcse-las} is enabled, the global common subexpression
4686 elimination pass eliminates redundant loads that come after stores to the
4687 same memory location (both partial and full redundancies).
4689 Not enabled at any optimization level.
4691 @item -fgcse-after-reload
4692 @opindex fgcse-after-reload
4693 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4694 pass is performed after reload. The purpose of this pass is to cleanup
4697 @item -floop-optimize
4698 @opindex floop-optimize
4699 Perform loop optimizations: move constant expressions out of loops, simplify
4700 exit test conditions and optionally do strength-reduction as well.
4702 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4704 @item -floop-optimize2
4705 @opindex floop-optimize2
4706 Perform loop optimizations using the new loop optimizer. The optimizations
4707 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4710 @item -fcrossjumping
4711 @opindex crossjumping
4712 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4713 resulting code may or may not perform better than without cross-jumping.
4715 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4717 @item -fif-conversion
4718 @opindex if-conversion
4719 Attempt to transform conditional jumps into branch-less equivalents. This
4720 include use of conditional moves, min, max, set flags and abs instructions, and
4721 some tricks doable by standard arithmetics. The use of conditional execution
4722 on chips where it is available is controlled by @code{if-conversion2}.
4724 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4726 @item -fif-conversion2
4727 @opindex if-conversion2
4728 Use conditional execution (where available) to transform conditional jumps into
4729 branch-less equivalents.
4731 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4733 @item -fdelete-null-pointer-checks
4734 @opindex fdelete-null-pointer-checks
4735 Use global dataflow analysis to identify and eliminate useless checks
4736 for null pointers. The compiler assumes that dereferencing a null
4737 pointer would have halted the program. If a pointer is checked after
4738 it has already been dereferenced, it cannot be null.
4740 In some environments, this assumption is not true, and programs can
4741 safely dereference null pointers. Use
4742 @option{-fno-delete-null-pointer-checks} to disable this optimization
4743 for programs which depend on that behavior.
4745 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4747 @item -fexpensive-optimizations
4748 @opindex fexpensive-optimizations
4749 Perform a number of minor optimizations that are relatively expensive.
4751 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4753 @item -foptimize-register-move
4755 @opindex foptimize-register-move
4757 Attempt to reassign register numbers in move instructions and as
4758 operands of other simple instructions in order to maximize the amount of
4759 register tying. This is especially helpful on machines with two-operand
4762 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4765 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4767 @item -fdelayed-branch
4768 @opindex fdelayed-branch
4769 If supported for the target machine, attempt to reorder instructions
4770 to exploit instruction slots available after delayed branch
4773 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4775 @item -fschedule-insns
4776 @opindex fschedule-insns
4777 If supported for the target machine, attempt to reorder instructions to
4778 eliminate execution stalls due to required data being unavailable. This
4779 helps machines that have slow floating point or memory load instructions
4780 by allowing other instructions to be issued until the result of the load
4781 or floating point instruction is required.
4783 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4785 @item -fschedule-insns2
4786 @opindex fschedule-insns2
4787 Similar to @option{-fschedule-insns}, but requests an additional pass of
4788 instruction scheduling after register allocation has been done. This is
4789 especially useful on machines with a relatively small number of
4790 registers and where memory load instructions take more than one cycle.
4792 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4794 @item -fno-sched-interblock
4795 @opindex fno-sched-interblock
4796 Don't schedule instructions across basic blocks. This is normally
4797 enabled by default when scheduling before register allocation, i.e.@:
4798 with @option{-fschedule-insns} or at @option{-O2} or higher.
4800 @item -fno-sched-spec
4801 @opindex fno-sched-spec
4802 Don't allow speculative motion of non-load instructions. This is normally
4803 enabled by default when scheduling before register allocation, i.e.@:
4804 with @option{-fschedule-insns} or at @option{-O2} or higher.
4806 @item -fsched-spec-load
4807 @opindex fsched-spec-load
4808 Allow speculative motion of some load instructions. This only makes
4809 sense when scheduling before register allocation, i.e.@: with
4810 @option{-fschedule-insns} or at @option{-O2} or higher.
4812 @item -fsched-spec-load-dangerous
4813 @opindex fsched-spec-load-dangerous
4814 Allow speculative motion of more load instructions. This only makes
4815 sense when scheduling before register allocation, i.e.@: with
4816 @option{-fschedule-insns} or at @option{-O2} or higher.
4818 @item -fsched-stalled-insns=@var{n}
4819 @opindex fsched-stalled-insns
4820 Define how many insns (if any) can be moved prematurely from the queue
4821 of stalled insns into the ready list, during the second scheduling pass.
4823 @item -fsched-stalled-insns-dep=@var{n}
4824 @opindex fsched-stalled-insns-dep
4825 Define how many insn groups (cycles) will be examined for a dependency
4826 on a stalled insn that is candidate for premature removal from the queue
4827 of stalled insns. Has an effect only during the second scheduling pass,
4828 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4830 @item -fsched2-use-superblocks
4831 @opindex fsched2-use-superblocks
4832 When scheduling after register allocation, do use superblock scheduling
4833 algorithm. Superblock scheduling allows motion across basic block boundaries
4834 resulting on faster schedules. This option is experimental, as not all machine
4835 descriptions used by GCC model the CPU closely enough to avoid unreliable
4836 results from the algorithm.
4838 This only makes sense when scheduling after register allocation, i.e.@: with
4839 @option{-fschedule-insns2} or at @option{-O2} or higher.
4841 @item -fsched2-use-traces
4842 @opindex fsched2-use-traces
4843 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4844 allocation and additionally perform code duplication in order to increase the
4845 size of superblocks using tracer pass. See @option{-ftracer} for details on
4848 This mode should produce faster but significantly longer programs. Also
4849 without @option{-fbranch-probabilities} the traces constructed may not
4850 match the reality and hurt the performance. This only makes
4851 sense when scheduling after register allocation, i.e.@: with
4852 @option{-fschedule-insns2} or at @option{-O2} or higher.
4854 @item -freschedule-modulo-scheduled-loops
4855 @opindex fscheduling-in-modulo-scheduled-loops
4856 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4857 we may want to prevent the later scheduling passes from changing its schedule, we use this
4858 option to control that.
4860 @item -fcaller-saves
4861 @opindex fcaller-saves
4862 Enable values to be allocated in registers that will be clobbered by
4863 function calls, by emitting extra instructions to save and restore the
4864 registers around such calls. Such allocation is done only when it
4865 seems to result in better code than would otherwise be produced.
4867 This option is always enabled by default on certain machines, usually
4868 those which have no call-preserved registers to use instead.
4870 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4873 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4874 enabled by default at @option{-O2} and @option{-O3}.
4877 Perform Full Redundancy Elimination (FRE) on trees. The difference
4878 between FRE and PRE is that FRE only considers expressions
4879 that are computed on all paths leading to the redundant computation.
4880 This analysis faster than PRE, though it exposes fewer redundancies.
4881 This flag is enabled by default at @option{-O} and higher.
4883 @item -ftree-copy-prop
4884 Perform copy propagation on trees. This pass eliminates unnecessary
4885 copy operations. This flag is enabled by default at @option{-O} and
4888 @item -ftree-store-copy-prop
4889 Perform copy propagation of memory loads and stores. This pass
4890 eliminates unnecessary copy operations in memory references
4891 (structures, global variables, arrays, etc). This flag is enabled by
4892 default at @option{-O2} and higher.
4895 Perform structural alias analysis on trees. This flag
4896 is enabled by default at @option{-O} and higher.
4899 Perform forward store motion on trees. This flag is
4900 enabled by default at @option{-O} and higher.
4903 Perform sparse conditional constant propagation (CCP) on trees. This
4904 pass only operates on local scalar variables and is enabled by default
4905 at @option{-O} and higher.
4907 @item -ftree-store-ccp
4908 Perform sparse conditional constant propagation (CCP) on trees. This
4909 pass operates on both local scalar variables and memory stores and
4910 loads (global variables, structures, arrays, etc). This flag is
4911 enabled by default at @option{-O2} and higher.
4914 Perform dead code elimination (DCE) on trees. This flag is enabled by
4915 default at @option{-O} and higher.
4917 @item -ftree-dominator-opts
4918 Perform dead code elimination (DCE) on trees. This flag is enabled by
4919 default at @option{-O} and higher.
4922 Perform loop header copying on trees. This is beneficial since it increases
4923 effectiveness of code motion optimizations. It also saves one jump. This flag
4924 is enabled by default at @option{-O} and higher. It is not enabled
4925 for @option{-Os}, since it usually increases code size.
4927 @item -ftree-loop-optimize
4928 Perform loop optimizations on trees. This flag is enabled by default
4929 at @option{-O} and higher.
4931 @item -ftree-loop-linear
4932 Perform linear loop transformations on tree. This flag can improve cache
4933 performance and allow further loop optimizations to take place.
4935 @item -ftree-loop-im
4936 Perform loop invariant motion on trees. This pass moves only invariants that
4937 would be hard to handle at RTL level (function calls, operations that expand to
4938 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
4939 operands of conditions that are invariant out of the loop, so that we can use
4940 just trivial invariantness analysis in loop unswitching. The pass also includes
4943 @item -ftree-loop-ivcanon
4944 Create a canonical counter for number of iterations in the loop for that
4945 determining number of iterations requires complicated analysis. Later
4946 optimizations then may determine the number easily. Useful especially
4947 in connection with unrolling.
4950 Perform induction variable optimizations (strength reduction, induction
4951 variable merging and induction variable elimination) on trees.
4954 Perform scalar replacement of aggregates. This pass replaces structure
4955 references with scalars to prevent committing structures to memory too
4956 early. This flag is enabled by default at @option{-O} and higher.
4958 @item -ftree-copyrename
4959 Perform copy renaming on trees. This pass attempts to rename compiler
4960 temporaries to other variables at copy locations, usually resulting in
4961 variable names which more closely resemble the original variables. This flag
4962 is enabled by default at @option{-O} and higher.
4965 Perform temporary expression replacement during the SSA->normal phase. Single
4966 use/single def temporaries are replaced at their use location with their
4967 defining expression. This results in non-GIMPLE code, but gives the expanders
4968 much more complex trees to work on resulting in better RTL generation. This is
4969 enabled by default at @option{-O} and higher.
4972 Perform live range splitting during the SSA->normal phase. Distinct live
4973 ranges of a variable are split into unique variables, allowing for better
4974 optimization later. This is enabled by default at @option{-O} and higher.
4976 @item -ftree-vectorize
4977 Perform loop vectorization on trees.
4980 Perform Value Range Propagation on trees. This is similar to the
4981 constant propagation pass, but instead of values, ranges of values are
4982 propagated. This allows the optimizers to remove unnecessary range
4983 checks like array bound checks and null pointer checks. This is
4984 enabled by default at @option{-O2} and higher. Null pointer check
4985 elimination is only done if @option{-fdelete-null-pointer-checks} is
4990 Perform tail duplication to enlarge superblock size. This transformation
4991 simplifies the control flow of the function allowing other optimizations to do
4994 @item -funroll-loops
4995 @opindex funroll-loops
4996 Unroll loops whose number of iterations can be determined at compile
4997 time or upon entry to the loop. @option{-funroll-loops} implies both
4998 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4999 option makes code larger, and may or may not make it run faster.
5001 @item -funroll-all-loops
5002 @opindex funroll-all-loops
5003 Unroll all loops, even if their number of iterations is uncertain when
5004 the loop is entered. This usually makes programs run more slowly.
5005 @option{-funroll-all-loops} implies the same options as
5006 @option{-funroll-loops},
5008 @item -fsplit-ivs-in-unroller
5009 @opindex -fsplit-ivs-in-unroller
5010 Enables expressing of values of induction variables in later iterations
5011 of the unrolled loop using the value in the first iteration. This breaks
5012 long dependency chains, thus improving efficiency of the scheduling passes
5013 (for best results, @option{-fweb} should be used as well).
5015 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5016 same effect. However in cases the loop body is more complicated than
5017 a single basic block, this is not reliable. It also does not work at all
5018 on some of the architectures due to restrictions in the CSE pass.
5020 This optimization is enabled by default.
5022 @item -fvariable-expansion-in-unroller
5023 @opindex -fvariable-expansion-in-unroller
5024 With this option, the compiler will create multiple copies of some
5025 local variables when unrolling a loop which can result in superior code.
5027 @item -fprefetch-loop-arrays
5028 @opindex fprefetch-loop-arrays
5029 If supported by the target machine, generate instructions to prefetch
5030 memory to improve the performance of loops that access large arrays.
5032 These options may generate better or worse code; results are highly
5033 dependent on the structure of loops within the source code.
5036 @itemx -fno-peephole2
5037 @opindex fno-peephole
5038 @opindex fno-peephole2
5039 Disable any machine-specific peephole optimizations. The difference
5040 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5041 are implemented in the compiler; some targets use one, some use the
5042 other, a few use both.
5044 @option{-fpeephole} is enabled by default.
5045 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5047 @item -fno-guess-branch-probability
5048 @opindex fno-guess-branch-probability
5049 Do not guess branch probabilities using heuristics.
5051 GCC will use heuristics to guess branch probabilities if they are
5052 not provided by profiling feedback (@option{-fprofile-arcs}). These
5053 heuristics are based on the control flow graph. If some branch probabilities
5054 are specified by @samp{__builtin_expect}, then the heuristics will be
5055 used to guess branch probabilities for the rest of the control flow graph,
5056 taking the @samp{__builtin_expect} info into account. The interactions
5057 between the heuristics and @samp{__builtin_expect} can be complex, and in
5058 some cases, it may be useful to disable the heuristics so that the effects
5059 of @samp{__builtin_expect} are easier to understand.
5061 The default is @option{-fguess-branch-probability} at levels
5062 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5064 @item -freorder-blocks
5065 @opindex freorder-blocks
5066 Reorder basic blocks in the compiled function in order to reduce number of
5067 taken branches and improve code locality.
5069 Enabled at levels @option{-O2}, @option{-O3}.
5071 @item -freorder-blocks-and-partition
5072 @opindex freorder-blocks-and-partition
5073 In addition to reordering basic blocks in the compiled function, in order
5074 to reduce number of taken branches, partitions hot and cold basic blocks
5075 into separate sections of the assembly and .o files, to improve
5076 paging and cache locality performance.
5078 This optimization is automatically turned off in the presence of
5079 exception handling, for linkonce sections, for functions with a user-defined
5080 section attribute and on any architecture that does not support named
5083 @item -freorder-functions
5084 @opindex freorder-functions
5085 Reorder functions in the object file in order to
5086 improve code locality. This is implemented by using special
5087 subsections @code{.text.hot} for most frequently executed functions and
5088 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5089 the linker so object file format must support named sections and linker must
5090 place them in a reasonable way.
5092 Also profile feedback must be available in to make this option effective. See
5093 @option{-fprofile-arcs} for details.
5095 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5097 @item -fstrict-aliasing
5098 @opindex fstrict-aliasing
5099 Allows the compiler to assume the strictest aliasing rules applicable to
5100 the language being compiled. For C (and C++), this activates
5101 optimizations based on the type of expressions. In particular, an
5102 object of one type is assumed never to reside at the same address as an
5103 object of a different type, unless the types are almost the same. For
5104 example, an @code{unsigned int} can alias an @code{int}, but not a
5105 @code{void*} or a @code{double}. A character type may alias any other
5108 Pay special attention to code like this:
5121 The practice of reading from a different union member than the one most
5122 recently written to (called ``type-punning'') is common. Even with
5123 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5124 is accessed through the union type. So, the code above will work as
5125 expected. However, this code might not:
5136 Every language that wishes to perform language-specific alias analysis
5137 should define a function that computes, given an @code{tree}
5138 node, an alias set for the node. Nodes in different alias sets are not
5139 allowed to alias. For an example, see the C front-end function
5140 @code{c_get_alias_set}.
5142 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5144 @item -falign-functions
5145 @itemx -falign-functions=@var{n}
5146 @opindex falign-functions
5147 Align the start of functions to the next power-of-two greater than
5148 @var{n}, skipping up to @var{n} bytes. For instance,
5149 @option{-falign-functions=32} aligns functions to the next 32-byte
5150 boundary, but @option{-falign-functions=24} would align to the next
5151 32-byte boundary only if this can be done by skipping 23 bytes or less.
5153 @option{-fno-align-functions} and @option{-falign-functions=1} are
5154 equivalent and mean that functions will not be aligned.
5156 Some assemblers only support this flag when @var{n} is a power of two;
5157 in that case, it is rounded up.
5159 If @var{n} is not specified or is zero, use a machine-dependent default.
5161 Enabled at levels @option{-O2}, @option{-O3}.
5163 @item -falign-labels
5164 @itemx -falign-labels=@var{n}
5165 @opindex falign-labels
5166 Align all branch targets to a power-of-two boundary, skipping up to
5167 @var{n} bytes like @option{-falign-functions}. This option can easily
5168 make code slower, because it must insert dummy operations for when the
5169 branch target is reached in the usual flow of the code.
5171 @option{-fno-align-labels} and @option{-falign-labels=1} are
5172 equivalent and mean that labels will not be aligned.
5174 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5175 are greater than this value, then their values are used instead.
5177 If @var{n} is not specified or is zero, use a machine-dependent default
5178 which is very likely to be @samp{1}, meaning no alignment.
5180 Enabled at levels @option{-O2}, @option{-O3}.
5183 @itemx -falign-loops=@var{n}
5184 @opindex falign-loops
5185 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5186 like @option{-falign-functions}. The hope is that the loop will be
5187 executed many times, which will make up for any execution of the dummy
5190 @option{-fno-align-loops} and @option{-falign-loops=1} are
5191 equivalent and mean that loops will not be aligned.
5193 If @var{n} is not specified or is zero, use a machine-dependent default.
5195 Enabled at levels @option{-O2}, @option{-O3}.
5198 @itemx -falign-jumps=@var{n}
5199 @opindex falign-jumps
5200 Align branch targets to a power-of-two boundary, for branch targets
5201 where the targets can only be reached by jumping, skipping up to @var{n}
5202 bytes like @option{-falign-functions}. In this case, no dummy operations
5205 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5206 equivalent and mean that loops will not be aligned.
5208 If @var{n} is not specified or is zero, use a machine-dependent default.
5210 Enabled at levels @option{-O2}, @option{-O3}.
5212 @item -funit-at-a-time
5213 @opindex funit-at-a-time
5214 Parse the whole compilation unit before starting to produce code.
5215 This allows some extra optimizations to take place but consumes
5216 more memory (in general). There are some compatibility issues
5217 with @emph{unit-at-at-time} mode:
5220 enabling @emph{unit-at-a-time} mode may change the order
5221 in which functions, variables, and top-level @code{asm} statements
5222 are emitted, and will likely break code relying on some particular
5223 ordering. The majority of such top-level @code{asm} statements,
5224 though, can be replaced by @code{section} attributes.
5227 @emph{unit-at-a-time} mode removes unreferenced static variables
5228 and functions are removed. This may result in undefined references
5229 when an @code{asm} statement refers directly to variables or functions
5230 that are otherwise unused. In that case either the variable/function
5231 shall be listed as an operand of the @code{asm} statement operand or,
5232 in the case of top-level @code{asm} statements the attribute @code{used}
5233 shall be used on the declaration.
5236 Static functions now can use non-standard passing conventions that
5237 may break @code{asm} statements calling functions directly. Again,
5238 attribute @code{used} will prevent this behavior.
5241 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5242 but this scheme may not be supported by future releases of GCC@.
5244 Enabled at levels @option{-O2}, @option{-O3}.
5248 Constructs webs as commonly used for register allocation purposes and assign
5249 each web individual pseudo register. This allows the register allocation pass
5250 to operate on pseudos directly, but also strengthens several other optimization
5251 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5252 however, make debugging impossible, since variables will no longer stay in a
5255 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os},
5256 on targets where the default format for debugging information supports
5259 @item -fwhole-program
5260 @opindex fwhole-program
5261 Assume that the current compilation unit represents whole program being
5262 compiled. All public functions and variables with the exception of @code{main}
5263 and those marged by attribute @code{externally_visible} become static functions
5264 and in a affect gets more aggresively optimized by interprocedural optimizers.
5265 While this option is equivalent to proper use of @code{static} keyword for
5266 programs consitsting of single file, in combination with option
5267 @option{--combine} this flag can be used to compile most of smaller scale C
5268 programs since the functions and variables become local for the whole combined
5269 compilation unit, not for the single source file itself.
5272 @item -fno-cprop-registers
5273 @opindex fno-cprop-registers
5274 After register allocation and post-register allocation instruction splitting,
5275 we perform a copy-propagation pass to try to reduce scheduling dependencies
5276 and occasionally eliminate the copy.
5278 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5280 @item -fprofile-generate
5281 @opindex fprofile-generate
5283 Enable options usually used for instrumenting application to produce
5284 profile useful for later recompilation with profile feedback based
5285 optimization. You must use @option{-fprofile-generate} both when
5286 compiling and when linking your program.
5288 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5291 @opindex fprofile-use
5292 Enable profile feedback directed optimizations, and optimizations
5293 generally profitable only with profile feedback available.
5295 The following options are enabled: @code{-fbranch-probabilities},
5296 @code{-fvpt}, @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}.
5300 The following options control compiler behavior regarding floating
5301 point arithmetic. These options trade off between speed and
5302 correctness. All must be specifically enabled.
5306 @opindex ffloat-store
5307 Do not store floating point variables in registers, and inhibit other
5308 options that might change whether a floating point value is taken from a
5311 @cindex floating point precision
5312 This option prevents undesirable excess precision on machines such as
5313 the 68000 where the floating registers (of the 68881) keep more
5314 precision than a @code{double} is supposed to have. Similarly for the
5315 x86 architecture. For most programs, the excess precision does only
5316 good, but a few programs rely on the precise definition of IEEE floating
5317 point. Use @option{-ffloat-store} for such programs, after modifying
5318 them to store all pertinent intermediate computations into variables.
5322 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5323 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5324 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5325 and @option{fcx-limited-range}.
5327 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5329 This option should never be turned on by any @option{-O} option since
5330 it can result in incorrect output for programs which depend on
5331 an exact implementation of IEEE or ISO rules/specifications for
5334 @item -fno-math-errno
5335 @opindex fno-math-errno
5336 Do not set ERRNO after calling math functions that are executed
5337 with a single instruction, e.g., sqrt. A program that relies on
5338 IEEE exceptions for math error handling may want to use this flag
5339 for speed while maintaining IEEE arithmetic compatibility.
5341 This option should never be turned on by any @option{-O} option since
5342 it can result in incorrect output for programs which depend on
5343 an exact implementation of IEEE or ISO rules/specifications for
5346 The default is @option{-fmath-errno}.
5348 On Darwin systems, the math library never sets @code{errno}. There is therefore
5349 no reason for the compiler to consider the possibility that it might,
5350 and @option{-fno-math-errno} is the default.
5352 @item -funsafe-math-optimizations
5353 @opindex funsafe-math-optimizations
5354 Allow optimizations for floating-point arithmetic that (a) assume
5355 that arguments and results are valid and (b) may violate IEEE or
5356 ANSI standards. When used at link-time, it may include libraries
5357 or startup files that change the default FPU control word or other
5358 similar optimizations.
5360 This option should never be turned on by any @option{-O} option since
5361 it can result in incorrect output for programs which depend on
5362 an exact implementation of IEEE or ISO rules/specifications for
5365 The default is @option{-fno-unsafe-math-optimizations}.
5367 @item -ffinite-math-only
5368 @opindex ffinite-math-only
5369 Allow optimizations for floating-point arithmetic that assume
5370 that arguments and results are not NaNs or +-Infs.
5372 This option should never be turned on by any @option{-O} option since
5373 it can result in incorrect output for programs which depend on
5374 an exact implementation of IEEE or ISO rules/specifications.
5376 The default is @option{-fno-finite-math-only}.
5378 @item -fno-trapping-math
5379 @opindex fno-trapping-math
5380 Compile code assuming that floating-point operations cannot generate
5381 user-visible traps. These traps include division by zero, overflow,
5382 underflow, inexact result and invalid operation. This option implies
5383 @option{-fno-signaling-nans}. Setting this option may allow faster
5384 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5386 This option should never be turned on by any @option{-O} option since
5387 it can result in incorrect output for programs which depend on
5388 an exact implementation of IEEE or ISO rules/specifications for
5391 The default is @option{-ftrapping-math}.
5393 @item -frounding-math
5394 @opindex frounding-math
5395 Disable transformations and optimizations that assume default floating
5396 point rounding behavior. This is round-to-zero for all floating point
5397 to integer conversions, and round-to-nearest for all other arithmetic
5398 truncations. This option should be specified for programs that change
5399 the FP rounding mode dynamically, or that may be executed with a
5400 non-default rounding mode. This option disables constant folding of
5401 floating point expressions at compile-time (which may be affected by
5402 rounding mode) and arithmetic transformations that are unsafe in the
5403 presence of sign-dependent rounding modes.
5405 The default is @option{-fno-rounding-math}.
5407 This option is experimental and does not currently guarantee to
5408 disable all GCC optimizations that are affected by rounding mode.
5409 Future versions of GCC may provide finer control of this setting
5410 using C99's @code{FENV_ACCESS} pragma. This command line option
5411 will be used to specify the default state for @code{FENV_ACCESS}.
5413 @item -fsignaling-nans
5414 @opindex fsignaling-nans
5415 Compile code assuming that IEEE signaling NaNs may generate user-visible
5416 traps during floating-point operations. Setting this option disables
5417 optimizations that may change the number of exceptions visible with
5418 signaling NaNs. This option implies @option{-ftrapping-math}.
5420 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5423 The default is @option{-fno-signaling-nans}.
5425 This option is experimental and does not currently guarantee to
5426 disable all GCC optimizations that affect signaling NaN behavior.
5428 @item -fsingle-precision-constant
5429 @opindex fsingle-precision-constant
5430 Treat floating point constant as single precision constant instead of
5431 implicitly converting it to double precision constant.
5433 @item -fcx-limited-range
5434 @itemx -fno-cx-limited-range
5435 @opindex fcx-limited-range
5436 @opindex fno-cx-limited-range
5437 When enabled, this option states that a range reduction step is not
5438 needed when performing complex division. The default is
5439 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5441 This option controls the default setting of the ISO C99
5442 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5447 The following options control optimizations that may improve
5448 performance, but are not enabled by any @option{-O} options. This
5449 section includes experimental options that may produce broken code.
5452 @item -fbranch-probabilities
5453 @opindex fbranch-probabilities
5454 After running a program compiled with @option{-fprofile-arcs}
5455 (@pxref{Debugging Options,, Options for Debugging Your Program or
5456 @command{gcc}}), you can compile it a second time using
5457 @option{-fbranch-probabilities}, to improve optimizations based on
5458 the number of times each branch was taken. When the program
5459 compiled with @option{-fprofile-arcs} exits it saves arc execution
5460 counts to a file called @file{@var{sourcename}.gcda} for each source
5461 file The information in this data file is very dependent on the
5462 structure of the generated code, so you must use the same source code
5463 and the same optimization options for both compilations.
5465 With @option{-fbranch-probabilities}, GCC puts a
5466 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5467 These can be used to improve optimization. Currently, they are only
5468 used in one place: in @file{reorg.c}, instead of guessing which path a
5469 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5470 exactly determine which path is taken more often.
5472 @item -fprofile-values
5473 @opindex fprofile-values
5474 If combined with @option{-fprofile-arcs}, it adds code so that some
5475 data about values of expressions in the program is gathered.
5477 With @option{-fbranch-probabilities}, it reads back the data gathered
5478 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5479 notes to instructions for their later usage in optimizations.
5481 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5485 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5486 a code to gather information about values of expressions.
5488 With @option{-fbranch-probabilities}, it reads back the data gathered
5489 and actually performs the optimizations based on them.
5490 Currently the optimizations include specialization of division operation
5491 using the knowledge about the value of the denominator.
5493 @item -fspeculative-prefetching
5494 @opindex fspeculative-prefetching
5495 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5496 a code to gather information about addresses of memory references in the
5499 With @option{-fbranch-probabilities}, it reads back the data gathered
5500 and issues prefetch instructions according to them. In addition to the opportunities
5501 noticed by @option{-fprefetch-loop-arrays}, it also notices more complicated
5502 memory access patterns---for example accesses to the data stored in linked
5503 list whose elements are usually allocated sequentially.
5505 In order to prevent issuing double prefetches, usage of
5506 @option{-fspeculative-prefetching} implies @option{-fno-prefetch-loop-arrays}.
5508 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5510 @item -frename-registers
5511 @opindex frename-registers
5512 Attempt to avoid false dependencies in scheduled code by making use
5513 of registers left over after register allocation. This optimization
5514 will most benefit processors with lots of registers. Depending on the
5515 debug information format adopted by the target, however, it can
5516 make debugging impossible, since variables will no longer stay in
5517 a ``home register''.
5519 Not enabled by default at any level because it has known bugs.
5523 Perform tail duplication to enlarge superblock size. This transformation
5524 simplifies the control flow of the function allowing other optimizations to do
5527 Enabled with @option{-fprofile-use}.
5529 @item -funroll-loops
5530 @opindex funroll-loops
5531 Unroll loops whose number of iterations can be determined at compile time or
5532 upon entry to the loop. @option{-funroll-loops} implies
5533 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
5534 (i.e.@: complete removal of loops with small constant number of iterations).
5535 This option makes code larger, and may or may not make it run faster.
5537 Enabled with @option{-fprofile-use}.
5539 @item -funroll-all-loops
5540 @opindex funroll-all-loops
5541 Unroll all loops, even if their number of iterations is uncertain when
5542 the loop is entered. This usually makes programs run more slowly.
5543 @option{-funroll-all-loops} implies the same options as
5544 @option{-funroll-loops}.
5547 @opindex fpeel-loops
5548 Peels the loops for that there is enough information that they do not
5549 roll much (from profile feedback). It also turns on complete loop peeling
5550 (i.e.@: complete removal of loops with small constant number of iterations).
5552 Enabled with @option{-fprofile-use}.
5554 @item -fmove-loop-invariants
5555 @opindex fmove-loop-invariants
5556 Enables the loop invariant motion pass in the new loop optimizer. Enabled
5557 at level @option{-O1}
5559 @item -funswitch-loops
5560 @opindex funswitch-loops
5561 Move branches with loop invariant conditions out of the loop, with duplicates
5562 of the loop on both branches (modified according to result of the condition).
5564 @item -fprefetch-loop-arrays
5565 @opindex fprefetch-loop-arrays
5566 If supported by the target machine, generate instructions to prefetch
5567 memory to improve the performance of loops that access large arrays.
5569 Disabled at level @option{-Os}.
5571 @item -ffunction-sections
5572 @itemx -fdata-sections
5573 @opindex ffunction-sections
5574 @opindex fdata-sections
5575 Place each function or data item into its own section in the output
5576 file if the target supports arbitrary sections. The name of the
5577 function or the name of the data item determines the section's name
5580 Use these options on systems where the linker can perform optimizations
5581 to improve locality of reference in the instruction space. Most systems
5582 using the ELF object format and SPARC processors running Solaris 2 have
5583 linkers with such optimizations. AIX may have these optimizations in
5586 Only use these options when there are significant benefits from doing
5587 so. When you specify these options, the assembler and linker will
5588 create larger object and executable files and will also be slower.
5589 You will not be able to use @code{gprof} on all systems if you
5590 specify this option and you may have problems with debugging if
5591 you specify both this option and @option{-g}.
5593 @item -fbranch-target-load-optimize
5594 @opindex fbranch-target-load-optimize
5595 Perform branch target register load optimization before prologue / epilogue
5597 The use of target registers can typically be exposed only during reload,
5598 thus hoisting loads out of loops and doing inter-block scheduling needs
5599 a separate optimization pass.
5601 @item -fbranch-target-load-optimize2
5602 @opindex fbranch-target-load-optimize2
5603 Perform branch target register load optimization after prologue / epilogue
5606 @item -fbtr-bb-exclusive
5607 @opindex fbtr-bb-exclusive
5608 When performing branch target register load optimization, don't reuse
5609 branch target registers in within any basic block.
5611 @item --param @var{name}=@var{value}
5613 In some places, GCC uses various constants to control the amount of
5614 optimization that is done. For example, GCC will not inline functions
5615 that contain more that a certain number of instructions. You can
5616 control some of these constants on the command-line using the
5617 @option{--param} option.
5619 The names of specific parameters, and the meaning of the values, are
5620 tied to the internals of the compiler, and are subject to change
5621 without notice in future releases.
