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,
106 @option{-fstrength-reduce}, @option{-Wformat} and so on. Most of
107 these have both positive and negative forms; the negative form of
108 @option{-ffoo} would be @option{-fno-foo}. This manual documents
109 only one of these two forms, whichever one is not the default.
113 @xref{Option Index}, for an index to GCC's options.
116 * Option Summary:: Brief list of all options, without explanations.
117 * Overall Options:: Controlling the kind of output:
118 an executable, object files, assembler files,
119 or preprocessed source.
120 * Invoking G++:: Compiling C++ programs.
121 * C Dialect Options:: Controlling the variant of C language compiled.
122 * C++ Dialect Options:: Variations on C++.
123 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
125 * Language Independent Options:: Controlling how diagnostics should be
127 * Warning Options:: How picky should the compiler be?
128 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
129 * Optimize Options:: How much optimization?
130 * Preprocessor Options:: Controlling header files and macro definitions.
131 Also, getting dependency information for Make.
132 * Assembler Options:: Passing options to the assembler.
133 * Link Options:: Specifying libraries and so on.
134 * Directory Options:: Where to find header files and libraries.
135 Where to find the compiler executable files.
136 * Spec Files:: How to pass switches to sub-processes.
137 * Target Options:: Running a cross-compiler, or an old version of GCC.
138 * Submodel Options:: Specifying minor hardware or convention variations,
139 such as 68010 vs 68020.
140 * Code Gen Options:: Specifying conventions for function calls, data layout
142 * Environment Variables:: Env vars that affect GCC.
143 * Precompiled Headers:: Compiling a header once, and using it many times.
144 * Running Protoize:: Automatically adding or removing function prototypes.
150 @section Option Summary
152 Here is a summary of all the options, grouped by type. Explanations are
153 in the following sections.
156 @item Overall Options
157 @xref{Overall Options,,Options Controlling the Kind of Output}.
158 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
159 -x @var{language} -v -### --help --target-help --version}
161 @item C Language Options
162 @xref{C Dialect Options,,Options Controlling C Dialect}.
163 @gccoptlist{-ansi -std=@var{standard} -aux-info @var{filename} @gol
164 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
165 -fhosted -ffreestanding -fms-extensions @gol
166 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
167 -fallow-single-precision -fcond-mismatch @gol
168 -fsigned-bitfields -fsigned-char @gol
169 -funsigned-bitfields -funsigned-char}
171 @item C++ Language Options
172 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
173 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
174 -fconserve-space -fno-const-strings @gol
175 -fno-elide-constructors @gol
176 -fno-enforce-eh-specs @gol
177 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
178 -fno-implicit-templates @gol
179 -fno-implicit-inline-templates @gol
180 -fno-implement-inlines -fms-extensions @gol
181 -fno-nonansi-builtins -fno-operator-names @gol
182 -fno-optional-diags -fpermissive @gol
183 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
184 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
185 -fno-default-inline -fvisibility-inlines-hidden @gol
186 -Wabi -Wctor-dtor-privacy @gol
187 -Wnon-virtual-dtor -Wreorder @gol
188 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel @gol
189 -Wno-non-template-friend -Wold-style-cast @gol
190 -Woverloaded-virtual -Wno-pmf-conversions @gol
193 @item Objective-C and Objective-C++ Language Options
194 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
195 Objective-C and Objective-C++ Dialects}.
197 -fconstant-string-class=@var{class-name} @gol
198 -fgnu-runtime -fnext-runtime @gol
199 -fno-nil-receivers @gol
200 -fobjc-call-cxx-cdtors @gol
201 -fobjc-direct-dispatch @gol
202 -fobjc-exceptions @gol
204 -freplace-objc-classes @gol
207 -Wassign-intercept @gol
208 -Wno-protocol -Wselector @gol
209 -Wstrict-selector-match @gol
210 -Wundeclared-selector}
212 @item Language Independent Options
213 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
214 @gccoptlist{-fmessage-length=@var{n} @gol
215 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}} @gol
216 -fdiagnostics-show-options
218 @item Warning Options
219 @xref{Warning Options,,Options to Request or Suppress Warnings}.
220 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
221 -w -Wextra -Wall -Waggregate-return -Wno-attributes @gol
222 -Wc++-compat -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
223 -Wconversion -Wno-deprecated-declarations @gol
224 -Wdisabled-optimization -Wno-div-by-zero -Wno-endif-labels @gol
225 -Werror -Werror-implicit-function-declaration @gol
226 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
227 -Wno-format-extra-args -Wformat-nonliteral @gol
228 -Wformat-security -Wformat-y2k @gol
229 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
230 -Wimport -Wno-import -Winit-self -Winline @gol
231 -Wno-int-to-pointer-cast @gol
232 -Wno-invalid-offsetof -Winvalid-pch @gol
233 -Wlarger-than-@var{len} -Wunsafe-loop-optimizations -Wlong-long @gol
234 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
235 -Wmissing-format-attribute -Wmissing-include-dirs @gol
236 -Wmissing-noreturn @gol
237 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
238 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
239 -Wredundant-decls @gol
240 -Wreturn-type -Wsequence-point -Wshadow @gol
241 -Wsign-compare -Wstrict-aliasing -Wstrict-aliasing=2 @gol
242 -Wswitch -Wswitch-default -Wswitch-enum @gol
243 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
244 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
245 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
246 -Wunused-value -Wunused-variable -Wwrite-strings @gol
249 @item C-only Warning Options
250 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
251 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
252 -Wstrict-prototypes -Wtraditional @gol
253 -Wdeclaration-after-statement -Wno-pointer-sign}
255 @item Debugging Options
256 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
257 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
258 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
259 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
260 -fdump-ipa-all -fdump-ipa-cgraph @gol
262 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
263 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
264 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
267 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
268 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
269 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
270 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
271 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
272 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
273 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
274 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
275 -fdump-tree-nrv -fdump-tree-vect @gol
276 -fdump-tree-sink @gol
277 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
278 -fdump-tree-salias @gol
279 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
280 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
281 -ftree-vectorizer-verbose=@var{n} @gol
282 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
283 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
284 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs @gol
285 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
286 -ftest-coverage -ftime-report -fvar-tracking @gol
287 -g -g@var{level} -gcoff -gdwarf-2 @gol
288 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
289 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
290 -print-multi-directory -print-multi-lib @gol
291 -print-prog-name=@var{program} -print-search-dirs -Q @gol
294 @item Optimization Options
295 @xref{Optimize Options,,Options that Control Optimization}.
296 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
297 -falign-labels=@var{n} -falign-loops=@var{n} @gol
298 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
299 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
300 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
301 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
302 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
303 -fdelayed-branch -fdelete-null-pointer-checks -fearly-inlining @gol
304 -fexpensive-optimizations -ffast-math -ffloat-store @gol
305 -fforce-addr -ffunction-sections @gol
306 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
307 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2 @gol
308 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
309 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
310 -fmodulo-sched -fno-branch-count-reg @gol
311 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants @gol
312 -fno-function-cse -fno-guess-branch-probability @gol
313 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
314 -funsafe-math-optimizations -funsafe-loop-optimizations -ffinite-math-only @gol
315 -fno-trapping-math -fno-zero-initialized-in-bss @gol
316 -fomit-frame-pointer -foptimize-register-move @gol
317 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
318 -fprofile-generate -fprofile-use @gol
319 -fregmove -frename-registers @gol
320 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
321 -frerun-cse-after-loop -frerun-loop-opt @gol
322 -frounding-math -fschedule-insns -fschedule-insns2 @gol
323 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
324 -fsched-spec-load-dangerous @gol
325 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
326 -fsched2-use-superblocks @gol
327 -fsched2-use-traces -freschedule-modulo-scheduled-loops @gol
328 -fsignaling-nans -fsingle-precision-constant @gol
329 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
330 -funroll-all-loops -funroll-loops -fpeel-loops @gol
331 -fsplit-ivs-in-unroller -funswitch-loops @gol
332 -fvariable-expansion-in-unroller @gol
333 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
334 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
335 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
336 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
337 -ftree-salias -fweb @gol
338 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
339 --param @var{name}=@var{value}
340 -O -O0 -O1 -O2 -O3 -Os}
342 @item Preprocessor Options
343 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
344 @gccoptlist{-A@var{question}=@var{answer} @gol
345 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
346 -C -dD -dI -dM -dN @gol
347 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
348 -idirafter @var{dir} @gol
349 -include @var{file} -imacros @var{file} @gol
350 -iprefix @var{file} -iwithprefix @var{dir} @gol
351 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
352 -isysroot @var{dir} @gol
353 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
354 -P -fworking-directory -remap @gol
355 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
356 -Xpreprocessor @var{option}}
358 @item Assembler Option
359 @xref{Assembler Options,,Passing Options to the Assembler}.
