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 @gol
189 -Wno-non-template-friend -Wold-style-cast @gol
190 -Woverloaded-virtual -Wno-pmf-conversions @gol
193 @item Objective-C and Objective-C++ Language Options
194 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
195 Objective-C and Objective-C++ Dialects}.
197 -fconstant-string-class=@var{class-name} @gol
198 -fgnu-runtime -fnext-runtime @gol
199 -fno-nil-receivers @gol
200 -fobjc-exceptions @gol
201 -freplace-objc-classes @gol
204 -Wno-protocol -Wselector -Wundeclared-selector}
206 @item Language Independent Options
207 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
208 @gccoptlist{-fmessage-length=@var{n} @gol
209 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}}
211 @item Warning Options
212 @xref{Warning Options,,Options to Request or Suppress Warnings}.
213 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
214 -w -Wextra -Wall -Waggregate-return @gol
215 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
216 -Wconversion -Wno-deprecated-declarations @gol
217 -Wdisabled-optimization -Wno-div-by-zero -Wno-endif-labels @gol
218 -Werror -Werror-implicit-function-declaration @gol
219 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
220 -Wno-format-extra-args -Wformat-nonliteral @gol
221 -Wformat-security -Wformat-y2k @gol
222 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
223 -Wimport -Wno-import -Winit-self -Winline @gol
224 -Wno-int-to-pointer-cast @gol
225 -Wno-invalid-offsetof -Winvalid-pch @gol
226 -Wlarger-than-@var{len} -Wlong-long @gol
227 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
228 -Wmissing-format-attribute -Wmissing-include-dirs @gol
229 -Wmissing-noreturn @gol
230 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
231 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
232 -Wredundant-decls @gol
233 -Wreturn-type -Wsequence-point -Wshadow @gol
234 -Wsign-compare -Wstrict-aliasing -Wstrict-aliasing=2 @gol
235 -Wswitch -Wswitch-default -Wswitch-enum @gol
236 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
237 -Wunknown-pragmas -Wunreachable-code @gol
238 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
239 -Wunused-value -Wunused-variable -Wwrite-strings @gol
242 @item C-only Warning Options
243 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
244 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
245 -Wstrict-prototypes -Wtraditional @gol
246 -Wdeclaration-after-statement -Wno-pointer-sign}
248 @item Debugging Options
249 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
250 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
251 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
252 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
253 -fdump-ipa-all -fdump-ipa-cgraph @gol
255 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
256 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
257 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
258 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
260 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
261 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
262 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
263 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
264 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
266 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
267 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
268 -fdump-tree-nrv -fdump-tree-vect @gol
269 -fdump-tree-sink @gol
270 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
271 -fdump-tree-salias @gol
272 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
273 -ftree-vectorizer-verbose=@var{n} @gol
274 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
275 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
276 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs -ftree-based-profiling @gol
277 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
278 -ftest-coverage -ftime-report -fvar-tracking @gol
279 -g -g@var{level} -gcoff -gdwarf-2 @gol
280 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
281 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
282 -print-multi-directory -print-multi-lib @gol
283 -print-prog-name=@var{program} -print-search-dirs -Q @gol
286 @item Optimization Options
287 @xref{Optimize Options,,Options that Control Optimization}.
288 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
289 -falign-labels=@var{n} -falign-loops=@var{n} @gol
290 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
291 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
292 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
293 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
294 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
295 -fdelayed-branch -fdelete-null-pointer-checks @gol
296 -fexpensive-optimizations -ffast-math -ffloat-store @gol
297 -fforce-addr -fforce-mem -ffunction-sections @gol
298 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
299 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2 @gol
300 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
301 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
302 -fmodulo-sched -fno-branch-count-reg @gol
303 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants @gol
304 -fno-function-cse -fno-guess-branch-probability @gol
305 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
306 -funsafe-math-optimizations -ffinite-math-only @gol
307 -fno-trapping-math -fno-zero-initialized-in-bss @gol
308 -fomit-frame-pointer -foptimize-register-move @gol
309 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
310 -fprofile-generate -fprofile-use @gol
311 -fregmove -frename-registers @gol
312 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
313 -frerun-cse-after-loop -frerun-loop-opt @gol
314 -frounding-math -fschedule-insns -fschedule-insns2 @gol
315 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
316 -fsched-spec-load-dangerous @gol
317 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
318 -fsched2-use-superblocks @gol
319 -fsched2-use-traces -freschedule-modulo-scheduled-loops @gol
320 -fsignaling-nans -fsingle-precision-constant -fspeculative-prefetching @gol
321 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
322 -funroll-all-loops -funroll-loops -fpeel-loops @gol
323 -fsplit-ivs-in-unroller -funswitch-loops @gol
324 -fvariable-expansion-in-unroller @gol
325 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
326 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
327 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
328 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
329 -ftree-salias -fweb @gol
330 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop @gol
331 --param @var{name}=@var{value}
332 -O -O0 -O1 -O2 -O3 -Os}
334 @item Preprocessor Options
335 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
336 @gccoptlist{-A@var{question}=@var{answer} @gol
337 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
338 -C -dD -dI -dM -dN @gol
339 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
340 -idirafter @var{dir} @gol
341 -include @var{file} -imacros @var{file} @gol
342 -iprefix @var{file} -iwithprefix @var{dir} @gol
343 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
344 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
345 -P -fworking-directory -remap @gol
346 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
347 -Xpreprocessor @var{option}}
349 @item Assembler Option
350 @xref{Assembler Options,,Passing Options to the Assembler}.
351 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
354 @xref{Link Options,,Options for Linking}.
355 @gccoptlist{@var{object-file-name} -l@var{library} @gol
356 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
357 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
358 -Wl,@var{option} -Xlinker @var{option} @gol
361 @item Directory Options
362 @xref{Directory Options,,Options for Directory Search}.
363 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-}
366 @c I wrote this xref this way to avoid overfull hbox. -- rms
367 @xref{Target Options}.
368 @gccoptlist{-V @var{version} -b @var{machine}}
370 @item Machine Dependent Options
371 @xref{Submodel Options,,Hardware Models and Configurations}.
372 @c This list is ordered alphanumerically by subsection name.
373 @c Try and put the significant identifier (CPU or system) first,
374 @c so users have a clue at guessing where the ones they want will be.
377 @gccoptlist{-EB -EL @gol
378 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
379 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
382 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
383 -mabi=@var{name} @gol
384 -mapcs-stack-check -mno-apcs-stack-check @gol
385 -mapcs-float -mno-apcs-float @gol
386 -mapcs-reentrant -mno-apcs-reentrant @gol
387 -msched-prolog -mno-sched-prolog @gol
388 -mlittle-endian -mbig-endian -mwords-little-endian @gol
389 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
390 -mthumb-interwork -mno-thumb-interwork @gol
391 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
392 -mstructure-size-boundary=@var{n} @gol
393 -mabort-on-noreturn @gol
394 -mlong-calls -mno-long-calls @gol
395 -msingle-pic-base -mno-single-pic-base @gol
396 -mpic-register=@var{reg} @gol
397 -mnop-fun-dllimport @gol
398 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
399 -mpoke-function-name @gol
401 -mtpcs-frame -mtpcs-leaf-frame @gol
402 -mcaller-super-interworking -mcallee-super-interworking}
405 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
406 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
408 @emph{Blackfin Options}
409 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer -mcsync @gol
410 -mno-csync -mlow-64k -mno-low64k -mid-shared-library @gol
411 -mno-id-shared-library -mshared-library-id=@var{n} @gol}
414 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
415 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
416 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
417 -mstack-align -mdata-align -mconst-align @gol
418 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
419 -melf -maout -melinux -mlinux -sim -sim2 @gol
420 -mmul-bug-workaround -mno-mul-bug-workaround}
422 @emph{Darwin Options}
423 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
424 -arch_only -bind_at_load -bundle -bundle_loader @gol
425 -client_name -compatibility_version -current_version @gol
427 -dependency-file -dylib_file -dylinker_install_name @gol
428 -dynamic -dynamiclib -exported_symbols_list @gol
429 -filelist -flat_namespace -force_cpusubtype_ALL @gol
430 -force_flat_namespace -headerpad_max_install_names @gol
431 -image_base -init -install_name -keep_private_externs @gol
432 -multi_module -multiply_defined -multiply_defined_unused @gol
433 -noall_load -no_dead_strip_inits_and_terms @gol
434 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
435 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
436 -private_bundle -read_only_relocs -sectalign @gol
437 -sectobjectsymbols -whyload -seg1addr @gol
438 -sectcreate -sectobjectsymbols -sectorder @gol
439 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
440 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
441 -segprot -segs_read_only_addr -segs_read_write_addr @gol
442 -single_module -static -sub_library -sub_umbrella @gol
443 -twolevel_namespace -umbrella -undefined @gol
444 -unexported_symbols_list -weak_reference_mismatches @gol
445 -whatsloaded -F -gused -gfull -mone-byte-bool}
447 @emph{DEC Alpha Options}
448 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
449 -mieee -mieee-with-inexact -mieee-conformant @gol
450 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
451 -mtrap-precision=@var{mode} -mbuild-constants @gol
452 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
453 -mbwx -mmax -mfix -mcix @gol
454 -mfloat-vax -mfloat-ieee @gol
455 -mexplicit-relocs -msmall-data -mlarge-data @gol
456 -msmall-text -mlarge-text @gol
457 -mmemory-latency=@var{time}}
459 @emph{DEC Alpha/VMS Options}
460 @gccoptlist{-mvms-return-codes}
463 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
464 -mhard-float -msoft-float @gol
465 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
466 -mdouble -mno-double @gol
467 -mmedia -mno-media -mmuladd -mno-muladd @gol
468 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
469 -mlinked-fp -mlong-calls -malign-labels @gol
470 -mlibrary-pic -macc-4 -macc-8 @gol
471 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
472 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
473 -mvliw-branch -mno-vliw-branch @gol
474 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
475 -mno-nested-cond-exec -mtomcat-stats @gol
479 @emph{H8/300 Options}
480 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
483 @gccoptlist{-march=@var{architecture-type} @gol
484 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
485 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
486 -mfixed-range=@var{register-range} @gol
487 -mjump-in-delay -mlinker-opt -mlong-calls @gol
488 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
489 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
490 -mno-jump-in-delay -mno-long-load-store @gol
491 -mno-portable-runtime -mno-soft-float @gol
492 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
493 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
494 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
495 -munix=@var{unix-std} -nolibdld -static -threads}
497 @emph{i386 and x86-64 Options}
498 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
499 -mfpmath=@var{unit} @gol
500 -masm=@var{dialect} -mno-fancy-math-387 @gol
501 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
502 -mno-wide-multiply -mrtd -malign-double @gol
503 -mpreferred-stack-boundary=@var{num} @gol
504 -mmmx -msse -msse2 -msse3 -m3dnow @gol
505 -mthreads -mno-align-stringops -minline-all-stringops @gol
506 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
507 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
508 -mno-red-zone -mno-tls-direct-seg-refs @gol
509 -mcmodel=@var{code-model} @gol
513 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
514 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
515 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
516 -minline-float-divide-max-throughput @gol
517 -minline-int-divide-min-latency @gol
518 -minline-int-divide-max-throughput @gol
519 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
520 -mno-dwarf2-asm -mearly-stop-bits @gol
521 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
522 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64}
524 @emph{M32R/D Options}
525 @gccoptlist{-m32r2 -m32rx -m32r @gol
527 -malign-loops -mno-align-loops @gol
528 -missue-rate=@var{number} @gol
529 -mbranch-cost=@var{number} @gol
530 -mmodel=@var{code-size-model-type} @gol
531 -msdata=@var{sdata-type} @gol
532 -mno-flush-func -mflush-func=@var{name} @gol
533 -mno-flush-trap -mflush-trap=@var{number} @gol
536 @emph{M680x0 Options}
537 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
538 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
539 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
540 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
541 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
543 @emph{M68hc1x Options}
544 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
545 -mauto-incdec -minmax -mlong-calls -mshort @gol
546 -msoft-reg-count=@var{count}}
549 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
550 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
551 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
552 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
553 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
556 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
557 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
558 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
559 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
560 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
561 -mpaired-single -mips3d @gol
562 -mint64 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
563 -G@var{num} -membedded-data -mno-embedded-data @gol
564 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
565 -msplit-addresses -mno-split-addresses @gol
566 -mexplicit-relocs -mno-explicit-relocs @gol
567 -mcheck-zero-division -mno-check-zero-division @gol
568 -mdivide-traps -mdivide-breaks @gol
569 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
570 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
571 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
572 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
573 -mfix-sb1 -mno-fix-sb1 @gol
574 -mflush-func=@var{func} -mno-flush-func @gol
575 -mbranch-likely -mno-branch-likely @gol
576 -mfp-exceptions -mno-fp-exceptions @gol
577 -mvr4130-align -mno-vr4130-align}
580 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
581 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
582 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
583 -mno-base-addresses -msingle-exit -mno-single-exit}
585 @emph{MN10300 Options}
586 @gccoptlist{-mmult-bug -mno-mult-bug @gol
587 -mam33 -mno-am33 @gol
588 -mam33-2 -mno-am33-2 @gol
592 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
593 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
594 -mregparam -mnoregparam -msb -mnosb @gol
595 -mbitfield -mnobitfield -mhimem -mnohimem}
597 @emph{PDP-11 Options}
598 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
599 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
600 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
601 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
602 -mbranch-expensive -mbranch-cheap @gol
603 -msplit -mno-split -munix-asm -mdec-asm}
605 @emph{PowerPC Options}
606 See RS/6000 and PowerPC Options.
608 @emph{RS/6000 and PowerPC Options}
609 @gccoptlist{-mcpu=@var{cpu-type} @gol
610 -mtune=@var{cpu-type} @gol
611 -mpower -mno-power -mpower2 -mno-power2 @gol
612 -mpowerpc -mpowerpc64 -mno-powerpc @gol
613 -maltivec -mno-altivec @gol
614 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
615 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
616 -mnew-mnemonics -mold-mnemonics @gol
617 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
618 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
619 -malign-power -malign-natural @gol
620 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
621 -mstring -mno-string -mupdate -mno-update @gol
622 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
623 -mstrict-align -mno-strict-align -mrelocatable @gol
624 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
625 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
626 -mdynamic-no-pic @gol
627 -mprioritize-restricted-insns=@var{priority} @gol
628 -msched-costly-dep=@var{dependence_type} @gol
629 -minsert-sched-nops=@var{scheme} @gol
630 -mcall-sysv -mcall-netbsd @gol
631 -maix-struct-return -msvr4-struct-return @gol
632 -mabi=altivec -mabi=no-altivec @gol
633 -mabi=spe -mabi=no-spe @gol
634 -misel=yes -misel=no @gol
635 -mspe=yes -mspe=no @gol
636 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
637 -mprototype -mno-prototype @gol
638 -msim -mmvme -mads -myellowknife -memb -msdata @gol
639 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
641 @emph{S/390 and zSeries Options}
642 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
643 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
644 -mpacked-stack -mno-packed-stack @gol
645 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
646 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
647 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
648 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
651 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
652 -m4-nofpu -m4-single-only -m4-single -m4 @gol
653 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
654 -m5-64media -m5-64media-nofpu @gol
655 -m5-32media -m5-32media-nofpu @gol
656 -m5-compact -m5-compact-nofpu @gol
657 -mb -ml -mdalign -mrelax @gol
658 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
659 -mieee -misize -mpadstruct -mspace @gol
660 -mprefergot -musermode}
663 @gccoptlist{-mcpu=@var{cpu-type} @gol
664 -mtune=@var{cpu-type} @gol
665 -mcmodel=@var{code-model} @gol
666 -m32 -m64 -mapp-regs -mno-app-regs @gol
667 -mfaster-structs -mno-faster-structs @gol
668 -mfpu -mno-fpu -mhard-float -msoft-float @gol
669 -mhard-quad-float -msoft-quad-float @gol
670 -mimpure-text -mno-impure-text -mlittle-endian @gol
671 -mstack-bias -mno-stack-bias @gol
672 -munaligned-doubles -mno-unaligned-doubles @gol
673 -mv8plus -mno-v8plus -mvis -mno-vis
676 @emph{System V Options}
677 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
679 @emph{TMS320C3x/C4x Options}
680 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
681 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
682 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
683 -mparallel-insns -mparallel-mpy -mpreserve-float}
686 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
687 -mprolog-function -mno-prolog-function -mspace @gol
688 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
689 -mapp-regs -mno-app-regs @gol
690 -mdisable-callt -mno-disable-callt @gol
696 @gccoptlist{-mg -mgnu -munix}
698 @emph{x86-64 Options}
699 See i386 and x86-64 Options.
701 @emph{Xstormy16 Options}
704 @emph{Xtensa Options}
705 @gccoptlist{-mconst16 -mno-const16 @gol
706 -mfused-madd -mno-fused-madd @gol
707 -mtext-section-literals -mno-text-section-literals @gol
708 -mtarget-align -mno-target-align @gol
709 -mlongcalls -mno-longcalls}
711 @emph{zSeries Options}
712 See S/390 and zSeries Options.
714 @item Code Generation Options
715 @xref{Code Gen Options,,Options for Code Generation Conventions}.
716 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
717 -ffixed-@var{reg} -fexceptions @gol
718 -fnon-call-exceptions -funwind-tables @gol
719 -fasynchronous-unwind-tables @gol
720 -finhibit-size-directive -finstrument-functions @gol
721 -fno-common -fno-ident @gol
722 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
723 -freg-struct-return -fshared-data -fshort-enums @gol
724 -fshort-double -fshort-wchar @gol
725 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
726 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
727 -fargument-alias -fargument-noalias @gol
728 -fargument-noalias-global -fleading-underscore @gol
729 -ftls-model=@var{model} @gol
730 -ftrapv -fwrapv -fbounds-check @gol
735 * Overall Options:: Controlling the kind of output:
736 an executable, object files, assembler files,
737 or preprocessed source.
738 * C Dialect Options:: Controlling the variant of C language compiled.
739 * C++ Dialect Options:: Variations on C++.
740 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
742 * Language Independent Options:: Controlling how diagnostics should be
744 * Warning Options:: How picky should the compiler be?
745 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
746 * Optimize Options:: How much optimization?
747 * Preprocessor Options:: Controlling header files and macro definitions.
748 Also, getting dependency information for Make.
749 * Assembler Options:: Passing options to the assembler.
750 * Link Options:: Specifying libraries and so on.
751 * Directory Options:: Where to find header files and libraries.
752 Where to find the compiler executable files.
753 * Spec Files:: How to pass switches to sub-processes.
754 * Target Options:: Running a cross-compiler, or an old version of GCC.
757 @node Overall Options
758 @section Options Controlling the Kind of Output
760 Compilation can involve up to four stages: preprocessing, compilation
761 proper, assembly and linking, always in that order. GCC is capable of
762 preprocessing and compiling several files either into several
763 assembler input files, or into one assembler input file; then each
764 assembler input file produces an object file, and linking combines all
765 the object files (those newly compiled, and those specified as input)
766 into an executable file.
768 @cindex file name suffix
769 For any given input file, the file name suffix determines what kind of
774 C source code which must be preprocessed.
777 C source code which should not be preprocessed.
780 C++ source code which should not be preprocessed.
783 Objective-C source code. Note that you must link with the @file{libobjc}
784 library to make an Objective-C program work.
787 Objective-C source code which should not be preprocessed.
791 Objective-C++ source code. Note that you must link with the @file{libobjc}
792 library to make an Objective-C++ program work. Note that @samp{.M} refers
793 to a literal capital M@.
796 Objective-C++ source code which should not be preprocessed.
799 C, C++, Objective-C or Objective-C++ header file to be turned into a
804 @itemx @var{file}.cxx
805 @itemx @var{file}.cpp
806 @itemx @var{file}.CPP
807 @itemx @var{file}.c++
809 C++ source code which must be preprocessed. Note that in @samp{.cxx},
810 the last two letters must both be literally @samp{x}. Likewise,
811 @samp{.C} refers to a literal capital C@.
815 C++ header file to be turned into a precompiled header.
818 @itemx @var{file}.for
819 @itemx @var{file}.FOR
820 Fortran source code which should not be preprocessed.
823 @itemx @var{file}.fpp
824 @itemx @var{file}.FPP
825 Fortran source code which must be preprocessed (with the traditional
829 Fortran source code which must be preprocessed with a RATFOR
830 preprocessor (not included with GCC)@.
833 @itemx @var{file}.f95
834 Fortran 90/95 source code which should not be preprocessed.
836 @c FIXME: Descriptions of Java file types.
843 Ada source code file which contains a library unit declaration (a
844 declaration of a package, subprogram, or generic, or a generic
845 instantiation), or a library unit renaming declaration (a package,
846 generic, or subprogram renaming declaration). Such files are also
849 @itemx @var{file}.adb
850 Ada source code file containing a library unit body (a subprogram or
851 package body). Such files are also called @dfn{bodies}.
853 @c GCC also knows about some suffixes for languages not yet included:
862 Assembler code which must be preprocessed.
865 An object file to be fed straight into linking.
866 Any file name with no recognized suffix is treated this way.
870 You can specify the input language explicitly with the @option{-x} option:
873 @item -x @var{language}
874 Specify explicitly the @var{language} for the following input files
875 (rather than letting the compiler choose a default based on the file
876 name suffix). This option applies to all following input files until
877 the next @option{-x} option. Possible values for @var{language} are:
879 c c-header c-cpp-output
880 c++ c++-header c++-cpp-output
881 objective-c objective-c-header objective-c-cpp-output
882 objective-c++ objective-c++-header objective-c++-cpp-output
883 assembler assembler-with-cpp
885 f77 f77-cpp-input ratfor
892 Turn off any specification of a language, so that subsequent files are
893 handled according to their file name suffixes (as they are if @option{-x}
894 has not been used at all).
896 @item -pass-exit-codes
897 @opindex pass-exit-codes
898 Normally the @command{gcc} program will exit with the code of 1 if any
899 phase of the compiler returns a non-success return code. If you specify
900 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
901 numerically highest error produced by any phase that returned an error
905 If you only want some of the stages of compilation, you can use
906 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
907 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
908 @command{gcc} is to stop. Note that some combinations (for example,
909 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
914 Compile or assemble the source files, but do not link. The linking
915 stage simply is not done. The ultimate output is in the form of an
916 object file for each source file.
918 By default, the object file name for a source file is made by replacing
919 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
921 Unrecognized input files, not requiring compilation or assembly, are
926 Stop after the stage of compilation proper; do not assemble. The output
927 is in the form of an assembler code file for each non-assembler input
930 By default, the assembler file name for a source file is made by
931 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
933 Input files that don't require compilation are ignored.
937 Stop after the preprocessing stage; do not run the compiler proper. The
938 output is in the form of preprocessed source code, which is sent to the
941 Input files which don't require preprocessing are ignored.
943 @cindex output file option
946 Place output in file @var{file}. This applies regardless to whatever
947 sort of output is being produced, whether it be an executable file,
948 an object file, an assembler file or preprocessed C code.
950 If @option{-o} is not specified, the default is to put an executable
951 file in @file{a.out}, the object file for
952 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
953 assembler file in @file{@var{source}.s}, a precompiled header file in
954 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
959 Print (on standard error output) the commands executed to run the stages
960 of compilation. Also print the version number of the compiler driver
961 program and of the preprocessor and the compiler proper.
965 Like @option{-v} except the commands are not executed and all command
966 arguments are quoted. This is useful for shell scripts to capture the
967 driver-generated command lines.
971 Use pipes rather than temporary files for communication between the
972 various stages of compilation. This fails to work on some systems where
973 the assembler is unable to read from a pipe; but the GNU assembler has
978 If you are compiling multiple source files, this option tells the driver
979 to pass all the source files to the compiler at once (for those
980 languages for which the compiler can handle this). This will allow
981 intermodule analysis (IMA) to be performed by the compiler. Currently the only
982 language for which this is supported is C@. If you pass source files for
983 multiple languages to the driver, using this option, the driver will invoke
984 the compiler(s) that support IMA once each, passing each compiler all the
985 source files appropriate for it. For those languages that do not support
986 IMA this option will be ignored, and the compiler will be invoked once for
987 each source file in that language. If you use this option in conjunction
988 with @option{-save-temps}, the compiler will generate multiple
990 (one for each source file), but only one (combined) @file{.o} or
995 Print (on the standard output) a description of the command line options
996 understood by @command{gcc}. If the @option{-v} option is also specified
997 then @option{--help} will also be passed on to the various processes
998 invoked by @command{gcc}, so that they can display the command line options
999 they accept. If the @option{-Wextra} option is also specified then command
1000 line options which have no documentation associated with them will also
1004 @opindex target-help
1005 Print (on the standard output) a description of target specific command
1006 line options for each tool.
1010 Display the version number and copyrights of the invoked GCC@.
1014 @section Compiling C++ Programs
1016 @cindex suffixes for C++ source
1017 @cindex C++ source file suffixes
1018 C++ source files conventionally use one of the suffixes @samp{.C},
1019 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1020 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1021 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1022 files with these names and compiles them as C++ programs even if you
1023 call the compiler the same way as for compiling C programs (usually
1024 with the name @command{gcc}).
1028 However, C++ programs often require class libraries as well as a
1029 compiler that understands the C++ language---and under some
1030 circumstances, you might want to compile programs or header files from
1031 standard input, or otherwise without a suffix that flags them as C++
1032 programs. You might also like to precompile a C header file with a
1033 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
1034 program that calls GCC with the default language set to C++, and
1035 automatically specifies linking against the C++ library. On many
1036 systems, @command{g++} is also installed with the name @command{c++}.
1038 @cindex invoking @command{g++}
1039 When you compile C++ programs, you may specify many of the same
1040 command-line options that you use for compiling programs in any
1041 language; or command-line options meaningful for C and related
1042 languages; or options that are meaningful only for C++ programs.
1043 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1044 explanations of options for languages related to C@.
1045 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1046 explanations of options that are meaningful only for C++ programs.
1048 @node C Dialect Options
1049 @section Options Controlling C Dialect
1050 @cindex dialect options
1051 @cindex language dialect options
1052 @cindex options, dialect
1054 The following options control the dialect of C (or languages derived
1055 from C, such as C++, Objective-C and Objective-C++) that the compiler
1059 @cindex ANSI support
1063 In C mode, support all ISO C90 programs. In C++ mode,
1064 remove GNU extensions that conflict with ISO C++.
1066 This turns off certain features of GCC that are incompatible with ISO
1067 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1068 such as the @code{asm} and @code{typeof} keywords, and
1069 predefined macros such as @code{unix} and @code{vax} that identify the
1070 type of system you are using. It also enables the undesirable and
1071 rarely used ISO trigraph feature. For the C compiler,
1072 it disables recognition of C++ style @samp{//} comments as well as
1073 the @code{inline} keyword.
1075 The alternate keywords @code{__asm__}, @code{__extension__},
1076 @code{__inline__} and @code{__typeof__} continue to work despite
1077 @option{-ansi}. You would not want to use them in an ISO C program, of
1078 course, but it is useful to put them in header files that might be included
1079 in compilations done with @option{-ansi}. Alternate predefined macros
1080 such as @code{__unix__} and @code{__vax__} are also available, with or
1081 without @option{-ansi}.
1083 The @option{-ansi} option does not cause non-ISO programs to be
1084 rejected gratuitously. For that, @option{-pedantic} is required in
1085 addition to @option{-ansi}. @xref{Warning Options}.
1087 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1088 option is used. Some header files may notice this macro and refrain
1089 from declaring certain functions or defining certain macros that the
1090 ISO standard doesn't call for; this is to avoid interfering with any
1091 programs that might use these names for other things.
1093 Functions which would normally be built in but do not have semantics
1094 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1095 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1096 built-in functions provided by GCC}, for details of the functions
1101 Determine the language standard. This option is currently only
1102 supported when compiling C or C++. A value for this option must be
1103 provided; possible values are
1108 ISO C90 (same as @option{-ansi}).
1110 @item iso9899:199409
1111 ISO C90 as modified in amendment 1.
1117 ISO C99. Note that this standard is not yet fully supported; see
1118 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1119 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1122 Default, ISO C90 plus GNU extensions (including some C99 features).
1126 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1127 this will become the default. The name @samp{gnu9x} is deprecated.
1130 The 1998 ISO C++ standard plus amendments.
1133 The same as @option{-std=c++98} plus GNU extensions. This is the
1134 default for C++ code.
1137 Even when this option is not specified, you can still use some of the
1138 features of newer standards in so far as they do not conflict with
1139 previous C standards. For example, you may use @code{__restrict__} even
1140 when @option{-std=c99} is not specified.
1142 The @option{-std} options specifying some version of ISO C have the same
1143 effects as @option{-ansi}, except that features that were not in ISO C90
1144 but are in the specified version (for example, @samp{//} comments and
1145 the @code{inline} keyword in ISO C99) are not disabled.
1147 @xref{Standards,,Language Standards Supported by GCC}, for details of
1148 these standard versions.
1150 @item -aux-info @var{filename}
1152 Output to the given filename prototyped declarations for all functions
1153 declared and/or defined in a translation unit, including those in header
1154 files. This option is silently ignored in any language other than C@.
1156 Besides declarations, the file indicates, in comments, the origin of
1157 each declaration (source file and line), whether the declaration was
1158 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1159 @samp{O} for old, respectively, in the first character after the line
1160 number and the colon), and whether it came from a declaration or a
1161 definition (@samp{C} or @samp{F}, respectively, in the following
1162 character). In the case of function definitions, a K&R-style list of
1163 arguments followed by their declarations is also provided, inside
1164 comments, after the declaration.
1168 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1169 keyword, so that code can use these words as identifiers. You can use
1170 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1171 instead. @option{-ansi} implies @option{-fno-asm}.
1173 In C++, this switch only affects the @code{typeof} keyword, since
1174 @code{asm} and @code{inline} are standard keywords. You may want to
1175 use the @option{-fno-gnu-keywords} flag instead, which has the same
1176 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1177 switch only affects the @code{asm} and @code{typeof} keywords, since
1178 @code{inline} is a standard keyword in ISO C99.
1181 @itemx -fno-builtin-@var{function}
1182 @opindex fno-builtin
1183 @cindex built-in functions
1184 Don't recognize built-in functions that do not begin with
1185 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1186 functions provided by GCC}, for details of the functions affected,
1187 including those which are not built-in functions when @option{-ansi} or
1188 @option{-std} options for strict ISO C conformance are used because they
1189 do not have an ISO standard meaning.
1191 GCC normally generates special code to handle certain built-in functions
1192 more efficiently; for instance, calls to @code{alloca} may become single
1193 instructions that adjust the stack directly, and calls to @code{memcpy}
1194 may become inline copy loops. The resulting code is often both smaller
1195 and faster, but since the function calls no longer appear as such, you
1196 cannot set a breakpoint on those calls, nor can you change the behavior
1197 of the functions by linking with a different library. In addition,
1198 when a function is recognized as a built-in function, GCC may use
1199 information about that function to warn about problems with calls to
1200 that function, or to generate more efficient code, even if the
1201 resulting code still contains calls to that function. For example,
1202 warnings are given with @option{-Wformat} for bad calls to
1203 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1204 known not to modify global memory.
1206 With the @option{-fno-builtin-@var{function}} option
1207 only the built-in function @var{function} is
1208 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1209 function is named this is not built-in in this version of GCC, this
1210 option is ignored. There is no corresponding
1211 @option{-fbuiltin-@var{function}} option; if you wish to enable
1212 built-in functions selectively when using @option{-fno-builtin} or
1213 @option{-ffreestanding}, you may define macros such as:
1216 #define abs(n) __builtin_abs ((n))
1217 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1222 @cindex hosted environment
1224 Assert that compilation takes place in a hosted environment. This implies
1225 @option{-fbuiltin}. A hosted environment is one in which the
1226 entire standard library is available, and in which @code{main} has a return
1227 type of @code{int}. Examples are nearly everything except a kernel.
1228 This is equivalent to @option{-fno-freestanding}.
1230 @item -ffreestanding
1231 @opindex ffreestanding
1232 @cindex hosted environment
1234 Assert that compilation takes place in a freestanding environment. This
1235 implies @option{-fno-builtin}. A freestanding environment
1236 is one in which the standard library may not exist, and program startup may
1237 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1238 This is equivalent to @option{-fno-hosted}.
1240 @xref{Standards,,Language Standards Supported by GCC}, for details of
1241 freestanding and hosted environments.
1243 @item -fms-extensions
1244 @opindex fms-extensions
1245 Accept some non-standard constructs used in Microsoft header files.
1247 Some cases of unnamed fields in structures and unions are only
1248 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1249 fields within structs/unions}, for details.
1253 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1254 options for strict ISO C conformance) implies @option{-trigraphs}.
1256 @item -no-integrated-cpp
1257 @opindex no-integrated-cpp
1258 Performs a compilation in two passes: preprocessing and compiling. This
1259 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1260 @option{-B} option. The user supplied compilation step can then add in
1261 an additional preprocessing step after normal preprocessing but before
1262 compiling. The default is to use the integrated cpp (internal cpp)
1264 The semantics of this option will change if "cc1", "cc1plus", and
1265 "cc1obj" are merged.
1267 @cindex traditional C language
1268 @cindex C language, traditional
1270 @itemx -traditional-cpp
1271 @opindex traditional-cpp
1272 @opindex traditional
1273 Formerly, these options caused GCC to attempt to emulate a pre-standard
1274 C compiler. They are now only supported with the @option{-E} switch.
1275 The preprocessor continues to support a pre-standard mode. See the GNU
1276 CPP manual for details.
1278 @item -fcond-mismatch
1279 @opindex fcond-mismatch
1280 Allow conditional expressions with mismatched types in the second and
1281 third arguments. The value of such an expression is void. This option
1282 is not supported for C++.
1284 @item -funsigned-char
1285 @opindex funsigned-char
1286 Let the type @code{char} be unsigned, like @code{unsigned char}.
1288 Each kind of machine has a default for what @code{char} should
1289 be. It is either like @code{unsigned char} by default or like
1290 @code{signed char} by default.
1292 Ideally, a portable program should always use @code{signed char} or
1293 @code{unsigned char} when it depends on the signedness of an object.
1294 But many programs have been written to use plain @code{char} and
1295 expect it to be signed, or expect it to be unsigned, depending on the
1296 machines they were written for. This option, and its inverse, let you
1297 make such a program work with the opposite default.
1299 The type @code{char} is always a distinct type from each of
1300 @code{signed char} or @code{unsigned char}, even though its behavior
1301 is always just like one of those two.
1304 @opindex fsigned-char
1305 Let the type @code{char} be signed, like @code{signed char}.
1307 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1308 the negative form of @option{-funsigned-char}. Likewise, the option
1309 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1311 @item -fsigned-bitfields
1312 @itemx -funsigned-bitfields
1313 @itemx -fno-signed-bitfields
1314 @itemx -fno-unsigned-bitfields
1315 @opindex fsigned-bitfields
1316 @opindex funsigned-bitfields
1317 @opindex fno-signed-bitfields
1318 @opindex fno-unsigned-bitfields
1319 These options control whether a bit-field is signed or unsigned, when the
1320 declaration does not use either @code{signed} or @code{unsigned}. By
1321 default, such a bit-field is signed, because this is consistent: the
1322 basic integer types such as @code{int} are signed types.
1325 @node C++ Dialect Options
1326 @section Options Controlling C++ Dialect
1328 @cindex compiler options, C++
1329 @cindex C++ options, command line
1330 @cindex options, C++
1331 This section describes the command-line options that are only meaningful
1332 for C++ programs; but you can also use most of the GNU compiler options
1333 regardless of what language your program is in. For example, you
1334 might compile a file @code{firstClass.C} like this:
1337 g++ -g -frepo -O -c firstClass.C
1341 In this example, only @option{-frepo} is an option meant
1342 only for C++ programs; you can use the other options with any
1343 language supported by GCC@.
1345 Here is a list of options that are @emph{only} for compiling C++ programs:
1349 @item -fabi-version=@var{n}
1350 @opindex fabi-version
1351 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1352 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1353 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1354 the version that conforms most closely to the C++ ABI specification.
1355 Therefore, the ABI obtained using version 0 will change as ABI bugs
1358 The default is version 2.
1360 @item -fno-access-control
1361 @opindex fno-access-control
1362 Turn off all access checking. This switch is mainly useful for working
1363 around bugs in the access control code.
1367 Check that the pointer returned by @code{operator new} is non-null
1368 before attempting to modify the storage allocated. This check is
1369 normally unnecessary because the C++ standard specifies that
1370 @code{operator new} will only return @code{0} if it is declared
1371 @samp{throw()}, in which case the compiler will always check the
1372 return value even without this option. In all other cases, when
1373 @code{operator new} has a non-empty exception specification, memory
1374 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1375 @samp{new (nothrow)}.
1377 @item -fconserve-space
1378 @opindex fconserve-space
1379 Put uninitialized or runtime-initialized global variables into the
1380 common segment, as C does. This saves space in the executable at the
1381 cost of not diagnosing duplicate definitions. If you compile with this
1382 flag and your program mysteriously crashes after @code{main()} has
1383 completed, you may have an object that is being destroyed twice because
1384 two definitions were merged.
1386 This option is no longer useful on most targets, now that support has
1387 been added for putting variables into BSS without making them common.
1389 @item -fno-const-strings
1390 @opindex fno-const-strings
1391 Give string constants type @code{char *} instead of type @code{const
1392 char *}. By default, G++ uses type @code{const char *} as required by
1393 the standard. Even if you use @option{-fno-const-strings}, you cannot
1394 actually modify the value of a string constant.
1396 This option might be removed in a future release of G++. For maximum
1397 portability, you should structure your code so that it works with
1398 string constants that have type @code{const char *}.
1400 @item -fno-elide-constructors
1401 @opindex fno-elide-constructors
1402 The C++ standard allows an implementation to omit creating a temporary
1403 which is only used to initialize another object of the same type.
1404 Specifying this option disables that optimization, and forces G++ to
1405 call the copy constructor in all cases.
1407 @item -fno-enforce-eh-specs
1408 @opindex fno-enforce-eh-specs
1409 Don't check for violation of exception specifications at runtime. This
1410 option violates the C++ standard, but may be useful for reducing code
1411 size in production builds, much like defining @samp{NDEBUG}. The compiler
1412 will still optimize based on the exception specifications.
1415 @itemx -fno-for-scope
1417 @opindex fno-for-scope
1418 If @option{-ffor-scope} is specified, the scope of variables declared in
1419 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1420 as specified by the C++ standard.
1421 If @option{-fno-for-scope} is specified, the scope of variables declared in
1422 a @i{for-init-statement} extends to the end of the enclosing scope,
1423 as was the case in old versions of G++, and other (traditional)
1424 implementations of C++.
1426 The default if neither flag is given to follow the standard,
1427 but to allow and give a warning for old-style code that would
1428 otherwise be invalid, or have different behavior.
1430 @item -fno-gnu-keywords
1431 @opindex fno-gnu-keywords
1432 Do not recognize @code{typeof} as a keyword, so that code can use this
1433 word as an identifier. You can use the keyword @code{__typeof__} instead.
1434 @option{-ansi} implies @option{-fno-gnu-keywords}.
1436 @item -fno-implicit-templates
1437 @opindex fno-implicit-templates
1438 Never emit code for non-inline templates which are instantiated
1439 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1440 @xref{Template Instantiation}, for more information.
1442 @item -fno-implicit-inline-templates
1443 @opindex fno-implicit-inline-templates
1444 Don't emit code for implicit instantiations of inline templates, either.
1445 The default is to handle inlines differently so that compiles with and
1446 without optimization will need the same set of explicit instantiations.
1448 @item -fno-implement-inlines
1449 @opindex fno-implement-inlines
1450 To save space, do not emit out-of-line copies of inline functions
1451 controlled by @samp{#pragma implementation}. This will cause linker
1452 errors if these functions are not inlined everywhere they are called.
1454 @item -fms-extensions
1455 @opindex fms-extensions
1456 Disable pedantic warnings about constructs used in MFC, such as implicit
1457 int and getting a pointer to member function via non-standard syntax.
1459 @item -fno-nonansi-builtins
1460 @opindex fno-nonansi-builtins
1461 Disable built-in declarations of functions that are not mandated by
1462 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1463 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1465 @item -fno-operator-names
1466 @opindex fno-operator-names
1467 Do not treat the operator name keywords @code{and}, @code{bitand},
1468 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1469 synonyms as keywords.
1471 @item -fno-optional-diags
1472 @opindex fno-optional-diags
1473 Disable diagnostics that the standard says a compiler does not need to
1474 issue. Currently, the only such diagnostic issued by G++ is the one for
1475 a name having multiple meanings within a class.
1478 @opindex fpermissive
1479 Downgrade some diagnostics about nonconformant code from errors to
1480 warnings. Thus, using @option{-fpermissive} will allow some
1481 nonconforming code to compile.
1485 Enable automatic template instantiation at link time. This option also
1486 implies @option{-fno-implicit-templates}. @xref{Template
1487 Instantiation}, for more information.
1491 Disable generation of information about every class with virtual
1492 functions for use by the C++ runtime type identification features
1493 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1494 of the language, you can save some space by using this flag. Note that
1495 exception handling uses the same information, but it will generate it as
1500 Emit statistics about front-end processing at the end of the compilation.
1501 This information is generally only useful to the G++ development team.
1503 @item -ftemplate-depth-@var{n}
1504 @opindex ftemplate-depth
1505 Set the maximum instantiation depth for template classes to @var{n}.
1506 A limit on the template instantiation depth is needed to detect
1507 endless recursions during template class instantiation. ANSI/ISO C++
1508 conforming programs must not rely on a maximum depth greater than 17.
1510 @item -fno-threadsafe-statics
1511 @opindex fno-threadsafe-statics
1512 Do not emit the extra code to use the routines specified in the C++
1513 ABI for thread-safe initialization of local statics. You can use this
1514 option to reduce code size slightly in code that doesn't need to be
1517 @item -fuse-cxa-atexit
1518 @opindex fuse-cxa-atexit
1519 Register destructors for objects with static storage duration with the
1520 @code{__cxa_atexit} function rather than the @code{atexit} function.
1521 This option is required for fully standards-compliant handling of static
1522 destructors, but will only work if your C library supports
1523 @code{__cxa_atexit}.
1525 @item -fvisibility-inlines-hidden
1526 @opindex fvisibility-inlines-hidden
1527 Causes all inlined methods to be marked with
1528 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1529 appear in the export table of a DSO and do not require a PLT indirection
1530 when used within the DSO@. Enabling this option can have a dramatic effect
1531 on load and link times of a DSO as it massively reduces the size of the
1532 dynamic export table when the library makes heavy use of templates. While
1533 it can cause bloating through duplication of code within each DSO where
1534 it is used, often the wastage is less than the considerable space occupied
1535 by a long symbol name in the export table which is typical when using
1536 templates and namespaces. For even more savings, combine with the
1537 @option{-fvisibility=hidden} switch.
1541 Do not use weak symbol support, even if it is provided by the linker.
1542 By default, G++ will use weak symbols if they are available. This
1543 option exists only for testing, and should not be used by end-users;
1544 it will result in inferior code and has no benefits. This option may
1545 be removed in a future release of G++.
1549 Do not search for header files in the standard directories specific to
1550 C++, but do still search the other standard directories. (This option
1551 is used when building the C++ library.)
1554 In addition, these optimization, warning, and code generation options
1555 have meanings only for C++ programs:
1558 @item -fno-default-inline
1559 @opindex fno-default-inline
1560 Do not assume @samp{inline} for functions defined inside a class scope.
1561 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1562 functions will have linkage like inline functions; they just won't be
1565 @item -Wabi @r{(C++ only)}
1567 Warn when G++ generates code that is probably not compatible with the
1568 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1569 all such cases, there are probably some cases that are not warned about,
1570 even though G++ is generating incompatible code. There may also be
1571 cases where warnings are emitted even though the code that is generated
1574 You should rewrite your code to avoid these warnings if you are
1575 concerned about the fact that code generated by G++ may not be binary
1576 compatible with code generated by other compilers.
1578 The known incompatibilities at this point include:
1583 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1584 pack data into the same byte as a base class. For example:
1587 struct A @{ virtual void f(); int f1 : 1; @};
1588 struct B : public A @{ int f2 : 1; @};
1592 In this case, G++ will place @code{B::f2} into the same byte
1593 as@code{A::f1}; other compilers will not. You can avoid this problem
1594 by explicitly padding @code{A} so that its size is a multiple of the
1595 byte size on your platform; that will cause G++ and other compilers to
1596 layout @code{B} identically.
1599 Incorrect handling of tail-padding for virtual bases. G++ does not use
1600 tail padding when laying out virtual bases. For example:
1603 struct A @{ virtual void f(); char c1; @};
1604 struct B @{ B(); char c2; @};
1605 struct C : public A, public virtual B @{@};
1609 In this case, G++ will not place @code{B} into the tail-padding for
1610 @code{A}; other compilers will. You can avoid this problem by
1611 explicitly padding @code{A} so that its size is a multiple of its
1612 alignment (ignoring virtual base classes); that will cause G++ and other
1613 compilers to layout @code{C} identically.
1616 Incorrect handling of bit-fields with declared widths greater than that
1617 of their underlying types, when the bit-fields appear in a union. For
1621 union U @{ int i : 4096; @};
1625 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1626 union too small by the number of bits in an @code{int}.
1629 Empty classes can be placed at incorrect offsets. For example:
1639 struct C : public B, public A @{@};
1643 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1644 it should be placed at offset zero. G++ mistakenly believes that the
1645 @code{A} data member of @code{B} is already at offset zero.
1648 Names of template functions whose types involve @code{typename} or
1649 template template parameters can be mangled incorrectly.
1652 template <typename Q>
1653 void f(typename Q::X) @{@}
1655 template <template <typename> class Q>
1656 void f(typename Q<int>::X) @{@}
1660 Instantiations of these templates may be mangled incorrectly.
1664 @item -Wctor-dtor-privacy @r{(C++ only)}
1665 @opindex Wctor-dtor-privacy
1666 Warn when a class seems unusable because all the constructors or
1667 destructors in that class are private, and it has neither friends nor
1668 public static member functions.
1670 @item -Wnon-virtual-dtor @r{(C++ only)}
1671 @opindex Wnon-virtual-dtor
1672 Warn when a class appears to be polymorphic, thereby requiring a virtual
1673 destructor, yet it declares a non-virtual one.
1674 This warning is enabled by @option{-Wall}.
1676 @item -Wreorder @r{(C++ only)}
1678 @cindex reordering, warning
1679 @cindex warning for reordering of member initializers
1680 Warn when the order of member initializers given in the code does not
1681 match the order in which they must be executed. For instance:
1687 A(): j (0), i (1) @{ @}
1691 The compiler will rearrange the member initializers for @samp{i}
1692 and @samp{j} to match the declaration order of the members, emitting
1693 a warning to that effect. This warning is enabled by @option{-Wall}.
1696 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1699 @item -Weffc++ @r{(C++ only)}
1701 Warn about violations of the following style guidelines from Scott Meyers'
1702 @cite{Effective C++} book:
1706 Item 11: Define a copy constructor and an assignment operator for classes
1707 with dynamically allocated memory.
1710 Item 12: Prefer initialization to assignment in constructors.
1713 Item 14: Make destructors virtual in base classes.
1716 Item 15: Have @code{operator=} return a reference to @code{*this}.
1719 Item 23: Don't try to return a reference when you must return an object.
1723 Also warn about violations of the following style guidelines from
1724 Scott Meyers' @cite{More Effective C++} book:
1728 Item 6: Distinguish between prefix and postfix forms of increment and
1729 decrement operators.
1732 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1736 When selecting this option, be aware that the standard library
1737 headers do not obey all of these guidelines; use @samp{grep -v}
1738 to filter out those warnings.
1740 @item -Wno-deprecated @r{(C++ only)}
1741 @opindex Wno-deprecated
1742 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1744 @item -Wno-non-template-friend @r{(C++ only)}
1745 @opindex Wno-non-template-friend
1746 Disable warnings when non-templatized friend functions are declared
1747 within a template. Since the advent of explicit template specification
1748 support in G++, if the name of the friend is an unqualified-id (i.e.,
1749 @samp{friend foo(int)}), the C++ language specification demands that the
1750 friend declare or define an ordinary, nontemplate function. (Section
1751 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1752 could be interpreted as a particular specialization of a templatized
1753 function. Because this non-conforming behavior is no longer the default
1754 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1755 check existing code for potential trouble spots and is on by default.
1756 This new compiler behavior can be turned off with
1757 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1758 but disables the helpful warning.
1760 @item -Wold-style-cast @r{(C++ only)}
1761 @opindex Wold-style-cast
1762 Warn if an old-style (C-style) cast to a non-void type is used within
1763 a C++ program. The new-style casts (@samp{static_cast},
1764 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1765 unintended effects and much easier to search for.
1767 @item -Woverloaded-virtual @r{(C++ only)}
1768 @opindex Woverloaded-virtual
1769 @cindex overloaded virtual fn, warning
1770 @cindex warning for overloaded virtual fn
1771 Warn when a function declaration hides virtual functions from a
1772 base class. For example, in:
1779 struct B: public A @{
1784 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1792 will fail to compile.