5623 In each case, the @var{value} is an integer. The allowable choices for
5624 @var{name} are given in the following table:
5627 @item salias-max-implicit-fields
5628 The maximum number of fields in a variable without direct
5629 structure accesses for which structure aliasing will consider trying
5630 to track each field. The default is 5
5632 @item sra-max-structure-size
5633 The maximum structure size, in bytes, at which the scalar replacement
5634 of aggregates (SRA) optimization will perform block copies. The
5635 default value, 0, implies that GCC will select the most appropriate
5638 @item sra-field-structure-ratio
5639 The threshold ratio (as a percentage) between instantiated fields and
5640 the complete structure size. We say that if the ratio of the number
5641 of bytes in instantiated fields to the number of bytes in the complete
5642 structure exceeds this parameter, then block copies are not used. The
5645 @item max-crossjump-edges
5646 The maximum number of incoming edges to consider for crossjumping.
5647 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5648 the number of edges incoming to each block. Increasing values mean
5649 more aggressive optimization, making the compile time increase with
5650 probably small improvement in executable size.
5652 @item min-crossjump-insns
5653 The minimum number of instructions which must be matched at the end
5654 of two blocks before crossjumping will be performed on them. This
5655 value is ignored in the case where all instructions in the block being
5656 crossjumped from are matched. The default value is 5.
5658 @item max-goto-duplication-insns
5659 The maximum number of instructions to duplicate to a block that jumps
5660 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5661 passes, GCC factors computed gotos early in the compilation process,
5662 and unfactors them as late as possible. Only computed jumps at the
5663 end of a basic blocks with no more than max-goto-duplication-insns are
5664 unfactored. The default value is 8.
5666 @item max-delay-slot-insn-search
5667 The maximum number of instructions to consider when looking for an
5668 instruction to fill a delay slot. If more than this arbitrary number of
5669 instructions is searched, the time savings from filling the delay slot
5670 will be minimal so stop searching. Increasing values mean more
5671 aggressive optimization, making the compile time increase with probably
5672 small improvement in executable run time.
5674 @item max-delay-slot-live-search
5675 When trying to fill delay slots, the maximum number of instructions to
5676 consider when searching for a block with valid live register
5677 information. Increasing this arbitrarily chosen value means more
5678 aggressive optimization, increasing the compile time. This parameter
5679 should be removed when the delay slot code is rewritten to maintain the
5682 @item max-gcse-memory
5683 The approximate maximum amount of memory that will be allocated in
5684 order to perform the global common subexpression elimination
5685 optimization. If more memory than specified is required, the
5686 optimization will not be done.
5688 @item max-gcse-passes
5689 The maximum number of passes of GCSE to run. The default is 1.
5691 @item max-pending-list-length
5692 The maximum number of pending dependencies scheduling will allow
5693 before flushing the current state and starting over. Large functions
5694 with few branches or calls can create excessively large lists which
5695 needlessly consume memory and resources.
5697 @item max-inline-insns-single
5698 Several parameters control the tree inliner used in gcc.
5699 This number sets the maximum number of instructions (counted in GCC's
5700 internal representation) in a single function that the tree inliner
5701 will consider for inlining. This only affects functions declared
5702 inline and methods implemented in a class declaration (C++).
5703 The default value is 450.
5705 @item max-inline-insns-auto
5706 When you use @option{-finline-functions} (included in @option{-O3}),
5707 a lot of functions that would otherwise not be considered for inlining
5708 by the compiler will be investigated. To those functions, a different
5709 (more restrictive) limit compared to functions declared inline can
5711 The default value is 90.
5713 @item large-function-insns
5714 The limit specifying really large functions. For functions larger than this
5715 limit after inlining inlining is constrained by
5716 @option{--param large-function-growth}. This parameter is useful primarily
5717 to avoid extreme compilation time caused by non-linear algorithms used by the
5719 This parameter is ignored when @option{-funit-at-a-time} is not used.
5720 The default value is 2700.
5722 @item large-function-growth
5723 Specifies maximal growth of large function caused by inlining in percents.
5724 This parameter is ignored when @option{-funit-at-a-time} is not used.
5725 The default value is 100 which limits large function growth to 2.0 times
5728 @item inline-unit-growth
5729 Specifies maximal overall growth of the compilation unit caused by inlining.
5730 This parameter is ignored when @option{-funit-at-a-time} is not used.
5731 The default value is 50 which limits unit growth to 1.5 times the original
5734 @item max-inline-insns-recursive
5735 @itemx max-inline-insns-recursive-auto
5736 Specifies maximum number of instructions out-of-line copy of self recursive inline
5737 function can grow into by performing recursive inlining.
5739 For functions declared inline @option{--param max-inline-insns-recursive} is
5740 taken into acount. For function not declared inline, recursive inlining
5741 happens only when @option{-finline-functions} (included in @option{-O3}) is
5742 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5743 default value is 450.
5745 @item max-inline-recursive-depth
5746 @itemx max-inline-recursive-depth-auto
5747 Specifies maximum recursion depth used by the recursive inlining.
5749 For functions declared inline @option{--param max-inline-recursive-depth} is
5750 taken into acount. For function not declared inline, recursive inlining
5751 happens only when @option{-finline-functions} (included in @option{-O3}) is
5752 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5753 default value is 450.
5755 @item inline-call-cost
5756 Specify cost of call instruction relative to simple arithmetics operations
5757 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
5758 functions and at the same time increases size of leaf function that is believed to
5759 reduce function size by being inlined. In effect it increases amount of
5760 inlining for code having large abstraction penalty (many functions that just
5761 pass the arguments to other functions) and decrease inlining for code with low
5762 abstraction penalty. The default value is 16.
5764 @item max-unrolled-insns
5765 The maximum number of instructions that a loop should have if that loop
5766 is unrolled, and if the loop is unrolled, it determines how many times
5767 the loop code is unrolled.
5769 @item max-average-unrolled-insns
5770 The maximum number of instructions biased by probabilities of their execution
5771 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5772 it determines how many times the loop code is unrolled.
5774 @item max-unroll-times
5775 The maximum number of unrollings of a single loop.
5777 @item max-peeled-insns
5778 The maximum number of instructions that a loop should have if that loop
5779 is peeled, and if the loop is peeled, it determines how many times
5780 the loop code is peeled.
5782 @item max-peel-times
5783 The maximum number of peelings of a single loop.
5785 @item max-completely-peeled-insns
5786 The maximum number of insns of a completely peeled loop.
5788 @item max-completely-peel-times
5789 The maximum number of iterations of a loop to be suitable for complete peeling.
5791 @item max-unswitch-insns
5792 The maximum number of insns of an unswitched loop.
5794 @item max-unswitch-level
5795 The maximum number of branches unswitched in a single loop.
5798 The minimum cost of an expensive expression in the loop invariant motion.
5800 @item iv-consider-all-candidates-bound
5801 Bound on number of candidates for induction variables below that
5802 all candidates are considered for each use in induction variable
5803 optimizations. Only the most relevant candidates are considered
5804 if there are more candidates, to avoid quadratic time complexity.
5806 @item iv-max-considered-uses
5807 The induction variable optimizations give up on loops that contain more
5808 induction variable uses.
5810 @item iv-always-prune-cand-set-bound
5811 If number of candidates in the set is smaller than this value,
5812 we always try to remove unnecessary ivs from the set during its
5813 optimization when a new iv is added to the set.
5815 @item scev-max-expr-size
5816 Bound on size of expressions used in the scalar evolutions analyzer.
5817 Large expressions slow the analyzer.
5819 @item max-iterations-to-track
5821 The maximum number of iterations of a loop the brute force algorithm
5822 for analysis of # of iterations of the loop tries to evaluate.
5824 @item hot-bb-count-fraction
5825 Select fraction of the maximal count of repetitions of basic block in program
5826 given basic block needs to have to be considered hot.
5828 @item hot-bb-frequency-fraction
5829 Select fraction of the maximal frequency of executions of basic block in
5830 function given basic block needs to have to be considered hot
5832 @item tracer-dynamic-coverage
5833 @itemx tracer-dynamic-coverage-feedback
5835 This value is used to limit superblock formation once the given percentage of
5836 executed instructions is covered. This limits unnecessary code size
5839 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5840 feedback is available. The real profiles (as opposed to statically estimated
5841 ones) are much less balanced allowing the threshold to be larger value.
5843 @item tracer-max-code-growth
5844 Stop tail duplication once code growth has reached given percentage. This is
5845 rather hokey argument, as most of the duplicates will be eliminated later in
5846 cross jumping, so it may be set to much higher values than is the desired code
5849 @item tracer-min-branch-ratio
5851 Stop reverse growth when the reverse probability of best edge is less than this
5852 threshold (in percent).
5854 @item tracer-min-branch-ratio
5855 @itemx tracer-min-branch-ratio-feedback
5857 Stop forward growth if the best edge do have probability lower than this
5860 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
5861 compilation for profile feedback and one for compilation without. The value
5862 for compilation with profile feedback needs to be more conservative (higher) in
5863 order to make tracer effective.
5865 @item max-cse-path-length
5867 Maximum number of basic blocks on path that cse considers. The default is 10.
5869 @item global-var-threshold
5871 Counts the number of function calls (@var{n}) and the number of
5872 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
5873 single artificial variable will be created to represent all the
5874 call-clobbered variables at function call sites. This artificial
5875 variable will then be made to alias every call-clobbered variable.
5876 (done as @code{int * size_t} on the host machine; beware overflow).
5878 @item max-aliased-vops
5880 Maximum number of virtual operands allowed to represent aliases
5881 before triggering the alias grouping heuristic. Alias grouping
5882 reduces compile times and memory consumption needed for aliasing at
5883 the expense of precision loss in alias information.
5885 @item ggc-min-expand
5887 GCC uses a garbage collector to manage its own memory allocation. This
5888 parameter specifies the minimum percentage by which the garbage
5889 collector's heap should be allowed to expand between collections.
5890 Tuning this may improve compilation speed; it has no effect on code
5893 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
5894 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
5895 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
5896 GCC is not able to calculate RAM on a particular platform, the lower
5897 bound of 30% is used. Setting this parameter and
5898 @option{ggc-min-heapsize} to zero causes a full collection to occur at
5899 every opportunity. This is extremely slow, but can be useful for
5902 @item ggc-min-heapsize
5904 Minimum size of the garbage collector's heap before it begins bothering
5905 to collect garbage. The first collection occurs after the heap expands
5906 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
5907 tuning this may improve compilation speed, and has no effect on code
5910 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
5911 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
5912 with a lower bound of 4096 (four megabytes) and an upper bound of
5913 131072 (128 megabytes). If GCC is not able to calculate RAM on a
5914 particular platform, the lower bound is used. Setting this parameter
5915 very large effectively disables garbage collection. Setting this
5916 parameter and @option{ggc-min-expand} to zero causes a full collection
5917 to occur at every opportunity.
5919 @item max-reload-search-insns
5920 The maximum number of instruction reload should look backward for equivalent
5921 register. Increasing values mean more aggressive optimization, making the
5922 compile time increase with probably slightly better performance. The default
5925 @item max-cselib-memory-location
5926 The maximum number of memory locations cselib should take into acount.
5927 Increasing values mean more aggressive optimization, making the compile time
5928 increase with probably slightly better performance. The default value is 500.
5930 @item reorder-blocks-duplicate
5931 @itemx reorder-blocks-duplicate-feedback
5933 Used by basic block reordering pass to decide whether to use unconditional
5934 branch or duplicate the code on its destination. Code is duplicated when its
5935 estimated size is smaller than this value multiplied by the estimated size of
5936 unconditional jump in the hot spots of the program.
5938 The @option{reorder-block-duplicate-feedback} is used only when profile
5939 feedback is available and may be set to higher values than
5940 @option{reorder-block-duplicate} since information about the hot spots is more
5943 @item max-sched-region-blocks
5944 The maximum number of blocks in a region to be considered for
5945 interblock scheduling. The default value is 10.
5947 @item max-sched-region-insns
5948 The maximum number of insns in a region to be considered for
5949 interblock scheduling. The default value is 100.
5951 @item max-last-value-rtl
5953 The maximum size measured as number of RTLs that can be recorded in an expression
5954 in combiner for a pseudo register as last known value of that register. The default
5957 @item integer-share-limit
5958 Small integer constants can use a shared data structure, reducing the
5959 compiler's memory usage and increasing its speed. This sets the maximum
5960 value of a shared integer constant's. The default value is 256.
5962 @item min-virtual-mappings
5963 Specifies the minimum number of virtual mappings in the incremental
5964 SSA updater that should be registered to trigger the virtual mappings
5965 heuristic defined by virtual-mappings-ratio. The default value is
5968 @item virtual-mappings-ratio
5969 If the number of virtual mappings is virtual-mappings-ratio bigger
5970 than the number of virtual symbols to be updated, then the incremental
5971 SSA updater switches to a full update for those symbols. The default
5977 @node Preprocessor Options
5978 @section Options Controlling the Preprocessor
5979 @cindex preprocessor options
5980 @cindex options, preprocessor
5982 These options control the C preprocessor, which is run on each C source
5983 file before actual compilation.
5985 If you use the @option{-E} option, nothing is done except preprocessing.
5986 Some of these options make sense only together with @option{-E} because
5987 they cause the preprocessor output to be unsuitable for actual
5992 You can use @option{-Wp,@var{option}} to bypass the compiler driver
5993 and pass @var{option} directly through to the preprocessor. If
5994 @var{option} contains commas, it is split into multiple options at the
5995 commas. However, many options are modified, translated or interpreted
5996 by the compiler driver before being passed to the preprocessor, and
5997 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
5998 interface is undocumented and subject to change, so whenever possible
5999 you should avoid using @option{-Wp} and let the driver handle the
6002 @item -Xpreprocessor @var{option}
6003 @opindex preprocessor
6004 Pass @var{option} as an option to the preprocessor. You can use this to
6005 supply system-specific preprocessor options which GCC does not know how to
6008 If you want to pass an option that takes an argument, you must use
6009 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6012 @include cppopts.texi
6014 @node Assembler Options
6015 @section Passing Options to the Assembler
6017 @c prevent bad page break with this line
6018 You can pass options to the assembler.
6021 @item -Wa,@var{option}
6023 Pass @var{option} as an option to the assembler. If @var{option}
6024 contains commas, it is split into multiple options at the commas.
6026 @item -Xassembler @var{option}
6028 Pass @var{option} as an option to the assembler. You can use this to
6029 supply system-specific assembler options which GCC does not know how to
6032 If you want to pass an option that takes an argument, you must use
6033 @option{-Xassembler} twice, once for the option and once for the argument.
6038 @section Options for Linking
6039 @cindex link options
6040 @cindex options, linking
6042 These options come into play when the compiler links object files into
6043 an executable output file. They are meaningless if the compiler is
6044 not doing a link step.
6048 @item @var{object-file-name}
6049 A file name that does not end in a special recognized suffix is
6050 considered to name an object file or library. (Object files are
6051 distinguished from libraries by the linker according to the file
6052 contents.) If linking is done, these object files are used as input
6061 If any of these options is used, then the linker is not run, and
6062 object file names should not be used as arguments. @xref{Overall
6066 @item -l@var{library}
6067 @itemx -l @var{library}
6069 Search the library named @var{library} when linking. (The second
6070 alternative with the library as a separate argument is only for
6071 POSIX compliance and is not recommended.)
6073 It makes a difference where in the command you write this option; the
6074 linker searches and processes libraries and object files in the order they
6075 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6076 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6077 to functions in @samp{z}, those functions may not be loaded.
6079 The linker searches a standard list of directories for the library,
6080 which is actually a file named @file{lib@var{library}.a}. The linker
6081 then uses this file as if it had been specified precisely by name.
6083 The directories searched include several standard system directories
6084 plus any that you specify with @option{-L}.
6086 Normally the files found this way are library files---archive files
6087 whose members are object files. The linker handles an archive file by
6088 scanning through it for members which define symbols that have so far
6089 been referenced but not defined. But if the file that is found is an
6090 ordinary object file, it is linked in the usual fashion. The only
6091 difference between using an @option{-l} option and specifying a file name
6092 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6093 and searches several directories.
6097 You need this special case of the @option{-l} option in order to
6098 link an Objective-C or Objective-C++ program.
6101 @opindex nostartfiles
6102 Do not use the standard system startup files when linking.
6103 The standard system libraries are used normally, unless @option{-nostdlib}
6104 or @option{-nodefaultlibs} is used.
6106 @item -nodefaultlibs
6107 @opindex nodefaultlibs
6108 Do not use the standard system libraries when linking.
6109 Only the libraries you specify will be passed to the linker.
6110 The standard startup files are used normally, unless @option{-nostartfiles}
6111 is used. The compiler may generate calls to @code{memcmp},
6112 @code{memset}, @code{memcpy} and @code{memmove}.
6113 These entries are usually resolved by entries in
6114 libc. These entry points should be supplied through some other
6115 mechanism when this option is specified.
6119 Do not use the standard system startup files or libraries when linking.
6120 No startup files and only the libraries you specify will be passed to
6121 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6122 @code{memcpy} and @code{memmove}.
6123 These entries are usually resolved by entries in
6124 libc. These entry points should be supplied through some other
6125 mechanism when this option is specified.
6127 @cindex @option{-lgcc}, use with @option{-nostdlib}
6128 @cindex @option{-nostdlib} and unresolved references
6129 @cindex unresolved references and @option{-nostdlib}
6130 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6131 @cindex @option{-nodefaultlibs} and unresolved references
6132 @cindex unresolved references and @option{-nodefaultlibs}
6133 One of the standard libraries bypassed by @option{-nostdlib} and
6134 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6135 that GCC uses to overcome shortcomings of particular machines, or special
6136 needs for some languages.
6137 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6138 Collection (GCC) Internals},
6139 for more discussion of @file{libgcc.a}.)
6140 In most cases, you need @file{libgcc.a} even when you want to avoid
6141 other standard libraries. In other words, when you specify @option{-nostdlib}
6142 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6143 This ensures that you have no unresolved references to internal GCC
6144 library subroutines. (For example, @samp{__main}, used to ensure C++
6145 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6146 GNU Compiler Collection (GCC) Internals}.)
6150 Produce a position independent executable on targets which support it.
6151 For predictable results, you must also specify the same set of options
6152 that were used to generate code (@option{-fpie}, @option{-fPIE},
6153 or model suboptions) when you specify this option.
6157 Remove all symbol table and relocation information from the executable.
6161 On systems that support dynamic linking, this prevents linking with the shared
6162 libraries. On other systems, this option has no effect.
6166 Produce a shared object which can then be linked with other objects to
6167 form an executable. Not all systems support this option. For predictable
6168 results, you must also specify the same set of options that were used to
6169 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6170 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6171 needs to build supplementary stub code for constructors to work. On
6172 multi-libbed systems, @samp{gcc -shared} must select the correct support
6173 libraries to link against. Failing to supply the correct flags may lead
6174 to subtle defects. Supplying them in cases where they are not necessary
6177 @item -shared-libgcc
6178 @itemx -static-libgcc
6179 @opindex shared-libgcc
6180 @opindex static-libgcc
6181 On systems that provide @file{libgcc} as a shared library, these options
6182 force the use of either the shared or static version respectively.
6183 If no shared version of @file{libgcc} was built when the compiler was
6184 configured, these options have no effect.
6186 There are several situations in which an application should use the
6187 shared @file{libgcc} instead of the static version. The most common
6188 of these is when the application wishes to throw and catch exceptions
6189 across different shared libraries. In that case, each of the libraries
6190 as well as the application itself should use the shared @file{libgcc}.
6192 Therefore, the G++ and GCJ drivers automatically add
6193 @option{-shared-libgcc} whenever you build a shared library or a main
6194 executable, because C++ and Java programs typically use exceptions, so
6195 this is the right thing to do.
6197 If, instead, you use the GCC driver to create shared libraries, you may
6198 find that they will not always be linked with the shared @file{libgcc}.
6199 If GCC finds, at its configuration time, that you have a non-GNU linker
6200 or a GNU linker that does not support option @option{--eh-frame-hdr},
6201 it will link the shared version of @file{libgcc} into shared libraries
6202 by default. Otherwise, it will take advantage of the linker and optimize
6203 away the linking with the shared version of @file{libgcc}, linking with
6204 the static version of libgcc by default. This allows exceptions to
6205 propagate through such shared libraries, without incurring relocation
6206 costs at library load time.
6208 However, if a library or main executable is supposed to throw or catch
6209 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6210 for the languages used in the program, or using the option
6211 @option{-shared-libgcc}, such that it is linked with the shared
6216 Bind references to global symbols when building a shared object. Warn
6217 about any unresolved references (unless overridden by the link editor
6218 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6221 @item -Xlinker @var{option}
6223 Pass @var{option} as an option to the linker. You can use this to
6224 supply system-specific linker options which GCC does not know how to
6227 If you want to pass an option that takes an argument, you must use
6228 @option{-Xlinker} twice, once for the option and once for the argument.
6229 For example, to pass @option{-assert definitions}, you must write
6230 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6231 @option{-Xlinker "-assert definitions"}, because this passes the entire
6232 string as a single argument, which is not what the linker expects.
6234 @item -Wl,@var{option}
6236 Pass @var{option} as an option to the linker. If @var{option} contains
6237 commas, it is split into multiple options at the commas.
6239 @item -u @var{symbol}
6241 Pretend the symbol @var{symbol} is undefined, to force linking of
6242 library modules to define it. You can use @option{-u} multiple times with
6243 different symbols to force loading of additional library modules.
6246 @node Directory Options
6247 @section Options for Directory Search
6248 @cindex directory options
6249 @cindex options, directory search
6252 These options specify directories to search for header files, for
6253 libraries and for parts of the compiler:
6258 Add the directory @var{dir} to the head of the list of directories to be
6259 searched for header files. This can be used to override a system header
6260 file, substituting your own version, since these directories are
6261 searched before the system header file directories. However, you should
6262 not use this option to add directories that contain vendor-supplied
6263 system header files (use @option{-isystem} for that). If you use more than
6264 one @option{-I} option, the directories are scanned in left-to-right
6265 order; the standard system directories come after.
6267 If a standard system include directory, or a directory specified with
6268 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6269 option will be ignored. The directory will still be searched but as a
6270 system directory at its normal position in the system include chain.
6271 This is to ensure that GCC's procedure to fix buggy system headers and
6272 the ordering for the include_next directive are not inadvertently changed.
6273 If you really need to change the search order for system directories,
6274 use the @option{-nostdinc} and/or @option{-isystem} options.
6276 @item -iquote@var{dir}
6278 Add the directory @var{dir} to the head of the list of directories to
6279 be searched for header files only for the case of @samp{#include
6280 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6281 otherwise just like @option{-I}.
6285 Add directory @var{dir} to the list of directories to be searched
6288 @item -B@var{prefix}
6290 This option specifies where to find the executables, libraries,
6291 include files, and data files of the compiler itself.
6293 The compiler driver program runs one or more of the subprograms
6294 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6295 @var{prefix} as a prefix for each program it tries to run, both with and
6296 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6298 For each subprogram to be run, the compiler driver first tries the
6299 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6300 was not specified, the driver tries two standard prefixes, which are
6301 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6302 those results in a file name that is found, the unmodified program
6303 name is searched for using the directories specified in your
6304 @env{PATH} environment variable.
6306 The compiler will check to see if the path provided by the @option{-B}
6307 refers to a directory, and if necessary it will add a directory
6308 separator character at the end of the path.
6310 @option{-B} prefixes that effectively specify directory names also apply
6311 to libraries in the linker, because the compiler translates these
6312 options into @option{-L} options for the linker. They also apply to
6313 includes files in the preprocessor, because the compiler translates these
6314 options into @option{-isystem} options for the preprocessor. In this case,
6315 the compiler appends @samp{include} to the prefix.
6317 The run-time support file @file{libgcc.a} can also be searched for using
6318 the @option{-B} prefix, if needed. If it is not found there, the two
6319 standard prefixes above are tried, and that is all. The file is left
6320 out of the link if it is not found by those means.
6322 Another way to specify a prefix much like the @option{-B} prefix is to use
6323 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6326 As a special kludge, if the path provided by @option{-B} is
6327 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6328 9, then it will be replaced by @file{[dir/]include}. This is to help
6329 with boot-strapping the compiler.
6331 @item -specs=@var{file}
6333 Process @var{file} after the compiler reads in the standard @file{specs}
6334 file, in order to override the defaults that the @file{gcc} driver
6335 program uses when determining what switches to pass to @file{cc1},
6336 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6337 @option{-specs=@var{file}} can be specified on the command line, and they
6338 are processed in order, from left to right.
6342 This option has been deprecated. Please use @option{-iquote} instead for
6343 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6344 Any directories you specify with @option{-I} options before the @option{-I-}
6345 option are searched only for the case of @samp{#include "@var{file}"};
6346 they are not searched for @samp{#include <@var{file}>}.
6348 If additional directories are specified with @option{-I} options after
6349 the @option{-I-}, these directories are searched for all @samp{#include}
6350 directives. (Ordinarily @emph{all} @option{-I} directories are used
6353 In addition, the @option{-I-} option inhibits the use of the current
6354 directory (where the current input file came from) as the first search
6355 directory for @samp{#include "@var{file}"}. There is no way to
6356 override this effect of @option{-I-}. With @option{-I.} you can specify
6357 searching the directory which was current when the compiler was
6358 invoked. That is not exactly the same as what the preprocessor does
6359 by default, but it is often satisfactory.
6361 @option{-I-} does not inhibit the use of the standard system directories
6362 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6369 @section Specifying subprocesses and the switches to pass to them
6372 @command{gcc} is a driver program. It performs its job by invoking a
6373 sequence of other programs to do the work of compiling, assembling and
6374 linking. GCC interprets its command-line parameters and uses these to
6375 deduce which programs it should invoke, and which command-line options
6376 it ought to place on their command lines. This behavior is controlled
6377 by @dfn{spec strings}. In most cases there is one spec string for each
6378 program that GCC can invoke, but a few programs have multiple spec
6379 strings to control their behavior. The spec strings built into GCC can
6380 be overridden by using the @option{-specs=} command-line switch to specify
6383 @dfn{Spec files} are plaintext files that are used to construct spec
6384 strings. They consist of a sequence of directives separated by blank
6385 lines. The type of directive is determined by the first non-whitespace
6386 character on the line and it can be one of the following:
6389 @item %@var{command}
6390 Issues a @var{command} to the spec file processor. The commands that can
6394 @item %include <@var{file}>
6396 Search for @var{file} and insert its text at the current point in the
6399 @item %include_noerr <@var{file}>
6400 @cindex %include_noerr
6401 Just like @samp{%include}, but do not generate an error message if the include
6402 file cannot be found.
6404 @item %rename @var{old_name} @var{new_name}
6406 Rename the spec string @var{old_name} to @var{new_name}.
6410 @item *[@var{spec_name}]:
6411 This tells the compiler to create, override or delete the named spec
6412 string. All lines after this directive up to the next directive or
6413 blank line are considered to be the text for the spec string. If this
6414 results in an empty string then the spec will be deleted. (Or, if the
6415 spec did not exist, then nothing will happened.) Otherwise, if the spec
6416 does not currently exist a new spec will be created. If the spec does
6417 exist then its contents will be overridden by the text of this
6418 directive, unless the first character of that text is the @samp{+}
6419 character, in which case the text will be appended to the spec.
6421 @item [@var{suffix}]:
6422 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6423 and up to the next directive or blank line are considered to make up the
6424 spec string for the indicated suffix. When the compiler encounters an
6425 input file with the named suffix, it will processes the spec string in
6426 order to work out how to compile that file. For example:
6433 This says that any input file whose name ends in @samp{.ZZ} should be
6434 passed to the program @samp{z-compile}, which should be invoked with the
6435 command-line switch @option{-input} and with the result of performing the
6436 @samp{%i} substitution. (See below.)
6438 As an alternative to providing a spec string, the text that follows a
6439 suffix directive can be one of the following:
6442 @item @@@var{language}
6443 This says that the suffix is an alias for a known @var{language}. This is
6444 similar to using the @option{-x} command-line switch to GCC to specify a
6445 language explicitly. For example:
6452 Says that .ZZ files are, in fact, C++ source files.
6455 This causes an error messages saying:
6458 @var{name} compiler not installed on this system.
6462 GCC already has an extensive list of suffixes built into it.
6463 This directive will add an entry to the end of the list of suffixes, but
6464 since the list is searched from the end backwards, it is effectively
6465 possible to override earlier entries using this technique.
6469 GCC has the following spec strings built into it. Spec files can
6470 override these strings or create their own. Note that individual
6471 targets can also add their own spec strings to this list.
6474 asm Options to pass to the assembler
6475 asm_final Options to pass to the assembler post-processor
6476 cpp Options to pass to the C preprocessor
6477 cc1 Options to pass to the C compiler
6478 cc1plus Options to pass to the C++ compiler
6479 endfile Object files to include at the end of the link
6480 link Options to pass to the linker
6481 lib Libraries to include on the command line to the linker
6482 libgcc Decides which GCC support library to pass to the linker
6483 linker Sets the name of the linker
6484 predefines Defines to be passed to the C preprocessor
6485 signed_char Defines to pass to CPP to say whether @code{char} is signed
6487 startfile Object files to include at the start of the link
6490 Here is a small example of a spec file:
6496 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6499 This example renames the spec called @samp{lib} to @samp{old_lib} and
6500 then overrides the previous definition of @samp{lib} with a new one.
6501 The new definition adds in some extra command-line options before
6502 including the text of the old definition.
6504 @dfn{Spec strings} are a list of command-line options to be passed to their
6505 corresponding program. In addition, the spec strings can contain
6506 @samp{%}-prefixed sequences to substitute variable text or to
6507 conditionally insert text into the command line. Using these constructs
6508 it is possible to generate quite complex command lines.
6510 Here is a table of all defined @samp{%}-sequences for spec
6511 strings. Note that spaces are not generated automatically around the
6512 results of expanding these sequences. Therefore you can concatenate them
6513 together or combine them with constant text in a single argument.
6517 Substitute one @samp{%} into the program name or argument.
6520 Substitute the name of the input file being processed.
6523 Substitute the basename of the input file being processed.
6524 This is the substring up to (and not including) the last period
6525 and not including the directory.
6528 This is the same as @samp{%b}, but include the file suffix (text after
6532 Marks the argument containing or following the @samp{%d} as a
6533 temporary file name, so that that file will be deleted if GCC exits
6534 successfully. Unlike @samp{%g}, this contributes no text to the
6537 @item %g@var{suffix}
6538 Substitute a file name that has suffix @var{suffix} and is chosen
6539 once per compilation, and mark the argument in the same way as
6540 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6541 name is now chosen in a way that is hard to predict even when previously
6542 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6543 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6544 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6545 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6546 was simply substituted with a file name chosen once per compilation,
6547 without regard to any appended suffix (which was therefore treated
6548 just like ordinary text), making such attacks more likely to succeed.
6550 @item %u@var{suffix}
6551 Like @samp{%g}, but generates a new temporary file name even if
6552 @samp{%u@var{suffix}} was already seen.
6554 @item %U@var{suffix}
6555 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6556 new one if there is no such last file name. In the absence of any
6557 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6558 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6559 would involve the generation of two distinct file names, one
6560 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6561 simply substituted with a file name chosen for the previous @samp{%u},
6562 without regard to any appended suffix.
6564 @item %j@var{suffix}
6565 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6566 writable, and if save-temps is off; otherwise, substitute the name
6567 of a temporary file, just like @samp{%u}. This temporary file is not
6568 meant for communication between processes, but rather as a junk
6571 @item %|@var{suffix}
6572 @itemx %m@var{suffix}
6573 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6574 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6575 all. These are the two most common ways to instruct a program that it
6576 should read from standard input or write to standard output. If you
6577 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6578 construct: see for example @file{f/lang-specs.h}.
6580 @item %.@var{SUFFIX}
6581 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6582 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6583 terminated by the next space or %.
6586 Marks the argument containing or following the @samp{%w} as the
6587 designated output file of this compilation. This puts the argument
6588 into the sequence of arguments that @samp{%o} will substitute later.
6591 Substitutes the names of all the output files, with spaces
6592 automatically placed around them. You should write spaces
6593 around the @samp{%o} as well or the results are undefined.
6594 @samp{%o} is for use in the specs for running the linker.
6595 Input files whose names have no recognized suffix are not compiled
6596 at all, but they are included among the output files, so they will
6600 Substitutes the suffix for object files. Note that this is
6601 handled specially when it immediately follows @samp{%g, %u, or %U},
6602 because of the need for those to form complete file names. The
6603 handling is such that @samp{%O} is treated exactly as if it had already
6604 been substituted, except that @samp{%g, %u, and %U} do not currently
6605 support additional @var{suffix} characters following @samp{%O} as they would
6606 following, for example, @samp{.o}.
6609 Substitutes the standard macro predefinitions for the
6610 current target machine. Use this when running @code{cpp}.
6613 Like @samp{%p}, but puts @samp{__} before and after the name of each
6614 predefined macro, except for macros that start with @samp{__} or with
6615 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6619 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6620 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
6621 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6625 Current argument is the name of a library or startup file of some sort.
6626 Search for that file in a standard list of directories and substitute
6627 the full name found.
6630 Print @var{str} as an error message. @var{str} is terminated by a newline.
6631 Use this when inconsistent options are detected.
6634 Substitute the contents of spec string @var{name} at this point.
6637 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6639 @item %x@{@var{option}@}
6640 Accumulate an option for @samp{%X}.
6643 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6647 Output the accumulated assembler options specified by @option{-Wa}.