360 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
363 @xref{Link Options,,Options for Linking}.
364 @gccoptlist{@var{object-file-name} -l@var{library} @gol
365 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
366 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
367 -Wl,@var{option} -Xlinker @var{option} @gol
370 @item Directory Options
371 @xref{Directory Options,,Options for Directory Search}.
372 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
373 -specs=@var{file} -I- --sysroot=@var{dir}}
376 @c I wrote this xref this way to avoid overfull hbox. -- rms
377 @xref{Target Options}.
378 @gccoptlist{-V @var{version} -b @var{machine}}
380 @item Machine Dependent Options
381 @xref{Submodel Options,,Hardware Models and Configurations}.
382 @c This list is ordered alphanumerically by subsection name.
383 @c Try and put the significant identifier (CPU or system) first,
384 @c so users have a clue at guessing where the ones they want will be.
387 @gccoptlist{-EB -EL @gol
388 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
389 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
392 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
393 -mabi=@var{name} @gol
394 -mapcs-stack-check -mno-apcs-stack-check @gol
395 -mapcs-float -mno-apcs-float @gol
396 -mapcs-reentrant -mno-apcs-reentrant @gol
397 -msched-prolog -mno-sched-prolog @gol
398 -mlittle-endian -mbig-endian -mwords-little-endian @gol
399 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
400 -mthumb-interwork -mno-thumb-interwork @gol
401 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
402 -mstructure-size-boundary=@var{n} @gol
403 -mabort-on-noreturn @gol
404 -mlong-calls -mno-long-calls @gol
405 -msingle-pic-base -mno-single-pic-base @gol
406 -mpic-register=@var{reg} @gol
407 -mnop-fun-dllimport @gol
408 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
409 -mpoke-function-name @gol
411 -mtpcs-frame -mtpcs-leaf-frame @gol
412 -mcaller-super-interworking -mcallee-super-interworking}
415 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
416 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
418 @emph{Blackfin Options}
419 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
420 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
421 -mlow-64k -mno-low64k -mid-shared-library @gol
422 -mno-id-shared-library -mshared-library-id=@var{n} @gol
423 -mlong-calls -mno-long-calls}
426 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
427 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
428 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
429 -mstack-align -mdata-align -mconst-align @gol
430 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
431 -melf -maout -melinux -mlinux -sim -sim2 @gol
432 -mmul-bug-workaround -mno-mul-bug-workaround}
434 @emph{Darwin Options}
435 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
436 -arch_only -bind_at_load -bundle -bundle_loader @gol
437 -client_name -compatibility_version -current_version @gol
439 -dependency-file -dylib_file -dylinker_install_name @gol
440 -dynamic -dynamiclib -exported_symbols_list @gol
441 -filelist -flat_namespace -force_cpusubtype_ALL @gol
442 -force_flat_namespace -headerpad_max_install_names @gol
443 -image_base -init -install_name -keep_private_externs @gol
444 -multi_module -multiply_defined -multiply_defined_unused @gol
445 -noall_load -no_dead_strip_inits_and_terms @gol
446 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
447 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
448 -private_bundle -read_only_relocs -sectalign @gol
449 -sectobjectsymbols -whyload -seg1addr @gol
450 -sectcreate -sectobjectsymbols -sectorder @gol
451 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
452 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
453 -segprot -segs_read_only_addr -segs_read_write_addr @gol
454 -single_module -static -sub_library -sub_umbrella @gol
455 -twolevel_namespace -umbrella -undefined @gol
456 -unexported_symbols_list -weak_reference_mismatches @gol
457 -whatsloaded -F -gused -gfull -mmacosx-min-version=@var{version} @gol
460 @emph{DEC Alpha Options}
461 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
462 -mieee -mieee-with-inexact -mieee-conformant @gol
463 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
464 -mtrap-precision=@var{mode} -mbuild-constants @gol
465 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
466 -mbwx -mmax -mfix -mcix @gol
467 -mfloat-vax -mfloat-ieee @gol
468 -mexplicit-relocs -msmall-data -mlarge-data @gol
469 -msmall-text -mlarge-text @gol
470 -mmemory-latency=@var{time}}
472 @emph{DEC Alpha/VMS Options}
473 @gccoptlist{-mvms-return-codes}
476 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
477 -mhard-float -msoft-float @gol
478 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
479 -mdouble -mno-double @gol
480 -mmedia -mno-media -mmuladd -mno-muladd @gol
481 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
482 -mlinked-fp -mlong-calls -malign-labels @gol
483 -mlibrary-pic -macc-4 -macc-8 @gol
484 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
485 -moptimize-membar -mno-optimize-membar @gol
486 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
487 -mvliw-branch -mno-vliw-branch @gol
488 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
489 -mno-nested-cond-exec -mtomcat-stats @gol
493 @emph{H8/300 Options}
494 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
497 @gccoptlist{-march=@var{architecture-type} @gol
498 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
499 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
500 -mfixed-range=@var{register-range} @gol
501 -mjump-in-delay -mlinker-opt -mlong-calls @gol
502 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
503 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
504 -mno-jump-in-delay -mno-long-load-store @gol
505 -mno-portable-runtime -mno-soft-float @gol
506 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
507 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
508 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
509 -munix=@var{unix-std} -nolibdld -static -threads}
511 @emph{i386 and x86-64 Options}
512 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
513 -mfpmath=@var{unit} @gol
514 -masm=@var{dialect} -mno-fancy-math-387 @gol
515 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
516 -mno-wide-multiply -mrtd -malign-double @gol
517 -mpreferred-stack-boundary=@var{num} @gol
518 -mmmx -msse -msse2 -msse3 -m3dnow @gol
519 -mthreads -mno-align-stringops -minline-all-stringops @gol
520 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
521 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
522 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
523 -mcmodel=@var{code-model} @gol
527 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
528 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
529 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
530 -minline-float-divide-max-throughput @gol
531 -minline-int-divide-min-latency @gol
532 -minline-int-divide-max-throughput @gol
533 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
534 -mno-dwarf2-asm -mearly-stop-bits @gol
535 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
536 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64}
538 @emph{M32R/D Options}
539 @gccoptlist{-m32r2 -m32rx -m32r @gol
541 -malign-loops -mno-align-loops @gol
542 -missue-rate=@var{number} @gol
543 -mbranch-cost=@var{number} @gol
544 -mmodel=@var{code-size-model-type} @gol
545 -msdata=@var{sdata-type} @gol
546 -mno-flush-func -mflush-func=@var{name} @gol
547 -mno-flush-trap -mflush-trap=@var{number} @gol
551 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
553 @emph{M680x0 Options}
554 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
555 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
556 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
557 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
558 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
560 @emph{M68hc1x Options}
561 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
562 -mauto-incdec -minmax -mlong-calls -mshort @gol
563 -msoft-reg-count=@var{count}}
566 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
567 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
568 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
569 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
570 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
573 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
574 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
575 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
576 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
577 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
578 -mdsp -mpaired-single -mips3d @gol
579 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
580 -G@var{num} -membedded-data -mno-embedded-data @gol
581 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
582 -msplit-addresses -mno-split-addresses @gol
583 -mexplicit-relocs -mno-explicit-relocs @gol
584 -mcheck-zero-division -mno-check-zero-division @gol
585 -mdivide-traps -mdivide-breaks @gol
586 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
587 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
588 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
589 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
590 -mfix-sb1 -mno-fix-sb1 @gol
591 -mflush-func=@var{func} -mno-flush-func @gol
592 -mbranch-likely -mno-branch-likely @gol
593 -mfp-exceptions -mno-fp-exceptions @gol
594 -mvr4130-align -mno-vr4130-align}
597 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
598 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
599 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
600 -mno-base-addresses -msingle-exit -mno-single-exit}
602 @emph{MN10300 Options}
603 @gccoptlist{-mmult-bug -mno-mult-bug @gol
604 -mam33 -mno-am33 @gol
605 -mam33-2 -mno-am33-2 @gol
608 @emph{PDP-11 Options}
609 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