1794 @item -Wno-pmf-conversions @r{(C++ only)}
1795 @opindex Wno-pmf-conversions
1796 Disable the diagnostic for converting a bound pointer to member function
1799 @item -Wsign-promo @r{(C++ only)}
1800 @opindex Wsign-promo
1801 Warn when overload resolution chooses a promotion from unsigned or
1802 enumerated type to a signed type, over a conversion to an unsigned type of
1803 the same size. Previous versions of G++ would try to preserve
1804 unsignedness, but the standard mandates the current behavior.
1809 A& operator = (int);
1819 In this example, G++ will synthesize a default @samp{A& operator =
1820 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1823 @node Objective-C and Objective-C++ Dialect Options
1824 @section Options Controlling Objective-C and Objective-C++ Dialects
1826 @cindex compiler options, Objective-C and Objective-C++
1827 @cindex Objective-C and Objective-C++ options, command line
1828 @cindex options, Objective-C and Objective-C++
1829 (NOTE: This manual does not describe the Objective-C and Objective-C++
1830 languages themselves. See @xref{Standards,,Language Standards
1831 Supported by GCC}, for references.)
1833 This section describes the command-line options that are only meaningful
1834 for Objective-C and Objective-C++ programs, but you can also use most of
1835 the language-independent GNU compiler options.
1836 For example, you might compile a file @code{some_class.m} like this:
1839 gcc -g -fgnu-runtime -O -c some_class.m
1843 In this example, @option{-fgnu-runtime} is an option meant only for
1844 Objective-C and Objective-C++ programs; you can use the other options with
1845 any language supported by GCC@.
1847 Note that since Objective-C is an extension of the C language, Objective-C
1848 compilations may also use options specific to the C front-end (e.g.,
1849 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1850 C++-specific options (e.g., @option{-Wabi}).
1852 Here is a list of options that are @emph{only} for compiling Objective-C
1853 and Objective-C++ programs:
1856 @item -fconstant-string-class=@var{class-name}
1857 @opindex fconstant-string-class
1858 Use @var{class-name} as the name of the class to instantiate for each
1859 literal string specified with the syntax @code{@@"@dots{}"}. The default
1860 class name is @code{NXConstantString} if the GNU runtime is being used, and
1861 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1862 @option{-fconstant-cfstrings} option, if also present, will override the
1863 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1864 to be laid out as constant CoreFoundation strings.
1867 @opindex fgnu-runtime
1868 Generate object code compatible with the standard GNU Objective-C
1869 runtime. This is the default for most types of systems.
1871 @item -fnext-runtime
1872 @opindex fnext-runtime
1873 Generate output compatible with the NeXT runtime. This is the default
1874 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1875 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1878 @item -fno-nil-receivers
1879 @opindex fno-nil-receivers
1880 Assume that all Objective-C message dispatches (e.g.,
1881 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1882 is not @code{nil}. This allows for more efficient entry points in the runtime
1883 to be used. Currently, this option is only available in conjunction with
1884 the NeXT runtime on Mac OS X 10.3 and later.
1886 @item -fobjc-exceptions
1887 @opindex fobjc-exceptions
1888 Enable syntactic support for structured exception handling in Objective-C,
1889 similar to what is offered by C++ and Java. Currently, this option is only
1890 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1898 @@catch (AnObjCClass *exc) @{
1905 @@catch (AnotherClass *exc) @{
1908 @@catch (id allOthers) @{
1918 The @code{@@throw} statement may appear anywhere in an Objective-C or
1919 Objective-C++ program; when used inside of a @code{@@catch} block, the
1920 @code{@@throw} may appear without an argument (as shown above), in which case
1921 the object caught by the @code{@@catch} will be rethrown.
1923 Note that only (pointers to) Objective-C objects may be thrown and
1924 caught using this scheme. When an object is thrown, it will be caught
1925 by the nearest @code{@@catch} clause capable of handling objects of that type,
1926 analogously to how @code{catch} blocks work in C++ and Java. A
1927 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1928 any and all Objective-C exceptions not caught by previous @code{@@catch}
1931 The @code{@@finally} clause, if present, will be executed upon exit from the
1932 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1933 regardless of whether any exceptions are thrown, caught or rethrown
1934 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1935 of the @code{finally} clause in Java.
1937 There are several caveats to using the new exception mechanism:
1941 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
1942 idioms provided by the @code{NSException} class, the new
1943 exceptions can only be used on Mac OS X 10.3 (Panther) and later
1944 systems, due to additional functionality needed in the (NeXT) Objective-C
1948 As mentioned above, the new exceptions do not support handling
1949 types other than Objective-C objects. Furthermore, when used from
1950 Objective-C++, the Objective-C exception model does not interoperate with C++
1951 exceptions at this time. This means you cannot @code{@@throw} an exception
1952 from Objective-C and @code{catch} it in C++, or vice versa
1953 (i.e., @code{throw @dots{} @@catch}).
1956 The @option{-fobjc-exceptions} switch also enables the use of synchronization
1957 blocks for thread-safe execution:
1960 @@synchronized (ObjCClass *guard) @{
1965 Upon entering the @code{@@synchronized} block, a thread of execution shall
1966 first check whether a lock has been placed on the corresponding @code{guard}
1967 object by another thread. If it has, the current thread shall wait until
1968 the other thread relinquishes its lock. Once @code{guard} becomes available,
1969 the current thread will place its own lock on it, execute the code contained in
1970 the @code{@@synchronized} block, and finally relinquish the lock (thereby
1971 making @code{guard} available to other threads).
1973 Unlike Java, Objective-C does not allow for entire methods to be marked
1974 @code{@@synchronized}. Note that throwing exceptions out of
1975 @code{@@synchronized} blocks is allowed, and will cause the guarding object
1976 to be unlocked properly.
1978 @item -freplace-objc-classes
1979 @opindex freplace-objc-classes
1980 Emit a special marker instructing @command{ld(1)} not to statically link in
1981 the resulting object file, and allow @command{dyld(1)} to load it in at
1982 run time instead. This is used in conjunction with the Fix-and-Continue
1983 debugging mode, where the object file in question may be recompiled and
1984 dynamically reloaded in the course of program execution, without the need
1985 to restart the program itself. Currently, Fix-and-Continue functionality
1986 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
1991 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
1992 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
1993 compile time) with static class references that get initialized at load time,
1994 which improves run-time performance. Specifying the @option{-fzero-link} flag
1995 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
1996 to be retained. This is useful in Zero-Link debugging mode, since it allows
1997 for individual class implementations to be modified during program execution.
2001 Dump interface declarations for all classes seen in the source file to a
2002 file named @file{@var{sourcename}.decl}.
2005 @opindex Wno-protocol
2006 If a class is declared to implement a protocol, a warning is issued for
2007 every method in the protocol that is not implemented by the class. The
2008 default behavior is to issue a warning for every method not explicitly
2009 implemented in the class, even if a method implementation is inherited
2010 from the superclass. If you use the @option{-Wno-protocol} option, then
2011 methods inherited from the superclass are considered to be implemented,
2012 and no warning is issued for them.
2016 Warn if multiple methods of different types for the same selector are
2017 found during compilation. The check is performed on the list of methods
2018 in the final stage of compilation. Additionally, a check is performed
2019 for each selector appearing in a @code{@@selector(@dots{})}
2020 expression, and a corresponding method for that selector has been found
2021 during compilation. Because these checks scan the method table only at
2022 the end of compilation, these warnings are not produced if the final
2023 stage of compilation is not reached, for example because an error is
2024 found during compilation, or because the @option{-fsyntax-only} option is
2027 @item -Wundeclared-selector
2028 @opindex Wundeclared-selector
2029 Warn if a @code{@@selector(@dots{})} expression referring to an
2030 undeclared selector is found. A selector is considered undeclared if no
2031 method with that name has been declared before the
2032 @code{@@selector(@dots{})} expression, either explicitly in an
2033 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2034 an @code{@@implementation} section. This option always performs its
2035 checks as soon as a @code{@@selector(@dots{})} expression is found,
2036 while @option{-Wselector} only performs its checks in the final stage of
2037 compilation. This also enforces the coding style convention
2038 that methods and selectors must be declared before being used.
2040 @item -print-objc-runtime-info
2041 @opindex print-objc-runtime-info
2042 Generate C header describing the largest structure that is passed by
2047 @node Language Independent Options
2048 @section Options to Control Diagnostic Messages Formatting
2049 @cindex options to control diagnostics formatting
2050 @cindex diagnostic messages
2051 @cindex message formatting
2053 Traditionally, diagnostic messages have been formatted irrespective of
2054 the output device's aspect (e.g.@: its width, @dots{}). The options described
2055 below can be used to control the diagnostic messages formatting
2056 algorithm, e.g.@: how many characters per line, how often source location
2057 information should be reported. Right now, only the C++ front end can
2058 honor these options. However it is expected, in the near future, that
2059 the remaining front ends would be able to digest them correctly.
2062 @item -fmessage-length=@var{n}
2063 @opindex fmessage-length
2064 Try to format error messages so that they fit on lines of about @var{n}
2065 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2066 the front ends supported by GCC@. If @var{n} is zero, then no
2067 line-wrapping will be done; each error message will appear on a single
2070 @opindex fdiagnostics-show-location
2071 @item -fdiagnostics-show-location=once
2072 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2073 reporter to emit @emph{once} source location information; that is, in
2074 case the message is too long to fit on a single physical line and has to
2075 be wrapped, the source location won't be emitted (as prefix) again,
2076 over and over, in subsequent continuation lines. This is the default
2079 @item -fdiagnostics-show-location=every-line
2080 Only meaningful in line-wrapping mode. Instructs the diagnostic
2081 messages reporter to emit the same source location information (as
2082 prefix) for physical lines that result from the process of breaking
2083 a message which is too long to fit on a single line.
2087 @node Warning Options
2088 @section Options to Request or Suppress Warnings
2089 @cindex options to control warnings
2090 @cindex warning messages
2091 @cindex messages, warning
2092 @cindex suppressing warnings
2094 Warnings are diagnostic messages that report constructions which
2095 are not inherently erroneous but which are risky or suggest there
2096 may have been an error.
2098 You can request many specific warnings with options beginning @samp{-W},
2099 for example @option{-Wimplicit} to request warnings on implicit
2100 declarations. Each of these specific warning options also has a
2101 negative form beginning @samp{-Wno-} to turn off warnings;
2102 for example, @option{-Wno-implicit}. This manual lists only one of the
2103 two forms, whichever is not the default.
2105 The following options control the amount and kinds of warnings produced
2106 by GCC; for further, language-specific options also refer to
2107 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2111 @cindex syntax checking
2113 @opindex fsyntax-only
2114 Check the code for syntax errors, but don't do anything beyond that.
2118 Issue all the warnings demanded by strict ISO C and ISO C++;
2119 reject all programs that use forbidden extensions, and some other
2120 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2121 version of the ISO C standard specified by any @option{-std} option used.
2123 Valid ISO C and ISO C++ programs should compile properly with or without
2124 this option (though a rare few will require @option{-ansi} or a
2125 @option{-std} option specifying the required version of ISO C)@. However,
2126 without this option, certain GNU extensions and traditional C and C++
2127 features are supported as well. With this option, they are rejected.
2129 @option{-pedantic} does not cause warning messages for use of the
2130 alternate keywords whose names begin and end with @samp{__}. Pedantic
2131 warnings are also disabled in the expression that follows
2132 @code{__extension__}. However, only system header files should use
2133 these escape routes; application programs should avoid them.
2134 @xref{Alternate Keywords}.
2136 Some users try to use @option{-pedantic} to check programs for strict ISO
2137 C conformance. They soon find that it does not do quite what they want:
2138 it finds some non-ISO practices, but not all---only those for which
2139 ISO C @emph{requires} a diagnostic, and some others for which
2140 diagnostics have been added.
2142 A feature to report any failure to conform to ISO C might be useful in
2143 some instances, but would require considerable additional work and would
2144 be quite different from @option{-pedantic}. We don't have plans to
2145 support such a feature in the near future.
2147 Where the standard specified with @option{-std} represents a GNU
2148 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2149 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2150 extended dialect is based. Warnings from @option{-pedantic} are given
2151 where they are required by the base standard. (It would not make sense
2152 for such warnings to be given only for features not in the specified GNU
2153 C dialect, since by definition the GNU dialects of C include all
2154 features the compiler supports with the given option, and there would be
2155 nothing to warn about.)
2157 @item -pedantic-errors
2158 @opindex pedantic-errors
2159 Like @option{-pedantic}, except that errors are produced rather than
2164 Inhibit all warning messages.
2168 Inhibit warning messages about the use of @samp{#import}.
2170 @item -Wchar-subscripts
2171 @opindex Wchar-subscripts
2172 Warn if an array subscript has type @code{char}. This is a common cause
2173 of error, as programmers often forget that this type is signed on some
2175 This warning is enabled by @option{-Wall}.
2179 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2180 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2181 This warning is enabled by @option{-Wall}.
2183 @item -Wfatal-errors
2184 @opindex Wfatal-errors
2185 This option causes the compiler to abort compilation on the first error
2186 occurred rather than trying to keep going and printing further error
2191 @opindex ffreestanding
2192 @opindex fno-builtin
2193 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2194 the arguments supplied have types appropriate to the format string
2195 specified, and that the conversions specified in the format string make
2196 sense. This includes standard functions, and others specified by format
2197 attributes (@pxref{Function Attributes}), in the @code{printf},
2198 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2199 not in the C standard) families (or other target-specific families).
2200 Which functions are checked without format attributes having been
2201 specified depends on the standard version selected, and such checks of
2202 functions without the attribute specified are disabled by
2203 @option{-ffreestanding} or @option{-fno-builtin}.
2205 The formats are checked against the format features supported by GNU
2206 libc version 2.2. These include all ISO C90 and C99 features, as well
2207 as features from the Single Unix Specification and some BSD and GNU
2208 extensions. Other library implementations may not support all these
2209 features; GCC does not support warning about features that go beyond a
2210 particular library's limitations. However, if @option{-pedantic} is used
2211 with @option{-Wformat}, warnings will be given about format features not
2212 in the selected standard version (but not for @code{strfmon} formats,
2213 since those are not in any version of the C standard). @xref{C Dialect
2214 Options,,Options Controlling C Dialect}.
2216 Since @option{-Wformat} also checks for null format arguments for
2217 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2219 @option{-Wformat} is included in @option{-Wall}. For more control over some
2220 aspects of format checking, the options @option{-Wformat-y2k},
2221 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2222 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2223 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2226 @opindex Wformat-y2k
2227 If @option{-Wformat} is specified, also warn about @code{strftime}
2228 formats which may yield only a two-digit year.
2230 @item -Wno-format-extra-args
2231 @opindex Wno-format-extra-args
2232 If @option{-Wformat} is specified, do not warn about excess arguments to a
2233 @code{printf} or @code{scanf} format function. The C standard specifies
2234 that such arguments are ignored.
2236 Where the unused arguments lie between used arguments that are
2237 specified with @samp{$} operand number specifications, normally
2238 warnings are still given, since the implementation could not know what
2239 type to pass to @code{va_arg} to skip the unused arguments. However,
2240 in the case of @code{scanf} formats, this option will suppress the
2241 warning if the unused arguments are all pointers, since the Single
2242 Unix Specification says that such unused arguments are allowed.
2244 @item -Wno-format-zero-length
2245 @opindex Wno-format-zero-length
2246 If @option{-Wformat} is specified, do not warn about zero-length formats.
2247 The C standard specifies that zero-length formats are allowed.
2249 @item -Wformat-nonliteral
2250 @opindex Wformat-nonliteral
2251 If @option{-Wformat} is specified, also warn if the format string is not a
2252 string literal and so cannot be checked, unless the format function
2253 takes its format arguments as a @code{va_list}.
2255 @item -Wformat-security
2256 @opindex Wformat-security
2257 If @option{-Wformat} is specified, also warn about uses of format
2258 functions that represent possible security problems. At present, this
2259 warns about calls to @code{printf} and @code{scanf} functions where the
2260 format string is not a string literal and there are no format arguments,
2261 as in @code{printf (foo);}. This may be a security hole if the format
2262 string came from untrusted input and contains @samp{%n}. (This is
2263 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2264 in future warnings may be added to @option{-Wformat-security} that are not
2265 included in @option{-Wformat-nonliteral}.)
2269 Enable @option{-Wformat} plus format checks not included in
2270 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2271 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2275 Warn about passing a null pointer for arguments marked as
2276 requiring a non-null value by the @code{nonnull} function attribute.
2278 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2279 can be disabled with the @option{-Wno-nonnull} option.
2281 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2283 Warn about uninitialized variables which are initialized with themselves.
2284 Note this option can only be used with the @option{-Wuninitialized} option,
2285 which in turn only works with @option{-O1} and above.
2287 For example, GCC will warn about @code{i} being uninitialized in the
2288 following snippet only when @option{-Winit-self} has been specified:
2299 @item -Wimplicit-int
2300 @opindex Wimplicit-int
2301 Warn when a declaration does not specify a type.
2302 This warning is enabled by @option{-Wall}.
2304 @item -Wimplicit-function-declaration
2305 @itemx -Werror-implicit-function-declaration
2306 @opindex Wimplicit-function-declaration
2307 @opindex Werror-implicit-function-declaration
2308 Give a warning (or error) whenever a function is used before being
2309 declared. The form @option{-Wno-error-implicit-function-declaration}
2311 This warning is enabled by @option{-Wall} (as a warning, not an error).
2315 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2316 This warning is enabled by @option{-Wall}.
2320 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2321 function with external linkage, returning int, taking either zero
2322 arguments, two, or three arguments of appropriate types.
2323 This warning is enabled by @option{-Wall}.
2325 @item -Wmissing-braces
2326 @opindex Wmissing-braces
2327 Warn if an aggregate or union initializer is not fully bracketed. In
2328 the following example, the initializer for @samp{a} is not fully
2329 bracketed, but that for @samp{b} is fully bracketed.
2332 int a[2][2] = @{ 0, 1, 2, 3 @};
2333 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2336 This warning is enabled by @option{-Wall}.
2338 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2339 @opindex Wmissing-include-dirs
2340 Warn if a user-supplied include directory does not exist.
2343 @opindex Wparentheses
2344 Warn if parentheses are omitted in certain contexts, such
2345 as when there is an assignment in a context where a truth value
2346 is expected, or when operators are nested whose precedence people
2347 often get confused about. Only the warning for an assignment used as
2348 a truth value is supported when compiling C++; the other warnings are
2349 only supported when compiling C@.
2351 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2352 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2353 interpretation from that of ordinary mathematical notation.
2355 Also warn about constructions where there may be confusion to which
2356 @code{if} statement an @code{else} branch belongs. Here is an example of
2371 In C, every @code{else} branch belongs to the innermost possible @code{if}
2372 statement, which in this example is @code{if (b)}. This is often not
2373 what the programmer expected, as illustrated in the above example by
2374 indentation the programmer chose. When there is the potential for this
2375 confusion, GCC will issue a warning when this flag is specified.
2376 To eliminate the warning, add explicit braces around the innermost
2377 @code{if} statement so there is no way the @code{else} could belong to
2378 the enclosing @code{if}. The resulting code would look like this:
2394 This warning is enabled by @option{-Wall}.
2396 @item -Wsequence-point
2397 @opindex Wsequence-point
2398 Warn about code that may have undefined semantics because of violations
2399 of sequence point rules in the C standard.
2401 The C standard defines the order in which expressions in a C program are
2402 evaluated in terms of @dfn{sequence points}, which represent a partial
2403 ordering between the execution of parts of the program: those executed
2404 before the sequence point, and those executed after it. These occur
2405 after the evaluation of a full expression (one which is not part of a
2406 larger expression), after the evaluation of the first operand of a
2407 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2408 function is called (but after the evaluation of its arguments and the
2409 expression denoting the called function), and in certain other places.
2410 Other than as expressed by the sequence point rules, the order of
2411 evaluation of subexpressions of an expression is not specified. All
2412 these rules describe only a partial order rather than a total order,
2413 since, for example, if two functions are called within one expression
2414 with no sequence point between them, the order in which the functions
2415 are called is not specified. However, the standards committee have
2416 ruled that function calls do not overlap.
2418 It is not specified when between sequence points modifications to the
2419 values of objects take effect. Programs whose behavior depends on this
2420 have undefined behavior; the C standard specifies that ``Between the
2421 previous and next sequence point an object shall have its stored value
2422 modified at most once by the evaluation of an expression. Furthermore,
2423 the prior value shall be read only to determine the value to be
2424 stored.''. If a program breaks these rules, the results on any
2425 particular implementation are entirely unpredictable.
2427 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2428 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2429 diagnosed by this option, and it may give an occasional false positive
2430 result, but in general it has been found fairly effective at detecting
2431 this sort of problem in programs.
2433 The present implementation of this option only works for C programs. A
2434 future implementation may also work for C++ programs.
2436 The C standard is worded confusingly, therefore there is some debate
2437 over the precise meaning of the sequence point rules in subtle cases.
2438 Links to discussions of the problem, including proposed formal
2439 definitions, may be found on the GCC readings page, at
2440 @w{@uref{http://gcc.gnu.org/readings.html}}.
2442 This warning is enabled by @option{-Wall}.
2445 @opindex Wreturn-type
2446 Warn whenever a function is defined with a return-type that defaults to
2447 @code{int}. Also warn about any @code{return} statement with no
2448 return-value in a function whose return-type is not @code{void}.
2450 For C, also warn if the return type of a function has a type qualifier
2451 such as @code{const}. Such a type qualifier has no effect, since the
2452 value returned by a function is not an lvalue. ISO C prohibits
2453 qualified @code{void} return types on function definitions, so such
2454 return types always receive a warning even without this option.
2456 For C++, a function without return type always produces a diagnostic
2457 message, even when @option{-Wno-return-type} is specified. The only
2458 exceptions are @samp{main} and functions defined in system headers.
2460 This warning is enabled by @option{-Wall}.
2464 Warn whenever a @code{switch} statement has an index of enumerated type
2465 and lacks a @code{case} for one or more of the named codes of that
2466 enumeration. (The presence of a @code{default} label prevents this
2467 warning.) @code{case} labels outside the enumeration range also
2468 provoke warnings when this option is used.
2469 This warning is enabled by @option{-Wall}.
2471 @item -Wswitch-default
2472 @opindex Wswitch-switch
2473 Warn whenever a @code{switch} statement does not have a @code{default}
2477 @opindex Wswitch-enum
2478 Warn whenever a @code{switch} statement has an index of enumerated type
2479 and lacks a @code{case} for one or more of the named codes of that
2480 enumeration. @code{case} labels outside the enumeration range also
2481 provoke warnings when this option is used.
2485 Warn if any trigraphs are encountered that might change the meaning of
2486 the program (trigraphs within comments are not warned about).
2487 This warning is enabled by @option{-Wall}.
2489 @item -Wunused-function
2490 @opindex Wunused-function
2491 Warn whenever a static function is declared but not defined or a
2492 non\-inline static function is unused.
2493 This warning is enabled by @option{-Wall}.
2495 @item -Wunused-label
2496 @opindex Wunused-label
2497 Warn whenever a label is declared but not used.
2498 This warning is enabled by @option{-Wall}.
2500 To suppress this warning use the @samp{unused} attribute
2501 (@pxref{Variable Attributes}).
2503 @item -Wunused-parameter
2504 @opindex Wunused-parameter
2505 Warn whenever a function parameter is unused aside from its declaration.
2507 To suppress this warning use the @samp{unused} attribute
2508 (@pxref{Variable Attributes}).
2510 @item -Wunused-variable
2511 @opindex Wunused-variable
2512 Warn whenever a local variable or non-constant static variable is unused
2513 aside from its declaration
2514 This warning is enabled by @option{-Wall}.
2516 To suppress this warning use the @samp{unused} attribute
2517 (@pxref{Variable Attributes}).
2519 @item -Wunused-value
2520 @opindex Wunused-value
2521 Warn whenever a statement computes a result that is explicitly not used.
2522 This warning is enabled by @option{-Wall}.
2524 To suppress this warning cast the expression to @samp{void}.
2528 All the above @option{-Wunused} options combined.
2530 In order to get a warning about an unused function parameter, you must
2531 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2532 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2534 @item -Wuninitialized
2535 @opindex Wuninitialized
2536 Warn if an automatic variable is used without first being initialized or
2537 if a variable may be clobbered by a @code{setjmp} call.
2539 These warnings are possible only in optimizing compilation,
2540 because they require data flow information that is computed only
2541 when optimizing. If you don't specify @option{-O}, you simply won't
2544 If you want to warn about code which uses the uninitialized value of the
2545 variable in its own initializer, use the @option{-Winit-self} option.
2547 These warnings occur only for variables that are candidates for
2548 register allocation. Therefore, they do not occur for a variable that
2549 is declared @code{volatile}, or whose address is taken, or whose size
2550 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
2551 structures, unions or arrays, even when they are in registers.
2553 Note that there may be no warning about a variable that is used only
2554 to compute a value that itself is never used, because such
2555 computations may be deleted by data flow analysis before the warnings
2558 These warnings are made optional because GCC is not smart
2559 enough to see all the reasons why the code might be correct
2560 despite appearing to have an error. Here is one example of how
2581 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2582 always initialized, but GCC doesn't know this. Here is
2583 another common case:
2588 if (change_y) save_y = y, y = new_y;
2590 if (change_y) y = save_y;
2595 This has no bug because @code{save_y} is used only if it is set.
2597 @cindex @code{longjmp} warnings
2598 This option also warns when a non-volatile automatic variable might be
2599 changed by a call to @code{longjmp}. These warnings as well are possible
2600 only in optimizing compilation.
2602 The compiler sees only the calls to @code{setjmp}. It cannot know
2603 where @code{longjmp} will be called; in fact, a signal handler could
2604 call it at any point in the code. As a result, you may get a warning
2605 even when there is in fact no problem because @code{longjmp} cannot
2606 in fact be called at the place which would cause a problem.
2608 Some spurious warnings can be avoided if you declare all the functions
2609 you use that never return as @code{noreturn}. @xref{Function
2612 This warning is enabled by @option{-Wall}.
2614 @item -Wunknown-pragmas
2615 @opindex Wunknown-pragmas
2616 @cindex warning for unknown pragmas
2617 @cindex unknown pragmas, warning
2618 @cindex pragmas, warning of unknown
2619 Warn when a #pragma directive is encountered which is not understood by
2620 GCC@. If this command line option is used, warnings will even be issued
2621 for unknown pragmas in system header files. This is not the case if
2622 the warnings were only enabled by the @option{-Wall} command line option.
2624 @item -Wstrict-aliasing
2625 @opindex Wstrict-aliasing
2626 This option is only active when @option{-fstrict-aliasing} is active.
2627 It warns about code which might break the strict aliasing rules that the
2628 compiler is using for optimization. The warning does not catch all
2629 cases, but does attempt to catch the more common pitfalls. It is
2630 included in @option{-Wall}.
2632 @item -Wstrict-aliasing=2
2633 @opindex Wstrict-aliasing=2
2634 This option is only active when @option{-fstrict-aliasing} is active.
2635 It warns about all code which might break the strict aliasing rules that the
2636 compiler is using for optimization. This warning catches all cases, but
2637 it will also give a warning for some ambiguous cases that are safe.
2641 All of the above @samp{-W} options combined. This enables all the
2642 warnings about constructions that some users consider questionable, and
2643 that are easy to avoid (or modify to prevent the warning), even in
2644 conjunction with macros. This also enables some language-specific
2645 warnings described in @ref{C++ Dialect Options} and
2646 @ref{Objective-C and Objective-C++ Dialect Options}.
2649 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2650 Some of them warn about constructions that users generally do not
2651 consider questionable, but which occasionally you might wish to check
2652 for; others warn about constructions that are necessary or hard to avoid
2653 in some cases, and there is no simple way to modify the code to suppress
2660 (This option used to be called @option{-W}. The older name is still
2661 supported, but the newer name is more descriptive.) Print extra warning
2662 messages for these events:
2666 A function can return either with or without a value. (Falling
2667 off the end of the function body is considered returning without
2668 a value.) For example, this function would evoke such a
2682 An expression-statement or the left-hand side of a comma expression
2683 contains no side effects.
2684 To suppress the warning, cast the unused expression to void.
2685 For example, an expression such as @samp{x[i,j]} will cause a warning,
2686 but @samp{x[(void)i,j]} will not.
2689 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2692 Storage-class specifiers like @code{static} are not the first things in
2693 a declaration. According to the C Standard, this usage is obsolescent.
2696 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2700 A comparison between signed and unsigned values could produce an
2701 incorrect result when the signed value is converted to unsigned.
2702 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2705 An aggregate has an initializer which does not initialize all members.
2706 This warning can be independently controlled by
2707 @option{-Wmissing-field-initializers}.
2710 A function parameter is declared without a type specifier in K&R-style
2718 An empty body occurs in an @samp{if} or @samp{else} statement.
2721 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2722 @samp{>}, or @samp{>=}.
2725 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2728 Any of several floating-point events that often indicate errors, such as
2729 overflow, underflow, loss of precision, etc.
2731 @item @r{(C++ only)}
2732 An enumerator and a non-enumerator both appear in a conditional expression.
2734 @item @r{(C++ only)}
2735 A non-static reference or non-static @samp{const} member appears in a
2736 class without constructors.
2738 @item @r{(C++ only)}
2739 Ambiguous virtual bases.
2741 @item @r{(C++ only)}
2742 Subscripting an array which has been declared @samp{register}.
2744 @item @r{(C++ only)}
2745 Taking the address of a variable which has been declared @samp{register}.
2747 @item @r{(C++ only)}
2748 A base class is not initialized in a derived class' copy constructor.
2751 @item -Wno-div-by-zero
2752 @opindex Wno-div-by-zero
2753 @opindex Wdiv-by-zero
2754 Do not warn about compile-time integer division by zero. Floating point
2755 division by zero is not warned about, as it can be a legitimate way of
2756 obtaining infinities and NaNs.
2758 @item -Wsystem-headers
2759 @opindex Wsystem-headers
2760 @cindex warnings from system headers
2761 @cindex system headers, warnings from
2762 Print warning messages for constructs found in system header files.
2763 Warnings from system headers are normally suppressed, on the assumption
2764 that they usually do not indicate real problems and would only make the
2765 compiler output harder to read. Using this command line option tells
2766 GCC to emit warnings from system headers as if they occurred in user
2767 code. However, note that using @option{-Wall} in conjunction with this
2768 option will @emph{not} warn about unknown pragmas in system
2769 headers---for that, @option{-Wunknown-pragmas} must also be used.
2772 @opindex Wfloat-equal
2773 Warn if floating point values are used in equality comparisons.
2775 The idea behind this is that sometimes it is convenient (for the
2776 programmer) to consider floating-point values as approximations to
2777 infinitely precise real numbers. If you are doing this, then you need
2778 to compute (by analyzing the code, or in some other way) the maximum or
2779 likely maximum error that the computation introduces, and allow for it
2780 when performing comparisons (and when producing output, but that's a
2781 different problem). In particular, instead of testing for equality, you
2782 would check to see whether the two values have ranges that overlap; and
2783 this is done with the relational operators, so equality comparisons are
2786 @item -Wtraditional @r{(C only)}
2787 @opindex Wtraditional
2788 Warn about certain constructs that behave differently in traditional and
2789 ISO C@. Also warn about ISO C constructs that have no traditional C
2790 equivalent, and/or problematic constructs which should be avoided.
2794 Macro parameters that appear within string literals in the macro body.
2795 In traditional C macro replacement takes place within string literals,
2796 but does not in ISO C@.
2799 In traditional C, some preprocessor directives did not exist.
2800 Traditional preprocessors would only consider a line to be a directive
2801 if the @samp{#} appeared in column 1 on the line. Therefore
2802 @option{-Wtraditional} warns about directives that traditional C
2803 understands but would ignore because the @samp{#} does not appear as the
2804 first character on the line. It also suggests you hide directives like
2805 @samp{#pragma} not understood by traditional C by indenting them. Some
2806 traditional implementations would not recognize @samp{#elif}, so it
2807 suggests avoiding it altogether.
2810 A function-like macro that appears without arguments.
2813 The unary plus operator.
2816 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2817 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2818 constants.) Note, these suffixes appear in macros defined in the system
2819 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2820 Use of these macros in user code might normally lead to spurious
2821 warnings, however GCC's integrated preprocessor has enough context to
2822 avoid warning in these cases.
2825 A function declared external in one block and then used after the end of
2829 A @code{switch} statement has an operand of type @code{long}.
2832 A non-@code{static} function declaration follows a @code{static} one.
2833 This construct is not accepted by some traditional C compilers.
2836 The ISO type of an integer constant has a different width or
2837 signedness from its traditional type. This warning is only issued if
2838 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2839 typically represent bit patterns, are not warned about.
2842 Usage of ISO string concatenation is detected.
2845 Initialization of automatic aggregates.
2848 Identifier conflicts with labels. Traditional C lacks a separate
2849 namespace for labels.
2852 Initialization of unions. If the initializer is zero, the warning is
2853 omitted. This is done under the assumption that the zero initializer in
2854 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2855 initializer warnings and relies on default initialization to zero in the
2859 Conversions by prototypes between fixed/floating point values and vice
2860 versa. The absence of these prototypes when compiling with traditional
2861 C would cause serious problems. This is a subset of the possible
2862 conversion warnings, for the full set use @option{-Wconversion}.
2865 Use of ISO C style function definitions. This warning intentionally is
2866 @emph{not} issued for prototype declarations or variadic functions
2867 because these ISO C features will appear in your code when using
2868 libiberty's traditional C compatibility macros, @code{PARAMS} and
2869 @code{VPARAMS}. This warning is also bypassed for nested functions
2870 because that feature is already a GCC extension and thus not relevant to
2871 traditional C compatibility.
2874 @item -Wdeclaration-after-statement @r{(C only)}
2875 @opindex Wdeclaration-after-statement
2876 Warn when a declaration is found after a statement in a block. This
2877 construct, known from C++, was introduced with ISO C99 and is by default
2878 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2879 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2883 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2885 @item -Wno-endif-labels
2886 @opindex Wno-endif-labels
2887 @opindex Wendif-labels
2888 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2892 Warn whenever a local variable shadows another local variable, parameter or
2893 global variable or whenever a built-in function is shadowed.
2895 @item -Wlarger-than-@var{len}
2896 @opindex Wlarger-than
2897 Warn whenever an object of larger than @var{len} bytes is defined.
2899 @item -Wpointer-arith
2900 @opindex Wpointer-arith
2901 Warn about anything that depends on the ``size of'' a function type or
2902 of @code{void}. GNU C assigns these types a size of 1, for
2903 convenience in calculations with @code{void *} pointers and pointers
2906 @item -Wbad-function-cast @r{(C only)}
2907 @opindex Wbad-function-cast
2908 Warn whenever a function call is cast to a non-matching type.
2909 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2913 Warn whenever a pointer is cast so as to remove a type qualifier from
2914 the target type. For example, warn if a @code{const char *} is cast
2915 to an ordinary @code{char *}.
2918 @opindex Wcast-align
2919 Warn whenever a pointer is cast such that the required alignment of the
2920 target is increased. For example, warn if a @code{char *} is cast to
2921 an @code{int *} on machines where integers can only be accessed at
2922 two- or four-byte boundaries.
2924 @item -Wwrite-strings
2925 @opindex Wwrite-strings
2926 When compiling C, give string constants the type @code{const
2927 char[@var{length}]} so that
2928 copying the address of one into a non-@code{const} @code{char *}
2929 pointer will get a warning; when compiling C++, warn about the
2930 deprecated conversion from string constants to @code{char *}.
2931 These warnings will help you find at
2932 compile time code that can try to write into a string constant, but
2933 only if you have been very careful about using @code{const} in
2934 declarations and prototypes. Otherwise, it will just be a nuisance;
2935 this is why we did not make @option{-Wall} request these warnings.
2938 @opindex Wconversion
2939 Warn if a prototype causes a type conversion that is different from what
2940 would happen to the same argument in the absence of a prototype. This
2941 includes conversions of fixed point to floating and vice versa, and
2942 conversions changing the width or signedness of a fixed point argument
2943 except when the same as the default promotion.
2945 Also, warn if a negative integer constant expression is implicitly
2946 converted to an unsigned type. For example, warn about the assignment
2947 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2948 casts like @code{(unsigned) -1}.
2950 @item -Wsign-compare
2951 @opindex Wsign-compare
2952 @cindex warning for comparison of signed and unsigned values
2953 @cindex comparison of signed and unsigned values, warning
2954 @cindex signed and unsigned values, comparison warning
2955 Warn when a comparison between signed and unsigned values could produce
2956 an incorrect result when the signed value is converted to unsigned.
2957 This warning is also enabled by @option{-Wextra}; to get the other warnings
2958 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2960 @item -Waggregate-return
2961 @opindex Waggregate-return
2962 Warn if any functions that return structures or unions are defined or
2963 called. (In languages where you can return an array, this also elicits
2966 @item -Wstrict-prototypes @r{(C only)}
2967 @opindex Wstrict-prototypes
2968 Warn if a function is declared or defined without specifying the
2969 argument types. (An old-style function definition is permitted without
2970 a warning if preceded by a declaration which specifies the argument
2973 @item -Wold-style-definition @r{(C only)}
2974 @opindex Wold-style-definition
2975 Warn if an old-style function definition is used. A warning is given
2976 even if there is a previous prototype.
2978 @item -Wmissing-prototypes @r{(C only)}
2979 @opindex Wmissing-prototypes
2980 Warn if a global function is defined without a previous prototype
2981 declaration. This warning is issued even if the definition itself
2982 provides a prototype. The aim is to detect global functions that fail
2983 to be declared in header files.
2985 @item -Wmissing-declarations @r{(C only)}
2986 @opindex Wmissing-declarations
2987 Warn if a global function is defined without a previous declaration.
2988 Do so even if the definition itself provides a prototype.
2989 Use this option to detect global functions that are not declared in
2992 @item -Wmissing-field-initializers
2993 @opindex Wmissing-field-initializers
2996 Warn if a structure's initializer has some fields missing. For
2997 example, the following code would cause such a warning, because
2998 @code{x.h} is implicitly zero:
3001 struct s @{ int f, g, h; @};
3002 struct s x = @{ 3, 4 @};
3005 This option does not warn about designated initializers, so the following
3006 modification would not trigger a warning:
3009 struct s @{ int f, g, h; @};
3010 struct s x = @{ .f = 3, .g = 4 @};
3013 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3014 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3016 @item -Wmissing-noreturn
3017 @opindex Wmissing-noreturn
3018 Warn about functions which might be candidates for attribute @code{noreturn}.
3019 Note these are only possible candidates, not absolute ones. Care should
3020 be taken to manually verify functions actually do not ever return before
3021 adding the @code{noreturn} attribute, otherwise subtle code generation
3022 bugs could be introduced. You will not get a warning for @code{main} in
3023 hosted C environments.
3025 @item -Wmissing-format-attribute
3026 @opindex Wmissing-format-attribute
3028 If @option{-Wformat} is enabled, also warn about functions which might be
3029 candidates for @code{format} attributes. Note these are only possible
3030 candidates, not absolute ones. GCC will guess that @code{format}
3031 attributes might be appropriate for any function that calls a function
3032 like @code{vprintf} or @code{vscanf}, but this might not always be the
3033 case, and some functions for which @code{format} attributes are
3034 appropriate may not be detected. This option has no effect unless
3035 @option{-Wformat} is enabled (possibly by @option{-Wall}).
3037 @item -Wno-multichar
3038 @opindex Wno-multichar
3040 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3041 Usually they indicate a typo in the user's code, as they have
3042 implementation-defined values, and should not be used in portable code.
3044 @item -Wnormalized=<none|id|nfc|nfkc>
3045 @opindex Wnormalized
3048 @cindex character set, input normalization
3049 In ISO C and ISO C++, two identifiers are different if they are
3050 different sequences of characters. However, sometimes when characters
3051 outside the basic ASCII character set are used, you can have two
3052 different character sequences that look the same. To avoid confusion,
3053 the ISO 10646 standard sets out some @dfn{normalization rules} which
3054 when applied ensure that two sequences that look the same are turned into
3055 the same sequence. GCC can warn you if you are using identifiers which
3056 have not been normalized; this option controls that warning.
3058 There are four levels of warning that GCC supports. The default is
3059 @option{-Wnormalized=nfc}, which warns about any identifier which is
3060 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3061 recommended form for most uses.
3063 Unfortunately, there are some characters which ISO C and ISO C++ allow
3064 in identifiers that when turned into NFC aren't allowable as
3065 identifiers. That is, there's no way to use these symbols in portable
3066 ISO C or C++ and have all your identifiers in NFC.
3067 @option{-Wnormalized=id} suppresses the warning for these characters.
3068 It is hoped that future versions of the standards involved will correct
3069 this, which is why this option is not the default.
3071 You can switch the warning off for all characters by writing
3072 @option{-Wnormalized=none}. You would only want to do this if you
3073 were using some other normalization scheme (like ``D''), because
3074 otherwise you can easily create bugs that are literally impossible to see.
3076 Some characters in ISO 10646 have distinct meanings but look identical
3077 in some fonts or display methodologies, especially once formatting has
3078 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3079 LETTER N'', will display just like a regular @code{n} which has been
3080 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3081 normalisation scheme to convert all these into a standard form as
3082 well, and GCC will warn if your code is not in NFKC if you use
3083 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3084 about every identifier that contains the letter O because it might be
3085 confused with the digit 0, and so is not the default, but may be
3086 useful as a local coding convention if the programming environment is
3087 unable to be fixed to display these characters distinctly.
3089 @item -Wno-deprecated-declarations
3090 @opindex Wno-deprecated-declarations
3091 Do not warn about uses of functions, variables, and types marked as
3092 deprecated by using the @code{deprecated} attribute.
3093 (@pxref{Function Attributes}, @pxref{Variable Attributes},
3094 @pxref{Type Attributes}.)
3098 Warn if a structure is given the packed attribute, but the packed
3099 attribute has no effect on the layout or size of the structure.
3100 Such structures may be mis-aligned for little benefit. For
3101 instance, in this code, the variable @code{f.x} in @code{struct bar}
3102 will be misaligned even though @code{struct bar} does not itself
3103 have the packed attribute:
3110 @} __attribute__((packed));
3120 Warn if padding is included in a structure, either to align an element
3121 of the structure or to align the whole structure. Sometimes when this
3122 happens it is possible to rearrange the fields of the structure to
3123 reduce the padding and so make the structure smaller.
3125 @item -Wredundant-decls
3126 @opindex Wredundant-decls
3127 Warn if anything is declared more than once in the same scope, even in
3128 cases where multiple declaration is valid and changes nothing.
3130 @item -Wnested-externs @r{(C only)}
3131 @opindex Wnested-externs
3132 Warn if an @code{extern} declaration is encountered within a function.
3134 @item -Wunreachable-code
3135 @opindex Wunreachable-code
3136 Warn if the compiler detects that code will never be executed.
3138 This option is intended to warn when the compiler detects that at
3139 least a whole line of source code will never be executed, because
3140 some condition is never satisfied or because it is after a
3141 procedure that never returns.
3143 It is possible for this option to produce a warning even though there
3144 are circumstances under which part of the affected line can be executed,
3145 so care should be taken when removing apparently-unreachable code.
3147 For instance, when a function is inlined, a warning may mean that the
3148 line is unreachable in only one inlined copy of the function.
3150 This option is not made part of @option{-Wall} because in a debugging
3151 version of a program there is often substantial code which checks
3152 correct functioning of the program and is, hopefully, unreachable
3153 because the program does work. Another common use of unreachable
3154 code is to provide behavior which is selectable at compile-time.
3158 Warn if a function can not be inlined and it was declared as inline.
3159 Even with this option, the compiler will not warn about failures to
3160 inline functions declared in system headers.
3162 The compiler uses a variety of heuristics to determine whether or not
3163 to inline a function. For example, the compiler takes into account
3164 the size of the function being inlined and the amount of inlining
3165 that has already been done in the current function. Therefore,
3166 seemingly insignificant changes in the source program can cause the
3167 warnings produced by @option{-Winline} to appear or disappear.
3169 @item -Wno-invalid-offsetof @r{(C++ only)}
3170 @opindex Wno-invalid-offsetof
3171 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3172 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3173 to a non-POD type is undefined. In existing C++ implementations,
3174 however, @samp{offsetof} typically gives meaningful results even when
3175 applied to certain kinds of non-POD types. (Such as a simple
3176 @samp{struct} that fails to be a POD type only by virtue of having a
3177 constructor.) This flag is for users who are aware that they are
3178 writing nonportable code and who have deliberately chosen to ignore the
3181 The restrictions on @samp{offsetof} may be relaxed in a future version
3182 of the C++ standard.
3184 @item -Wno-int-to-pointer-cast @r{(C only)}
3185 @opindex Wno-int-to-pointer-cast
3186 Suppress warnings from casts to pointer type of an integer of a
3189 @item -Wno-pointer-to-int-cast @r{(C only)}
3190 @opindex Wno-pointer-to-int-cast
3191 Suppress warnings from casts from a pointer to an integer type of a
3195 @opindex Winvalid-pch
3196 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3197 the search path but can't be used.
3201 @opindex Wno-long-long
3202 Warn if @samp{long long} type is used. This is default. To inhibit
3203 the warning messages, use @option{-Wno-long-long}. Flags
3204 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3205 only when @option{-pedantic} flag is used.
3207 @item -Wvariadic-macros
3208 @opindex Wvariadic-macros
3209 @opindex Wno-variadic-macros
3210 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3211 alternate syntax when in pedantic ISO C99 mode. This is default.
3212 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3214 @item -Wdisabled-optimization
3215 @opindex Wdisabled-optimization
3216 Warn if a requested optimization pass is disabled. This warning does
3217 not generally indicate that there is anything wrong with your code; it
3218 merely indicates that GCC's optimizers were unable to handle the code
3219 effectively. Often, the problem is that your code is too big or too
3220 complex; GCC will refuse to optimize programs when the optimization
3221 itself is likely to take inordinate amounts of time.
3223 @item -Wno-pointer-sign
3224 @opindex Wno-pointer-sign
3225 Don't warn for pointer argument passing or assignment with different signedness.
3226 Only useful in the negative form since this warning is enabled by default.
3227 This option is only supported for C and Objective-C@.
3231 Make all warnings into errors.
3234 @node Debugging Options
3235 @section Options for Debugging Your Program or GCC
3236 @cindex options, debugging
3237 @cindex debugging information options
3239 GCC has various special options that are used for debugging
3240 either your program or GCC:
3245 Produce debugging information in the operating system's native format
3246 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3249 On most systems that use stabs format, @option{-g} enables use of extra
3250 debugging information that only GDB can use; this extra information
3251 makes debugging work better in GDB but will probably make other debuggers
3253 refuse to read the program. If you want to control for certain whether
3254 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3255 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3257 GCC allows you to use @option{-g} with
3258 @option{-O}. The shortcuts taken by optimized code may occasionally
3259 produce surprising results: some variables you declared may not exist
3260 at all; flow of control may briefly move where you did not expect it;
3261 some statements may not be executed because they compute constant
3262 results or their values were already at hand; some statements may
3263 execute in different places because they were moved out of loops.
3265 Nevertheless it proves possible to debug optimized output. This makes
3266 it reasonable to use the optimizer for programs that might have bugs.
3268 The following options are useful when GCC is generated with the
3269 capability for more than one debugging format.
3273 Produce debugging information for use by GDB@. This means to use the
3274 most expressive format available (DWARF 2, stabs, or the native format
3275 if neither of those are supported), including GDB extensions if at all
3280 Produce debugging information in stabs format (if that is supported),
3281 without GDB extensions. This is the format used by DBX on most BSD
3282 systems. On MIPS, Alpha and System V Release 4 systems this option
3283 produces stabs debugging output which is not understood by DBX or SDB@.
3284 On System V Release 4 systems this option requires the GNU assembler.
3286 @item -feliminate-unused-debug-symbols
3287 @opindex feliminate-unused-debug-symbols
3288 Produce debugging information in stabs format (if that is supported),
3289 for only symbols that are actually used.
3293 Produce debugging information in stabs format (if that is supported),
3294 using GNU extensions understood only by the GNU debugger (GDB)@. The
3295 use of these extensions is likely to make other debuggers crash or
3296 refuse to read the program.
3300 Produce debugging information in COFF format (if that is supported).
3301 This is the format used by SDB on most System V systems prior to
3306 Produce debugging information in XCOFF format (if that is supported).
3307 This is the format used by the DBX debugger on IBM RS/6000 systems.
3311 Produce debugging information in XCOFF format (if that is supported),
3312 using GNU extensions understood only by the GNU debugger (GDB)@. The
3313 use of these extensions is likely to make other debuggers crash or
3314 refuse to read the program, and may cause assemblers other than the GNU
3315 assembler (GAS) to fail with an error.