6650 Output the accumulated preprocessor options specified by @option{-Wp}.
6653 Process the @code{asm} spec. This is used to compute the
6654 switches to be passed to the assembler.
6657 Process the @code{asm_final} spec. This is a spec string for
6658 passing switches to an assembler post-processor, if such a program is
6662 Process the @code{link} spec. This is the spec for computing the
6663 command line passed to the linker. Typically it will make use of the
6664 @samp{%L %G %S %D and %E} sequences.
6667 Dump out a @option{-L} option for each directory that GCC believes might
6668 contain startup files. If the target supports multilibs then the
6669 current multilib directory will be prepended to each of these paths.
6672 Process the @code{lib} spec. This is a spec string for deciding which
6673 libraries should be included on the command line to the linker.
6676 Process the @code{libgcc} spec. This is a spec string for deciding
6677 which GCC support library should be included on the command line to the linker.
6680 Process the @code{startfile} spec. This is a spec for deciding which
6681 object files should be the first ones passed to the linker. Typically
6682 this might be a file named @file{crt0.o}.
6685 Process the @code{endfile} spec. This is a spec string that specifies
6686 the last object files that will be passed to the linker.
6689 Process the @code{cpp} spec. This is used to construct the arguments
6690 to be passed to the C preprocessor.
6693 Process the @code{cc1} spec. This is used to construct the options to be
6694 passed to the actual C compiler (@samp{cc1}).
6697 Process the @code{cc1plus} spec. This is used to construct the options to be
6698 passed to the actual C++ compiler (@samp{cc1plus}).
6701 Substitute the variable part of a matched option. See below.
6702 Note that each comma in the substituted string is replaced by
6706 Remove all occurrences of @code{-S} from the command line. Note---this
6707 command is position dependent. @samp{%} commands in the spec string
6708 before this one will see @code{-S}, @samp{%} commands in the spec string
6709 after this one will not.
6711 @item %:@var{function}(@var{args})
6712 Call the named function @var{function}, passing it @var{args}.
6713 @var{args} is first processed as a nested spec string, then split
6714 into an argument vector in the usual fashion. The function returns
6715 a string which is processed as if it had appeared literally as part
6716 of the current spec.
6718 The following built-in spec functions are provided:
6721 @item @code{if-exists}
6722 The @code{if-exists} spec function takes one argument, an absolute
6723 pathname to a file. If the file exists, @code{if-exists} returns the
6724 pathname. Here is a small example of its usage:
6728 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6731 @item @code{if-exists-else}
6732 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6733 spec function, except that it takes two arguments. The first argument is
6734 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6735 returns the pathname. If it does not exist, it returns the second argument.
6736 This way, @code{if-exists-else} can be used to select one file or another,
6737 based on the existence of the first. Here is a small example of its usage:
6741 crt0%O%s %:if-exists(crti%O%s) \
6742 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6745 @item @code{replace-outfile}
6746 The @code{replace-outfile} spec function takes two arguments. It looks for the
6747 first argument in the outfiles array and replaces it with the second argument. Here
6748 is a small example of its usage:
6751 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
6757 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6758 If that switch was not specified, this substitutes nothing. Note that
6759 the leading dash is omitted when specifying this option, and it is
6760 automatically inserted if the substitution is performed. Thus the spec
6761 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6762 and would output the command line option @option{-foo}.
6764 @item %W@{@code{S}@}
6765 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6768 @item %@{@code{S}*@}
6769 Substitutes all the switches specified to GCC whose names start
6770 with @code{-S}, but which also take an argument. This is used for
6771 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6772 GCC considers @option{-o foo} as being
6773 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6774 text, including the space. Thus two arguments would be generated.
6776 @item %@{@code{S}*&@code{T}*@}
6777 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6778 (the order of @code{S} and @code{T} in the spec is not significant).
6779 There can be any number of ampersand-separated variables; for each the
6780 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6782 @item %@{@code{S}:@code{X}@}
6783 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6785 @item %@{!@code{S}:@code{X}@}
6786 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6788 @item %@{@code{S}*:@code{X}@}
6789 Substitutes @code{X} if one or more switches whose names start with
6790 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6791 once, no matter how many such switches appeared. However, if @code{%*}
6792 appears somewhere in @code{X}, then @code{X} will be substituted once
6793 for each matching switch, with the @code{%*} replaced by the part of
6794 that switch that matched the @code{*}.
6796 @item %@{.@code{S}:@code{X}@}
6797 Substitutes @code{X}, if processing a file with suffix @code{S}.
6799 @item %@{!.@code{S}:@code{X}@}
6800 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6802 @item %@{@code{S}|@code{P}:@code{X}@}
6803 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6804 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
6805 although they have a stronger binding than the @samp{|}. If @code{%*}
6806 appears in @code{X}, all of the alternatives must be starred, and only
6807 the first matching alternative is substituted.
6809 For example, a spec string like this:
6812 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
6815 will output the following command-line options from the following input
6816 command-line options:
6821 -d fred.c -foo -baz -boggle
6822 -d jim.d -bar -baz -boggle
6825 @item %@{S:X; T:Y; :D@}
6827 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
6828 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
6829 be as many clauses as you need. This may be combined with @code{.},
6830 @code{!}, @code{|}, and @code{*} as needed.
6835 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
6836 construct may contain other nested @samp{%} constructs or spaces, or
6837 even newlines. They are processed as usual, as described above.
6838 Trailing white space in @code{X} is ignored. White space may also
6839 appear anywhere on the left side of the colon in these constructs,
6840 except between @code{.} or @code{*} and the corresponding word.
6842 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
6843 handled specifically in these constructs. If another value of
6844 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
6845 @option{-W} switch is found later in the command line, the earlier
6846 switch value is ignored, except with @{@code{S}*@} where @code{S} is
6847 just one letter, which passes all matching options.
6849 The character @samp{|} at the beginning of the predicate text is used to
6850 indicate that a command should be piped to the following command, but
6851 only if @option{-pipe} is specified.
6853 It is built into GCC which switches take arguments and which do not.
6854 (You might think it would be useful to generalize this to allow each
6855 compiler's spec to say which switches take arguments. But this cannot
6856 be done in a consistent fashion. GCC cannot even decide which input
6857 files have been specified without knowing which switches take arguments,
6858 and it must know which input files to compile in order to tell which
6861 GCC also knows implicitly that arguments starting in @option{-l} are to be
6862 treated as compiler output files, and passed to the linker in their
6863 proper position among the other output files.
6865 @c man begin OPTIONS
6867 @node Target Options
6868 @section Specifying Target Machine and Compiler Version
6869 @cindex target options
6870 @cindex cross compiling
6871 @cindex specifying machine version
6872 @cindex specifying compiler version and target machine
6873 @cindex compiler version, specifying
6874 @cindex target machine, specifying
6876 The usual way to run GCC is to run the executable called @file{gcc}, or
6877 @file{<machine>-gcc} when cross-compiling, or
6878 @file{<machine>-gcc-<version>} to run a version other than the one that
6879 was installed last. Sometimes this is inconvenient, so GCC provides
6880 options that will switch to another cross-compiler or version.
6883 @item -b @var{machine}
6885 The argument @var{machine} specifies the target machine for compilation.
6887 The value to use for @var{machine} is the same as was specified as the
6888 machine type when configuring GCC as a cross-compiler. For
6889 example, if a cross-compiler was configured with @samp{configure
6890 i386v}, meaning to compile for an 80386 running System V, then you
6891 would specify @option{-b i386v} to run that cross compiler.
6893 @item -V @var{version}
6895 The argument @var{version} specifies which version of GCC to run.
6896 This is useful when multiple versions are installed. For example,
6897 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
6900 The @option{-V} and @option{-b} options work by running the
6901 @file{<machine>-gcc-<version>} executable, so there's no real reason to
6902 use them if you can just run that directly.
6904 @node Submodel Options
6905 @section Hardware Models and Configurations
6906 @cindex submodel options
6907 @cindex specifying hardware config
6908 @cindex hardware models and configurations, specifying
6909 @cindex machine dependent options
6911 Earlier we discussed the standard option @option{-b} which chooses among
6912 different installed compilers for completely different target
6913 machines, such as VAX vs.@: 68000 vs.@: 80386.
6915 In addition, each of these target machine types can have its own
6916 special options, starting with @samp{-m}, to choose among various
6917 hardware models or configurations---for example, 68010 vs 68020,
6918 floating coprocessor or none. A single installed version of the
6919 compiler can compile for any model or configuration, according to the
6922 Some configurations of the compiler also support additional special
6923 options, usually for compatibility with other compilers on the same
6926 @c This list is ordered alphanumerically by subsection name.
6927 @c It should be the same order and spelling as these options are listed
6928 @c in Machine Dependent Options
6934 * Blackfin Options::
6937 * DEC Alpha Options::
6938 * DEC Alpha/VMS Options::
6942 * i386 and x86-64 Options::
6954 * RS/6000 and PowerPC Options::
6955 * S/390 and zSeries Options::
6958 * System V Options::
6959 * TMS320C3x/C4x Options::
6963 * Xstormy16 Options::
6969 @subsection ARC Options
6972 These options are defined for ARC implementations:
6977 Compile code for little endian mode. This is the default.
6981 Compile code for big endian mode.
6984 @opindex mmangle-cpu
6985 Prepend the name of the cpu to all public symbol names.
6986 In multiple-processor systems, there are many ARC variants with different
6987 instruction and register set characteristics. This flag prevents code
6988 compiled for one cpu to be linked with code compiled for another.
6989 No facility exists for handling variants that are ``almost identical''.
6990 This is an all or nothing option.
6992 @item -mcpu=@var{cpu}
6994 Compile code for ARC variant @var{cpu}.
6995 Which variants are supported depend on the configuration.
6996 All variants support @option{-mcpu=base}, this is the default.
6998 @item -mtext=@var{text-section}
6999 @itemx -mdata=@var{data-section}
7000 @itemx -mrodata=@var{readonly-data-section}
7004 Put functions, data, and readonly data in @var{text-section},
7005 @var{data-section}, and @var{readonly-data-section} respectively
7006 by default. This can be overridden with the @code{section} attribute.
7007 @xref{Variable Attributes}.
7012 @subsection ARM Options
7015 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7019 @item -mabi=@var{name}
7021 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7022 @samp{atpcs}, @samp{aapcs} and @samp{iwmmxt}.
7025 @opindex mapcs-frame
7026 Generate a stack frame that is compliant with the ARM Procedure Call
7027 Standard for all functions, even if this is not strictly necessary for
7028 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7029 with this option will cause the stack frames not to be generated for
7030 leaf functions. The default is @option{-mno-apcs-frame}.
7034 This is a synonym for @option{-mapcs-frame}.
7037 @c not currently implemented
7038 @item -mapcs-stack-check
7039 @opindex mapcs-stack-check
7040 Generate code to check the amount of stack space available upon entry to
7041 every function (that actually uses some stack space). If there is
7042 insufficient space available then either the function
7043 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7044 called, depending upon the amount of stack space required. The run time
7045 system is required to provide these functions. The default is
7046 @option{-mno-apcs-stack-check}, since this produces smaller code.
7048 @c not currently implemented
7050 @opindex mapcs-float
7051 Pass floating point arguments using the float point registers. This is
7052 one of the variants of the APCS@. This option is recommended if the
7053 target hardware has a floating point unit or if a lot of floating point
7054 arithmetic is going to be performed by the code. The default is
7055 @option{-mno-apcs-float}, since integer only code is slightly increased in
7056 size if @option{-mapcs-float} is used.
7058 @c not currently implemented
7059 @item -mapcs-reentrant
7060 @opindex mapcs-reentrant
7061 Generate reentrant, position independent code. The default is
7062 @option{-mno-apcs-reentrant}.
7065 @item -mthumb-interwork
7066 @opindex mthumb-interwork
7067 Generate code which supports calling between the ARM and Thumb
7068 instruction sets. Without this option the two instruction sets cannot
7069 be reliably used inside one program. The default is
7070 @option{-mno-thumb-interwork}, since slightly larger code is generated
7071 when @option{-mthumb-interwork} is specified.
7073 @item -mno-sched-prolog
7074 @opindex mno-sched-prolog
7075 Prevent the reordering of instructions in the function prolog, or the
7076 merging of those instruction with the instructions in the function's
7077 body. This means that all functions will start with a recognizable set
7078 of instructions (or in fact one of a choice from a small set of
7079 different function prologues), and this information can be used to
7080 locate the start if functions inside an executable piece of code. The
7081 default is @option{-msched-prolog}.
7084 @opindex mhard-float
7085 Generate output containing floating point instructions. This is the
7089 @opindex msoft-float
7090 Generate output containing library calls for floating point.
7091 @strong{Warning:} the requisite libraries are not available for all ARM
7092 targets. Normally the facilities of the machine's usual C compiler are
7093 used, but this cannot be done directly in cross-compilation. You must make
7094 your own arrangements to provide suitable library functions for
7097 @option{-msoft-float} changes the calling convention in the output file;
7098 therefore, it is only useful if you compile @emph{all} of a program with
7099 this option. In particular, you need to compile @file{libgcc.a}, the
7100 library that comes with GCC, with @option{-msoft-float} in order for
7103 @item -mfloat-abi=@var{name}
7105 Specifies which ABI to use for floating point values. Permissible values
7106 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7108 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7109 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7110 of floating point instructions, but still uses the soft-float calling
7113 @item -mlittle-endian
7114 @opindex mlittle-endian
7115 Generate code for a processor running in little-endian mode. This is
7116 the default for all standard configurations.
7119 @opindex mbig-endian
7120 Generate code for a processor running in big-endian mode; the default is
7121 to compile code for a little-endian processor.
7123 @item -mwords-little-endian
7124 @opindex mwords-little-endian
7125 This option only applies when generating code for big-endian processors.
7126 Generate code for a little-endian word order but a big-endian byte
7127 order. That is, a byte order of the form @samp{32107654}. Note: this
7128 option should only be used if you require compatibility with code for
7129 big-endian ARM processors generated by versions of the compiler prior to
7132 @item -mcpu=@var{name}
7134 This specifies the name of the target ARM processor. GCC uses this name
7135 to determine what kind of instructions it can emit when generating
7136 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7137 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7138 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7139 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7140 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7141 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7142 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7143 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7144 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7145 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7146 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7147 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7148 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7149 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
7152 @itemx -mtune=@var{name}
7154 This option is very similar to the @option{-mcpu=} option, except that
7155 instead of specifying the actual target processor type, and hence
7156 restricting which instructions can be used, it specifies that GCC should
7157 tune the performance of the code as if the target were of the type
7158 specified in this option, but still choosing the instructions that it
7159 will generate based on the cpu specified by a @option{-mcpu=} option.
7160 For some ARM implementations better performance can be obtained by using
7163 @item -march=@var{name}
7165 This specifies the name of the target ARM architecture. GCC uses this
7166 name to determine what kind of instructions it can emit when generating
7167 assembly code. This option can be used in conjunction with or instead
7168 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7169 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7170 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7171 @samp{iwmmxt}, @samp{ep9312}.
7173 @item -mfpu=@var{name}
7174 @itemx -mfpe=@var{number}
7175 @itemx -mfp=@var{number}
7179 This specifies what floating point hardware (or hardware emulation) is
7180 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7181 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7182 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7183 with older versions of GCC@.
7185 If @option{-msoft-float} is specified this specifies the format of
7186 floating point values.
7188 @item -mstructure-size-boundary=@var{n}
7189 @opindex mstructure-size-boundary
7190 The size of all structures and unions will be rounded up to a multiple
7191 of the number of bits set by this option. Permissible values are 8, 32
7192 and 64. The default value varies for different toolchains. For the COFF
7193 targeted toolchain the default value is 8. A value of 64 is only allowed
7194 if the underlying ABI supports it.
7196 Specifying the larger number can produce faster, more efficient code, but
7197 can also increase the size of the program. Different values are potentially
7198 incompatible. Code compiled with one value cannot necessarily expect to
7199 work with code or libraries compiled with another value, if they exchange
7200 information using structures or unions.
7202 @item -mabort-on-noreturn
7203 @opindex mabort-on-noreturn
7204 Generate a call to the function @code{abort} at the end of a
7205 @code{noreturn} function. It will be executed if the function tries to
7209 @itemx -mno-long-calls
7210 @opindex mlong-calls
7211 @opindex mno-long-calls
7212 Tells the compiler to perform function calls by first loading the
7213 address of the function into a register and then performing a subroutine
7214 call on this register. This switch is needed if the target function
7215 will lie outside of the 64 megabyte addressing range of the offset based
7216 version of subroutine call instruction.
7218 Even if this switch is enabled, not all function calls will be turned
7219 into long calls. The heuristic is that static functions, functions
7220 which have the @samp{short-call} attribute, functions that are inside
7221 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7222 definitions have already been compiled within the current compilation
7223 unit, will not be turned into long calls. The exception to this rule is
7224 that weak function definitions, functions with the @samp{long-call}
7225 attribute or the @samp{section} attribute, and functions that are within
7226 the scope of a @samp{#pragma long_calls} directive, will always be
7227 turned into long calls.
7229 This feature is not enabled by default. Specifying
7230 @option{-mno-long-calls} will restore the default behavior, as will
7231 placing the function calls within the scope of a @samp{#pragma
7232 long_calls_off} directive. Note these switches have no effect on how
7233 the compiler generates code to handle function calls via function
7236 @item -mnop-fun-dllimport
7237 @opindex mnop-fun-dllimport
7238 Disable support for the @code{dllimport} attribute.
7240 @item -msingle-pic-base
7241 @opindex msingle-pic-base
7242 Treat the register used for PIC addressing as read-only, rather than
7243 loading it in the prologue for each function. The run-time system is
7244 responsible for initializing this register with an appropriate value
7245 before execution begins.
7247 @item -mpic-register=@var{reg}
7248 @opindex mpic-register
7249 Specify the register to be used for PIC addressing. The default is R10
7250 unless stack-checking is enabled, when R9 is used.
7252 @item -mcirrus-fix-invalid-insns
7253 @opindex mcirrus-fix-invalid-insns
7254 @opindex mno-cirrus-fix-invalid-insns
7255 Insert NOPs into the instruction stream to in order to work around
7256 problems with invalid Maverick instruction combinations. This option
7257 is only valid if the @option{-mcpu=ep9312} option has been used to
7258 enable generation of instructions for the Cirrus Maverick floating
7259 point co-processor. This option is not enabled by default, since the
7260 problem is only present in older Maverick implementations. The default
7261 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7264 @item -mpoke-function-name
7265 @opindex mpoke-function-name
7266 Write the name of each function into the text section, directly
7267 preceding the function prologue. The generated code is similar to this:
7271 .ascii "arm_poke_function_name", 0
7274 .word 0xff000000 + (t1 - t0)
7275 arm_poke_function_name
7277 stmfd sp!, @{fp, ip, lr, pc@}
7281 When performing a stack backtrace, code can inspect the value of
7282 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7283 location @code{pc - 12} and the top 8 bits are set, then we know that
7284 there is a function name embedded immediately preceding this location
7285 and has length @code{((pc[-3]) & 0xff000000)}.
7289 Generate code for the 16-bit Thumb instruction set. The default is to
7290 use the 32-bit ARM instruction set.
7293 @opindex mtpcs-frame
7294 Generate a stack frame that is compliant with the Thumb Procedure Call
7295 Standard for all non-leaf functions. (A leaf function is one that does
7296 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7298 @item -mtpcs-leaf-frame
7299 @opindex mtpcs-leaf-frame
7300 Generate a stack frame that is compliant with the Thumb Procedure Call
7301 Standard for all leaf functions. (A leaf function is one that does
7302 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7304 @item -mcallee-super-interworking
7305 @opindex mcallee-super-interworking
7306 Gives all externally visible functions in the file being compiled an ARM
7307 instruction set header which switches to Thumb mode before executing the
7308 rest of the function. This allows these functions to be called from
7309 non-interworking code.
7311 @item -mcaller-super-interworking
7312 @opindex mcaller-super-interworking
7313 Allows calls via function pointers (including virtual functions) to
7314 execute correctly regardless of whether the target code has been
7315 compiled for interworking or not. There is a small overhead in the cost
7316 of executing a function pointer if this option is enabled.
7321 @subsection AVR Options
7324 These options are defined for AVR implementations:
7327 @item -mmcu=@var{mcu}
7329 Specify ATMEL AVR instruction set or MCU type.
7331 Instruction set avr1 is for the minimal AVR core, not supported by the C
7332 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7333 attiny11, attiny12, attiny15, attiny28).
7335 Instruction set avr2 (default) is for the classic AVR core with up to
7336 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7337 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7338 at90c8534, at90s8535).
7340 Instruction set avr3 is for the classic AVR core with up to 128K program
7341 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7343 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7344 memory space (MCU types: atmega8, atmega83, atmega85).
7346 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7347 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7348 atmega64, atmega128, at43usb355, at94k).
7352 Output instruction sizes to the asm file.
7354 @item -minit-stack=@var{N}
7355 @opindex minit-stack
7356 Specify the initial stack address, which may be a symbol or numeric value,
7357 @samp{__stack} is the default.
7359 @item -mno-interrupts
7360 @opindex mno-interrupts
7361 Generated code is not compatible with hardware interrupts.
7362 Code size will be smaller.
7364 @item -mcall-prologues
7365 @opindex mcall-prologues
7366 Functions prologues/epilogues expanded as call to appropriate
7367 subroutines. Code size will be smaller.
7369 @item -mno-tablejump
7370 @opindex mno-tablejump
7371 Do not generate tablejump insns which sometimes increase code size.
7374 @opindex mtiny-stack
7375 Change only the low 8 bits of the stack pointer.
7379 Assume int to be 8 bit integer. This affects the sizes of all types: A
7380 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7381 and long long will be 4 bytes. Please note that this option does not
7382 comply to the C standards, but it will provide you with smaller code
7386 @node Blackfin Options
7387 @subsection Blackfin Options
7388 @cindex Blackfin Options
7391 @item -momit-leaf-frame-pointer
7392 @opindex momit-leaf-frame-pointer
7393 Don't keep the frame pointer in a register for leaf functions. This
7394 avoids the instructions to save, set up and restore frame pointers and
7395 makes an extra register available in leaf functions. The option
7396 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7397 which might make debugging harder.
7401 When enabled, the compiler will ensure that the generated code does not
7402 contain speculative loads after jump instructions. This option is enabled
7407 Don't generate extra code to prevent speculative loads from occurring.
7411 When enabled, the compiler is free to take advantage of the knowledge that
7412 the entire program fits into the low 64k of memory.
7415 @opindex mno-low-64k
7416 Assume that the program is arbitrarily large. This is the default.
7418 @item -mid-shared-library
7419 @opindex mid-shared-library
7420 Generate code that supports shared libraries via the library ID method.
7421 This allows for execute in place and shared libraries in an environment
7422 without virtual memory management. This option implies @option{-fPIC}.
7424 @item -mno-id-shared-library
7425 @opindex mno-id-shared-library
7426 Generate code that doesn't assume ID based shared libraries are being used.
7427 This is the default.
7429 @item -mshared-library-id=n
7430 @opindex mshared-library-id
7431 Specified the identification number of the ID based shared library being
7432 compiled. Specifying a value of 0 will generate more compact code, specifying
7433 other values will force the allocation of that number to the current
7434 library but is no more space or time efficient than omitting this option.
7437 @itemx -mno-long-calls
7438 @opindex mlong-calls
7439 @opindex mno-long-calls
7440 Tells the compiler to perform function calls by first loading the
7441 address of the function into a register and then performing a subroutine
7442 call on this register. This switch is needed if the target function
7443 will lie outside of the 24 bit addressing range of the offset based
7444 version of subroutine call instruction.
7446 This feature is not enabled by default. Specifying
7447 @option{-mno-long-calls} will restore the default behavior. Note these
7448 switches have no effect on how the compiler generates code to handle
7449 function calls via function pointers.
7453 @subsection CRIS Options
7454 @cindex CRIS Options
7456 These options are defined specifically for the CRIS ports.
7459 @item -march=@var{architecture-type}
7460 @itemx -mcpu=@var{architecture-type}
7463 Generate code for the specified architecture. The choices for
7464 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7465 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7466 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7469 @item -mtune=@var{architecture-type}
7471 Tune to @var{architecture-type} everything applicable about the generated
7472 code, except for the ABI and the set of available instructions. The
7473 choices for @var{architecture-type} are the same as for
7474 @option{-march=@var{architecture-type}}.
7476 @item -mmax-stack-frame=@var{n}
7477 @opindex mmax-stack-frame
7478 Warn when the stack frame of a function exceeds @var{n} bytes.
7480 @item -melinux-stacksize=@var{n}
7481 @opindex melinux-stacksize
7482 Only available with the @samp{cris-axis-aout} target. Arranges for
7483 indications in the program to the kernel loader that the stack of the
7484 program should be set to @var{n} bytes.
7490 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7491 @option{-march=v3} and @option{-march=v8} respectively.
7493 @item -mmul-bug-workaround
7494 @itemx -mno-mul-bug-workaround
7495 @opindex mmul-bug-workaround
7496 @opindex mno-mul-bug-workaround
7497 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7498 models where it applies. This option is active by default.
7502 Enable CRIS-specific verbose debug-related information in the assembly
7503 code. This option also has the effect to turn off the @samp{#NO_APP}
7504 formatted-code indicator to the assembler at the beginning of the
7509 Do not use condition-code results from previous instruction; always emit
7510 compare and test instructions before use of condition codes.
7512 @item -mno-side-effects
7513 @opindex mno-side-effects
7514 Do not emit instructions with side-effects in addressing modes other than
7518 @itemx -mno-stack-align
7520 @itemx -mno-data-align
7521 @itemx -mconst-align
7522 @itemx -mno-const-align
7523 @opindex mstack-align
7524 @opindex mno-stack-align
7525 @opindex mdata-align
7526 @opindex mno-data-align
7527 @opindex mconst-align
7528 @opindex mno-const-align
7529 These options (no-options) arranges (eliminate arrangements) for the
7530 stack-frame, individual data and constants to be aligned for the maximum
7531 single data access size for the chosen CPU model. The default is to
7532 arrange for 32-bit alignment. ABI details such as structure layout are
7533 not affected by these options.
7541 Similar to the stack- data- and const-align options above, these options
7542 arrange for stack-frame, writable data and constants to all be 32-bit,
7543 16-bit or 8-bit aligned. The default is 32-bit alignment.
7545 @item -mno-prologue-epilogue
7546 @itemx -mprologue-epilogue
7547 @opindex mno-prologue-epilogue
7548 @opindex mprologue-epilogue
7549 With @option{-mno-prologue-epilogue}, the normal function prologue and
7550 epilogue that sets up the stack-frame are omitted and no return
7551 instructions or return sequences are generated in the code. Use this
7552 option only together with visual inspection of the compiled code: no
7553 warnings or errors are generated when call-saved registers must be saved,
7554 or storage for local variable needs to be allocated.
7560 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7561 instruction sequences that load addresses for functions from the PLT part
7562 of the GOT rather than (traditional on other architectures) calls to the
7563 PLT@. The default is @option{-mgotplt}.
7567 Legacy no-op option only recognized with the cris-axis-aout target.
7571 Legacy no-op option only recognized with the cris-axis-elf and
7572 cris-axis-linux-gnu targets.
7576 Only recognized with the cris-axis-aout target, where it selects a
7577 GNU/linux-like multilib, include files and instruction set for
7582 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7586 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7587 to link with input-output functions from a simulator library. Code,
7588 initialized data and zero-initialized data are allocated consecutively.
7592 Like @option{-sim}, but pass linker options to locate initialized data at
7593 0x40000000 and zero-initialized data at 0x80000000.
7596 @node Darwin Options
7597 @subsection Darwin Options
7598 @cindex Darwin options
7600 These options are defined for all architectures running the Darwin operating
7603 FSF GCC on Darwin does not create ``fat'' object files; it will create
7604 an object file for the single architecture that it was built to
7605 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7606 @option{-arch} options are used; it does so by running the compiler or
7607 linker multiple times and joining the results together with
7610 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7611 @samp{i686}) is determined by the flags that specify the ISA
7612 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7613 @option{-force_cpusubtype_ALL} option can be used to override this.
7615 The Darwin tools vary in their behavior when presented with an ISA
7616 mismatch. The assembler, @file{as}, will only permit instructions to
7617 be used that are valid for the subtype of the file it is generating,
7618 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7619 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
7620 and print an error if asked to create a shared library with a less
7621 restrictive subtype than its input files (for instance, trying to put
7622 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
7623 for executables, @file{ld}, will quietly give the executable the most
7624 restrictive subtype of any of its input files.
7629 Add the framework directory @var{dir} to the head of the list of
7630 directories to be searched for header files. These directories are
7631 interleaved with those specified by @option{-I} options and are
7632 scanned in a left-to-right order.
7634 A framework directory is a directory with frameworks in it. A
7635 framework is a directory with a @samp{"Headers"} and/or
7636 @samp{"PrivateHeaders"} directory contained directly in it that ends
7637 in @samp{".framework"}. The name of a framework is the name of this
7638 directory excluding the @samp{".framework"}. Headers associated with
7639 the framework are found in one of those two directories, with
7640 @samp{"Headers"} being searched first. A subframework is a framework
7641 directory that is in a framework's @samp{"Frameworks"} directory.
7642 Includes of subframework headers can only appear in a header of a
7643 framework that contains the subframework, or in a sibling subframework
7644 header. Two subframeworks are siblings if they occur in the same
7645 framework. A subframework should not have the same name as a
7646 framework, a warning will be issued if this is violated. Currently a
7647 subframework cannot have subframeworks, in the future, the mechanism
7648 may be extended to support this. The standard frameworks can be found
7649 in @samp{"/System/Library/Frameworks"} and
7650 @samp{"/Library/Frameworks"}. An example include looks like
7651 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
7652 the name of the framework and header.h is found in the
7653 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
7657 Emit debugging information for symbols that are used. For STABS
7658 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
7659 This is by default ON@.
7663 Emit debugging information for all symbols and types.
7665 @item -mone-byte-bool
7666 @opindex -mone-byte-bool
7667 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7668 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7669 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7670 option has no effect on x86.
7672 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
7673 to generate code that is not binary compatible with code generated
7674 without that switch. Using this switch may require recompiling all
7675 other modules in a program, including system libraries. Use this
7676 switch to conform to a non-default data model.
7678 @item -mfix-and-continue
7679 @itemx -ffix-and-continue
7680 @itemx -findirect-data
7681 @opindex mfix-and-continue
7682 @opindex ffix-and-continue
7683 @opindex findirect-data
7684 Generate code suitable for fast turn around development. Needed to
7685 enable gdb to dynamically load @code{.o} files into already running
7686 programs. @option{-findirect-data} and @option{-ffix-and-continue}
7687 are provided for backwards compatibility.
7691 Loads all members of static archive libraries.
7692 See man ld(1) for more information.
7694 @item -arch_errors_fatal
7695 @opindex arch_errors_fatal
7696 Cause the errors having to do with files that have the wrong architecture
7700 @opindex bind_at_load
7701 Causes the output file to be marked such that the dynamic linker will
7702 bind all undefined references when the file is loaded or launched.
7706 Produce a Mach-o bundle format file.
7707 See man ld(1) for more information.
7709 @item -bundle_loader @var{executable}
7710 @opindex bundle_loader
7711 This option specifies the @var{executable} that will be loading the build
7712 output file being linked. See man ld(1) for more information.
7715 @opindex -dynamiclib
7716 When passed this option, GCC will produce a dynamic library instead of
7717 an executable when linking, using the Darwin @file{libtool} command.
7719 @item -force_cpusubtype_ALL
7720 @opindex -force_cpusubtype_ALL
7721 This causes GCC's output file to have the @var{ALL} subtype, instead of
7722 one controlled by the @option{-mcpu} or @option{-march} option.