610 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
611 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
612 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
613 -mbranch-expensive -mbranch-cheap @gol
614 -msplit -mno-split -munix-asm -mdec-asm}
616 @emph{PowerPC Options}
617 See RS/6000 and PowerPC Options.
619 @emph{RS/6000 and PowerPC Options}
620 @gccoptlist{-mcpu=@var{cpu-type} @gol
621 -mtune=@var{cpu-type} @gol
622 -mpower -mno-power -mpower2 -mno-power2 @gol
623 -mpowerpc -mpowerpc64 -mno-powerpc @gol
624 -maltivec -mno-altivec @gol
625 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
626 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
627 -mnew-mnemonics -mold-mnemonics @gol
628 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
629 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
630 -malign-power -malign-natural @gol
631 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
632 -mstring -mno-string -mupdate -mno-update @gol
633 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
634 -mstrict-align -mno-strict-align -mrelocatable @gol
635 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
636 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
637 -mdynamic-no-pic -maltivec -mswdiv @gol
638 -mprioritize-restricted-insns=@var{priority} @gol
639 -msched-costly-dep=@var{dependence_type} @gol
640 -minsert-sched-nops=@var{scheme} @gol
641 -mcall-sysv -mcall-netbsd @gol
642 -maix-struct-return -msvr4-struct-return @gol
643 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
644 -misel -mno-isel @gol
645 -misel=yes -misel=no @gol
647 -mspe=yes -mspe=no @gol
648 -mvrsave -mno-vrsave @gol
649 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
650 -mprototype -mno-prototype @gol
651 -msim -mmvme -mads -myellowknife -memb -msdata @gol
652 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
654 @emph{S/390 and zSeries Options}
655 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
656 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
657 -mpacked-stack -mno-packed-stack @gol
658 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
659 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
660 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
661 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
664 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
665 -m4-nofpu -m4-single-only -m4-single -m4 @gol
666 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
667 -m5-64media -m5-64media-nofpu @gol
668 -m5-32media -m5-32media-nofpu @gol
669 -m5-compact -m5-compact-nofpu @gol
670 -mb -ml -mdalign -mrelax @gol
671 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
672 -mieee -misize -mpadstruct -mspace @gol
673 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
674 -mdivsi3_libfunc=@var{name} @gol
675 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
679 @gccoptlist{-mcpu=@var{cpu-type} @gol
680 -mtune=@var{cpu-type} @gol
681 -mcmodel=@var{code-model} @gol
682 -m32 -m64 -mapp-regs -mno-app-regs @gol
683 -mfaster-structs -mno-faster-structs @gol
684 -mfpu -mno-fpu -mhard-float -msoft-float @gol
685 -mhard-quad-float -msoft-quad-float @gol
686 -mimpure-text -mno-impure-text -mlittle-endian @gol
687 -mstack-bias -mno-stack-bias @gol
688 -munaligned-doubles -mno-unaligned-doubles @gol
689 -mv8plus -mno-v8plus -mvis -mno-vis
692 @emph{System V Options}
693 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
695 @emph{TMS320C3x/C4x Options}
696 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
697 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
698 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
699 -mparallel-insns -mparallel-mpy -mpreserve-float}
702 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
703 -mprolog-function -mno-prolog-function -mspace @gol
704 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
705 -mapp-regs -mno-app-regs @gol
706 -mdisable-callt -mno-disable-callt @gol
712 @gccoptlist{-mg -mgnu -munix}
714 @emph{x86-64 Options}
715 See i386 and x86-64 Options.
717 @emph{Xstormy16 Options}
720 @emph{Xtensa Options}
721 @gccoptlist{-mconst16 -mno-const16 @gol
722 -mfused-madd -mno-fused-madd @gol
723 -mtext-section-literals -mno-text-section-literals @gol
724 -mtarget-align -mno-target-align @gol
725 -mlongcalls -mno-longcalls}
727 @emph{zSeries Options}
728 See S/390 and zSeries Options.
730 @item Code Generation Options
731 @xref{Code Gen Options,,Options for Code Generation Conventions}.
732 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
733 -ffixed-@var{reg} -fexceptions @gol
734 -fnon-call-exceptions -funwind-tables @gol
735 -fasynchronous-unwind-tables @gol
736 -finhibit-size-directive -finstrument-functions @gol
737 -fno-common -fno-ident @gol
738 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
739 -fno-jump-tables @gol
740 -freg-struct-return -fshared-data -fshort-enums @gol
741 -fshort-double -fshort-wchar @gol
742 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
743 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
744 -fargument-alias -fargument-noalias @gol
745 -fargument-noalias-global -fleading-underscore @gol
746 -ftls-model=@var{model} @gol
747 -ftrapv -fwrapv -fbounds-check @gol
752 * Overall Options:: Controlling the kind of output:
753 an executable, object files, assembler files,
754 or preprocessed source.
755 * C Dialect Options:: Controlling the variant of C language compiled.
756 * C++ Dialect Options:: Variations on C++.
757 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
759 * Language Independent Options:: Controlling how diagnostics should be
761 * Warning Options:: How picky should the compiler be?
762 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
763 * Optimize Options:: How much optimization?
764 * Preprocessor Options:: Controlling header files and macro definitions.
765 Also, getting dependency information for Make.
766 * Assembler Options:: Passing options to the assembler.
767 * Link Options:: Specifying libraries and so on.
768 * Directory Options:: Where to find header files and libraries.
769 Where to find the compiler executable files.
770 * Spec Files:: How to pass switches to sub-processes.
771 * Target Options:: Running a cross-compiler, or an old version of GCC.
774 @node Overall Options
775 @section Options Controlling the Kind of Output
777 Compilation can involve up to four stages: preprocessing, compilation
778 proper, assembly and linking, always in that order. GCC is capable of
779 preprocessing and compiling several files either into several
780 assembler input files, or into one assembler input file; then each
781 assembler input file produces an object file, and linking combines all
782 the object files (those newly compiled, and those specified as input)
783 into an executable file.
785 @cindex file name suffix
786 For any given input file, the file name suffix determines what kind of
791 C source code which must be preprocessed.
794 C source code which should not be preprocessed.
797 C++ source code which should not be preprocessed.
800 Objective-C source code. Note that you must link with the @file{libobjc}
801 library to make an Objective-C program work.
804 Objective-C source code which should not be preprocessed.
808 Objective-C++ source code. Note that you must link with the @file{libobjc}
809 library to make an Objective-C++ program work. Note that @samp{.M} refers
810 to a literal capital M@.
813 Objective-C++ source code which should not be preprocessed.
816 C, C++, Objective-C or Objective-C++ header file to be turned into a
821 @itemx @var{file}.cxx
822 @itemx @var{file}.cpp
823 @itemx @var{file}.CPP
824 @itemx @var{file}.c++
826 C++ source code which must be preprocessed. Note that in @samp{.cxx},
827 the last two letters must both be literally @samp{x}. Likewise,
828 @samp{.C} refers to a literal capital C@.
832 Objective-C++ source code which must be preprocessed.
835 Objective-C++ source code which should not be preprocessed.
839 C++ header file to be turned into a precompiled header.
842 @itemx @var{file}.for
843 @itemx @var{file}.FOR
844 Fortran source code which should not be preprocessed.
847 @itemx @var{file}.fpp
848 @itemx @var{file}.FPP
849 Fortran source code which must be preprocessed (with the traditional
853 Fortran source code which must be preprocessed with a RATFOR
854 preprocessor (not included with GCC)@.
857 @itemx @var{file}.f95
858 Fortran 90/95 source code which should not be preprocessed.
860 @c FIXME: Descriptions of Java file types.
867 Ada source code file which contains a library unit declaration (a
868 declaration of a package, subprogram, or generic, or a generic
869 instantiation), or a library unit renaming declaration (a package,
870 generic, or subprogram renaming declaration). Such files are also
873 @itemx @var{file}.adb
874 Ada source code file containing a library unit body (a subprogram or
875 package body). Such files are also called @dfn{bodies}.
877 @c GCC also knows about some suffixes for languages not yet included:
886 Assembler code which must be preprocessed.
889 An object file to be fed straight into linking.
890 Any file name with no recognized suffix is treated this way.
894 You can specify the input language explicitly with the @option{-x} option:
897 @item -x @var{language}
898 Specify explicitly the @var{language} for the following input files
899 (rather than letting the compiler choose a default based on the file
900 name suffix). This option applies to all following input files until
901 the next @option{-x} option. Possible values for @var{language} are:
903 c c-header c-cpp-output
904 c++ c++-header c++-cpp-output
905 objective-c objective-c-header objective-c-cpp-output
906 objective-c++ objective-c++-header objective-c++-cpp-output
907 assembler assembler-with-cpp
909 f77 f77-cpp-input ratfor
916 Turn off any specification of a language, so that subsequent files are
917 handled according to their file name suffixes (as they are if @option{-x}
918 has not been used at all).