3319 Produce debugging information in DWARF version 2 format (if that is
3320 supported). This is the format used by DBX on IRIX 6. With this
3321 option, GCC uses features of DWARF version 3 when they are useful;
3322 version 3 is upward compatible with version 2, but may still cause
3323 problems for older debuggers.
3327 Produce debugging information in VMS debug format (if that is
3328 supported). This is the format used by DEBUG on VMS systems.
3331 @itemx -ggdb@var{level}
3332 @itemx -gstabs@var{level}
3333 @itemx -gcoff@var{level}
3334 @itemx -gxcoff@var{level}
3335 @itemx -gvms@var{level}
3336 Request debugging information and also use @var{level} to specify how
3337 much information. The default level is 2.
3339 Level 1 produces minimal information, enough for making backtraces in
3340 parts of the program that you don't plan to debug. This includes
3341 descriptions of functions and external variables, but no information
3342 about local variables and no line numbers.
3344 Level 3 includes extra information, such as all the macro definitions
3345 present in the program. Some debuggers support macro expansion when
3346 you use @option{-g3}.
3348 @option{-gdwarf-2} does not accept a concatenated debug level, because
3349 GCC used to support an option @option{-gdwarf} that meant to generate
3350 debug information in version 1 of the DWARF format (which is very
3351 different from version 2), and it would have been too confusing. That
3352 debug format is long obsolete, but the option cannot be changed now.
3353 Instead use an additional @option{-g@var{level}} option to change the
3354 debug level for DWARF2.
3356 @item -feliminate-dwarf2-dups
3357 @opindex feliminate-dwarf2-dups
3358 Compress DWARF2 debugging information by eliminating duplicated
3359 information about each symbol. This option only makes sense when
3360 generating DWARF2 debugging information with @option{-gdwarf-2}.
3362 @cindex @command{prof}
3365 Generate extra code to write profile information suitable for the
3366 analysis program @command{prof}. You must use this option when compiling
3367 the source files you want data about, and you must also use it when
3370 @cindex @command{gprof}
3373 Generate extra code to write profile information suitable for the
3374 analysis program @command{gprof}. You must use this option when compiling
3375 the source files you want data about, and you must also use it when
3380 Makes the compiler print out each function name as it is compiled, and
3381 print some statistics about each pass when it finishes.
3384 @opindex ftime-report
3385 Makes the compiler print some statistics about the time consumed by each
3386 pass when it finishes.
3389 @opindex fmem-report
3390 Makes the compiler print some statistics about permanent memory
3391 allocation when it finishes.
3393 @item -fprofile-arcs
3394 @opindex fprofile-arcs
3395 Add code so that program flow @dfn{arcs} are instrumented. During
3396 execution the program records how many times each branch and call is
3397 executed and how many times it is taken or returns. When the compiled
3398 program exits it saves this data to a file called
3399 @file{@var{auxname}.gcda} for each source file. The data may be used for
3400 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3401 test coverage analysis (@option{-ftest-coverage}). Each object file's
3402 @var{auxname} is generated from the name of the output file, if
3403 explicitly specified and it is not the final executable, otherwise it is
3404 the basename of the source file. In both cases any suffix is removed
3405 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3406 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3408 @cindex @command{gcov}
3412 This option is used to compile and link code instrumented for coverage
3413 analysis. The option is a synonym for @option{-fprofile-arcs}
3414 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3415 linking). See the documentation for those options for more details.
3420 Compile the source files with @option{-fprofile-arcs} plus optimization
3421 and code generation options. For test coverage analysis, use the
3422 additional @option{-ftest-coverage} option. You do not need to profile
3423 every source file in a program.
3426 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3427 (the latter implies the former).
3430 Run the program on a representative workload to generate the arc profile
3431 information. This may be repeated any number of times. You can run
3432 concurrent instances of your program, and provided that the file system
3433 supports locking, the data files will be correctly updated. Also
3434 @code{fork} calls are detected and correctly handled (double counting
3438 For profile-directed optimizations, compile the source files again with
3439 the same optimization and code generation options plus
3440 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3441 Control Optimization}).
3444 For test coverage analysis, use @command{gcov} to produce human readable
3445 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3446 @command{gcov} documentation for further information.
3450 With @option{-fprofile-arcs}, for each function of your program GCC
3451 creates a program flow graph, then finds a spanning tree for the graph.
3452 Only arcs that are not on the spanning tree have to be instrumented: the
3453 compiler adds code to count the number of times that these arcs are
3454 executed. When an arc is the only exit or only entrance to a block, the
3455 instrumentation code can be added to the block; otherwise, a new basic
3456 block must be created to hold the instrumentation code.
3458 @item -ftree-based-profiling
3459 @opindex ftree-based-profiling
3460 This option is used in addition to @option{-fprofile-arcs} or
3461 @option{-fbranch-probabilities} to control whether those optimizations
3462 are performed on a tree-based or rtl-based internal representation.
3463 If you use this option when compiling with @option{-fprofile-arcs},
3464 you must also use it when compiling later with @option{-fbranch-probabilities}.
3465 Currently the tree-based optimization is in an early stage of
3466 development, and this option is recommended only for those people
3467 working on improving it.
3470 @item -ftest-coverage
3471 @opindex ftest-coverage
3472 Produce a notes file that the @command{gcov} code-coverage utility
3473 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3474 show program coverage. Each source file's note file is called
3475 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3476 above for a description of @var{auxname} and instructions on how to
3477 generate test coverage data. Coverage data will match the source files
3478 more closely, if you do not optimize.
3480 @item -d@var{letters}
3481 @item -fdump-rtl-@var{pass}
3483 Says to make debugging dumps during compilation at times specified by
3484 @var{letters}. This is used for debugging the RTL-based passes of the
3485 compiler. The file names for most of the dumps are made by appending a
3486 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3487 from the name of the output file, if explicitly specified and it is not
3488 an executable, otherwise it is the basename of the source file.
3490 Most debug dumps can be enabled either passing a letter to the @option{-d}
3491 option, or with a long @option{-fdump-rtl} switch; here are the possible
3492 letters for use in @var{letters} and @var{pass}, and their meanings:
3497 Annotate the assembler output with miscellaneous debugging information.
3500 @itemx -fdump-rtl-bp
3502 @opindex fdump-rtl-bp
3503 Dump after computing branch probabilities, to @file{@var{file}.09.bp}.
3506 @itemx -fdump-rtl-bbro
3508 @opindex fdump-rtl-bbro
3509 Dump after block reordering, to @file{@var{file}.30.bbro}.
3512 @itemx -fdump-rtl-combine
3514 @opindex fdump-rtl-combine
3515 Dump after instruction combination, to the file @file{@var{file}.17.combine}.
3518 @itemx -fdump-rtl-ce1
3519 @itemx -fdump-rtl-ce2
3521 @opindex fdump-rtl-ce1
3522 @opindex fdump-rtl-ce2
3523 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3524 first if conversion, to the file @file{@var{file}.11.ce1}. @option{-dC}
3525 and @option{-fdump-rtl-ce2} enable dumping after the second if
3526 conversion, to the file @file{@var{file}.18.ce2}.
3529 @itemx -fdump-rtl-btl
3530 @itemx -fdump-rtl-dbr
3532 @opindex fdump-rtl-btl
3533 @opindex fdump-rtl-dbr
3534 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3535 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3536 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3537 scheduling, to @file{@var{file}.36.dbr}.
3541 Dump all macro definitions, at the end of preprocessing, in addition to
3545 @itemx -fdump-rtl-ce3
3547 @opindex fdump-rtl-ce3
3548 Dump after the third if conversion, to @file{@var{file}.28.ce3}.
3551 @itemx -fdump-rtl-cfg
3552 @itemx -fdump-rtl-life
3554 @opindex fdump-rtl-cfg
3555 @opindex fdump-rtl-life
3556 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3557 and data flow analysis, to @file{@var{file}.08.cfg}. @option{-df}
3558 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3559 to @file{@var{file}.16.life}.
3562 @itemx -fdump-rtl-greg
3564 @opindex fdump-rtl-greg
3565 Dump after global register allocation, to @file{@var{file}.23.greg}.
3568 @itemx -fdump-rtl-gcse
3569 @itemx -fdump-rtl-bypass
3571 @opindex fdump-rtl-gcse
3572 @opindex fdump-rtl-bypass
3573 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3574 @file{@var{file}.05.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3575 enable dumping after jump bypassing and control flow optimizations, to
3576 @file{@var{file}.07.bypass}.
3579 @itemx -fdump-rtl-eh
3581 @opindex fdump-rtl-eh
3582 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3585 @itemx -fdump-rtl-sibling
3587 @opindex fdump-rtl-sibling
3588 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
3591 @itemx -fdump-rtl-jump
3593 @opindex fdump-rtl-jump
3594 Dump after the first jump optimization, to @file{@var{file}.03.jump}.
3597 @itemx -fdump-rtl-stack
3599 @opindex fdump-rtl-stack
3600 Dump after conversion from registers to stack, to @file{@var{file}.33.stack}.
3603 @itemx -fdump-rtl-lreg
3605 @opindex fdump-rtl-lreg
3606 Dump after local register allocation, to @file{@var{file}.22.lreg}.
3609 @itemx -fdump-rtl-loop
3610 @itemx -fdump-rtl-loop2
3612 @opindex fdump-rtl-loop
3613 @opindex fdump-rtl-loop2
3614 @option{-dL} and @option{-fdump-rtl-loop} enable dumping after the first
3615 loop optimization pass, to @file{@var{file}.06.loop}. @option{-dL} and
3616 @option{-fdump-rtl-loop2} enable dumping after the second pass, to
3617 @file{@var{file}.13.loop2}.
3620 @itemx -fdump-rtl-sms
3622 @opindex fdump-rtl-sms
3623 Dump after modulo scheduling, to @file{@var{file}.20.sms}.
3626 @itemx -fdump-rtl-mach
3628 @opindex fdump-rtl-mach
3629 Dump after performing the machine dependent reorganization pass, to
3630 @file{@var{file}.35.mach}.
3633 @itemx -fdump-rtl-rnreg
3635 @opindex fdump-rtl-rnreg
3636 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3639 @itemx -fdump-rtl-regmove
3641 @opindex fdump-rtl-regmove
3642 Dump after the register move pass, to @file{@var{file}.19.regmove}.
3645 @itemx -fdump-rtl-postreload
3647 @opindex fdump-rtl-postreload
3648 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3651 @itemx -fdump-rtl-expand
3653 @opindex fdump-rtl-expand
3654 Dump after RTL generation, to @file{@var{file}.00.expand}.
3657 @itemx -fdump-rtl-sched2
3659 @opindex fdump-rtl-sched2
3660 Dump after the second scheduling pass, to @file{@var{file}.32.sched2}.
3663 @itemx -fdump-rtl-cse
3665 @opindex fdump-rtl-cse
3666 Dump after CSE (including the jump optimization that sometimes follows
3667 CSE), to @file{@var{file}.04.cse}.
3670 @itemx -fdump-rtl-sched
3672 @opindex fdump-rtl-sched
3673 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3676 @itemx -fdump-rtl-cse2
3678 @opindex fdump-rtl-cse2
3679 Dump after the second CSE pass (including the jump optimization that
3680 sometimes follows CSE), to @file{@var{file}.15.cse2}.
3683 @itemx -fdump-rtl-tracer
3685 @opindex fdump-rtl-tracer
3686 Dump after running tracer, to @file{@var{file}.12.tracer}.
3689 @itemx -fdump-rtl-vpt
3690 @itemx -fdump-rtl-vartrack
3692 @opindex fdump-rtl-vpt
3693 @opindex fdump-rtl-vartrack
3694 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3695 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3696 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3697 to @file{@var{file}.34.vartrack}.
3700 @itemx -fdump-rtl-flow2
3702 @opindex fdump-rtl-flow2
3703 Dump after the second flow pass, to @file{@var{file}.26.flow2}.
3706 @itemx -fdump-rtl-peephole2
3708 @opindex fdump-rtl-peephole2
3709 Dump after the peephole pass, to @file{@var{file}.27.peephole2}.
3712 @itemx -fdump-rtl-web
3714 @opindex fdump-rtl-web
3715 Dump after live range splitting, to @file{@var{file}.14.web}.
3718 @itemx -fdump-rtl-all
3720 @opindex fdump-rtl-all
3721 Produce all the dumps listed above.
3725 Produce a core dump whenever an error occurs.
3729 Print statistics on memory usage, at the end of the run, to
3734 Annotate the assembler output with a comment indicating which
3735 pattern and alternative was used. The length of each instruction is
3740 Dump the RTL in the assembler output as a comment before each instruction.
3741 Also turns on @option{-dp} annotation.
3745 For each of the other indicated dump files (either with @option{-d} or
3746 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3747 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3751 Just generate RTL for a function instead of compiling it. Usually used
3752 with @samp{r} (@option{-fdump-rtl-expand}).
3756 Dump debugging information during parsing, to standard error.
3759 @item -fdump-unnumbered
3760 @opindex fdump-unnumbered
3761 When doing debugging dumps (see @option{-d} option above), suppress instruction
3762 numbers and line number note output. This makes it more feasible to
3763 use diff on debugging dumps for compiler invocations with different
3764 options, in particular with and without @option{-g}.
3766 @item -fdump-translation-unit @r{(C and C++ only)}
3767 @itemx -fdump-translation-unit-@var{options} @r{(C and C++ only)}
3768 @opindex fdump-translation-unit
3769 Dump a representation of the tree structure for the entire translation
3770 unit to a file. The file name is made by appending @file{.tu} to the
3771 source file name. If the @samp{-@var{options}} form is used, @var{options}
3772 controls the details of the dump as described for the
3773 @option{-fdump-tree} options.
3775 @item -fdump-class-hierarchy @r{(C++ only)}
3776 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3777 @opindex fdump-class-hierarchy
3778 Dump a representation of each class's hierarchy and virtual function
3779 table layout to a file. The file name is made by appending @file{.class}
3780 to the source file name. If the @samp{-@var{options}} form is used,
3781 @var{options} controls the details of the dump as described for the
3782 @option{-fdump-tree} options.
3784 @item -fdump-ipa-@var{switch}
3786 Control the dumping at various stages of inter-procedural analysis
3787 language tree to a file. The file name is generated by appending a switch
3788 specific suffix to the source file name. The following dumps are possible:
3792 Enables all inter-procedural analysis dumps; currently the only produced
3793 dump is the @samp{cgraph} dump.
3796 Dumps information about call-graph optimization, unused function removal,
3797 and inlining decisions.
3800 @item -fdump-tree-@var{switch} @r{(C and C++ only)}
3801 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C and C++ only)}
3803 Control the dumping at various stages of processing the intermediate
3804 language tree to a file. The file name is generated by appending a switch
3805 specific suffix to the source file name. If the @samp{-@var{options}}
3806 form is used, @var{options} is a list of @samp{-} separated options that
3807 control the details of the dump. Not all options are applicable to all
3808 dumps, those which are not meaningful will be ignored. The following
3809 options are available
3813 Print the address of each node. Usually this is not meaningful as it
3814 changes according to the environment and source file. Its primary use
3815 is for tying up a dump file with a debug environment.
3817 Inhibit dumping of members of a scope or body of a function merely
3818 because that scope has been reached. Only dump such items when they
3819 are directly reachable by some other path. When dumping pretty-printed
3820 trees, this option inhibits dumping the bodies of control structures.
3822 Print a raw representation of the tree. By default, trees are
3823 pretty-printed into a C-like representation.
3825 Enable more detailed dumps (not honored by every dump option).
3827 Enable dumping various statistics about the pass (not honored by every dump
3830 Enable showing basic block boundaries (disabled in raw dumps).
3832 Enable showing virtual operands for every statement.
3834 Enable showing line numbers for statements.
3836 Enable showing the unique ID (@code{DECL_UID}) for each variable.
3838 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
3841 The following tree dumps are possible:
3845 Dump before any tree based optimization, to @file{@var{file}.original}.
3848 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3851 Dump after function inlining, to @file{@var{file}.inlined}.
3854 @opindex fdump-tree-gimple
3855 Dump each function before and after the gimplification pass to a file. The
3856 file name is made by appending @file{.gimple} to the source file name.
3859 @opindex fdump-tree-cfg
3860 Dump the control flow graph of each function to a file. The file name is
3861 made by appending @file{.cfg} to the source file name.
3864 @opindex fdump-tree-vcg
3865 Dump the control flow graph of each function to a file in VCG format. The
3866 file name is made by appending @file{.vcg} to the source file name. Note
3867 that if the file contains more than one function, the generated file cannot
3868 be used directly by VCG@. You will need to cut and paste each function's
3869 graph into its own separate file first.
3872 @opindex fdump-tree-ch
3873 Dump each function after copying loop headers. The file name is made by
3874 appending @file{.ch} to the source file name.
3877 @opindex fdump-tree-ssa
3878 Dump SSA related information to a file. The file name is made by appending
3879 @file{.ssa} to the source file name.
3882 @opindex fdump-tree-salias
3883 Dump structure aliasing variable information to a file. This file name
3884 is made by appending @file{.salias} to the source file name.
3887 @opindex fdump-tree-alias
3888 Dump aliasing information for each function. The file name is made by
3889 appending @file{.alias} to the source file name.
3892 @opindex fdump-tree-ccp
3893 Dump each function after CCP@. The file name is made by appending
3894 @file{.ccp} to the source file name.
3897 @opindex fdump-tree-storeccp
3898 Dump each function after STORE-CCP. The file name is made by appending
3899 @file{.storeccp} to the source file name.
3902 @opindex fdump-tree-pre
3903 Dump trees after partial redundancy elimination. The file name is made
3904 by appending @file{.pre} to the source file name.
3907 @opindex fdump-tree-fre
3908 Dump trees after full redundancy elimination. The file name is made
3909 by appending @file{.fre} to the source file name.
3912 @opindex fdump-tree-copyprop
3913 Dump trees after copy propagation. The file name is made
3914 by appending @file{.copyprop} to the source file name.
3916 @item store_copyprop
3917 @opindex fdump-tree-store_copyprop
3918 Dump trees after store copy-propagation. The file name is made
3919 by appending @file{.store_copyprop} to the source file name.
3922 @opindex fdump-tree-dce
3923 Dump each function after dead code elimination. The file name is made by
3924 appending @file{.dce} to the source file name.
3927 @opindex fdump-tree-mudflap
3928 Dump each function after adding mudflap instrumentation. The file name is
3929 made by appending @file{.mudflap} to the source file name.
3932 @opindex fdump-tree-sra
3933 Dump each function after performing scalar replacement of aggregates. The
3934 file name is made by appending @file{.sra} to the source file name.
3937 @opindex fdump-tree-sink
3938 Dump each function after performing code sinking. The file name is made
3939 by appending @file{.sink} to the source file name.
3942 @opindex fdump-tree-dom
3943 Dump each function after applying dominator tree optimizations. The file
3944 name is made by appending @file{.dom} to the source file name.
3947 @opindex fdump-tree-dse
3948 Dump each function after applying dead store elimination. The file
3949 name is made by appending @file{.dse} to the source file name.
3952 @opindex fdump-tree-phiopt
3953 Dump each function after optimizing PHI nodes into straightline code. The file
3954 name is made by appending @file{.phiopt} to the source file name.
3957 @opindex fdump-tree-forwprop
3958 Dump each function after forward propagating single use variables. The file
3959 name is made by appending @file{.forwprop} to the source file name.
3962 @opindex fdump-tree-copyrename
3963 Dump each function after applying the copy rename optimization. The file
3964 name is made by appending @file{.copyrename} to the source file name.
3967 @opindex fdump-tree-nrv
3968 Dump each function after applying the named return value optimization on
3969 generic trees. The file name is made by appending @file{.nrv} to the source
3973 @opindex fdump-tree-vect
3974 Dump each function after applying vectorization of loops. The file name is
3975 made by appending @file{.vect} to the source file name.
3978 @opindex fdump-tree-all
3979 Enable all the available tree dumps with the flags provided in this option.
3982 @item -ftree-vectorizer-verbose=@var{n}
3983 @opindex ftree-vectorizer-verbose
3984 This option controls the amount of debugging output the vectorizer prints.
3985 This information is written to standard error, unless @option{-fdump-tree-all}
3986 or @option{-fdump-tree-vect} is specified, in which case it is output to the
3987 usual dump listing file, @file{.vect}.
3989 @item -frandom-seed=@var{string}
3990 @opindex frandom-string
3991 This option provides a seed that GCC uses when it would otherwise use
3992 random numbers. It is used to generate certain symbol names
3993 that have to be different in every compiled file. It is also used to
3994 place unique stamps in coverage data files and the object files that
3995 produce them. You can use the @option{-frandom-seed} option to produce
3996 reproducibly identical object files.
3998 The @var{string} should be different for every file you compile.
4000 @item -fsched-verbose=@var{n}
4001 @opindex fsched-verbose
4002 On targets that use instruction scheduling, this option controls the
4003 amount of debugging output the scheduler prints. This information is
4004 written to standard error, unless @option{-dS} or @option{-dR} is
4005 specified, in which case it is output to the usual dump
4006 listing file, @file{.sched} or @file{.sched2} respectively. However
4007 for @var{n} greater than nine, the output is always printed to standard
4010 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4011 same information as @option{-dRS}. For @var{n} greater than one, it
4012 also output basic block probabilities, detailed ready list information
4013 and unit/insn info. For @var{n} greater than two, it includes RTL
4014 at abort point, control-flow and regions info. And for @var{n} over
4015 four, @option{-fsched-verbose} also includes dependence info.
4019 Store the usual ``temporary'' intermediate files permanently; place them
4020 in the current directory and name them based on the source file. Thus,
4021 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4022 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4023 preprocessed @file{foo.i} output file even though the compiler now
4024 normally uses an integrated preprocessor.
4026 When used in combination with the @option{-x} command line option,
4027 @option{-save-temps} is sensible enough to avoid over writing an
4028 input source file with the same extension as an intermediate file.
4029 The corresponding intermediate file may be obtained by renaming the
4030 source file before using @option{-save-temps}.
4034 Report the CPU time taken by each subprocess in the compilation
4035 sequence. For C source files, this is the compiler proper and assembler
4036 (plus the linker if linking is done). The output looks like this:
4043 The first number on each line is the ``user time'', that is time spent
4044 executing the program itself. The second number is ``system time'',
4045 time spent executing operating system routines on behalf of the program.
4046 Both numbers are in seconds.
4048 @item -fvar-tracking
4049 @opindex fvar-tracking
4050 Run variable tracking pass. It computes where variables are stored at each
4051 position in code. Better debugging information is then generated
4052 (if the debugging information format supports this information).
4054 It is enabled by default when compiling with optimization (@option{-Os},
4055 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4056 the debug info format supports it.
4058 @item -print-file-name=@var{library}
4059 @opindex print-file-name
4060 Print the full absolute name of the library file @var{library} that
4061 would be used when linking---and don't do anything else. With this
4062 option, GCC does not compile or link anything; it just prints the
4065 @item -print-multi-directory
4066 @opindex print-multi-directory
4067 Print the directory name corresponding to the multilib selected by any
4068 other switches present in the command line. This directory is supposed
4069 to exist in @env{GCC_EXEC_PREFIX}.
4071 @item -print-multi-lib
4072 @opindex print-multi-lib
4073 Print the mapping from multilib directory names to compiler switches
4074 that enable them. The directory name is separated from the switches by
4075 @samp{;}, and each switch starts with an @samp{@@} instead of the
4076 @samp{-}, without spaces between multiple switches. This is supposed to
4077 ease shell-processing.
4079 @item -print-prog-name=@var{program}
4080 @opindex print-prog-name
4081 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4083 @item -print-libgcc-file-name
4084 @opindex print-libgcc-file-name
4085 Same as @option{-print-file-name=libgcc.a}.
4087 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4088 but you do want to link with @file{libgcc.a}. You can do
4091 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4094 @item -print-search-dirs
4095 @opindex print-search-dirs
4096 Print the name of the configured installation directory and a list of
4097 program and library directories @command{gcc} will search---and don't do anything else.
4099 This is useful when @command{gcc} prints the error message
4100 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4101 To resolve this you either need to put @file{cpp0} and the other compiler
4102 components where @command{gcc} expects to find them, or you can set the environment
4103 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4104 Don't forget the trailing @samp{/}.
4105 @xref{Environment Variables}.
4108 @opindex dumpmachine
4109 Print the compiler's target machine (for example,
4110 @samp{i686-pc-linux-gnu})---and don't do anything else.
4113 @opindex dumpversion
4114 Print the compiler version (for example, @samp{3.0})---and don't do
4119 Print the compiler's built-in specs---and don't do anything else. (This
4120 is used when GCC itself is being built.) @xref{Spec Files}.
4122 @item -feliminate-unused-debug-types
4123 @opindex feliminate-unused-debug-types
4124 Normally, when producing DWARF2 output, GCC will emit debugging
4125 information for all types declared in a compilation
4126 unit, regardless of whether or not they are actually used
4127 in that compilation unit. Sometimes this is useful, such as
4128 if, in the debugger, you want to cast a value to a type that is
4129 not actually used in your program (but is declared). More often,
4130 however, this results in a significant amount of wasted space.
4131 With this option, GCC will avoid producing debug symbol output
4132 for types that are nowhere used in the source file being compiled.
4135 @node Optimize Options
4136 @section Options That Control Optimization
4137 @cindex optimize options
4138 @cindex options, optimization
4140 These options control various sorts of optimizations.
4142 Without any optimization option, the compiler's goal is to reduce the
4143 cost of compilation and to make debugging produce the expected
4144 results. Statements are independent: if you stop the program with a
4145 breakpoint between statements, you can then assign a new value to any
4146 variable or change the program counter to any other statement in the
4147 function and get exactly the results you would expect from the source
4150 Turning on optimization flags makes the compiler attempt to improve
4151 the performance and/or code size at the expense of compilation time
4152 and possibly the ability to debug the program.
4154 The compiler performs optimization based on the knowledge it has of
4155 the program. Optimization levels @option{-O2} and above, in
4156 particular, enable @emph{unit-at-a-time} mode, which allows the
4157 compiler to consider information gained from later functions in
4158 the file when compiling a function. Compiling multiple files at
4159 once to a single output file in @emph{unit-at-a-time} mode allows
4160 the compiler to use information gained from all of the files when
4161 compiling each of them.
4163 Not all optimizations are controlled directly by a flag. Only
4164 optimizations that have a flag are listed.
4171 Optimize. Optimizing compilation takes somewhat more time, and a lot
4172 more memory for a large function.
4174 With @option{-O}, the compiler tries to reduce code size and execution
4175 time, without performing any optimizations that take a great deal of
4178 @option{-O} turns on the following optimization flags:
4179 @gccoptlist{-fdefer-pop @gol
4180 -fdelayed-branch @gol
4181 -fguess-branch-probability @gol
4182 -fcprop-registers @gol
4183 -floop-optimize @gol
4184 -fif-conversion @gol
4185 -fif-conversion2 @gol
4188 -ftree-dominator-opts @gol
4193 -ftree-copyrename @gol
4198 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4199 where doing so does not interfere with debugging.
4203 Optimize even more. GCC performs nearly all supported optimizations
4204 that do not involve a space-speed tradeoff. The compiler does not
4205 perform loop unrolling or function inlining when you specify @option{-O2}.
4206 As compared to @option{-O}, this option increases both compilation time
4207 and the performance of the generated code.
4209 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4210 also turns on the following optimization flags:
4211 @gccoptlist{-fthread-jumps @gol
4213 -foptimize-sibling-calls @gol
4214 -fcse-follow-jumps -fcse-skip-blocks @gol
4215 -fgcse -fgcse-lm @gol
4216 -fexpensive-optimizations @gol
4217 -fstrength-reduce @gol
4218 -frerun-cse-after-loop -frerun-loop-opt @gol
4222 -fschedule-insns -fschedule-insns2 @gol
4223 -fsched-interblock -fsched-spec @gol
4225 -fstrict-aliasing @gol
4226 -fdelete-null-pointer-checks @gol
4227 -freorder-blocks -freorder-functions @gol
4228 -funit-at-a-time @gol
4229 -falign-functions -falign-jumps @gol
4230 -falign-loops -falign-labels @gol
4233 Please note the warning under @option{-fgcse} about
4234 invoking @option{-O2} on programs that use computed gotos.
4238 Optimize yet more. @option{-O3} turns on all optimizations specified by
4239 @option{-O2} and also turns on the @option{-finline-functions},
4240 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4244 Do not optimize. This is the default.
4248 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4249 do not typically increase code size. It also performs further
4250 optimizations designed to reduce code size.
4252 @option{-Os} disables the following optimization flags:
4253 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4254 -falign-labels -freorder-blocks -freorder-blocks-and-partition -fprefetch-loop-arrays}
4256 If you use multiple @option{-O} options, with or without level numbers,
4257 the last such option is the one that is effective.
4260 Options of the form @option{-f@var{flag}} specify machine-independent
4261 flags. Most flags have both positive and negative forms; the negative
4262 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4263 below, only one of the forms is listed---the one you typically will
4264 use. You can figure out the other form by either removing @samp{no-}
4267 The following options control specific optimizations. They are either
4268 activated by @option{-O} options or are related to ones that are. You
4269 can use the following flags in the rare cases when ``fine-tuning'' of
4270 optimizations to be performed is desired.
4273 @item -fno-default-inline
4274 @opindex fno-default-inline
4275 Do not make member functions inline by default merely because they are
4276 defined inside the class scope (C++ only). Otherwise, when you specify
4277 @w{@option{-O}}, member functions defined inside class scope are compiled
4278 inline by default; i.e., you don't need to add @samp{inline} in front of
4279 the member function name.
4281 @item -fno-defer-pop
4282 @opindex fno-defer-pop
4283 Always pop the arguments to each function call as soon as that function
4284 returns. For machines which must pop arguments after a function call,
4285 the compiler normally lets arguments accumulate on the stack for several
4286 function calls and pops them all at once.
4288 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4292 Force memory operands to be copied into registers before doing
4293 arithmetic on them. This produces better code by making all memory
4294 references potential common subexpressions. When they are not common
4295 subexpressions, instruction combination should eliminate the separate
4298 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4301 @opindex fforce-addr
4302 Force memory address constants to be copied into registers before
4303 doing arithmetic on them. This may produce better code just as
4304 @option{-fforce-mem} may.
4306 @item -fomit-frame-pointer
4307 @opindex fomit-frame-pointer
4308 Don't keep the frame pointer in a register for functions that
4309 don't need one. This avoids the instructions to save, set up and
4310 restore frame pointers; it also makes an extra register available
4311 in many functions. @strong{It also makes debugging impossible on
4314 On some machines, such as the VAX, this flag has no effect, because
4315 the standard calling sequence automatically handles the frame pointer
4316 and nothing is saved by pretending it doesn't exist. The
4317 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4318 whether a target machine supports this flag. @xref{Registers,,Register
4319 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4321 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4323 @item -foptimize-sibling-calls
4324 @opindex foptimize-sibling-calls
4325 Optimize sibling and tail recursive calls.
4327 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4331 Don't pay attention to the @code{inline} keyword. Normally this option
4332 is used to keep the compiler from expanding any functions inline.
4333 Note that if you are not optimizing, no functions can be expanded inline.
4335 @item -finline-functions
4336 @opindex finline-functions
4337 Integrate all simple functions into their callers. The compiler
4338 heuristically decides which functions are simple enough to be worth
4339 integrating in this way.
4341 If all calls to a given function are integrated, and the function is
4342 declared @code{static}, then the function is normally not output as
4343 assembler code in its own right.
4345 Enabled at level @option{-O3}.
4347 @item -finline-limit=@var{n}
4348 @opindex finline-limit
4349 By default, GCC limits the size of functions that can be inlined. This flag
4350 allows the control of this limit for functions that are explicitly marked as
4351 inline (i.e., marked with the inline keyword or defined within the class
4352 definition in c++). @var{n} is the size of functions that can be inlined in
4353 number of pseudo instructions (not counting parameter handling). The default
4354 value of @var{n} is 600.
4355 Increasing this value can result in more inlined code at
4356 the cost of compilation time and memory consumption. Decreasing usually makes
4357 the compilation faster and less code will be inlined (which presumably
4358 means slower programs). This option is particularly useful for programs that
4359 use inlining heavily such as those based on recursive templates with C++.
4361 Inlining is actually controlled by a number of parameters, which may be
4362 specified individually by using @option{--param @var{name}=@var{value}}.
4363 The @option{-finline-limit=@var{n}} option sets some of these parameters
4367 @item max-inline-insns-single
4368 is set to @var{n}/2.
4369 @item max-inline-insns-auto
4370 is set to @var{n}/2.
4371 @item min-inline-insns
4372 is set to 130 or @var{n}/4, whichever is smaller.
4373 @item max-inline-insns-rtl
4377 See below for a documentation of the individual
4378 parameters controlling inlining.
4380 @emph{Note:} pseudo instruction represents, in this particular context, an
4381 abstract measurement of function's size. In no way, it represents a count
4382 of assembly instructions and as such its exact meaning might change from one
4383 release to an another.
4385 @item -fkeep-inline-functions
4386 @opindex fkeep-inline-functions
4387 In C, emit @code{static} functions that are declared @code{inline}
4388 into the object file, even if the function has been inlined into all
4389 of its callers. This switch does not affect functions using the
4390 @code{extern inline} extension in GNU C@. In C++, emit any and all
4391 inline functions into the object file.
4393 @item -fkeep-static-consts
4394 @opindex fkeep-static-consts
4395 Emit variables declared @code{static const} when optimization isn't turned
4396 on, even if the variables aren't referenced.
4398 GCC enables this option by default. If you want to force the compiler to
4399 check if the variable was referenced, regardless of whether or not
4400 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4402 @item -fmerge-constants
4403 Attempt to merge identical constants (string constants and floating point
4404 constants) across compilation units.
4406 This option is the default for optimized compilation if the assembler and
4407 linker support it. Use @option{-fno-merge-constants} to inhibit this
4410 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4412 @item -fmerge-all-constants
4413 Attempt to merge identical constants and identical variables.
4415 This option implies @option{-fmerge-constants}. In addition to
4416 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4417 arrays or initialized constant variables with integral or floating point
4418 types. Languages like C or C++ require each non-automatic variable to
4419 have distinct location, so using this option will result in non-conforming
4422 @item -fmodulo-sched
4423 @opindex fmodulo-sched
4424 Perform swing modulo scheduling immediately before the first scheduling
4425 pass. This pass looks at innermost loops and reorders their
4426 instructions by overlapping different iterations.
4428 @item -fno-branch-count-reg
4429 @opindex fno-branch-count-reg
4430 Do not use ``decrement and branch'' instructions on a count register,
4431 but instead generate a sequence of instructions that decrement a
4432 register, compare it against zero, then branch based upon the result.
4433 This option is only meaningful on architectures that support such
4434 instructions, which include x86, PowerPC, IA-64 and S/390.
4436 The default is @option{-fbranch-count-reg}, enabled when
4437 @option{-fstrength-reduce} is enabled.
4439 @item -fno-function-cse
4440 @opindex fno-function-cse
4441 Do not put function addresses in registers; make each instruction that
4442 calls a constant function contain the function's address explicitly.
4444 This option results in less efficient code, but some strange hacks
4445 that alter the assembler output may be confused by the optimizations
4446 performed when this option is not used.
4448 The default is @option{-ffunction-cse}
4450 @item -fno-zero-initialized-in-bss
4451 @opindex fno-zero-initialized-in-bss
4452 If the target supports a BSS section, GCC by default puts variables that
4453 are initialized to zero into BSS@. This can save space in the resulting
4456 This option turns off this behavior because some programs explicitly
4457 rely on variables going to the data section. E.g., so that the
4458 resulting executable can find the beginning of that section and/or make
4459 assumptions based on that.
4461 The default is @option{-fzero-initialized-in-bss}.
4463 @item -fbounds-check
4464 @opindex fbounds-check
4465 For front-ends that support it, generate additional code to check that
4466 indices used to access arrays are within the declared range. This is
4467 currently only supported by the Java and Fortran front-ends, where
4468 this option defaults to true and false respectively.
4470 @item -fmudflap -fmudflapth -fmudflapir
4474 @cindex bounds checking
4476 For front-ends that support it (C and C++), instrument all risky
4477 pointer/array dereferencing operations, some standard library
4478 string/heap functions, and some other associated constructs with
4479 range/validity tests. Modules so instrumented should be immune to
4480 buffer overflows, invalid heap use, and some other classes of C/C++
4481 programming errors. The instrumentation relies on a separate runtime
4482 library (@file{libmudflap}), which will be linked into a program if
4483 @option{-fmudflap} is given at link time. Run-time behavior of the
4484 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4485 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4488 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4489 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4490 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4491 instrumentation should ignore pointer reads. This produces less
4492 instrumentation (and therefore faster execution) and still provides
4493 some protection against outright memory corrupting writes, but allows
4494 erroneously read data to propagate within a program.
4496 @item -fstrength-reduce
4497 @opindex fstrength-reduce
4498 Perform the optimizations of loop strength reduction and
4499 elimination of iteration variables.
4501 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4503 @item -fthread-jumps
4504 @opindex fthread-jumps
4505 Perform optimizations where we check to see if a jump branches to a
4506 location where another comparison subsumed by the first is found. If
4507 so, the first branch is redirected to either the destination of the
4508 second branch or a point immediately following it, depending on whether
4509 the condition is known to be true or false.
4511 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4513 @item -fcse-follow-jumps
4514 @opindex fcse-follow-jumps
4515 In common subexpression elimination, scan through jump instructions
4516 when the target of the jump is not reached by any other path. For
4517 example, when CSE encounters an @code{if} statement with an
4518 @code{else} clause, CSE will follow the jump when the condition
4521 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4523 @item -fcse-skip-blocks
4524 @opindex fcse-skip-blocks
4525 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4526 follow jumps which conditionally skip over blocks. When CSE
4527 encounters a simple @code{if} statement with no else clause,
4528 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4529 body of the @code{if}.
4531 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4533 @item -frerun-cse-after-loop
4534 @opindex frerun-cse-after-loop
4535 Re-run common subexpression elimination after loop optimizations has been
4538 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4540 @item -frerun-loop-opt
4541 @opindex frerun-loop-opt
4542 Run the loop optimizer twice.
4544 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4548 Perform a global common subexpression elimination pass.
4549 This pass also performs global constant and copy propagation.
4551 @emph{Note:} When compiling a program using computed gotos, a GCC
4552 extension, you may get better runtime performance if you disable
4553 the global common subexpression elimination pass by adding
4554 @option{-fno-gcse} to the command line.
4556 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4560 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4561 attempt to move loads which are only killed by stores into themselves. This
4562 allows a loop containing a load/store sequence to be changed to a load outside
4563 the loop, and a copy/store within the loop.
4565 Enabled by default when gcse is enabled.
4569 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4570 global common subexpression elimination. This pass will attempt to move
4571 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4572 loops containing a load/store sequence can be changed to a load before
4573 the loop and a store after the loop.
4575 Not enabled at any optimization level.
4579 When @option{-fgcse-las} is enabled, the global common subexpression
4580 elimination pass eliminates redundant loads that come after stores to the
4581 same memory location (both partial and full redundancies).
4583 Not enabled at any optimization level.
4585 @item -fgcse-after-reload
4586 @opindex fgcse-after-reload
4587 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4588 pass is performed after reload. The purpose of this pass is to cleanup
4591 @item -floop-optimize
4592 @opindex floop-optimize
4593 Perform loop optimizations: move constant expressions out of loops, simplify
4594 exit test conditions and optionally do strength-reduction as well.
4596 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4598 @item -floop-optimize2
4599 @opindex floop-optimize2
4600 Perform loop optimizations using the new loop optimizer. The optimizations
4601 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4604 @item -fcrossjumping
4605 @opindex crossjumping
4606 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4607 resulting code may or may not perform better than without cross-jumping.
4609 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4611 @item -fif-conversion
4612 @opindex if-conversion
4613 Attempt to transform conditional jumps into branch-less equivalents. This
4614 include use of conditional moves, min, max, set flags and abs instructions, and
4615 some tricks doable by standard arithmetics. The use of conditional execution
4616 on chips where it is available is controlled by @code{if-conversion2}.
4618 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4620 @item -fif-conversion2
4621 @opindex if-conversion2
4622 Use conditional execution (where available) to transform conditional jumps into
4623 branch-less equivalents.
4625 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4627 @item -fdelete-null-pointer-checks
4628 @opindex fdelete-null-pointer-checks
4629 Use global dataflow analysis to identify and eliminate useless checks
4630 for null pointers. The compiler assumes that dereferencing a null
4631 pointer would have halted the program. If a pointer is checked after
4632 it has already been dereferenced, it cannot be null.
4634 In some environments, this assumption is not true, and programs can
4635 safely dereference null pointers. Use
4636 @option{-fno-delete-null-pointer-checks} to disable this optimization
4637 for programs which depend on that behavior.
4639 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4641 @item -fexpensive-optimizations
4642 @opindex fexpensive-optimizations
4643 Perform a number of minor optimizations that are relatively expensive.
4645 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4647 @item -foptimize-register-move
4649 @opindex foptimize-register-move
4651 Attempt to reassign register numbers in move instructions and as
4652 operands of other simple instructions in order to maximize the amount of
4653 register tying. This is especially helpful on machines with two-operand
4656 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4659 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4661 @item -fdelayed-branch
4662 @opindex fdelayed-branch
4663 If supported for the target machine, attempt to reorder instructions
4664 to exploit instruction slots available after delayed branch
4667 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4669 @item -fschedule-insns
4670 @opindex fschedule-insns
4671 If supported for the target machine, attempt to reorder instructions to
4672 eliminate execution stalls due to required data being unavailable. This
4673 helps machines that have slow floating point or memory load instructions
4674 by allowing other instructions to be issued until the result of the load
4675 or floating point instruction is required.
4677 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4679 @item -fschedule-insns2
4680 @opindex fschedule-insns2
4681 Similar to @option{-fschedule-insns}, but requests an additional pass of
4682 instruction scheduling after register allocation has been done. This is
4683 especially useful on machines with a relatively small number of
4684 registers and where memory load instructions take more than one cycle.
4686 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4688 @item -fno-sched-interblock
4689 @opindex fno-sched-interblock
4690 Don't schedule instructions across basic blocks. This is normally
4691 enabled by default when scheduling before register allocation, i.e.@:
4692 with @option{-fschedule-insns} or at @option{-O2} or higher.
4694 @item -fno-sched-spec
4695 @opindex fno-sched-spec
4696 Don't allow speculative motion of non-load instructions. This is normally
4697 enabled by default when scheduling before register allocation, i.e.@:
4698 with @option{-fschedule-insns} or at @option{-O2} or higher.
4700 @item -fsched-spec-load
4701 @opindex fsched-spec-load
4702 Allow speculative motion of some load instructions. This only makes
4703 sense when scheduling before register allocation, i.e.@: with
4704 @option{-fschedule-insns} or at @option{-O2} or higher.
4706 @item -fsched-spec-load-dangerous
4707 @opindex fsched-spec-load-dangerous
4708 Allow speculative motion of more load instructions. This only makes
4709 sense when scheduling before register allocation, i.e.@: with
4710 @option{-fschedule-insns} or at @option{-O2} or higher.
4712 @item -fsched-stalled-insns=@var{n}
4713 @opindex fsched-stalled-insns
4714 Define how many insns (if any) can be moved prematurely from the queue
4715 of stalled insns into the ready list, during the second scheduling pass.
4717 @item -fsched-stalled-insns-dep=@var{n}
4718 @opindex fsched-stalled-insns-dep
4719 Define how many insn groups (cycles) will be examined for a dependency
4720 on a stalled insn that is candidate for premature removal from the queue
4721 of stalled insns. Has an effect only during the second scheduling pass,
4722 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4724 @item -fsched2-use-superblocks
4725 @opindex fsched2-use-superblocks
4726 When scheduling after register allocation, do use superblock scheduling
4727 algorithm. Superblock scheduling allows motion across basic block boundaries
4728 resulting on faster schedules. This option is experimental, as not all machine
4729 descriptions used by GCC model the CPU closely enough to avoid unreliable
4730 results from the algorithm.
4732 This only makes sense when scheduling after register allocation, i.e.@: with
4733 @option{-fschedule-insns2} or at @option{-O2} or higher.
4735 @item -fsched2-use-traces
4736 @opindex fsched2-use-traces
4737 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4738 allocation and additionally perform code duplication in order to increase the
4739 size of superblocks using tracer pass. See @option{-ftracer} for details on
4742 This mode should produce faster but significantly longer programs. Also
4743 without @option{-fbranch-probabilities} the traces constructed may not
4744 match the reality and hurt the performance. This only makes
4745 sense when scheduling after register allocation, i.e.@: with
4746 @option{-fschedule-insns2} or at @option{-O2} or higher.
4748 @item -freschedule-modulo-scheduled-loops
4749 @opindex fscheduling-in-modulo-scheduled-loops
4750 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4751 we may want to prevent the later scheduling passes from changing its schedule, we use this
4752 option to control that.
4754 @item -fcaller-saves
4755 @opindex fcaller-saves
4756 Enable values to be allocated in registers that will be clobbered by
4757 function calls, by emitting extra instructions to save and restore the
4758 registers around such calls. Such allocation is done only when it
4759 seems to result in better code than would otherwise be produced.
4761 This option is always enabled by default on certain machines, usually
4762 those which have no call-preserved registers to use instead.
4764 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4767 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4768 enabled by default at @option{-O2} and @option{-O3}.
4771 Perform Full Redundancy Elimination (FRE) on trees. The difference
4772 between FRE and PRE is that FRE only considers expressions
4773 that are computed on all paths leading to the redundant computation.
4774 This analysis faster than PRE, though it exposes fewer redundancies.
4775 This flag is enabled by default at @option{-O} and higher.
4777 @item -ftree-copy-prop
4778 Perform copy propagation on trees. This pass eliminates unnecessary
4779 copy operations. This flag is enabled by default at @option{-O} and
4782 @item -ftree-store-copy-prop
4783 Perform copy propagation of memory loads and stores. This pass
4784 eliminates unnecessary copy operations in memory references
4785 (structures, global variables, arrays, etc). This flag is enabled by
4786 default at @option{-O2} and higher.
4789 Perform structural alias analysis on trees. This flag
4790 is enabled by default at @option{-O} and higher.
4793 Perform forward store motion on trees. This flag is
4794 enabled by default at @option{-O} and higher.
4797 Perform sparse conditional constant propagation (CCP) on trees. This
4798 pass only operates on local scalar variables and is enabled by default
4799 at @option{-O} and higher.
4801 @item -ftree-store-ccp
4802 Perform sparse conditional constant propagation (CCP) on trees. This
4803 pass operates on both local scalar variables and memory stores and
4804 loads (global variables, structures, arrays, etc). This flag is
4805 enabled by default at @option{-O2} and higher.
4808 Perform dead code elimination (DCE) on trees. This flag is enabled by
4809 default at @option{-O} and higher.
4811 @item -ftree-dominator-opts
4812 Perform dead code elimination (DCE) on trees. This flag is enabled by
4813 default at @option{-O} and higher.
4816 Perform loop header copying on trees. This is beneficial since it increases
4817 effectiveness of code motion optimizations. It also saves one jump. This flag
4818 is enabled by default at @option{-O} and higher. It is not enabled
4819 for @option{-Os}, since it usually increases code size.
4821 @item -ftree-loop-optimize
4822 Perform loop optimizations on trees. This flag is enabled by default
4823 at @option{-O} and higher.
4825 @item -ftree-loop-linear
4826 Perform linear loop transformations on tree. This flag can improve cache
4827 performance and allow further loop optimizations to take place.
4829 @item -ftree-loop-im
4830 Perform loop invariant motion on trees. This pass moves only invariants that
4831 would be hard to handle at RTL level (function calls, operations that expand to
4832 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
4833 operands of conditions that are invariant out of the loop, so that we can use
4834 just trivial invariantness analysis in loop unswitching. The pass also includes
4837 @item -ftree-loop-ivcanon
4838 Create a canonical counter for number of iterations in the loop for that
4839 determining number of iterations requires complicated analysis. Later
4840 optimizations then may determine the number easily. Useful especially
4841 in connection with unrolling.
4844 Perform induction variable optimizations (strength reduction, induction
4845 variable merging and induction variable elimination) on trees.
4848 Perform scalar replacement of aggregates. This pass replaces structure
4849 references with scalars to prevent committing structures to memory too
4850 early. This flag is enabled by default at @option{-O} and higher.
4852 @item -ftree-copyrename
4853 Perform copy renaming on trees. This pass attempts to rename compiler
4854 temporaries to other variables at copy locations, usually resulting in
4855 variable names which more closely resemble the original variables. This flag
4856 is enabled by default at @option{-O} and higher.
4859 Perform temporary expression replacement during the SSA->normal phase. Single
4860 use/single def temporaries are replaced at their use location with their
4861 defining expression. This results in non-GIMPLE code, but gives the expanders
4862 much more complex trees to work on resulting in better RTL generation. This is
4863 enabled by default at @option{-O} and higher.