7724 @item -allowable_client @var{client_name}
7726 @itemx -compatibility_version
7727 @itemx -current_version
7729 @itemx -dependency-file
7731 @itemx -dylinker_install_name
7733 @itemx -exported_symbols_list
7735 @itemx -flat_namespace
7736 @itemx -force_flat_namespace
7737 @itemx -headerpad_max_install_names
7740 @itemx -install_name
7741 @itemx -keep_private_externs
7742 @itemx -multi_module
7743 @itemx -multiply_defined
7744 @itemx -multiply_defined_unused
7746 @itemx -no_dead_strip_inits_and_terms
7747 @itemx -nofixprebinding
7750 @itemx -noseglinkedit
7751 @itemx -pagezero_size
7753 @itemx -prebind_all_twolevel_modules
7754 @itemx -private_bundle
7755 @itemx -read_only_relocs
7757 @itemx -sectobjectsymbols
7761 @itemx -sectobjectsymbols
7764 @itemx -segs_read_only_addr
7765 @itemx -segs_read_write_addr
7766 @itemx -seg_addr_table
7767 @itemx -seg_addr_table_filename
7770 @itemx -segs_read_only_addr
7771 @itemx -segs_read_write_addr
7772 @itemx -single_module
7775 @itemx -sub_umbrella
7776 @itemx -twolevel_namespace
7779 @itemx -unexported_symbols_list
7780 @itemx -weak_reference_mismatches
7783 @opindex allowable_client
7784 @opindex client_name
7785 @opindex compatibility_version
7786 @opindex current_version
7788 @opindex dependency-file
7790 @opindex dylinker_install_name
7792 @opindex exported_symbols_list
7794 @opindex flat_namespace
7795 @opindex force_flat_namespace
7796 @opindex headerpad_max_install_names
7799 @opindex install_name
7800 @opindex keep_private_externs
7801 @opindex multi_module
7802 @opindex multiply_defined
7803 @opindex multiply_defined_unused
7805 @opindex no_dead_strip_inits_and_terms
7806 @opindex nofixprebinding
7807 @opindex nomultidefs
7809 @opindex noseglinkedit
7810 @opindex pagezero_size
7812 @opindex prebind_all_twolevel_modules
7813 @opindex private_bundle
7814 @opindex read_only_relocs
7816 @opindex sectobjectsymbols
7820 @opindex sectobjectsymbols
7823 @opindex segs_read_only_addr
7824 @opindex segs_read_write_addr
7825 @opindex seg_addr_table
7826 @opindex seg_addr_table_filename
7827 @opindex seglinkedit
7829 @opindex segs_read_only_addr
7830 @opindex segs_read_write_addr
7831 @opindex single_module
7833 @opindex sub_library
7834 @opindex sub_umbrella
7835 @opindex twolevel_namespace
7838 @opindex unexported_symbols_list
7839 @opindex weak_reference_mismatches
7840 @opindex whatsloaded
7842 These options are passed to the Darwin linker. The Darwin linker man page
7843 describes them in detail.
7846 @node DEC Alpha Options
7847 @subsection DEC Alpha Options
7849 These @samp{-m} options are defined for the DEC Alpha implementations:
7852 @item -mno-soft-float
7854 @opindex mno-soft-float
7855 @opindex msoft-float
7856 Use (do not use) the hardware floating-point instructions for
7857 floating-point operations. When @option{-msoft-float} is specified,
7858 functions in @file{libgcc.a} will be used to perform floating-point
7859 operations. Unless they are replaced by routines that emulate the
7860 floating-point operations, or compiled in such a way as to call such
7861 emulations routines, these routines will issue floating-point
7862 operations. If you are compiling for an Alpha without floating-point
7863 operations, you must ensure that the library is built so as not to call
7866 Note that Alpha implementations without floating-point operations are
7867 required to have floating-point registers.
7872 @opindex mno-fp-regs
7873 Generate code that uses (does not use) the floating-point register set.
7874 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
7875 register set is not used, floating point operands are passed in integer
7876 registers as if they were integers and floating-point results are passed
7877 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
7878 so any function with a floating-point argument or return value called by code
7879 compiled with @option{-mno-fp-regs} must also be compiled with that
7882 A typical use of this option is building a kernel that does not use,
7883 and hence need not save and restore, any floating-point registers.
7887 The Alpha architecture implements floating-point hardware optimized for
7888 maximum performance. It is mostly compliant with the IEEE floating
7889 point standard. However, for full compliance, software assistance is
7890 required. This option generates code fully IEEE compliant code
7891 @emph{except} that the @var{inexact-flag} is not maintained (see below).
7892 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
7893 defined during compilation. The resulting code is less efficient but is
7894 able to correctly support denormalized numbers and exceptional IEEE
7895 values such as not-a-number and plus/minus infinity. Other Alpha
7896 compilers call this option @option{-ieee_with_no_inexact}.
7898 @item -mieee-with-inexact
7899 @opindex mieee-with-inexact
7900 This is like @option{-mieee} except the generated code also maintains
7901 the IEEE @var{inexact-flag}. Turning on this option causes the
7902 generated code to implement fully-compliant IEEE math. In addition to
7903 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
7904 macro. On some Alpha implementations the resulting code may execute
7905 significantly slower than the code generated by default. Since there is
7906 very little code that depends on the @var{inexact-flag}, you should
7907 normally not specify this option. Other Alpha compilers call this
7908 option @option{-ieee_with_inexact}.
7910 @item -mfp-trap-mode=@var{trap-mode}
7911 @opindex mfp-trap-mode
7912 This option controls what floating-point related traps are enabled.
7913 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
7914 The trap mode can be set to one of four values:
7918 This is the default (normal) setting. The only traps that are enabled
7919 are the ones that cannot be disabled in software (e.g., division by zero
7923 In addition to the traps enabled by @samp{n}, underflow traps are enabled
7927 Like @samp{su}, but the instructions are marked to be safe for software
7928 completion (see Alpha architecture manual for details).
7931 Like @samp{su}, but inexact traps are enabled as well.
7934 @item -mfp-rounding-mode=@var{rounding-mode}
7935 @opindex mfp-rounding-mode
7936 Selects the IEEE rounding mode. Other Alpha compilers call this option
7937 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
7942 Normal IEEE rounding mode. Floating point numbers are rounded towards
7943 the nearest machine number or towards the even machine number in case
7947 Round towards minus infinity.
7950 Chopped rounding mode. Floating point numbers are rounded towards zero.
7953 Dynamic rounding mode. A field in the floating point control register
7954 (@var{fpcr}, see Alpha architecture reference manual) controls the
7955 rounding mode in effect. The C library initializes this register for
7956 rounding towards plus infinity. Thus, unless your program modifies the
7957 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
7960 @item -mtrap-precision=@var{trap-precision}
7961 @opindex mtrap-precision
7962 In the Alpha architecture, floating point traps are imprecise. This
7963 means without software assistance it is impossible to recover from a
7964 floating trap and program execution normally needs to be terminated.
7965 GCC can generate code that can assist operating system trap handlers
7966 in determining the exact location that caused a floating point trap.
7967 Depending on the requirements of an application, different levels of
7968 precisions can be selected:
7972 Program precision. This option is the default and means a trap handler
7973 can only identify which program caused a floating point exception.
7976 Function precision. The trap handler can determine the function that
7977 caused a floating point exception.
7980 Instruction precision. The trap handler can determine the exact
7981 instruction that caused a floating point exception.
7984 Other Alpha compilers provide the equivalent options called
7985 @option{-scope_safe} and @option{-resumption_safe}.
7987 @item -mieee-conformant
7988 @opindex mieee-conformant
7989 This option marks the generated code as IEEE conformant. You must not
7990 use this option unless you also specify @option{-mtrap-precision=i} and either
7991 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
7992 is to emit the line @samp{.eflag 48} in the function prologue of the
7993 generated assembly file. Under DEC Unix, this has the effect that
7994 IEEE-conformant math library routines will be linked in.
7996 @item -mbuild-constants
7997 @opindex mbuild-constants
7998 Normally GCC examines a 32- or 64-bit integer constant to
7999 see if it can construct it from smaller constants in two or three
8000 instructions. If it cannot, it will output the constant as a literal and
8001 generate code to load it from the data segment at runtime.
8003 Use this option to require GCC to construct @emph{all} integer constants
8004 using code, even if it takes more instructions (the maximum is six).
8006 You would typically use this option to build a shared library dynamic
8007 loader. Itself a shared library, it must relocate itself in memory
8008 before it can find the variables and constants in its own data segment.
8014 Select whether to generate code to be assembled by the vendor-supplied
8015 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8033 Indicate whether GCC should generate code to use the optional BWX,
8034 CIX, FIX and MAX instruction sets. The default is to use the instruction
8035 sets supported by the CPU type specified via @option{-mcpu=} option or that
8036 of the CPU on which GCC was built if none was specified.
8041 @opindex mfloat-ieee
8042 Generate code that uses (does not use) VAX F and G floating point
8043 arithmetic instead of IEEE single and double precision.
8045 @item -mexplicit-relocs
8046 @itemx -mno-explicit-relocs
8047 @opindex mexplicit-relocs
8048 @opindex mno-explicit-relocs
8049 Older Alpha assemblers provided no way to generate symbol relocations
8050 except via assembler macros. Use of these macros does not allow
8051 optimal instruction scheduling. GNU binutils as of version 2.12
8052 supports a new syntax that allows the compiler to explicitly mark
8053 which relocations should apply to which instructions. This option
8054 is mostly useful for debugging, as GCC detects the capabilities of
8055 the assembler when it is built and sets the default accordingly.
8059 @opindex msmall-data
8060 @opindex mlarge-data
8061 When @option{-mexplicit-relocs} is in effect, static data is
8062 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8063 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8064 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8065 16-bit relocations off of the @code{$gp} register. This limits the
8066 size of the small data area to 64KB, but allows the variables to be
8067 directly accessed via a single instruction.
8069 The default is @option{-mlarge-data}. With this option the data area
8070 is limited to just below 2GB@. Programs that require more than 2GB of
8071 data must use @code{malloc} or @code{mmap} to allocate the data in the
8072 heap instead of in the program's data segment.
8074 When generating code for shared libraries, @option{-fpic} implies
8075 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8079 @opindex msmall-text
8080 @opindex mlarge-text
8081 When @option{-msmall-text} is used, the compiler assumes that the
8082 code of the entire program (or shared library) fits in 4MB, and is
8083 thus reachable with a branch instruction. When @option{-msmall-data}
8084 is used, the compiler can assume that all local symbols share the
8085 same @code{$gp} value, and thus reduce the number of instructions
8086 required for a function call from 4 to 1.
8088 The default is @option{-mlarge-text}.
8090 @item -mcpu=@var{cpu_type}
8092 Set the instruction set and instruction scheduling parameters for
8093 machine type @var{cpu_type}. You can specify either the @samp{EV}
8094 style name or the corresponding chip number. GCC supports scheduling
8095 parameters for the EV4, EV5 and EV6 family of processors and will
8096 choose the default values for the instruction set from the processor
8097 you specify. If you do not specify a processor type, GCC will default
8098 to the processor on which the compiler was built.
8100 Supported values for @var{cpu_type} are
8106 Schedules as an EV4 and has no instruction set extensions.
8110 Schedules as an EV5 and has no instruction set extensions.
8114 Schedules as an EV5 and supports the BWX extension.
8119 Schedules as an EV5 and supports the BWX and MAX extensions.
8123 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8127 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8130 @item -mtune=@var{cpu_type}
8132 Set only the instruction scheduling parameters for machine type
8133 @var{cpu_type}. The instruction set is not changed.
8135 @item -mmemory-latency=@var{time}
8136 @opindex mmemory-latency
8137 Sets the latency the scheduler should assume for typical memory
8138 references as seen by the application. This number is highly
8139 dependent on the memory access patterns used by the application
8140 and the size of the external cache on the machine.
8142 Valid options for @var{time} are
8146 A decimal number representing clock cycles.
8152 The compiler contains estimates of the number of clock cycles for
8153 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8154 (also called Dcache, Scache, and Bcache), as well as to main memory.
8155 Note that L3 is only valid for EV5.
8160 @node DEC Alpha/VMS Options
8161 @subsection DEC Alpha/VMS Options
8163 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8166 @item -mvms-return-codes
8167 @opindex mvms-return-codes
8168 Return VMS condition codes from main. The default is to return POSIX
8169 style condition (e.g.@ error) codes.
8173 @subsection FRV Options
8180 Only use the first 32 general purpose registers.
8185 Use all 64 general purpose registers.
8190 Use only the first 32 floating point registers.
8195 Use all 64 floating point registers
8198 @opindex mhard-float
8200 Use hardware instructions for floating point operations.
8203 @opindex msoft-float
8205 Use library routines for floating point operations.
8210 Dynamically allocate condition code registers.
8215 Do not try to dynamically allocate condition code registers, only
8216 use @code{icc0} and @code{fcc0}.
8221 Change ABI to use double word insns.
8226 Do not use double word instructions.
8231 Use floating point double instructions.
8236 Do not use floating point double instructions.
8241 Use media instructions.
8246 Do not use media instructions.
8251 Use multiply and add/subtract instructions.
8256 Do not use multiply and add/subtract instructions.
8261 Select the FDPIC ABI, that uses function descriptors to represent
8262 pointers to functions. Without any PIC/PIE-related options, it
8263 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8264 assumes GOT entries and small data are within a 12-bit range from the
8265 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8266 are computed with 32 bits.
8269 @opindex minline-plt
8271 Enable inlining of PLT entries in function calls to functions that are
8272 not known to bind locally. It has no effect without @option{-mfdpic}.
8273 It's enabled by default if optimizing for speed and compiling for
8274 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8275 optimization option such as @option{-O3} or above is present in the
8281 Assume a large TLS segment when generating thread-local code.
8286 Do not assume a large TLS segment when generating thread-local code.
8291 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8292 that is known to be in read-only sections. It's enabled by default,
8293 except for @option{-fpic} or @option{-fpie}: even though it may help
8294 make the global offset table smaller, it trades 1 instruction for 4.
8295 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8296 one of which may be shared by multiple symbols, and it avoids the need
8297 for a GOT entry for the referenced symbol, so it's more likely to be a
8298 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8300 @item -multilib-library-pic
8301 @opindex multilib-library-pic
8303 Link with the (library, not FD) pic libraries. It's implied by
8304 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8305 @option{-fpic} without @option{-mfdpic}. You should never have to use
8311 Follow the EABI requirement of always creating a frame pointer whenever
8312 a stack frame is allocated. This option is enabled by default and can
8313 be disabled with @option{-mno-linked-fp}.
8316 @opindex mlong-calls
8318 Use indirect addressing to call functions outside the current
8319 compilation unit. This allows the functions to be placed anywhere
8320 within the 32-bit address space.
8322 @item -malign-labels
8323 @opindex malign-labels
8325 Try to align labels to an 8-byte boundary by inserting nops into the
8326 previous packet. This option only has an effect when VLIW packing
8327 is enabled. It doesn't create new packets; it merely adds nops to
8331 @opindex mlibrary-pic
8333 Generate position-independent EABI code.
8338 Use only the first four media accumulator registers.
8343 Use all eight media accumulator registers.
8348 Pack VLIW instructions.
8353 Do not pack VLIW instructions.
8358 Do not mark ABI switches in e_flags.
8363 Enable the use of conditional-move instructions (default).
8365 This switch is mainly for debugging the compiler and will likely be removed
8366 in a future version.
8368 @item -mno-cond-move
8369 @opindex mno-cond-move
8371 Disable the use of conditional-move instructions.
8373 This switch is mainly for debugging the compiler and will likely be removed
8374 in a future version.
8379 Enable the use of conditional set instructions (default).
8381 This switch is mainly for debugging the compiler and will likely be removed
8382 in a future version.
8387 Disable the use of conditional set instructions.
8389 This switch is mainly for debugging the compiler and will likely be removed
8390 in a future version.
8395 Enable the use of conditional execution (default).
8397 This switch is mainly for debugging the compiler and will likely be removed
8398 in a future version.
8400 @item -mno-cond-exec
8401 @opindex mno-cond-exec
8403 Disable the use of conditional execution.
8405 This switch is mainly for debugging the compiler and will likely be removed
8406 in a future version.
8409 @opindex mvliw-branch
8411 Run a pass to pack branches into VLIW instructions (default).
8413 This switch is mainly for debugging the compiler and will likely be removed
8414 in a future version.
8416 @item -mno-vliw-branch
8417 @opindex mno-vliw-branch
8419 Do not run a pass to pack branches into VLIW instructions.
8421 This switch is mainly for debugging the compiler and will likely be removed
8422 in a future version.
8424 @item -mmulti-cond-exec
8425 @opindex mmulti-cond-exec
8427 Enable optimization of @code{&&} and @code{||} in conditional execution
8430 This switch is mainly for debugging the compiler and will likely be removed
8431 in a future version.
8433 @item -mno-multi-cond-exec
8434 @opindex mno-multi-cond-exec
8436 Disable optimization of @code{&&} and @code{||} in conditional execution.
8438 This switch is mainly for debugging the compiler and will likely be removed
8439 in a future version.
8441 @item -mnested-cond-exec
8442 @opindex mnested-cond-exec
8444 Enable nested conditional execution optimizations (default).
8446 This switch is mainly for debugging the compiler and will likely be removed
8447 in a future version.
8449 @item -mno-nested-cond-exec
8450 @opindex mno-nested-cond-exec
8452 Disable nested conditional execution optimizations.
8454 This switch is mainly for debugging the compiler and will likely be removed
8455 in a future version.
8457 @item -mtomcat-stats
8458 @opindex mtomcat-stats
8460 Cause gas to print out tomcat statistics.
8462 @item -mcpu=@var{cpu}
8465 Select the processor type for which to generate code. Possible values are
8466 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8467 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8471 @node H8/300 Options
8472 @subsection H8/300 Options
8474 These @samp{-m} options are defined for the H8/300 implementations:
8479 Shorten some address references at link time, when possible; uses the
8480 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8481 ld, Using ld}, for a fuller description.
8485 Generate code for the H8/300H@.
8489 Generate code for the H8S@.
8493 Generate code for the H8S and H8/300H in the normal mode. This switch
8494 must be used either with @option{-mh} or @option{-ms}.
8498 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8502 Make @code{int} data 32 bits by default.
8506 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8507 The default for the H8/300H and H8S is to align longs and floats on 4
8509 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8510 This option has no effect on the H8/300.
8514 @subsection HPPA Options
8515 @cindex HPPA Options
8517 These @samp{-m} options are defined for the HPPA family of computers:
8520 @item -march=@var{architecture-type}
8522 Generate code for the specified architecture. The choices for
8523 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8524 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8525 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8526 architecture option for your machine. Code compiled for lower numbered
8527 architectures will run on higher numbered architectures, but not the
8531 @itemx -mpa-risc-1-1
8532 @itemx -mpa-risc-2-0
8533 @opindex mpa-risc-1-0
8534 @opindex mpa-risc-1-1
8535 @opindex mpa-risc-2-0
8536 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8539 @opindex mbig-switch
8540 Generate code suitable for big switch tables. Use this option only if
8541 the assembler/linker complain about out of range branches within a switch
8544 @item -mjump-in-delay
8545 @opindex mjump-in-delay
8546 Fill delay slots of function calls with unconditional jump instructions
8547 by modifying the return pointer for the function call to be the target
8548 of the conditional jump.
8550 @item -mdisable-fpregs
8551 @opindex mdisable-fpregs
8552 Prevent floating point registers from being used in any manner. This is
8553 necessary for compiling kernels which perform lazy context switching of
8554 floating point registers. If you use this option and attempt to perform
8555 floating point operations, the compiler will abort.
8557 @item -mdisable-indexing
8558 @opindex mdisable-indexing
8559 Prevent the compiler from using indexing address modes. This avoids some
8560 rather obscure problems when compiling MIG generated code under MACH@.
8562 @item -mno-space-regs
8563 @opindex mno-space-regs
8564 Generate code that assumes the target has no space registers. This allows
8565 GCC to generate faster indirect calls and use unscaled index address modes.
8567 Such code is suitable for level 0 PA systems and kernels.
8569 @item -mfast-indirect-calls
8570 @opindex mfast-indirect-calls
8571 Generate code that assumes calls never cross space boundaries. This
8572 allows GCC to emit code which performs faster indirect calls.
8574 This option will not work in the presence of shared libraries or nested
8577 @item -mfixed-range=@var{register-range}
8578 @opindex mfixed-range
8579 Generate code treating the given register range as fixed registers.
8580 A fixed register is one that the register allocator can not use. This is
8581 useful when compiling kernel code. A register range is specified as
8582 two registers separated by a dash. Multiple register ranges can be
8583 specified separated by a comma.
8585 @item -mlong-load-store
8586 @opindex mlong-load-store
8587 Generate 3-instruction load and store sequences as sometimes required by
8588 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8591 @item -mportable-runtime
8592 @opindex mportable-runtime
8593 Use the portable calling conventions proposed by HP for ELF systems.
8597 Enable the use of assembler directives only GAS understands.
8599 @item -mschedule=@var{cpu-type}
8601 Schedule code according to the constraints for the machine type
8602 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8603 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8604 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8605 proper scheduling option for your machine. The default scheduling is
8609 @opindex mlinker-opt
8610 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8611 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8612 linkers in which they give bogus error messages when linking some programs.
8615 @opindex msoft-float
8616 Generate output containing library calls for floating point.
8617 @strong{Warning:} the requisite libraries are not available for all HPPA
8618 targets. Normally the facilities of the machine's usual C compiler are
8619 used, but this cannot be done directly in cross-compilation. You must make
8620 your own arrangements to provide suitable library functions for
8621 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8622 does provide software floating point support.
8624 @option{-msoft-float} changes the calling convention in the output file;
8625 therefore, it is only useful if you compile @emph{all} of a program with
8626 this option. In particular, you need to compile @file{libgcc.a}, the
8627 library that comes with GCC, with @option{-msoft-float} in order for
8632 Generate the predefine, @code{_SIO}, for server IO@. The default is
8633 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8634 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
8635 options are available under HP-UX and HI-UX@.
8639 Use GNU ld specific options. This passes @option{-shared} to ld when
8640 building a shared library. It is the default when GCC is configured,
8641 explicitly or implicitly, with the GNU linker. This option does not
8642 have any affect on which ld is called, it only changes what parameters
8643 are passed to that ld. The ld that is called is determined by the
8644 @option{--with-ld} configure option, GCC's program search path, and
8645 finally by the user's @env{PATH}. The linker used by GCC can be printed
8646 using @samp{which `gcc -print-prog-name=ld`}.
8650 Use HP ld specific options. This passes @option{-b} to ld when building
8651 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8652 links. It is the default when GCC is configured, explicitly or
8653 implicitly, with the HP linker. This option does not have any affect on
8654 which ld is called, it only changes what parameters are passed to that
8655 ld. The ld that is called is determined by the @option{--with-ld}
8656 configure option, GCC's program search path, and finally by the user's
8657 @env{PATH}. The linker used by GCC can be printed using @samp{which
8658 `gcc -print-prog-name=ld`}.
8661 @opindex mno-long-calls
8662 Generate code that uses long call sequences. This ensures that a call
8663 is always able to reach linker generated stubs. The default is to generate
8664 long calls only when the distance from the call site to the beginning
8665 of the function or translation unit, as the case may be, exceeds a
8666 predefined limit set by the branch type being used. The limits for
8667 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8668 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8671 Distances are measured from the beginning of functions when using the
8672 @option{-ffunction-sections} option, or when using the @option{-mgas}
8673 and @option{-mno-portable-runtime} options together under HP-UX with
8676 It is normally not desirable to use this option as it will degrade
8677 performance. However, it may be useful in large applications,
8678 particularly when partial linking is used to build the application.
8680 The types of long calls used depends on the capabilities of the
8681 assembler and linker, and the type of code being generated. The
8682 impact on systems that support long absolute calls, and long pic
8683 symbol-difference or pc-relative calls should be relatively small.
8684 However, an indirect call is used on 32-bit ELF systems in pic code
8685 and it is quite long.
8687 @item -munix=@var{unix-std}
8689 Generate compiler predefines and select a startfile for the specified
8690 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
8691 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
8692 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
8693 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
8694 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
8697 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
8698 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
8699 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
8700 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
8701 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
8702 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
8704 It is @emph{important} to note that this option changes the interfaces
8705 for various library routines. It also affects the operational behavior
8706 of the C library. Thus, @emph{extreme} care is needed in using this
8709 Library code that is intended to operate with more than one UNIX
8710 standard must test, set and restore the variable @var{__xpg4_extended_mask}
8711 as appropriate. Most GNU software doesn't provide this capability.
8715 Suppress the generation of link options to search libdld.sl when the
8716 @option{-static} option is specified on HP-UX 10 and later.
8720 The HP-UX implementation of setlocale in libc has a dependency on
8721 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8722 when the @option{-static} option is specified, special link options
8723 are needed to resolve this dependency.
8725 On HP-UX 10 and later, the GCC driver adds the necessary options to
8726 link with libdld.sl when the @option{-static} option is specified.
8727 This causes the resulting binary to be dynamic. On the 64-bit port,
8728 the linkers generate dynamic binaries by default in any case. The
8729 @option{-nolibdld} option can be used to prevent the GCC driver from
8730 adding these link options.
8734 Add support for multithreading with the @dfn{dce thread} library
8735 under HP-UX@. This option sets flags for both the preprocessor and
8739 @node i386 and x86-64 Options
8740 @subsection Intel 386 and AMD x86-64 Options
8741 @cindex i386 Options
8742 @cindex x86-64 Options
8743 @cindex Intel 386 Options
8744 @cindex AMD x86-64 Options
8746 These @samp{-m} options are defined for the i386 and x86-64 family of
8750 @item -mtune=@var{cpu-type}
8752 Tune to @var{cpu-type} everything applicable about the generated code, except
8753 for the ABI and the set of available instructions. The choices for
8757 Original Intel's i386 CPU@.
8759 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
8761 Intel Pentium CPU with no MMX support.
8763 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
8764 @item i686, pentiumpro
8765 Intel PentiumPro CPU@.
8767 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
8768 @item pentium3, pentium3m
8769 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
8772 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
8773 support. Used by Centrino notebooks.
8774 @item pentium4, pentium4m
8775 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
8777 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
8780 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
8781 SSE2 and SSE3 instruction set support.
8783 AMD K6 CPU with MMX instruction set support.
8785 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
8786 @item athlon, athlon-tbird
8787 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
8789 @item athlon-4, athlon-xp, athlon-mp
8790 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
8791 instruction set support.
8792 @item k8, opteron, athlon64, athlon-fx
8793 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
8794 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
8796 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
8799 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
8800 instruction set support.
8802 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
8803 implemented for this chip.)
8805 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
8806 implemented for this chip.)
8809 While picking a specific @var{cpu-type} will schedule things appropriately
8810 for that particular chip, the compiler will not generate any code that
8811 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8814 @item -march=@var{cpu-type}
8816 Generate instructions for the machine type @var{cpu-type}. The choices
8817 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8818 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8820 @item -mcpu=@var{cpu-type}
8822 A deprecated synonym for @option{-mtune}.
8831 @opindex mpentiumpro
8832 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8833 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8834 These synonyms are deprecated.
8836 @item -mfpmath=@var{unit}
8838 Generate floating point arithmetics for selected unit @var{unit}. The choices
8843 Use the standard 387 floating point coprocessor present majority of chips and
8844 emulated otherwise. Code compiled with this option will run almost everywhere.
8845 The temporary results are computed in 80bit precision instead of precision
8846 specified by the type resulting in slightly different results compared to most
8847 of other chips. See @option{-ffloat-store} for more detailed description.
8849 This is the default choice for i386 compiler.
8852 Use scalar floating point instructions present in the SSE instruction set.
8853 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8854 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8855 instruction set supports only single precision arithmetics, thus the double and
8856 extended precision arithmetics is still done using 387. Later version, present
8857 only in Pentium4 and the future AMD x86-64 chips supports double precision
8860 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
8861 or @option{-msse2} switches to enable SSE extensions and make this option
8862 effective. For the x86-64 compiler, these extensions are enabled by default.
8864 The resulting code should be considerably faster in the majority of cases and avoid
8865 the numerical instability problems of 387 code, but may break some existing
8866 code that expects temporaries to be 80bit.
8868 This is the default choice for the x86-64 compiler.
8871 Attempt to utilize both instruction sets at once. This effectively double the
8872 amount of available registers and on chips with separate execution units for
8873 387 and SSE the execution resources too. Use this option with care, as it is
8874 still experimental, because the GCC register allocator does not model separate
8875 functional units well resulting in instable performance.
8878 @item -masm=@var{dialect}
8879 @opindex masm=@var{dialect}
8880 Output asm instructions using selected @var{dialect}. Supported choices are
8881 @samp{intel} or @samp{att} (the default one).
8886 @opindex mno-ieee-fp
8887 Control whether or not the compiler uses IEEE floating point
8888 comparisons. These handle correctly the case where the result of a
8889 comparison is unordered.
8892 @opindex msoft-float
8893 Generate output containing library calls for floating point.
8894 @strong{Warning:} the requisite libraries are not part of GCC@.
8895 Normally the facilities of the machine's usual C compiler are used, but
8896 this can't be done directly in cross-compilation. You must make your
8897 own arrangements to provide suitable library functions for
8900 On machines where a function returns floating point results in the 80387
8901 register stack, some floating point opcodes may be emitted even if
8902 @option{-msoft-float} is used.
8904 @item -mno-fp-ret-in-387
8905 @opindex mno-fp-ret-in-387
8906 Do not use the FPU registers for return values of functions.
8908 The usual calling convention has functions return values of types
8909 @code{float} and @code{double} in an FPU register, even if there
8910 is no FPU@. The idea is that the operating system should emulate
8913 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8914 in ordinary CPU registers instead.
8916 @item -mno-fancy-math-387
8917 @opindex mno-fancy-math-387
8918 Some 387 emulators do not support the @code{sin}, @code{cos} and
8919 @code{sqrt} instructions for the 387. Specify this option to avoid
8920 generating those instructions. This option is the default on FreeBSD,
8921 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8922 indicates that the target cpu will always have an FPU and so the
8923 instruction will not need emulation. As of revision 2.6.1, these
8924 instructions are not generated unless you also use the
8925 @option{-funsafe-math-optimizations} switch.
8927 @item -malign-double
8928 @itemx -mno-align-double
8929 @opindex malign-double
8930 @opindex mno-align-double
8931 Control whether GCC aligns @code{double}, @code{long double}, and
8932 @code{long long} variables on a two word boundary or a one word
8933 boundary. Aligning @code{double} variables on a two word boundary will
8934 produce code that runs somewhat faster on a @samp{Pentium} at the
8935 expense of more memory.
8937 @strong{Warning:} if you use the @option{-malign-double} switch,
8938 structures containing the above types will be aligned differently than
8939 the published application binary interface specifications for the 386
8940 and will not be binary compatible with structures in code compiled
8941 without that switch.
8943 @item -m96bit-long-double
8944 @itemx -m128bit-long-double
8945 @opindex m96bit-long-double
8946 @opindex m128bit-long-double
8947 These switches control the size of @code{long double} type. The i386
8948 application binary interface specifies the size to be 96 bits,
8949 so @option{-m96bit-long-double} is the default in 32 bit mode.
8951 Modern architectures (Pentium and newer) would prefer @code{long double}
8952 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8953 conforming to the ABI, this would not be possible. So specifying a
8954 @option{-m128bit-long-double} will align @code{long double}
8955 to a 16 byte boundary by padding the @code{long double} with an additional
8958 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8959 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8961 Notice that neither of these options enable any extra precision over the x87
8962 standard of 80 bits for a @code{long double}.
8964 @strong{Warning:} if you override the default value for your target ABI, the
8965 structures and arrays containing @code{long double} variables will change
8966 their size as well as function calling convention for function taking
8967 @code{long double} will be modified. Hence they will not be binary
8968 compatible with arrays or structures in code compiled without that switch.
8972 @itemx -mno-svr3-shlib
8973 @opindex msvr3-shlib
8974 @opindex mno-svr3-shlib
8975 Control whether GCC places uninitialized local variables into the
8976 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8977 into @code{bss}. These options are meaningful only on System V Release 3.
8981 Use a different function-calling convention, in which functions that
8982 take a fixed number of arguments return with the @code{ret} @var{num}
8983 instruction, which pops their arguments while returning. This saves one
8984 instruction in the caller since there is no need to pop the arguments
8987 You can specify that an individual function is called with this calling
8988 sequence with the function attribute @samp{stdcall}. You can also
8989 override the @option{-mrtd} option by using the function attribute
8990 @samp{cdecl}. @xref{Function Attributes}.
8992 @strong{Warning:} this calling convention is incompatible with the one
8993 normally used on Unix, so you cannot use it if you need to call
8994 libraries compiled with the Unix compiler.
8996 Also, you must provide function prototypes for all functions that
8997 take variable numbers of arguments (including @code{printf});
8998 otherwise incorrect code will be generated for calls to those
9001 In addition, seriously incorrect code will result if you call a
9002 function with too many arguments. (Normally, extra arguments are
9003 harmlessly ignored.)
9005 @item -mregparm=@var{num}
9007 Control how many registers are used to pass integer arguments. By
9008 default, no registers are used to pass arguments, and at most 3
9009 registers can be used. You can control this behavior for a specific
9010 function by using the function attribute @samp{regparm}.
9011 @xref{Function Attributes}.
9013 @strong{Warning:} if you use this switch, and
9014 @var{num} is nonzero, then you must build all modules with the same
9015 value, including any libraries. This includes the system libraries and
9018 @item -mpreferred-stack-boundary=@var{num}
9019 @opindex mpreferred-stack-boundary
9020 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
9021 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
9022 the default is 4 (16 bytes or 128 bits), except when optimizing for code
9023 size (@option{-Os}), in which case the default is the minimum correct
9024 alignment (4 bytes for x86, and 8 bytes for x86-64).