920 @item -pass-exit-codes
921 @opindex pass-exit-codes
922 Normally the @command{gcc} program will exit with the code of 1 if any
923 phase of the compiler returns a non-success return code. If you specify
924 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
925 numerically highest error produced by any phase that returned an error
929 If you only want some of the stages of compilation, you can use
930 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
931 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
932 @command{gcc} is to stop. Note that some combinations (for example,
933 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
938 Compile or assemble the source files, but do not link. The linking
939 stage simply is not done. The ultimate output is in the form of an
940 object file for each source file.
942 By default, the object file name for a source file is made by replacing
943 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
945 Unrecognized input files, not requiring compilation or assembly, are
950 Stop after the stage of compilation proper; do not assemble. The output
951 is in the form of an assembler code file for each non-assembler input
954 By default, the assembler file name for a source file is made by
955 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
957 Input files that don't require compilation are ignored.
961 Stop after the preprocessing stage; do not run the compiler proper. The
962 output is in the form of preprocessed source code, which is sent to the
965 Input files which don't require preprocessing are ignored.
967 @cindex output file option
970 Place output in file @var{file}. This applies regardless to whatever
971 sort of output is being produced, whether it be an executable file,
972 an object file, an assembler file or preprocessed C code.
974 If @option{-o} is not specified, the default is to put an executable
975 file in @file{a.out}, the object file for
976 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
977 assembler file in @file{@var{source}.s}, a precompiled header file in
978 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
983 Print (on standard error output) the commands executed to run the stages
984 of compilation. Also print the version number of the compiler driver
985 program and of the preprocessor and the compiler proper.
989 Like @option{-v} except the commands are not executed and all command
990 arguments are quoted. This is useful for shell scripts to capture the
991 driver-generated command lines.
995 Use pipes rather than temporary files for communication between the
996 various stages of compilation. This fails to work on some systems where
997 the assembler is unable to read from a pipe; but the GNU assembler has
1002 If you are compiling multiple source files, this option tells the driver
1003 to pass all the source files to the compiler at once (for those
1004 languages for which the compiler can handle this). This will allow
1005 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1006 language for which this is supported is C@. If you pass source files for
1007 multiple languages to the driver, using this option, the driver will invoke
1008 the compiler(s) that support IMA once each, passing each compiler all the
1009 source files appropriate for it. For those languages that do not support
1010 IMA this option will be ignored, and the compiler will be invoked once for
1011 each source file in that language. If you use this option in conjunction
1012 with @option{-save-temps}, the compiler will generate multiple
1014 (one for each source file), but only one (combined) @file{.o} or
1019 Print (on the standard output) a description of the command line options
1020 understood by @command{gcc}. If the @option{-v} option is also specified
1021 then @option{--help} will also be passed on to the various processes
1022 invoked by @command{gcc}, so that they can display the command line options
1023 they accept. If the @option{-Wextra} option is also specified then command
1024 line options which have no documentation associated with them will also
1028 @opindex target-help
1029 Print (on the standard output) a description of target specific command
1030 line options for each tool.
1034 Display the version number and copyrights of the invoked GCC@.
1038 @section Compiling C++ Programs
1040 @cindex suffixes for C++ source
1041 @cindex C++ source file suffixes
1042 C++ source files conventionally use one of the suffixes @samp{.C},
1043 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1044 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1045 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1046 files with these names and compiles them as C++ programs even if you
1047 call the compiler the same way as for compiling C programs (usually
1048 with the name @command{gcc}).
1052 However, C++ programs often require class libraries as well as a
1053 compiler that understands the C++ language---and under some
1054 circumstances, you might want to compile programs or header files from
1055 standard input, or otherwise without a suffix that flags them as C++
1056 programs. You might also like to precompile a C header file with a
1057 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
1058 program that calls GCC with the default language set to C++, and
1059 automatically specifies linking against the C++ library. On many
1060 systems, @command{g++} is also installed with the name @command{c++}.
1062 @cindex invoking @command{g++}
1063 When you compile C++ programs, you may specify many of the same
1064 command-line options that you use for compiling programs in any
1065 language; or command-line options meaningful for C and related
1066 languages; or options that are meaningful only for C++ programs.
1067 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1068 explanations of options for languages related to C@.
1069 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1070 explanations of options that are meaningful only for C++ programs.
1072 @node C Dialect Options
1073 @section Options Controlling C Dialect
1074 @cindex dialect options
1075 @cindex language dialect options
1076 @cindex options, dialect
1078 The following options control the dialect of C (or languages derived
1079 from C, such as C++, Objective-C and Objective-C++) that the compiler
1083 @cindex ANSI support
1087 In C mode, support all ISO C90 programs. In C++ mode,
1088 remove GNU extensions that conflict with ISO C++.
1090 This turns off certain features of GCC that are incompatible with ISO
1091 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1092 such as the @code{asm} and @code{typeof} keywords, and
1093 predefined macros such as @code{unix} and @code{vax} that identify the
1094 type of system you are using. It also enables the undesirable and
1095 rarely used ISO trigraph feature. For the C compiler,
1096 it disables recognition of C++ style @samp{//} comments as well as
1097 the @code{inline} keyword.
1099 The alternate keywords @code{__asm__}, @code{__extension__},
1100 @code{__inline__} and @code{__typeof__} continue to work despite
1101 @option{-ansi}. You would not want to use them in an ISO C program, of
1102 course, but it is useful to put them in header files that might be included
1103 in compilations done with @option{-ansi}. Alternate predefined macros
1104 such as @code{__unix__} and @code{__vax__} are also available, with or
1105 without @option{-ansi}.
1107 The @option{-ansi} option does not cause non-ISO programs to be
1108 rejected gratuitously. For that, @option{-pedantic} is required in
1109 addition to @option{-ansi}. @xref{Warning Options}.
1111 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1112 option is used. Some header files may notice this macro and refrain
1113 from declaring certain functions or defining certain macros that the
1114 ISO standard doesn't call for; this is to avoid interfering with any
1115 programs that might use these names for other things.
1117 Functions which would normally be built in but do not have semantics
1118 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1119 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1120 built-in functions provided by GCC}, for details of the functions
1125 Determine the language standard. This option is currently only
1126 supported when compiling C or C++. A value for this option must be
1127 provided; possible values are
1132 ISO C90 (same as @option{-ansi}).
1134 @item iso9899:199409
1135 ISO C90 as modified in amendment 1.
1141 ISO C99. Note that this standard is not yet fully supported; see
1142 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1143 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1146 Default, ISO C90 plus GNU extensions (including some C99 features).
1150 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1151 this will become the default. The name @samp{gnu9x} is deprecated.
1154 The 1998 ISO C++ standard plus amendments.
1157 The same as @option{-std=c++98} plus GNU extensions. This is the
1158 default for C++ code.
1161 Even when this option is not specified, you can still use some of the
1162 features of newer standards in so far as they do not conflict with
1163 previous C standards. For example, you may use @code{__restrict__} even
1164 when @option{-std=c99} is not specified.
1166 The @option{-std} options specifying some version of ISO C have the same
1167 effects as @option{-ansi}, except that features that were not in ISO C90
1168 but are in the specified version (for example, @samp{//} comments and
1169 the @code{inline} keyword in ISO C99) are not disabled.
1171 @xref{Standards,,Language Standards Supported by GCC}, for details of
1172 these standard versions.
1174 @item -aux-info @var{filename}
1176 Output to the given filename prototyped declarations for all functions
1177 declared and/or defined in a translation unit, including those in header
1178 files. This option is silently ignored in any language other than C@.
1180 Besides declarations, the file indicates, in comments, the origin of
1181 each declaration (source file and line), whether the declaration was
1182 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1183 @samp{O} for old, respectively, in the first character after the line
1184 number and the colon), and whether it came from a declaration or a
1185 definition (@samp{C} or @samp{F}, respectively, in the following
1186 character). In the case of function definitions, a K&R-style list of
1187 arguments followed by their declarations is also provided, inside
1188 comments, after the declaration.
1192 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1193 keyword, so that code can use these words as identifiers. You can use
1194 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1195 instead. @option{-ansi} implies @option{-fno-asm}.
1197 In C++, this switch only affects the @code{typeof} keyword, since
1198 @code{asm} and @code{inline} are standard keywords. You may want to
1199 use the @option{-fno-gnu-keywords} flag instead, which has the same
1200 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1201 switch only affects the @code{asm} and @code{typeof} keywords, since
1202 @code{inline} is a standard keyword in ISO C99.