4866 Perform live range splitting during the SSA->normal phase. Distinct live
4867 ranges of a variable are split into unique variables, allowing for better
4868 optimization later. This is enabled by default at @option{-O} and higher.
4870 @item -ftree-vectorize
4871 Perform loop vectorization on trees.
4875 Perform tail duplication to enlarge superblock size. This transformation
4876 simplifies the control flow of the function allowing other optimizations to do
4879 @item -funroll-loops
4880 @opindex funroll-loops
4881 Unroll loops whose number of iterations can be determined at compile
4882 time or upon entry to the loop. @option{-funroll-loops} implies both
4883 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4884 option makes code larger, and may or may not make it run faster.
4886 @item -funroll-all-loops
4887 @opindex funroll-all-loops
4888 Unroll all loops, even if their number of iterations is uncertain when
4889 the loop is entered. This usually makes programs run more slowly.
4890 @option{-funroll-all-loops} implies the same options as
4891 @option{-funroll-loops},
4893 @item -fsplit-ivs-in-unroller
4894 @opindex -fsplit-ivs-in-unroller
4895 Enables expressing of values of induction variables in later iterations
4896 of the unrolled loop using the value in the first iteration. This breaks
4897 long dependency chains, thus improving efficiency of the scheduling passes
4898 (for best results, @option{-fweb} should be used as well).
4900 Combination of @option{-fweb} and CSE is often sufficient to obtain the
4901 same effect. However in cases the loop body is more complicated than
4902 a single basic block, this is not reliable. It also does not work at all
4903 on some of the architectures due to restrictions in the CSE pass.
4905 This optimization is enabled by default.
4907 @item -fvariable-expansion-in-unroller
4908 @opindex -fvariable-expansion-in-unroller
4909 With this option, the compiler will create multiple copies of some
4910 local variables when unrolling a loop which can result in superior code.
4912 @item -fprefetch-loop-arrays
4913 @opindex fprefetch-loop-arrays
4914 If supported by the target machine, generate instructions to prefetch
4915 memory to improve the performance of loops that access large arrays.
4917 These options may generate better or worse code; results are highly
4918 dependent on the structure of loops within the source code.
4921 @itemx -fno-peephole2
4922 @opindex fno-peephole
4923 @opindex fno-peephole2
4924 Disable any machine-specific peephole optimizations. The difference
4925 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
4926 are implemented in the compiler; some targets use one, some use the
4927 other, a few use both.
4929 @option{-fpeephole} is enabled by default.
4930 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4932 @item -fno-guess-branch-probability
4933 @opindex fno-guess-branch-probability
4934 Do not guess branch probabilities using heuristics.
4936 GCC will use heuristics to guess branch probabilities if they are
4937 not provided by profiling feedback (@option{-fprofile-arcs}). These
4938 heuristics are based on the control flow graph. If some branch probabilities
4939 are specified by @samp{__builtin_expect}, then the heuristics will be
4940 used to guess branch probabilities for the rest of the control flow graph,
4941 taking the @samp{__builtin_expect} info into account. The interactions
4942 between the heuristics and @samp{__builtin_expect} can be complex, and in
4943 some cases, it may be useful to disable the heuristics so that the effects
4944 of @samp{__builtin_expect} are easier to understand.
4946 The default is @option{-fguess-branch-probability} at levels
4947 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4949 @item -freorder-blocks
4950 @opindex freorder-blocks
4951 Reorder basic blocks in the compiled function in order to reduce number of
4952 taken branches and improve code locality.
4954 Enabled at levels @option{-O2}, @option{-O3}.
4956 @item -freorder-blocks-and-partition
4957 @opindex freorder-blocks-and-partition
4958 In addition to reordering basic blocks in the compiled function, in order
4959 to reduce number of taken branches, partitions hot and cold basic blocks
4960 into separate sections of the assembly and .o files, to improve
4961 paging and cache locality performance.
4963 This optimization is automatically turned off in the presence of
4964 exception handling, for linkonce sections, for functions with a user-defined
4965 section attribute and on any architecture that does not support named
4968 @item -freorder-functions
4969 @opindex freorder-functions
4970 Reorder functions in the object file in order to
4971 improve code locality. This is implemented by using special
4972 subsections @code{.text.hot} for most frequently executed functions and
4973 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
4974 the linker so object file format must support named sections and linker must
4975 place them in a reasonable way.
4977 Also profile feedback must be available in to make this option effective. See
4978 @option{-fprofile-arcs} for details.
4980 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4982 @item -fstrict-aliasing
4983 @opindex fstrict-aliasing
4984 Allows the compiler to assume the strictest aliasing rules applicable to
4985 the language being compiled. For C (and C++), this activates
4986 optimizations based on the type of expressions. In particular, an
4987 object of one type is assumed never to reside at the same address as an
4988 object of a different type, unless the types are almost the same. For
4989 example, an @code{unsigned int} can alias an @code{int}, but not a
4990 @code{void*} or a @code{double}. A character type may alias any other
4993 Pay special attention to code like this:
5006 The practice of reading from a different union member than the one most
5007 recently written to (called ``type-punning'') is common. Even with
5008 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5009 is accessed through the union type. So, the code above will work as
5010 expected. However, this code might not:
5021 Every language that wishes to perform language-specific alias analysis
5022 should define a function that computes, given an @code{tree}
5023 node, an alias set for the node. Nodes in different alias sets are not
5024 allowed to alias. For an example, see the C front-end function
5025 @code{c_get_alias_set}.
5027 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5029 @item -falign-functions
5030 @itemx -falign-functions=@var{n}
5031 @opindex falign-functions
5032 Align the start of functions to the next power-of-two greater than
5033 @var{n}, skipping up to @var{n} bytes. For instance,
5034 @option{-falign-functions=32} aligns functions to the next 32-byte
5035 boundary, but @option{-falign-functions=24} would align to the next
5036 32-byte boundary only if this can be done by skipping 23 bytes or less.
5038 @option{-fno-align-functions} and @option{-falign-functions=1} are
5039 equivalent and mean that functions will not be aligned.
5041 Some assemblers only support this flag when @var{n} is a power of two;
5042 in that case, it is rounded up.
5044 If @var{n} is not specified or is zero, use a machine-dependent default.
5046 Enabled at levels @option{-O2}, @option{-O3}.
5048 @item -falign-labels
5049 @itemx -falign-labels=@var{n}
5050 @opindex falign-labels
5051 Align all branch targets to a power-of-two boundary, skipping up to
5052 @var{n} bytes like @option{-falign-functions}. This option can easily
5053 make code slower, because it must insert dummy operations for when the
5054 branch target is reached in the usual flow of the code.
5056 @option{-fno-align-labels} and @option{-falign-labels=1} are
5057 equivalent and mean that labels will not be aligned.
5059 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5060 are greater than this value, then their values are used instead.
5062 If @var{n} is not specified or is zero, use a machine-dependent default
5063 which is very likely to be @samp{1}, meaning no alignment.
5065 Enabled at levels @option{-O2}, @option{-O3}.
5068 @itemx -falign-loops=@var{n}
5069 @opindex falign-loops
5070 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5071 like @option{-falign-functions}. The hope is that the loop will be
5072 executed many times, which will make up for any execution of the dummy
5075 @option{-fno-align-loops} and @option{-falign-loops=1} are
5076 equivalent and mean that loops will not be aligned.
5078 If @var{n} is not specified or is zero, use a machine-dependent default.
5080 Enabled at levels @option{-O2}, @option{-O3}.
5083 @itemx -falign-jumps=@var{n}
5084 @opindex falign-jumps
5085 Align branch targets to a power-of-two boundary, for branch targets
5086 where the targets can only be reached by jumping, skipping up to @var{n}
5087 bytes like @option{-falign-functions}. In this case, no dummy operations
5090 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5091 equivalent and mean that loops will not be aligned.
5093 If @var{n} is not specified or is zero, use a machine-dependent default.
5095 Enabled at levels @option{-O2}, @option{-O3}.
5097 @item -funit-at-a-time
5098 @opindex funit-at-a-time
5099 Parse the whole compilation unit before starting to produce code.
5100 This allows some extra optimizations to take place but consumes
5101 more memory (in general). There are some compatibility issues
5102 with @emph{unit-at-at-time} mode:
5105 enabling @emph{unit-at-a-time} mode may change the order
5106 in which functions, variables, and top-level @code{asm} statements
5107 are emitted, and will likely break code relying on some particular
5108 ordering. The majority of such top-level @code{asm} statements,
5109 though, can be replaced by @code{section} attributes.
5112 @emph{unit-at-a-time} mode removes unreferenced static variables
5113 and functions are removed. This may result in undefined references
5114 when an @code{asm} statement refers directly to variables or functions
5115 that are otherwise unused. In that case either the variable/function
5116 shall be listed as an operand of the @code{asm} statement operand or,
5117 in the case of top-level @code{asm} statements the attribute @code{used}
5118 shall be used on the declaration.
5121 Static functions now can use non-standard passing conventions that
5122 may break @code{asm} statements calling functions directly. Again,
5123 attribute @code{used} will prevent this behavior.
5126 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5127 but this scheme may not be supported by future releases of GCC@.
5129 Enabled at levels @option{-O2}, @option{-O3}.
5133 Constructs webs as commonly used for register allocation purposes and assign
5134 each web individual pseudo register. This allows the register allocation pass
5135 to operate on pseudos directly, but also strengthens several other optimization
5136 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5137 however, make debugging impossible, since variables will no longer stay in a
5140 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os},
5141 on targets where the default format for debugging information supports
5144 @item -fno-cprop-registers
5145 @opindex fno-cprop-registers
5146 After register allocation and post-register allocation instruction splitting,
5147 we perform a copy-propagation pass to try to reduce scheduling dependencies
5148 and occasionally eliminate the copy.
5150 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5152 @item -fprofile-generate
5153 @opindex fprofile-generate
5155 Enable options usually used for instrumenting application to produce
5156 profile useful for later recompilation with profile feedback based
5157 optimization. You must use @option{-fprofile-generate} both when
5158 compiling and when linking your program.
5160 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5163 @opindex fprofile-use
5164 Enable profile feedback directed optimizations, and optimizations
5165 generally profitable only with profile feedback available.
5167 The following options are enabled: @code{-fbranch-probabilities},
5168 @code{-fvpt}, @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}.
5172 The following options control compiler behavior regarding floating
5173 point arithmetic. These options trade off between speed and
5174 correctness. All must be specifically enabled.
5178 @opindex ffloat-store
5179 Do not store floating point variables in registers, and inhibit other
5180 options that might change whether a floating point value is taken from a
5183 @cindex floating point precision
5184 This option prevents undesirable excess precision on machines such as
5185 the 68000 where the floating registers (of the 68881) keep more
5186 precision than a @code{double} is supposed to have. Similarly for the
5187 x86 architecture. For most programs, the excess precision does only
5188 good, but a few programs rely on the precise definition of IEEE floating
5189 point. Use @option{-ffloat-store} for such programs, after modifying
5190 them to store all pertinent intermediate computations into variables.
5194 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5195 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5196 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5197 and @option{fcx-limited-range}.
5199 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5201 This option should never be turned on by any @option{-O} option since
5202 it can result in incorrect output for programs which depend on
5203 an exact implementation of IEEE or ISO rules/specifications for
5206 @item -fno-math-errno
5207 @opindex fno-math-errno
5208 Do not set ERRNO after calling math functions that are executed
5209 with a single instruction, e.g., sqrt. A program that relies on
5210 IEEE exceptions for math error handling may want to use this flag
5211 for speed while maintaining IEEE arithmetic compatibility.
5213 This option should never be turned on by any @option{-O} option since
5214 it can result in incorrect output for programs which depend on
5215 an exact implementation of IEEE or ISO rules/specifications for
5218 The default is @option{-fmath-errno}.
5220 @item -funsafe-math-optimizations
5221 @opindex funsafe-math-optimizations
5222 Allow optimizations for floating-point arithmetic that (a) assume
5223 that arguments and results are valid and (b) may violate IEEE or
5224 ANSI standards. When used at link-time, it may include libraries
5225 or startup files that change the default FPU control word or other
5226 similar optimizations.
5228 This option should never be turned on by any @option{-O} option since
5229 it can result in incorrect output for programs which depend on
5230 an exact implementation of IEEE or ISO rules/specifications for
5233 The default is @option{-fno-unsafe-math-optimizations}.
5235 @item -ffinite-math-only
5236 @opindex ffinite-math-only
5237 Allow optimizations for floating-point arithmetic that assume
5238 that arguments and results are not NaNs or +-Infs.
5240 This option should never be turned on by any @option{-O} option since
5241 it can result in incorrect output for programs which depend on
5242 an exact implementation of IEEE or ISO rules/specifications.
5244 The default is @option{-fno-finite-math-only}.
5246 @item -fno-trapping-math
5247 @opindex fno-trapping-math
5248 Compile code assuming that floating-point operations cannot generate
5249 user-visible traps. These traps include division by zero, overflow,
5250 underflow, inexact result and invalid operation. This option implies
5251 @option{-fno-signaling-nans}. Setting this option may allow faster
5252 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5254 This option should never be turned on by any @option{-O} option since
5255 it can result in incorrect output for programs which depend on
5256 an exact implementation of IEEE or ISO rules/specifications for
5259 The default is @option{-ftrapping-math}.
5261 @item -frounding-math
5262 @opindex frounding-math
5263 Disable transformations and optimizations that assume default floating
5264 point rounding behavior. This is round-to-zero for all floating point
5265 to integer conversions, and round-to-nearest for all other arithmetic
5266 truncations. This option should be specified for programs that change
5267 the FP rounding mode dynamically, or that may be executed with a
5268 non-default rounding mode. This option disables constant folding of
5269 floating point expressions at compile-time (which may be affected by
5270 rounding mode) and arithmetic transformations that are unsafe in the
5271 presence of sign-dependent rounding modes.
5273 The default is @option{-fno-rounding-math}.
5275 This option is experimental and does not currently guarantee to
5276 disable all GCC optimizations that are affected by rounding mode.
5277 Future versions of GCC may provide finer control of this setting
5278 using C99's @code{FENV_ACCESS} pragma. This command line option
5279 will be used to specify the default state for @code{FENV_ACCESS}.
5281 @item -fsignaling-nans
5282 @opindex fsignaling-nans
5283 Compile code assuming that IEEE signaling NaNs may generate user-visible
5284 traps during floating-point operations. Setting this option disables
5285 optimizations that may change the number of exceptions visible with
5286 signaling NaNs. This option implies @option{-ftrapping-math}.
5288 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5291 The default is @option{-fno-signaling-nans}.
5293 This option is experimental and does not currently guarantee to
5294 disable all GCC optimizations that affect signaling NaN behavior.
5296 @item -fsingle-precision-constant
5297 @opindex fsingle-precision-constant
5298 Treat floating point constant as single precision constant instead of
5299 implicitly converting it to double precision constant.
5301 @item -fcx-limited-range
5302 @itemx -fno-cx-limited-range
5303 @opindex fcx-limited-range
5304 @opindex fno-cx-limited-range
5305 When enabled, this option states that a range reduction step is not
5306 needed when performing complex division. The default is
5307 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5309 This option controls the default setting of the ISO C99
5310 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5315 The following options control optimizations that may improve
5316 performance, but are not enabled by any @option{-O} options. This
5317 section includes experimental options that may produce broken code.
5320 @item -fbranch-probabilities
5321 @opindex fbranch-probabilities
5322 After running a program compiled with @option{-fprofile-arcs}
5323 (@pxref{Debugging Options,, Options for Debugging Your Program or
5324 @command{gcc}}), you can compile it a second time using
5325 @option{-fbranch-probabilities}, to improve optimizations based on
5326 the number of times each branch was taken. When the program
5327 compiled with @option{-fprofile-arcs} exits it saves arc execution
5328 counts to a file called @file{@var{sourcename}.gcda} for each source
5329 file The information in this data file is very dependent on the
5330 structure of the generated code, so you must use the same source code
5331 and the same optimization options for both compilations.
5333 With @option{-fbranch-probabilities}, GCC puts a
5334 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5335 These can be used to improve optimization. Currently, they are only
5336 used in one place: in @file{reorg.c}, instead of guessing which path a
5337 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5338 exactly determine which path is taken more often.
5340 @item -fprofile-values
5341 @opindex fprofile-values
5342 If combined with @option{-fprofile-arcs}, it adds code so that some
5343 data about values of expressions in the program is gathered.
5345 With @option{-fbranch-probabilities}, it reads back the data gathered
5346 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5347 notes to instructions for their later usage in optimizations.
5349 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5353 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5354 a code to gather information about values of expressions.
5356 With @option{-fbranch-probabilities}, it reads back the data gathered
5357 and actually performs the optimizations based on them.
5358 Currently the optimizations include specialization of division operation
5359 using the knowledge about the value of the denominator.
5361 @item -fspeculative-prefetching
5362 @opindex fspeculative-prefetching
5363 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5364 a code to gather information about addresses of memory references in the
5367 With @option{-fbranch-probabilities}, it reads back the data gathered
5368 and issues prefetch instructions according to them. In addition to the opportunities
5369 noticed by @option{-fprefetch-loop-arrays}, it also notices more complicated
5370 memory access patterns---for example accesses to the data stored in linked
5371 list whose elements are usually allocated sequentially.
5373 In order to prevent issuing double prefetches, usage of
5374 @option{-fspeculative-prefetching} implies @option{-fno-prefetch-loop-arrays}.
5376 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5378 @item -frename-registers
5379 @opindex frename-registers
5380 Attempt to avoid false dependencies in scheduled code by making use
5381 of registers left over after register allocation. This optimization
5382 will most benefit processors with lots of registers. Depending on the
5383 debug information format adopted by the target, however, it can
5384 make debugging impossible, since variables will no longer stay in
5385 a ``home register''.
5387 Not enabled by default at any level because it has known bugs.
5391 Perform tail duplication to enlarge superblock size. This transformation
5392 simplifies the control flow of the function allowing other optimizations to do
5395 Enabled with @option{-fprofile-use}.
5397 @item -funroll-loops
5398 @opindex funroll-loops
5399 Unroll loops whose number of iterations can be determined at compile time or
5400 upon entry to the loop. @option{-funroll-loops} implies
5401 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
5402 (i.e.@: complete removal of loops with small constant number of iterations).
5403 This option makes code larger, and may or may not make it run faster.
5405 Enabled with @option{-fprofile-use}.
5407 @item -funroll-all-loops
5408 @opindex funroll-all-loops
5409 Unroll all loops, even if their number of iterations is uncertain when
5410 the loop is entered. This usually makes programs run more slowly.
5411 @option{-funroll-all-loops} implies the same options as
5412 @option{-funroll-loops}.
5415 @opindex fpeel-loops
5416 Peels the loops for that there is enough information that they do not
5417 roll much (from profile feedback). It also turns on complete loop peeling
5418 (i.e.@: complete removal of loops with small constant number of iterations).
5420 Enabled with @option{-fprofile-use}.
5422 @item -fmove-loop-invariants
5423 @opindex fmove-loop-invariants
5424 Enables the loop invariant motion pass in the new loop optimizer. Enabled
5425 at level @option{-O1}
5427 @item -funswitch-loops
5428 @opindex funswitch-loops
5429 Move branches with loop invariant conditions out of the loop, with duplicates
5430 of the loop on both branches (modified according to result of the condition).
5432 @item -fprefetch-loop-arrays
5433 @opindex fprefetch-loop-arrays
5434 If supported by the target machine, generate instructions to prefetch
5435 memory to improve the performance of loops that access large arrays.
5437 Disabled at level @option{-Os}.
5439 @item -ffunction-sections
5440 @itemx -fdata-sections
5441 @opindex ffunction-sections
5442 @opindex fdata-sections
5443 Place each function or data item into its own section in the output
5444 file if the target supports arbitrary sections. The name of the
5445 function or the name of the data item determines the section's name
5448 Use these options on systems where the linker can perform optimizations
5449 to improve locality of reference in the instruction space. Most systems
5450 using the ELF object format and SPARC processors running Solaris 2 have
5451 linkers with such optimizations. AIX may have these optimizations in
5454 Only use these options when there are significant benefits from doing
5455 so. When you specify these options, the assembler and linker will
5456 create larger object and executable files and will also be slower.
5457 You will not be able to use @code{gprof} on all systems if you
5458 specify this option and you may have problems with debugging if
5459 you specify both this option and @option{-g}.
5461 @item -fbranch-target-load-optimize
5462 @opindex fbranch-target-load-optimize
5463 Perform branch target register load optimization before prologue / epilogue
5465 The use of target registers can typically be exposed only during reload,
5466 thus hoisting loads out of loops and doing inter-block scheduling needs
5467 a separate optimization pass.
5469 @item -fbranch-target-load-optimize2
5470 @opindex fbranch-target-load-optimize2
5471 Perform branch target register load optimization after prologue / epilogue
5474 @item -fbtr-bb-exclusive
5475 @opindex fbtr-bb-exclusive
5476 When performing branch target register load optimization, don't reuse
5477 branch target registers in within any basic block.
5479 @item --param @var{name}=@var{value}
5481 In some places, GCC uses various constants to control the amount of
5482 optimization that is done. For example, GCC will not inline functions
5483 that contain more that a certain number of instructions. You can
5484 control some of these constants on the command-line using the
5485 @option{--param} option.
5487 The names of specific parameters, and the meaning of the values, are
5488 tied to the internals of the compiler, and are subject to change
5489 without notice in future releases.
5491 In each case, the @var{value} is an integer. The allowable choices for
5492 @var{name} are given in the following table:
5495 @item salias-max-implicit-fields
5496 The maximum number of fields in a variable without direct
5497 structure accesses for which structure aliasing will consider trying
5498 to track each field. The default is 5
5500 @item sra-max-structure-size
5501 The maximum structure size, in bytes, at which the scalar replacement
5502 of aggregates (SRA) optimization will perform block copies. The
5503 default value, 0, implies that GCC will select the most appropriate
5506 @item sra-field-structure-ratio
5507 The threshold ratio (as a percentage) between instantiated fields and
5508 the complete structure size. We say that if the ratio of the number
5509 of bytes in instantiated fields to the number of bytes in the complete
5510 structure exceeds this parameter, then block copies are not used. The
5513 @item max-crossjump-edges
5514 The maximum number of incoming edges to consider for crossjumping.
5515 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5516 the number of edges incoming to each block. Increasing values mean
5517 more aggressive optimization, making the compile time increase with
5518 probably small improvement in executable size.
5520 @item min-crossjump-insns
5521 The minimum number of instructions which must be matched at the end
5522 of two blocks before crossjumping will be performed on them. This
5523 value is ignored in the case where all instructions in the block being
5524 crossjumped from are matched. The default value is 5.
5526 @item max-goto-duplication-insns
5527 The maximum number of instructions to duplicate to a block that jumps
5528 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5529 passes, GCC factors computed gotos early in the compilation process,
5530 and unfactors them as late as possible. Only computed jumps at the
5531 end of a basic blocks with no more than max-goto-duplication-insns are
5532 unfactored. The default value is 8.
5534 @item max-delay-slot-insn-search
5535 The maximum number of instructions to consider when looking for an
5536 instruction to fill a delay slot. If more than this arbitrary number of
5537 instructions is searched, the time savings from filling the delay slot
5538 will be minimal so stop searching. Increasing values mean more
5539 aggressive optimization, making the compile time increase with probably
5540 small improvement in executable run time.
5542 @item max-delay-slot-live-search
5543 When trying to fill delay slots, the maximum number of instructions to
5544 consider when searching for a block with valid live register
5545 information. Increasing this arbitrarily chosen value means more
5546 aggressive optimization, increasing the compile time. This parameter
5547 should be removed when the delay slot code is rewritten to maintain the
5550 @item max-gcse-memory
5551 The approximate maximum amount of memory that will be allocated in
5552 order to perform the global common subexpression elimination
5553 optimization. If more memory than specified is required, the
5554 optimization will not be done.
5556 @item max-gcse-passes
5557 The maximum number of passes of GCSE to run. The default is 1.
5559 @item max-pending-list-length
5560 The maximum number of pending dependencies scheduling will allow
5561 before flushing the current state and starting over. Large functions
5562 with few branches or calls can create excessively large lists which
5563 needlessly consume memory and resources.
5565 @item max-inline-insns-single
5566 Several parameters control the tree inliner used in gcc.
5567 This number sets the maximum number of instructions (counted in GCC's
5568 internal representation) in a single function that the tree inliner
5569 will consider for inlining. This only affects functions declared
5570 inline and methods implemented in a class declaration (C++).
5571 The default value is 450.
5573 @item max-inline-insns-auto
5574 When you use @option{-finline-functions} (included in @option{-O3}),
5575 a lot of functions that would otherwise not be considered for inlining
5576 by the compiler will be investigated. To those functions, a different
5577 (more restrictive) limit compared to functions declared inline can
5579 The default value is 90.
5581 @item large-function-insns
5582 The limit specifying really large functions. For functions larger than this
5583 limit after inlining inlining is constrained by
5584 @option{--param large-function-growth}. This parameter is useful primarily
5585 to avoid extreme compilation time caused by non-linear algorithms used by the
5587 This parameter is ignored when @option{-funit-at-a-time} is not used.
5588 The default value is 2700.
5590 @item large-function-growth
5591 Specifies maximal growth of large function caused by inlining in percents.
5592 This parameter is ignored when @option{-funit-at-a-time} is not used.
5593 The default value is 100 which limits large function growth to 2.0 times
5596 @item inline-unit-growth
5597 Specifies maximal overall growth of the compilation unit caused by inlining.
5598 This parameter is ignored when @option{-funit-at-a-time} is not used.
5599 The default value is 50 which limits unit growth to 1.5 times the original
5602 @item max-inline-insns-recursive
5603 @itemx max-inline-insns-recursive-auto
5604 Specifies maximum number of instructions out-of-line copy of self recursive inline
5605 function can grow into by performing recursive inlining.
5607 For functions declared inline @option{--param max-inline-insns-recursive} is
5608 taken into acount. For function not declared inline, recursive inlining
5609 happens only when @option{-finline-functions} (included in @option{-O3}) is
5610 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5611 default value is 450.
5613 @item max-inline-recursive-depth
5614 @itemx max-inline-recursive-depth-auto
5615 Specifies maximum recursion depth used by the recursive inlining.
5617 For functions declared inline @option{--param max-inline-recursive-depth} is
5618 taken into acount. For function not declared inline, recursive inlining
5619 happens only when @option{-finline-functions} (included in @option{-O3}) is
5620 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5621 default value is 450.
5623 @item inline-call-cost
5624 Specify cost of call instruction relative to simple arithmetics operations
5625 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
5626 functions and at the same time increases size of leaf function that is believed to
5627 reduce function size by being inlined. In effect it increases amount of
5628 inlining for code having large abstraction penalty (many functions that just
5629 pass the arguments to other functions) and decrease inlining for code with low
5630 abstraction penalty. The default value is 16.
5632 @item max-unrolled-insns
5633 The maximum number of instructions that a loop should have if that loop
5634 is unrolled, and if the loop is unrolled, it determines how many times
5635 the loop code is unrolled.
5637 @item max-average-unrolled-insns
5638 The maximum number of instructions biased by probabilities of their execution
5639 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5640 it determines how many times the loop code is unrolled.
5642 @item max-unroll-times
5643 The maximum number of unrollings of a single loop.
5645 @item max-peeled-insns
5646 The maximum number of instructions that a loop should have if that loop
5647 is peeled, and if the loop is peeled, it determines how many times
5648 the loop code is peeled.
5650 @item max-peel-times
5651 The maximum number of peelings of a single loop.
5653 @item max-completely-peeled-insns
5654 The maximum number of insns of a completely peeled loop.
5656 @item max-completely-peel-times
5657 The maximum number of iterations of a loop to be suitable for complete peeling.
5659 @item max-unswitch-insns
5660 The maximum number of insns of an unswitched loop.
5662 @item max-unswitch-level
5663 The maximum number of branches unswitched in a single loop.
5666 The minimum cost of an expensive expression in the loop invariant motion.
5668 @item iv-consider-all-candidates-bound
5669 Bound on number of candidates for induction variables below that
5670 all candidates are considered for each use in induction variable
5671 optimizations. Only the most relevant candidates are considered
5672 if there are more candidates, to avoid quadratic time complexity.
5674 @item iv-max-considered-uses
5675 The induction variable optimizations give up on loops that contain more
5676 induction variable uses.
5678 @item iv-always-prune-cand-set-bound
5679 If number of candidates in the set is smaller than this value,
5680 we always try to remove unnecessary ivs from the set during its
5681 optimization when a new iv is added to the set.
5683 @item max-iterations-to-track
5685 The maximum number of iterations of a loop the brute force algorithm
5686 for analysis of # of iterations of the loop tries to evaluate.
5688 @item hot-bb-count-fraction
5689 Select fraction of the maximal count of repetitions of basic block in program
5690 given basic block needs to have to be considered hot.
5692 @item hot-bb-frequency-fraction
5693 Select fraction of the maximal frequency of executions of basic block in
5694 function given basic block needs to have to be considered hot
5696 @item tracer-dynamic-coverage
5697 @itemx tracer-dynamic-coverage-feedback
5699 This value is used to limit superblock formation once the given percentage of
5700 executed instructions is covered. This limits unnecessary code size
5703 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5704 feedback is available. The real profiles (as opposed to statically estimated
5705 ones) are much less balanced allowing the threshold to be larger value.
5707 @item tracer-max-code-growth
5708 Stop tail duplication once code growth has reached given percentage. This is
5709 rather hokey argument, as most of the duplicates will be eliminated later in
5710 cross jumping, so it may be set to much higher values than is the desired code
5713 @item tracer-min-branch-ratio
5715 Stop reverse growth when the reverse probability of best edge is less than this
5716 threshold (in percent).
5718 @item tracer-min-branch-ratio
5719 @itemx tracer-min-branch-ratio-feedback
5721 Stop forward growth if the best edge do have probability lower than this
5724 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
5725 compilation for profile feedback and one for compilation without. The value
5726 for compilation with profile feedback needs to be more conservative (higher) in
5727 order to make tracer effective.
5729 @item max-cse-path-length
5731 Maximum number of basic blocks on path that cse considers. The default is 10.
5733 @item global-var-threshold
5735 Counts the number of function calls (@var{n}) and the number of
5736 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
5737 single artificial variable will be created to represent all the
5738 call-clobbered variables at function call sites. This artificial
5739 variable will then be made to alias every call-clobbered variable.
5740 (done as @code{int * size_t} on the host machine; beware overflow).
5742 @item max-aliased-vops
5744 Maximum number of virtual operands allowed to represent aliases
5745 before triggering the alias grouping heuristic. Alias grouping
5746 reduces compile times and memory consumption needed for aliasing at
5747 the expense of precision loss in alias information.
5749 @item ggc-min-expand
5751 GCC uses a garbage collector to manage its own memory allocation. This
5752 parameter specifies the minimum percentage by which the garbage
5753 collector's heap should be allowed to expand between collections.
5754 Tuning this may improve compilation speed; it has no effect on code
5757 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
5758 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
5759 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
5760 GCC is not able to calculate RAM on a particular platform, the lower
5761 bound of 30% is used. Setting this parameter and
5762 @option{ggc-min-heapsize} to zero causes a full collection to occur at
5763 every opportunity. This is extremely slow, but can be useful for
5766 @item ggc-min-heapsize
5768 Minimum size of the garbage collector's heap before it begins bothering
5769 to collect garbage. The first collection occurs after the heap expands
5770 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
5771 tuning this may improve compilation speed, and has no effect on code
5774 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
5775 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
5776 with a lower bound of 4096 (four megabytes) and an upper bound of
5777 131072 (128 megabytes). If GCC is not able to calculate RAM on a
5778 particular platform, the lower bound is used. Setting this parameter
5779 very large effectively disables garbage collection. Setting this
5780 parameter and @option{ggc-min-expand} to zero causes a full collection
5781 to occur at every opportunity.
5783 @item max-reload-search-insns
5784 The maximum number of instruction reload should look backward for equivalent
5785 register. Increasing values mean more aggressive optimization, making the
5786 compile time increase with probably slightly better performance. The default
5789 @item max-cselib-memory-location
5790 The maximum number of memory locations cselib should take into acount.
5791 Increasing values mean more aggressive optimization, making the compile time
5792 increase with probably slightly better performance. The default value is 500.
5794 @item reorder-blocks-duplicate
5795 @itemx reorder-blocks-duplicate-feedback
5797 Used by basic block reordering pass to decide whether to use unconditional
5798 branch or duplicate the code on its destination. Code is duplicated when its
5799 estimated size is smaller than this value multiplied by the estimated size of
5800 unconditional jump in the hot spots of the program.
5802 The @option{reorder-block-duplicate-feedback} is used only when profile
5803 feedback is available and may be set to higher values than
5804 @option{reorder-block-duplicate} since information about the hot spots is more
5807 @item max-sched-region-blocks
5808 The maximum number of blocks in a region to be considered for
5809 interblock scheduling. The default value is 10.
5811 @item max-sched-region-insns
5812 The maximum number of insns in a region to be considered for
5813 interblock scheduling. The default value is 100.
5815 @item max-last-value-rtl
5817 The maximum size measured as number of RTLs that can be recorded in an expression
5818 in combiner for a pseudo register as last known value of that register. The default
5821 @item integer-share-limit
5822 Small integer constants can use a shared data structure, reducing the
5823 compiler's memory usage and increasing its speed. This sets the maximum
5824 value of a shared integer constant's. The default value is 256.
5829 @node Preprocessor Options
5830 @section Options Controlling the Preprocessor
5831 @cindex preprocessor options
5832 @cindex options, preprocessor
5834 These options control the C preprocessor, which is run on each C source
5835 file before actual compilation.
5837 If you use the @option{-E} option, nothing is done except preprocessing.
5838 Some of these options make sense only together with @option{-E} because
5839 they cause the preprocessor output to be unsuitable for actual
5844 You can use @option{-Wp,@var{option}} to bypass the compiler driver
5845 and pass @var{option} directly through to the preprocessor. If
5846 @var{option} contains commas, it is split into multiple options at the
5847 commas. However, many options are modified, translated or interpreted
5848 by the compiler driver before being passed to the preprocessor, and
5849 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
5850 interface is undocumented and subject to change, so whenever possible
5851 you should avoid using @option{-Wp} and let the driver handle the
5854 @item -Xpreprocessor @var{option}
5855 @opindex preprocessor
5856 Pass @var{option} as an option to the preprocessor. You can use this to
5857 supply system-specific preprocessor options which GCC does not know how to
5860 If you want to pass an option that takes an argument, you must use
5861 @option{-Xpreprocessor} twice, once for the option and once for the argument.
5864 @include cppopts.texi
5866 @node Assembler Options
5867 @section Passing Options to the Assembler
5869 @c prevent bad page break with this line
5870 You can pass options to the assembler.
5873 @item -Wa,@var{option}
5875 Pass @var{option} as an option to the assembler. If @var{option}
5876 contains commas, it is split into multiple options at the commas.
5878 @item -Xassembler @var{option}
5880 Pass @var{option} as an option to the assembler. You can use this to
5881 supply system-specific assembler options which GCC does not know how to
5884 If you want to pass an option that takes an argument, you must use
5885 @option{-Xassembler} twice, once for the option and once for the argument.
5890 @section Options for Linking
5891 @cindex link options
5892 @cindex options, linking
5894 These options come into play when the compiler links object files into
5895 an executable output file. They are meaningless if the compiler is
5896 not doing a link step.
5900 @item @var{object-file-name}
5901 A file name that does not end in a special recognized suffix is
5902 considered to name an object file or library. (Object files are
5903 distinguished from libraries by the linker according to the file
5904 contents.) If linking is done, these object files are used as input
5913 If any of these options is used, then the linker is not run, and
5914 object file names should not be used as arguments. @xref{Overall
5918 @item -l@var{library}
5919 @itemx -l @var{library}
5921 Search the library named @var{library} when linking. (The second
5922 alternative with the library as a separate argument is only for
5923 POSIX compliance and is not recommended.)
5925 It makes a difference where in the command you write this option; the
5926 linker searches and processes libraries and object files in the order they
5927 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
5928 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
5929 to functions in @samp{z}, those functions may not be loaded.
5931 The linker searches a standard list of directories for the library,
5932 which is actually a file named @file{lib@var{library}.a}. The linker
5933 then uses this file as if it had been specified precisely by name.
5935 The directories searched include several standard system directories
5936 plus any that you specify with @option{-L}.
5938 Normally the files found this way are library files---archive files
5939 whose members are object files. The linker handles an archive file by
5940 scanning through it for members which define symbols that have so far
5941 been referenced but not defined. But if the file that is found is an
5942 ordinary object file, it is linked in the usual fashion. The only
5943 difference between using an @option{-l} option and specifying a file name
5944 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
5945 and searches several directories.
5949 You need this special case of the @option{-l} option in order to
5950 link an Objective-C or Objective-C++ program.
5953 @opindex nostartfiles
5954 Do not use the standard system startup files when linking.
5955 The standard system libraries are used normally, unless @option{-nostdlib}
5956 or @option{-nodefaultlibs} is used.
5958 @item -nodefaultlibs
5959 @opindex nodefaultlibs
5960 Do not use the standard system libraries when linking.
5961 Only the libraries you specify will be passed to the linker.
5962 The standard startup files are used normally, unless @option{-nostartfiles}
5963 is used. The compiler may generate calls to @code{memcmp},
5964 @code{memset}, @code{memcpy} and @code{memmove}.
5965 These entries are usually resolved by entries in
5966 libc. These entry points should be supplied through some other
5967 mechanism when this option is specified.
5971 Do not use the standard system startup files or libraries when linking.
5972 No startup files and only the libraries you specify will be passed to
5973 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
5974 @code{memcpy} and @code{memmove}.
5975 These entries are usually resolved by entries in
5976 libc. These entry points should be supplied through some other
5977 mechanism when this option is specified.
5979 @cindex @option{-lgcc}, use with @option{-nostdlib}
5980 @cindex @option{-nostdlib} and unresolved references
5981 @cindex unresolved references and @option{-nostdlib}
5982 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
5983 @cindex @option{-nodefaultlibs} and unresolved references
5984 @cindex unresolved references and @option{-nodefaultlibs}
5985 One of the standard libraries bypassed by @option{-nostdlib} and
5986 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
5987 that GCC uses to overcome shortcomings of particular machines, or special
5988 needs for some languages.
5989 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
5990 Collection (GCC) Internals},
5991 for more discussion of @file{libgcc.a}.)
5992 In most cases, you need @file{libgcc.a} even when you want to avoid
5993 other standard libraries. In other words, when you specify @option{-nostdlib}
5994 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
5995 This ensures that you have no unresolved references to internal GCC
5996 library subroutines. (For example, @samp{__main}, used to ensure C++
5997 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
5998 GNU Compiler Collection (GCC) Internals}.)
6002 Produce a position independent executable on targets which support it.
6003 For predictable results, you must also specify the same set of options
6004 that were used to generate code (@option{-fpie}, @option{-fPIE},
6005 or model suboptions) when you specify this option.
6009 Remove all symbol table and relocation information from the executable.
6013 On systems that support dynamic linking, this prevents linking with the shared
6014 libraries. On other systems, this option has no effect.
6018 Produce a shared object which can then be linked with other objects to
6019 form an executable. Not all systems support this option. For predictable
6020 results, you must also specify the same set of options that were used to
6021 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6022 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6023 needs to build supplementary stub code for constructors to work. On
6024 multi-libbed systems, @samp{gcc -shared} must select the correct support
6025 libraries to link against. Failing to supply the correct flags may lead
6026 to subtle defects. Supplying them in cases where they are not necessary
6029 @item -shared-libgcc
6030 @itemx -static-libgcc
6031 @opindex shared-libgcc
6032 @opindex static-libgcc
6033 On systems that provide @file{libgcc} as a shared library, these options
6034 force the use of either the shared or static version respectively.
6035 If no shared version of @file{libgcc} was built when the compiler was
6036 configured, these options have no effect.
6038 There are several situations in which an application should use the
6039 shared @file{libgcc} instead of the static version. The most common
6040 of these is when the application wishes to throw and catch exceptions
6041 across different shared libraries. In that case, each of the libraries
6042 as well as the application itself should use the shared @file{libgcc}.
6044 Therefore, the G++ and GCJ drivers automatically add
6045 @option{-shared-libgcc} whenever you build a shared library or a main
6046 executable, because C++ and Java programs typically use exceptions, so
6047 this is the right thing to do.
6049 If, instead, you use the GCC driver to create shared libraries, you may
6050 find that they will not always be linked with the shared @file{libgcc}.
6051 If GCC finds, at its configuration time, that you have a non-GNU linker
6052 or a GNU linker that does not support option @option{--eh-frame-hdr},
6053 it will link the shared version of @file{libgcc} into shared libraries
6054 by default. Otherwise, it will take advantage of the linker and optimize
6055 away the linking with the shared version of @file{libgcc}, linking with
6056 the static version of libgcc by default. This allows exceptions to
6057 propagate through such shared libraries, without incurring relocation
6058 costs at library load time.
6060 However, if a library or main executable is supposed to throw or catch
6061 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6062 for the languages used in the program, or using the option
6063 @option{-shared-libgcc}, such that it is linked with the shared
6068 Bind references to global symbols when building a shared object. Warn
6069 about any unresolved references (unless overridden by the link editor
6070 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6073 @item -Xlinker @var{option}
6075 Pass @var{option} as an option to the linker. You can use this to
6076 supply system-specific linker options which GCC does not know how to
6079 If you want to pass an option that takes an argument, you must use
6080 @option{-Xlinker} twice, once for the option and once for the argument.
6081 For example, to pass @option{-assert definitions}, you must write
6082 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6083 @option{-Xlinker "-assert definitions"}, because this passes the entire
6084 string as a single argument, which is not what the linker expects.
6086 @item -Wl,@var{option}
6088 Pass @var{option} as an option to the linker. If @var{option} contains
6089 commas, it is split into multiple options at the commas.
6091 @item -u @var{symbol}
6093 Pretend the symbol @var{symbol} is undefined, to force linking of
6094 library modules to define it. You can use @option{-u} multiple times with
6095 different symbols to force loading of additional library modules.
6098 @node Directory Options
6099 @section Options for Directory Search
6100 @cindex directory options
6101 @cindex options, directory search
6104 These options specify directories to search for header files, for
6105 libraries and for parts of the compiler:
6110 Add the directory @var{dir} to the head of the list of directories to be
6111 searched for header files. This can be used to override a system header
6112 file, substituting your own version, since these directories are
6113 searched before the system header file directories. However, you should
6114 not use this option to add directories that contain vendor-supplied
6115 system header files (use @option{-isystem} for that). If you use more than
6116 one @option{-I} option, the directories are scanned in left-to-right
6117 order; the standard system directories come after.
6119 If a standard system include directory, or a directory specified with
6120 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6121 option will be ignored. The directory will still be searched but as a
6122 system directory at its normal position in the system include chain.
6123 This is to ensure that GCC's procedure to fix buggy system headers and
6124 the ordering for the include_next directive are not inadvertently changed.
6125 If you really need to change the search order for system directories,
6126 use the @option{-nostdinc} and/or @option{-isystem} options.
6128 @item -iquote@var{dir}
6130 Add the directory @var{dir} to the head of the list of directories to
6131 be searched for header files only for the case of @samp{#include
6132 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6133 otherwise just like @option{-I}.
6137 Add directory @var{dir} to the list of directories to be searched
6140 @item -B@var{prefix}
6142 This option specifies where to find the executables, libraries,
6143 include files, and data files of the compiler itself.
6145 The compiler driver program runs one or more of the subprograms
6146 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6147 @var{prefix} as a prefix for each program it tries to run, both with and
6148 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6150 For each subprogram to be run, the compiler driver first tries the
6151 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6152 was not specified, the driver tries two standard prefixes, which are
6153 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6154 those results in a file name that is found, the unmodified program
6155 name is searched for using the directories specified in your
6156 @env{PATH} environment variable.
6158 The compiler will check to see if the path provided by the @option{-B}
6159 refers to a directory, and if necessary it will add a directory
6160 separator character at the end of the path.
6162 @option{-B} prefixes that effectively specify directory names also apply
6163 to libraries in the linker, because the compiler translates these
6164 options into @option{-L} options for the linker. They also apply to
6165 includes files in the preprocessor, because the compiler translates these
6166 options into @option{-isystem} options for the preprocessor. In this case,
6167 the compiler appends @samp{include} to the prefix.
6169 The run-time support file @file{libgcc.a} can also be searched for using
6170 the @option{-B} prefix, if needed. If it is not found there, the two
6171 standard prefixes above are tried, and that is all. The file is left
6172 out of the link if it is not found by those means.
6174 Another way to specify a prefix much like the @option{-B} prefix is to use
6175 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6178 As a special kludge, if the path provided by @option{-B} is
6179 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6180 9, then it will be replaced by @file{[dir/]include}. This is to help
6181 with boot-strapping the compiler.
6183 @item -specs=@var{file}
6185 Process @var{file} after the compiler reads in the standard @file{specs}
6186 file, in order to override the defaults that the @file{gcc} driver
6187 program uses when determining what switches to pass to @file{cc1},
6188 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6189 @option{-specs=@var{file}} can be specified on the command line, and they
6190 are processed in order, from left to right.
6194 This option has been deprecated. Please use @option{-iquote} instead for
6195 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6196 Any directories you specify with @option{-I} options before the @option{-I-}
6197 option are searched only for the case of @samp{#include "@var{file}"};
6198 they are not searched for @samp{#include <@var{file}>}.
6200 If additional directories are specified with @option{-I} options after
6201 the @option{-I-}, these directories are searched for all @samp{#include}
6202 directives. (Ordinarily @emph{all} @option{-I} directories are used
6205 In addition, the @option{-I-} option inhibits the use of the current
6206 directory (where the current input file came from) as the first search
6207 directory for @samp{#include "@var{file}"}. There is no way to
6208 override this effect of @option{-I-}. With @option{-I.} you can specify
6209 searching the directory which was current when the compiler was
6210 invoked. That is not exactly the same as what the preprocessor does
6211 by default, but it is often satisfactory.
6213 @option{-I-} does not inhibit the use of the standard system directories
6214 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6221 @section Specifying subprocesses and the switches to pass to them
6224 @command{gcc} is a driver program. It performs its job by invoking a
6225 sequence of other programs to do the work of compiling, assembling and
6226 linking. GCC interprets its command-line parameters and uses these to
6227 deduce which programs it should invoke, and which command-line options
6228 it ought to place on their command lines. This behavior is controlled
6229 by @dfn{spec strings}. In most cases there is one spec string for each
6230 program that GCC can invoke, but a few programs have multiple spec
6231 strings to control their behavior. The spec strings built into GCC can
6232 be overridden by using the @option{-specs=} command-line switch to specify
6235 @dfn{Spec files} are plaintext files that are used to construct spec
6236 strings. They consist of a sequence of directives separated by blank
6237 lines. The type of directive is determined by the first non-whitespace
6238 character on the line and it can be one of the following:
6241 @item %@var{command}
6242 Issues a @var{command} to the spec file processor. The commands that can
6246 @item %include <@var{file}>
6248 Search for @var{file} and insert its text at the current point in the
6251 @item %include_noerr <@var{file}>
6252 @cindex %include_noerr
6253 Just like @samp{%include}, but do not generate an error message if the include
6254 file cannot be found.
6256 @item %rename @var{old_name} @var{new_name}
6258 Rename the spec string @var{old_name} to @var{new_name}.
6262 @item *[@var{spec_name}]:
6263 This tells the compiler to create, override or delete the named spec
6264 string. All lines after this directive up to the next directive or
6265 blank line are considered to be the text for the spec string. If this
6266 results in an empty string then the spec will be deleted. (Or, if the
6267 spec did not exist, then nothing will happened.) Otherwise, if the spec
6268 does not currently exist a new spec will be created. If the spec does
6269 exist then its contents will be overridden by the text of this
6270 directive, unless the first character of that text is the @samp{+}
6271 character, in which case the text will be appended to the spec.
6273 @item [@var{suffix}]:
6274 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6275 and up to the next directive or blank line are considered to make up the
6276 spec string for the indicated suffix. When the compiler encounters an
6277 input file with the named suffix, it will processes the spec string in
6278 order to work out how to compile that file. For example:
6285 This says that any input file whose name ends in @samp{.ZZ} should be
6286 passed to the program @samp{z-compile}, which should be invoked with the
6287 command-line switch @option{-input} and with the result of performing the
6288 @samp{%i} substitution. (See below.)
6290 As an alternative to providing a spec string, the text that follows a
6291 suffix directive can be one of the following:
6294 @item @@@var{language}
6295 This says that the suffix is an alias for a known @var{language}. This is
6296 similar to using the @option{-x} command-line switch to GCC to specify a
6297 language explicitly. For example:
6304 Says that .ZZ files are, in fact, C++ source files.
6307 This causes an error messages saying:
6310 @var{name} compiler not installed on this system.
6314 GCC already has an extensive list of suffixes built into it.
6315 This directive will add an entry to the end of the list of suffixes, but
6316 since the list is searched from the end backwards, it is effectively
6317 possible to override earlier entries using this technique.