9026 On Pentium and PentiumPro, @code{double} and @code{long double} values
9027 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
9028 suffer significant run time performance penalties. On Pentium III, the
9029 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
9030 penalties if it is not 16 byte aligned.
9032 To ensure proper alignment of this values on the stack, the stack boundary
9033 must be as aligned as that required by any value stored on the stack.
9034 Further, every function must be generated such that it keeps the stack
9035 aligned. Thus calling a function compiled with a higher preferred
9036 stack boundary from a function compiled with a lower preferred stack
9037 boundary will most likely misalign the stack. It is recommended that
9038 libraries that use callbacks always use the default setting.
9040 This extra alignment does consume extra stack space, and generally
9041 increases code size. Code that is sensitive to stack space usage, such
9042 as embedded systems and operating system kernels, may want to reduce the
9043 preferred alignment to @option{-mpreferred-stack-boundary=2}.
9061 These switches enable or disable the use of built-in functions that allow
9062 direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow extensions of the
9065 @xref{X86 Built-in Functions}, for details of the functions enabled
9066 and disabled by these switches.
9068 To have SSE/SSE2 instructions generated automatically from floating-point
9069 code, see @option{-mfpmath=sse}.
9072 @itemx -mno-push-args
9074 @opindex mno-push-args
9075 Use PUSH operations to store outgoing parameters. This method is shorter
9076 and usually equally fast as method using SUB/MOV operations and is enabled
9077 by default. In some cases disabling it may improve performance because of
9078 improved scheduling and reduced dependencies.
9080 @item -maccumulate-outgoing-args
9081 @opindex maccumulate-outgoing-args
9082 If enabled, the maximum amount of space required for outgoing arguments will be
9083 computed in the function prologue. This is faster on most modern CPUs
9084 because of reduced dependencies, improved scheduling and reduced stack usage
9085 when preferred stack boundary is not equal to 2. The drawback is a notable
9086 increase in code size. This switch implies @option{-mno-push-args}.
9090 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
9091 on thread-safe exception handling must compile and link all code with the
9092 @option{-mthreads} option. When compiling, @option{-mthreads} defines
9093 @option{-D_MT}; when linking, it links in a special thread helper library
9094 @option{-lmingwthrd} which cleans up per thread exception handling data.
9096 @item -mno-align-stringops
9097 @opindex mno-align-stringops
9098 Do not align destination of inlined string operations. This switch reduces
9099 code size and improves performance in case the destination is already aligned,
9100 but GCC doesn't know about it.
9102 @item -minline-all-stringops
9103 @opindex minline-all-stringops
9104 By default GCC inlines string operations only when destination is known to be
9105 aligned at least to 4 byte boundary. This enables more inlining, increase code
9106 size, but may improve performance of code that depends on fast memcpy, strlen
9107 and memset for short lengths.
9109 @item -momit-leaf-frame-pointer
9110 @opindex momit-leaf-frame-pointer
9111 Don't keep the frame pointer in a register for leaf functions. This
9112 avoids the instructions to save, set up and restore frame pointers and
9113 makes an extra register available in leaf functions. The option
9114 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9115 which might make debugging harder.
9117 @item -mtls-direct-seg-refs
9118 @itemx -mno-tls-direct-seg-refs
9119 @opindex mtls-direct-seg-refs
9120 Controls whether TLS variables may be accessed with offsets from the
9121 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
9122 or whether the thread base pointer must be added. Whether or not this
9123 is legal depends on the operating system, and whether it maps the
9124 segment to cover the entire TLS area.
9126 For systems that use GNU libc, the default is on.
9129 These @samp{-m} switches are supported in addition to the above
9130 on AMD x86-64 processors in 64-bit environments.
9137 Generate code for a 32-bit or 64-bit environment.
9138 The 32-bit environment sets int, long and pointer to 32 bits and
9139 generates code that runs on any i386 system.
9140 The 64-bit environment sets int to 32 bits and long and pointer
9141 to 64 bits and generates code for AMD's x86-64 architecture.
9144 @opindex no-red-zone
9145 Do not use a so called red zone for x86-64 code. The red zone is mandated
9146 by the x86-64 ABI, it is a 128-byte area beyond the location of the
9147 stack pointer that will not be modified by signal or interrupt handlers
9148 and therefore can be used for temporary data without adjusting the stack
9149 pointer. The flag @option{-mno-red-zone} disables this red zone.
9151 @item -mcmodel=small
9152 @opindex mcmodel=small
9153 Generate code for the small code model: the program and its symbols must
9154 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
9155 Programs can be statically or dynamically linked. This is the default
9158 @item -mcmodel=kernel
9159 @opindex mcmodel=kernel
9160 Generate code for the kernel code model. The kernel runs in the
9161 negative 2 GB of the address space.
9162 This model has to be used for Linux kernel code.
9164 @item -mcmodel=medium
9165 @opindex mcmodel=medium
9166 Generate code for the medium model: The program is linked in the lower 2
9167 GB of the address space but symbols can be located anywhere in the
9168 address space. Programs can be statically or dynamically linked, but
9169 building of shared libraries are not supported with the medium model.
9171 @item -mcmodel=large
9172 @opindex mcmodel=large
9173 Generate code for the large model: This model makes no assumptions
9174 about addresses and sizes of sections. Currently GCC does not implement
9179 @subsection IA-64 Options
9180 @cindex IA-64 Options
9182 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9186 @opindex mbig-endian
9187 Generate code for a big endian target. This is the default for HP-UX@.
9189 @item -mlittle-endian
9190 @opindex mlittle-endian
9191 Generate code for a little endian target. This is the default for AIX5
9198 Generate (or don't) code for the GNU assembler. This is the default.
9199 @c Also, this is the default if the configure option @option{--with-gnu-as}
9206 Generate (or don't) code for the GNU linker. This is the default.
9207 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9212 Generate code that does not use a global pointer register. The result
9213 is not position independent code, and violates the IA-64 ABI@.
9215 @item -mvolatile-asm-stop
9216 @itemx -mno-volatile-asm-stop
9217 @opindex mvolatile-asm-stop
9218 @opindex mno-volatile-asm-stop
9219 Generate (or don't) a stop bit immediately before and after volatile asm
9222 @item -mregister-names
9223 @itemx -mno-register-names
9224 @opindex mregister-names
9225 @opindex mno-register-names
9226 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9227 the stacked registers. This may make assembler output more readable.
9233 Disable (or enable) optimizations that use the small data section. This may
9234 be useful for working around optimizer bugs.
9237 @opindex mconstant-gp
9238 Generate code that uses a single constant global pointer value. This is
9239 useful when compiling kernel code.
9243 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9244 This is useful when compiling firmware code.
9246 @item -minline-float-divide-min-latency
9247 @opindex minline-float-divide-min-latency
9248 Generate code for inline divides of floating point values
9249 using the minimum latency algorithm.
9251 @item -minline-float-divide-max-throughput
9252 @opindex minline-float-divide-max-throughput
9253 Generate code for inline divides of floating point values
9254 using the maximum throughput algorithm.
9256 @item -minline-int-divide-min-latency
9257 @opindex minline-int-divide-min-latency
9258 Generate code for inline divides of integer values
9259 using the minimum latency algorithm.
9261 @item -minline-int-divide-max-throughput
9262 @opindex minline-int-divide-max-throughput
9263 Generate code for inline divides of integer values
9264 using the maximum throughput algorithm.
9266 @item -minline-sqrt-min-latency
9267 @opindex minline-sqrt-min-latency
9268 Generate code for inline square roots
9269 using the minimum latency algorithm.
9271 @item -minline-sqrt-max-throughput
9272 @opindex minline-sqrt-max-throughput
9273 Generate code for inline square roots
9274 using the maximum throughput algorithm.
9276 @item -mno-dwarf2-asm
9278 @opindex mno-dwarf2-asm
9279 @opindex mdwarf2-asm
9280 Don't (or do) generate assembler code for the DWARF2 line number debugging
9281 info. This may be useful when not using the GNU assembler.
9283 @item -mearly-stop-bits
9284 @itemx -mno-early-stop-bits
9285 @opindex mearly-stop-bits
9286 @opindex mno-early-stop-bits
9287 Allow stop bits to be placed earlier than immediately preceding the
9288 instruction that triggered the stop bit. This can improve instruction
9289 scheduling, but does not always do so.
9291 @item -mfixed-range=@var{register-range}
9292 @opindex mfixed-range
9293 Generate code treating the given register range as fixed registers.
9294 A fixed register is one that the register allocator can not use. This is
9295 useful when compiling kernel code. A register range is specified as
9296 two registers separated by a dash. Multiple register ranges can be
9297 specified separated by a comma.
9299 @item -mtls-size=@var{tls-size}
9301 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9304 @item -mtune-arch=@var{cpu-type}
9306 Tune the instruction scheduling for a particular CPU, Valid values are
9307 itanium, itanium1, merced, itanium2, and mckinley.
9313 Add support for multithreading using the POSIX threads library. This
9314 option sets flags for both the preprocessor and linker. It does
9315 not affect the thread safety of object code produced by the compiler or
9316 that of libraries supplied with it. These are HP-UX specific flags.
9322 Generate code for a 32-bit or 64-bit environment.
9323 The 32-bit environment sets int, long and pointer to 32 bits.
9324 The 64-bit environment sets int to 32 bits and long and pointer
9325 to 64 bits. These are HP-UX specific flags.
9329 @node M32R/D Options
9330 @subsection M32R/D Options
9331 @cindex M32R/D options
9333 These @option{-m} options are defined for Renesas M32R/D architectures:
9338 Generate code for the M32R/2@.
9342 Generate code for the M32R/X@.
9346 Generate code for the M32R@. This is the default.
9349 @opindex mmodel=small
9350 Assume all objects live in the lower 16MB of memory (so that their addresses
9351 can be loaded with the @code{ld24} instruction), and assume all subroutines
9352 are reachable with the @code{bl} instruction.
9353 This is the default.
9355 The addressability of a particular object can be set with the
9356 @code{model} attribute.
9358 @item -mmodel=medium
9359 @opindex mmodel=medium
9360 Assume objects may be anywhere in the 32-bit address space (the compiler
9361 will generate @code{seth/add3} instructions to load their addresses), and
9362 assume all subroutines are reachable with the @code{bl} instruction.
9365 @opindex mmodel=large
9366 Assume objects may be anywhere in the 32-bit address space (the compiler
9367 will generate @code{seth/add3} instructions to load their addresses), and
9368 assume subroutines may not be reachable with the @code{bl} instruction
9369 (the compiler will generate the much slower @code{seth/add3/jl}
9370 instruction sequence).
9373 @opindex msdata=none
9374 Disable use of the small data area. Variables will be put into
9375 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
9376 @code{section} attribute has been specified).
9377 This is the default.
9379 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
9380 Objects may be explicitly put in the small data area with the
9381 @code{section} attribute using one of these sections.
9384 @opindex msdata=sdata
9385 Put small global and static data in the small data area, but do not
9386 generate special code to reference them.
9390 Put small global and static data in the small data area, and generate
9391 special instructions to reference them.
9395 @cindex smaller data references
9396 Put global and static objects less than or equal to @var{num} bytes
9397 into the small data or bss sections instead of the normal data or bss
9398 sections. The default value of @var{num} is 8.
9399 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
9400 for this option to have any effect.
9402 All modules should be compiled with the same @option{-G @var{num}} value.
9403 Compiling with different values of @var{num} may or may not work; if it
9404 doesn't the linker will give an error message---incorrect code will not be
9409 Makes the M32R specific code in the compiler display some statistics
9410 that might help in debugging programs.
9413 @opindex malign-loops
9414 Align all loops to a 32-byte boundary.
9416 @item -mno-align-loops
9417 @opindex mno-align-loops
9418 Do not enforce a 32-byte alignment for loops. This is the default.
9420 @item -missue-rate=@var{number}
9421 @opindex missue-rate=@var{number}
9422 Issue @var{number} instructions per cycle. @var{number} can only be 1
9425 @item -mbranch-cost=@var{number}
9426 @opindex mbranch-cost=@var{number}
9427 @var{number} can only be 1 or 2. If it is 1 then branches will be
9428 preferred over conditional code, if it is 2, then the opposite will
9431 @item -mflush-trap=@var{number}
9432 @opindex mflush-trap=@var{number}
9433 Specifies the trap number to use to flush the cache. The default is
9434 12. Valid numbers are between 0 and 15 inclusive.
9436 @item -mno-flush-trap
9437 @opindex mno-flush-trap
9438 Specifies that the cache cannot be flushed by using a trap.
9440 @item -mflush-func=@var{name}
9441 @opindex mflush-func=@var{name}
9442 Specifies the name of the operating system function to call to flush
9443 the cache. The default is @emph{_flush_cache}, but a function call
9444 will only be used if a trap is not available.
9446 @item -mno-flush-func
9447 @opindex mno-flush-func
9448 Indicates that there is no OS function for flushing the cache.
9452 @node M680x0 Options
9453 @subsection M680x0 Options
9454 @cindex M680x0 options
9456 These are the @samp{-m} options defined for the 68000 series. The default
9457 values for these options depends on which style of 68000 was selected when
9458 the compiler was configured; the defaults for the most common choices are
9466 Generate output for a 68000. This is the default
9467 when the compiler is configured for 68000-based systems.
9469 Use this option for microcontrollers with a 68000 or EC000 core,
9470 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
9476 Generate output for a 68020. This is the default
9477 when the compiler is configured for 68020-based systems.
9481 Generate output containing 68881 instructions for floating point.
9482 This is the default for most 68020 systems unless @option{--nfp} was
9483 specified when the compiler was configured.
9487 Generate output for a 68030. This is the default when the compiler is
9488 configured for 68030-based systems.
9492 Generate output for a 68040. This is the default when the compiler is
9493 configured for 68040-based systems.
9495 This option inhibits the use of 68881/68882 instructions that have to be
9496 emulated by software on the 68040. Use this option if your 68040 does not
9497 have code to emulate those instructions.
9501 Generate output for a 68060. This is the default when the compiler is
9502 configured for 68060-based systems.
9504 This option inhibits the use of 68020 and 68881/68882 instructions that
9505 have to be emulated by software on the 68060. Use this option if your 68060
9506 does not have code to emulate those instructions.
9510 Generate output for a CPU32. This is the default
9511 when the compiler is configured for CPU32-based systems.
9513 Use this option for microcontrollers with a
9514 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
9515 68336, 68340, 68341, 68349 and 68360.
9519 Generate output for a 520X ``coldfire'' family cpu. This is the default
9520 when the compiler is configured for 520X-based systems.
9522 Use this option for microcontroller with a 5200 core, including
9523 the MCF5202, MCF5203, MCF5204 and MCF5202.
9528 Generate output for a 68040, without using any of the new instructions.
9529 This results in code which can run relatively efficiently on either a
9530 68020/68881 or a 68030 or a 68040. The generated code does use the
9531 68881 instructions that are emulated on the 68040.
9535 Generate output for a 68060, without using any of the new instructions.
9536 This results in code which can run relatively efficiently on either a
9537 68020/68881 or a 68030 or a 68040. The generated code does use the
9538 68881 instructions that are emulated on the 68060.
9541 @opindex msoft-float
9542 Generate output containing library calls for floating point.
9543 @strong{Warning:} the requisite libraries are not available for all m68k
9544 targets. Normally the facilities of the machine's usual C compiler are
9545 used, but this can't be done directly in cross-compilation. You must
9546 make your own arrangements to provide suitable library functions for
9547 cross-compilation. The embedded targets @samp{m68k-*-aout} and
9548 @samp{m68k-*-coff} do provide software floating point support.
9552 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9553 Additionally, parameters passed on the stack are also aligned to a
9554 16-bit boundary even on targets whose API mandates promotion to 32-bit.
9557 @opindex mnobitfield
9558 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
9559 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
9563 Do use the bit-field instructions. The @option{-m68020} option implies
9564 @option{-mbitfield}. This is the default if you use a configuration
9565 designed for a 68020.
9569 Use a different function-calling convention, in which functions
9570 that take a fixed number of arguments return with the @code{rtd}
9571 instruction, which pops their arguments while returning. This
9572 saves one instruction in the caller since there is no need to pop
9573 the arguments there.
9575 This calling convention is incompatible with the one normally
9576 used on Unix, so you cannot use it if you need to call libraries
9577 compiled with the Unix compiler.
9579 Also, you must provide function prototypes for all functions that
9580 take variable numbers of arguments (including @code{printf});
9581 otherwise incorrect code will be generated for calls to those
9584 In addition, seriously incorrect code will result if you call a
9585 function with too many arguments. (Normally, extra arguments are
9586 harmlessly ignored.)
9588 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
9589 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
9592 @itemx -mno-align-int
9594 @opindex mno-align-int
9595 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
9596 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
9597 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
9598 Aligning variables on 32-bit boundaries produces code that runs somewhat
9599 faster on processors with 32-bit busses at the expense of more memory.
9601 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
9602 align structures containing the above types differently than
9603 most published application binary interface specifications for the m68k.
9607 Use the pc-relative addressing mode of the 68000 directly, instead of
9608 using a global offset table. At present, this option implies @option{-fpic},
9609 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
9610 not presently supported with @option{-mpcrel}, though this could be supported for
9611 68020 and higher processors.
9613 @item -mno-strict-align
9614 @itemx -mstrict-align
9615 @opindex mno-strict-align
9616 @opindex mstrict-align
9617 Do not (do) assume that unaligned memory references will be handled by
9621 Generate code that allows the data segment to be located in a different
9622 area of memory from the text segment. This allows for execute in place in
9623 an environment without virtual memory management. This option implies
9627 Generate code that assumes that the data segment follows the text segment.
9628 This is the default.
9630 @item -mid-shared-library
9631 Generate code that supports shared libraries via the library ID method.
9632 This allows for execute in place and shared libraries in an environment
9633 without virtual memory management. This option implies @option{-fPIC}.
9635 @item -mno-id-shared-library
9636 Generate code that doesn't assume ID based shared libraries are being used.
9637 This is the default.
9639 @item -mshared-library-id=n
9640 Specified the identification number of the ID based shared library being
9641 compiled. Specifying a value of 0 will generate more compact code, specifying
9642 other values will force the allocation of that number to the current
9643 library but is no more space or time efficient than omitting this option.
9647 @node M68hc1x Options
9648 @subsection M68hc1x Options
9649 @cindex M68hc1x options
9651 These are the @samp{-m} options defined for the 68hc11 and 68hc12
9652 microcontrollers. The default values for these options depends on
9653 which style of microcontroller was selected when the compiler was configured;
9654 the defaults for the most common choices are given below.
9661 Generate output for a 68HC11. This is the default
9662 when the compiler is configured for 68HC11-based systems.
9668 Generate output for a 68HC12. This is the default
9669 when the compiler is configured for 68HC12-based systems.
9675 Generate output for a 68HCS12.
9678 @opindex mauto-incdec
9679 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
9686 Enable the use of 68HC12 min and max instructions.
9689 @itemx -mno-long-calls
9690 @opindex mlong-calls
9691 @opindex mno-long-calls
9692 Treat all calls as being far away (near). If calls are assumed to be
9693 far away, the compiler will use the @code{call} instruction to
9694 call a function and the @code{rtc} instruction for returning.
9698 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9700 @item -msoft-reg-count=@var{count}
9701 @opindex msoft-reg-count
9702 Specify the number of pseudo-soft registers which are used for the
9703 code generation. The maximum number is 32. Using more pseudo-soft
9704 register may or may not result in better code depending on the program.
9705 The default is 4 for 68HC11 and 2 for 68HC12.
9710 @subsection MCore Options
9711 @cindex MCore options
9713 These are the @samp{-m} options defined for the Motorola M*Core
9721 @opindex mno-hardlit
9722 Inline constants into the code stream if it can be done in two
9723 instructions or less.
9729 Use the divide instruction. (Enabled by default).
9731 @item -mrelax-immediate
9732 @itemx -mno-relax-immediate
9733 @opindex mrelax-immediate
9734 @opindex mno-relax-immediate
9735 Allow arbitrary sized immediates in bit operations.
9737 @item -mwide-bitfields
9738 @itemx -mno-wide-bitfields
9739 @opindex mwide-bitfields
9740 @opindex mno-wide-bitfields
9741 Always treat bit-fields as int-sized.
9743 @item -m4byte-functions
9744 @itemx -mno-4byte-functions
9745 @opindex m4byte-functions
9746 @opindex mno-4byte-functions
9747 Force all functions to be aligned to a four byte boundary.
9749 @item -mcallgraph-data
9750 @itemx -mno-callgraph-data
9751 @opindex mcallgraph-data
9752 @opindex mno-callgraph-data
9753 Emit callgraph information.
9756 @itemx -mno-slow-bytes
9757 @opindex mslow-bytes
9758 @opindex mno-slow-bytes
9759 Prefer word access when reading byte quantities.
9761 @item -mlittle-endian
9763 @opindex mlittle-endian
9764 @opindex mbig-endian
9765 Generate code for a little endian target.
9771 Generate code for the 210 processor.
9775 @subsection MIPS Options
9776 @cindex MIPS options
9782 Generate big-endian code.
9786 Generate little-endian code. This is the default for @samp{mips*el-*-*}
9789 @item -march=@var{arch}
9791 Generate code that will run on @var{arch}, which can be the name of a
9792 generic MIPS ISA, or the name of a particular processor.
9794 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
9795 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
9796 The processor names are:
9797 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
9799 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
9800 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
9804 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
9805 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
9806 The special value @samp{from-abi} selects the
9807 most compatible architecture for the selected ABI (that is,
9808 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
9810 In processor names, a final @samp{000} can be abbreviated as @samp{k}
9811 (for example, @samp{-march=r2k}). Prefixes are optional, and
9812 @samp{vr} may be written @samp{r}.
9814 GCC defines two macros based on the value of this option. The first
9815 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
9816 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
9817 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
9818 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
9819 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
9821 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
9822 above. In other words, it will have the full prefix and will not
9823 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
9824 the macro names the resolved architecture (either @samp{"mips1"} or
9825 @samp{"mips3"}). It names the default architecture when no
9826 @option{-march} option is given.
9828 @item -mtune=@var{arch}
9830 Optimize for @var{arch}. Among other things, this option controls
9831 the way instructions are scheduled, and the perceived cost of arithmetic
9832 operations. The list of @var{arch} values is the same as for
9835 When this option is not used, GCC will optimize for the processor
9836 specified by @option{-march}. By using @option{-march} and
9837 @option{-mtune} together, it is possible to generate code that will
9838 run on a family of processors, but optimize the code for one
9839 particular member of that family.
9841 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
9842 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
9843 @samp{-march} ones described above.
9847 Equivalent to @samp{-march=mips1}.
9851 Equivalent to @samp{-march=mips2}.
9855 Equivalent to @samp{-march=mips3}.
9859 Equivalent to @samp{-march=mips4}.
9863 Equivalent to @samp{-march=mips32}.
9867 Equivalent to @samp{-march=mips32r2}.
9871 Equivalent to @samp{-march=mips64}.
9877 Generate (do not generate) MIPS16 code. If GCC is targetting a
9878 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
9890 Generate code for the given ABI@.
9892 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
9893 generates 64-bit code when you select a 64-bit architecture, but you
9894 can use @option{-mgp32} to get 32-bit code instead.
9896 For information about the O64 ABI, see
9897 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
9900 @itemx -mno-abicalls
9902 @opindex mno-abicalls
9903 Generate (do not generate) SVR4-style position-independent code.
9904 @option{-mabicalls} is the default for SVR4-based systems.
9910 Lift (do not lift) the usual restrictions on the size of the global
9913 GCC normally uses a single instruction to load values from the GOT@.
9914 While this is relatively efficient, it will only work if the GOT
9915 is smaller than about 64k. Anything larger will cause the linker
9916 to report an error such as:
9918 @cindex relocation truncated to fit (MIPS)
9920 relocation truncated to fit: R_MIPS_GOT16 foobar
9923 If this happens, you should recompile your code with @option{-mxgot}.
9924 It should then work with very large GOTs, although it will also be
9925 less efficient, since it will take three instructions to fetch the
9926 value of a global symbol.
9928 Note that some linkers can create multiple GOTs. If you have such a
9929 linker, you should only need to use @option{-mxgot} when a single object
9930 file accesses more than 64k's worth of GOT entries. Very few do.
9932 These options have no effect unless GCC is generating position
9937 Assume that general-purpose registers are 32 bits wide.
9941 Assume that general-purpose registers are 64 bits wide.
9945 Assume that floating-point registers are 32 bits wide.
9949 Assume that floating-point registers are 64 bits wide.
9952 @opindex mhard-float
9953 Use floating-point coprocessor instructions.
9956 @opindex msoft-float
9957 Do not use floating-point coprocessor instructions. Implement
9958 floating-point calculations using library calls instead.
9960 @item -msingle-float
9961 @opindex msingle-float
9962 Assume that the floating-point coprocessor only supports single-precision
9965 @itemx -mdouble-float
9966 @opindex mdouble-float
9967 Assume that the floating-point coprocessor supports double-precision
9968 operations. This is the default.
9970 @itemx -mpaired-single
9971 @itemx -mno-paired-single
9972 @opindex mpaired-single
9973 @opindex mno-paired-single
9974 Use (do not use) paired-single floating-point instructions.
9975 @xref{MIPS Paired-Single Support}. This option can only be used
9976 when generating 64-bit code and requires hardware floating-point
9977 support to be enabled.
9983 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
9984 The option @option{-mips3d} implies @option{-mpaired-single}.
9988 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
9989 an explanation of the default and the way that the pointer size is
9994 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
9996 The default size of @code{int}s, @code{long}s and pointers depends on
9997 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
9998 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
9999 32-bit @code{long}s. Pointers are the same size as @code{long}s,
10000 or the same size as integer registers, whichever is smaller.
10006 Assume (do not assume) that all symbols have 32-bit values, regardless
10007 of the selected ABI@. This option is useful in combination with
10008 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
10009 to generate shorter and faster references to symbolic addresses.
10013 @cindex smaller data references (MIPS)
10014 @cindex gp-relative references (MIPS)
10015 Put global and static items less than or equal to @var{num} bytes into
10016 the small data or bss section instead of the normal data or bss section.
10017 This allows the data to be accessed using a single instruction.
10019 All modules should be compiled with the same @option{-G @var{num}}
10022 @item -membedded-data
10023 @itemx -mno-embedded-data
10024 @opindex membedded-data
10025 @opindex mno-embedded-data
10026 Allocate variables to the read-only data section first if possible, then
10027 next in the small data section if possible, otherwise in data. This gives
10028 slightly slower code than the default, but reduces the amount of RAM required
10029 when executing, and thus may be preferred for some embedded systems.
10031 @item -muninit-const-in-rodata
10032 @itemx -mno-uninit-const-in-rodata
10033 @opindex muninit-const-in-rodata
10034 @opindex mno-uninit-const-in-rodata
10035 Put uninitialized @code{const} variables in the read-only data section.
10036 This option is only meaningful in conjunction with @option{-membedded-data}.
10038 @item -msplit-addresses
10039 @itemx -mno-split-addresses
10040 @opindex msplit-addresses
10041 @opindex mno-split-addresses
10042 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
10043 relocation operators. This option has been superseded by
10044 @option{-mexplicit-relocs} but is retained for backwards compatibility.
10046 @item -mexplicit-relocs
10047 @itemx -mno-explicit-relocs
10048 @opindex mexplicit-relocs
10049 @opindex mno-explicit-relocs
10050 Use (do not use) assembler relocation operators when dealing with symbolic
10051 addresses. The alternative, selected by @option{-mno-explicit-relocs},
10052 is to use assembler macros instead.
10054 @option{-mexplicit-relocs} is the default if GCC was configured
10055 to use an assembler that supports relocation operators.
10057 @item -mcheck-zero-division
10058 @itemx -mno-check-zero-division
10059 @opindex mcheck-zero-division
10060 @opindex mno-check-zero-division
10061 Trap (do not trap) on integer division by zero. The default is
10062 @option{-mcheck-zero-division}.
10064 @item -mdivide-traps
10065 @itemx -mdivide-breaks
10066 @opindex mdivide-traps
10067 @opindex mdivide-breaks
10068 MIPS systems check for division by zero by generating either a
10069 conditional trap or a break instruction. Using traps results in
10070 smaller code, but is only supported on MIPS II and later. Also, some
10071 versions of the Linux kernel have a bug that prevents trap from
10072 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
10073 allow conditional traps on architectures that support them and
10074 @option{-mdivide-breaks} to force the use of breaks.
10076 The default is usually @option{-mdivide-traps}, but this can be
10077 overridden at configure time using @option{--with-divide=breaks}.
10078 Divide-by-zero checks can be completely disabled using
10079 @option{-mno-check-zero-division}.
10084 @opindex mno-memcpy
10085 Force (do not force) the use of @code{memcpy()} for non-trivial block
10086 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
10087 most constant-sized copies.
10090 @itemx -mno-long-calls
10091 @opindex mlong-calls
10092 @opindex mno-long-calls
10093 Disable (do not disable) use of the @code{jal} instruction. Calling
10094 functions using @code{jal} is more efficient but requires the caller
10095 and callee to be in the same 256 megabyte segment.
10097 This option has no effect on abicalls code. The default is
10098 @option{-mno-long-calls}.
10104 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
10105 instructions, as provided by the R4650 ISA@.
10108 @itemx -mno-fused-madd
10109 @opindex mfused-madd
10110 @opindex mno-fused-madd
10111 Enable (disable) use of the floating point multiply-accumulate
10112 instructions, when they are available. The default is
10113 @option{-mfused-madd}.
10115 When multiply-accumulate instructions are used, the intermediate
10116 product is calculated to infinite precision and is not subject to
10117 the FCSR Flush to Zero bit. This may be undesirable in some
10122 Tell the MIPS assembler to not run its preprocessor over user
10123 assembler files (with a @samp{.s} suffix) when assembling them.
10126 @itemx -mno-fix-r4000
10127 @opindex mfix-r4000
10128 @opindex mno-fix-r4000
10129 Work around certain R4000 CPU errata:
10132 A double-word or a variable shift may give an incorrect result if executed
10133 immediately after starting an integer division.
10135 A double-word or a variable shift may give an incorrect result if executed
10136 while an integer multiplication is in progress.
10138 An integer division may give an incorrect result if started in a delay slot
10139 of a taken branch or a jump.
10143 @itemx -mno-fix-r4400
10144 @opindex mfix-r4400
10145 @opindex mno-fix-r4400
10146 Work around certain R4400 CPU errata:
10149 A double-word or a variable shift may give an incorrect result if executed
10150 immediately after starting an integer division.
10154 @itemx -mno-fix-vr4120
10155 @opindex mfix-vr4120
10156 Work around certain VR4120 errata:
10159 @code{dmultu} does not always produce the correct result.
10161 @code{div} and @code{ddiv} do not always produce the correct result if one
10162 of the operands is negative.
10164 The workarounds for the division errata rely on special functions in
10165 @file{libgcc.a}. At present, these functions are only provided by
10166 the @code{mips64vr*-elf} configurations.
10168 Other VR4120 errata require a nop to be inserted between certain pairs of
10169 instructions. These errata are handled by the assembler, not by GCC itself.
10172 @opindex mfix-vr4130
10173 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
10174 workarounds are implemented by the assembler rather than by GCC,
10175 although GCC will avoid using @code{mflo} and @code{mfhi} if the
10176 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
10177 instructions are available instead.
10180 @itemx -mno-fix-sb1
10182 Work around certain SB-1 CPU core errata.
10183 (This flag currently works around the SB-1 revision 2
10184 ``F1'' and ``F2'' floating point errata.)
10186 @item -mflush-func=@var{func}
10187 @itemx -mno-flush-func
10188 @opindex mflush-func
10189 Specifies the function to call to flush the I and D caches, or to not
10190 call any such function. If called, the function must take the same
10191 arguments as the common @code{_flush_func()}, that is, the address of the
10192 memory range for which the cache is being flushed, the size of the
10193 memory range, and the number 3 (to flush both caches). The default
10194 depends on the target GCC was configured for, but commonly is either
10195 @samp{_flush_func} or @samp{__cpu_flush}.
10197 @item -mbranch-likely
10198 @itemx -mno-branch-likely
10199 @opindex mbranch-likely
10200 @opindex mno-branch-likely
10201 Enable or disable use of Branch Likely instructions, regardless of the
10202 default for the selected architecture. By default, Branch Likely
10203 instructions may be generated if they are supported by the selected
10204 architecture. An exception is for the MIPS32 and MIPS64 architectures
10205 and processors which implement those architectures; for those, Branch
10206 Likely instructions will not be generated by default because the MIPS32
10207 and MIPS64 architectures specifically deprecate their use.