1205 @itemx -fno-builtin-@var{function}
1206 @opindex fno-builtin
1207 @cindex built-in functions
1208 Don't recognize built-in functions that do not begin with
1209 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1210 functions provided by GCC}, for details of the functions affected,
1211 including those which are not built-in functions when @option{-ansi} or
1212 @option{-std} options for strict ISO C conformance are used because they
1213 do not have an ISO standard meaning.
1215 GCC normally generates special code to handle certain built-in functions
1216 more efficiently; for instance, calls to @code{alloca} may become single
1217 instructions that adjust the stack directly, and calls to @code{memcpy}
1218 may become inline copy loops. The resulting code is often both smaller
1219 and faster, but since the function calls no longer appear as such, you
1220 cannot set a breakpoint on those calls, nor can you change the behavior
1221 of the functions by linking with a different library. In addition,
1222 when a function is recognized as a built-in function, GCC may use
1223 information about that function to warn about problems with calls to
1224 that function, or to generate more efficient code, even if the
1225 resulting code still contains calls to that function. For example,
1226 warnings are given with @option{-Wformat} for bad calls to
1227 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1228 known not to modify global memory.
1230 With the @option{-fno-builtin-@var{function}} option
1231 only the built-in function @var{function} is
1232 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1233 function is named this is not built-in in this version of GCC, this
1234 option is ignored. There is no corresponding
1235 @option{-fbuiltin-@var{function}} option; if you wish to enable
1236 built-in functions selectively when using @option{-fno-builtin} or
1237 @option{-ffreestanding}, you may define macros such as:
1240 #define abs(n) __builtin_abs ((n))
1241 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1246 @cindex hosted environment
1248 Assert that compilation takes place in a hosted environment. This implies
1249 @option{-fbuiltin}. A hosted environment is one in which the
1250 entire standard library is available, and in which @code{main} has a return
1251 type of @code{int}. Examples are nearly everything except a kernel.
1252 This is equivalent to @option{-fno-freestanding}.
1254 @item -ffreestanding
1255 @opindex ffreestanding
1256 @cindex hosted environment
1258 Assert that compilation takes place in a freestanding environment. This
1259 implies @option{-fno-builtin}. A freestanding environment
1260 is one in which the standard library may not exist, and program startup may
1261 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1262 This is equivalent to @option{-fno-hosted}.
1264 @xref{Standards,,Language Standards Supported by GCC}, for details of
1265 freestanding and hosted environments.
1267 @item -fms-extensions
1268 @opindex fms-extensions
1269 Accept some non-standard constructs used in Microsoft header files.
1271 Some cases of unnamed fields in structures and unions are only
1272 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1273 fields within structs/unions}, for details.
1277 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1278 options for strict ISO C conformance) implies @option{-trigraphs}.
1280 @item -no-integrated-cpp
1281 @opindex no-integrated-cpp
1282 Performs a compilation in two passes: preprocessing and compiling. This
1283 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1284 @option{-B} option. The user supplied compilation step can then add in
1285 an additional preprocessing step after normal preprocessing but before
1286 compiling. The default is to use the integrated cpp (internal cpp)
1288 The semantics of this option will change if "cc1", "cc1plus", and
1289 "cc1obj" are merged.
1291 @cindex traditional C language
1292 @cindex C language, traditional
1294 @itemx -traditional-cpp
1295 @opindex traditional-cpp
1296 @opindex traditional
1297 Formerly, these options caused GCC to attempt to emulate a pre-standard
1298 C compiler. They are now only supported with the @option{-E} switch.
1299 The preprocessor continues to support a pre-standard mode. See the GNU
1300 CPP manual for details.
1302 @item -fcond-mismatch
1303 @opindex fcond-mismatch
1304 Allow conditional expressions with mismatched types in the second and
1305 third arguments. The value of such an expression is void. This option
1306 is not supported for C++.
1308 @item -funsigned-char
1309 @opindex funsigned-char
1310 Let the type @code{char} be unsigned, like @code{unsigned char}.
1312 Each kind of machine has a default for what @code{char} should
1313 be. It is either like @code{unsigned char} by default or like
1314 @code{signed char} by default.
1316 Ideally, a portable program should always use @code{signed char} or
1317 @code{unsigned char} when it depends on the signedness of an object.
1318 But many programs have been written to use plain @code{char} and
1319 expect it to be signed, or expect it to be unsigned, depending on the
1320 machines they were written for. This option, and its inverse, let you
1321 make such a program work with the opposite default.
1323 The type @code{char} is always a distinct type from each of
1324 @code{signed char} or @code{unsigned char}, even though its behavior
1325 is always just like one of those two.
1328 @opindex fsigned-char
1329 Let the type @code{char} be signed, like @code{signed char}.
1331 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1332 the negative form of @option{-funsigned-char}. Likewise, the option
1333 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1335 @item -fsigned-bitfields
1336 @itemx -funsigned-bitfields
1337 @itemx -fno-signed-bitfields
1338 @itemx -fno-unsigned-bitfields
1339 @opindex fsigned-bitfields
1340 @opindex funsigned-bitfields
1341 @opindex fno-signed-bitfields
1342 @opindex fno-unsigned-bitfields
1343 These options control whether a bit-field is signed or unsigned, when the
1344 declaration does not use either @code{signed} or @code{unsigned}. By
1345 default, such a bit-field is signed, because this is consistent: the
1346 basic integer types such as @code{int} are signed types.
1349 @node C++ Dialect Options
1350 @section Options Controlling C++ Dialect
1352 @cindex compiler options, C++
1353 @cindex C++ options, command line
1354 @cindex options, C++
1355 This section describes the command-line options that are only meaningful
1356 for C++ programs; but you can also use most of the GNU compiler options
1357 regardless of what language your program is in. For example, you
1358 might compile a file @code{firstClass.C} like this:
1361 g++ -g -frepo -O -c firstClass.C
1365 In this example, only @option{-frepo} is an option meant
1366 only for C++ programs; you can use the other options with any
1367 language supported by GCC@.
1369 Here is a list of options that are @emph{only} for compiling C++ programs:
1373 @item -fabi-version=@var{n}
1374 @opindex fabi-version
1375 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1376 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1377 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1378 the version that conforms most closely to the C++ ABI specification.
1379 Therefore, the ABI obtained using version 0 will change as ABI bugs
1382 The default is version 2.
1384 @item -fno-access-control
1385 @opindex fno-access-control
1386 Turn off all access checking. This switch is mainly useful for working
1387 around bugs in the access control code.
1391 Check that the pointer returned by @code{operator new} is non-null
1392 before attempting to modify the storage allocated. This check is
1393 normally unnecessary because the C++ standard specifies that
1394 @code{operator new} will only return @code{0} if it is declared
1395 @samp{throw()}, in which case the compiler will always check the
1396 return value even without this option. In all other cases, when
1397 @code{operator new} has a non-empty exception specification, memory
1398 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1399 @samp{new (nothrow)}.
1401 @item -fconserve-space
1402 @opindex fconserve-space
1403 Put uninitialized or runtime-initialized global variables into the
1404 common segment, as C does. This saves space in the executable at the
1405 cost of not diagnosing duplicate definitions. If you compile with this
1406 flag and your program mysteriously crashes after @code{main()} has
1407 completed, you may have an object that is being destroyed twice because
1408 two definitions were merged.
1410 This option is no longer useful on most targets, now that support has
1411 been added for putting variables into BSS without making them common.
1413 @item -fno-const-strings
1414 @opindex fno-const-strings
1415 Give string constants type @code{char *} instead of type @code{const
1416 char *}. By default, G++ uses type @code{const char *} as required by
1417 the standard. Even if you use @option{-fno-const-strings}, you cannot
1418 actually modify the value of a string constant.
1420 This option might be removed in a future release of G++. For maximum
1421 portability, you should structure your code so that it works with
1422 string constants that have type @code{const char *}.
1424 @item -fno-elide-constructors
1425 @opindex fno-elide-constructors
1426 The C++ standard allows an implementation to omit creating a temporary
1427 which is only used to initialize another object of the same type.
1428 Specifying this option disables that optimization, and forces G++ to
1429 call the copy constructor in all cases.
1431 @item -fno-enforce-eh-specs
1432 @opindex fno-enforce-eh-specs
1433 Don't check for violation of exception specifications at runtime. This
1434 option violates the C++ standard, but may be useful for reducing code
1435 size in production builds, much like defining @samp{NDEBUG}. The compiler
1436 will still optimize based on the exception specifications.
1439 @itemx -fno-for-scope
1441 @opindex fno-for-scope
1442 If @option{-ffor-scope} is specified, the scope of variables declared in
1443 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1444 as specified by the C++ standard.