6321 GCC has the following spec strings built into it. Spec files can
6322 override these strings or create their own. Note that individual
6323 targets can also add their own spec strings to this list.
6326 asm Options to pass to the assembler
6327 asm_final Options to pass to the assembler post-processor
6328 cpp Options to pass to the C preprocessor
6329 cc1 Options to pass to the C compiler
6330 cc1plus Options to pass to the C++ compiler
6331 endfile Object files to include at the end of the link
6332 link Options to pass to the linker
6333 lib Libraries to include on the command line to the linker
6334 libgcc Decides which GCC support library to pass to the linker
6335 linker Sets the name of the linker
6336 predefines Defines to be passed to the C preprocessor
6337 signed_char Defines to pass to CPP to say whether @code{char} is signed
6339 startfile Object files to include at the start of the link
6342 Here is a small example of a spec file:
6348 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6351 This example renames the spec called @samp{lib} to @samp{old_lib} and
6352 then overrides the previous definition of @samp{lib} with a new one.
6353 The new definition adds in some extra command-line options before
6354 including the text of the old definition.
6356 @dfn{Spec strings} are a list of command-line options to be passed to their
6357 corresponding program. In addition, the spec strings can contain
6358 @samp{%}-prefixed sequences to substitute variable text or to
6359 conditionally insert text into the command line. Using these constructs
6360 it is possible to generate quite complex command lines.
6362 Here is a table of all defined @samp{%}-sequences for spec
6363 strings. Note that spaces are not generated automatically around the
6364 results of expanding these sequences. Therefore you can concatenate them
6365 together or combine them with constant text in a single argument.
6369 Substitute one @samp{%} into the program name or argument.
6372 Substitute the name of the input file being processed.
6375 Substitute the basename of the input file being processed.
6376 This is the substring up to (and not including) the last period
6377 and not including the directory.
6380 This is the same as @samp{%b}, but include the file suffix (text after
6384 Marks the argument containing or following the @samp{%d} as a
6385 temporary file name, so that that file will be deleted if GCC exits
6386 successfully. Unlike @samp{%g}, this contributes no text to the
6389 @item %g@var{suffix}
6390 Substitute a file name that has suffix @var{suffix} and is chosen
6391 once per compilation, and mark the argument in the same way as
6392 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6393 name is now chosen in a way that is hard to predict even when previously
6394 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6395 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6396 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6397 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6398 was simply substituted with a file name chosen once per compilation,
6399 without regard to any appended suffix (which was therefore treated
6400 just like ordinary text), making such attacks more likely to succeed.
6402 @item %u@var{suffix}
6403 Like @samp{%g}, but generates a new temporary file name even if
6404 @samp{%u@var{suffix}} was already seen.
6406 @item %U@var{suffix}
6407 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6408 new one if there is no such last file name. In the absence of any
6409 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6410 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6411 would involve the generation of two distinct file names, one
6412 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6413 simply substituted with a file name chosen for the previous @samp{%u},
6414 without regard to any appended suffix.
6416 @item %j@var{suffix}
6417 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6418 writable, and if save-temps is off; otherwise, substitute the name
6419 of a temporary file, just like @samp{%u}. This temporary file is not
6420 meant for communication between processes, but rather as a junk
6423 @item %|@var{suffix}
6424 @itemx %m@var{suffix}
6425 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6426 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6427 all. These are the two most common ways to instruct a program that it
6428 should read from standard input or write to standard output. If you
6429 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6430 construct: see for example @file{f/lang-specs.h}.
6432 @item %.@var{SUFFIX}
6433 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6434 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6435 terminated by the next space or %.
6438 Marks the argument containing or following the @samp{%w} as the
6439 designated output file of this compilation. This puts the argument
6440 into the sequence of arguments that @samp{%o} will substitute later.
6443 Substitutes the names of all the output files, with spaces
6444 automatically placed around them. You should write spaces
6445 around the @samp{%o} as well or the results are undefined.
6446 @samp{%o} is for use in the specs for running the linker.
6447 Input files whose names have no recognized suffix are not compiled
6448 at all, but they are included among the output files, so they will
6452 Substitutes the suffix for object files. Note that this is
6453 handled specially when it immediately follows @samp{%g, %u, or %U},
6454 because of the need for those to form complete file names. The
6455 handling is such that @samp{%O} is treated exactly as if it had already
6456 been substituted, except that @samp{%g, %u, and %U} do not currently
6457 support additional @var{suffix} characters following @samp{%O} as they would
6458 following, for example, @samp{.o}.
6461 Substitutes the standard macro predefinitions for the
6462 current target machine. Use this when running @code{cpp}.
6465 Like @samp{%p}, but puts @samp{__} before and after the name of each
6466 predefined macro, except for macros that start with @samp{__} or with
6467 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6471 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6472 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
6473 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6477 Current argument is the name of a library or startup file of some sort.
6478 Search for that file in a standard list of directories and substitute
6479 the full name found.
6482 Print @var{str} as an error message. @var{str} is terminated by a newline.
6483 Use this when inconsistent options are detected.
6486 Substitute the contents of spec string @var{name} at this point.
6489 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6491 @item %x@{@var{option}@}
6492 Accumulate an option for @samp{%X}.
6495 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6499 Output the accumulated assembler options specified by @option{-Wa}.
6502 Output the accumulated preprocessor options specified by @option{-Wp}.
6505 Process the @code{asm} spec. This is used to compute the
6506 switches to be passed to the assembler.
6509 Process the @code{asm_final} spec. This is a spec string for
6510 passing switches to an assembler post-processor, if such a program is
6514 Process the @code{link} spec. This is the spec for computing the
6515 command line passed to the linker. Typically it will make use of the
6516 @samp{%L %G %S %D and %E} sequences.
6519 Dump out a @option{-L} option for each directory that GCC believes might
6520 contain startup files. If the target supports multilibs then the
6521 current multilib directory will be prepended to each of these paths.
6524 Process the @code{lib} spec. This is a spec string for deciding which
6525 libraries should be included on the command line to the linker.
6528 Process the @code{libgcc} spec. This is a spec string for deciding
6529 which GCC support library should be included on the command line to the linker.
6532 Process the @code{startfile} spec. This is a spec for deciding which
6533 object files should be the first ones passed to the linker. Typically
6534 this might be a file named @file{crt0.o}.
6537 Process the @code{endfile} spec. This is a spec string that specifies
6538 the last object files that will be passed to the linker.
6541 Process the @code{cpp} spec. This is used to construct the arguments
6542 to be passed to the C preprocessor.
6545 Process the @code{cc1} spec. This is used to construct the options to be
6546 passed to the actual C compiler (@samp{cc1}).
6549 Process the @code{cc1plus} spec. This is used to construct the options to be
6550 passed to the actual C++ compiler (@samp{cc1plus}).
6553 Substitute the variable part of a matched option. See below.
6554 Note that each comma in the substituted string is replaced by
6558 Remove all occurrences of @code{-S} from the command line. Note---this
6559 command is position dependent. @samp{%} commands in the spec string
6560 before this one will see @code{-S}, @samp{%} commands in the spec string
6561 after this one will not.
6563 @item %:@var{function}(@var{args})
6564 Call the named function @var{function}, passing it @var{args}.
6565 @var{args} is first processed as a nested spec string, then split
6566 into an argument vector in the usual fashion. The function returns
6567 a string which is processed as if it had appeared literally as part
6568 of the current spec.
6570 The following built-in spec functions are provided:
6573 @item @code{if-exists}
6574 The @code{if-exists} spec function takes one argument, an absolute
6575 pathname to a file. If the file exists, @code{if-exists} returns the
6576 pathname. Here is a small example of its usage:
6580 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6583 @item @code{if-exists-else}
6584 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6585 spec function, except that it takes two arguments. The first argument is
6586 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6587 returns the pathname. If it does not exist, it returns the second argument.
6588 This way, @code{if-exists-else} can be used to select one file or another,
6589 based on the existence of the first. Here is a small example of its usage:
6593 crt0%O%s %:if-exists(crti%O%s) \
6594 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6597 @item @code{replace-outfile}
6598 The @code{replace-outfile} spec function takes two arguments. It looks for the
6599 first argument in the outfiles array and replaces it with the second argument. Here
6600 is a small example of its usage:
6603 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
6609 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6610 If that switch was not specified, this substitutes nothing. Note that
6611 the leading dash is omitted when specifying this option, and it is
6612 automatically inserted if the substitution is performed. Thus the spec
6613 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6614 and would output the command line option @option{-foo}.
6616 @item %W@{@code{S}@}
6617 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6620 @item %@{@code{S}*@}
6621 Substitutes all the switches specified to GCC whose names start
6622 with @code{-S}, but which also take an argument. This is used for
6623 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6624 GCC considers @option{-o foo} as being
6625 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6626 text, including the space. Thus two arguments would be generated.
6628 @item %@{@code{S}*&@code{T}*@}
6629 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6630 (the order of @code{S} and @code{T} in the spec is not significant).
6631 There can be any number of ampersand-separated variables; for each the
6632 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6634 @item %@{@code{S}:@code{X}@}
6635 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6637 @item %@{!@code{S}:@code{X}@}
6638 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6640 @item %@{@code{S}*:@code{X}@}
6641 Substitutes @code{X} if one or more switches whose names start with
6642 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6643 once, no matter how many such switches appeared. However, if @code{%*}
6644 appears somewhere in @code{X}, then @code{X} will be substituted once
6645 for each matching switch, with the @code{%*} replaced by the part of
6646 that switch that matched the @code{*}.
6648 @item %@{.@code{S}:@code{X}@}
6649 Substitutes @code{X}, if processing a file with suffix @code{S}.
6651 @item %@{!.@code{S}:@code{X}@}
6652 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6654 @item %@{@code{S}|@code{P}:@code{X}@}
6655 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6656 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
6657 although they have a stronger binding than the @samp{|}. If @code{%*}
6658 appears in @code{X}, all of the alternatives must be starred, and only
6659 the first matching alternative is substituted.
6661 For example, a spec string like this:
6664 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
6667 will output the following command-line options from the following input
6668 command-line options:
6673 -d fred.c -foo -baz -boggle
6674 -d jim.d -bar -baz -boggle
6677 @item %@{S:X; T:Y; :D@}
6679 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
6680 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
6681 be as many clauses as you need. This may be combined with @code{.},
6682 @code{!}, @code{|}, and @code{*} as needed.
6687 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
6688 construct may contain other nested @samp{%} constructs or spaces, or
6689 even newlines. They are processed as usual, as described above.
6690 Trailing white space in @code{X} is ignored. White space may also
6691 appear anywhere on the left side of the colon in these constructs,
6692 except between @code{.} or @code{*} and the corresponding word.
6694 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
6695 handled specifically in these constructs. If another value of
6696 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
6697 @option{-W} switch is found later in the command line, the earlier
6698 switch value is ignored, except with @{@code{S}*@} where @code{S} is
6699 just one letter, which passes all matching options.
6701 The character @samp{|} at the beginning of the predicate text is used to
6702 indicate that a command should be piped to the following command, but
6703 only if @option{-pipe} is specified.
6705 It is built into GCC which switches take arguments and which do not.
6706 (You might think it would be useful to generalize this to allow each
6707 compiler's spec to say which switches take arguments. But this cannot
6708 be done in a consistent fashion. GCC cannot even decide which input
6709 files have been specified without knowing which switches take arguments,
6710 and it must know which input files to compile in order to tell which
6713 GCC also knows implicitly that arguments starting in @option{-l} are to be
6714 treated as compiler output files, and passed to the linker in their
6715 proper position among the other output files.
6717 @c man begin OPTIONS
6719 @node Target Options
6720 @section Specifying Target Machine and Compiler Version
6721 @cindex target options
6722 @cindex cross compiling
6723 @cindex specifying machine version
6724 @cindex specifying compiler version and target machine
6725 @cindex compiler version, specifying
6726 @cindex target machine, specifying
6728 The usual way to run GCC is to run the executable called @file{gcc}, or
6729 @file{<machine>-gcc} when cross-compiling, or
6730 @file{<machine>-gcc-<version>} to run a version other than the one that
6731 was installed last. Sometimes this is inconvenient, so GCC provides
6732 options that will switch to another cross-compiler or version.
6735 @item -b @var{machine}
6737 The argument @var{machine} specifies the target machine for compilation.
6739 The value to use for @var{machine} is the same as was specified as the
6740 machine type when configuring GCC as a cross-compiler. For
6741 example, if a cross-compiler was configured with @samp{configure
6742 i386v}, meaning to compile for an 80386 running System V, then you
6743 would specify @option{-b i386v} to run that cross compiler.
6745 @item -V @var{version}
6747 The argument @var{version} specifies which version of GCC to run.
6748 This is useful when multiple versions are installed. For example,
6749 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
6752 The @option{-V} and @option{-b} options work by running the
6753 @file{<machine>-gcc-<version>} executable, so there's no real reason to
6754 use them if you can just run that directly.
6756 @node Submodel Options
6757 @section Hardware Models and Configurations
6758 @cindex submodel options
6759 @cindex specifying hardware config
6760 @cindex hardware models and configurations, specifying
6761 @cindex machine dependent options
6763 Earlier we discussed the standard option @option{-b} which chooses among
6764 different installed compilers for completely different target
6765 machines, such as VAX vs.@: 68000 vs.@: 80386.
6767 In addition, each of these target machine types can have its own
6768 special options, starting with @samp{-m}, to choose among various
6769 hardware models or configurations---for example, 68010 vs 68020,
6770 floating coprocessor or none. A single installed version of the
6771 compiler can compile for any model or configuration, according to the
6774 Some configurations of the compiler also support additional special
6775 options, usually for compatibility with other compilers on the same
6778 These options are defined by the macro @code{TARGET_SWITCHES} in the
6779 machine description. The default for the options is also defined by
6780 that macro, which enables you to change the defaults.
6782 @c This list is ordered alphanumerically by subsection name.
6783 @c It should be the same order and spelling as these options are listed
6784 @c in Machine Dependent Options
6790 * Blackfin Options::
6793 * DEC Alpha Options::
6794 * DEC Alpha/VMS Options::
6798 * i386 and x86-64 Options::
6810 * RS/6000 and PowerPC Options::
6811 * S/390 and zSeries Options::
6814 * System V Options::
6815 * TMS320C3x/C4x Options::
6819 * Xstormy16 Options::
6825 @subsection ARC Options
6828 These options are defined for ARC implementations:
6833 Compile code for little endian mode. This is the default.
6837 Compile code for big endian mode.
6840 @opindex mmangle-cpu
6841 Prepend the name of the cpu to all public symbol names.
6842 In multiple-processor systems, there are many ARC variants with different
6843 instruction and register set characteristics. This flag prevents code
6844 compiled for one cpu to be linked with code compiled for another.
6845 No facility exists for handling variants that are ``almost identical''.
6846 This is an all or nothing option.
6848 @item -mcpu=@var{cpu}
6850 Compile code for ARC variant @var{cpu}.
6851 Which variants are supported depend on the configuration.
6852 All variants support @option{-mcpu=base}, this is the default.
6854 @item -mtext=@var{text-section}
6855 @itemx -mdata=@var{data-section}
6856 @itemx -mrodata=@var{readonly-data-section}
6860 Put functions, data, and readonly data in @var{text-section},
6861 @var{data-section}, and @var{readonly-data-section} respectively
6862 by default. This can be overridden with the @code{section} attribute.
6863 @xref{Variable Attributes}.
6868 @subsection ARM Options
6871 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6875 @item -mabi=@var{name}
6877 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
6878 @samp{atpcs}, @samp{aapcs} and @samp{iwmmxt}.
6881 @opindex mapcs-frame
6882 Generate a stack frame that is compliant with the ARM Procedure Call
6883 Standard for all functions, even if this is not strictly necessary for
6884 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6885 with this option will cause the stack frames not to be generated for
6886 leaf functions. The default is @option{-mno-apcs-frame}.
6890 This is a synonym for @option{-mapcs-frame}.
6893 @c not currently implemented
6894 @item -mapcs-stack-check
6895 @opindex mapcs-stack-check
6896 Generate code to check the amount of stack space available upon entry to
6897 every function (that actually uses some stack space). If there is
6898 insufficient space available then either the function
6899 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
6900 called, depending upon the amount of stack space required. The run time
6901 system is required to provide these functions. The default is
6902 @option{-mno-apcs-stack-check}, since this produces smaller code.
6904 @c not currently implemented
6906 @opindex mapcs-float
6907 Pass floating point arguments using the float point registers. This is
6908 one of the variants of the APCS@. This option is recommended if the
6909 target hardware has a floating point unit or if a lot of floating point
6910 arithmetic is going to be performed by the code. The default is
6911 @option{-mno-apcs-float}, since integer only code is slightly increased in
6912 size if @option{-mapcs-float} is used.
6914 @c not currently implemented
6915 @item -mapcs-reentrant
6916 @opindex mapcs-reentrant
6917 Generate reentrant, position independent code. The default is
6918 @option{-mno-apcs-reentrant}.
6921 @item -mthumb-interwork
6922 @opindex mthumb-interwork
6923 Generate code which supports calling between the ARM and Thumb
6924 instruction sets. Without this option the two instruction sets cannot
6925 be reliably used inside one program. The default is
6926 @option{-mno-thumb-interwork}, since slightly larger code is generated
6927 when @option{-mthumb-interwork} is specified.
6929 @item -mno-sched-prolog
6930 @opindex mno-sched-prolog
6931 Prevent the reordering of instructions in the function prolog, or the
6932 merging of those instruction with the instructions in the function's
6933 body. This means that all functions will start with a recognizable set
6934 of instructions (or in fact one of a choice from a small set of
6935 different function prologues), and this information can be used to
6936 locate the start if functions inside an executable piece of code. The
6937 default is @option{-msched-prolog}.
6940 @opindex mhard-float
6941 Generate output containing floating point instructions. This is the
6945 @opindex msoft-float
6946 Generate output containing library calls for floating point.
6947 @strong{Warning:} the requisite libraries are not available for all ARM
6948 targets. Normally the facilities of the machine's usual C compiler are
6949 used, but this cannot be done directly in cross-compilation. You must make
6950 your own arrangements to provide suitable library functions for
6953 @option{-msoft-float} changes the calling convention in the output file;
6954 therefore, it is only useful if you compile @emph{all} of a program with
6955 this option. In particular, you need to compile @file{libgcc.a}, the
6956 library that comes with GCC, with @option{-msoft-float} in order for
6959 @item -mfloat-abi=@var{name}
6961 Specifies which ABI to use for floating point values. Permissible values
6962 are: @samp{soft}, @samp{softfp} and @samp{hard}.
6964 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
6965 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
6966 of floating point instructions, but still uses the soft-float calling
6969 @item -mlittle-endian
6970 @opindex mlittle-endian
6971 Generate code for a processor running in little-endian mode. This is
6972 the default for all standard configurations.
6975 @opindex mbig-endian
6976 Generate code for a processor running in big-endian mode; the default is
6977 to compile code for a little-endian processor.
6979 @item -mwords-little-endian
6980 @opindex mwords-little-endian
6981 This option only applies when generating code for big-endian processors.
6982 Generate code for a little-endian word order but a big-endian byte
6983 order. That is, a byte order of the form @samp{32107654}. Note: this
6984 option should only be used if you require compatibility with code for
6985 big-endian ARM processors generated by versions of the compiler prior to
6988 @item -mcpu=@var{name}
6990 This specifies the name of the target ARM processor. GCC uses this name
6991 to determine what kind of instructions it can emit when generating
6992 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
6993 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
6994 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
6995 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
6996 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
6997 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
6998 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
6999 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7000 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7001 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7002 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7003 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7004 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7005 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
7008 @itemx -mtune=@var{name}
7010 This option is very similar to the @option{-mcpu=} option, except that
7011 instead of specifying the actual target processor type, and hence
7012 restricting which instructions can be used, it specifies that GCC should
7013 tune the performance of the code as if the target were of the type
7014 specified in this option, but still choosing the instructions that it
7015 will generate based on the cpu specified by a @option{-mcpu=} option.
7016 For some ARM implementations better performance can be obtained by using
7019 @item -march=@var{name}
7021 This specifies the name of the target ARM architecture. GCC uses this
7022 name to determine what kind of instructions it can emit when generating
7023 assembly code. This option can be used in conjunction with or instead
7024 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7025 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7026 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7027 @samp{iwmmxt}, @samp{ep9312}.
7029 @item -mfpu=@var{name}
7030 @itemx -mfpe=@var{number}
7031 @itemx -mfp=@var{number}
7035 This specifies what floating point hardware (or hardware emulation) is
7036 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7037 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7038 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7039 with older versions of GCC@.
7041 If @option{-msoft-float} is specified this specifies the format of
7042 floating point values.
7044 @item -mstructure-size-boundary=@var{n}
7045 @opindex mstructure-size-boundary
7046 The size of all structures and unions will be rounded up to a multiple
7047 of the number of bits set by this option. Permissible values are 8, 32
7048 and 64. The default value varies for different toolchains. For the COFF
7049 targeted toolchain the default value is 8. A value of 64 is only allowed
7050 if the underlying ABI supports it.
7052 Specifying the larger number can produce faster, more efficient code, but
7053 can also increase the size of the program. Different values are potentially
7054 incompatible. Code compiled with one value cannot necessarily expect to
7055 work with code or libraries compiled with another value, if they exchange
7056 information using structures or unions.
7058 @item -mabort-on-noreturn
7059 @opindex mabort-on-noreturn
7060 Generate a call to the function @code{abort} at the end of a
7061 @code{noreturn} function. It will be executed if the function tries to
7065 @itemx -mno-long-calls
7066 @opindex mlong-calls
7067 @opindex mno-long-calls
7068 Tells the compiler to perform function calls by first loading the
7069 address of the function into a register and then performing a subroutine
7070 call on this register. This switch is needed if the target function
7071 will lie outside of the 64 megabyte addressing range of the offset based
7072 version of subroutine call instruction.
7074 Even if this switch is enabled, not all function calls will be turned
7075 into long calls. The heuristic is that static functions, functions
7076 which have the @samp{short-call} attribute, functions that are inside
7077 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7078 definitions have already been compiled within the current compilation
7079 unit, will not be turned into long calls. The exception to this rule is
7080 that weak function definitions, functions with the @samp{long-call}
7081 attribute or the @samp{section} attribute, and functions that are within
7082 the scope of a @samp{#pragma long_calls} directive, will always be
7083 turned into long calls.
7085 This feature is not enabled by default. Specifying
7086 @option{-mno-long-calls} will restore the default behavior, as will
7087 placing the function calls within the scope of a @samp{#pragma
7088 long_calls_off} directive. Note these switches have no effect on how
7089 the compiler generates code to handle function calls via function
7092 @item -mnop-fun-dllimport
7093 @opindex mnop-fun-dllimport
7094 Disable support for the @code{dllimport} attribute.
7096 @item -msingle-pic-base
7097 @opindex msingle-pic-base
7098 Treat the register used for PIC addressing as read-only, rather than
7099 loading it in the prologue for each function. The run-time system is
7100 responsible for initializing this register with an appropriate value
7101 before execution begins.
7103 @item -mpic-register=@var{reg}
7104 @opindex mpic-register
7105 Specify the register to be used for PIC addressing. The default is R10
7106 unless stack-checking is enabled, when R9 is used.
7108 @item -mcirrus-fix-invalid-insns
7109 @opindex mcirrus-fix-invalid-insns
7110 @opindex mno-cirrus-fix-invalid-insns
7111 Insert NOPs into the instruction stream to in order to work around
7112 problems with invalid Maverick instruction combinations. This option
7113 is only valid if the @option{-mcpu=ep9312} option has been used to
7114 enable generation of instructions for the Cirrus Maverick floating
7115 point co-processor. This option is not enabled by default, since the
7116 problem is only present in older Maverick implementations. The default
7117 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7120 @item -mpoke-function-name
7121 @opindex mpoke-function-name
7122 Write the name of each function into the text section, directly
7123 preceding the function prologue. The generated code is similar to this:
7127 .ascii "arm_poke_function_name", 0
7130 .word 0xff000000 + (t1 - t0)
7131 arm_poke_function_name
7133 stmfd sp!, @{fp, ip, lr, pc@}
7137 When performing a stack backtrace, code can inspect the value of
7138 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7139 location @code{pc - 12} and the top 8 bits are set, then we know that
7140 there is a function name embedded immediately preceding this location
7141 and has length @code{((pc[-3]) & 0xff000000)}.
7145 Generate code for the 16-bit Thumb instruction set. The default is to
7146 use the 32-bit ARM instruction set.
7149 @opindex mtpcs-frame
7150 Generate a stack frame that is compliant with the Thumb Procedure Call
7151 Standard for all non-leaf functions. (A leaf function is one that does
7152 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7154 @item -mtpcs-leaf-frame
7155 @opindex mtpcs-leaf-frame
7156 Generate a stack frame that is compliant with the Thumb Procedure Call
7157 Standard for all leaf functions. (A leaf function is one that does
7158 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7160 @item -mcallee-super-interworking
7161 @opindex mcallee-super-interworking
7162 Gives all externally visible functions in the file being compiled an ARM
7163 instruction set header which switches to Thumb mode before executing the
7164 rest of the function. This allows these functions to be called from
7165 non-interworking code.
7167 @item -mcaller-super-interworking
7168 @opindex mcaller-super-interworking
7169 Allows calls via function pointers (including virtual functions) to
7170 execute correctly regardless of whether the target code has been
7171 compiled for interworking or not. There is a small overhead in the cost
7172 of executing a function pointer if this option is enabled.
7177 @subsection AVR Options
7180 These options are defined for AVR implementations:
7183 @item -mmcu=@var{mcu}
7185 Specify ATMEL AVR instruction set or MCU type.
7187 Instruction set avr1 is for the minimal AVR core, not supported by the C
7188 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7189 attiny11, attiny12, attiny15, attiny28).
7191 Instruction set avr2 (default) is for the classic AVR core with up to
7192 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7193 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7194 at90c8534, at90s8535).
7196 Instruction set avr3 is for the classic AVR core with up to 128K program
7197 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7199 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7200 memory space (MCU types: atmega8, atmega83, atmega85).
7202 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7203 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7204 atmega64, atmega128, at43usb355, at94k).
7208 Output instruction sizes to the asm file.
7210 @item -minit-stack=@var{N}
7211 @opindex minit-stack
7212 Specify the initial stack address, which may be a symbol or numeric value,
7213 @samp{__stack} is the default.
7215 @item -mno-interrupts
7216 @opindex mno-interrupts
7217 Generated code is not compatible with hardware interrupts.
7218 Code size will be smaller.
7220 @item -mcall-prologues
7221 @opindex mcall-prologues
7222 Functions prologues/epilogues expanded as call to appropriate
7223 subroutines. Code size will be smaller.
7225 @item -mno-tablejump
7226 @opindex mno-tablejump
7227 Do not generate tablejump insns which sometimes increase code size.
7230 @opindex mtiny-stack
7231 Change only the low 8 bits of the stack pointer.
7235 Assume int to be 8 bit integer. This affects the sizes of all types: A
7236 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7237 and long long will be 4 bytes. Please note that this option does not
7238 comply to the C standards, but it will provide you with smaller code
7242 @node Blackfin Options
7243 @subsection Blackfin Options
7244 @cindex Blackfin Options
7247 @item -momit-leaf-frame-pointer
7248 @opindex momit-leaf-frame-pointer
7249 Don't keep the frame pointer in a register for leaf functions. This
7250 avoids the instructions to save, set up and restore frame pointers and
7251 makes an extra register available in leaf functions. The option
7252 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7253 which might make debugging harder.
7257 When enabled, the compiler will ensure that the generated code does not
7258 contain speculative loads after jump instructions. This option is enabled
7263 Don't generate extra code to prevent speculative loads from occurring.
7267 When enabled, the compiler is free to take advantage of the knowledge that
7268 the entire program fits into the low 64k of memory.
7271 @opindex mno-low-64k
7272 Assume that the program is arbitrarily large. This is the default.
7274 @item -mid-shared-library
7275 @opindex mid-shared-library
7276 Generate code that supports shared libraries via the library ID method.
7277 This allows for execute in place and shared libraries in an environment
7278 without virtual memory management. This option implies @option{-fPIC}.
7280 @item -mno-id-shared-library
7281 @opindex mno-id-shared-library
7282 Generate code that doesn't assume ID based shared libraries are being used.
7283 This is the default.
7285 @item -mshared-library-id=n
7286 @opindex mshared-library-id
7287 Specified the identification number of the ID based shared library being
7288 compiled. Specifying a value of 0 will generate more compact code, specifying
7289 other values will force the allocation of that number to the current
7290 library but is no more space or time efficient than omitting this option.
7294 @subsection CRIS Options
7295 @cindex CRIS Options
7297 These options are defined specifically for the CRIS ports.
7300 @item -march=@var{architecture-type}
7301 @itemx -mcpu=@var{architecture-type}
7304 Generate code for the specified architecture. The choices for
7305 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7306 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7307 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7310 @item -mtune=@var{architecture-type}
7312 Tune to @var{architecture-type} everything applicable about the generated
7313 code, except for the ABI and the set of available instructions. The
7314 choices for @var{architecture-type} are the same as for
7315 @option{-march=@var{architecture-type}}.
7317 @item -mmax-stack-frame=@var{n}
7318 @opindex mmax-stack-frame
7319 Warn when the stack frame of a function exceeds @var{n} bytes.
7321 @item -melinux-stacksize=@var{n}
7322 @opindex melinux-stacksize
7323 Only available with the @samp{cris-axis-aout} target. Arranges for
7324 indications in the program to the kernel loader that the stack of the
7325 program should be set to @var{n} bytes.
7331 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7332 @option{-march=v3} and @option{-march=v8} respectively.
7334 @item -mmul-bug-workaround
7335 @itemx -mno-mul-bug-workaround
7336 @opindex mmul-bug-workaround
7337 @opindex mno-mul-bug-workaround
7338 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7339 models where it applies. This option is active by default.
7343 Enable CRIS-specific verbose debug-related information in the assembly
7344 code. This option also has the effect to turn off the @samp{#NO_APP}
7345 formatted-code indicator to the assembler at the beginning of the
7350 Do not use condition-code results from previous instruction; always emit
7351 compare and test instructions before use of condition codes.
7353 @item -mno-side-effects
7354 @opindex mno-side-effects
7355 Do not emit instructions with side-effects in addressing modes other than
7359 @itemx -mno-stack-align
7361 @itemx -mno-data-align
7362 @itemx -mconst-align
7363 @itemx -mno-const-align
7364 @opindex mstack-align
7365 @opindex mno-stack-align
7366 @opindex mdata-align
7367 @opindex mno-data-align
7368 @opindex mconst-align
7369 @opindex mno-const-align
7370 These options (no-options) arranges (eliminate arrangements) for the
7371 stack-frame, individual data and constants to be aligned for the maximum
7372 single data access size for the chosen CPU model. The default is to
7373 arrange for 32-bit alignment. ABI details such as structure layout are
7374 not affected by these options.
7382 Similar to the stack- data- and const-align options above, these options
7383 arrange for stack-frame, writable data and constants to all be 32-bit,
7384 16-bit or 8-bit aligned. The default is 32-bit alignment.
7386 @item -mno-prologue-epilogue
7387 @itemx -mprologue-epilogue
7388 @opindex mno-prologue-epilogue
7389 @opindex mprologue-epilogue
7390 With @option{-mno-prologue-epilogue}, the normal function prologue and
7391 epilogue that sets up the stack-frame are omitted and no return
7392 instructions or return sequences are generated in the code. Use this
7393 option only together with visual inspection of the compiled code: no
7394 warnings or errors are generated when call-saved registers must be saved,
7395 or storage for local variable needs to be allocated.
7401 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7402 instruction sequences that load addresses for functions from the PLT part
7403 of the GOT rather than (traditional on other architectures) calls to the
7404 PLT@. The default is @option{-mgotplt}.
7408 Legacy no-op option only recognized with the cris-axis-aout target.
7412 Legacy no-op option only recognized with the cris-axis-elf and
7413 cris-axis-linux-gnu targets.
7417 Only recognized with the cris-axis-aout target, where it selects a
7418 GNU/linux-like multilib, include files and instruction set for
7423 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7427 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7428 to link with input-output functions from a simulator library. Code,
7429 initialized data and zero-initialized data are allocated consecutively.
7433 Like @option{-sim}, but pass linker options to locate initialized data at
7434 0x40000000 and zero-initialized data at 0x80000000.
7437 @node Darwin Options
7438 @subsection Darwin Options
7439 @cindex Darwin options
7441 These options are defined for all architectures running the Darwin operating
7444 FSF GCC on Darwin does not create ``fat'' object files; it will create
7445 an object file for the single architecture that it was built to
7446 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7447 @option{-arch} options are used; it does so by running the compiler or
7448 linker multiple times and joining the results together with
7451 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7452 @samp{i686}) is determined by the flags that specify the ISA
7453 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7454 @option{-force_cpusubtype_ALL} option can be used to override this.
7456 The Darwin tools vary in their behavior when presented with an ISA
7457 mismatch. The assembler, @file{as}, will only permit instructions to
7458 be used that are valid for the subtype of the file it is generating,
7459 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7460 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
7461 and print an error if asked to create a shared library with a less
7462 restrictive subtype than its input files (for instance, trying to put
7463 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
7464 for executables, @file{ld}, will quietly give the executable the most
7465 restrictive subtype of any of its input files.
7470 Add the framework directory @var{dir} to the head of the list of
7471 directories to be searched for header files. These directories are
7472 interleaved with those specified by @option{-I} options and are
7473 scanned in a left-to-right order.
7475 A framework directory is a directory with frameworks in it. A
7476 framework is a directory with a @samp{"Headers"} and/or
7477 @samp{"PrivateHeaders"} directory contained directly in it that ends
7478 in @samp{".framework"}. The name of a framework is the name of this
7479 directory excluding the @samp{".framework"}. Headers associated with
7480 the framework are found in one of those two directories, with
7481 @samp{"Headers"} being searched first. A subframework is a framework
7482 directory that is in a framework's @samp{"Frameworks"} directory.
7483 Includes of subframework headers can only appear in a header of a
7484 framework that contains the subframework, or in a sibling subframework
7485 header. Two subframeworks are siblings if they occur in the same
7486 framework. A subframework should not have the same name as a
7487 framework, a warning will be issued if this is violated. Currently a
7488 subframework cannot have subframeworks, in the future, the mechanism
7489 may be extended to support this. The standard frameworks can be found
7490 in @samp{"/System/Library/Frameworks"} and
7491 @samp{"/Library/Frameworks"}. An example include looks like
7492 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
7493 the name of the framework and header.h is found in the
7494 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
7498 Emit debugging information for symbols that are used. For STABS
7499 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
7500 This is by default ON@.
7504 Emit debugging information for all symbols and types.
7506 @item -mone-byte-bool
7507 @opindex -mone-byte-bool
7508 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7509 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7510 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7511 option has no effect on x86.
7513 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
7514 to generate code that is not binary compatible with code generated
7515 without that switch. Using this switch may require recompiling all
7516 other modules in a program, including system libraries. Use this
7517 switch to conform to a non-default data model.
7519 @item -mfix-and-continue
7520 @itemx -ffix-and-continue
7521 @itemx -findirect-data
7522 @opindex mfix-and-continue
7523 @opindex ffix-and-continue
7524 @opindex findirect-data
7525 Generate code suitable for fast turn around development. Needed to
7526 enable gdb to dynamically load @code{.o} files into already running
7527 programs. @option{-findirect-data} and @option{-ffix-and-continue}
7528 are provided for backwards compatibility.
7532 Loads all members of static archive libraries.
7533 See man ld(1) for more information.
7535 @item -arch_errors_fatal
7536 @opindex arch_errors_fatal
7537 Cause the errors having to do with files that have the wrong architecture
7541 @opindex bind_at_load
7542 Causes the output file to be marked such that the dynamic linker will
7543 bind all undefined references when the file is loaded or launched.
7547 Produce a Mach-o bundle format file.
7548 See man ld(1) for more information.
7550 @item -bundle_loader @var{executable}
7551 @opindex bundle_loader
7552 This option specifies the @var{executable} that will be loading the build
7553 output file being linked. See man ld(1) for more information.
7556 @opindex -dynamiclib
7557 When passed this option, GCC will produce a dynamic library instead of
7558 an executable when linking, using the Darwin @file{libtool} command.
7560 @item -force_cpusubtype_ALL
7561 @opindex -force_cpusubtype_ALL
7562 This causes GCC's output file to have the @var{ALL} subtype, instead of
7563 one controlled by the @option{-mcpu} or @option{-march} option.
7565 @item -allowable_client @var{client_name}
7567 @itemx -compatibility_version
7568 @itemx -current_version
7570 @itemx -dependency-file
7572 @itemx -dylinker_install_name
7574 @itemx -exported_symbols_list
7576 @itemx -flat_namespace
7577 @itemx -force_flat_namespace
7578 @itemx -headerpad_max_install_names
7581 @itemx -install_name
7582 @itemx -keep_private_externs
7583 @itemx -multi_module
7584 @itemx -multiply_defined
7585 @itemx -multiply_defined_unused
7587 @itemx -no_dead_strip_inits_and_terms
7588 @itemx -nofixprebinding
7591 @itemx -noseglinkedit
7592 @itemx -pagezero_size
7594 @itemx -prebind_all_twolevel_modules
7595 @itemx -private_bundle
7596 @itemx -read_only_relocs
7598 @itemx -sectobjectsymbols
7602 @itemx -sectobjectsymbols
7605 @itemx -segs_read_only_addr
7606 @itemx -segs_read_write_addr
7607 @itemx -seg_addr_table
7608 @itemx -seg_addr_table_filename
7611 @itemx -segs_read_only_addr
7612 @itemx -segs_read_write_addr
7613 @itemx -single_module
7616 @itemx -sub_umbrella
7617 @itemx -twolevel_namespace
7620 @itemx -unexported_symbols_list
7621 @itemx -weak_reference_mismatches
7624 @opindex allowable_client
7625 @opindex client_name
7626 @opindex compatibility_version
7627 @opindex current_version
7629 @opindex dependency-file
7631 @opindex dylinker_install_name
7633 @opindex exported_symbols_list
7635 @opindex flat_namespace
7636 @opindex force_flat_namespace
7637 @opindex headerpad_max_install_names
7640 @opindex install_name
7641 @opindex keep_private_externs
7642 @opindex multi_module
7643 @opindex multiply_defined
7644 @opindex multiply_defined_unused
7646 @opindex no_dead_strip_inits_and_terms
7647 @opindex nofixprebinding
7648 @opindex nomultidefs
7650 @opindex noseglinkedit
7651 @opindex pagezero_size
7653 @opindex prebind_all_twolevel_modules
7654 @opindex private_bundle
7655 @opindex read_only_relocs
7657 @opindex sectobjectsymbols
7661 @opindex sectobjectsymbols
7664 @opindex segs_read_only_addr
7665 @opindex segs_read_write_addr
7666 @opindex seg_addr_table
7667 @opindex seg_addr_table_filename
7668 @opindex seglinkedit
7670 @opindex segs_read_only_addr
7671 @opindex segs_read_write_addr
7672 @opindex single_module
7674 @opindex sub_library
7675 @opindex sub_umbrella
7676 @opindex twolevel_namespace
7679 @opindex unexported_symbols_list
7680 @opindex weak_reference_mismatches
7681 @opindex whatsloaded
7683 These options are passed to the Darwin linker. The Darwin linker man page
7684 describes them in detail.
7687 @node DEC Alpha Options
7688 @subsection DEC Alpha Options
7690 These @samp{-m} options are defined for the DEC Alpha implementations:
7693 @item -mno-soft-float
7695 @opindex mno-soft-float
7696 @opindex msoft-float
7697 Use (do not use) the hardware floating-point instructions for
7698 floating-point operations. When @option{-msoft-float} is specified,
7699 functions in @file{libgcc.a} will be used to perform floating-point
7700 operations. Unless they are replaced by routines that emulate the
7701 floating-point operations, or compiled in such a way as to call such
7702 emulations routines, these routines will issue floating-point
7703 operations. If you are compiling for an Alpha without floating-point
7704 operations, you must ensure that the library is built so as not to call
7707 Note that Alpha implementations without floating-point operations are
7708 required to have floating-point registers.
7713 @opindex mno-fp-regs
7714 Generate code that uses (does not use) the floating-point register set.
7715 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
7716 register set is not used, floating point operands are passed in integer
7717 registers as if they were integers and floating-point results are passed
7718 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
7719 so any function with a floating-point argument or return value called by code
7720 compiled with @option{-mno-fp-regs} must also be compiled with that
7723 A typical use of this option is building a kernel that does not use,
7724 and hence need not save and restore, any floating-point registers.
7728 The Alpha architecture implements floating-point hardware optimized for
7729 maximum performance. It is mostly compliant with the IEEE floating
7730 point standard. However, for full compliance, software assistance is
7731 required. This option generates code fully IEEE compliant code
7732 @emph{except} that the @var{inexact-flag} is not maintained (see below).
7733 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
7734 defined during compilation. The resulting code is less efficient but is
7735 able to correctly support denormalized numbers and exceptional IEEE
7736 values such as not-a-number and plus/minus infinity. Other Alpha
7737 compilers call this option @option{-ieee_with_no_inexact}.
7739 @item -mieee-with-inexact
7740 @opindex mieee-with-inexact
7741 This is like @option{-mieee} except the generated code also maintains
7742 the IEEE @var{inexact-flag}. Turning on this option causes the
7743 generated code to implement fully-compliant IEEE math. In addition to
7744 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
7745 macro. On some Alpha implementations the resulting code may execute
7746 significantly slower than the code generated by default. Since there is
7747 very little code that depends on the @var{inexact-flag}, you should
7748 normally not specify this option. Other Alpha compilers call this
7749 option @option{-ieee_with_inexact}.
7751 @item -mfp-trap-mode=@var{trap-mode}
7752 @opindex mfp-trap-mode
7753 This option controls what floating-point related traps are enabled.
7754 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
7755 The trap mode can be set to one of four values:
7759 This is the default (normal) setting. The only traps that are enabled
7760 are the ones that cannot be disabled in software (e.g., division by zero
7764 In addition to the traps enabled by @samp{n}, underflow traps are enabled
7768 Like @samp{su}, but the instructions are marked to be safe for software
7769 completion (see Alpha architecture manual for details).
7772 Like @samp{su}, but inexact traps are enabled as well.
7775 @item -mfp-rounding-mode=@var{rounding-mode}
7776 @opindex mfp-rounding-mode
7777 Selects the IEEE rounding mode. Other Alpha compilers call this option
7778 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
7783 Normal IEEE rounding mode. Floating point numbers are rounded towards
7784 the nearest machine number or towards the even machine number in case
7788 Round towards minus infinity.
7791 Chopped rounding mode. Floating point numbers are rounded towards zero.
7794 Dynamic rounding mode. A field in the floating point control register
7795 (@var{fpcr}, see Alpha architecture reference manual) controls the
7796 rounding mode in effect. The C library initializes this register for
7797 rounding towards plus infinity. Thus, unless your program modifies the
7798 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
7801 @item -mtrap-precision=@var{trap-precision}
7802 @opindex mtrap-precision
7803 In the Alpha architecture, floating point traps are imprecise. This
7804 means without software assistance it is impossible to recover from a
7805 floating trap and program execution normally needs to be terminated.
7806 GCC can generate code that can assist operating system trap handlers
7807 in determining the exact location that caused a floating point trap.
7808 Depending on the requirements of an application, different levels of
7809 precisions can be selected:
7813 Program precision. This option is the default and means a trap handler
7814 can only identify which program caused a floating point exception.
7817 Function precision. The trap handler can determine the function that
7818 caused a floating point exception.
7821 Instruction precision. The trap handler can determine the exact
7822 instruction that caused a floating point exception.
7825 Other Alpha compilers provide the equivalent options called
7826 @option{-scope_safe} and @option{-resumption_safe}.
7828 @item -mieee-conformant
7829 @opindex mieee-conformant
7830 This option marks the generated code as IEEE conformant. You must not
7831 use this option unless you also specify @option{-mtrap-precision=i} and either
7832 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
7833 is to emit the line @samp{.eflag 48} in the function prologue of the
7834 generated assembly file. Under DEC Unix, this has the effect that
7835 IEEE-conformant math library routines will be linked in.
7837 @item -mbuild-constants
7838 @opindex mbuild-constants
7839 Normally GCC examines a 32- or 64-bit integer constant to
7840 see if it can construct it from smaller constants in two or three
7841 instructions. If it cannot, it will output the constant as a literal and
7842 generate code to load it from the data segment at runtime.
7844 Use this option to require GCC to construct @emph{all} integer constants
7845 using code, even if it takes more instructions (the maximum is six).
7847 You would typically use this option to build a shared library dynamic
7848 loader. Itself a shared library, it must relocate itself in memory
7849 before it can find the variables and constants in its own data segment.
7855 Select whether to generate code to be assembled by the vendor-supplied
7856 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
7874 Indicate whether GCC should generate code to use the optional BWX,
7875 CIX, FIX and MAX instruction sets. The default is to use the instruction
7876 sets supported by the CPU type specified via @option{-mcpu=} option or that
7877 of the CPU on which GCC was built if none was specified.
7882 @opindex mfloat-ieee
7883 Generate code that uses (does not use) VAX F and G floating point
7884 arithmetic instead of IEEE single and double precision.
7886 @item -mexplicit-relocs
7887 @itemx -mno-explicit-relocs
7888 @opindex mexplicit-relocs
7889 @opindex mno-explicit-relocs
7890 Older Alpha assemblers provided no way to generate symbol relocations
7891 except via assembler macros. Use of these macros does not allow
7892 optimal instruction scheduling. GNU binutils as of version 2.12
7893 supports a new syntax that allows the compiler to explicitly mark
7894 which relocations should apply to which instructions. This option
7895 is mostly useful for debugging, as GCC detects the capabilities of
7896 the assembler when it is built and sets the default accordingly.
7900 @opindex msmall-data
7901 @opindex mlarge-data
7902 When @option{-mexplicit-relocs} is in effect, static data is
7903 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
7904 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
7905 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
7906 16-bit relocations off of the @code{$gp} register. This limits the
7907 size of the small data area to 64KB, but allows the variables to be
7908 directly accessed via a single instruction.
7910 The default is @option{-mlarge-data}. With this option the data area
7911 is limited to just below 2GB@. Programs that require more than 2GB of
7912 data must use @code{malloc} or @code{mmap} to allocate the data in the
7913 heap instead of in the program's data segment.
7915 When generating code for shared libraries, @option{-fpic} implies
7916 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
7920 @opindex msmall-text
7921 @opindex mlarge-text
7922 When @option{-msmall-text} is used, the compiler assumes that the
7923 code of the entire program (or shared library) fits in 4MB, and is
7924 thus reachable with a branch instruction. When @option{-msmall-data}
7925 is used, the compiler can assume that all local symbols share the
7926 same @code{$gp} value, and thus reduce the number of instructions
7927 required for a function call from 4 to 1.
7929 The default is @option{-mlarge-text}.
7931 @item -mcpu=@var{cpu_type}
7933 Set the instruction set and instruction scheduling parameters for
7934 machine type @var{cpu_type}. You can specify either the @samp{EV}
7935 style name or the corresponding chip number. GCC supports scheduling
7936 parameters for the EV4, EV5 and EV6 family of processors and will
7937 choose the default values for the instruction set from the processor
7938 you specify. If you do not specify a processor type, GCC will default
7939 to the processor on which the compiler was built.
7941 Supported values for @var{cpu_type} are
7947 Schedules as an EV4 and has no instruction set extensions.
7951 Schedules as an EV5 and has no instruction set extensions.
7955 Schedules as an EV5 and supports the BWX extension.
7960 Schedules as an EV5 and supports the BWX and MAX extensions.
7964 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
7968 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
7971 @item -mtune=@var{cpu_type}
7973 Set only the instruction scheduling parameters for machine type
7974 @var{cpu_type}. The instruction set is not changed.
7976 @item -mmemory-latency=@var{time}
7977 @opindex mmemory-latency
7978 Sets the latency the scheduler should assume for typical memory
7979 references as seen by the application. This number is highly
7980 dependent on the memory access patterns used by the application
7981 and the size of the external cache on the machine.
7983 Valid options for @var{time} are
7987 A decimal number representing clock cycles.
7993 The compiler contains estimates of the number of clock cycles for
7994 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
7995 (also called Dcache, Scache, and Bcache), as well as to main memory.
7996 Note that L3 is only valid for EV5.
8001 @node DEC Alpha/VMS Options
8002 @subsection DEC Alpha/VMS Options
8004 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8007 @item -mvms-return-codes
8008 @opindex mvms-return-codes
8009 Return VMS condition codes from main. The default is to return POSIX
8010 style condition (e.g.@ error) codes.