10209 @item -mfp-exceptions
10210 @itemx -mno-fp-exceptions
10211 @opindex mfp-exceptions
10212 Specifies whether FP exceptions are enabled. This affects how we schedule
10213 FP instructions for some processors. The default is that FP exceptions are
10216 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
10217 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
10220 @item -mvr4130-align
10221 @itemx -mno-vr4130-align
10222 @opindex mvr4130-align
10223 The VR4130 pipeline is two-way superscalar, but can only issue two
10224 instructions together if the first one is 8-byte aligned. When this
10225 option is enabled, GCC will align pairs of instructions that it
10226 thinks should execute in parallel.
10228 This option only has an effect when optimizing for the VR4130.
10229 It normally makes code faster, but at the expense of making it bigger.
10230 It is enabled by default at optimization level @option{-O3}.
10234 @subsection MMIX Options
10235 @cindex MMIX Options
10237 These options are defined for the MMIX:
10241 @itemx -mno-libfuncs
10243 @opindex mno-libfuncs
10244 Specify that intrinsic library functions are being compiled, passing all
10245 values in registers, no matter the size.
10248 @itemx -mno-epsilon
10250 @opindex mno-epsilon
10251 Generate floating-point comparison instructions that compare with respect
10252 to the @code{rE} epsilon register.
10254 @item -mabi=mmixware
10256 @opindex mabi-mmixware
10258 Generate code that passes function parameters and return values that (in
10259 the called function) are seen as registers @code{$0} and up, as opposed to
10260 the GNU ABI which uses global registers @code{$231} and up.
10262 @item -mzero-extend
10263 @itemx -mno-zero-extend
10264 @opindex mzero-extend
10265 @opindex mno-zero-extend
10266 When reading data from memory in sizes shorter than 64 bits, use (do not
10267 use) zero-extending load instructions by default, rather than
10268 sign-extending ones.
10271 @itemx -mno-knuthdiv
10273 @opindex mno-knuthdiv
10274 Make the result of a division yielding a remainder have the same sign as
10275 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10276 remainder follows the sign of the dividend. Both methods are
10277 arithmetically valid, the latter being almost exclusively used.
10279 @item -mtoplevel-symbols
10280 @itemx -mno-toplevel-symbols
10281 @opindex mtoplevel-symbols
10282 @opindex mno-toplevel-symbols
10283 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10284 code can be used with the @code{PREFIX} assembly directive.
10288 Generate an executable in the ELF format, rather than the default
10289 @samp{mmo} format used by the @command{mmix} simulator.
10291 @item -mbranch-predict
10292 @itemx -mno-branch-predict
10293 @opindex mbranch-predict
10294 @opindex mno-branch-predict
10295 Use (do not use) the probable-branch instructions, when static branch
10296 prediction indicates a probable branch.
10298 @item -mbase-addresses
10299 @itemx -mno-base-addresses
10300 @opindex mbase-addresses
10301 @opindex mno-base-addresses
10302 Generate (do not generate) code that uses @emph{base addresses}. Using a
10303 base address automatically generates a request (handled by the assembler
10304 and the linker) for a constant to be set up in a global register. The
10305 register is used for one or more base address requests within the range 0
10306 to 255 from the value held in the register. The generally leads to short
10307 and fast code, but the number of different data items that can be
10308 addressed is limited. This means that a program that uses lots of static
10309 data may require @option{-mno-base-addresses}.
10311 @item -msingle-exit
10312 @itemx -mno-single-exit
10313 @opindex msingle-exit
10314 @opindex mno-single-exit
10315 Force (do not force) generated code to have a single exit point in each
10319 @node MN10300 Options
10320 @subsection MN10300 Options
10321 @cindex MN10300 options
10323 These @option{-m} options are defined for Matsushita MN10300 architectures:
10328 Generate code to avoid bugs in the multiply instructions for the MN10300
10329 processors. This is the default.
10331 @item -mno-mult-bug
10332 @opindex mno-mult-bug
10333 Do not generate code to avoid bugs in the multiply instructions for the
10334 MN10300 processors.
10338 Generate code which uses features specific to the AM33 processor.
10342 Do not generate code which uses features specific to the AM33 processor. This
10347 Do not link in the C run-time initialization object file.
10351 Indicate to the linker that it should perform a relaxation optimization pass
10352 to shorten branches, calls and absolute memory addresses. This option only
10353 has an effect when used on the command line for the final link step.
10355 This option makes symbolic debugging impossible.
10358 @node NS32K Options
10359 @subsection NS32K Options
10360 @cindex NS32K options
10362 These are the @samp{-m} options defined for the 32000 series. The default
10363 values for these options depends on which style of 32000 was selected when
10364 the compiler was configured; the defaults for the most common choices are
10372 Generate output for a 32032. This is the default
10373 when the compiler is configured for 32032 and 32016 based systems.
10379 Generate output for a 32332. This is the default
10380 when the compiler is configured for 32332-based systems.
10386 Generate output for a 32532. This is the default
10387 when the compiler is configured for 32532-based systems.
10391 Generate output containing 32081 instructions for floating point.
10392 This is the default for all systems.
10396 Generate output containing 32381 instructions for floating point. This
10397 also implies @option{-m32081}. The 32381 is only compatible with the 32332
10398 and 32532 cpus. This is the default for the pc532-netbsd configuration.
10401 @opindex mmulti-add
10402 Try and generate multiply-add floating point instructions @code{polyF}
10403 and @code{dotF}. This option is only available if the @option{-m32381}
10404 option is in effect. Using these instructions requires changes to
10405 register allocation which generally has a negative impact on
10406 performance. This option should only be enabled when compiling code
10407 particularly likely to make heavy use of multiply-add instructions.
10409 @item -mnomulti-add
10410 @opindex mnomulti-add
10411 Do not try and generate multiply-add floating point instructions
10412 @code{polyF} and @code{dotF}. This is the default on all platforms.
10415 @opindex msoft-float
10416 Generate output containing library calls for floating point.
10417 @strong{Warning:} the requisite libraries may not be available.
10419 @item -mieee-compare
10420 @itemx -mno-ieee-compare
10421 @opindex mieee-compare
10422 @opindex mno-ieee-compare
10423 Control whether or not the compiler uses IEEE floating point
10424 comparisons. These handle correctly the case where the result of a
10425 comparison is unordered.
10426 @strong{Warning:} the requisite kernel support may not be available.
10429 @opindex mnobitfield
10430 Do not use the bit-field instructions. On some machines it is faster to
10431 use shifting and masking operations. This is the default for the pc532.
10435 Do use the bit-field instructions. This is the default for all platforms
10440 Use a different function-calling convention, in which functions
10441 that take a fixed number of arguments return pop their
10442 arguments on return with the @code{ret} instruction.
10444 This calling convention is incompatible with the one normally
10445 used on Unix, so you cannot use it if you need to call libraries
10446 compiled with the Unix compiler.
10448 Also, you must provide function prototypes for all functions that
10449 take variable numbers of arguments (including @code{printf});
10450 otherwise incorrect code will be generated for calls to those
10453 In addition, seriously incorrect code will result if you call a
10454 function with too many arguments. (Normally, extra arguments are
10455 harmlessly ignored.)
10457 This option takes its name from the 680x0 @code{rtd} instruction.
10462 Use a different function-calling convention where the first two arguments
10463 are passed in registers.
10465 This calling convention is incompatible with the one normally
10466 used on Unix, so you cannot use it if you need to call libraries
10467 compiled with the Unix compiler.
10470 @opindex mnoregparam
10471 Do not pass any arguments in registers. This is the default for all
10476 It is OK to use the sb as an index register which is always loaded with
10477 zero. This is the default for the pc532-netbsd target.
10481 The sb register is not available for use or has not been initialized to
10482 zero by the run time system. This is the default for all targets except
10483 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
10484 @option{-fpic} is set.
10488 Many ns32000 series addressing modes use displacements of up to 512MB@.
10489 If an address is above 512MB then displacements from zero can not be used.
10490 This option causes code to be generated which can be loaded above 512MB@.
10491 This may be useful for operating systems or ROM code.
10495 Assume code will be loaded in the first 512MB of virtual address space.
10496 This is the default for all platforms.
10500 @node PDP-11 Options
10501 @subsection PDP-11 Options
10502 @cindex PDP-11 Options
10504 These options are defined for the PDP-11:
10509 Use hardware FPP floating point. This is the default. (FIS floating
10510 point on the PDP-11/40 is not supported.)
10513 @opindex msoft-float
10514 Do not use hardware floating point.
10518 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10522 Return floating-point results in memory. This is the default.
10526 Generate code for a PDP-11/40.
10530 Generate code for a PDP-11/45. This is the default.
10534 Generate code for a PDP-11/10.
10536 @item -mbcopy-builtin
10537 @opindex bcopy-builtin
10538 Use inline @code{movmemhi} patterns for copying memory. This is the
10543 Do not use inline @code{movmemhi} patterns for copying memory.
10549 Use 16-bit @code{int}. This is the default.
10555 Use 32-bit @code{int}.
10558 @itemx -mno-float32
10560 @opindex mno-float32
10561 Use 64-bit @code{float}. This is the default.
10564 @itemx -mno-float64
10566 @opindex mno-float64
10567 Use 32-bit @code{float}.
10571 Use @code{abshi2} pattern. This is the default.
10575 Do not use @code{abshi2} pattern.
10577 @item -mbranch-expensive
10578 @opindex mbranch-expensive
10579 Pretend that branches are expensive. This is for experimenting with
10580 code generation only.
10582 @item -mbranch-cheap
10583 @opindex mbranch-cheap
10584 Do not pretend that branches are expensive. This is the default.
10588 Generate code for a system with split I&D@.
10592 Generate code for a system without split I&D@. This is the default.
10596 Use Unix assembler syntax. This is the default when configured for
10597 @samp{pdp11-*-bsd}.
10601 Use DEC assembler syntax. This is the default when configured for any
10602 PDP-11 target other than @samp{pdp11-*-bsd}.
10605 @node PowerPC Options
10606 @subsection PowerPC Options
10607 @cindex PowerPC options
10609 These are listed under @xref{RS/6000 and PowerPC Options}.
10611 @node RS/6000 and PowerPC Options
10612 @subsection IBM RS/6000 and PowerPC Options
10613 @cindex RS/6000 and PowerPC Options
10614 @cindex IBM RS/6000 and PowerPC Options
10616 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
10623 @itemx -mno-powerpc
10624 @itemx -mpowerpc-gpopt
10625 @itemx -mno-powerpc-gpopt
10626 @itemx -mpowerpc-gfxopt
10627 @itemx -mno-powerpc-gfxopt
10629 @itemx -mno-powerpc64
10633 @opindex mno-power2
10635 @opindex mno-powerpc
10636 @opindex mpowerpc-gpopt
10637 @opindex mno-powerpc-gpopt
10638 @opindex mpowerpc-gfxopt
10639 @opindex mno-powerpc-gfxopt
10640 @opindex mpowerpc64
10641 @opindex mno-powerpc64
10642 GCC supports two related instruction set architectures for the
10643 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
10644 instructions supported by the @samp{rios} chip set used in the original
10645 RS/6000 systems and the @dfn{PowerPC} instruction set is the
10646 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
10647 the IBM 4xx microprocessors.
10649 Neither architecture is a subset of the other. However there is a
10650 large common subset of instructions supported by both. An MQ
10651 register is included in processors supporting the POWER architecture.
10653 You use these options to specify which instructions are available on the
10654 processor you are using. The default value of these options is
10655 determined when configuring GCC@. Specifying the
10656 @option{-mcpu=@var{cpu_type}} overrides the specification of these
10657 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
10658 rather than the options listed above.
10660 The @option{-mpower} option allows GCC to generate instructions that
10661 are found only in the POWER architecture and to use the MQ register.
10662 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
10663 to generate instructions that are present in the POWER2 architecture but
10664 not the original POWER architecture.
10666 The @option{-mpowerpc} option allows GCC to generate instructions that
10667 are found only in the 32-bit subset of the PowerPC architecture.
10668 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
10669 GCC to use the optional PowerPC architecture instructions in the
10670 General Purpose group, including floating-point square root. Specifying
10671 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
10672 use the optional PowerPC architecture instructions in the Graphics
10673 group, including floating-point select.
10675 The @option{-mpowerpc64} option allows GCC to generate the additional
10676 64-bit instructions that are found in the full PowerPC64 architecture
10677 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
10678 @option{-mno-powerpc64}.
10680 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
10681 will use only the instructions in the common subset of both
10682 architectures plus some special AIX common-mode calls, and will not use
10683 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
10684 permits GCC to use any instruction from either architecture and to
10685 allow use of the MQ register; specify this for the Motorola MPC601.
10687 @item -mnew-mnemonics
10688 @itemx -mold-mnemonics
10689 @opindex mnew-mnemonics
10690 @opindex mold-mnemonics
10691 Select which mnemonics to use in the generated assembler code. With
10692 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
10693 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
10694 assembler mnemonics defined for the POWER architecture. Instructions
10695 defined in only one architecture have only one mnemonic; GCC uses that
10696 mnemonic irrespective of which of these options is specified.
10698 GCC defaults to the mnemonics appropriate for the architecture in
10699 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
10700 value of these option. Unless you are building a cross-compiler, you
10701 should normally not specify either @option{-mnew-mnemonics} or
10702 @option{-mold-mnemonics}, but should instead accept the default.
10704 @item -mcpu=@var{cpu_type}
10706 Set architecture type, register usage, choice of mnemonics, and
10707 instruction scheduling parameters for machine type @var{cpu_type}.
10708 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
10709 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
10710 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
10711 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
10712 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
10713 @samp{860}, @samp{970}, @samp{8540}, @samp{common}, @samp{ec603e}, @samp{G3},
10714 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
10715 @samp{power4}, @samp{power5}, @samp{powerpc}, @samp{powerpc64},
10716 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
10718 @option{-mcpu=common} selects a completely generic processor. Code
10719 generated under this option will run on any POWER or PowerPC processor.
10720 GCC will use only the instructions in the common subset of both
10721 architectures, and will not use the MQ register. GCC assumes a generic
10722 processor model for scheduling purposes.
10724 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
10725 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
10726 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
10727 types, with an appropriate, generic processor model assumed for
10728 scheduling purposes.
10730 The other options specify a specific processor. Code generated under
10731 those options will run best on that processor, and may not run at all on
10734 The @option{-mcpu} options automatically enable or disable the
10735 following options: @option{-maltivec}, @option{-mhard-float},
10736 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
10737 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
10738 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
10739 @option{-mstring}. The particular options set for any particular CPU
10740 will vary between compiler versions, depending on what setting seems
10741 to produce optimal code for that CPU; it doesn't necessarily reflect
10742 the actual hardware's capabilities. If you wish to set an individual
10743 option to a particular value, you may specify it after the
10744 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
10746 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
10747 not enabled or disabled by the @option{-mcpu} option at present, since
10748 AIX does not have full support for these options. You may still
10749 enable or disable them individually if you're sure it'll work in your
10752 @item -mtune=@var{cpu_type}
10754 Set the instruction scheduling parameters for machine type
10755 @var{cpu_type}, but do not set the architecture type, register usage, or
10756 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
10757 values for @var{cpu_type} are used for @option{-mtune} as for
10758 @option{-mcpu}. If both are specified, the code generated will use the
10759 architecture, registers, and mnemonics set by @option{-mcpu}, but the
10760 scheduling parameters set by @option{-mtune}.
10766 Generate code to compute division as reciprocal estimate and iterative
10767 refinement, creating opportunities for increased throughput. This
10768 feature requires: optional PowerPC Graphics instruction set for single
10769 precision and FRE instruction for double precision, assuming divides
10770 cannot generate user-visible traps, and the domain values not include
10771 Infinities, denormals or zero denominator.
10774 @itemx -mno-altivec
10776 @opindex mno-altivec
10777 Generate code that uses (does not use) AltiVec instructions, and also
10778 enable the use of built-in functions that allow more direct access to
10779 the AltiVec instruction set. You may also need to set
10780 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
10786 @opindex mno-vrsave
10787 Generate VRSAVE instructions when generating AltiVec code.
10791 Extend the current ABI with SPE ABI extensions. This does not change
10792 the default ABI, instead it adds the SPE ABI extensions to the current
10796 @opindex mabi=no-spe
10797 Disable Booke SPE ABI extensions for the current ABI@.
10800 @opindex msecure-plt
10801 Generate code that allows ld and ld.so to build executables and shared
10802 libraries with non-exec .plt and .got sections. This is a PowerPC
10803 32-bit SYSV ABI option.
10807 Generate code that uses a BSS .plt section that ld.so fills in, and
10808 requires .plt and .got sections that are both writable and executable.
10809 This is a PowerPC 32-bit SYSV ABI option.
10815 This switch enables or disables the generation of ISEL instructions.
10817 @item -misel=@var{yes/no}
10818 This switch has been deprecated. Use @option{-misel} and
10819 @option{-mno-isel} instead.
10825 This switch enables or disables the generation of SPE simd
10828 @item -mspe=@var{yes/no}
10829 This option has been deprecated. Use @option{-mspe} and
10830 @option{-mno-spe} instead.
10832 @item -mfloat-gprs=@var{yes/single/double/no}
10833 @itemx -mfloat-gprs
10834 @opindex mfloat-gprs
10835 This switch enables or disables the generation of floating point
10836 operations on the general purpose registers for architectures that
10839 The argument @var{yes} or @var{single} enables the use of
10840 single-precision floating point operations.
10842 The argument @var{double} enables the use of single and
10843 double-precision floating point operations.
10845 The argument @var{no} disables floating point operations on the
10846 general purpose registers.
10848 This option is currently only available on the MPC854x.
10854 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
10855 targets (including GNU/Linux). The 32-bit environment sets int, long
10856 and pointer to 32 bits and generates code that runs on any PowerPC
10857 variant. The 64-bit environment sets int to 32 bits and long and
10858 pointer to 64 bits, and generates code for PowerPC64, as for
10859 @option{-mpowerpc64}.
10862 @itemx -mno-fp-in-toc
10863 @itemx -mno-sum-in-toc
10864 @itemx -mminimal-toc
10866 @opindex mno-fp-in-toc
10867 @opindex mno-sum-in-toc
10868 @opindex mminimal-toc
10869 Modify generation of the TOC (Table Of Contents), which is created for
10870 every executable file. The @option{-mfull-toc} option is selected by
10871 default. In that case, GCC will allocate at least one TOC entry for
10872 each unique non-automatic variable reference in your program. GCC
10873 will also place floating-point constants in the TOC@. However, only
10874 16,384 entries are available in the TOC@.
10876 If you receive a linker error message that saying you have overflowed
10877 the available TOC space, you can reduce the amount of TOC space used
10878 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
10879 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
10880 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
10881 generate code to calculate the sum of an address and a constant at
10882 run-time instead of putting that sum into the TOC@. You may specify one
10883 or both of these options. Each causes GCC to produce very slightly
10884 slower and larger code at the expense of conserving TOC space.
10886 If you still run out of space in the TOC even when you specify both of
10887 these options, specify @option{-mminimal-toc} instead. This option causes
10888 GCC to make only one TOC entry for every file. When you specify this
10889 option, GCC will produce code that is slower and larger but which
10890 uses extremely little TOC space. You may wish to use this option
10891 only on files that contain less frequently executed code.
10897 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
10898 @code{long} type, and the infrastructure needed to support them.
10899 Specifying @option{-maix64} implies @option{-mpowerpc64} and
10900 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
10901 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
10904 @itemx -mno-xl-compat
10905 @opindex mxl-compat
10906 @opindex mno-xl-compat
10907 Produce code that conforms more closely to IBM XLC semantics when using
10908 AIX-compatible ABI. Pass floating-point arguments to prototyped
10909 functions beyond the register save area (RSA) on the stack in addition
10910 to argument FPRs. Do not assume that most significant double in 128
10911 bit long double value is properly rounded when comparing values.
10913 The AIX calling convention was extended but not initially documented to
10914 handle an obscure K&R C case of calling a function that takes the
10915 address of its arguments with fewer arguments than declared. AIX XL
10916 compilers access floating point arguments which do not fit in the
10917 RSA from the stack when a subroutine is compiled without
10918 optimization. Because always storing floating-point arguments on the
10919 stack is inefficient and rarely needed, this option is not enabled by
10920 default and only is necessary when calling subroutines compiled by AIX
10921 XL compilers without optimization.
10925 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
10926 application written to use message passing with special startup code to
10927 enable the application to run. The system must have PE installed in the
10928 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
10929 must be overridden with the @option{-specs=} option to specify the
10930 appropriate directory location. The Parallel Environment does not
10931 support threads, so the @option{-mpe} option and the @option{-pthread}
10932 option are incompatible.
10934 @item -malign-natural
10935 @itemx -malign-power
10936 @opindex malign-natural
10937 @opindex malign-power
10938 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
10939 @option{-malign-natural} overrides the ABI-defined alignment of larger
10940 types, such as floating-point doubles, on their natural size-based boundary.
10941 The option @option{-malign-power} instructs GCC to follow the ABI-specified
10942 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
10944 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
10948 @itemx -mhard-float
10949 @opindex msoft-float
10950 @opindex mhard-float
10951 Generate code that does not use (uses) the floating-point register set.
10952 Software floating point emulation is provided if you use the
10953 @option{-msoft-float} option, and pass the option to GCC when linking.
10956 @itemx -mno-multiple
10958 @opindex mno-multiple
10959 Generate code that uses (does not use) the load multiple word
10960 instructions and the store multiple word instructions. These
10961 instructions are generated by default on POWER systems, and not
10962 generated on PowerPC systems. Do not use @option{-mmultiple} on little
10963 endian PowerPC systems, since those instructions do not work when the
10964 processor is in little endian mode. The exceptions are PPC740 and
10965 PPC750 which permit the instructions usage in little endian mode.
10970 @opindex mno-string
10971 Generate code that uses (does not use) the load string instructions
10972 and the store string word instructions to save multiple registers and
10973 do small block moves. These instructions are generated by default on
10974 POWER systems, and not generated on PowerPC systems. Do not use
10975 @option{-mstring} on little endian PowerPC systems, since those
10976 instructions do not work when the processor is in little endian mode.
10977 The exceptions are PPC740 and PPC750 which permit the instructions
10978 usage in little endian mode.
10983 @opindex mno-update
10984 Generate code that uses (does not use) the load or store instructions
10985 that update the base register to the address of the calculated memory
10986 location. These instructions are generated by default. If you use
10987 @option{-mno-update}, there is a small window between the time that the
10988 stack pointer is updated and the address of the previous frame is
10989 stored, which means code that walks the stack frame across interrupts or
10990 signals may get corrupted data.
10993 @itemx -mno-fused-madd
10994 @opindex mfused-madd
10995 @opindex mno-fused-madd
10996 Generate code that uses (does not use) the floating point multiply and
10997 accumulate instructions. These instructions are generated by default if
10998 hardware floating is used.
11000 @item -mno-bit-align
11002 @opindex mno-bit-align
11003 @opindex mbit-align
11004 On System V.4 and embedded PowerPC systems do not (do) force structures
11005 and unions that contain bit-fields to be aligned to the base type of the
11008 For example, by default a structure containing nothing but 8
11009 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
11010 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
11011 the structure would be aligned to a 1 byte boundary and be one byte in
11014 @item -mno-strict-align
11015 @itemx -mstrict-align
11016 @opindex mno-strict-align
11017 @opindex mstrict-align
11018 On System V.4 and embedded PowerPC systems do not (do) assume that
11019 unaligned memory references will be handled by the system.
11021 @item -mrelocatable
11022 @itemx -mno-relocatable
11023 @opindex mrelocatable
11024 @opindex mno-relocatable
11025 On embedded PowerPC systems generate code that allows (does not allow)
11026 the program to be relocated to a different address at runtime. If you
11027 use @option{-mrelocatable} on any module, all objects linked together must
11028 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
11030 @item -mrelocatable-lib
11031 @itemx -mno-relocatable-lib
11032 @opindex mrelocatable-lib
11033 @opindex mno-relocatable-lib
11034 On embedded PowerPC systems generate code that allows (does not allow)
11035 the program to be relocated to a different address at runtime. Modules
11036 compiled with @option{-mrelocatable-lib} can be linked with either modules
11037 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
11038 with modules compiled with the @option{-mrelocatable} options.
11044 On System V.4 and embedded PowerPC systems do not (do) assume that
11045 register 2 contains a pointer to a global area pointing to the addresses
11046 used in the program.
11049 @itemx -mlittle-endian
11051 @opindex mlittle-endian
11052 On System V.4 and embedded PowerPC systems compile code for the
11053 processor in little endian mode. The @option{-mlittle-endian} option is
11054 the same as @option{-mlittle}.
11057 @itemx -mbig-endian
11059 @opindex mbig-endian
11060 On System V.4 and embedded PowerPC systems compile code for the
11061 processor in big endian mode. The @option{-mbig-endian} option is
11062 the same as @option{-mbig}.
11064 @item -mdynamic-no-pic
11065 @opindex mdynamic-no-pic
11066 On Darwin and Mac OS X systems, compile code so that it is not
11067 relocatable, but that its external references are relocatable. The
11068 resulting code is suitable for applications, but not shared
11071 @item -mprioritize-restricted-insns=@var{priority}
11072 @opindex mprioritize-restricted-insns
11073 This option controls the priority that is assigned to
11074 dispatch-slot restricted instructions during the second scheduling
11075 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
11076 @var{no/highest/second-highest} priority to dispatch slot restricted
11079 @item -msched-costly-dep=@var{dependence_type}
11080 @opindex msched-costly-dep
11081 This option controls which dependences are considered costly
11082 by the target during instruction scheduling. The argument
11083 @var{dependence_type} takes one of the following values:
11084 @var{no}: no dependence is costly,
11085 @var{all}: all dependences are costly,
11086 @var{true_store_to_load}: a true dependence from store to load is costly,
11087 @var{store_to_load}: any dependence from store to load is costly,
11088 @var{number}: any dependence which latency >= @var{number} is costly.
11090 @item -minsert-sched-nops=@var{scheme}
11091 @opindex minsert-sched-nops
11092 This option controls which nop insertion scheme will be used during
11093 the second scheduling pass. The argument @var{scheme} takes one of the
11095 @var{no}: Don't insert nops.
11096 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
11097 according to the scheduler's grouping.
11098 @var{regroup_exact}: Insert nops to force costly dependent insns into
11099 separate groups. Insert exactly as many nops as needed to force an insn
11100 to a new group, according to the estimated processor grouping.
11101 @var{number}: Insert nops to force costly dependent insns into
11102 separate groups. Insert @var{number} nops to force an insn to a new group.
11105 @opindex mcall-sysv
11106 On System V.4 and embedded PowerPC systems compile code using calling
11107 conventions that adheres to the March 1995 draft of the System V
11108 Application Binary Interface, PowerPC processor supplement. This is the
11109 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
11111 @item -mcall-sysv-eabi
11112 @opindex mcall-sysv-eabi
11113 Specify both @option{-mcall-sysv} and @option{-meabi} options.
11115 @item -mcall-sysv-noeabi
11116 @opindex mcall-sysv-noeabi
11117 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
11119 @item -mcall-solaris
11120 @opindex mcall-solaris
11121 On System V.4 and embedded PowerPC systems compile code for the Solaris
11125 @opindex mcall-linux
11126 On System V.4 and embedded PowerPC systems compile code for the
11127 Linux-based GNU system.
11131 On System V.4 and embedded PowerPC systems compile code for the
11132 Hurd-based GNU system.
11134 @item -mcall-netbsd
11135 @opindex mcall-netbsd
11136 On System V.4 and embedded PowerPC systems compile code for the
11137 NetBSD operating system.
11139 @item -maix-struct-return
11140 @opindex maix-struct-return
11141 Return all structures in memory (as specified by the AIX ABI)@.
11143 @item -msvr4-struct-return
11144 @opindex msvr4-struct-return
11145 Return structures smaller than 8 bytes in registers (as specified by the
11148 @item -mabi=@var{abi-type}
11150 Extend the current ABI with a particular extension, or remove such extension.
11151 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
11155 @itemx -mno-prototype
11156 @opindex mprototype
11157 @opindex mno-prototype
11158 On System V.4 and embedded PowerPC systems assume that all calls to
11159 variable argument functions are properly prototyped. Otherwise, the
11160 compiler must insert an instruction before every non prototyped call to
11161 set or clear bit 6 of the condition code register (@var{CR}) to
11162 indicate whether floating point values were passed in the floating point
11163 registers in case the function takes a variable arguments. With
11164 @option{-mprototype}, only calls to prototyped variable argument functions
11165 will set or clear the bit.
11169 On embedded PowerPC systems, assume that the startup module is called
11170 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
11171 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
11176 On embedded PowerPC systems, assume that the startup module is called
11177 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
11182 On embedded PowerPC systems, assume that the startup module is called
11183 @file{crt0.o} and the standard C libraries are @file{libads.a} and
11186 @item -myellowknife
11187 @opindex myellowknife
11188 On embedded PowerPC systems, assume that the startup module is called
11189 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
11194 On System V.4 and embedded PowerPC systems, specify that you are
11195 compiling for a VxWorks system.
11199 Specify that you are compiling for the WindISS simulation environment.
11203 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
11204 header to indicate that @samp{eabi} extended relocations are used.
11210 On System V.4 and embedded PowerPC systems do (do not) adhere to the
11211 Embedded Applications Binary Interface (eabi) which is a set of
11212 modifications to the System V.4 specifications. Selecting @option{-meabi}
11213 means that the stack is aligned to an 8 byte boundary, a function
11214 @code{__eabi} is called to from @code{main} to set up the eabi
11215 environment, and the @option{-msdata} option can use both @code{r2} and
11216 @code{r13} to point to two separate small data areas. Selecting
11217 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
11218 do not call an initialization function from @code{main}, and the
11219 @option{-msdata} option will only use @code{r13} to point to a single
11220 small data area. The @option{-meabi} option is on by default if you
11221 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
11224 @opindex msdata=eabi
11225 On System V.4 and embedded PowerPC systems, put small initialized
11226 @code{const} global and static data in the @samp{.sdata2} section, which
11227 is pointed to by register @code{r2}. Put small initialized
11228 non-@code{const} global and static data in the @samp{.sdata} section,
11229 which is pointed to by register @code{r13}. Put small uninitialized
11230 global and static data in the @samp{.sbss} section, which is adjacent to
11231 the @samp{.sdata} section. The @option{-msdata=eabi} option is
11232 incompatible with the @option{-mrelocatable} option. The
11233 @option{-msdata=eabi} option also sets the @option{-memb} option.
11236 @opindex msdata=sysv
11237 On System V.4 and embedded PowerPC systems, put small global and static
11238 data in the @samp{.sdata} section, which is pointed to by register
11239 @code{r13}. Put small uninitialized global and static data in the
11240 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11241 The @option{-msdata=sysv} option is incompatible with the
11242 @option{-mrelocatable} option.
11244 @item -msdata=default
11246 @opindex msdata=default
11248 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
11249 compile code the same as @option{-msdata=eabi}, otherwise compile code the
11250 same as @option{-msdata=sysv}.
11253 @opindex msdata-data
11254 On System V.4 and embedded PowerPC systems, put small global and static
11255 data in the @samp{.sdata} section. Put small uninitialized global and
11256 static data in the @samp{.sbss} section. Do not use register @code{r13}
11257 to address small data however. This is the default behavior unless
11258 other @option{-msdata} options are used.
11262 @opindex msdata=none
11264 On embedded PowerPC systems, put all initialized global and static data
11265 in the @samp{.data} section, and all uninitialized data in the
11266 @samp{.bss} section.
11270 @cindex smaller data references (PowerPC)
11271 @cindex .sdata/.sdata2 references (PowerPC)
11272 On embedded PowerPC systems, put global and static items less than or
11273 equal to @var{num} bytes into the small data or bss sections instead of
11274 the normal data or bss section. By default, @var{num} is 8. The
11275 @option{-G @var{num}} switch is also passed to the linker.
11276 All modules should be compiled with the same @option{-G @var{num}} value.
11279 @itemx -mno-regnames
11281 @opindex mno-regnames
11282 On System V.4 and embedded PowerPC systems do (do not) emit register
11283 names in the assembly language output using symbolic forms.
11286 @itemx -mno-longcall
11288 @opindex mno-longcall
11289 Default to making all function calls indirectly, using a register, so
11290 that functions which reside further than 32 megabytes (33,554,432
11291 bytes) from the current location can be called. This setting can be
11292 overridden by the @code{shortcall} function attribute, or by
11293 @code{#pragma longcall(0)}.
11295 Some linkers are capable of detecting out-of-range calls and generating
11296 glue code on the fly. On these systems, long calls are unnecessary and
11297 generate slower code. As of this writing, the AIX linker can do this,
11298 as can the GNU linker for PowerPC/64. It is planned to add this feature
11299 to the GNU linker for 32-bit PowerPC systems as well.
11301 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11302 callee, L42'', plus a ``branch island'' (glue code). The two target
11303 addresses represent the callee and the ``branch island''. The
11304 Darwin/PPC linker will prefer the first address and generate a ``bl
11305 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11306 otherwise, the linker will generate ``bl L42'' to call the ``branch
11307 island''. The ``branch island'' is appended to the body of the
11308 calling function; it computes the full 32-bit address of the callee
11311 On Mach-O (Darwin) systems, this option directs the compiler emit to
11312 the glue for every direct call, and the Darwin linker decides whether
11313 to use or discard it.