1445 If @option{-fno-for-scope} is specified, the scope of variables declared in
1446 a @i{for-init-statement} extends to the end of the enclosing scope,
1447 as was the case in old versions of G++, and other (traditional)
1448 implementations of C++.
1450 The default if neither flag is given to follow the standard,
1451 but to allow and give a warning for old-style code that would
1452 otherwise be invalid, or have different behavior.
1454 @item -fno-gnu-keywords
1455 @opindex fno-gnu-keywords
1456 Do not recognize @code{typeof} as a keyword, so that code can use this
1457 word as an identifier. You can use the keyword @code{__typeof__} instead.
1458 @option{-ansi} implies @option{-fno-gnu-keywords}.
1460 @item -fno-implicit-templates
1461 @opindex fno-implicit-templates
1462 Never emit code for non-inline templates which are instantiated
1463 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1464 @xref{Template Instantiation}, for more information.
1466 @item -fno-implicit-inline-templates
1467 @opindex fno-implicit-inline-templates
1468 Don't emit code for implicit instantiations of inline templates, either.
1469 The default is to handle inlines differently so that compiles with and
1470 without optimization will need the same set of explicit instantiations.
1472 @item -fno-implement-inlines
1473 @opindex fno-implement-inlines
1474 To save space, do not emit out-of-line copies of inline functions
1475 controlled by @samp{#pragma implementation}. This will cause linker
1476 errors if these functions are not inlined everywhere they are called.
1478 @item -fms-extensions
1479 @opindex fms-extensions
1480 Disable pedantic warnings about constructs used in MFC, such as implicit
1481 int and getting a pointer to member function via non-standard syntax.
1483 @item -fno-nonansi-builtins
1484 @opindex fno-nonansi-builtins
1485 Disable built-in declarations of functions that are not mandated by
1486 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1487 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1489 @item -fno-operator-names
1490 @opindex fno-operator-names
1491 Do not treat the operator name keywords @code{and}, @code{bitand},
1492 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1493 synonyms as keywords.
1495 @item -fno-optional-diags
1496 @opindex fno-optional-diags
1497 Disable diagnostics that the standard says a compiler does not need to
1498 issue. Currently, the only such diagnostic issued by G++ is the one for
1499 a name having multiple meanings within a class.
1502 @opindex fpermissive
1503 Downgrade some diagnostics about nonconformant code from errors to
1504 warnings. Thus, using @option{-fpermissive} will allow some
1505 nonconforming code to compile.
1509 Enable automatic template instantiation at link time. This option also
1510 implies @option{-fno-implicit-templates}. @xref{Template
1511 Instantiation}, for more information.
1515 Disable generation of information about every class with virtual
1516 functions for use by the C++ runtime type identification features
1517 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1518 of the language, you can save some space by using this flag. Note that
1519 exception handling uses the same information, but it will generate it as
1524 Emit statistics about front-end processing at the end of the compilation.
1525 This information is generally only useful to the G++ development team.
1527 @item -ftemplate-depth-@var{n}
1528 @opindex ftemplate-depth
1529 Set the maximum instantiation depth for template classes to @var{n}.
1530 A limit on the template instantiation depth is needed to detect
1531 endless recursions during template class instantiation. ANSI/ISO C++
1532 conforming programs must not rely on a maximum depth greater than 17.
1534 @item -fno-threadsafe-statics
1535 @opindex fno-threadsafe-statics
1536 Do not emit the extra code to use the routines specified in the C++
1537 ABI for thread-safe initialization of local statics. You can use this
1538 option to reduce code size slightly in code that doesn't need to be
1541 @item -fuse-cxa-atexit
1542 @opindex fuse-cxa-atexit
1543 Register destructors for objects with static storage duration with the
1544 @code{__cxa_atexit} function rather than the @code{atexit} function.
1545 This option is required for fully standards-compliant handling of static
1546 destructors, but will only work if your C library supports
1547 @code{__cxa_atexit}.
1549 @item -fvisibility-inlines-hidden
1550 @opindex fvisibility-inlines-hidden
1551 Causes all inlined methods to be marked with
1552 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1553 appear in the export table of a DSO and do not require a PLT indirection
1554 when used within the DSO@. Enabling this option can have a dramatic effect
1555 on load and link times of a DSO as it massively reduces the size of the
1556 dynamic export table when the library makes heavy use of templates. While
1557 it can cause bloating through duplication of code within each DSO where
1558 it is used, often the wastage is less than the considerable space occupied
1559 by a long symbol name in the export table which is typical when using
1560 templates and namespaces. For even more savings, combine with the
1561 @option{-fvisibility=hidden} switch.
1565 Do not use weak symbol support, even if it is provided by the linker.
1566 By default, G++ will use weak symbols if they are available. This
1567 option exists only for testing, and should not be used by end-users;
1568 it will result in inferior code and has no benefits. This option may
1569 be removed in a future release of G++.
1573 Do not search for header files in the standard directories specific to
1574 C++, but do still search the other standard directories. (This option
1575 is used when building the C++ library.)
1578 In addition, these optimization, warning, and code generation options
1579 have meanings only for C++ programs:
1582 @item -fno-default-inline
1583 @opindex fno-default-inline
1584 Do not assume @samp{inline} for functions defined inside a class scope.
1585 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1586 functions will have linkage like inline functions; they just won't be
1589 @item -Wabi @r{(C++ only)}
1591 Warn when G++ generates code that is probably not compatible with the
1592 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1593 all such cases, there are probably some cases that are not warned about,
1594 even though G++ is generating incompatible code. There may also be
1595 cases where warnings are emitted even though the code that is generated
1598 You should rewrite your code to avoid these warnings if you are
1599 concerned about the fact that code generated by G++ may not be binary
1600 compatible with code generated by other compilers.
1602 The known incompatibilities at this point include:
1607 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1608 pack data into the same byte as a base class. For example:
1611 struct A @{ virtual void f(); int f1 : 1; @};
1612 struct B : public A @{ int f2 : 1; @};
1616 In this case, G++ will place @code{B::f2} into the same byte
1617 as@code{A::f1}; other compilers will not. You can avoid this problem
1618 by explicitly padding @code{A} so that its size is a multiple of the
1619 byte size on your platform; that will cause G++ and other compilers to
1620 layout @code{B} identically.
1623 Incorrect handling of tail-padding for virtual bases. G++ does not use
1624 tail padding when laying out virtual bases. For example:
1627 struct A @{ virtual void f(); char c1; @};
1628 struct B @{ B(); char c2; @};
1629 struct C : public A, public virtual B @{@};
1633 In this case, G++ will not place @code{B} into the tail-padding for
1634 @code{A}; other compilers will. You can avoid this problem by
1635 explicitly padding @code{A} so that its size is a multiple of its
1636 alignment (ignoring virtual base classes); that will cause G++ and other
1637 compilers to layout @code{C} identically.
1640 Incorrect handling of bit-fields with declared widths greater than that
1641 of their underlying types, when the bit-fields appear in a union. For
1645 union U @{ int i : 4096; @};
1649 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1650 union too small by the number of bits in an @code{int}.
1653 Empty classes can be placed at incorrect offsets. For example:
1663 struct C : public B, public A @{@};
1667 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1668 it should be placed at offset zero. G++ mistakenly believes that the
1669 @code{A} data member of @code{B} is already at offset zero.
1672 Names of template functions whose types involve @code{typename} or
1673 template template parameters can be mangled incorrectly.
1676 template <typename Q>
1677 void f(typename Q::X) @{@}
1679 template <template <typename> class Q>
1680 void f(typename Q<int>::X) @{@}
1684 Instantiations of these templates may be mangled incorrectly.
1688 @item -Wctor-dtor-privacy @r{(C++ only)}
1689 @opindex Wctor-dtor-privacy
1690 Warn when a class seems unusable because all the constructors or
1691 destructors in that class are private, and it has neither friends nor
1692 public static member functions.
1694 @item -Wnon-virtual-dtor @r{(C++ only)}
1695 @opindex Wnon-virtual-dtor
1696 Warn when a class appears to be polymorphic, thereby requiring a virtual
1697 destructor, yet it declares a non-virtual one.
1698 This warning is enabled by @option{-Wall}.
1700 @item -Wreorder @r{(C++ only)}
1702 @cindex reordering, warning
1703 @cindex warning for reordering of member initializers
1704 Warn when the order of member initializers given in the code does not
1705 match the order in which they must be executed. For instance:
1711 A(): j (0), i (1) @{ @}
1715 The compiler will rearrange the member initializers for @samp{i}
1716 and @samp{j} to match the declaration order of the members, emitting
1717 a warning to that effect. This warning is enabled by @option{-Wall}.