8014 @subsection FRV Options
8021 Only use the first 32 general purpose registers.
8026 Use all 64 general purpose registers.
8031 Use only the first 32 floating point registers.
8036 Use all 64 floating point registers
8039 @opindex mhard-float
8041 Use hardware instructions for floating point operations.
8044 @opindex msoft-float
8046 Use library routines for floating point operations.
8051 Dynamically allocate condition code registers.
8056 Do not try to dynamically allocate condition code registers, only
8057 use @code{icc0} and @code{fcc0}.
8062 Change ABI to use double word insns.
8067 Do not use double word instructions.
8072 Use floating point double instructions.
8077 Do not use floating point double instructions.
8082 Use media instructions.
8087 Do not use media instructions.
8092 Use multiply and add/subtract instructions.
8097 Do not use multiply and add/subtract instructions.
8102 Select the FDPIC ABI, that uses function descriptors to represent
8103 pointers to functions. Without any PIC/PIE-related options, it
8104 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8105 assumes GOT entries and small data are within a 12-bit range from the
8106 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8107 are computed with 32 bits.
8110 @opindex minline-plt
8112 Enable inlining of PLT entries in function calls to functions that are
8113 not known to bind locally. It has no effect without @option{-mfdpic}.
8114 It's enabled by default if optimizing for speed and compiling for
8115 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8116 optimization option such as @option{-O3} or above is present in the
8122 Assume a large TLS segment when generating thread-local code.
8127 Do not assume a large TLS segment when generating thread-local code.
8132 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8133 that is known to be in read-only sections. It's enabled by default,
8134 except for @option{-fpic} or @option{-fpie}: even though it may help
8135 make the global offset table smaller, it trades 1 instruction for 4.
8136 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8137 one of which may be shared by multiple symbols, and it avoids the need
8138 for a GOT entry for the referenced symbol, so it's more likely to be a
8139 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8141 @item -multilib-library-pic
8142 @opindex multilib-library-pic
8144 Link with the (library, not FD) pic libraries. It's implied by
8145 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8146 @option{-fpic} without @option{-mfdpic}. You should never have to use
8152 Follow the EABI requirement of always creating a frame pointer whenever
8153 a stack frame is allocated. This option is enabled by default and can
8154 be disabled with @option{-mno-linked-fp}.
8157 @opindex mlong-calls
8159 Use indirect addressing to call functions outside the current
8160 compilation unit. This allows the functions to be placed anywhere
8161 within the 32-bit address space.
8163 @item -malign-labels
8164 @opindex malign-labels
8166 Try to align labels to an 8-byte boundary by inserting nops into the
8167 previous packet. This option only has an effect when VLIW packing
8168 is enabled. It doesn't create new packets; it merely adds nops to
8172 @opindex mlibrary-pic
8174 Generate position-independent EABI code.
8179 Use only the first four media accumulator registers.
8184 Use all eight media accumulator registers.
8189 Pack VLIW instructions.
8194 Do not pack VLIW instructions.
8199 Do not mark ABI switches in e_flags.
8204 Enable the use of conditional-move instructions (default).
8206 This switch is mainly for debugging the compiler and will likely be removed
8207 in a future version.
8209 @item -mno-cond-move
8210 @opindex mno-cond-move
8212 Disable the use of conditional-move instructions.
8214 This switch is mainly for debugging the compiler and will likely be removed
8215 in a future version.
8220 Enable the use of conditional set instructions (default).
8222 This switch is mainly for debugging the compiler and will likely be removed
8223 in a future version.
8228 Disable the use of conditional set instructions.
8230 This switch is mainly for debugging the compiler and will likely be removed
8231 in a future version.
8236 Enable the use of conditional execution (default).
8238 This switch is mainly for debugging the compiler and will likely be removed
8239 in a future version.
8241 @item -mno-cond-exec
8242 @opindex mno-cond-exec
8244 Disable the use of conditional execution.
8246 This switch is mainly for debugging the compiler and will likely be removed
8247 in a future version.
8250 @opindex mvliw-branch
8252 Run a pass to pack branches into VLIW instructions (default).
8254 This switch is mainly for debugging the compiler and will likely be removed
8255 in a future version.
8257 @item -mno-vliw-branch
8258 @opindex mno-vliw-branch
8260 Do not run a pass to pack branches into VLIW instructions.
8262 This switch is mainly for debugging the compiler and will likely be removed
8263 in a future version.
8265 @item -mmulti-cond-exec
8266 @opindex mmulti-cond-exec
8268 Enable optimization of @code{&&} and @code{||} in conditional execution
8271 This switch is mainly for debugging the compiler and will likely be removed
8272 in a future version.
8274 @item -mno-multi-cond-exec
8275 @opindex mno-multi-cond-exec
8277 Disable optimization of @code{&&} and @code{||} in conditional execution.
8279 This switch is mainly for debugging the compiler and will likely be removed
8280 in a future version.
8282 @item -mnested-cond-exec
8283 @opindex mnested-cond-exec
8285 Enable nested conditional execution optimizations (default).
8287 This switch is mainly for debugging the compiler and will likely be removed
8288 in a future version.
8290 @item -mno-nested-cond-exec
8291 @opindex mno-nested-cond-exec
8293 Disable nested conditional execution optimizations.
8295 This switch is mainly for debugging the compiler and will likely be removed
8296 in a future version.
8298 @item -mtomcat-stats
8299 @opindex mtomcat-stats
8301 Cause gas to print out tomcat statistics.
8303 @item -mcpu=@var{cpu}
8306 Select the processor type for which to generate code. Possible values are
8307 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8308 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8312 @node H8/300 Options
8313 @subsection H8/300 Options
8315 These @samp{-m} options are defined for the H8/300 implementations:
8320 Shorten some address references at link time, when possible; uses the
8321 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8322 ld, Using ld}, for a fuller description.
8326 Generate code for the H8/300H@.
8330 Generate code for the H8S@.
8334 Generate code for the H8S and H8/300H in the normal mode. This switch
8335 must be used either with @option{-mh} or @option{-ms}.
8339 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8343 Make @code{int} data 32 bits by default.
8347 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8348 The default for the H8/300H and H8S is to align longs and floats on 4
8350 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8351 This option has no effect on the H8/300.
8355 @subsection HPPA Options
8356 @cindex HPPA Options
8358 These @samp{-m} options are defined for the HPPA family of computers:
8361 @item -march=@var{architecture-type}
8363 Generate code for the specified architecture. The choices for
8364 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8365 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8366 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8367 architecture option for your machine. Code compiled for lower numbered
8368 architectures will run on higher numbered architectures, but not the
8372 @itemx -mpa-risc-1-1
8373 @itemx -mpa-risc-2-0
8374 @opindex mpa-risc-1-0
8375 @opindex mpa-risc-1-1
8376 @opindex mpa-risc-2-0
8377 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8380 @opindex mbig-switch
8381 Generate code suitable for big switch tables. Use this option only if
8382 the assembler/linker complain about out of range branches within a switch
8385 @item -mjump-in-delay
8386 @opindex mjump-in-delay
8387 Fill delay slots of function calls with unconditional jump instructions
8388 by modifying the return pointer for the function call to be the target
8389 of the conditional jump.
8391 @item -mdisable-fpregs
8392 @opindex mdisable-fpregs
8393 Prevent floating point registers from being used in any manner. This is
8394 necessary for compiling kernels which perform lazy context switching of
8395 floating point registers. If you use this option and attempt to perform
8396 floating point operations, the compiler will abort.
8398 @item -mdisable-indexing
8399 @opindex mdisable-indexing
8400 Prevent the compiler from using indexing address modes. This avoids some
8401 rather obscure problems when compiling MIG generated code under MACH@.
8403 @item -mno-space-regs
8404 @opindex mno-space-regs
8405 Generate code that assumes the target has no space registers. This allows
8406 GCC to generate faster indirect calls and use unscaled index address modes.
8408 Such code is suitable for level 0 PA systems and kernels.
8410 @item -mfast-indirect-calls
8411 @opindex mfast-indirect-calls
8412 Generate code that assumes calls never cross space boundaries. This
8413 allows GCC to emit code which performs faster indirect calls.
8415 This option will not work in the presence of shared libraries or nested
8418 @item -mfixed-range=@var{register-range}
8419 @opindex mfixed-range
8420 Generate code treating the given register range as fixed registers.
8421 A fixed register is one that the register allocator can not use. This is
8422 useful when compiling kernel code. A register range is specified as
8423 two registers separated by a dash. Multiple register ranges can be
8424 specified separated by a comma.
8426 @item -mlong-load-store
8427 @opindex mlong-load-store
8428 Generate 3-instruction load and store sequences as sometimes required by
8429 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8432 @item -mportable-runtime
8433 @opindex mportable-runtime
8434 Use the portable calling conventions proposed by HP for ELF systems.
8438 Enable the use of assembler directives only GAS understands.
8440 @item -mschedule=@var{cpu-type}
8442 Schedule code according to the constraints for the machine type
8443 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8444 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8445 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8446 proper scheduling option for your machine. The default scheduling is
8450 @opindex mlinker-opt
8451 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8452 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8453 linkers in which they give bogus error messages when linking some programs.
8456 @opindex msoft-float
8457 Generate output containing library calls for floating point.
8458 @strong{Warning:} the requisite libraries are not available for all HPPA
8459 targets. Normally the facilities of the machine's usual C compiler are
8460 used, but this cannot be done directly in cross-compilation. You must make
8461 your own arrangements to provide suitable library functions for
8462 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8463 does provide software floating point support.
8465 @option{-msoft-float} changes the calling convention in the output file;
8466 therefore, it is only useful if you compile @emph{all} of a program with
8467 this option. In particular, you need to compile @file{libgcc.a}, the
8468 library that comes with GCC, with @option{-msoft-float} in order for
8473 Generate the predefine, @code{_SIO}, for server IO@. The default is
8474 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8475 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
8476 options are available under HP-UX and HI-UX@.
8480 Use GNU ld specific options. This passes @option{-shared} to ld when
8481 building a shared library. It is the default when GCC is configured,
8482 explicitly or implicitly, with the GNU linker. This option does not
8483 have any affect on which ld is called, it only changes what parameters
8484 are passed to that ld. The ld that is called is determined by the
8485 @option{--with-ld} configure option, GCC's program search path, and
8486 finally by the user's @env{PATH}. The linker used by GCC can be printed
8487 using @samp{which `gcc -print-prog-name=ld`}.
8491 Use HP ld specific options. This passes @option{-b} to ld when building
8492 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8493 links. It is the default when GCC is configured, explicitly or
8494 implicitly, with the HP linker. This option does not have any affect on
8495 which ld is called, it only changes what parameters are passed to that
8496 ld. The ld that is called is determined by the @option{--with-ld}
8497 configure option, GCC's program search path, and finally by the user's
8498 @env{PATH}. The linker used by GCC can be printed using @samp{which
8499 `gcc -print-prog-name=ld`}.
8502 @opindex mno-long-calls
8503 Generate code that uses long call sequences. This ensures that a call
8504 is always able to reach linker generated stubs. The default is to generate
8505 long calls only when the distance from the call site to the beginning
8506 of the function or translation unit, as the case may be, exceeds a
8507 predefined limit set by the branch type being used. The limits for
8508 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8509 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8512 Distances are measured from the beginning of functions when using the
8513 @option{-ffunction-sections} option, or when using the @option{-mgas}
8514 and @option{-mno-portable-runtime} options together under HP-UX with
8517 It is normally not desirable to use this option as it will degrade
8518 performance. However, it may be useful in large applications,
8519 particularly when partial linking is used to build the application.
8521 The types of long calls used depends on the capabilities of the
8522 assembler and linker, and the type of code being generated. The
8523 impact on systems that support long absolute calls, and long pic
8524 symbol-difference or pc-relative calls should be relatively small.
8525 However, an indirect call is used on 32-bit ELF systems in pic code
8526 and it is quite long.
8528 @item -munix=@var{unix-std}
8530 Generate compiler predefines and select a startfile for the specified
8531 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
8532 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
8533 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
8534 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
8535 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
8538 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
8539 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
8540 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
8541 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
8542 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
8543 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
8545 It is @emph{important} to note that this option changes the interfaces
8546 for various library routines. It also affects the operational behavior
8547 of the C library. Thus, @emph{extreme} care is needed in using this
8550 Library code that is intended to operate with more than one UNIX
8551 standard must test, set and restore the variable @var{__xpg4_extended_mask}
8552 as appropriate. Most GNU software doesn't provide this capability.
8556 Suppress the generation of link options to search libdld.sl when the
8557 @option{-static} option is specified on HP-UX 10 and later.
8561 The HP-UX implementation of setlocale in libc has a dependency on
8562 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8563 when the @option{-static} option is specified, special link options
8564 are needed to resolve this dependency.
8566 On HP-UX 10 and later, the GCC driver adds the necessary options to
8567 link with libdld.sl when the @option{-static} option is specified.
8568 This causes the resulting binary to be dynamic. On the 64-bit port,
8569 the linkers generate dynamic binaries by default in any case. The
8570 @option{-nolibdld} option can be used to prevent the GCC driver from
8571 adding these link options.
8575 Add support for multithreading with the @dfn{dce thread} library
8576 under HP-UX@. This option sets flags for both the preprocessor and
8580 @node i386 and x86-64 Options
8581 @subsection Intel 386 and AMD x86-64 Options
8582 @cindex i386 Options
8583 @cindex x86-64 Options
8584 @cindex Intel 386 Options
8585 @cindex AMD x86-64 Options
8587 These @samp{-m} options are defined for the i386 and x86-64 family of
8591 @item -mtune=@var{cpu-type}
8593 Tune to @var{cpu-type} everything applicable about the generated code, except
8594 for the ABI and the set of available instructions. The choices for
8598 Original Intel's i386 CPU@.
8600 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
8602 Intel Pentium CPU with no MMX support.
8604 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
8605 @item i686, pentiumpro
8606 Intel PentiumPro CPU@.
8608 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
8609 @item pentium3, pentium3m
8610 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
8613 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
8614 support. Used by Centrino notebooks.
8615 @item pentium4, pentium4m
8616 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
8618 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
8621 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
8622 SSE2 and SSE3 instruction set support.
8624 AMD K6 CPU with MMX instruction set support.
8626 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
8627 @item athlon, athlon-tbird
8628 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
8630 @item athlon-4, athlon-xp, athlon-mp
8631 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
8632 instruction set support.
8633 @item k8, opteron, athlon64, athlon-fx
8634 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
8635 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
8637 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
8640 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
8641 instruction set support.
8643 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
8644 implemented for this chip.)
8646 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
8647 implemented for this chip.)
8650 While picking a specific @var{cpu-type} will schedule things appropriately
8651 for that particular chip, the compiler will not generate any code that
8652 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8655 @item -march=@var{cpu-type}
8657 Generate instructions for the machine type @var{cpu-type}. The choices
8658 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8659 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8661 @item -mcpu=@var{cpu-type}
8663 A deprecated synonym for @option{-mtune}.
8672 @opindex mpentiumpro
8673 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8674 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8675 These synonyms are deprecated.
8677 @item -mfpmath=@var{unit}
8679 Generate floating point arithmetics for selected unit @var{unit}. The choices
8684 Use the standard 387 floating point coprocessor present majority of chips and
8685 emulated otherwise. Code compiled with this option will run almost everywhere.
8686 The temporary results are computed in 80bit precision instead of precision
8687 specified by the type resulting in slightly different results compared to most
8688 of other chips. See @option{-ffloat-store} for more detailed description.
8690 This is the default choice for i386 compiler.
8693 Use scalar floating point instructions present in the SSE instruction set.
8694 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8695 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8696 instruction set supports only single precision arithmetics, thus the double and
8697 extended precision arithmetics is still done using 387. Later version, present
8698 only in Pentium4 and the future AMD x86-64 chips supports double precision
8701 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
8702 or @option{-msse2} switches to enable SSE extensions and make this option
8703 effective. For the x86-64 compiler, these extensions are enabled by default.
8705 The resulting code should be considerably faster in the majority of cases and avoid
8706 the numerical instability problems of 387 code, but may break some existing
8707 code that expects temporaries to be 80bit.
8709 This is the default choice for the x86-64 compiler.
8712 Attempt to utilize both instruction sets at once. This effectively double the
8713 amount of available registers and on chips with separate execution units for
8714 387 and SSE the execution resources too. Use this option with care, as it is
8715 still experimental, because the GCC register allocator does not model separate
8716 functional units well resulting in instable performance.
8719 @item -masm=@var{dialect}
8720 @opindex masm=@var{dialect}
8721 Output asm instructions using selected @var{dialect}. Supported choices are
8722 @samp{intel} or @samp{att} (the default one).
8727 @opindex mno-ieee-fp
8728 Control whether or not the compiler uses IEEE floating point
8729 comparisons. These handle correctly the case where the result of a
8730 comparison is unordered.
8733 @opindex msoft-float
8734 Generate output containing library calls for floating point.
8735 @strong{Warning:} the requisite libraries are not part of GCC@.
8736 Normally the facilities of the machine's usual C compiler are used, but
8737 this can't be done directly in cross-compilation. You must make your
8738 own arrangements to provide suitable library functions for
8741 On machines where a function returns floating point results in the 80387
8742 register stack, some floating point opcodes may be emitted even if
8743 @option{-msoft-float} is used.
8745 @item -mno-fp-ret-in-387
8746 @opindex mno-fp-ret-in-387
8747 Do not use the FPU registers for return values of functions.
8749 The usual calling convention has functions return values of types
8750 @code{float} and @code{double} in an FPU register, even if there
8751 is no FPU@. The idea is that the operating system should emulate
8754 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8755 in ordinary CPU registers instead.
8757 @item -mno-fancy-math-387
8758 @opindex mno-fancy-math-387
8759 Some 387 emulators do not support the @code{sin}, @code{cos} and
8760 @code{sqrt} instructions for the 387. Specify this option to avoid
8761 generating those instructions. This option is the default on FreeBSD,
8762 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8763 indicates that the target cpu will always have an FPU and so the
8764 instruction will not need emulation. As of revision 2.6.1, these
8765 instructions are not generated unless you also use the
8766 @option{-funsafe-math-optimizations} switch.
8768 @item -malign-double
8769 @itemx -mno-align-double
8770 @opindex malign-double
8771 @opindex mno-align-double
8772 Control whether GCC aligns @code{double}, @code{long double}, and
8773 @code{long long} variables on a two word boundary or a one word
8774 boundary. Aligning @code{double} variables on a two word boundary will
8775 produce code that runs somewhat faster on a @samp{Pentium} at the
8776 expense of more memory.
8778 @strong{Warning:} if you use the @option{-malign-double} switch,
8779 structures containing the above types will be aligned differently than
8780 the published application binary interface specifications for the 386
8781 and will not be binary compatible with structures in code compiled
8782 without that switch.
8784 @item -m96bit-long-double
8785 @itemx -m128bit-long-double
8786 @opindex m96bit-long-double
8787 @opindex m128bit-long-double
8788 These switches control the size of @code{long double} type. The i386
8789 application binary interface specifies the size to be 96 bits,
8790 so @option{-m96bit-long-double} is the default in 32 bit mode.
8792 Modern architectures (Pentium and newer) would prefer @code{long double}
8793 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8794 conforming to the ABI, this would not be possible. So specifying a
8795 @option{-m128bit-long-double} will align @code{long double}
8796 to a 16 byte boundary by padding the @code{long double} with an additional
8799 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8800 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8802 Notice that neither of these options enable any extra precision over the x87
8803 standard of 80 bits for a @code{long double}.
8805 @strong{Warning:} if you override the default value for your target ABI, the
8806 structures and arrays containing @code{long double} variables will change
8807 their size as well as function calling convention for function taking
8808 @code{long double} will be modified. Hence they will not be binary
8809 compatible with arrays or structures in code compiled without that switch.
8813 @itemx -mno-svr3-shlib
8814 @opindex msvr3-shlib
8815 @opindex mno-svr3-shlib
8816 Control whether GCC places uninitialized local variables into the
8817 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8818 into @code{bss}. These options are meaningful only on System V Release 3.
8822 Use a different function-calling convention, in which functions that
8823 take a fixed number of arguments return with the @code{ret} @var{num}
8824 instruction, which pops their arguments while returning. This saves one
8825 instruction in the caller since there is no need to pop the arguments
8828 You can specify that an individual function is called with this calling
8829 sequence with the function attribute @samp{stdcall}. You can also
8830 override the @option{-mrtd} option by using the function attribute
8831 @samp{cdecl}. @xref{Function Attributes}.
8833 @strong{Warning:} this calling convention is incompatible with the one
8834 normally used on Unix, so you cannot use it if you need to call
8835 libraries compiled with the Unix compiler.
8837 Also, you must provide function prototypes for all functions that
8838 take variable numbers of arguments (including @code{printf});
8839 otherwise incorrect code will be generated for calls to those
8842 In addition, seriously incorrect code will result if you call a
8843 function with too many arguments. (Normally, extra arguments are
8844 harmlessly ignored.)
8846 @item -mregparm=@var{num}
8848 Control how many registers are used to pass integer arguments. By
8849 default, no registers are used to pass arguments, and at most 3
8850 registers can be used. You can control this behavior for a specific
8851 function by using the function attribute @samp{regparm}.
8852 @xref{Function Attributes}.
8854 @strong{Warning:} if you use this switch, and
8855 @var{num} is nonzero, then you must build all modules with the same
8856 value, including any libraries. This includes the system libraries and
8859 @item -mpreferred-stack-boundary=@var{num}
8860 @opindex mpreferred-stack-boundary
8861 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8862 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8863 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8864 size (@option{-Os}), in which case the default is the minimum correct
8865 alignment (4 bytes for x86, and 8 bytes for x86-64).
8867 On Pentium and PentiumPro, @code{double} and @code{long double} values
8868 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8869 suffer significant run time performance penalties. On Pentium III, the
8870 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8871 penalties if it is not 16 byte aligned.
8873 To ensure proper alignment of this values on the stack, the stack boundary
8874 must be as aligned as that required by any value stored on the stack.
8875 Further, every function must be generated such that it keeps the stack
8876 aligned. Thus calling a function compiled with a higher preferred
8877 stack boundary from a function compiled with a lower preferred stack
8878 boundary will most likely misalign the stack. It is recommended that
8879 libraries that use callbacks always use the default setting.
8881 This extra alignment does consume extra stack space, and generally
8882 increases code size. Code that is sensitive to stack space usage, such
8883 as embedded systems and operating system kernels, may want to reduce the
8884 preferred alignment to @option{-mpreferred-stack-boundary=2}.
8902 These switches enable or disable the use of built-in functions that allow
8903 direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow extensions of the
8906 @xref{X86 Built-in Functions}, for details of the functions enabled
8907 and disabled by these switches.
8909 To have SSE/SSE2 instructions generated automatically from floating-point
8910 code, see @option{-mfpmath=sse}.
8913 @itemx -mno-push-args
8915 @opindex mno-push-args
8916 Use PUSH operations to store outgoing parameters. This method is shorter
8917 and usually equally fast as method using SUB/MOV operations and is enabled
8918 by default. In some cases disabling it may improve performance because of
8919 improved scheduling and reduced dependencies.
8921 @item -maccumulate-outgoing-args
8922 @opindex maccumulate-outgoing-args
8923 If enabled, the maximum amount of space required for outgoing arguments will be
8924 computed in the function prologue. This is faster on most modern CPUs
8925 because of reduced dependencies, improved scheduling and reduced stack usage
8926 when preferred stack boundary is not equal to 2. The drawback is a notable
8927 increase in code size. This switch implies @option{-mno-push-args}.
8931 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8932 on thread-safe exception handling must compile and link all code with the
8933 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8934 @option{-D_MT}; when linking, it links in a special thread helper library
8935 @option{-lmingwthrd} which cleans up per thread exception handling data.
8937 @item -mno-align-stringops
8938 @opindex mno-align-stringops
8939 Do not align destination of inlined string operations. This switch reduces
8940 code size and improves performance in case the destination is already aligned,
8941 but GCC doesn't know about it.
8943 @item -minline-all-stringops
8944 @opindex minline-all-stringops
8945 By default GCC inlines string operations only when destination is known to be
8946 aligned at least to 4 byte boundary. This enables more inlining, increase code
8947 size, but may improve performance of code that depends on fast memcpy, strlen
8948 and memset for short lengths.
8950 @item -momit-leaf-frame-pointer
8951 @opindex momit-leaf-frame-pointer
8952 Don't keep the frame pointer in a register for leaf functions. This
8953 avoids the instructions to save, set up and restore frame pointers and
8954 makes an extra register available in leaf functions. The option
8955 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8956 which might make debugging harder.
8958 @item -mtls-direct-seg-refs
8959 @itemx -mno-tls-direct-seg-refs
8960 @opindex mtls-direct-seg-refs
8961 Controls whether TLS variables may be accessed with offsets from the
8962 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
8963 or whether the thread base pointer must be added. Whether or not this
8964 is legal depends on the operating system, and whether it maps the
8965 segment to cover the entire TLS area.
8967 For systems that use GNU libc, the default is on.
8970 These @samp{-m} switches are supported in addition to the above
8971 on AMD x86-64 processors in 64-bit environments.
8978 Generate code for a 32-bit or 64-bit environment.
8979 The 32-bit environment sets int, long and pointer to 32 bits and
8980 generates code that runs on any i386 system.
8981 The 64-bit environment sets int to 32 bits and long and pointer
8982 to 64 bits and generates code for AMD's x86-64 architecture.
8985 @opindex no-red-zone
8986 Do not use a so called red zone for x86-64 code. The red zone is mandated
8987 by the x86-64 ABI, it is a 128-byte area beyond the location of the
8988 stack pointer that will not be modified by signal or interrupt handlers
8989 and therefore can be used for temporary data without adjusting the stack
8990 pointer. The flag @option{-mno-red-zone} disables this red zone.
8992 @item -mcmodel=small
8993 @opindex mcmodel=small
8994 Generate code for the small code model: the program and its symbols must
8995 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
8996 Programs can be statically or dynamically linked. This is the default
8999 @item -mcmodel=kernel
9000 @opindex mcmodel=kernel
9001 Generate code for the kernel code model. The kernel runs in the
9002 negative 2 GB of the address space.
9003 This model has to be used for Linux kernel code.
9005 @item -mcmodel=medium
9006 @opindex mcmodel=medium
9007 Generate code for the medium model: The program is linked in the lower 2
9008 GB of the address space but symbols can be located anywhere in the
9009 address space. Programs can be statically or dynamically linked, but
9010 building of shared libraries are not supported with the medium model.
9012 @item -mcmodel=large
9013 @opindex mcmodel=large
9014 Generate code for the large model: This model makes no assumptions
9015 about addresses and sizes of sections. Currently GCC does not implement
9020 @subsection IA-64 Options
9021 @cindex IA-64 Options
9023 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9027 @opindex mbig-endian
9028 Generate code for a big endian target. This is the default for HP-UX@.
9030 @item -mlittle-endian
9031 @opindex mlittle-endian
9032 Generate code for a little endian target. This is the default for AIX5
9039 Generate (or don't) code for the GNU assembler. This is the default.
9040 @c Also, this is the default if the configure option @option{--with-gnu-as}
9047 Generate (or don't) code for the GNU linker. This is the default.
9048 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9053 Generate code that does not use a global pointer register. The result
9054 is not position independent code, and violates the IA-64 ABI@.
9056 @item -mvolatile-asm-stop
9057 @itemx -mno-volatile-asm-stop
9058 @opindex mvolatile-asm-stop
9059 @opindex mno-volatile-asm-stop
9060 Generate (or don't) a stop bit immediately before and after volatile asm
9063 @item -mregister-names
9064 @itemx -mno-register-names
9065 @opindex mregister-names
9066 @opindex mno-register-names
9067 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9068 the stacked registers. This may make assembler output more readable.
9074 Disable (or enable) optimizations that use the small data section. This may
9075 be useful for working around optimizer bugs.
9078 @opindex mconstant-gp
9079 Generate code that uses a single constant global pointer value. This is
9080 useful when compiling kernel code.
9084 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9085 This is useful when compiling firmware code.
9087 @item -minline-float-divide-min-latency
9088 @opindex minline-float-divide-min-latency
9089 Generate code for inline divides of floating point values
9090 using the minimum latency algorithm.
9092 @item -minline-float-divide-max-throughput
9093 @opindex minline-float-divide-max-throughput
9094 Generate code for inline divides of floating point values
9095 using the maximum throughput algorithm.
9097 @item -minline-int-divide-min-latency
9098 @opindex minline-int-divide-min-latency
9099 Generate code for inline divides of integer values
9100 using the minimum latency algorithm.
9102 @item -minline-int-divide-max-throughput
9103 @opindex minline-int-divide-max-throughput
9104 Generate code for inline divides of integer values
9105 using the maximum throughput algorithm.
9107 @item -minline-sqrt-min-latency
9108 @opindex minline-sqrt-min-latency
9109 Generate code for inline square roots
9110 using the minimum latency algorithm.
9112 @item -minline-sqrt-max-throughput
9113 @opindex minline-sqrt-max-throughput
9114 Generate code for inline square roots
9115 using the maximum throughput algorithm.
9117 @item -mno-dwarf2-asm
9119 @opindex mno-dwarf2-asm
9120 @opindex mdwarf2-asm
9121 Don't (or do) generate assembler code for the DWARF2 line number debugging
9122 info. This may be useful when not using the GNU assembler.
9124 @item -mearly-stop-bits
9125 @itemx -mno-early-stop-bits
9126 @opindex mearly-stop-bits
9127 @opindex mno-early-stop-bits
9128 Allow stop bits to be placed earlier than immediately preceding the
9129 instruction that triggered the stop bit. This can improve instruction
9130 scheduling, but does not always do so.
9132 @item -mfixed-range=@var{register-range}
9133 @opindex mfixed-range
9134 Generate code treating the given register range as fixed registers.
9135 A fixed register is one that the register allocator can not use. This is
9136 useful when compiling kernel code. A register range is specified as
9137 two registers separated by a dash. Multiple register ranges can be
9138 specified separated by a comma.
9140 @item -mtls-size=@var{tls-size}
9142 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9145 @item -mtune-arch=@var{cpu-type}
9147 Tune the instruction scheduling for a particular CPU, Valid values are
9148 itanium, itanium1, merced, itanium2, and mckinley.
9154 Add support for multithreading using the POSIX threads library. This
9155 option sets flags for both the preprocessor and linker. It does
9156 not affect the thread safety of object code produced by the compiler or
9157 that of libraries supplied with it. These are HP-UX specific flags.
9163 Generate code for a 32-bit or 64-bit environment.
9164 The 32-bit environment sets int, long and pointer to 32 bits.
9165 The 64-bit environment sets int to 32 bits and long and pointer
9166 to 64 bits. These are HP-UX specific flags.
9170 @node M32R/D Options
9171 @subsection M32R/D Options
9172 @cindex M32R/D options
9174 These @option{-m} options are defined for Renesas M32R/D architectures:
9179 Generate code for the M32R/2@.
9183 Generate code for the M32R/X@.
9187 Generate code for the M32R@. This is the default.
9190 @opindex mmodel=small
9191 Assume all objects live in the lower 16MB of memory (so that their addresses
9192 can be loaded with the @code{ld24} instruction), and assume all subroutines
9193 are reachable with the @code{bl} instruction.
9194 This is the default.
9196 The addressability of a particular object can be set with the
9197 @code{model} attribute.
9199 @item -mmodel=medium
9200 @opindex mmodel=medium
9201 Assume objects may be anywhere in the 32-bit address space (the compiler
9202 will generate @code{seth/add3} instructions to load their addresses), and
9203 assume all subroutines are reachable with the @code{bl} instruction.
9206 @opindex mmodel=large
9207 Assume objects may be anywhere in the 32-bit address space (the compiler
9208 will generate @code{seth/add3} instructions to load their addresses), and
9209 assume subroutines may not be reachable with the @code{bl} instruction
9210 (the compiler will generate the much slower @code{seth/add3/jl}
9211 instruction sequence).
9214 @opindex msdata=none
9215 Disable use of the small data area. Variables will be put into
9216 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
9217 @code{section} attribute has been specified).
9218 This is the default.
9220 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
9221 Objects may be explicitly put in the small data area with the
9222 @code{section} attribute using one of these sections.
9225 @opindex msdata=sdata
9226 Put small global and static data in the small data area, but do not
9227 generate special code to reference them.
9231 Put small global and static data in the small data area, and generate
9232 special instructions to reference them.
9236 @cindex smaller data references
9237 Put global and static objects less than or equal to @var{num} bytes
9238 into the small data or bss sections instead of the normal data or bss
9239 sections. The default value of @var{num} is 8.
9240 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
9241 for this option to have any effect.
9243 All modules should be compiled with the same @option{-G @var{num}} value.
9244 Compiling with different values of @var{num} may or may not work; if it
9245 doesn't the linker will give an error message---incorrect code will not be
9250 Makes the M32R specific code in the compiler display some statistics
9251 that might help in debugging programs.
9254 @opindex malign-loops
9255 Align all loops to a 32-byte boundary.
9257 @item -mno-align-loops
9258 @opindex mno-align-loops
9259 Do not enforce a 32-byte alignment for loops. This is the default.
9261 @item -missue-rate=@var{number}
9262 @opindex missue-rate=@var{number}
9263 Issue @var{number} instructions per cycle. @var{number} can only be 1
9266 @item -mbranch-cost=@var{number}
9267 @opindex mbranch-cost=@var{number}
9268 @var{number} can only be 1 or 2. If it is 1 then branches will be
9269 preferred over conditional code, if it is 2, then the opposite will
9272 @item -mflush-trap=@var{number}
9273 @opindex mflush-trap=@var{number}
9274 Specifies the trap number to use to flush the cache. The default is
9275 12. Valid numbers are between 0 and 15 inclusive.
9277 @item -mno-flush-trap
9278 @opindex mno-flush-trap
9279 Specifies that the cache cannot be flushed by using a trap.
9281 @item -mflush-func=@var{name}
9282 @opindex mflush-func=@var{name}
9283 Specifies the name of the operating system function to call to flush
9284 the cache. The default is @emph{_flush_cache}, but a function call
9285 will only be used if a trap is not available.
9287 @item -mno-flush-func
9288 @opindex mno-flush-func
9289 Indicates that there is no OS function for flushing the cache.
9293 @node M680x0 Options
9294 @subsection M680x0 Options
9295 @cindex M680x0 options
9297 These are the @samp{-m} options defined for the 68000 series. The default
9298 values for these options depends on which style of 68000 was selected when
9299 the compiler was configured; the defaults for the most common choices are
9307 Generate output for a 68000. This is the default
9308 when the compiler is configured for 68000-based systems.
9310 Use this option for microcontrollers with a 68000 or EC000 core,
9311 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
9317 Generate output for a 68020. This is the default
9318 when the compiler is configured for 68020-based systems.
9322 Generate output containing 68881 instructions for floating point.
9323 This is the default for most 68020 systems unless @option{--nfp} was
9324 specified when the compiler was configured.
9328 Generate output for a 68030. This is the default when the compiler is
9329 configured for 68030-based systems.
9333 Generate output for a 68040. This is the default when the compiler is
9334 configured for 68040-based systems.
9336 This option inhibits the use of 68881/68882 instructions that have to be
9337 emulated by software on the 68040. Use this option if your 68040 does not
9338 have code to emulate those instructions.
9342 Generate output for a 68060. This is the default when the compiler is
9343 configured for 68060-based systems.
9345 This option inhibits the use of 68020 and 68881/68882 instructions that
9346 have to be emulated by software on the 68060. Use this option if your 68060
9347 does not have code to emulate those instructions.
9351 Generate output for a CPU32. This is the default
9352 when the compiler is configured for CPU32-based systems.
9354 Use this option for microcontrollers with a
9355 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
9356 68336, 68340, 68341, 68349 and 68360.
9360 Generate output for a 520X ``coldfire'' family cpu. This is the default
9361 when the compiler is configured for 520X-based systems.
9363 Use this option for microcontroller with a 5200 core, including
9364 the MCF5202, MCF5203, MCF5204 and MCF5202.
9369 Generate output for a 68040, without using any of the new instructions.
9370 This results in code which can run relatively efficiently on either a
9371 68020/68881 or a 68030 or a 68040. The generated code does use the
9372 68881 instructions that are emulated on the 68040.
9376 Generate output for a 68060, without using any of the new instructions.
9377 This results in code which can run relatively efficiently on either a
9378 68020/68881 or a 68030 or a 68040. The generated code does use the
9379 68881 instructions that are emulated on the 68060.
9382 @opindex msoft-float
9383 Generate output containing library calls for floating point.
9384 @strong{Warning:} the requisite libraries are not available for all m68k
9385 targets. Normally the facilities of the machine's usual C compiler are
9386 used, but this can't be done directly in cross-compilation. You must
9387 make your own arrangements to provide suitable library functions for
9388 cross-compilation. The embedded targets @samp{m68k-*-aout} and
9389 @samp{m68k-*-coff} do provide software floating point support.
9393 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9394 Additionally, parameters passed on the stack are also aligned to a
9395 16-bit boundary even on targets whose API mandates promotion to 32-bit.
9398 @opindex mnobitfield
9399 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
9400 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
9404 Do use the bit-field instructions. The @option{-m68020} option implies
9405 @option{-mbitfield}. This is the default if you use a configuration
9406 designed for a 68020.
9410 Use a different function-calling convention, in which functions
9411 that take a fixed number of arguments return with the @code{rtd}
9412 instruction, which pops their arguments while returning. This
9413 saves one instruction in the caller since there is no need to pop
9414 the arguments there.
9416 This calling convention is incompatible with the one normally
9417 used on Unix, so you cannot use it if you need to call libraries
9418 compiled with the Unix compiler.
9420 Also, you must provide function prototypes for all functions that
9421 take variable numbers of arguments (including @code{printf});
9422 otherwise incorrect code will be generated for calls to those
9425 In addition, seriously incorrect code will result if you call a
9426 function with too many arguments. (Normally, extra arguments are
9427 harmlessly ignored.)
9429 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
9430 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
9433 @itemx -mno-align-int
9435 @opindex mno-align-int
9436 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
9437 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
9438 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
9439 Aligning variables on 32-bit boundaries produces code that runs somewhat
9440 faster on processors with 32-bit busses at the expense of more memory.
9442 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
9443 align structures containing the above types differently than
9444 most published application binary interface specifications for the m68k.
9448 Use the pc-relative addressing mode of the 68000 directly, instead of
9449 using a global offset table. At present, this option implies @option{-fpic},
9450 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
9451 not presently supported with @option{-mpcrel}, though this could be supported for
9452 68020 and higher processors.
9454 @item -mno-strict-align
9455 @itemx -mstrict-align
9456 @opindex mno-strict-align
9457 @opindex mstrict-align
9458 Do not (do) assume that unaligned memory references will be handled by
9462 Generate code that allows the data segment to be located in a different
9463 area of memory from the text segment. This allows for execute in place in
9464 an environment without virtual memory management. This option implies
9468 Generate code that assumes that the data segment follows the text segment.
9469 This is the default.
9471 @item -mid-shared-library
9472 Generate code that supports shared libraries via the library ID method.
9473 This allows for execute in place and shared libraries in an environment
9474 without virtual memory management. This option implies @option{-fPIC}.
9476 @item -mno-id-shared-library
9477 Generate code that doesn't assume ID based shared libraries are being used.
9478 This is the default.
9480 @item -mshared-library-id=n
9481 Specified the identification number of the ID based shared library being
9482 compiled. Specifying a value of 0 will generate more compact code, specifying
9483 other values will force the allocation of that number to the current
9484 library but is no more space or time efficient than omitting this option.
9488 @node M68hc1x Options
9489 @subsection M68hc1x Options
9490 @cindex M68hc1x options
9492 These are the @samp{-m} options defined for the 68hc11 and 68hc12
9493 microcontrollers. The default values for these options depends on
9494 which style of microcontroller was selected when the compiler was configured;
9495 the defaults for the most common choices are given below.
9502 Generate output for a 68HC11. This is the default
9503 when the compiler is configured for 68HC11-based systems.
9509 Generate output for a 68HC12. This is the default
9510 when the compiler is configured for 68HC12-based systems.
9516 Generate output for a 68HCS12.
9519 @opindex mauto-incdec
9520 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
9527 Enable the use of 68HC12 min and max instructions.
9530 @itemx -mno-long-calls
9531 @opindex mlong-calls
9532 @opindex mno-long-calls
9533 Treat all calls as being far away (near). If calls are assumed to be
9534 far away, the compiler will use the @code{call} instruction to
9535 call a function and the @code{rtc} instruction for returning.
9539 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9541 @item -msoft-reg-count=@var{count}
9542 @opindex msoft-reg-count
9543 Specify the number of pseudo-soft registers which are used for the
9544 code generation. The maximum number is 32. Using more pseudo-soft
9545 register may or may not result in better code depending on the program.
9546 The default is 4 for 68HC11 and 2 for 68HC12.
9551 @subsection MCore Options
9552 @cindex MCore options
9554 These are the @samp{-m} options defined for the Motorola M*Core
9562 @opindex mno-hardlit
9563 Inline constants into the code stream if it can be done in two
9564 instructions or less.
9570 Use the divide instruction. (Enabled by default).
9572 @item -mrelax-immediate
9573 @itemx -mno-relax-immediate
9574 @opindex mrelax-immediate
9575 @opindex mno-relax-immediate
9576 Allow arbitrary sized immediates in bit operations.
9578 @item -mwide-bitfields
9579 @itemx -mno-wide-bitfields
9580 @opindex mwide-bitfields
9581 @opindex mno-wide-bitfields
9582 Always treat bit-fields as int-sized.
9584 @item -m4byte-functions
9585 @itemx -mno-4byte-functions
9586 @opindex m4byte-functions
9587 @opindex mno-4byte-functions
9588 Force all functions to be aligned to a four byte boundary.
9590 @item -mcallgraph-data
9591 @itemx -mno-callgraph-data
9592 @opindex mcallgraph-data
9593 @opindex mno-callgraph-data
9594 Emit callgraph information.
9597 @itemx -mno-slow-bytes
9598 @opindex mslow-bytes
9599 @opindex mno-slow-bytes
9600 Prefer word access when reading byte quantities.
9602 @item -mlittle-endian
9604 @opindex mlittle-endian
9605 @opindex mbig-endian
9606 Generate code for a little endian target.
9612 Generate code for the 210 processor.
9616 @subsection MIPS Options
9617 @cindex MIPS options
9623 Generate big-endian code.
9627 Generate little-endian code. This is the default for @samp{mips*el-*-*}
9630 @item -march=@var{arch}
9632 Generate code that will run on @var{arch}, which can be the name of a
9633 generic MIPS ISA, or the name of a particular processor.
9635 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
9636 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
9637 The processor names are:
9638 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
9640 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
9641 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
9645 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
9646 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
9647 The special value @samp{from-abi} selects the
9648 most compatible architecture for the selected ABI (that is,
9649 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
9651 In processor names, a final @samp{000} can be abbreviated as @samp{k}
9652 (for example, @samp{-march=r2k}). Prefixes are optional, and
9653 @samp{vr} may be written @samp{r}.
9655 GCC defines two macros based on the value of this option. The first
9656 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
9657 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
9658 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
9659 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
9660 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
9662 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
9663 above. In other words, it will have the full prefix and will not
9664 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
9665 the macro names the resolved architecture (either @samp{"mips1"} or
9666 @samp{"mips3"}). It names the default architecture when no
9667 @option{-march} option is given.
9669 @item -mtune=@var{arch}
9671 Optimize for @var{arch}. Among other things, this option controls
9672 the way instructions are scheduled, and the perceived cost of arithmetic
9673 operations. The list of @var{arch} values is the same as for
9676 When this option is not used, GCC will optimize for the processor
9677 specified by @option{-march}. By using @option{-march} and
9678 @option{-mtune} together, it is possible to generate code that will
9679 run on a family of processors, but optimize the code for one
9680 particular member of that family.
9682 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
9683 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
9684 @samp{-march} ones described above.
9688 Equivalent to @samp{-march=mips1}.
9692 Equivalent to @samp{-march=mips2}.
9696 Equivalent to @samp{-march=mips3}.
9700 Equivalent to @samp{-march=mips4}.
9704 Equivalent to @samp{-march=mips32}.
9708 Equivalent to @samp{-march=mips32r2}.
9712 Equivalent to @samp{-march=mips64}.
9718 Use (do not use) the MIPS16 ISA@.
9730 Generate code for the given ABI@.
9732 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
9733 generates 64-bit code when you select a 64-bit architecture, but you
9734 can use @option{-mgp32} to get 32-bit code instead.
9736 For information about the O64 ABI, see
9737 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
9740 @itemx -mno-abicalls
9742 @opindex mno-abicalls
9743 Generate (do not generate) SVR4-style position-independent code.
9744 @option{-mabicalls} is the default for SVR4-based systems.
9750 Lift (do not lift) the usual restrictions on the size of the global
9753 GCC normally uses a single instruction to load values from the GOT@.
9754 While this is relatively efficient, it will only work if the GOT
9755 is smaller than about 64k. Anything larger will cause the linker
9756 to report an error such as:
9758 @cindex relocation truncated to fit (MIPS)
9760 relocation truncated to fit: R_MIPS_GOT16 foobar
9763 If this happens, you should recompile your code with @option{-mxgot}.
9764 It should then work with very large GOTs, although it will also be
9765 less efficient, since it will take three instructions to fetch the
9766 value of a global symbol.
9768 Note that some linkers can create multiple GOTs. If you have such a
9769 linker, you should only need to use @option{-mxgot} when a single object
9770 file accesses more than 64k's worth of GOT entries. Very few do.
9772 These options have no effect unless GCC is generating position
9777 Assume that general-purpose registers are 32 bits wide.
9781 Assume that general-purpose registers are 64 bits wide.
9785 Assume that floating-point registers are 32 bits wide.
9789 Assume that floating-point registers are 64 bits wide.
9792 @opindex mhard-float
9793 Use floating-point coprocessor instructions.
9796 @opindex msoft-float
9797 Do not use floating-point coprocessor instructions. Implement
9798 floating-point calculations using library calls instead.
9800 @item -msingle-float
9801 @opindex msingle-float
9802 Assume that the floating-point coprocessor only supports single-precision
9805 @itemx -mdouble-float
9806 @opindex mdouble-float
9807 Assume that the floating-point coprocessor supports double-precision
9808 operations. This is the default.
9810 @itemx -mpaired-single
9811 @itemx -mno-paired-single
9812 @opindex mpaired-single
9813 @opindex mno-paired-single
9814 Use (do not use) paired-single floating-point instructions.
9815 @xref{MIPS Paired-Single Support}. This option can only be used
9816 when generating 64-bit code and requires hardware floating-point
9817 support to be enabled.
9823 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
9824 The option @option{-mips3d} implies @option{-mpaired-single}.
9828 Force @code{int} and @code{long} types to be 64 bits wide. See
9829 @option{-mlong32} for an explanation of the default and the way
9830 that the pointer size is determined.
9832 This option has been deprecated and will be removed in a future release.
9836 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
9837 an explanation of the default and the way that the pointer size is
9842 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
9844 The default size of @code{int}s, @code{long}s and pointers depends on
9845 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
9846 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
9847 32-bit @code{long}s. Pointers are the same size as @code{long}s,
9848 or the same size as integer registers, whichever is smaller.
9854 Assume (do not assume) that all symbols have 32-bit values, regardless
9855 of the selected ABI@. This option is useful in combination with
9856 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
9857 to generate shorter and faster references to symbolic addresses.
9861 @cindex smaller data references (MIPS)
9862 @cindex gp-relative references (MIPS)
9863 Put global and static items less than or equal to @var{num} bytes into
9864 the small data or bss section instead of the normal data or bss section.
9865 This allows the data to be accessed using a single instruction.
9867 All modules should be compiled with the same @option{-G @var{num}}
9870 @item -membedded-data
9871 @itemx -mno-embedded-data
9872 @opindex membedded-data
9873 @opindex mno-embedded-data
9874 Allocate variables to the read-only data section first if possible, then
9875 next in the small data section if possible, otherwise in data. This gives
9876 slightly slower code than the default, but reduces the amount of RAM required
9877 when executing, and thus may be preferred for some embedded systems.
9879 @item -muninit-const-in-rodata
9880 @itemx -mno-uninit-const-in-rodata
9881 @opindex muninit-const-in-rodata
9882 @opindex mno-uninit-const-in-rodata
9883 Put uninitialized @code{const} variables in the read-only data section.
9884 This option is only meaningful in conjunction with @option{-membedded-data}.
9886 @item -msplit-addresses
9887 @itemx -mno-split-addresses
9888 @opindex msplit-addresses
9889 @opindex mno-split-addresses
9890 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
9891 relocation operators. This option has been superseded by
9892 @option{-mexplicit-relocs} but is retained for backwards compatibility.
9894 @item -mexplicit-relocs
9895 @itemx -mno-explicit-relocs
9896 @opindex mexplicit-relocs
9897 @opindex mno-explicit-relocs
9898 Use (do not use) assembler relocation operators when dealing with symbolic
9899 addresses. The alternative, selected by @option{-mno-explicit-relocs},
9900 is to use assembler macros instead.