11315 In the future, we may cause GCC to ignore all longcall specifications
11316 when the linker is known to generate glue.
11320 Adds support for multithreading with the @dfn{pthreads} library.
11321 This option sets flags for both the preprocessor and linker.
11325 @node S/390 and zSeries Options
11326 @subsection S/390 and zSeries Options
11327 @cindex S/390 and zSeries Options
11329 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11333 @itemx -msoft-float
11334 @opindex mhard-float
11335 @opindex msoft-float
11336 Use (do not use) the hardware floating-point instructions and registers
11337 for floating-point operations. When @option{-msoft-float} is specified,
11338 functions in @file{libgcc.a} will be used to perform floating-point
11339 operations. When @option{-mhard-float} is specified, the compiler
11340 generates IEEE floating-point instructions. This is the default.
11343 @itemx -mno-backchain
11344 @opindex mbackchain
11345 @opindex mno-backchain
11346 Store (do not store) the address of the caller's frame as backchain pointer
11347 into the callee's stack frame.
11348 A backchain may be needed to allow debugging using tools that do not understand
11349 DWARF-2 call frame information.
11350 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
11351 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
11352 the backchain is placed into the topmost word of the 96/160 byte register
11355 In general, code compiled with @option{-mbackchain} is call-compatible with
11356 code compiled with @option{-mmo-backchain}; however, use of the backchain
11357 for debugging purposes usually requires that the whole binary is built with
11358 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
11359 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11360 to build a linux kernel use @option{-msoft-float}.
11362 The default is to not maintain the backchain.
11364 @item -mpacked-stack
11365 @item -mno-packed-stack
11366 @opindex mpacked-stack
11367 @opindex mno-packed-stack
11368 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
11369 specified, the compiler uses the all fields of the 96/160 byte register save
11370 area only for their default purpose; unused fields still take up stack space.
11371 When @option{-mpacked-stack} is specified, register save slots are densely
11372 packed at the top of the register save area; unused space is reused for other
11373 purposes, allowing for more efficient use of the available stack space.
11374 However, when @option{-mbackchain} is also in effect, the topmost word of
11375 the save area is always used to store the backchain, and the return address
11376 register is always saved two words below the backchain.
11378 As long as the stack frame backchain is not used, code generated with
11379 @option{-mpacked-stack} is call-compatible with code generated with
11380 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
11381 S/390 or zSeries generated code that uses the stack frame backchain at run
11382 time, not just for debugging purposes. Such code is not call-compatible
11383 with code compiled with @option{-mpacked-stack}. Also, note that the
11384 combination of @option{-mbackchain},
11385 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11386 to build a linux kernel use @option{-msoft-float}.
11388 The default is to not use the packed stack layout.
11391 @itemx -mno-small-exec
11392 @opindex msmall-exec
11393 @opindex mno-small-exec
11394 Generate (or do not generate) code using the @code{bras} instruction
11395 to do subroutine calls.
11396 This only works reliably if the total executable size does not
11397 exceed 64k. The default is to use the @code{basr} instruction instead,
11398 which does not have this limitation.
11404 When @option{-m31} is specified, generate code compliant to the
11405 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
11406 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
11407 particular to generate 64-bit instructions. For the @samp{s390}
11408 targets, the default is @option{-m31}, while the @samp{s390x}
11409 targets default to @option{-m64}.
11415 When @option{-mzarch} is specified, generate code using the
11416 instructions available on z/Architecture.
11417 When @option{-mesa} is specified, generate code using the
11418 instructions available on ESA/390. Note that @option{-mesa} is
11419 not possible with @option{-m64}.
11420 When generating code compliant to the GNU/Linux for S/390 ABI,
11421 the default is @option{-mesa}. When generating code compliant
11422 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
11428 Generate (or do not generate) code using the @code{mvcle} instruction
11429 to perform block moves. When @option{-mno-mvcle} is specified,
11430 use a @code{mvc} loop instead. This is the default.
11436 Print (or do not print) additional debug information when compiling.
11437 The default is to not print debug information.
11439 @item -march=@var{cpu-type}
11441 Generate code that will run on @var{cpu-type}, which is the name of a system
11442 representing a certain processor type. Possible values for
11443 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
11444 When generating code using the instructions available on z/Architecture,
11445 the default is @option{-march=z900}. Otherwise, the default is
11446 @option{-march=g5}.
11448 @item -mtune=@var{cpu-type}
11450 Tune to @var{cpu-type} everything applicable about the generated code,
11451 except for the ABI and the set of available instructions.
11452 The list of @var{cpu-type} values is the same as for @option{-march}.
11453 The default is the value used for @option{-march}.
11456 @itemx -mno-tpf-trace
11457 @opindex mtpf-trace
11458 @opindex mno-tpf-trace
11459 Generate code that adds (does not add) in TPF OS specific branches to trace
11460 routines in the operating system. This option is off by default, even
11461 when compiling for the TPF OS@.
11464 @itemx -mno-fused-madd
11465 @opindex mfused-madd
11466 @opindex mno-fused-madd
11467 Generate code that uses (does not use) the floating point multiply and
11468 accumulate instructions. These instructions are generated by default if
11469 hardware floating point is used.
11471 @item -mwarn-framesize=@var{framesize}
11472 @opindex mwarn-framesize
11473 Emit a warning if the current function exceeds the given frame size. Because
11474 this is a compile time check it doesn't need to be a real problem when the program
11475 runs. It is intended to identify functions which most probably cause
11476 a stack overflow. It is useful to be used in an environment with limited stack
11477 size e.g.@: the linux kernel.
11479 @item -mwarn-dynamicstack
11480 @opindex mwarn-dynamicstack
11481 Emit a warning if the function calls alloca or uses dynamically
11482 sized arrays. This is generally a bad idea with a limited stack size.
11484 @item -mstack-guard=@var{stack-guard}
11485 @item -mstack-size=@var{stack-size}
11486 @opindex mstack-guard
11487 @opindex mstack-size
11488 These arguments always have to be used in conjunction. If they are present the s390
11489 back end emits additional instructions in the function prologue which trigger a trap
11490 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
11491 (remember that the stack on s390 grows downward). These options are intended to
11492 be used to help debugging stack overflow problems. The additionally emitted code
11493 cause only little overhead and hence can also be used in production like systems
11494 without greater performance degradation. The given values have to be exact
11495 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard}.
11496 In order to be efficient the extra code makes the assumption that the stack starts
11497 at an address aligned to the value given by @var{stack-size}.
11501 @subsection SH Options
11503 These @samp{-m} options are defined for the SH implementations:
11508 Generate code for the SH1.
11512 Generate code for the SH2.
11515 Generate code for the SH2e.
11519 Generate code for the SH3.
11523 Generate code for the SH3e.
11527 Generate code for the SH4 without a floating-point unit.
11529 @item -m4-single-only
11530 @opindex m4-single-only
11531 Generate code for the SH4 with a floating-point unit that only
11532 supports single-precision arithmetic.
11536 Generate code for the SH4 assuming the floating-point unit is in
11537 single-precision mode by default.
11541 Generate code for the SH4.
11545 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
11546 floating-point unit is not used.
11548 @item -m4a-single-only
11549 @opindex m4a-single-only
11550 Generate code for the SH4a, in such a way that no double-precision
11551 floating point operations are used.
11554 @opindex m4a-single
11555 Generate code for the SH4a assuming the floating-point unit is in
11556 single-precision mode by default.
11560 Generate code for the SH4a.
11564 Same as @option{-m4a-nofpu}, except that it implicitly passes
11565 @option{-dsp} to the assembler. GCC doesn't generate any DSP
11566 instructions at the moment.
11570 Compile code for the processor in big endian mode.
11574 Compile code for the processor in little endian mode.
11578 Align doubles at 64-bit boundaries. Note that this changes the calling
11579 conventions, and thus some functions from the standard C library will
11580 not work unless you recompile it first with @option{-mdalign}.
11584 Shorten some address references at link time, when possible; uses the
11585 linker option @option{-relax}.
11589 Use 32-bit offsets in @code{switch} tables. The default is to use
11594 Enable the use of the instruction @code{fmovd}.
11598 Comply with the calling conventions defined by Renesas.
11602 Comply with the calling conventions defined by Renesas.
11606 Comply with the calling conventions defined for GCC before the Renesas
11607 conventions were available. This option is the default for all
11608 targets of the SH toolchain except for @samp{sh-symbianelf}.
11611 @opindex mnomacsave
11612 Mark the @code{MAC} register as call-clobbered, even if
11613 @option{-mhitachi} is given.
11617 Increase IEEE-compliance of floating-point code.
11618 At the moment, this is equivalent to @option{-fno-finite-math-only}.
11619 When generating 16 bit SH opcodes, getting IEEE-conforming results for
11620 comparisons of NANs / infinities incurs extra overhead in every
11621 floating point comparison, therefore the default is set to
11622 @option{-ffinite-math-only}.
11626 Dump instruction size and location in the assembly code.
11629 @opindex mpadstruct
11630 This option is deprecated. It pads structures to multiple of 4 bytes,
11631 which is incompatible with the SH ABI@.
11635 Optimize for space instead of speed. Implied by @option{-Os}.
11638 @opindex mprefergot
11639 When generating position-independent code, emit function calls using
11640 the Global Offset Table instead of the Procedure Linkage Table.
11644 Generate a library function call to invalidate instruction cache
11645 entries, after fixing up a trampoline. This library function call
11646 doesn't assume it can write to the whole memory address space. This
11647 is the default when the target is @code{sh-*-linux*}.
11649 @item -multcost=@var{number}
11650 @opindex multcost=@var{number}
11651 Set the cost to assume for a multiply insn.
11653 @item -mdiv=@var{strategy}
11654 @opindex mdiv=@var{strategy}
11655 Set the division strategy to use for SHmedia code. @var{strategy} must be
11656 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
11657 inv:call2, inv:fp .
11658 "fp" performs the operation in floating point. This has a very high latency,
11659 but needs only a few instructions, so it might be a good choice if
11660 your code has enough easily exploitable ILP to allow the compiler to
11661 schedule the floating point instructions together with other instructions.
11662 Division by zero causes a floating point exception.
11663 "inv" uses integer operations to calculate the inverse of the divisor,
11664 and then multiplies the dividend with the inverse. This strategy allows
11665 cse and hoisting of the inverse calculation. Division by zero calculates
11666 an unspecified result, but does not trap.
11667 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
11668 have been found, or if the entire operation has been hoisted to the same
11669 place, the last stages of the inverse calculation are intertwined with the
11670 final multiply to reduce the overall latency, at the expense of using a few
11671 more instructions, and thus offering fewer scheduling opportunities with
11673 "call" calls a library function that usually implements the inv:minlat
11675 This gives high code density for m5-*media-nofpu compilations.
11676 "call2" uses a different entry point of the same library function, where it
11677 assumes that a pointer to a lookup table has already been set up, which
11678 exposes the pointer load to cse / code hoisting optimizations.
11679 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
11680 code generation, but if the code stays unoptimized, revert to the "call",
11681 "call2", or "fp" strategies, respectively. Note that the
11682 potentially-trapping side effect of division by zero is carried by a
11683 separate instruction, so it is possible that all the integer instructions
11684 are hoisted out, but the marker for the side effect stays where it is.
11685 A recombination to fp operations or a call is not possible in that case.
11686 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
11687 that the inverse calculation was nor separated from the multiply, they speed
11688 up division where the dividend fits into 20 bits (plus sign where applicable),
11689 by inserting a test to skip a number of operations in this case; this test
11690 slows down the case of larger dividends. inv20u assumes the case of a such
11691 a small dividend to be unlikely, and inv20l assumes it to be likely.
11693 @item -mdivsi3_libfunc=@var{name}
11694 @opindex mdivsi3_libfunc=@var{name}
11695 Set the name of the library function used for 32 bit signed division to
11696 @var{name}. This only affect the name used in the call and inv:call
11697 division strategies, and the compiler will still expect the same
11698 sets of input/output/clobbered registers as if this option was not present.
11700 @item -madjust-unroll
11701 @opindex madjust-unroll
11702 Throttle unrolling to avoid thrashing target registers.
11703 This option only has an effect if the gcc code base supports the
11704 TARGET_ADJUST_UNROLL_MAX target hook.
11706 @item -mindexed-addressing
11707 @opindex mindexed-addressing
11708 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
11709 This is only safe if the hardware and/or OS implement 32 bit wrap-around
11710 semantics for the indexed addressing mode. The architecture allows the
11711 implementation of processors with 64 bit MMU, which the OS could use to
11712 get 32 bit addressing, but since no current hardware implementation supports
11713 this or any other way to make the indexed addressing mode safe to use in
11714 the 32 bit ABI, the default is -mno-indexed-addressing.
11716 @item -mgettrcost=@var{number}
11717 @opindex mgettrcost=@var{number}
11718 Set the cost assumed for the gettr instruction to @var{number}.
11719 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
11723 Assume pt* instructions won't trap. This will generally generate better
11724 scheduled code, but is unsafe on current hardware. The current architecture
11725 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
11726 This has the unintentional effect of making it unsafe to schedule ptabs /
11727 ptrel before a branch, or hoist it out of a loop. For example,
11728 __do_global_ctors, a part of libgcc that runs constructors at program
11729 startup, calls functions in a list which is delimited by -1. With the
11730 -mpt-fixed option, the ptabs will be done before testing against -1.
11731 That means that all the constructors will be run a bit quicker, but when
11732 the loop comes to the end of the list, the program crashes because ptabs
11733 loads -1 into a target register. Since this option is unsafe for any
11734 hardware implementing the current architecture specification, the default
11735 is -mno-pt-fixed. Unless the user specifies a specific cost with
11736 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
11737 this deters register allocation using target registers for storing
11740 @item -minvalid-symbols
11741 @opindex minvalid-symbols
11742 Assume symbols might be invalid. Ordinary function symbols generated by
11743 the compiler will always be valid to load with movi/shori/ptabs or
11744 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
11745 to generate symbols that will cause ptabs / ptrel to trap.
11746 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
11747 It will then prevent cross-basic-block cse, hoisting and most scheduling
11748 of symbol loads. The default is @option{-mno-invalid-symbols}.
11751 @node SPARC Options
11752 @subsection SPARC Options
11753 @cindex SPARC options
11755 These @samp{-m} options are supported on the SPARC:
11758 @item -mno-app-regs
11760 @opindex mno-app-regs
11762 Specify @option{-mapp-regs} to generate output using the global registers
11763 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
11766 To be fully SVR4 ABI compliant at the cost of some performance loss,
11767 specify @option{-mno-app-regs}. You should compile libraries and system
11768 software with this option.
11771 @itemx -mhard-float
11773 @opindex mhard-float
11774 Generate output containing floating point instructions. This is the
11778 @itemx -msoft-float
11780 @opindex msoft-float
11781 Generate output containing library calls for floating point.
11782 @strong{Warning:} the requisite libraries are not available for all SPARC
11783 targets. Normally the facilities of the machine's usual C compiler are
11784 used, but this cannot be done directly in cross-compilation. You must make
11785 your own arrangements to provide suitable library functions for
11786 cross-compilation. The embedded targets @samp{sparc-*-aout} and
11787 @samp{sparclite-*-*} do provide software floating point support.
11789 @option{-msoft-float} changes the calling convention in the output file;
11790 therefore, it is only useful if you compile @emph{all} of a program with
11791 this option. In particular, you need to compile @file{libgcc.a}, the
11792 library that comes with GCC, with @option{-msoft-float} in order for
11795 @item -mhard-quad-float
11796 @opindex mhard-quad-float
11797 Generate output containing quad-word (long double) floating point
11800 @item -msoft-quad-float
11801 @opindex msoft-quad-float
11802 Generate output containing library calls for quad-word (long double)
11803 floating point instructions. The functions called are those specified
11804 in the SPARC ABI@. This is the default.
11806 As of this writing, there are no SPARC implementations that have hardware
11807 support for the quad-word floating point instructions. They all invoke
11808 a trap handler for one of these instructions, and then the trap handler
11809 emulates the effect of the instruction. Because of the trap handler overhead,
11810 this is much slower than calling the ABI library routines. Thus the
11811 @option{-msoft-quad-float} option is the default.
11813 @item -mno-unaligned-doubles
11814 @itemx -munaligned-doubles
11815 @opindex mno-unaligned-doubles
11816 @opindex munaligned-doubles
11817 Assume that doubles have 8 byte alignment. This is the default.
11819 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
11820 alignment only if they are contained in another type, or if they have an
11821 absolute address. Otherwise, it assumes they have 4 byte alignment.
11822 Specifying this option avoids some rare compatibility problems with code
11823 generated by other compilers. It is not the default because it results
11824 in a performance loss, especially for floating point code.
11826 @item -mno-faster-structs
11827 @itemx -mfaster-structs
11828 @opindex mno-faster-structs
11829 @opindex mfaster-structs
11830 With @option{-mfaster-structs}, the compiler assumes that structures
11831 should have 8 byte alignment. This enables the use of pairs of
11832 @code{ldd} and @code{std} instructions for copies in structure
11833 assignment, in place of twice as many @code{ld} and @code{st} pairs.
11834 However, the use of this changed alignment directly violates the SPARC
11835 ABI@. Thus, it's intended only for use on targets where the developer
11836 acknowledges that their resulting code will not be directly in line with
11837 the rules of the ABI@.
11839 @item -mimpure-text
11840 @opindex mimpure-text
11841 @option{-mimpure-text}, used in addition to @option{-shared}, tells
11842 the compiler to not pass @option{-z text} to the linker when linking a
11843 shared object. Using this option, you can link position-dependent
11844 code into a shared object.
11846 @option{-mimpure-text} suppresses the ``relocations remain against
11847 allocatable but non-writable sections'' linker error message.
11848 However, the necessary relocations will trigger copy-on-write, and the
11849 shared object is not actually shared across processes. Instead of
11850 using @option{-mimpure-text}, you should compile all source code with
11851 @option{-fpic} or @option{-fPIC}.
11853 This option is only available on SunOS and Solaris.
11855 @item -mcpu=@var{cpu_type}
11857 Set the instruction set, register set, and instruction scheduling parameters
11858 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
11859 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
11860 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
11861 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
11862 @samp{ultrasparc3}.
11864 Default instruction scheduling parameters are used for values that select
11865 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
11866 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
11868 Here is a list of each supported architecture and their supported
11873 v8: supersparc, hypersparc
11874 sparclite: f930, f934, sparclite86x
11876 v9: ultrasparc, ultrasparc3
11879 By default (unless configured otherwise), GCC generates code for the V7
11880 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
11881 additionally optimizes it for the Cypress CY7C602 chip, as used in the
11882 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
11883 SPARCStation 1, 2, IPX etc.
11885 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
11886 architecture. The only difference from V7 code is that the compiler emits
11887 the integer multiply and integer divide instructions which exist in SPARC-V8
11888 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
11889 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
11892 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
11893 the SPARC architecture. This adds the integer multiply, integer divide step
11894 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
11895 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
11896 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
11897 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
11898 MB86934 chip, which is the more recent SPARClite with FPU@.
11900 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
11901 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
11902 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
11903 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
11904 optimizes it for the TEMIC SPARClet chip.
11906 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
11907 architecture. This adds 64-bit integer and floating-point move instructions,
11908 3 additional floating-point condition code registers and conditional move
11909 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
11910 optimizes it for the Sun UltraSPARC I/II chips. With
11911 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
11912 Sun UltraSPARC III chip.
11914 @item -mtune=@var{cpu_type}
11916 Set the instruction scheduling parameters for machine type
11917 @var{cpu_type}, but do not set the instruction set or register set that the
11918 option @option{-mcpu=@var{cpu_type}} would.
11920 The same values for @option{-mcpu=@var{cpu_type}} can be used for
11921 @option{-mtune=@var{cpu_type}}, but the only useful values are those
11922 that select a particular cpu implementation. Those are @samp{cypress},
11923 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
11924 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
11925 @samp{ultrasparc3}.
11930 @opindex mno-v8plus
11931 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
11932 difference from the V8 ABI is that the global and out registers are
11933 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
11934 mode for all SPARC-V9 processors.
11940 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
11941 Visual Instruction Set extensions. The default is @option{-mno-vis}.
11944 These @samp{-m} options are supported in addition to the above
11945 on SPARC-V9 processors in 64-bit environments:
11948 @item -mlittle-endian
11949 @opindex mlittle-endian
11950 Generate code for a processor running in little-endian mode. It is only
11951 available for a few configurations and most notably not on Solaris and Linux.
11957 Generate code for a 32-bit or 64-bit environment.
11958 The 32-bit environment sets int, long and pointer to 32 bits.
11959 The 64-bit environment sets int to 32 bits and long and pointer
11962 @item -mcmodel=medlow
11963 @opindex mcmodel=medlow
11964 Generate code for the Medium/Low code model: 64-bit addresses, programs
11965 must be linked in the low 32 bits of memory. Programs can be statically
11966 or dynamically linked.
11968 @item -mcmodel=medmid
11969 @opindex mcmodel=medmid
11970 Generate code for the Medium/Middle code model: 64-bit addresses, programs
11971 must be linked in the low 44 bits of memory, the text and data segments must
11972 be less than 2GB in size and the data segment must be located within 2GB of
11975 @item -mcmodel=medany
11976 @opindex mcmodel=medany
11977 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
11978 may be linked anywhere in memory, the text and data segments must be less
11979 than 2GB in size and the data segment must be located within 2GB of the
11982 @item -mcmodel=embmedany
11983 @opindex mcmodel=embmedany
11984 Generate code for the Medium/Anywhere code model for embedded systems:
11985 64-bit addresses, the text and data segments must be less than 2GB in
11986 size, both starting anywhere in memory (determined at link time). The
11987 global register %g4 points to the base of the data segment. Programs
11988 are statically linked and PIC is not supported.
11991 @itemx -mno-stack-bias
11992 @opindex mstack-bias
11993 @opindex mno-stack-bias
11994 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
11995 frame pointer if present, are offset by @minus{}2047 which must be added back
11996 when making stack frame references. This is the default in 64-bit mode.
11997 Otherwise, assume no such offset is present.
12000 These switches are supported in addition to the above on Solaris:
12005 Add support for multithreading using the Solaris threads library. This
12006 option sets flags for both the preprocessor and linker. This option does
12007 not affect the thread safety of object code produced by the compiler or
12008 that of libraries supplied with it.
12012 Add support for multithreading using the POSIX threads library. This
12013 option sets flags for both the preprocessor and linker. This option does
12014 not affect the thread safety of object code produced by the compiler or
12015 that of libraries supplied with it.
12018 @node System V Options
12019 @subsection Options for System V
12021 These additional options are available on System V Release 4 for
12022 compatibility with other compilers on those systems:
12027 Create a shared object.
12028 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
12032 Identify the versions of each tool used by the compiler, in a
12033 @code{.ident} assembler directive in the output.
12037 Refrain from adding @code{.ident} directives to the output file (this is
12040 @item -YP,@var{dirs}
12042 Search the directories @var{dirs}, and no others, for libraries
12043 specified with @option{-l}.
12045 @item -Ym,@var{dir}
12047 Look in the directory @var{dir} to find the M4 preprocessor.
12048 The assembler uses this option.
12049 @c This is supposed to go with a -Yd for predefined M4 macro files, but
12050 @c the generic assembler that comes with Solaris takes just -Ym.
12053 @node TMS320C3x/C4x Options
12054 @subsection TMS320C3x/C4x Options
12055 @cindex TMS320C3x/C4x Options
12057 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
12061 @item -mcpu=@var{cpu_type}
12063 Set the instruction set, register set, and instruction scheduling
12064 parameters for machine type @var{cpu_type}. Supported values for
12065 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
12066 @samp{c44}. The default is @samp{c40} to generate code for the
12071 @itemx -msmall-memory
12073 @opindex mbig-memory
12075 @opindex msmall-memory
12077 Generates code for the big or small memory model. The small memory
12078 model assumed that all data fits into one 64K word page. At run-time
12079 the data page (DP) register must be set to point to the 64K page
12080 containing the .bss and .data program sections. The big memory model is
12081 the default and requires reloading of the DP register for every direct
12088 Allow (disallow) allocation of general integer operands into the block
12089 count register BK@.
12095 Enable (disable) generation of code using decrement and branch,
12096 DBcond(D), instructions. This is enabled by default for the C4x. To be
12097 on the safe side, this is disabled for the C3x, since the maximum
12098 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
12099 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
12100 that it can utilize the decrement and branch instruction, but will give
12101 up if there is more than one memory reference in the loop. Thus a loop
12102 where the loop counter is decremented can generate slightly more
12103 efficient code, in cases where the RPTB instruction cannot be utilized.
12105 @item -mdp-isr-reload
12107 @opindex mdp-isr-reload
12109 Force the DP register to be saved on entry to an interrupt service
12110 routine (ISR), reloaded to point to the data section, and restored on
12111 exit from the ISR@. This should not be required unless someone has
12112 violated the small memory model by modifying the DP register, say within
12119 For the C3x use the 24-bit MPYI instruction for integer multiplies
12120 instead of a library call to guarantee 32-bit results. Note that if one
12121 of the operands is a constant, then the multiplication will be performed
12122 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
12123 then squaring operations are performed inline instead of a library call.
12126 @itemx -mno-fast-fix
12128 @opindex mno-fast-fix
12129 The C3x/C4x FIX instruction to convert a floating point value to an
12130 integer value chooses the nearest integer less than or equal to the
12131 floating point value rather than to the nearest integer. Thus if the
12132 floating point number is negative, the result will be incorrectly
12133 truncated an additional code is necessary to detect and correct this
12134 case. This option can be used to disable generation of the additional
12135 code required to correct the result.
12141 Enable (disable) generation of repeat block sequences using the RPTB
12142 instruction for zero overhead looping. The RPTB construct is only used
12143 for innermost loops that do not call functions or jump across the loop
12144 boundaries. There is no advantage having nested RPTB loops due to the
12145 overhead required to save and restore the RC, RS, and RE registers.
12146 This is enabled by default with @option{-O2}.
12148 @item -mrpts=@var{count}
12152 Enable (disable) the use of the single instruction repeat instruction
12153 RPTS@. If a repeat block contains a single instruction, and the loop
12154 count can be guaranteed to be less than the value @var{count}, GCC will
12155 emit a RPTS instruction instead of a RPTB@. If no value is specified,
12156 then a RPTS will be emitted even if the loop count cannot be determined
12157 at compile time. Note that the repeated instruction following RPTS does
12158 not have to be reloaded from memory each iteration, thus freeing up the
12159 CPU buses for operands. However, since interrupts are blocked by this
12160 instruction, it is disabled by default.
12162 @item -mloop-unsigned
12163 @itemx -mno-loop-unsigned
12164 @opindex mloop-unsigned
12165 @opindex mno-loop-unsigned
12166 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
12167 is @math{2^{31} + 1} since these instructions test if the iteration count is
12168 negative to terminate the loop. If the iteration count is unsigned
12169 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
12170 exceeded. This switch allows an unsigned iteration count.
12174 Try to emit an assembler syntax that the TI assembler (asm30) is happy
12175 with. This also enforces compatibility with the API employed by the TI
12176 C3x C compiler. For example, long doubles are passed as structures
12177 rather than in floating point registers.
12183 Generate code that uses registers (stack) for passing arguments to functions.
12184 By default, arguments are passed in registers where possible rather
12185 than by pushing arguments on to the stack.
12187 @item -mparallel-insns
12188 @itemx -mno-parallel-insns
12189 @opindex mparallel-insns
12190 @opindex mno-parallel-insns
12191 Allow the generation of parallel instructions. This is enabled by
12192 default with @option{-O2}.
12194 @item -mparallel-mpy
12195 @itemx -mno-parallel-mpy
12196 @opindex mparallel-mpy
12197 @opindex mno-parallel-mpy
12198 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
12199 provided @option{-mparallel-insns} is also specified. These instructions have
12200 tight register constraints which can pessimize the code generation
12201 of large functions.
12206 @subsection V850 Options
12207 @cindex V850 Options
12209 These @samp{-m} options are defined for V850 implementations:
12213 @itemx -mno-long-calls
12214 @opindex mlong-calls
12215 @opindex mno-long-calls
12216 Treat all calls as being far away (near). If calls are assumed to be
12217 far away, the compiler will always load the functions address up into a
12218 register, and call indirect through the pointer.
12224 Do not optimize (do optimize) basic blocks that use the same index
12225 pointer 4 or more times to copy pointer into the @code{ep} register, and
12226 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
12227 option is on by default if you optimize.
12229 @item -mno-prolog-function
12230 @itemx -mprolog-function
12231 @opindex mno-prolog-function
12232 @opindex mprolog-function
12233 Do not use (do use) external functions to save and restore registers
12234 at the prologue and epilogue of a function. The external functions
12235 are slower, but use less code space if more than one function saves
12236 the same number of registers. The @option{-mprolog-function} option
12237 is on by default if you optimize.
12241 Try to make the code as small as possible. At present, this just turns
12242 on the @option{-mep} and @option{-mprolog-function} options.
12244 @item -mtda=@var{n}
12246 Put static or global variables whose size is @var{n} bytes or less into
12247 the tiny data area that register @code{ep} points to. The tiny data
12248 area can hold up to 256 bytes in total (128 bytes for byte references).
12250 @item -msda=@var{n}
12252 Put static or global variables whose size is @var{n} bytes or less into
12253 the small data area that register @code{gp} points to. The small data
12254 area can hold up to 64 kilobytes.
12256 @item -mzda=@var{n}
12258 Put static or global variables whose size is @var{n} bytes or less into
12259 the first 32 kilobytes of memory.
12263 Specify that the target processor is the V850.
12266 @opindex mbig-switch
12267 Generate code suitable for big switch tables. Use this option only if
12268 the assembler/linker complain about out of range branches within a switch
12273 This option will cause r2 and r5 to be used in the code generated by
12274 the compiler. This setting is the default.
12276 @item -mno-app-regs
12277 @opindex mno-app-regs
12278 This option will cause r2 and r5 to be treated as fixed registers.
12282 Specify that the target processor is the V850E1. The preprocessor
12283 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
12284 this option is used.
12288 Specify that the target processor is the V850E@. The preprocessor
12289 constant @samp{__v850e__} will be defined if this option is used.
12291 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
12292 are defined then a default target processor will be chosen and the
12293 relevant @samp{__v850*__} preprocessor constant will be defined.
12295 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
12296 defined, regardless of which processor variant is the target.
12298 @item -mdisable-callt
12299 @opindex mdisable-callt
12300 This option will suppress generation of the CALLT instruction for the
12301 v850e and v850e1 flavors of the v850 architecture. The default is
12302 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
12307 @subsection VAX Options
12308 @cindex VAX options
12310 These @samp{-m} options are defined for the VAX:
12315 Do not output certain jump instructions (@code{aobleq} and so on)
12316 that the Unix assembler for the VAX cannot handle across long
12321 Do output those jump instructions, on the assumption that you
12322 will assemble with the GNU assembler.
12326 Output code for g-format floating point numbers instead of d-format.
12329 @node x86-64 Options
12330 @subsection x86-64 Options
12331 @cindex x86-64 options
12333 These are listed under @xref{i386 and x86-64 Options}.
12335 @node Xstormy16 Options
12336 @subsection Xstormy16 Options
12337 @cindex Xstormy16 Options
12339 These options are defined for Xstormy16:
12344 Choose startup files and linker script suitable for the simulator.
12347 @node Xtensa Options
12348 @subsection Xtensa Options
12349 @cindex Xtensa Options
12351 These options are supported for Xtensa targets:
12355 @itemx -mno-const16
12357 @opindex mno-const16
12358 Enable or disable use of @code{CONST16} instructions for loading
12359 constant values. The @code{CONST16} instruction is currently not a
12360 standard option from Tensilica. When enabled, @code{CONST16}
12361 instructions are always used in place of the standard @code{L32R}
12362 instructions. The use of @code{CONST16} is enabled by default only if
12363 the @code{L32R} instruction is not available.