1720 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1723 @item -Weffc++ @r{(C++ only)}
1725 Warn about violations of the following style guidelines from Scott Meyers'
1726 @cite{Effective C++} book:
1730 Item 11: Define a copy constructor and an assignment operator for classes
1731 with dynamically allocated memory.
1734 Item 12: Prefer initialization to assignment in constructors.
1737 Item 14: Make destructors virtual in base classes.
1740 Item 15: Have @code{operator=} return a reference to @code{*this}.
1743 Item 23: Don't try to return a reference when you must return an object.
1747 Also warn about violations of the following style guidelines from
1748 Scott Meyers' @cite{More Effective C++} book:
1752 Item 6: Distinguish between prefix and postfix forms of increment and
1753 decrement operators.
1756 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1760 When selecting this option, be aware that the standard library
1761 headers do not obey all of these guidelines; use @samp{grep -v}
1762 to filter out those warnings.
1764 @item -Wno-deprecated @r{(C++ only)}
1765 @opindex Wno-deprecated
1766 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1768 @item -Wstrict-null-sentinel @r{(C++ only)}
1769 @opindex Wstrict-null-sentinel
1770 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1771 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1772 to @code{__null}. Although it is a null pointer constant not a null pointer,
1773 it is guaranteed to of the same size as a pointer. But this use is
1774 not portable across different compilers.
1776 @item -Wno-non-template-friend @r{(C++ only)}
1777 @opindex Wno-non-template-friend
1778 Disable warnings when non-templatized friend functions are declared
1779 within a template. Since the advent of explicit template specification
1780 support in G++, if the name of the friend is an unqualified-id (i.e.,
1781 @samp{friend foo(int)}), the C++ language specification demands that the
1782 friend declare or define an ordinary, nontemplate function. (Section
1783 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1784 could be interpreted as a particular specialization of a templatized
1785 function. Because this non-conforming behavior is no longer the default
1786 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1787 check existing code for potential trouble spots and is on by default.
1788 This new compiler behavior can be turned off with
1789 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1790 but disables the helpful warning.
1792 @item -Wold-style-cast @r{(C++ only)}
1793 @opindex Wold-style-cast
1794 Warn if an old-style (C-style) cast to a non-void type is used within
1795 a C++ program. The new-style casts (@samp{static_cast},
1796 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1797 unintended effects and much easier to search for.
1799 @item -Woverloaded-virtual @r{(C++ only)}
1800 @opindex Woverloaded-virtual
1801 @cindex overloaded virtual fn, warning
1802 @cindex warning for overloaded virtual fn
1803 Warn when a function declaration hides virtual functions from a
1804 base class. For example, in:
1811 struct B: public A @{
1816 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1824 will fail to compile.
1826 @item -Wno-pmf-conversions @r{(C++ only)}
1827 @opindex Wno-pmf-conversions
1828 Disable the diagnostic for converting a bound pointer to member function
1831 @item -Wsign-promo @r{(C++ only)}
1832 @opindex Wsign-promo
1833 Warn when overload resolution chooses a promotion from unsigned or
1834 enumerated type to a signed type, over a conversion to an unsigned type of
1835 the same size. Previous versions of G++ would try to preserve
1836 unsignedness, but the standard mandates the current behavior.
1841 A& operator = (int);
1851 In this example, G++ will synthesize a default @samp{A& operator =
1852 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1855 @node Objective-C and Objective-C++ Dialect Options
1856 @section Options Controlling Objective-C and Objective-C++ Dialects
1858 @cindex compiler options, Objective-C and Objective-C++
1859 @cindex Objective-C and Objective-C++ options, command line
1860 @cindex options, Objective-C and Objective-C++
1861 (NOTE: This manual does not describe the Objective-C and Objective-C++
1862 languages themselves. See @xref{Standards,,Language Standards
1863 Supported by GCC}, for references.)
1865 This section describes the command-line options that are only meaningful
1866 for Objective-C and Objective-C++ programs, but you can also use most of
1867 the language-independent GNU compiler options.
1868 For example, you might compile a file @code{some_class.m} like this:
1871 gcc -g -fgnu-runtime -O -c some_class.m
1875 In this example, @option{-fgnu-runtime} is an option meant only for
1876 Objective-C and Objective-C++ programs; you can use the other options with
1877 any language supported by GCC@.
1879 Note that since Objective-C is an extension of the C language, Objective-C
1880 compilations may also use options specific to the C front-end (e.g.,
1881 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1882 C++-specific options (e.g., @option{-Wabi}).
1884 Here is a list of options that are @emph{only} for compiling Objective-C
1885 and Objective-C++ programs:
1888 @item -fconstant-string-class=@var{class-name}
1889 @opindex fconstant-string-class
1890 Use @var{class-name} as the name of the class to instantiate for each
1891 literal string specified with the syntax @code{@@"@dots{}"}. The default
1892 class name is @code{NXConstantString} if the GNU runtime is being used, and
1893 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1894 @option{-fconstant-cfstrings} option, if also present, will override the
1895 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1896 to be laid out as constant CoreFoundation strings.
1899 @opindex fgnu-runtime
1900 Generate object code compatible with the standard GNU Objective-C
1901 runtime. This is the default for most types of systems.
1903 @item -fnext-runtime
1904 @opindex fnext-runtime
1905 Generate output compatible with the NeXT runtime. This is the default
1906 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1907 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1910 @item -fno-nil-receivers
1911 @opindex fno-nil-receivers
1912 Assume that all Objective-C message dispatches (e.g.,
1913 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1914 is not @code{nil}. This allows for more efficient entry points in the runtime
1915 to be used. Currently, this option is only available in conjunction with
1916 the NeXT runtime on Mac OS X 10.3 and later.
1918 @item -fobjc-call-cxx-cdtors
1919 @opindex fobjc-call-cxx-cdtors
1920 For each Objective-C class, check if any of its instance variables is a
1921 C++ object with a non-trivial default constructor. If so, synthesize a
1922 special @code{- (id) .cxx_construct} instance method that will run
1923 non-trivial default constructors on any such instance variables, in order,
1924 and then return @code{self}. Similarly, check if any instance variable
1925 is a C++ object with a non-trivial destructor, and if so, synthesize a
1926 special @code{- (void) .cxx_destruct} method that will run
1927 all such default destructors, in reverse order.
1929 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
1930 thusly generated will only operate on instance variables declared in the
1931 current Objective-C class, and not those inherited from superclasses. It
1932 is the responsibility of the Objective-C runtime to invoke all such methods
1933 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
1934 will be invoked by the runtime immediately after a new object
1935 instance is allocated; the @code{- (void) .cxx_destruct} methods will
1936 be invoked immediately before the runtime deallocates an object instance.
1938 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
1939 support for invoking the @code{- (id) .cxx_construct} and
1940 @code{- (void) .cxx_destruct} methods.
1942 @item -fobjc-direct-dispatch
1943 @opindex fobjc-direct-dispatch
1944 Allow fast jumps to the message dispatcher. On Darwin this is
1945 accomplished via the comm page.
1947 @item -fobjc-exceptions
1948 @opindex fobjc-exceptions
1949 Enable syntactic support for structured exception handling in Objective-C,
1950 similar to what is offered by C++ and Java. Currently, this option is only
1951 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1959 @@catch (AnObjCClass *exc) @{
1966 @@catch (AnotherClass *exc) @{
1969 @@catch (id allOthers) @{
1979 The @code{@@throw} statement may appear anywhere in an Objective-C or
1980 Objective-C++ program; when used inside of a @code{@@catch} block, the
1981 @code{@@throw} may appear without an argument (as shown above), in which case
1982 the object caught by the @code{@@catch} will be rethrown.
1984 Note that only (pointers to) Objective-C objects may be thrown and
1985 caught using this scheme. When an object is thrown, it will be caught
1986 by the nearest @code{@@catch} clause capable of handling objects of that type,
1987 analogously to how @code{catch} blocks work in C++ and Java. A
1988 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1989 any and all Objective-C exceptions not caught by previous @code{@@catch}
1992 The @code{@@finally} clause, if present, will be executed upon exit from the
1993 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1994 regardless of whether any exceptions are thrown, caught or rethrown
1995 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1996 of the @code{finally} clause in Java.