9902 @option{-mexplicit-relocs} is the default if GCC was configured
9903 to use an assembler that supports relocation operators.
9905 @item -mcheck-zero-division
9906 @itemx -mno-check-zero-division
9907 @opindex mcheck-zero-division
9908 @opindex mno-check-zero-division
9909 Trap (do not trap) on integer division by zero. The default is
9910 @option{-mcheck-zero-division}.
9912 @item -mdivide-traps
9913 @itemx -mdivide-breaks
9914 @opindex mdivide-traps
9915 @opindex mdivide-breaks
9916 MIPS systems check for division by zero by generating either a
9917 conditional trap or a break instruction. Using traps results in
9918 smaller code, but is only supported on MIPS II and later. Also, some
9919 versions of the Linux kernel have a bug that prevents trap from
9920 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
9921 allow conditional traps on architectures that support them and
9922 @option{-mdivide-breaks} to force the use of breaks.
9924 The default is usually @option{-mdivide-traps}, but this can be
9925 overridden at configure time using @option{--with-divide=breaks}.
9926 Divide-by-zero checks can be completely disabled using
9927 @option{-mno-check-zero-division}.
9933 Force (do not force) the use of @code{memcpy()} for non-trivial block
9934 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
9935 most constant-sized copies.
9938 @itemx -mno-long-calls
9939 @opindex mlong-calls
9940 @opindex mno-long-calls
9941 Disable (do not disable) use of the @code{jal} instruction. Calling
9942 functions using @code{jal} is more efficient but requires the caller
9943 and callee to be in the same 256 megabyte segment.
9945 This option has no effect on abicalls code. The default is
9946 @option{-mno-long-calls}.
9952 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
9953 instructions, as provided by the R4650 ISA@.
9956 @itemx -mno-fused-madd
9957 @opindex mfused-madd
9958 @opindex mno-fused-madd
9959 Enable (disable) use of the floating point multiply-accumulate
9960 instructions, when they are available. The default is
9961 @option{-mfused-madd}.
9963 When multiply-accumulate instructions are used, the intermediate
9964 product is calculated to infinite precision and is not subject to
9965 the FCSR Flush to Zero bit. This may be undesirable in some
9970 Tell the MIPS assembler to not run its preprocessor over user
9971 assembler files (with a @samp{.s} suffix) when assembling them.
9974 @itemx -mno-fix-r4000
9976 @opindex mno-fix-r4000
9977 Work around certain R4000 CPU errata:
9980 A double-word or a variable shift may give an incorrect result if executed
9981 immediately after starting an integer division.
9983 A double-word or a variable shift may give an incorrect result if executed
9984 while an integer multiplication is in progress.
9986 An integer division may give an incorrect result if started in a delay slot
9987 of a taken branch or a jump.
9991 @itemx -mno-fix-r4400
9993 @opindex mno-fix-r4400
9994 Work around certain R4400 CPU errata:
9997 A double-word or a variable shift may give an incorrect result if executed
9998 immediately after starting an integer division.
10002 @itemx -mno-fix-vr4120
10003 @opindex mfix-vr4120
10004 Work around certain VR4120 errata:
10007 @code{dmultu} does not always produce the correct result.
10009 @code{div} and @code{ddiv} do not always produce the correct result if one
10010 of the operands is negative.
10012 The workarounds for the division errata rely on special functions in
10013 @file{libgcc.a}. At present, these functions are only provided by
10014 the @code{mips64vr*-elf} configurations.
10016 Other VR4120 errata require a nop to be inserted between certain pairs of
10017 instructions. These errata are handled by the assembler, not by GCC itself.
10020 @opindex mfix-vr4130
10021 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
10022 workarounds are implemented by the assembler rather than by GCC,
10023 although GCC will avoid using @code{mflo} and @code{mfhi} if the
10024 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
10025 instructions are available instead.
10028 @itemx -mno-fix-sb1
10030 Work around certain SB-1 CPU core errata.
10031 (This flag currently works around the SB-1 revision 2
10032 ``F1'' and ``F2'' floating point errata.)
10034 @item -mflush-func=@var{func}
10035 @itemx -mno-flush-func
10036 @opindex mflush-func
10037 Specifies the function to call to flush the I and D caches, or to not
10038 call any such function. If called, the function must take the same
10039 arguments as the common @code{_flush_func()}, that is, the address of the
10040 memory range for which the cache is being flushed, the size of the
10041 memory range, and the number 3 (to flush both caches). The default
10042 depends on the target GCC was configured for, but commonly is either
10043 @samp{_flush_func} or @samp{__cpu_flush}.
10045 @item -mbranch-likely
10046 @itemx -mno-branch-likely
10047 @opindex mbranch-likely
10048 @opindex mno-branch-likely
10049 Enable or disable use of Branch Likely instructions, regardless of the
10050 default for the selected architecture. By default, Branch Likely
10051 instructions may be generated if they are supported by the selected
10052 architecture. An exception is for the MIPS32 and MIPS64 architectures
10053 and processors which implement those architectures; for those, Branch
10054 Likely instructions will not be generated by default because the MIPS32
10055 and MIPS64 architectures specifically deprecate their use.
10057 @item -mfp-exceptions
10058 @itemx -mno-fp-exceptions
10059 @opindex mfp-exceptions
10060 Specifies whether FP exceptions are enabled. This affects how we schedule
10061 FP instructions for some processors. The default is that FP exceptions are
10064 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
10065 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
10068 @item -mvr4130-align
10069 @itemx -mno-vr4130-align
10070 @opindex mvr4130-align
10071 The VR4130 pipeline is two-way superscalar, but can only issue two
10072 instructions together if the first one is 8-byte aligned. When this
10073 option is enabled, GCC will align pairs of instructions that it
10074 thinks should execute in parallel.
10076 This option only has an effect when optimizing for the VR4130.
10077 It normally makes code faster, but at the expense of making it bigger.
10078 It is enabled by default at optimization level @option{-O3}.
10082 @subsection MMIX Options
10083 @cindex MMIX Options
10085 These options are defined for the MMIX:
10089 @itemx -mno-libfuncs
10091 @opindex mno-libfuncs
10092 Specify that intrinsic library functions are being compiled, passing all
10093 values in registers, no matter the size.
10096 @itemx -mno-epsilon
10098 @opindex mno-epsilon
10099 Generate floating-point comparison instructions that compare with respect
10100 to the @code{rE} epsilon register.
10102 @item -mabi=mmixware
10104 @opindex mabi-mmixware
10106 Generate code that passes function parameters and return values that (in
10107 the called function) are seen as registers @code{$0} and up, as opposed to
10108 the GNU ABI which uses global registers @code{$231} and up.
10110 @item -mzero-extend
10111 @itemx -mno-zero-extend
10112 @opindex mzero-extend
10113 @opindex mno-zero-extend
10114 When reading data from memory in sizes shorter than 64 bits, use (do not
10115 use) zero-extending load instructions by default, rather than
10116 sign-extending ones.
10119 @itemx -mno-knuthdiv
10121 @opindex mno-knuthdiv
10122 Make the result of a division yielding a remainder have the same sign as
10123 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10124 remainder follows the sign of the dividend. Both methods are
10125 arithmetically valid, the latter being almost exclusively used.
10127 @item -mtoplevel-symbols
10128 @itemx -mno-toplevel-symbols
10129 @opindex mtoplevel-symbols
10130 @opindex mno-toplevel-symbols
10131 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10132 code can be used with the @code{PREFIX} assembly directive.
10136 Generate an executable in the ELF format, rather than the default
10137 @samp{mmo} format used by the @command{mmix} simulator.
10139 @item -mbranch-predict
10140 @itemx -mno-branch-predict
10141 @opindex mbranch-predict
10142 @opindex mno-branch-predict
10143 Use (do not use) the probable-branch instructions, when static branch
10144 prediction indicates a probable branch.
10146 @item -mbase-addresses
10147 @itemx -mno-base-addresses
10148 @opindex mbase-addresses
10149 @opindex mno-base-addresses
10150 Generate (do not generate) code that uses @emph{base addresses}. Using a
10151 base address automatically generates a request (handled by the assembler
10152 and the linker) for a constant to be set up in a global register. The
10153 register is used for one or more base address requests within the range 0
10154 to 255 from the value held in the register. The generally leads to short
10155 and fast code, but the number of different data items that can be
10156 addressed is limited. This means that a program that uses lots of static
10157 data may require @option{-mno-base-addresses}.
10159 @item -msingle-exit
10160 @itemx -mno-single-exit
10161 @opindex msingle-exit
10162 @opindex mno-single-exit
10163 Force (do not force) generated code to have a single exit point in each
10167 @node MN10300 Options
10168 @subsection MN10300 Options
10169 @cindex MN10300 options
10171 These @option{-m} options are defined for Matsushita MN10300 architectures:
10176 Generate code to avoid bugs in the multiply instructions for the MN10300
10177 processors. This is the default.
10179 @item -mno-mult-bug
10180 @opindex mno-mult-bug
10181 Do not generate code to avoid bugs in the multiply instructions for the
10182 MN10300 processors.
10186 Generate code which uses features specific to the AM33 processor.
10190 Do not generate code which uses features specific to the AM33 processor. This
10195 Do not link in the C run-time initialization object file.
10199 Indicate to the linker that it should perform a relaxation optimization pass
10200 to shorten branches, calls and absolute memory addresses. This option only
10201 has an effect when used on the command line for the final link step.
10203 This option makes symbolic debugging impossible.
10206 @node NS32K Options
10207 @subsection NS32K Options
10208 @cindex NS32K options
10210 These are the @samp{-m} options defined for the 32000 series. The default
10211 values for these options depends on which style of 32000 was selected when
10212 the compiler was configured; the defaults for the most common choices are
10220 Generate output for a 32032. This is the default
10221 when the compiler is configured for 32032 and 32016 based systems.
10227 Generate output for a 32332. This is the default
10228 when the compiler is configured for 32332-based systems.
10234 Generate output for a 32532. This is the default
10235 when the compiler is configured for 32532-based systems.
10239 Generate output containing 32081 instructions for floating point.
10240 This is the default for all systems.
10244 Generate output containing 32381 instructions for floating point. This
10245 also implies @option{-m32081}. The 32381 is only compatible with the 32332
10246 and 32532 cpus. This is the default for the pc532-netbsd configuration.
10249 @opindex mmulti-add
10250 Try and generate multiply-add floating point instructions @code{polyF}
10251 and @code{dotF}. This option is only available if the @option{-m32381}
10252 option is in effect. Using these instructions requires changes to
10253 register allocation which generally has a negative impact on
10254 performance. This option should only be enabled when compiling code
10255 particularly likely to make heavy use of multiply-add instructions.
10257 @item -mnomulti-add
10258 @opindex mnomulti-add
10259 Do not try and generate multiply-add floating point instructions
10260 @code{polyF} and @code{dotF}. This is the default on all platforms.
10263 @opindex msoft-float
10264 Generate output containing library calls for floating point.
10265 @strong{Warning:} the requisite libraries may not be available.
10267 @item -mieee-compare
10268 @itemx -mno-ieee-compare
10269 @opindex mieee-compare
10270 @opindex mno-ieee-compare
10271 Control whether or not the compiler uses IEEE floating point
10272 comparisons. These handle correctly the case where the result of a
10273 comparison is unordered.
10274 @strong{Warning:} the requisite kernel support may not be available.
10277 @opindex mnobitfield
10278 Do not use the bit-field instructions. On some machines it is faster to
10279 use shifting and masking operations. This is the default for the pc532.
10283 Do use the bit-field instructions. This is the default for all platforms
10288 Use a different function-calling convention, in which functions
10289 that take a fixed number of arguments return pop their
10290 arguments on return with the @code{ret} instruction.
10292 This calling convention is incompatible with the one normally
10293 used on Unix, so you cannot use it if you need to call libraries
10294 compiled with the Unix compiler.
10296 Also, you must provide function prototypes for all functions that
10297 take variable numbers of arguments (including @code{printf});
10298 otherwise incorrect code will be generated for calls to those
10301 In addition, seriously incorrect code will result if you call a
10302 function with too many arguments. (Normally, extra arguments are
10303 harmlessly ignored.)
10305 This option takes its name from the 680x0 @code{rtd} instruction.
10310 Use a different function-calling convention where the first two arguments
10311 are passed in registers.
10313 This calling convention is incompatible with the one normally
10314 used on Unix, so you cannot use it if you need to call libraries
10315 compiled with the Unix compiler.
10318 @opindex mnoregparam
10319 Do not pass any arguments in registers. This is the default for all
10324 It is OK to use the sb as an index register which is always loaded with
10325 zero. This is the default for the pc532-netbsd target.
10329 The sb register is not available for use or has not been initialized to
10330 zero by the run time system. This is the default for all targets except
10331 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
10332 @option{-fpic} is set.
10336 Many ns32000 series addressing modes use displacements of up to 512MB@.
10337 If an address is above 512MB then displacements from zero can not be used.
10338 This option causes code to be generated which can be loaded above 512MB@.
10339 This may be useful for operating systems or ROM code.
10343 Assume code will be loaded in the first 512MB of virtual address space.
10344 This is the default for all platforms.
10348 @node PDP-11 Options
10349 @subsection PDP-11 Options
10350 @cindex PDP-11 Options
10352 These options are defined for the PDP-11:
10357 Use hardware FPP floating point. This is the default. (FIS floating
10358 point on the PDP-11/40 is not supported.)
10361 @opindex msoft-float
10362 Do not use hardware floating point.
10366 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10370 Return floating-point results in memory. This is the default.
10374 Generate code for a PDP-11/40.
10378 Generate code for a PDP-11/45. This is the default.
10382 Generate code for a PDP-11/10.
10384 @item -mbcopy-builtin
10385 @opindex bcopy-builtin
10386 Use inline @code{movmemhi} patterns for copying memory. This is the
10391 Do not use inline @code{movmemhi} patterns for copying memory.
10397 Use 16-bit @code{int}. This is the default.
10403 Use 32-bit @code{int}.
10406 @itemx -mno-float32
10408 @opindex mno-float32
10409 Use 64-bit @code{float}. This is the default.
10412 @itemx -mno-float64
10414 @opindex mno-float64
10415 Use 32-bit @code{float}.
10419 Use @code{abshi2} pattern. This is the default.
10423 Do not use @code{abshi2} pattern.
10425 @item -mbranch-expensive
10426 @opindex mbranch-expensive
10427 Pretend that branches are expensive. This is for experimenting with
10428 code generation only.
10430 @item -mbranch-cheap
10431 @opindex mbranch-cheap
10432 Do not pretend that branches are expensive. This is the default.
10436 Generate code for a system with split I&D@.
10440 Generate code for a system without split I&D@. This is the default.
10444 Use Unix assembler syntax. This is the default when configured for
10445 @samp{pdp11-*-bsd}.
10449 Use DEC assembler syntax. This is the default when configured for any
10450 PDP-11 target other than @samp{pdp11-*-bsd}.
10453 @node PowerPC Options
10454 @subsection PowerPC Options
10455 @cindex PowerPC options
10457 These are listed under @xref{RS/6000 and PowerPC Options}.
10459 @node RS/6000 and PowerPC Options
10460 @subsection IBM RS/6000 and PowerPC Options
10461 @cindex RS/6000 and PowerPC Options
10462 @cindex IBM RS/6000 and PowerPC Options
10464 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
10471 @itemx -mno-powerpc
10472 @itemx -mpowerpc-gpopt
10473 @itemx -mno-powerpc-gpopt
10474 @itemx -mpowerpc-gfxopt
10475 @itemx -mno-powerpc-gfxopt
10477 @itemx -mno-powerpc64
10481 @opindex mno-power2
10483 @opindex mno-powerpc
10484 @opindex mpowerpc-gpopt
10485 @opindex mno-powerpc-gpopt
10486 @opindex mpowerpc-gfxopt
10487 @opindex mno-powerpc-gfxopt
10488 @opindex mpowerpc64
10489 @opindex mno-powerpc64
10490 GCC supports two related instruction set architectures for the
10491 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
10492 instructions supported by the @samp{rios} chip set used in the original
10493 RS/6000 systems and the @dfn{PowerPC} instruction set is the
10494 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
10495 the IBM 4xx microprocessors.
10497 Neither architecture is a subset of the other. However there is a
10498 large common subset of instructions supported by both. An MQ
10499 register is included in processors supporting the POWER architecture.
10501 You use these options to specify which instructions are available on the
10502 processor you are using. The default value of these options is
10503 determined when configuring GCC@. Specifying the
10504 @option{-mcpu=@var{cpu_type}} overrides the specification of these
10505 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
10506 rather than the options listed above.
10508 The @option{-mpower} option allows GCC to generate instructions that
10509 are found only in the POWER architecture and to use the MQ register.
10510 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
10511 to generate instructions that are present in the POWER2 architecture but
10512 not the original POWER architecture.
10514 The @option{-mpowerpc} option allows GCC to generate instructions that
10515 are found only in the 32-bit subset of the PowerPC architecture.
10516 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
10517 GCC to use the optional PowerPC architecture instructions in the
10518 General Purpose group, including floating-point square root. Specifying
10519 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
10520 use the optional PowerPC architecture instructions in the Graphics
10521 group, including floating-point select.
10523 The @option{-mpowerpc64} option allows GCC to generate the additional
10524 64-bit instructions that are found in the full PowerPC64 architecture
10525 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
10526 @option{-mno-powerpc64}.
10528 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
10529 will use only the instructions in the common subset of both
10530 architectures plus some special AIX common-mode calls, and will not use
10531 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
10532 permits GCC to use any instruction from either architecture and to
10533 allow use of the MQ register; specify this for the Motorola MPC601.
10535 @item -mnew-mnemonics
10536 @itemx -mold-mnemonics
10537 @opindex mnew-mnemonics
10538 @opindex mold-mnemonics
10539 Select which mnemonics to use in the generated assembler code. With
10540 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
10541 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
10542 assembler mnemonics defined for the POWER architecture. Instructions
10543 defined in only one architecture have only one mnemonic; GCC uses that
10544 mnemonic irrespective of which of these options is specified.
10546 GCC defaults to the mnemonics appropriate for the architecture in
10547 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
10548 value of these option. Unless you are building a cross-compiler, you
10549 should normally not specify either @option{-mnew-mnemonics} or
10550 @option{-mold-mnemonics}, but should instead accept the default.
10552 @item -mcpu=@var{cpu_type}
10554 Set architecture type, register usage, choice of mnemonics, and
10555 instruction scheduling parameters for machine type @var{cpu_type}.
10556 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
10557 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
10558 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
10559 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
10560 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
10561 @samp{860}, @samp{970}, @samp{8540}, @samp{common}, @samp{ec603e}, @samp{G3},
10562 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
10563 @samp{power4}, @samp{power5}, @samp{powerpc}, @samp{powerpc64},
10564 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64a}.
10566 @option{-mcpu=common} selects a completely generic processor. Code
10567 generated under this option will run on any POWER or PowerPC processor.
10568 GCC will use only the instructions in the common subset of both
10569 architectures, and will not use the MQ register. GCC assumes a generic
10570 processor model for scheduling purposes.
10572 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
10573 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
10574 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
10575 types, with an appropriate, generic processor model assumed for
10576 scheduling purposes.
10578 The other options specify a specific processor. Code generated under
10579 those options will run best on that processor, and may not run at all on
10582 The @option{-mcpu} options automatically enable or disable the
10583 following options: @option{-maltivec}, @option{-mhard-float},
10584 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
10585 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
10586 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
10587 @option{-mstring}. The particular options set for any particular CPU
10588 will vary between compiler versions, depending on what setting seems
10589 to produce optimal code for that CPU; it doesn't necessarily reflect
10590 the actual hardware's capabilities. If you wish to set an individual
10591 option to a particular value, you may specify it after the
10592 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
10594 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
10595 not enabled or disabled by the @option{-mcpu} option at present, since
10596 AIX does not have full support for these options. You may still
10597 enable or disable them individually if you're sure it'll work in your
10600 @item -mtune=@var{cpu_type}
10602 Set the instruction scheduling parameters for machine type
10603 @var{cpu_type}, but do not set the architecture type, register usage, or
10604 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
10605 values for @var{cpu_type} are used for @option{-mtune} as for
10606 @option{-mcpu}. If both are specified, the code generated will use the
10607 architecture, registers, and mnemonics set by @option{-mcpu}, but the
10608 scheduling parameters set by @option{-mtune}.
10611 @itemx -mno-altivec
10613 @opindex mno-altivec
10614 Generate code that uses (does not use) AltiVec instructions, and also
10615 enable the use of built-in functions that allow more direct access to
10616 the AltiVec instruction set. You may also need to set
10617 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
10622 Extend the current ABI with SPE ABI extensions. This does not change
10623 the default ABI, instead it adds the SPE ABI extensions to the current
10627 @opindex mabi=no-spe
10628 Disable Booke SPE ABI extensions for the current ABI@.
10630 @item -misel=@var{yes/no}
10633 This switch enables or disables the generation of ISEL instructions.
10635 @item -mspe=@var{yes/no}
10638 This switch enables or disables the generation of SPE simd
10641 @item -mfloat-gprs=@var{yes/single/double/no}
10642 @itemx -mfloat-gprs
10643 @opindex mfloat-gprs
10644 This switch enables or disables the generation of floating point
10645 operations on the general purpose registers for architectures that
10648 The argument @var{yes} or @var{single} enables the use of
10649 single-precision floating point operations.
10651 The argument @var{double} enables the use of single and
10652 double-precision floating point operations.
10654 The argument @var{no} disables floating point operations on the
10655 general purpose registers.
10657 This option is currently only available on the MPC854x.
10663 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
10664 targets (including GNU/Linux). The 32-bit environment sets int, long
10665 and pointer to 32 bits and generates code that runs on any PowerPC
10666 variant. The 64-bit environment sets int to 32 bits and long and
10667 pointer to 64 bits, and generates code for PowerPC64, as for
10668 @option{-mpowerpc64}.
10671 @itemx -mno-fp-in-toc
10672 @itemx -mno-sum-in-toc
10673 @itemx -mminimal-toc
10675 @opindex mno-fp-in-toc
10676 @opindex mno-sum-in-toc
10677 @opindex mminimal-toc
10678 Modify generation of the TOC (Table Of Contents), which is created for
10679 every executable file. The @option{-mfull-toc} option is selected by
10680 default. In that case, GCC will allocate at least one TOC entry for
10681 each unique non-automatic variable reference in your program. GCC
10682 will also place floating-point constants in the TOC@. However, only
10683 16,384 entries are available in the TOC@.
10685 If you receive a linker error message that saying you have overflowed
10686 the available TOC space, you can reduce the amount of TOC space used
10687 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
10688 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
10689 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
10690 generate code to calculate the sum of an address and a constant at
10691 run-time instead of putting that sum into the TOC@. You may specify one
10692 or both of these options. Each causes GCC to produce very slightly
10693 slower and larger code at the expense of conserving TOC space.
10695 If you still run out of space in the TOC even when you specify both of
10696 these options, specify @option{-mminimal-toc} instead. This option causes
10697 GCC to make only one TOC entry for every file. When you specify this
10698 option, GCC will produce code that is slower and larger but which
10699 uses extremely little TOC space. You may wish to use this option
10700 only on files that contain less frequently executed code.
10706 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
10707 @code{long} type, and the infrastructure needed to support them.
10708 Specifying @option{-maix64} implies @option{-mpowerpc64} and
10709 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
10710 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
10713 @itemx -mno-xl-compat
10714 @opindex mxl-compat
10715 @opindex mno-xl-compat
10716 Produce code that conforms more closely to IBM XLC semantics when using
10717 AIX-compatible ABI. Pass floating-point arguments to prototyped
10718 functions beyond the register save area (RSA) on the stack in addition
10719 to argument FPRs. Do not assume that most significant double in 128
10720 bit long double value is properly rounded when comparing values.
10722 The AIX calling convention was extended but not initially documented to
10723 handle an obscure K&R C case of calling a function that takes the
10724 address of its arguments with fewer arguments than declared. AIX XL
10725 compilers access floating point arguments which do not fit in the
10726 RSA from the stack when a subroutine is compiled without
10727 optimization. Because always storing floating-point arguments on the
10728 stack is inefficient and rarely needed, this option is not enabled by
10729 default and only is necessary when calling subroutines compiled by AIX
10730 XL compilers without optimization.
10734 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
10735 application written to use message passing with special startup code to
10736 enable the application to run. The system must have PE installed in the
10737 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
10738 must be overridden with the @option{-specs=} option to specify the
10739 appropriate directory location. The Parallel Environment does not
10740 support threads, so the @option{-mpe} option and the @option{-pthread}
10741 option are incompatible.
10743 @item -malign-natural
10744 @itemx -malign-power
10745 @opindex malign-natural
10746 @opindex malign-power
10747 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
10748 @option{-malign-natural} overrides the ABI-defined alignment of larger
10749 types, such as floating-point doubles, on their natural size-based boundary.
10750 The option @option{-malign-power} instructs GCC to follow the ABI-specified
10751 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
10753 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
10757 @itemx -mhard-float
10758 @opindex msoft-float
10759 @opindex mhard-float
10760 Generate code that does not use (uses) the floating-point register set.
10761 Software floating point emulation is provided if you use the
10762 @option{-msoft-float} option, and pass the option to GCC when linking.
10765 @itemx -mno-multiple
10767 @opindex mno-multiple
10768 Generate code that uses (does not use) the load multiple word
10769 instructions and the store multiple word instructions. These
10770 instructions are generated by default on POWER systems, and not
10771 generated on PowerPC systems. Do not use @option{-mmultiple} on little
10772 endian PowerPC systems, since those instructions do not work when the
10773 processor is in little endian mode. The exceptions are PPC740 and
10774 PPC750 which permit the instructions usage in little endian mode.
10779 @opindex mno-string
10780 Generate code that uses (does not use) the load string instructions
10781 and the store string word instructions to save multiple registers and
10782 do small block moves. These instructions are generated by default on
10783 POWER systems, and not generated on PowerPC systems. Do not use
10784 @option{-mstring} on little endian PowerPC systems, since those
10785 instructions do not work when the processor is in little endian mode.
10786 The exceptions are PPC740 and PPC750 which permit the instructions
10787 usage in little endian mode.
10792 @opindex mno-update
10793 Generate code that uses (does not use) the load or store instructions
10794 that update the base register to the address of the calculated memory
10795 location. These instructions are generated by default. If you use
10796 @option{-mno-update}, there is a small window between the time that the
10797 stack pointer is updated and the address of the previous frame is
10798 stored, which means code that walks the stack frame across interrupts or
10799 signals may get corrupted data.
10802 @itemx -mno-fused-madd
10803 @opindex mfused-madd
10804 @opindex mno-fused-madd
10805 Generate code that uses (does not use) the floating point multiply and
10806 accumulate instructions. These instructions are generated by default if
10807 hardware floating is used.
10809 @item -mno-bit-align
10811 @opindex mno-bit-align
10812 @opindex mbit-align
10813 On System V.4 and embedded PowerPC systems do not (do) force structures
10814 and unions that contain bit-fields to be aligned to the base type of the
10817 For example, by default a structure containing nothing but 8
10818 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
10819 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
10820 the structure would be aligned to a 1 byte boundary and be one byte in
10823 @item -mno-strict-align
10824 @itemx -mstrict-align
10825 @opindex mno-strict-align
10826 @opindex mstrict-align
10827 On System V.4 and embedded PowerPC systems do not (do) assume that
10828 unaligned memory references will be handled by the system.
10830 @item -mrelocatable
10831 @itemx -mno-relocatable
10832 @opindex mrelocatable
10833 @opindex mno-relocatable
10834 On embedded PowerPC systems generate code that allows (does not allow)
10835 the program to be relocated to a different address at runtime. If you
10836 use @option{-mrelocatable} on any module, all objects linked together must
10837 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
10839 @item -mrelocatable-lib
10840 @itemx -mno-relocatable-lib
10841 @opindex mrelocatable-lib
10842 @opindex mno-relocatable-lib
10843 On embedded PowerPC systems generate code that allows (does not allow)
10844 the program to be relocated to a different address at runtime. Modules
10845 compiled with @option{-mrelocatable-lib} can be linked with either modules
10846 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
10847 with modules compiled with the @option{-mrelocatable} options.
10853 On System V.4 and embedded PowerPC systems do not (do) assume that
10854 register 2 contains a pointer to a global area pointing to the addresses
10855 used in the program.
10858 @itemx -mlittle-endian
10860 @opindex mlittle-endian
10861 On System V.4 and embedded PowerPC systems compile code for the
10862 processor in little endian mode. The @option{-mlittle-endian} option is
10863 the same as @option{-mlittle}.
10866 @itemx -mbig-endian
10868 @opindex mbig-endian
10869 On System V.4 and embedded PowerPC systems compile code for the
10870 processor in big endian mode. The @option{-mbig-endian} option is
10871 the same as @option{-mbig}.
10873 @item -mdynamic-no-pic
10874 @opindex mdynamic-no-pic
10875 On Darwin and Mac OS X systems, compile code so that it is not
10876 relocatable, but that its external references are relocatable. The
10877 resulting code is suitable for applications, but not shared
10880 @item -mprioritize-restricted-insns=@var{priority}
10881 @opindex mprioritize-restricted-insns
10882 This option controls the priority that is assigned to
10883 dispatch-slot restricted instructions during the second scheduling
10884 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
10885 @var{no/highest/second-highest} priority to dispatch slot restricted
10888 @item -msched-costly-dep=@var{dependence_type}
10889 @opindex msched-costly-dep
10890 This option controls which dependences are considered costly
10891 by the target during instruction scheduling. The argument
10892 @var{dependence_type} takes one of the following values:
10893 @var{no}: no dependence is costly,
10894 @var{all}: all dependences are costly,
10895 @var{true_store_to_load}: a true dependence from store to load is costly,
10896 @var{store_to_load}: any dependence from store to load is costly,
10897 @var{number}: any dependence which latency >= @var{number} is costly.
10899 @item -minsert-sched-nops=@var{scheme}
10900 @opindex minsert-sched-nops
10901 This option controls which nop insertion scheme will be used during
10902 the second scheduling pass. The argument @var{scheme} takes one of the
10904 @var{no}: Don't insert nops.
10905 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
10906 according to the scheduler's grouping.
10907 @var{regroup_exact}: Insert nops to force costly dependent insns into
10908 separate groups. Insert exactly as many nops as needed to force an insn
10909 to a new group, according to the estimated processor grouping.
10910 @var{number}: Insert nops to force costly dependent insns into
10911 separate groups. Insert @var{number} nops to force an insn to a new group.
10914 @opindex mcall-sysv
10915 On System V.4 and embedded PowerPC systems compile code using calling
10916 conventions that adheres to the March 1995 draft of the System V
10917 Application Binary Interface, PowerPC processor supplement. This is the
10918 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
10920 @item -mcall-sysv-eabi
10921 @opindex mcall-sysv-eabi
10922 Specify both @option{-mcall-sysv} and @option{-meabi} options.
10924 @item -mcall-sysv-noeabi
10925 @opindex mcall-sysv-noeabi
10926 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
10928 @item -mcall-solaris
10929 @opindex mcall-solaris
10930 On System V.4 and embedded PowerPC systems compile code for the Solaris
10934 @opindex mcall-linux
10935 On System V.4 and embedded PowerPC systems compile code for the
10936 Linux-based GNU system.
10940 On System V.4 and embedded PowerPC systems compile code for the
10941 Hurd-based GNU system.
10943 @item -mcall-netbsd
10944 @opindex mcall-netbsd
10945 On System V.4 and embedded PowerPC systems compile code for the
10946 NetBSD operating system.
10948 @item -maix-struct-return
10949 @opindex maix-struct-return
10950 Return all structures in memory (as specified by the AIX ABI)@.
10952 @item -msvr4-struct-return
10953 @opindex msvr4-struct-return
10954 Return structures smaller than 8 bytes in registers (as specified by the
10957 @item -mabi=altivec
10958 @opindex mabi=altivec
10959 Extend the current ABI with AltiVec ABI extensions. This does not
10960 change the default ABI, instead it adds the AltiVec ABI extensions to
10963 @item -mabi=no-altivec
10964 @opindex mabi=no-altivec
10965 Disable AltiVec ABI extensions for the current ABI@.
10968 @itemx -mno-prototype
10969 @opindex mprototype
10970 @opindex mno-prototype
10971 On System V.4 and embedded PowerPC systems assume that all calls to
10972 variable argument functions are properly prototyped. Otherwise, the
10973 compiler must insert an instruction before every non prototyped call to
10974 set or clear bit 6 of the condition code register (@var{CR}) to
10975 indicate whether floating point values were passed in the floating point
10976 registers in case the function takes a variable arguments. With
10977 @option{-mprototype}, only calls to prototyped variable argument functions
10978 will set or clear the bit.
10982 On embedded PowerPC systems, assume that the startup module is called
10983 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
10984 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
10989 On embedded PowerPC systems, assume that the startup module is called
10990 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
10995 On embedded PowerPC systems, assume that the startup module is called
10996 @file{crt0.o} and the standard C libraries are @file{libads.a} and
10999 @item -myellowknife
11000 @opindex myellowknife
11001 On embedded PowerPC systems, assume that the startup module is called
11002 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
11007 On System V.4 and embedded PowerPC systems, specify that you are
11008 compiling for a VxWorks system.
11012 Specify that you are compiling for the WindISS simulation environment.
11016 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
11017 header to indicate that @samp{eabi} extended relocations are used.
11023 On System V.4 and embedded PowerPC systems do (do not) adhere to the
11024 Embedded Applications Binary Interface (eabi) which is a set of
11025 modifications to the System V.4 specifications. Selecting @option{-meabi}
11026 means that the stack is aligned to an 8 byte boundary, a function
11027 @code{__eabi} is called to from @code{main} to set up the eabi
11028 environment, and the @option{-msdata} option can use both @code{r2} and
11029 @code{r13} to point to two separate small data areas. Selecting
11030 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
11031 do not call an initialization function from @code{main}, and the
11032 @option{-msdata} option will only use @code{r13} to point to a single
11033 small data area. The @option{-meabi} option is on by default if you
11034 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
11037 @opindex msdata=eabi
11038 On System V.4 and embedded PowerPC systems, put small initialized
11039 @code{const} global and static data in the @samp{.sdata2} section, which
11040 is pointed to by register @code{r2}. Put small initialized
11041 non-@code{const} global and static data in the @samp{.sdata} section,
11042 which is pointed to by register @code{r13}. Put small uninitialized
11043 global and static data in the @samp{.sbss} section, which is adjacent to
11044 the @samp{.sdata} section. The @option{-msdata=eabi} option is
11045 incompatible with the @option{-mrelocatable} option. The
11046 @option{-msdata=eabi} option also sets the @option{-memb} option.
11049 @opindex msdata=sysv
11050 On System V.4 and embedded PowerPC systems, put small global and static
11051 data in the @samp{.sdata} section, which is pointed to by register
11052 @code{r13}. Put small uninitialized global and static data in the
11053 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11054 The @option{-msdata=sysv} option is incompatible with the
11055 @option{-mrelocatable} option.
11057 @item -msdata=default
11059 @opindex msdata=default
11061 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
11062 compile code the same as @option{-msdata=eabi}, otherwise compile code the
11063 same as @option{-msdata=sysv}.
11066 @opindex msdata-data
11067 On System V.4 and embedded PowerPC systems, put small global and static
11068 data in the @samp{.sdata} section. Put small uninitialized global and
11069 static data in the @samp{.sbss} section. Do not use register @code{r13}
11070 to address small data however. This is the default behavior unless
11071 other @option{-msdata} options are used.
11075 @opindex msdata=none
11077 On embedded PowerPC systems, put all initialized global and static data
11078 in the @samp{.data} section, and all uninitialized data in the
11079 @samp{.bss} section.
11083 @cindex smaller data references (PowerPC)
11084 @cindex .sdata/.sdata2 references (PowerPC)
11085 On embedded PowerPC systems, put global and static items less than or
11086 equal to @var{num} bytes into the small data or bss sections instead of
11087 the normal data or bss section. By default, @var{num} is 8. The
11088 @option{-G @var{num}} switch is also passed to the linker.
11089 All modules should be compiled with the same @option{-G @var{num}} value.
11092 @itemx -mno-regnames
11094 @opindex mno-regnames
11095 On System V.4 and embedded PowerPC systems do (do not) emit register
11096 names in the assembly language output using symbolic forms.
11099 @itemx -mno-longcall
11101 @opindex mno-longcall
11102 Default to making all function calls indirectly, using a register, so
11103 that functions which reside further than 32 megabytes (33,554,432
11104 bytes) from the current location can be called. This setting can be
11105 overridden by the @code{shortcall} function attribute, or by
11106 @code{#pragma longcall(0)}.
11108 Some linkers are capable of detecting out-of-range calls and generating
11109 glue code on the fly. On these systems, long calls are unnecessary and
11110 generate slower code. As of this writing, the AIX linker can do this,
11111 as can the GNU linker for PowerPC/64. It is planned to add this feature
11112 to the GNU linker for 32-bit PowerPC systems as well.
11114 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11115 callee, L42'', plus a ``branch island'' (glue code). The two target
11116 addresses represent the callee and the ``branch island''. The
11117 Darwin/PPC linker will prefer the first address and generate a ``bl
11118 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11119 otherwise, the linker will generate ``bl L42'' to call the ``branch
11120 island''. The ``branch island'' is appended to the body of the
11121 calling function; it computes the full 32-bit address of the callee
11124 On Mach-O (Darwin) systems, this option directs the compiler emit to
11125 the glue for every direct call, and the Darwin linker decides whether
11126 to use or discard it.
11128 In the future, we may cause GCC to ignore all longcall specifications
11129 when the linker is known to generate glue.
11133 Adds support for multithreading with the @dfn{pthreads} library.
11134 This option sets flags for both the preprocessor and linker.
11138 @node S/390 and zSeries Options
11139 @subsection S/390 and zSeries Options
11140 @cindex S/390 and zSeries Options
11142 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11146 @itemx -msoft-float
11147 @opindex mhard-float
11148 @opindex msoft-float
11149 Use (do not use) the hardware floating-point instructions and registers
11150 for floating-point operations. When @option{-msoft-float} is specified,
11151 functions in @file{libgcc.a} will be used to perform floating-point
11152 operations. When @option{-mhard-float} is specified, the compiler
11153 generates IEEE floating-point instructions. This is the default.
11156 @itemx -mno-backchain
11157 @opindex mbackchain
11158 @opindex mno-backchain
11159 Store (do not store) the address of the caller's frame as backchain pointer
11160 into the callee's stack frame.
11161 A backchain may be needed to allow debugging using tools that do not understand
11162 DWARF-2 call frame information.
11163 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
11164 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
11165 the backchain is placed into the topmost word of the 96/160 byte register
11168 In general, code compiled with @option{-mbackchain} is call-compatible with
11169 code compiled with @option{-mmo-backchain}; however, use of the backchain
11170 for debugging purposes usually requires that the whole binary is built with
11171 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
11172 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11173 to build a linux kernel use @option{-msoft-float}.
11175 The default is to not maintain the backchain.
11177 @item -mpacked-stack
11178 @item -mno-packed-stack
11179 @opindex mpacked-stack
11180 @opindex mno-packed-stack
11181 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
11182 specified, the compiler uses the all fields of the 96/160 byte register save
11183 area only for their default purpose; unused fields still take up stack space.
11184 When @option{-mpacked-stack} is specified, register save slots are densely
11185 packed at the top of the register save area; unused space is reused for other
11186 purposes, allowing for more efficient use of the available stack space.
11187 However, when @option{-mbackchain} is also in effect, the topmost word of
11188 the save area is always used to store the backchain, and the return address
11189 register is always saved two words below the backchain.
11191 As long as the stack frame backchain is not used, code generated with
11192 @option{-mpacked-stack} is call-compatible with code generated with
11193 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
11194 S/390 or zSeries generated code that uses the stack frame backchain at run
11195 time, not just for debugging purposes. Such code is not call-compatible
11196 with code compiled with @option{-mpacked-stack}. Also, note that the
11197 combination of @option{-mbackchain},
11198 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11199 to build a linux kernel use @option{-msoft-float}.
11201 The default is to not use the packed stack layout.
11204 @itemx -mno-small-exec
11205 @opindex msmall-exec
11206 @opindex mno-small-exec
11207 Generate (or do not generate) code using the @code{bras} instruction
11208 to do subroutine calls.
11209 This only works reliably if the total executable size does not
11210 exceed 64k. The default is to use the @code{basr} instruction instead,
11211 which does not have this limitation.
11217 When @option{-m31} is specified, generate code compliant to the
11218 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
11219 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
11220 particular to generate 64-bit instructions. For the @samp{s390}
11221 targets, the default is @option{-m31}, while the @samp{s390x}
11222 targets default to @option{-m64}.
11228 When @option{-mzarch} is specified, generate code using the
11229 instructions available on z/Architecture.
11230 When @option{-mesa} is specified, generate code using the
11231 instructions available on ESA/390. Note that @option{-mesa} is
11232 not possible with @option{-m64}.
11233 When generating code compliant to the GNU/Linux for S/390 ABI,
11234 the default is @option{-mesa}. When generating code compliant
11235 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
11241 Generate (or do not generate) code using the @code{mvcle} instruction
11242 to perform block moves. When @option{-mno-mvcle} is specified,
11243 use a @code{mvc} loop instead. This is the default.
11249 Print (or do not print) additional debug information when compiling.
11250 The default is to not print debug information.
11252 @item -march=@var{cpu-type}
11254 Generate code that will run on @var{cpu-type}, which is the name of a system
11255 representing a certain processor type. Possible values for
11256 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
11257 When generating code using the instructions available on z/Architecture,
11258 the default is @option{-march=z900}. Otherwise, the default is
11259 @option{-march=g5}.
11261 @item -mtune=@var{cpu-type}
11263 Tune to @var{cpu-type} everything applicable about the generated code,
11264 except for the ABI and the set of available instructions.
11265 The list of @var{cpu-type} values is the same as for @option{-march}.
11266 The default is the value used for @option{-march}.
11269 @itemx -mno-tpf-trace
11270 @opindex mtpf-trace
11271 @opindex mno-tpf-trace
11272 Generate code that adds (does not add) in TPF OS specific branches to trace
11273 routines in the operating system. This option is off by default, even
11274 when compiling for the TPF OS@.
11277 @itemx -mno-fused-madd
11278 @opindex mfused-madd
11279 @opindex mno-fused-madd
11280 Generate code that uses (does not use) the floating point multiply and
11281 accumulate instructions. These instructions are generated by default if
11282 hardware floating point is used.
11284 @item -mwarn-framesize=@var{framesize}
11285 @opindex mwarn-framesize
11286 Emit a warning if the current function exceeds the given frame size. Because
11287 this is a compile time check it doesn't need to be a real problem when the program
11288 runs. It is intended to identify functions which most probably cause
11289 a stack overflow. It is useful to be used in an environment with limited stack
11290 size e.g.@: the linux kernel.
11292 @item -mwarn-dynamicstack
11293 @opindex mwarn-dynamicstack
11294 Emit a warning if the function calls alloca or uses dynamically
11295 sized arrays. This is generally a bad idea with a limited stack size.
11297 @item -mstack-guard=@var{stack-guard}
11298 @item -mstack-size=@var{stack-size}
11299 @opindex mstack-guard
11300 @opindex mstack-size
11301 These arguments always have to be used in conjunction. If they are present the s390
11302 back end emits additional instructions in the function prologue which trigger a trap
11303 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
11304 (remember that the stack on s390 grows downward). These options are intended to
11305 be used to help debugging stack overflow problems. The additionally emitted code
11306 cause only little overhead and hence can also be used in production like systems
11307 without greater performance degradation. The given values have to be exact
11308 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard}.
11309 In order to be efficient the extra code makes the assumption that the stack starts
11310 at an address aligned to the value given by @var{stack-size}.
11314 @subsection SH Options
11316 These @samp{-m} options are defined for the SH implementations:
11321 Generate code for the SH1.
11325 Generate code for the SH2.
11328 Generate code for the SH2e.
11332 Generate code for the SH3.
11336 Generate code for the SH3e.
11340 Generate code for the SH4 without a floating-point unit.
11342 @item -m4-single-only
11343 @opindex m4-single-only
11344 Generate code for the SH4 with a floating-point unit that only
11345 supports single-precision arithmetic.
11349 Generate code for the SH4 assuming the floating-point unit is in
11350 single-precision mode by default.
11354 Generate code for the SH4.
11358 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
11359 floating-point unit is not used.
11361 @item -m4a-single-only
11362 @opindex m4a-single-only
11363 Generate code for the SH4a, in such a way that no double-precision
11364 floating point operations are used.
11367 @opindex m4a-single
11368 Generate code for the SH4a assuming the floating-point unit is in
11369 single-precision mode by default.
11373 Generate code for the SH4a.
11377 Same as @option{-m4a-nofpu}, except that it implicitly passes
11378 @option{-dsp} to the assembler. GCC doesn't generate any DSP
11379 instructions at the moment.
11383 Compile code for the processor in big endian mode.
11387 Compile code for the processor in little endian mode.
11391 Align doubles at 64-bit boundaries. Note that this changes the calling
11392 conventions, and thus some functions from the standard C library will
11393 not work unless you recompile it first with @option{-mdalign}.
11397 Shorten some address references at link time, when possible; uses the
11398 linker option @option{-relax}.
11402 Use 32-bit offsets in @code{switch} tables. The default is to use
11407 Enable the use of the instruction @code{fmovd}.
11411 Comply with the calling conventions defined by Renesas.
11415 Comply with the calling conventions defined by Renesas.
11419 Comply with the calling conventions defined for GCC before the Renesas
11420 conventions were available. This option is the default for all
11421 targets of the SH toolchain except for @samp{sh-symbianelf}.
11424 @opindex mnomacsave
11425 Mark the @code{MAC} register as call-clobbered, even if
11426 @option{-mhitachi} is given.
11430 Increase IEEE-compliance of floating-point code.
11434 Dump instruction size and location in the assembly code.
11437 @opindex mpadstruct
11438 This option is deprecated. It pads structures to multiple of 4 bytes,
11439 which is incompatible with the SH ABI@.
11443 Optimize for space instead of speed. Implied by @option{-Os}.
11446 @opindex mprefergot
11447 When generating position-independent code, emit function calls using
11448 the Global Offset Table instead of the Procedure Linkage Table.
11452 Generate a library function call to invalidate instruction cache
11453 entries, after fixing up a trampoline. This library function call
11454 doesn't assume it can write to the whole memory address space. This
11455 is the default when the target is @code{sh-*-linux*}.
11458 @node SPARC Options
11459 @subsection SPARC Options
11460 @cindex SPARC options
11462 These @samp{-m} options are supported on the SPARC:
11465 @item -mno-app-regs
11467 @opindex mno-app-regs
11469 Specify @option{-mapp-regs} to generate output using the global registers
11470 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
11473 To be fully SVR4 ABI compliant at the cost of some performance loss,
11474 specify @option{-mno-app-regs}. You should compile libraries and system
11475 software with this option.
11478 @itemx -mhard-float
11480 @opindex mhard-float
11481 Generate output containing floating point instructions. This is the
11485 @itemx -msoft-float
11487 @opindex msoft-float
11488 Generate output containing library calls for floating point.
11489 @strong{Warning:} the requisite libraries are not available for all SPARC
11490 targets. Normally the facilities of the machine's usual C compiler are
11491 used, but this cannot be done directly in cross-compilation. You must make
11492 your own arrangements to provide suitable library functions for
11493 cross-compilation. The embedded targets @samp{sparc-*-aout} and
11494 @samp{sparclite-*-*} do provide software floating point support.
11496 @option{-msoft-float} changes the calling convention in the output file;
11497 therefore, it is only useful if you compile @emph{all} of a program with
11498 this option. In particular, you need to compile @file{libgcc.a}, the
11499 library that comes with GCC, with @option{-msoft-float} in order for
11502 @item -mhard-quad-float
11503 @opindex mhard-quad-float
11504 Generate output containing quad-word (long double) floating point
11507 @item -msoft-quad-float
11508 @opindex msoft-quad-float
11509 Generate output containing library calls for quad-word (long double)
11510 floating point instructions. The functions called are those specified
11511 in the SPARC ABI@. This is the default.
11513 As of this writing, there are no SPARC implementations that have hardware
11514 support for the quad-word floating point instructions. They all invoke
11515 a trap handler for one of these instructions, and then the trap handler
11516 emulates the effect of the instruction. Because of the trap handler overhead,
11517 this is much slower than calling the ABI library routines. Thus the
11518 @option{-msoft-quad-float} option is the default.
11520 @item -mno-unaligned-doubles
11521 @itemx -munaligned-doubles
11522 @opindex mno-unaligned-doubles
11523 @opindex munaligned-doubles
11524 Assume that doubles have 8 byte alignment. This is the default.
11526 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
11527 alignment only if they are contained in another type, or if they have an
11528 absolute address. Otherwise, it assumes they have 4 byte alignment.
11529 Specifying this option avoids some rare compatibility problems with code
11530 generated by other compilers. It is not the default because it results
11531 in a performance loss, especially for floating point code.