12366 @itemx -mno-fused-madd
12367 @opindex mfused-madd
12368 @opindex mno-fused-madd
12369 Enable or disable use of fused multiply/add and multiply/subtract
12370 instructions in the floating-point option. This has no effect if the
12371 floating-point option is not also enabled. Disabling fused multiply/add
12372 and multiply/subtract instructions forces the compiler to use separate
12373 instructions for the multiply and add/subtract operations. This may be
12374 desirable in some cases where strict IEEE 754-compliant results are
12375 required: the fused multiply add/subtract instructions do not round the
12376 intermediate result, thereby producing results with @emph{more} bits of
12377 precision than specified by the IEEE standard. Disabling fused multiply
12378 add/subtract instructions also ensures that the program output is not
12379 sensitive to the compiler's ability to combine multiply and add/subtract
12382 @item -mtext-section-literals
12383 @itemx -mno-text-section-literals
12384 @opindex mtext-section-literals
12385 @opindex mno-text-section-literals
12386 Control the treatment of literal pools. The default is
12387 @option{-mno-text-section-literals}, which places literals in a separate
12388 section in the output file. This allows the literal pool to be placed
12389 in a data RAM/ROM, and it also allows the linker to combine literal
12390 pools from separate object files to remove redundant literals and
12391 improve code size. With @option{-mtext-section-literals}, the literals
12392 are interspersed in the text section in order to keep them as close as
12393 possible to their references. This may be necessary for large assembly
12396 @item -mtarget-align
12397 @itemx -mno-target-align
12398 @opindex mtarget-align
12399 @opindex mno-target-align
12400 When this option is enabled, GCC instructs the assembler to
12401 automatically align instructions to reduce branch penalties at the
12402 expense of some code density. The assembler attempts to widen density
12403 instructions to align branch targets and the instructions following call
12404 instructions. If there are not enough preceding safe density
12405 instructions to align a target, no widening will be performed. The
12406 default is @option{-mtarget-align}. These options do not affect the
12407 treatment of auto-aligned instructions like @code{LOOP}, which the
12408 assembler will always align, either by widening density instructions or
12409 by inserting no-op instructions.
12412 @itemx -mno-longcalls
12413 @opindex mlongcalls
12414 @opindex mno-longcalls
12415 When this option is enabled, GCC instructs the assembler to translate
12416 direct calls to indirect calls unless it can determine that the target
12417 of a direct call is in the range allowed by the call instruction. This
12418 translation typically occurs for calls to functions in other source
12419 files. Specifically, the assembler translates a direct @code{CALL}
12420 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
12421 The default is @option{-mno-longcalls}. This option should be used in
12422 programs where the call target can potentially be out of range. This
12423 option is implemented in the assembler, not the compiler, so the
12424 assembly code generated by GCC will still show direct call
12425 instructions---look at the disassembled object code to see the actual
12426 instructions. Note that the assembler will use an indirect call for
12427 every cross-file call, not just those that really will be out of range.
12430 @node zSeries Options
12431 @subsection zSeries Options
12432 @cindex zSeries options
12434 These are listed under @xref{S/390 and zSeries Options}.
12436 @node Code Gen Options
12437 @section Options for Code Generation Conventions
12438 @cindex code generation conventions
12439 @cindex options, code generation
12440 @cindex run-time options
12442 These machine-independent options control the interface conventions
12443 used in code generation.
12445 Most of them have both positive and negative forms; the negative form
12446 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
12447 one of the forms is listed---the one which is not the default. You
12448 can figure out the other form by either removing @samp{no-} or adding
12452 @item -fbounds-check
12453 @opindex fbounds-check
12454 For front-ends that support it, generate additional code to check that
12455 indices used to access arrays are within the declared range. This is
12456 currently only supported by the Java and Fortran 77 front-ends, where
12457 this option defaults to true and false respectively.
12461 This option generates traps for signed overflow on addition, subtraction,
12462 multiplication operations.
12466 This option instructs the compiler to assume that signed arithmetic
12467 overflow of addition, subtraction and multiplication wraps around
12468 using twos-complement representation. This flag enables some optimizations
12469 and disables other. This option is enabled by default for the Java
12470 front-end, as required by the Java language specification.
12473 @opindex fexceptions
12474 Enable exception handling. Generates extra code needed to propagate
12475 exceptions. For some targets, this implies GCC will generate frame
12476 unwind information for all functions, which can produce significant data
12477 size overhead, although it does not affect execution. If you do not
12478 specify this option, GCC will enable it by default for languages like
12479 C++ which normally require exception handling, and disable it for
12480 languages like C that do not normally require it. However, you may need
12481 to enable this option when compiling C code that needs to interoperate
12482 properly with exception handlers written in C++. You may also wish to
12483 disable this option if you are compiling older C++ programs that don't
12484 use exception handling.
12486 @item -fnon-call-exceptions
12487 @opindex fnon-call-exceptions
12488 Generate code that allows trapping instructions to throw exceptions.
12489 Note that this requires platform-specific runtime support that does
12490 not exist everywhere. Moreover, it only allows @emph{trapping}
12491 instructions to throw exceptions, i.e.@: memory references or floating
12492 point instructions. It does not allow exceptions to be thrown from
12493 arbitrary signal handlers such as @code{SIGALRM}.
12495 @item -funwind-tables
12496 @opindex funwind-tables
12497 Similar to @option{-fexceptions}, except that it will just generate any needed
12498 static data, but will not affect the generated code in any other way.
12499 You will normally not enable this option; instead, a language processor
12500 that needs this handling would enable it on your behalf.
12502 @item -fasynchronous-unwind-tables
12503 @opindex fasynchronous-unwind-tables
12504 Generate unwind table in dwarf2 format, if supported by target machine. The
12505 table is exact at each instruction boundary, so it can be used for stack
12506 unwinding from asynchronous events (such as debugger or garbage collector).
12508 @item -fpcc-struct-return
12509 @opindex fpcc-struct-return
12510 Return ``short'' @code{struct} and @code{union} values in memory like
12511 longer ones, rather than in registers. This convention is less
12512 efficient, but it has the advantage of allowing intercallability between
12513 GCC-compiled files and files compiled with other compilers, particularly
12514 the Portable C Compiler (pcc).
12516 The precise convention for returning structures in memory depends
12517 on the target configuration macros.
12519 Short structures and unions are those whose size and alignment match
12520 that of some integer type.
12522 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12523 switch is not binary compatible with code compiled with the
12524 @option{-freg-struct-return} switch.
12525 Use it to conform to a non-default application binary interface.
12527 @item -freg-struct-return
12528 @opindex freg-struct-return
12529 Return @code{struct} and @code{union} values in registers when possible.
12530 This is more efficient for small structures than
12531 @option{-fpcc-struct-return}.
12533 If you specify neither @option{-fpcc-struct-return} nor
12534 @option{-freg-struct-return}, GCC defaults to whichever convention is
12535 standard for the target. If there is no standard convention, GCC
12536 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12537 the principal compiler. In those cases, we can choose the standard, and
12538 we chose the more efficient register return alternative.
12540 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12541 switch is not binary compatible with code compiled with the
12542 @option{-fpcc-struct-return} switch.
12543 Use it to conform to a non-default application binary interface.
12545 @item -fshort-enums
12546 @opindex fshort-enums
12547 Allocate to an @code{enum} type only as many bytes as it needs for the
12548 declared range of possible values. Specifically, the @code{enum} type
12549 will be equivalent to the smallest integer type which has enough room.
12551 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12552 code that is not binary compatible with code generated without that switch.
12553 Use it to conform to a non-default application binary interface.
12555 @item -fshort-double
12556 @opindex fshort-double
12557 Use the same size for @code{double} as for @code{float}.
12559 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
12560 code that is not binary compatible with code generated without that switch.
12561 Use it to conform to a non-default application binary interface.
12563 @item -fshort-wchar
12564 @opindex fshort-wchar
12565 Override the underlying type for @samp{wchar_t} to be @samp{short
12566 unsigned int} instead of the default for the target. This option is
12567 useful for building programs to run under WINE@.
12569 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12570 code that is not binary compatible with code generated without that switch.
12571 Use it to conform to a non-default application binary interface.
12573 @item -fshared-data
12574 @opindex fshared-data
12575 Requests that the data and non-@code{const} variables of this
12576 compilation be shared data rather than private data. The distinction
12577 makes sense only on certain operating systems, where shared data is
12578 shared between processes running the same program, while private data
12579 exists in one copy per process.
12582 @opindex fno-common
12583 In C, allocate even uninitialized global variables in the data section of the
12584 object file, rather than generating them as common blocks. This has the
12585 effect that if the same variable is declared (without @code{extern}) in
12586 two different compilations, you will get an error when you link them.
12587 The only reason this might be useful is if you wish to verify that the
12588 program will work on other systems which always work this way.
12592 Ignore the @samp{#ident} directive.
12594 @item -finhibit-size-directive
12595 @opindex finhibit-size-directive
12596 Don't output a @code{.size} assembler directive, or anything else that
12597 would cause trouble if the function is split in the middle, and the
12598 two halves are placed at locations far apart in memory. This option is
12599 used when compiling @file{crtstuff.c}; you should not need to use it
12602 @item -fverbose-asm
12603 @opindex fverbose-asm
12604 Put extra commentary information in the generated assembly code to
12605 make it more readable. This option is generally only of use to those
12606 who actually need to read the generated assembly code (perhaps while
12607 debugging the compiler itself).
12609 @option{-fno-verbose-asm}, the default, causes the
12610 extra information to be omitted and is useful when comparing two assembler
12615 @cindex global offset table
12617 Generate position-independent code (PIC) suitable for use in a shared
12618 library, if supported for the target machine. Such code accesses all
12619 constant addresses through a global offset table (GOT)@. The dynamic
12620 loader resolves the GOT entries when the program starts (the dynamic
12621 loader is not part of GCC; it is part of the operating system). If
12622 the GOT size for the linked executable exceeds a machine-specific
12623 maximum size, you get an error message from the linker indicating that
12624 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12625 instead. (These maximums are 8k on the SPARC and 32k
12626 on the m68k and RS/6000. The 386 has no such limit.)
12628 Position-independent code requires special support, and therefore works
12629 only on certain machines. For the 386, GCC supports PIC for System V
12630 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12631 position-independent.
12635 If supported for the target machine, emit position-independent code,
12636 suitable for dynamic linking and avoiding any limit on the size of the
12637 global offset table. This option makes a difference on the m68k,
12638 PowerPC and SPARC@.
12640 Position-independent code requires special support, and therefore works
12641 only on certain machines.
12647 These options are similar to @option{-fpic} and @option{-fPIC}, but
12648 generated position independent code can be only linked into executables.
12649 Usually these options are used when @option{-pie} GCC option will be
12650 used during linking.
12652 @item -fno-jump-tables
12653 @opindex fno-jump-tables
12654 Do not use jump tables for switch statements even where it would be
12655 more efficient than other code generation strategies. This option is
12656 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12657 building code which forms part of a dynamic linker and cannot
12658 reference the address of a jump table. On some targets, jump tables
12659 do not require a GOT and this option is not needed.
12661 @item -ffixed-@var{reg}
12663 Treat the register named @var{reg} as a fixed register; generated code
12664 should never refer to it (except perhaps as a stack pointer, frame
12665 pointer or in some other fixed role).
12667 @var{reg} must be the name of a register. The register names accepted
12668 are machine-specific and are defined in the @code{REGISTER_NAMES}
12669 macro in the machine description macro file.
12671 This flag does not have a negative form, because it specifies a
12674 @item -fcall-used-@var{reg}
12675 @opindex fcall-used
12676 Treat the register named @var{reg} as an allocable register that is
12677 clobbered by function calls. It may be allocated for temporaries or
12678 variables that do not live across a call. Functions compiled this way
12679 will not save and restore the register @var{reg}.
12681 It is an error to used this flag with the frame pointer or stack pointer.
12682 Use of this flag for other registers that have fixed pervasive roles in
12683 the machine's execution model will produce disastrous results.
12685 This flag does not have a negative form, because it specifies a
12688 @item -fcall-saved-@var{reg}
12689 @opindex fcall-saved
12690 Treat the register named @var{reg} as an allocable register saved by
12691 functions. It may be allocated even for temporaries or variables that
12692 live across a call. Functions compiled this way will save and restore
12693 the register @var{reg} if they use it.
12695 It is an error to used this flag with the frame pointer or stack pointer.
12696 Use of this flag for other registers that have fixed pervasive roles in
12697 the machine's execution model will produce disastrous results.
12699 A different sort of disaster will result from the use of this flag for
12700 a register in which function values may be returned.
12702 This flag does not have a negative form, because it specifies a
12705 @item -fpack-struct[=@var{n}]
12706 @opindex fpack-struct
12707 Without a value specified, pack all structure members together without
12708 holes. When a value is specified (which must be a small power of two), pack
12709 structure members according to this value, representing the maximum
12710 alignment (that is, objects with default alignment requirements larger than
12711 this will be output potentially unaligned at the next fitting location.
12713 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12714 code that is not binary compatible with code generated without that switch.
12715 Additionally, it makes the code suboptimal.
12716 Use it to conform to a non-default application binary interface.
12718 @item -finstrument-functions
12719 @opindex finstrument-functions
12720 Generate instrumentation calls for entry and exit to functions. Just
12721 after function entry and just before function exit, the following
12722 profiling functions will be called with the address of the current
12723 function and its call site. (On some platforms,
12724 @code{__builtin_return_address} does not work beyond the current
12725 function, so the call site information may not be available to the
12726 profiling functions otherwise.)
12729 void __cyg_profile_func_enter (void *this_fn,
12731 void __cyg_profile_func_exit (void *this_fn,
12735 The first argument is the address of the start of the current function,
12736 which may be looked up exactly in the symbol table.
12738 This instrumentation is also done for functions expanded inline in other
12739 functions. The profiling calls will indicate where, conceptually, the
12740 inline function is entered and exited. This means that addressable
12741 versions of such functions must be available. If all your uses of a
12742 function are expanded inline, this may mean an additional expansion of
12743 code size. If you use @samp{extern inline} in your C code, an
12744 addressable version of such functions must be provided. (This is
12745 normally the case anyways, but if you get lucky and the optimizer always
12746 expands the functions inline, you might have gotten away without
12747 providing static copies.)
12749 A function may be given the attribute @code{no_instrument_function}, in
12750 which case this instrumentation will not be done. This can be used, for
12751 example, for the profiling functions listed above, high-priority
12752 interrupt routines, and any functions from which the profiling functions
12753 cannot safely be called (perhaps signal handlers, if the profiling
12754 routines generate output or allocate memory).
12756 @item -fstack-check
12757 @opindex fstack-check
12758 Generate code to verify that you do not go beyond the boundary of the
12759 stack. You should specify this flag if you are running in an
12760 environment with multiple threads, but only rarely need to specify it in
12761 a single-threaded environment since stack overflow is automatically
12762 detected on nearly all systems if there is only one stack.
12764 Note that this switch does not actually cause checking to be done; the
12765 operating system must do that. The switch causes generation of code
12766 to ensure that the operating system sees the stack being extended.
12768 @item -fstack-limit-register=@var{reg}
12769 @itemx -fstack-limit-symbol=@var{sym}
12770 @itemx -fno-stack-limit
12771 @opindex fstack-limit-register
12772 @opindex fstack-limit-symbol
12773 @opindex fno-stack-limit
12774 Generate code to ensure that the stack does not grow beyond a certain value,
12775 either the value of a register or the address of a symbol. If the stack
12776 would grow beyond the value, a signal is raised. For most targets,
12777 the signal is raised before the stack overruns the boundary, so
12778 it is possible to catch the signal without taking special precautions.
12780 For instance, if the stack starts at absolute address @samp{0x80000000}
12781 and grows downwards, you can use the flags
12782 @option{-fstack-limit-symbol=__stack_limit} and
12783 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12784 of 128KB@. Note that this may only work with the GNU linker.
12786 @cindex aliasing of parameters
12787 @cindex parameters, aliased
12788 @item -fargument-alias
12789 @itemx -fargument-noalias
12790 @itemx -fargument-noalias-global
12791 @opindex fargument-alias
12792 @opindex fargument-noalias
12793 @opindex fargument-noalias-global
12794 Specify the possible relationships among parameters and between
12795 parameters and global data.
12797 @option{-fargument-alias} specifies that arguments (parameters) may
12798 alias each other and may alias global storage.@*
12799 @option{-fargument-noalias} specifies that arguments do not alias
12800 each other, but may alias global storage.@*
12801 @option{-fargument-noalias-global} specifies that arguments do not
12802 alias each other and do not alias global storage.
12804 Each language will automatically use whatever option is required by
12805 the language standard. You should not need to use these options yourself.
12807 @item -fleading-underscore
12808 @opindex fleading-underscore
12809 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12810 change the way C symbols are represented in the object file. One use
12811 is to help link with legacy assembly code.
12813 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12814 generate code that is not binary compatible with code generated without that
12815 switch. Use it to conform to a non-default application binary interface.
12816 Not all targets provide complete support for this switch.
12818 @item -ftls-model=@var{model}
12819 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12820 The @var{model} argument should be one of @code{global-dynamic},
12821 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
12823 The default without @option{-fpic} is @code{initial-exec}; with
12824 @option{-fpic} the default is @code{global-dynamic}.
12826 @item -fvisibility=@var{default|internal|hidden|protected}
12827 @opindex fvisibility
12828 Set the default ELF image symbol visibility to the specified option---all
12829 symbols will be marked with this unless overridden within the code.
12830 Using this feature can very substantially improve linking and
12831 load times of shared object libraries, produce more optimized
12832 code, provide near-perfect API export and prevent symbol clashes.
12833 It is @strong{strongly} recommended that you use this in any shared objects
12836 Despite the nomenclature, @code{default} always means public ie;
12837 available to be linked against from outside the shared object.
12838 @code{protected} and @code{internal} are pretty useless in real-world
12839 usage so the only other commonly used option will be @code{hidden}.
12840 The default if @option{-fvisibility} isn't specified is
12841 @code{default}, i.e., make every
12842 symbol public---this causes the same behavior as previous versions of
12845 A good explanation of the benefits offered by ensuring ELF
12846 symbols have the correct visibility is given by ``How To Write
12847 Shared Libraries'' by Ulrich Drepper (which can be found at
12848 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
12849 solution made possible by this option to marking things hidden when
12850 the default is public is to make the default hidden and mark things
12851 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
12852 and @code{__attribute__ ((visibility("default")))} instead of
12853 @code{__declspec(dllexport)} you get almost identical semantics with
12854 identical syntax. This is a great boon to those working with
12855 cross-platform projects.
12857 For those adding visibility support to existing code, you may find
12858 @samp{#pragma GCC visibility} of use. This works by you enclosing
12859 the declarations you wish to set visibility for with (for example)
12860 @samp{#pragma GCC visibility push(hidden)} and
12861 @samp{#pragma GCC visibility pop}.
12862 Bear in mind that symbol visibility should be viewed @strong{as
12863 part of the API interface contract} and thus all new code should
12864 always specify visibility when it is not the default ie; declarations
12865 only for use within the local DSO should @strong{always} be marked explicitly
12866 as hidden as so to avoid PLT indirection overheads---making this
12867 abundantly clear also aids readability and self-documentation of the code.
12868 Note that due to ISO C++ specification requirements, operator new and
12869 operator delete must always be of default visibility.
12871 An overview of these techniques, their benefits and how to use them
12872 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
12878 @node Environment Variables
12879 @section Environment Variables Affecting GCC
12880 @cindex environment variables
12882 @c man begin ENVIRONMENT
12883 This section describes several environment variables that affect how GCC
12884 operates. Some of them work by specifying directories or prefixes to use
12885 when searching for various kinds of files. Some are used to specify other
12886 aspects of the compilation environment.
12888 Note that you can also specify places to search using options such as
12889 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
12890 take precedence over places specified using environment variables, which
12891 in turn take precedence over those specified by the configuration of GCC@.
12892 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
12893 GNU Compiler Collection (GCC) Internals}.
12898 @c @itemx LC_COLLATE
12900 @c @itemx LC_MONETARY
12901 @c @itemx LC_NUMERIC
12906 @c @findex LC_COLLATE
12907 @findex LC_MESSAGES
12908 @c @findex LC_MONETARY
12909 @c @findex LC_NUMERIC
12913 These environment variables control the way that GCC uses
12914 localization information that allow GCC to work with different
12915 national conventions. GCC inspects the locale categories
12916 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
12917 so. These locale categories can be set to any value supported by your
12918 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
12919 Kingdom encoded in UTF-8.
12921 The @env{LC_CTYPE} environment variable specifies character
12922 classification. GCC uses it to determine the character boundaries in
12923 a string; this is needed for some multibyte encodings that contain quote
12924 and escape characters that would otherwise be interpreted as a string
12927 The @env{LC_MESSAGES} environment variable specifies the language to
12928 use in diagnostic messages.
12930 If the @env{LC_ALL} environment variable is set, it overrides the value
12931 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
12932 and @env{LC_MESSAGES} default to the value of the @env{LANG}
12933 environment variable. If none of these variables are set, GCC
12934 defaults to traditional C English behavior.
12938 If @env{TMPDIR} is set, it specifies the directory to use for temporary
12939 files. GCC uses temporary files to hold the output of one stage of
12940 compilation which is to be used as input to the next stage: for example,
12941 the output of the preprocessor, which is the input to the compiler
12944 @item GCC_EXEC_PREFIX
12945 @findex GCC_EXEC_PREFIX
12946 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
12947 names of the subprograms executed by the compiler. No slash is added
12948 when this prefix is combined with the name of a subprogram, but you can
12949 specify a prefix that ends with a slash if you wish.
12951 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
12952 an appropriate prefix to use based on the pathname it was invoked with.
12954 If GCC cannot find the subprogram using the specified prefix, it
12955 tries looking in the usual places for the subprogram.
12957 The default value of @env{GCC_EXEC_PREFIX} is
12958 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
12959 of @code{prefix} when you ran the @file{configure} script.
12961 Other prefixes specified with @option{-B} take precedence over this prefix.
12963 This prefix is also used for finding files such as @file{crt0.o} that are
12966 In addition, the prefix is used in an unusual way in finding the
12967 directories to search for header files. For each of the standard
12968 directories whose name normally begins with @samp{/usr/local/lib/gcc}
12969 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
12970 replacing that beginning with the specified prefix to produce an
12971 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
12972 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
12973 These alternate directories are searched first; the standard directories
12976 @item COMPILER_PATH
12977 @findex COMPILER_PATH
12978 The value of @env{COMPILER_PATH} is a colon-separated list of
12979 directories, much like @env{PATH}. GCC tries the directories thus
12980 specified when searching for subprograms, if it can't find the
12981 subprograms using @env{GCC_EXEC_PREFIX}.
12984 @findex LIBRARY_PATH
12985 The value of @env{LIBRARY_PATH} is a colon-separated list of
12986 directories, much like @env{PATH}. When configured as a native compiler,
12987 GCC tries the directories thus specified when searching for special
12988 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
12989 using GCC also uses these directories when searching for ordinary
12990 libraries for the @option{-l} option (but directories specified with
12991 @option{-L} come first).
12995 @cindex locale definition
12996 This variable is used to pass locale information to the compiler. One way in
12997 which this information is used is to determine the character set to be used
12998 when character literals, string literals and comments are parsed in C and C++.
12999 When the compiler is configured to allow multibyte characters,
13000 the following values for @env{LANG} are recognized:
13004 Recognize JIS characters.
13006 Recognize SJIS characters.
13008 Recognize EUCJP characters.
13011 If @env{LANG} is not defined, or if it has some other value, then the
13012 compiler will use mblen and mbtowc as defined by the default locale to
13013 recognize and translate multibyte characters.
13017 Some additional environments variables affect the behavior of the
13020 @include cppenv.texi
13024 @node Precompiled Headers
13025 @section Using Precompiled Headers
13026 @cindex precompiled headers
13027 @cindex speed of compilation
13029 Often large projects have many header files that are included in every
13030 source file. The time the compiler takes to process these header files
13031 over and over again can account for nearly all of the time required to
13032 build the project. To make builds faster, GCC allows users to
13033 `precompile' a header file; then, if builds can use the precompiled
13034 header file they will be much faster.
13036 To create a precompiled header file, simply compile it as you would any
13037 other file, if necessary using the @option{-x} option to make the driver
13038 treat it as a C or C++ header file. You will probably want to use a
13039 tool like @command{make} to keep the precompiled header up-to-date when
13040 the headers it contains change.
13042 A precompiled header file will be searched for when @code{#include} is
13043 seen in the compilation. As it searches for the included file
13044 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
13045 compiler looks for a precompiled header in each directory just before it
13046 looks for the include file in that directory. The name searched for is
13047 the name specified in the @code{#include} with @samp{.gch} appended. If
13048 the precompiled header file can't be used, it is ignored.
13050 For instance, if you have @code{#include "all.h"}, and you have
13051 @file{all.h.gch} in the same directory as @file{all.h}, then the
13052 precompiled header file will be used if possible, and the original
13053 header will be used otherwise.
13055 Alternatively, you might decide to put the precompiled header file in a
13056 directory and use @option{-I} to ensure that directory is searched
13057 before (or instead of) the directory containing the original header.
13058 Then, if you want to check that the precompiled header file is always
13059 used, you can put a file of the same name as the original header in this
13060 directory containing an @code{#error} command.
13062 This also works with @option{-include}. So yet another way to use
13063 precompiled headers, good for projects not designed with precompiled
13064 header files in mind, is to simply take most of the header files used by
13065 a project, include them from another header file, precompile that header
13066 file, and @option{-include} the precompiled header. If the header files
13067 have guards against multiple inclusion, they will be skipped because
13068 they've already been included (in the precompiled header).
13070 If you need to precompile the same header file for different
13071 languages, targets, or compiler options, you can instead make a
13072 @emph{directory} named like @file{all.h.gch}, and put each precompiled
13073 header in the directory, perhaps using @option{-o}. It doesn't matter
13074 what you call the files in the directory, every precompiled header in
13075 the directory will be considered. The first precompiled header
13076 encountered in the directory that is valid for this compilation will
13077 be used; they're searched in no particular order.
13079 There are many other possibilities, limited only by your imagination,
13080 good sense, and the constraints of your build system.
13082 A precompiled header file can be used only when these conditions apply:
13086 Only one precompiled header can be used in a particular compilation.
13089 A precompiled header can't be used once the first C token is seen. You
13090 can have preprocessor directives before a precompiled header; you can
13091 even include a precompiled header from inside another header, so long as
13092 there are no C tokens before the @code{#include}.
13095 The precompiled header file must be produced for the same language as
13096 the current compilation. You can't use a C precompiled header for a C++
13100 The precompiled header file must have been produced by the same compiler
13101 binary as the current compilation is using.
13104 Any macros defined before the precompiled header is included must
13105 either be defined in the same way as when the precompiled header was
13106 generated, or must not affect the precompiled header, which usually
13107 means that they don't appear in the precompiled header at all.
13109 The @option{-D} option is one way to define a macro before a
13110 precompiled header is included; using a @code{#define} can also do it.
13111 There are also some options that define macros implicitly, like
13112 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
13115 @item If debugging information is output when using the precompiled
13116 header, using @option{-g} or similar, the same kind of debugging information
13117 must have been output when building the precompiled header. However,
13118 a precompiled header built using @option{-g} can be used in a compilation
13119 when no debugging information is being output.
13121 @item The same @option{-m} options must generally be used when building
13122 and using the precompiled header. @xref{Submodel Options},
13123 for any cases where this rule is relaxed.
13125 @item Each of the following options must be the same when building and using
13126 the precompiled header:
13128 @gccoptlist{-fexceptions -funit-at-a-time}
13131 Some other command-line options starting with @option{-f},
13132 @option{-p}, or @option{-O} must be defined in the same way as when
13133 the precompiled header was generated. At present, it's not clear
13134 which options are safe to change and which are not; the safest choice
13135 is to use exactly the same options when generating and using the
13136 precompiled header. The following are known to be safe:
13138 @gccoptlist{-fpreprocessed
13139 -fsched-interblock -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
13140 -fsched-verbose=<number> -fschedule-insns
13145 For all of these except the last, the compiler will automatically
13146 ignore the precompiled header if the conditions aren't met. If you
13147 find an option combination that doesn't work and doesn't cause the
13148 precompiled header to be ignored, please consider filing a bug report,
13151 If you do use differing options when generating and using the
13152 precompiled header, the actual behavior will be a mixture of the
13153 behavior for the options. For instance, if you use @option{-g} to
13154 generate the precompiled header but not when using it, you may or may
13155 not get debugging information for routines in the precompiled header.
13157 @node Running Protoize
13158 @section Running Protoize
13160 The program @code{protoize} is an optional part of GCC@. You can use
13161 it to add prototypes to a program, thus converting the program to ISO
13162 C in one respect. The companion program @code{unprotoize} does the
13163 reverse: it removes argument types from any prototypes that are found.
13165 When you run these programs, you must specify a set of source files as
13166 command line arguments. The conversion programs start out by compiling
13167 these files to see what functions they define. The information gathered
13168 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
13170 After scanning comes actual conversion. The specified files are all
13171 eligible to be converted; any files they include (whether sources or
13172 just headers) are eligible as well.
13174 But not all the eligible files are converted. By default,
13175 @code{protoize} and @code{unprotoize} convert only source and header
13176 files in the current directory. You can specify additional directories
13177 whose files should be converted with the @option{-d @var{directory}}
13178 option. You can also specify particular files to exclude with the
13179 @option{-x @var{file}} option. A file is converted if it is eligible, its
13180 directory name matches one of the specified directory names, and its
13181 name within the directory has not been excluded.
13183 Basic conversion with @code{protoize} consists of rewriting most
13184 function definitions and function declarations to specify the types of
13185 the arguments. The only ones not rewritten are those for varargs
13188 @code{protoize} optionally inserts prototype declarations at the
13189 beginning of the source file, to make them available for any calls that
13190 precede the function's definition. Or it can insert prototype
13191 declarations with block scope in the blocks where undeclared functions
13194 Basic conversion with @code{unprotoize} consists of rewriting most
13195 function declarations to remove any argument types, and rewriting
13196 function definitions to the old-style pre-ISO form.
13198 Both conversion programs print a warning for any function declaration or
13199 definition that they can't convert. You can suppress these warnings
13202 The output from @code{protoize} or @code{unprotoize} replaces the
13203 original source file. The original file is renamed to a name ending
13204 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
13205 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
13206 for DOS) file already exists, then the source file is simply discarded.
13208 @code{protoize} and @code{unprotoize} both depend on GCC itself to
13209 scan the program and collect information about the functions it uses.
13210 So neither of these programs will work until GCC is installed.
13212 Here is a table of the options you can use with @code{protoize} and
13213 @code{unprotoize}. Each option works with both programs unless
13217 @item -B @var{directory}
13218 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
13219 usual directory (normally @file{/usr/local/lib}). This file contains
13220 prototype information about standard system functions. This option
13221 applies only to @code{protoize}.
13223 @item -c @var{compilation-options}
13224 Use @var{compilation-options} as the options when running @command{gcc} to
13225 produce the @samp{.X} files. The special option @option{-aux-info} is
13226 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
13228 Note that the compilation options must be given as a single argument to
13229 @code{protoize} or @code{unprotoize}. If you want to specify several
13230 @command{gcc} options, you must quote the entire set of compilation options
13231 to make them a single word in the shell.
13233 There are certain @command{gcc} arguments that you cannot use, because they
13234 would produce the wrong kind of output. These include @option{-g},
13235 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
13236 the @var{compilation-options}, they are ignored.
13239 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
13240 systems) instead of @samp{.c}. This is convenient if you are converting
13241 a C program to C++. This option applies only to @code{protoize}.
13244 Add explicit global declarations. This means inserting explicit
13245 declarations at the beginning of each source file for each function
13246 that is called in the file and was not declared. These declarations
13247 precede the first function definition that contains a call to an
13248 undeclared function. This option applies only to @code{protoize}.
13250 @item -i @var{string}
13251 Indent old-style parameter declarations with the string @var{string}.
13252 This option applies only to @code{protoize}.
13254 @code{unprotoize} converts prototyped function definitions to old-style
13255 function definitions, where the arguments are declared between the
13256 argument list and the initial @samp{@{}. By default, @code{unprotoize}
13257 uses five spaces as the indentation. If you want to indent with just
13258 one space instead, use @option{-i " "}.
13261 Keep the @samp{.X} files. Normally, they are deleted after conversion
13265 Add explicit local declarations. @code{protoize} with @option{-l} inserts
13266 a prototype declaration for each function in each block which calls the
13267 function without any declaration. This option applies only to
13271 Make no real changes. This mode just prints information about the conversions
13272 that would have been done without @option{-n}.
13275 Make no @samp{.save} files. The original files are simply deleted.
13276 Use this option with caution.
13278 @item -p @var{program}
13279 Use the program @var{program} as the compiler. Normally, the name
13280 @file{gcc} is used.
13283 Work quietly. Most warnings are suppressed.
13286 Print the version number, just like @option{-v} for @command{gcc}.
13289 If you need special compiler options to compile one of your program's
13290 source files, then you should generate that file's @samp{.X} file
13291 specially, by running @command{gcc} on that source file with the
13292 appropriate options and the option @option{-aux-info}. Then run
13293 @code{protoize} on the entire set of files. @code{protoize} will use
13294 the existing @samp{.X} file because it is newer than the source file.
13298 gcc -Dfoo=bar file1.c -aux-info file1.X
13303 You need to include the special files along with the rest in the
13304 @code{protoize} command, even though their @samp{.X} files already
13305 exist, because otherwise they won't get converted.
13307 @xref{Protoize Caveats}, for more information on how to use
13308 @code{protoize} successfully.