11533 @item -mno-faster-structs
11534 @itemx -mfaster-structs
11535 @opindex mno-faster-structs
11536 @opindex mfaster-structs
11537 With @option{-mfaster-structs}, the compiler assumes that structures
11538 should have 8 byte alignment. This enables the use of pairs of
11539 @code{ldd} and @code{std} instructions for copies in structure
11540 assignment, in place of twice as many @code{ld} and @code{st} pairs.
11541 However, the use of this changed alignment directly violates the SPARC
11542 ABI@. Thus, it's intended only for use on targets where the developer
11543 acknowledges that their resulting code will not be directly in line with
11544 the rules of the ABI@.
11546 @item -mimpure-text
11547 @opindex mimpure-text
11548 @option{-mimpure-text}, used in addition to @option{-shared}, tells
11549 the compiler to not pass @option{-z text} to the linker when linking a
11550 shared object. Using this option, you can link position-dependent
11551 code into a shared object.
11553 @option{-mimpure-text} suppresses the ``relocations remain against
11554 allocatable but non-writable sections'' linker error message.
11555 However, the necessary relocations will trigger copy-on-write, and the
11556 shared object is not actually shared across processes. Instead of
11557 using @option{-mimpure-text}, you should compile all source code with
11558 @option{-fpic} or @option{-fPIC}.
11560 This option is only available on SunOS and Solaris.
11562 @item -mcpu=@var{cpu_type}
11564 Set the instruction set, register set, and instruction scheduling parameters
11565 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
11566 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
11567 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
11568 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
11569 @samp{ultrasparc3}.
11571 Default instruction scheduling parameters are used for values that select
11572 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
11573 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
11575 Here is a list of each supported architecture and their supported
11580 v8: supersparc, hypersparc
11581 sparclite: f930, f934, sparclite86x
11583 v9: ultrasparc, ultrasparc3
11586 By default (unless configured otherwise), GCC generates code for the V7
11587 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
11588 additionally optimizes it for the Cypress CY7C602 chip, as used in the
11589 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
11590 SPARCStation 1, 2, IPX etc.
11592 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
11593 architecture. The only difference from V7 code is that the compiler emits
11594 the integer multiply and integer divide instructions which exist in SPARC-V8
11595 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
11596 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
11599 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
11600 the SPARC architecture. This adds the integer multiply, integer divide step
11601 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
11602 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
11603 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
11604 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
11605 MB86934 chip, which is the more recent SPARClite with FPU@.
11607 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
11608 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
11609 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
11610 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
11611 optimizes it for the TEMIC SPARClet chip.
11613 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
11614 architecture. This adds 64-bit integer and floating-point move instructions,
11615 3 additional floating-point condition code registers and conditional move
11616 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
11617 optimizes it for the Sun UltraSPARC I/II chips. With
11618 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
11619 Sun UltraSPARC III chip.
11621 @item -mtune=@var{cpu_type}
11623 Set the instruction scheduling parameters for machine type
11624 @var{cpu_type}, but do not set the instruction set or register set that the
11625 option @option{-mcpu=@var{cpu_type}} would.
11627 The same values for @option{-mcpu=@var{cpu_type}} can be used for
11628 @option{-mtune=@var{cpu_type}}, but the only useful values are those
11629 that select a particular cpu implementation. Those are @samp{cypress},
11630 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
11631 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
11632 @samp{ultrasparc3}.
11637 @opindex mno-v8plus
11638 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
11639 difference from the V8 ABI is that the global and out registers are
11640 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
11641 mode for all SPARC-V9 processors.
11647 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
11648 Visual Instruction Set extensions. The default is @option{-mno-vis}.
11651 These @samp{-m} options are supported in addition to the above
11652 on SPARC-V9 processors in 64-bit environments:
11655 @item -mlittle-endian
11656 @opindex mlittle-endian
11657 Generate code for a processor running in little-endian mode. It is only
11658 available for a few configurations and most notably not on Solaris and Linux.
11664 Generate code for a 32-bit or 64-bit environment.
11665 The 32-bit environment sets int, long and pointer to 32 bits.
11666 The 64-bit environment sets int to 32 bits and long and pointer
11669 @item -mcmodel=medlow
11670 @opindex mcmodel=medlow
11671 Generate code for the Medium/Low code model: 64-bit addresses, programs
11672 must be linked in the low 32 bits of memory. Programs can be statically
11673 or dynamically linked.
11675 @item -mcmodel=medmid
11676 @opindex mcmodel=medmid
11677 Generate code for the Medium/Middle code model: 64-bit addresses, programs
11678 must be linked in the low 44 bits of memory, the text and data segments must
11679 be less than 2GB in size and the data segment must be located within 2GB of
11682 @item -mcmodel=medany
11683 @opindex mcmodel=medany
11684 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
11685 may be linked anywhere in memory, the text and data segments must be less
11686 than 2GB in size and the data segment must be located within 2GB of the
11689 @item -mcmodel=embmedany
11690 @opindex mcmodel=embmedany
11691 Generate code for the Medium/Anywhere code model for embedded systems:
11692 64-bit addresses, the text and data segments must be less than 2GB in
11693 size, both starting anywhere in memory (determined at link time). The
11694 global register %g4 points to the base of the data segment. Programs
11695 are statically linked and PIC is not supported.
11698 @itemx -mno-stack-bias
11699 @opindex mstack-bias
11700 @opindex mno-stack-bias
11701 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
11702 frame pointer if present, are offset by @minus{}2047 which must be added back
11703 when making stack frame references. This is the default in 64-bit mode.
11704 Otherwise, assume no such offset is present.
11707 These switches are supported in addition to the above on Solaris:
11712 Add support for multithreading using the Solaris threads library. This
11713 option sets flags for both the preprocessor and linker. This option does
11714 not affect the thread safety of object code produced by the compiler or
11715 that of libraries supplied with it.
11719 Add support for multithreading using the POSIX threads library. This
11720 option sets flags for both the preprocessor and linker. This option does
11721 not affect the thread safety of object code produced by the compiler or
11722 that of libraries supplied with it.
11725 @node System V Options
11726 @subsection Options for System V
11728 These additional options are available on System V Release 4 for
11729 compatibility with other compilers on those systems:
11734 Create a shared object.
11735 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
11739 Identify the versions of each tool used by the compiler, in a
11740 @code{.ident} assembler directive in the output.
11744 Refrain from adding @code{.ident} directives to the output file (this is
11747 @item -YP,@var{dirs}
11749 Search the directories @var{dirs}, and no others, for libraries
11750 specified with @option{-l}.
11752 @item -Ym,@var{dir}
11754 Look in the directory @var{dir} to find the M4 preprocessor.
11755 The assembler uses this option.
11756 @c This is supposed to go with a -Yd for predefined M4 macro files, but
11757 @c the generic assembler that comes with Solaris takes just -Ym.
11760 @node TMS320C3x/C4x Options
11761 @subsection TMS320C3x/C4x Options
11762 @cindex TMS320C3x/C4x Options
11764 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
11768 @item -mcpu=@var{cpu_type}
11770 Set the instruction set, register set, and instruction scheduling
11771 parameters for machine type @var{cpu_type}. Supported values for
11772 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
11773 @samp{c44}. The default is @samp{c40} to generate code for the
11778 @itemx -msmall-memory
11780 @opindex mbig-memory
11782 @opindex msmall-memory
11784 Generates code for the big or small memory model. The small memory
11785 model assumed that all data fits into one 64K word page. At run-time
11786 the data page (DP) register must be set to point to the 64K page
11787 containing the .bss and .data program sections. The big memory model is
11788 the default and requires reloading of the DP register for every direct
11795 Allow (disallow) allocation of general integer operands into the block
11796 count register BK@.
11802 Enable (disable) generation of code using decrement and branch,
11803 DBcond(D), instructions. This is enabled by default for the C4x. To be
11804 on the safe side, this is disabled for the C3x, since the maximum
11805 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
11806 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
11807 that it can utilize the decrement and branch instruction, but will give
11808 up if there is more than one memory reference in the loop. Thus a loop
11809 where the loop counter is decremented can generate slightly more
11810 efficient code, in cases where the RPTB instruction cannot be utilized.
11812 @item -mdp-isr-reload
11814 @opindex mdp-isr-reload
11816 Force the DP register to be saved on entry to an interrupt service
11817 routine (ISR), reloaded to point to the data section, and restored on
11818 exit from the ISR@. This should not be required unless someone has
11819 violated the small memory model by modifying the DP register, say within
11826 For the C3x use the 24-bit MPYI instruction for integer multiplies
11827 instead of a library call to guarantee 32-bit results. Note that if one
11828 of the operands is a constant, then the multiplication will be performed
11829 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
11830 then squaring operations are performed inline instead of a library call.
11833 @itemx -mno-fast-fix
11835 @opindex mno-fast-fix
11836 The C3x/C4x FIX instruction to convert a floating point value to an
11837 integer value chooses the nearest integer less than or equal to the
11838 floating point value rather than to the nearest integer. Thus if the
11839 floating point number is negative, the result will be incorrectly
11840 truncated an additional code is necessary to detect and correct this
11841 case. This option can be used to disable generation of the additional
11842 code required to correct the result.
11848 Enable (disable) generation of repeat block sequences using the RPTB
11849 instruction for zero overhead looping. The RPTB construct is only used
11850 for innermost loops that do not call functions or jump across the loop
11851 boundaries. There is no advantage having nested RPTB loops due to the
11852 overhead required to save and restore the RC, RS, and RE registers.
11853 This is enabled by default with @option{-O2}.
11855 @item -mrpts=@var{count}
11859 Enable (disable) the use of the single instruction repeat instruction
11860 RPTS@. If a repeat block contains a single instruction, and the loop
11861 count can be guaranteed to be less than the value @var{count}, GCC will
11862 emit a RPTS instruction instead of a RPTB@. If no value is specified,
11863 then a RPTS will be emitted even if the loop count cannot be determined
11864 at compile time. Note that the repeated instruction following RPTS does
11865 not have to be reloaded from memory each iteration, thus freeing up the
11866 CPU buses for operands. However, since interrupts are blocked by this
11867 instruction, it is disabled by default.
11869 @item -mloop-unsigned
11870 @itemx -mno-loop-unsigned
11871 @opindex mloop-unsigned
11872 @opindex mno-loop-unsigned
11873 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
11874 is @math{2^{31} + 1} since these instructions test if the iteration count is
11875 negative to terminate the loop. If the iteration count is unsigned
11876 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
11877 exceeded. This switch allows an unsigned iteration count.
11881 Try to emit an assembler syntax that the TI assembler (asm30) is happy
11882 with. This also enforces compatibility with the API employed by the TI
11883 C3x C compiler. For example, long doubles are passed as structures
11884 rather than in floating point registers.
11890 Generate code that uses registers (stack) for passing arguments to functions.
11891 By default, arguments are passed in registers where possible rather
11892 than by pushing arguments on to the stack.
11894 @item -mparallel-insns
11895 @itemx -mno-parallel-insns
11896 @opindex mparallel-insns
11897 @opindex mno-parallel-insns
11898 Allow the generation of parallel instructions. This is enabled by
11899 default with @option{-O2}.
11901 @item -mparallel-mpy
11902 @itemx -mno-parallel-mpy
11903 @opindex mparallel-mpy
11904 @opindex mno-parallel-mpy
11905 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
11906 provided @option{-mparallel-insns} is also specified. These instructions have
11907 tight register constraints which can pessimize the code generation
11908 of large functions.
11913 @subsection V850 Options
11914 @cindex V850 Options
11916 These @samp{-m} options are defined for V850 implementations:
11920 @itemx -mno-long-calls
11921 @opindex mlong-calls
11922 @opindex mno-long-calls
11923 Treat all calls as being far away (near). If calls are assumed to be
11924 far away, the compiler will always load the functions address up into a
11925 register, and call indirect through the pointer.
11931 Do not optimize (do optimize) basic blocks that use the same index
11932 pointer 4 or more times to copy pointer into the @code{ep} register, and
11933 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
11934 option is on by default if you optimize.
11936 @item -mno-prolog-function
11937 @itemx -mprolog-function
11938 @opindex mno-prolog-function
11939 @opindex mprolog-function
11940 Do not use (do use) external functions to save and restore registers
11941 at the prologue and epilogue of a function. The external functions
11942 are slower, but use less code space if more than one function saves
11943 the same number of registers. The @option{-mprolog-function} option
11944 is on by default if you optimize.
11948 Try to make the code as small as possible. At present, this just turns
11949 on the @option{-mep} and @option{-mprolog-function} options.
11951 @item -mtda=@var{n}
11953 Put static or global variables whose size is @var{n} bytes or less into
11954 the tiny data area that register @code{ep} points to. The tiny data
11955 area can hold up to 256 bytes in total (128 bytes for byte references).
11957 @item -msda=@var{n}
11959 Put static or global variables whose size is @var{n} bytes or less into
11960 the small data area that register @code{gp} points to. The small data
11961 area can hold up to 64 kilobytes.
11963 @item -mzda=@var{n}
11965 Put static or global variables whose size is @var{n} bytes or less into
11966 the first 32 kilobytes of memory.
11970 Specify that the target processor is the V850.
11973 @opindex mbig-switch
11974 Generate code suitable for big switch tables. Use this option only if
11975 the assembler/linker complain about out of range branches within a switch
11980 This option will cause r2 and r5 to be used in the code generated by
11981 the compiler. This setting is the default.
11983 @item -mno-app-regs
11984 @opindex mno-app-regs
11985 This option will cause r2 and r5 to be treated as fixed registers.
11989 Specify that the target processor is the V850E1. The preprocessor
11990 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
11991 this option is used.
11995 Specify that the target processor is the V850E@. The preprocessor
11996 constant @samp{__v850e__} will be defined if this option is used.
11998 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
11999 are defined then a default target processor will be chosen and the
12000 relevant @samp{__v850*__} preprocessor constant will be defined.
12002 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
12003 defined, regardless of which processor variant is the target.
12005 @item -mdisable-callt
12006 @opindex mdisable-callt
12007 This option will suppress generation of the CALLT instruction for the
12008 v850e and v850e1 flavors of the v850 architecture. The default is
12009 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
12014 @subsection VAX Options
12015 @cindex VAX options
12017 These @samp{-m} options are defined for the VAX:
12022 Do not output certain jump instructions (@code{aobleq} and so on)
12023 that the Unix assembler for the VAX cannot handle across long
12028 Do output those jump instructions, on the assumption that you
12029 will assemble with the GNU assembler.
12033 Output code for g-format floating point numbers instead of d-format.
12036 @node x86-64 Options
12037 @subsection x86-64 Options
12038 @cindex x86-64 options
12040 These are listed under @xref{i386 and x86-64 Options}.
12042 @node Xstormy16 Options
12043 @subsection Xstormy16 Options
12044 @cindex Xstormy16 Options
12046 These options are defined for Xstormy16:
12051 Choose startup files and linker script suitable for the simulator.
12054 @node Xtensa Options
12055 @subsection Xtensa Options
12056 @cindex Xtensa Options
12058 These options are supported for Xtensa targets:
12062 @itemx -mno-const16
12064 @opindex mno-const16
12065 Enable or disable use of @code{CONST16} instructions for loading
12066 constant values. The @code{CONST16} instruction is currently not a
12067 standard option from Tensilica. When enabled, @code{CONST16}
12068 instructions are always used in place of the standard @code{L32R}
12069 instructions. The use of @code{CONST16} is enabled by default only if
12070 the @code{L32R} instruction is not available.
12073 @itemx -mno-fused-madd
12074 @opindex mfused-madd
12075 @opindex mno-fused-madd
12076 Enable or disable use of fused multiply/add and multiply/subtract
12077 instructions in the floating-point option. This has no effect if the
12078 floating-point option is not also enabled. Disabling fused multiply/add
12079 and multiply/subtract instructions forces the compiler to use separate
12080 instructions for the multiply and add/subtract operations. This may be
12081 desirable in some cases where strict IEEE 754-compliant results are
12082 required: the fused multiply add/subtract instructions do not round the
12083 intermediate result, thereby producing results with @emph{more} bits of
12084 precision than specified by the IEEE standard. Disabling fused multiply
12085 add/subtract instructions also ensures that the program output is not
12086 sensitive to the compiler's ability to combine multiply and add/subtract
12089 @item -mtext-section-literals
12090 @itemx -mno-text-section-literals
12091 @opindex mtext-section-literals
12092 @opindex mno-text-section-literals
12093 Control the treatment of literal pools. The default is
12094 @option{-mno-text-section-literals}, which places literals in a separate
12095 section in the output file. This allows the literal pool to be placed
12096 in a data RAM/ROM, and it also allows the linker to combine literal
12097 pools from separate object files to remove redundant literals and
12098 improve code size. With @option{-mtext-section-literals}, the literals
12099 are interspersed in the text section in order to keep them as close as
12100 possible to their references. This may be necessary for large assembly
12103 @item -mtarget-align
12104 @itemx -mno-target-align
12105 @opindex mtarget-align
12106 @opindex mno-target-align
12107 When this option is enabled, GCC instructs the assembler to
12108 automatically align instructions to reduce branch penalties at the
12109 expense of some code density. The assembler attempts to widen density
12110 instructions to align branch targets and the instructions following call
12111 instructions. If there are not enough preceding safe density
12112 instructions to align a target, no widening will be performed. The
12113 default is @option{-mtarget-align}. These options do not affect the
12114 treatment of auto-aligned instructions like @code{LOOP}, which the
12115 assembler will always align, either by widening density instructions or
12116 by inserting no-op instructions.
12119 @itemx -mno-longcalls
12120 @opindex mlongcalls
12121 @opindex mno-longcalls
12122 When this option is enabled, GCC instructs the assembler to translate
12123 direct calls to indirect calls unless it can determine that the target
12124 of a direct call is in the range allowed by the call instruction. This
12125 translation typically occurs for calls to functions in other source
12126 files. Specifically, the assembler translates a direct @code{CALL}
12127 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
12128 The default is @option{-mno-longcalls}. This option should be used in
12129 programs where the call target can potentially be out of range. This
12130 option is implemented in the assembler, not the compiler, so the
12131 assembly code generated by GCC will still show direct call
12132 instructions---look at the disassembled object code to see the actual
12133 instructions. Note that the assembler will use an indirect call for
12134 every cross-file call, not just those that really will be out of range.
12137 @node zSeries Options
12138 @subsection zSeries Options
12139 @cindex zSeries options
12141 These are listed under @xref{S/390 and zSeries Options}.
12143 @node Code Gen Options
12144 @section Options for Code Generation Conventions
12145 @cindex code generation conventions
12146 @cindex options, code generation
12147 @cindex run-time options
12149 These machine-independent options control the interface conventions
12150 used in code generation.
12152 Most of them have both positive and negative forms; the negative form
12153 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
12154 one of the forms is listed---the one which is not the default. You
12155 can figure out the other form by either removing @samp{no-} or adding
12159 @item -fbounds-check
12160 @opindex fbounds-check
12161 For front-ends that support it, generate additional code to check that
12162 indices used to access arrays are within the declared range. This is
12163 currently only supported by the Java and Fortran 77 front-ends, where
12164 this option defaults to true and false respectively.
12168 This option generates traps for signed overflow on addition, subtraction,
12169 multiplication operations.
12173 This option instructs the compiler to assume that signed arithmetic
12174 overflow of addition, subtraction and multiplication wraps around
12175 using twos-complement representation. This flag enables some optimizations
12176 and disables other. This option is enabled by default for the Java
12177 front-end, as required by the Java language specification.
12180 @opindex fexceptions
12181 Enable exception handling. Generates extra code needed to propagate
12182 exceptions. For some targets, this implies GCC will generate frame
12183 unwind information for all functions, which can produce significant data
12184 size overhead, although it does not affect execution. If you do not
12185 specify this option, GCC will enable it by default for languages like
12186 C++ which normally require exception handling, and disable it for
12187 languages like C that do not normally require it. However, you may need
12188 to enable this option when compiling C code that needs to interoperate
12189 properly with exception handlers written in C++. You may also wish to
12190 disable this option if you are compiling older C++ programs that don't
12191 use exception handling.
12193 @item -fnon-call-exceptions
12194 @opindex fnon-call-exceptions
12195 Generate code that allows trapping instructions to throw exceptions.
12196 Note that this requires platform-specific runtime support that does
12197 not exist everywhere. Moreover, it only allows @emph{trapping}
12198 instructions to throw exceptions, i.e.@: memory references or floating
12199 point instructions. It does not allow exceptions to be thrown from
12200 arbitrary signal handlers such as @code{SIGALRM}.
12202 @item -funwind-tables
12203 @opindex funwind-tables
12204 Similar to @option{-fexceptions}, except that it will just generate any needed
12205 static data, but will not affect the generated code in any other way.
12206 You will normally not enable this option; instead, a language processor
12207 that needs this handling would enable it on your behalf.
12209 @item -fasynchronous-unwind-tables
12210 @opindex fasynchronous-unwind-tables
12211 Generate unwind table in dwarf2 format, if supported by target machine. The
12212 table is exact at each instruction boundary, so it can be used for stack
12213 unwinding from asynchronous events (such as debugger or garbage collector).
12215 @item -fpcc-struct-return
12216 @opindex fpcc-struct-return
12217 Return ``short'' @code{struct} and @code{union} values in memory like
12218 longer ones, rather than in registers. This convention is less
12219 efficient, but it has the advantage of allowing intercallability between
12220 GCC-compiled files and files compiled with other compilers, particularly
12221 the Portable C Compiler (pcc).
12223 The precise convention for returning structures in memory depends
12224 on the target configuration macros.
12226 Short structures and unions are those whose size and alignment match
12227 that of some integer type.
12229 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12230 switch is not binary compatible with code compiled with the
12231 @option{-freg-struct-return} switch.
12232 Use it to conform to a non-default application binary interface.
12234 @item -freg-struct-return
12235 @opindex freg-struct-return
12236 Return @code{struct} and @code{union} values in registers when possible.
12237 This is more efficient for small structures than
12238 @option{-fpcc-struct-return}.
12240 If you specify neither @option{-fpcc-struct-return} nor
12241 @option{-freg-struct-return}, GCC defaults to whichever convention is
12242 standard for the target. If there is no standard convention, GCC
12243 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12244 the principal compiler. In those cases, we can choose the standard, and
12245 we chose the more efficient register return alternative.
12247 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12248 switch is not binary compatible with code compiled with the
12249 @option{-fpcc-struct-return} switch.
12250 Use it to conform to a non-default application binary interface.
12252 @item -fshort-enums
12253 @opindex fshort-enums
12254 Allocate to an @code{enum} type only as many bytes as it needs for the
12255 declared range of possible values. Specifically, the @code{enum} type
12256 will be equivalent to the smallest integer type which has enough room.
12258 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12259 code that is not binary compatible with code generated without that switch.
12260 Use it to conform to a non-default application binary interface.
12262 @item -fshort-double
12263 @opindex fshort-double
12264 Use the same size for @code{double} as for @code{float}.
12266 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
12267 code that is not binary compatible with code generated without that switch.
12268 Use it to conform to a non-default application binary interface.
12270 @item -fshort-wchar
12271 @opindex fshort-wchar
12272 Override the underlying type for @samp{wchar_t} to be @samp{short
12273 unsigned int} instead of the default for the target. This option is
12274 useful for building programs to run under WINE@.
12276 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12277 code that is not binary compatible with code generated without that switch.
12278 Use it to conform to a non-default application binary interface.
12280 @item -fshared-data
12281 @opindex fshared-data
12282 Requests that the data and non-@code{const} variables of this
12283 compilation be shared data rather than private data. The distinction
12284 makes sense only on certain operating systems, where shared data is
12285 shared between processes running the same program, while private data
12286 exists in one copy per process.
12289 @opindex fno-common
12290 In C, allocate even uninitialized global variables in the data section of the
12291 object file, rather than generating them as common blocks. This has the
12292 effect that if the same variable is declared (without @code{extern}) in
12293 two different compilations, you will get an error when you link them.
12294 The only reason this might be useful is if you wish to verify that the
12295 program will work on other systems which always work this way.
12299 Ignore the @samp{#ident} directive.
12301 @item -finhibit-size-directive
12302 @opindex finhibit-size-directive
12303 Don't output a @code{.size} assembler directive, or anything else that
12304 would cause trouble if the function is split in the middle, and the
12305 two halves are placed at locations far apart in memory. This option is
12306 used when compiling @file{crtstuff.c}; you should not need to use it
12309 @item -fverbose-asm
12310 @opindex fverbose-asm
12311 Put extra commentary information in the generated assembly code to
12312 make it more readable. This option is generally only of use to those
12313 who actually need to read the generated assembly code (perhaps while
12314 debugging the compiler itself).
12316 @option{-fno-verbose-asm}, the default, causes the
12317 extra information to be omitted and is useful when comparing two assembler
12322 @cindex global offset table
12324 Generate position-independent code (PIC) suitable for use in a shared
12325 library, if supported for the target machine. Such code accesses all
12326 constant addresses through a global offset table (GOT)@. The dynamic
12327 loader resolves the GOT entries when the program starts (the dynamic
12328 loader is not part of GCC; it is part of the operating system). If
12329 the GOT size for the linked executable exceeds a machine-specific
12330 maximum size, you get an error message from the linker indicating that
12331 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12332 instead. (These maximums are 8k on the SPARC and 32k
12333 on the m68k and RS/6000. The 386 has no such limit.)
12335 Position-independent code requires special support, and therefore works
12336 only on certain machines. For the 386, GCC supports PIC for System V
12337 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12338 position-independent.
12342 If supported for the target machine, emit position-independent code,
12343 suitable for dynamic linking and avoiding any limit on the size of the
12344 global offset table. This option makes a difference on the m68k,
12345 PowerPC and SPARC@.
12347 Position-independent code requires special support, and therefore works
12348 only on certain machines.
12354 These options are similar to @option{-fpic} and @option{-fPIC}, but
12355 generated position independent code can be only linked into executables.
12356 Usually these options are used when @option{-pie} GCC option will be
12357 used during linking.
12359 @item -ffixed-@var{reg}
12361 Treat the register named @var{reg} as a fixed register; generated code
12362 should never refer to it (except perhaps as a stack pointer, frame
12363 pointer or in some other fixed role).
12365 @var{reg} must be the name of a register. The register names accepted
12366 are machine-specific and are defined in the @code{REGISTER_NAMES}
12367 macro in the machine description macro file.
12369 This flag does not have a negative form, because it specifies a
12372 @item -fcall-used-@var{reg}
12373 @opindex fcall-used
12374 Treat the register named @var{reg} as an allocable register that is
12375 clobbered by function calls. It may be allocated for temporaries or
12376 variables that do not live across a call. Functions compiled this way
12377 will not save and restore the register @var{reg}.
12379 It is an error to used this flag with the frame pointer or stack pointer.
12380 Use of this flag for other registers that have fixed pervasive roles in
12381 the machine's execution model will produce disastrous results.
12383 This flag does not have a negative form, because it specifies a
12386 @item -fcall-saved-@var{reg}
12387 @opindex fcall-saved
12388 Treat the register named @var{reg} as an allocable register saved by
12389 functions. It may be allocated even for temporaries or variables that
12390 live across a call. Functions compiled this way will save and restore
12391 the register @var{reg} if they use it.
12393 It is an error to used this flag with the frame pointer or stack pointer.
12394 Use of this flag for other registers that have fixed pervasive roles in
12395 the machine's execution model will produce disastrous results.
12397 A different sort of disaster will result from the use of this flag for
12398 a register in which function values may be returned.
12400 This flag does not have a negative form, because it specifies a
12403 @item -fpack-struct[=@var{n}]
12404 @opindex fpack-struct
12405 Without a value specified, pack all structure members together without
12406 holes. When a value is specified (which must be a small power of two), pack
12407 structure members according to this value, representing the maximum
12408 alignment (that is, objects with default alignment requirements larger than
12409 this will be output potentially unaligned at the next fitting location.
12411 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12412 code that is not binary compatible with code generated without that switch.
12413 Additionally, it makes the code suboptimal.
12414 Use it to conform to a non-default application binary interface.
12416 @item -finstrument-functions
12417 @opindex finstrument-functions
12418 Generate instrumentation calls for entry and exit to functions. Just
12419 after function entry and just before function exit, the following
12420 profiling functions will be called with the address of the current
12421 function and its call site. (On some platforms,
12422 @code{__builtin_return_address} does not work beyond the current
12423 function, so the call site information may not be available to the
12424 profiling functions otherwise.)
12427 void __cyg_profile_func_enter (void *this_fn,
12429 void __cyg_profile_func_exit (void *this_fn,
12433 The first argument is the address of the start of the current function,
12434 which may be looked up exactly in the symbol table.
12436 This instrumentation is also done for functions expanded inline in other
12437 functions. The profiling calls will indicate where, conceptually, the
12438 inline function is entered and exited. This means that addressable
12439 versions of such functions must be available. If all your uses of a
12440 function are expanded inline, this may mean an additional expansion of
12441 code size. If you use @samp{extern inline} in your C code, an
12442 addressable version of such functions must be provided. (This is
12443 normally the case anyways, but if you get lucky and the optimizer always
12444 expands the functions inline, you might have gotten away without
12445 providing static copies.)
12447 A function may be given the attribute @code{no_instrument_function}, in
12448 which case this instrumentation will not be done. This can be used, for
12449 example, for the profiling functions listed above, high-priority
12450 interrupt routines, and any functions from which the profiling functions
12451 cannot safely be called (perhaps signal handlers, if the profiling
12452 routines generate output or allocate memory).
12454 @item -fstack-check
12455 @opindex fstack-check
12456 Generate code to verify that you do not go beyond the boundary of the
12457 stack. You should specify this flag if you are running in an
12458 environment with multiple threads, but only rarely need to specify it in
12459 a single-threaded environment since stack overflow is automatically
12460 detected on nearly all systems if there is only one stack.
12462 Note that this switch does not actually cause checking to be done; the
12463 operating system must do that. The switch causes generation of code
12464 to ensure that the operating system sees the stack being extended.
12466 @item -fstack-limit-register=@var{reg}
12467 @itemx -fstack-limit-symbol=@var{sym}
12468 @itemx -fno-stack-limit
12469 @opindex fstack-limit-register
12470 @opindex fstack-limit-symbol
12471 @opindex fno-stack-limit
12472 Generate code to ensure that the stack does not grow beyond a certain value,
12473 either the value of a register or the address of a symbol. If the stack
12474 would grow beyond the value, a signal is raised. For most targets,
12475 the signal is raised before the stack overruns the boundary, so
12476 it is possible to catch the signal without taking special precautions.
12478 For instance, if the stack starts at absolute address @samp{0x80000000}
12479 and grows downwards, you can use the flags
12480 @option{-fstack-limit-symbol=__stack_limit} and
12481 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12482 of 128KB@. Note that this may only work with the GNU linker.
12484 @cindex aliasing of parameters
12485 @cindex parameters, aliased
12486 @item -fargument-alias
12487 @itemx -fargument-noalias
12488 @itemx -fargument-noalias-global
12489 @opindex fargument-alias
12490 @opindex fargument-noalias
12491 @opindex fargument-noalias-global
12492 Specify the possible relationships among parameters and between
12493 parameters and global data.
12495 @option{-fargument-alias} specifies that arguments (parameters) may
12496 alias each other and may alias global storage.@*
12497 @option{-fargument-noalias} specifies that arguments do not alias
12498 each other, but may alias global storage.@*
12499 @option{-fargument-noalias-global} specifies that arguments do not
12500 alias each other and do not alias global storage.
12502 Each language will automatically use whatever option is required by
12503 the language standard. You should not need to use these options yourself.
12505 @item -fleading-underscore
12506 @opindex fleading-underscore
12507 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12508 change the way C symbols are represented in the object file. One use
12509 is to help link with legacy assembly code.
12511 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12512 generate code that is not binary compatible with code generated without that
12513 switch. Use it to conform to a non-default application binary interface.
12514 Not all targets provide complete support for this switch.
12516 @item -ftls-model=@var{model}
12517 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12518 The @var{model} argument should be one of @code{global-dynamic},
12519 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
12521 The default without @option{-fpic} is @code{initial-exec}; with
12522 @option{-fpic} the default is @code{global-dynamic}.
12524 @item -fvisibility=@var{default|internal|hidden|protected}
12525 @opindex fvisibility
12526 Set the default ELF image symbol visibility to the specified option---all
12527 symbols will be marked with this unless overridden within the code.
12528 Using this feature can very substantially improve linking and
12529 load times of shared object libraries, produce more optimized
12530 code, provide near-perfect API export and prevent symbol clashes.
12531 It is @strong{strongly} recommended that you use this in any shared objects
12534 Despite the nomenclature, @code{default} always means public ie;
12535 available to be linked against from outside the shared object.
12536 @code{protected} and @code{internal} are pretty useless in real-world
12537 usage so the only other commonly used option will be @code{hidden}.
12538 The default if @option{-fvisibility} isn't specified is
12539 @code{default}, i.e., make every
12540 symbol public---this causes the same behavior as previous versions of
12543 A good explanation of the benefits offered by ensuring ELF
12544 symbols have the correct visibility is given by ``How To Write
12545 Shared Libraries'' by Ulrich Drepper (which can be found at
12546 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
12547 solution made possible by this option to marking things hidden when
12548 the default is public is to make the default hidden and mark things
12549 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
12550 and @code{__attribute__ ((visibility("default")))} instead of
12551 @code{__declspec(dllexport)} you get almost identical semantics with
12552 identical syntax. This is a great boon to those working with
12553 cross-platform projects.
12555 For those adding visibility support to existing code, you may find
12556 @samp{#pragma GCC visibility} of use. This works by you enclosing
12557 the declarations you wish to set visibility for with (for example)
12558 @samp{#pragma GCC visibility push(hidden)} and
12559 @samp{#pragma GCC visibility pop}. These can be nested up to sixteen
12560 times. Bear in mind that symbol visibility should be viewed @strong{as
12561 part of the API interface contract} and thus all new code should
12562 always specify visibility when it is not the default ie; declarations
12563 only for use within the local DSO should @strong{always} be marked explicitly
12564 as hidden as so to avoid PLT indirection overheads---making this
12565 abundantly clear also aids readability and self-documentation of the code.
12566 Note that due to ISO C++ specification requirements, operator new and
12567 operator delete must always be of default visibility.
12569 An overview of these techniques, their benefits and how to use them
12570 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
12576 @node Environment Variables
12577 @section Environment Variables Affecting GCC
12578 @cindex environment variables
12580 @c man begin ENVIRONMENT
12581 This section describes several environment variables that affect how GCC
12582 operates. Some of them work by specifying directories or prefixes to use
12583 when searching for various kinds of files. Some are used to specify other
12584 aspects of the compilation environment.
12586 Note that you can also specify places to search using options such as
12587 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
12588 take precedence over places specified using environment variables, which
12589 in turn take precedence over those specified by the configuration of GCC@.
12590 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
12591 GNU Compiler Collection (GCC) Internals}.
12596 @c @itemx LC_COLLATE
12598 @c @itemx LC_MONETARY
12599 @c @itemx LC_NUMERIC
12604 @c @findex LC_COLLATE
12605 @findex LC_MESSAGES
12606 @c @findex LC_MONETARY
12607 @c @findex LC_NUMERIC
12611 These environment variables control the way that GCC uses
12612 localization information that allow GCC to work with different
12613 national conventions. GCC inspects the locale categories
12614 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
12615 so. These locale categories can be set to any value supported by your
12616 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
12617 Kingdom encoded in UTF-8.
12619 The @env{LC_CTYPE} environment variable specifies character
12620 classification. GCC uses it to determine the character boundaries in
12621 a string; this is needed for some multibyte encodings that contain quote
12622 and escape characters that would otherwise be interpreted as a string
12625 The @env{LC_MESSAGES} environment variable specifies the language to
12626 use in diagnostic messages.
12628 If the @env{LC_ALL} environment variable is set, it overrides the value
12629 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
12630 and @env{LC_MESSAGES} default to the value of the @env{LANG}
12631 environment variable. If none of these variables are set, GCC
12632 defaults to traditional C English behavior.
12636 If @env{TMPDIR} is set, it specifies the directory to use for temporary
12637 files. GCC uses temporary files to hold the output of one stage of
12638 compilation which is to be used as input to the next stage: for example,
12639 the output of the preprocessor, which is the input to the compiler
12642 @item GCC_EXEC_PREFIX
12643 @findex GCC_EXEC_PREFIX
12644 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
12645 names of the subprograms executed by the compiler. No slash is added
12646 when this prefix is combined with the name of a subprogram, but you can
12647 specify a prefix that ends with a slash if you wish.
12649 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
12650 an appropriate prefix to use based on the pathname it was invoked with.
12652 If GCC cannot find the subprogram using the specified prefix, it
12653 tries looking in the usual places for the subprogram.
12655 The default value of @env{GCC_EXEC_PREFIX} is
12656 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
12657 of @code{prefix} when you ran the @file{configure} script.
12659 Other prefixes specified with @option{-B} take precedence over this prefix.
12661 This prefix is also used for finding files such as @file{crt0.o} that are
12664 In addition, the prefix is used in an unusual way in finding the
12665 directories to search for header files. For each of the standard
12666 directories whose name normally begins with @samp{/usr/local/lib/gcc}
12667 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
12668 replacing that beginning with the specified prefix to produce an
12669 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
12670 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
12671 These alternate directories are searched first; the standard directories
12674 @item COMPILER_PATH
12675 @findex COMPILER_PATH
12676 The value of @env{COMPILER_PATH} is a colon-separated list of
12677 directories, much like @env{PATH}. GCC tries the directories thus
12678 specified when searching for subprograms, if it can't find the
12679 subprograms using @env{GCC_EXEC_PREFIX}.
12682 @findex LIBRARY_PATH
12683 The value of @env{LIBRARY_PATH} is a colon-separated list of
12684 directories, much like @env{PATH}. When configured as a native compiler,
12685 GCC tries the directories thus specified when searching for special
12686 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
12687 using GCC also uses these directories when searching for ordinary
12688 libraries for the @option{-l} option (but directories specified with
12689 @option{-L} come first).
12693 @cindex locale definition
12694 This variable is used to pass locale information to the compiler. One way in
12695 which this information is used is to determine the character set to be used
12696 when character literals, string literals and comments are parsed in C and C++.
12697 When the compiler is configured to allow multibyte characters,
12698 the following values for @env{LANG} are recognized:
12702 Recognize JIS characters.
12704 Recognize SJIS characters.
12706 Recognize EUCJP characters.
12709 If @env{LANG} is not defined, or if it has some other value, then the
12710 compiler will use mblen and mbtowc as defined by the default locale to
12711 recognize and translate multibyte characters.
12715 Some additional environments variables affect the behavior of the
12718 @include cppenv.texi
12722 @node Precompiled Headers
12723 @section Using Precompiled Headers
12724 @cindex precompiled headers
12725 @cindex speed of compilation
12727 Often large projects have many header files that are included in every
12728 source file. The time the compiler takes to process these header files
12729 over and over again can account for nearly all of the time required to
12730 build the project. To make builds faster, GCC allows users to
12731 `precompile' a header file; then, if builds can use the precompiled
12732 header file they will be much faster.
12734 @strong{Caution:} There are a few known situations where GCC will
12735 crash when trying to use a precompiled header. If you have trouble
12736 with a precompiled header, you should remove the precompiled header
12737 and compile without it. In addition, please use GCC's on-line
12738 defect-tracking system to report any problems you encounter with
12739 precompiled headers. @xref{Bugs}.
12741 To create a precompiled header file, simply compile it as you would any
12742 other file, if necessary using the @option{-x} option to make the driver
12743 treat it as a C or C++ header file. You will probably want to use a
12744 tool like @command{make} to keep the precompiled header up-to-date when
12745 the headers it contains change.
12747 A precompiled header file will be searched for when @code{#include} is
12748 seen in the compilation. As it searches for the included file
12749 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
12750 compiler looks for a precompiled header in each directory just before it
12751 looks for the include file in that directory. The name searched for is
12752 the name specified in the @code{#include} with @samp{.gch} appended. If
12753 the precompiled header file can't be used, it is ignored.
12755 For instance, if you have @code{#include "all.h"}, and you have
12756 @file{all.h.gch} in the same directory as @file{all.h}, then the
12757 precompiled header file will be used if possible, and the original
12758 header will be used otherwise.
12760 Alternatively, you might decide to put the precompiled header file in a
12761 directory and use @option{-I} to ensure that directory is searched
12762 before (or instead of) the directory containing the original header.
12763 Then, if you want to check that the precompiled header file is always
12764 used, you can put a file of the same name as the original header in this
12765 directory containing an @code{#error} command.
12767 This also works with @option{-include}. So yet another way to use
12768 precompiled headers, good for projects not designed with precompiled
12769 header files in mind, is to simply take most of the header files used by
12770 a project, include them from another header file, precompile that header
12771 file, and @option{-include} the precompiled header. If the header files
12772 have guards against multiple inclusion, they will be skipped because
12773 they've already been included (in the precompiled header).
12775 If you need to precompile the same header file for different
12776 languages, targets, or compiler options, you can instead make a
12777 @emph{directory} named like @file{all.h.gch}, and put each precompiled
12778 header in the directory, perhaps using @option{-o}. It doesn't matter
12779 what you call the files in the directory, every precompiled header in
12780 the directory will be considered. The first precompiled header
12781 encountered in the directory that is valid for this compilation will
12782 be used; they're searched in no particular order.
12784 There are many other possibilities, limited only by your imagination,
12785 good sense, and the constraints of your build system.
12787 A precompiled header file can be used only when these conditions apply:
12791 Only one precompiled header can be used in a particular compilation.
12794 A precompiled header can't be used once the first C token is seen. You
12795 can have preprocessor directives before a precompiled header; you can
12796 even include a precompiled header from inside another header, so long as
12797 there are no C tokens before the @code{#include}.
12800 The precompiled header file must be produced for the same language as
12801 the current compilation. You can't use a C precompiled header for a C++
12805 The precompiled header file must be produced by the same compiler
12806 version and configuration as the current compilation is using.
12807 The easiest way to guarantee this is to use the same compiler binary
12808 for creating and using precompiled headers.
12811 Any macros defined before the precompiled header is included must
12812 either be defined in the same way as when the precompiled header was
12813 generated, or must not affect the precompiled header, which usually
12814 means that they don't appear in the precompiled header at all.
12816 The @option{-D} option is one way to define a macro before a
12817 precompiled header is included; using a @code{#define} can also do it.
12818 There are also some options that define macros implicitly, like
12819 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
12822 @item If debugging information is output when using the precompiled
12823 header, using @option{-g} or similar, the same kind of debugging information
12824 must have been output when building the precompiled header. However,
12825 a precompiled header built using @option{-g} can be used in a compilation
12826 when no debugging information is being output.
12828 @item The same @option{-m} options must generally be used when building
12829 and using the precompiled header. @xref{Submodel Options},
12830 for any cases where this rule is relaxed.
12832 @item Each of the following options must be the same when building and using
12833 the precompiled header:
12835 @gccoptlist{-fexceptions -funit-at-a-time}
12838 Some other command-line options starting with @option{-f},
12839 @option{-p}, or @option{-O} must be defined in the same way as when
12840 the precompiled header was generated. At present, it's not clear
12841 which options are safe to change and which are not; the safest choice
12842 is to use exactly the same options when generating and using the
12843 precompiled header. The following are known to be safe:
12845 @gccoptlist{-fpreprocessed -pedantic-errors}
12849 For all of these except the last, the compiler will automatically
12850 ignore the precompiled header if the conditions aren't met. If you
12851 find an option combination that doesn't work and doesn't cause the
12852 precompiled header to be ignored, please consider filing a bug report,
12855 If you do use differing options when generating and using the
12856 precompiled header, the actual behavior will be a mixture of the
12857 behavior for the options. For instance, if you use @option{-g} to
12858 generate the precompiled header but not when using it, you may or may
12859 not get debugging information for routines in the precompiled header.
12861 @node Running Protoize
12862 @section Running Protoize
12864 The program @code{protoize} is an optional part of GCC@. You can use
12865 it to add prototypes to a program, thus converting the program to ISO
12866 C in one respect. The companion program @code{unprotoize} does the
12867 reverse: it removes argument types from any prototypes that are found.
12869 When you run these programs, you must specify a set of source files as
12870 command line arguments. The conversion programs start out by compiling
12871 these files to see what functions they define. The information gathered
12872 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
12874 After scanning comes actual conversion. The specified files are all
12875 eligible to be converted; any files they include (whether sources or
12876 just headers) are eligible as well.
12878 But not all the eligible files are converted. By default,
12879 @code{protoize} and @code{unprotoize} convert only source and header
12880 files in the current directory. You can specify additional directories
12881 whose files should be converted with the @option{-d @var{directory}}
12882 option. You can also specify particular files to exclude with the
12883 @option{-x @var{file}} option. A file is converted if it is eligible, its
12884 directory name matches one of the specified directory names, and its
12885 name within the directory has not been excluded.
12887 Basic conversion with @code{protoize} consists of rewriting most
12888 function definitions and function declarations to specify the types of
12889 the arguments. The only ones not rewritten are those for varargs
12892 @code{protoize} optionally inserts prototype declarations at the
12893 beginning of the source file, to make them available for any calls that
12894 precede the function's definition. Or it can insert prototype
12895 declarations with block scope in the blocks where undeclared functions
12898 Basic conversion with @code{unprotoize} consists of rewriting most
12899 function declarations to remove any argument types, and rewriting
12900 function definitions to the old-style pre-ISO form.
12902 Both conversion programs print a warning for any function declaration or
12903 definition that they can't convert. You can suppress these warnings
12906 The output from @code{protoize} or @code{unprotoize} replaces the
12907 original source file. The original file is renamed to a name ending
12908 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
12909 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
12910 for DOS) file already exists, then the source file is simply discarded.
12912 @code{protoize} and @code{unprotoize} both depend on GCC itself to
12913 scan the program and collect information about the functions it uses.
12914 So neither of these programs will work until GCC is installed.
12916 Here is a table of the options you can use with @code{protoize} and
12917 @code{unprotoize}. Each option works with both programs unless
12921 @item -B @var{directory}
12922 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
12923 usual directory (normally @file{/usr/local/lib}). This file contains
12924 prototype information about standard system functions. This option
12925 applies only to @code{protoize}.
12927 @item -c @var{compilation-options}
12928 Use @var{compilation-options} as the options when running @command{gcc} to
12929 produce the @samp{.X} files. The special option @option{-aux-info} is
12930 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
12932 Note that the compilation options must be given as a single argument to
12933 @code{protoize} or @code{unprotoize}. If you want to specify several
12934 @command{gcc} options, you must quote the entire set of compilation options
12935 to make them a single word in the shell.
12937 There are certain @command{gcc} arguments that you cannot use, because they
12938 would produce the wrong kind of output. These include @option{-g},
12939 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
12940 the @var{compilation-options}, they are ignored.
12943 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
12944 systems) instead of @samp{.c}. This is convenient if you are converting
12945 a C program to C++. This option applies only to @code{protoize}.
12948 Add explicit global declarations. This means inserting explicit
12949 declarations at the beginning of each source file for each function
12950 that is called in the file and was not declared. These declarations
12951 precede the first function definition that contains a call to an
12952 undeclared function. This option applies only to @code{protoize}.
12954 @item -i @var{string}
12955 Indent old-style parameter declarations with the string @var{string}.
12956 This option applies only to @code{protoize}.
12958 @code{unprotoize} converts prototyped function definitions to old-style
12959 function definitions, where the arguments are declared between the
12960 argument list and the initial @samp{@{}. By default, @code{unprotoize}
12961 uses five spaces as the indentation. If you want to indent with just
12962 one space instead, use @option{-i " "}.
12965 Keep the @samp{.X} files. Normally, they are deleted after conversion
12969 Add explicit local declarations. @code{protoize} with @option{-l} inserts
12970 a prototype declaration for each function in each block which calls the
12971 function without any declaration. This option applies only to
12975 Make no real changes. This mode just prints information about the conversions
12976 that would have been done without @option{-n}.
12979 Make no @samp{.save} files. The original files are simply deleted.
12980 Use this option with caution.
12982 @item -p @var{program}
12983 Use the program @var{program} as the compiler. Normally, the name
12984 @file{gcc} is used.
12987 Work quietly. Most warnings are suppressed.
12990 Print the version number, just like @option{-v} for @command{gcc}.
12993 If you need special compiler options to compile one of your program's
12994 source files, then you should generate that file's @samp{.X} file
12995 specially, by running @command{gcc} on that source file with the
12996 appropriate options and the option @option{-aux-info}. Then run
12997 @code{protoize} on the entire set of files. @code{protoize} will use
12998 the existing @samp{.X} file because it is newer than the source file.
13002 gcc -Dfoo=bar file1.c -aux-info file1.X
13007 You need to include the special files along with the rest in the
13008 @code{protoize} command, even though their @samp{.X} files already
13009 exist, because otherwise they won't get converted.
13011 @xref{Protoize Caveats}, for more information on how to use
13012 @code{protoize} successfully.