1998 There are several caveats to using the new exception mechanism:
2002 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2003 idioms provided by the @code{NSException} class, the new
2004 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2005 systems, due to additional functionality needed in the (NeXT) Objective-C
2009 As mentioned above, the new exceptions do not support handling
2010 types other than Objective-C objects. Furthermore, when used from
2011 Objective-C++, the Objective-C exception model does not interoperate with C++
2012 exceptions at this time. This means you cannot @code{@@throw} an exception
2013 from Objective-C and @code{catch} it in C++, or vice versa
2014 (i.e., @code{throw @dots{} @@catch}).
2017 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2018 blocks for thread-safe execution:
2021 @@synchronized (ObjCClass *guard) @{
2026 Upon entering the @code{@@synchronized} block, a thread of execution shall
2027 first check whether a lock has been placed on the corresponding @code{guard}
2028 object by another thread. If it has, the current thread shall wait until
2029 the other thread relinquishes its lock. Once @code{guard} becomes available,
2030 the current thread will place its own lock on it, execute the code contained in
2031 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2032 making @code{guard} available to other threads).
2034 Unlike Java, Objective-C does not allow for entire methods to be marked
2035 @code{@@synchronized}. Note that throwing exceptions out of
2036 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2037 to be unlocked properly.
2041 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2043 @item -freplace-objc-classes
2044 @opindex freplace-objc-classes
2045 Emit a special marker instructing @command{ld(1)} not to statically link in
2046 the resulting object file, and allow @command{dyld(1)} to load it in at
2047 run time instead. This is used in conjunction with the Fix-and-Continue
2048 debugging mode, where the object file in question may be recompiled and
2049 dynamically reloaded in the course of program execution, without the need
2050 to restart the program itself. Currently, Fix-and-Continue functionality
2051 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2056 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2057 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2058 compile time) with static class references that get initialized at load time,
2059 which improves run-time performance. Specifying the @option{-fzero-link} flag
2060 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2061 to be retained. This is useful in Zero-Link debugging mode, since it allows
2062 for individual class implementations to be modified during program execution.
2066 Dump interface declarations for all classes seen in the source file to a
2067 file named @file{@var{sourcename}.decl}.
2069 @item -Wassign-intercept
2070 @opindex Wassign-intercept
2071 Warn whenever an Objective-C assignment is being intercepted by the
2075 @opindex Wno-protocol
2076 If a class is declared to implement a protocol, a warning is issued for
2077 every method in the protocol that is not implemented by the class. The
2078 default behavior is to issue a warning for every method not explicitly
2079 implemented in the class, even if a method implementation is inherited
2080 from the superclass. If you use the @option{-Wno-protocol} option, then
2081 methods inherited from the superclass are considered to be implemented,
2082 and no warning is issued for them.
2086 Warn if multiple methods of different types for the same selector are
2087 found during compilation. The check is performed on the list of methods
2088 in the final stage of compilation. Additionally, a check is performed
2089 for each selector appearing in a @code{@@selector(@dots{})}
2090 expression, and a corresponding method for that selector has been found
2091 during compilation. Because these checks scan the method table only at
2092 the end of compilation, these warnings are not produced if the final
2093 stage of compilation is not reached, for example because an error is
2094 found during compilation, or because the @option{-fsyntax-only} option is
2097 @item -Wstrict-selector-match
2098 @opindex Wstrict-selector-match
2099 Warn if multiple methods with differing argument and/or return types are
2100 found for a given selector when attempting to send a message using this
2101 selector to a receiver of type @code{id} or @code{Class}. When this flag
2102 is off (which is the default behavior), the compiler will omit such warnings
2103 if any differences found are confined to types which share the same size
2106 @item -Wundeclared-selector
2107 @opindex Wundeclared-selector
2108 Warn if a @code{@@selector(@dots{})} expression referring to an
2109 undeclared selector is found. A selector is considered undeclared if no
2110 method with that name has been declared before the
2111 @code{@@selector(@dots{})} expression, either explicitly in an
2112 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2113 an @code{@@implementation} section. This option always performs its
2114 checks as soon as a @code{@@selector(@dots{})} expression is found,
2115 while @option{-Wselector} only performs its checks in the final stage of
2116 compilation. This also enforces the coding style convention
2117 that methods and selectors must be declared before being used.
2119 @item -print-objc-runtime-info
2120 @opindex print-objc-runtime-info
2121 Generate C header describing the largest structure that is passed by
2126 @node Language Independent Options
2127 @section Options to Control Diagnostic Messages Formatting
2128 @cindex options to control diagnostics formatting
2129 @cindex diagnostic messages
2130 @cindex message formatting
2132 Traditionally, diagnostic messages have been formatted irrespective of
2133 the output device's aspect (e.g.@: its width, @dots{}). The options described
2134 below can be used to control the diagnostic messages formatting
2135 algorithm, e.g.@: how many characters per line, how often source location
2136 information should be reported. Right now, only the C++ front end can
2137 honor these options. However it is expected, in the near future, that
2138 the remaining front ends would be able to digest them correctly.
2141 @item -fmessage-length=@var{n}
2142 @opindex fmessage-length
2143 Try to format error messages so that they fit on lines of about @var{n}
2144 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2145 the front ends supported by GCC@. If @var{n} is zero, then no
2146 line-wrapping will be done; each error message will appear on a single
2149 @opindex fdiagnostics-show-location
2150 @item -fdiagnostics-show-location=once
2151 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2152 reporter to emit @emph{once} source location information; that is, in
2153 case the message is too long to fit on a single physical line and has to
2154 be wrapped, the source location won't be emitted (as prefix) again,
2155 over and over, in subsequent continuation lines. This is the default
2158 @item -fdiagnostics-show-location=every-line
2159 Only meaningful in line-wrapping mode. Instructs the diagnostic
2160 messages reporter to emit the same source location information (as
2161 prefix) for physical lines that result from the process of breaking
2162 a message which is too long to fit on a single line.
2164 @item -fdiagnostics-show-options
2165 @opindex fdiagnostics-show-options
2166 This option instructs the diagnostic machinery to add text to each
2167 diagnostic emitted, which indicates which command line option directly
2168 controls that diagnostic, when such an option is known to the
2169 diagnostic machinery.
2173 @node Warning Options
2174 @section Options to Request or Suppress Warnings
2175 @cindex options to control warnings
2176 @cindex warning messages
2177 @cindex messages, warning
2178 @cindex suppressing warnings
2180 Warnings are diagnostic messages that report constructions which
2181 are not inherently erroneous but which are risky or suggest there
2182 may have been an error.
2184 You can request many specific warnings with options beginning @samp{-W},
2185 for example @option{-Wimplicit} to request warnings on implicit
2186 declarations. Each of these specific warning options also has a
2187 negative form beginning @samp{-Wno-} to turn off warnings;
2188 for example, @option{-Wno-implicit}. This manual lists only one of the
2189 two forms, whichever is not the default.
2191 The following options control the amount and kinds of warnings produced
2192 by GCC; for further, language-specific options also refer to
2193 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2197 @cindex syntax checking
2199 @opindex fsyntax-only
2200 Check the code for syntax errors, but don't do anything beyond that.
2204 Issue all the warnings demanded by strict ISO C and ISO C++;
2205 reject all programs that use forbidden extensions, and some other
2206 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2207 version of the ISO C standard specified by any @option{-std} option used.
2209 Valid ISO C and ISO C++ programs should compile properly with or without
2210 this option (though a rare few will require @option{-ansi} or a
2211 @option{-std} option specifying the required version of ISO C)@. However,
2212 without this option, certain GNU extensions and traditional C and C++
2213 features are supported as well. With this option, they are rejected.
2215 @option{-pedantic} does not cause warning messages for use of the
2216 alternate keywords whose names begin and end with @samp{__}. Pedantic
2217 warnings are also disabled in the expression that follows
2218 @code{__extension__}. However, only system header files should use
2219 these escape routes; application programs should avoid them.
2220 @xref{Alternate Keywords}.
2222 Some users try to use @option{-pedantic} to check programs for strict ISO
2223 C conformance. They soon find that it does not do quite what they want:
2224 it finds some non-ISO practices, but not all---only those for which
2225 ISO C @emph{requires} a diagnostic, and some others for which
2226 diagnostics have been added.
2228 A feature to report any failure to conform to ISO C might be useful in
2229 some instances, but would require considerable additional work and would
2230 be quite different from @option{-pedantic}. We don't have plans to
2231 support such a feature in the near future.
2233 Where the standard specified with @option{-std} represents a GNU
2234 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2235 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2236 extended dialect is based. Warnings from @option{-pedantic} are given
2237 where they are required by the base standard. (It would not make sense
2238 for such warnings to be given only for features not in the specified GNU
2239 C dialect, since by definition the GNU dialects of C include all
2240 features the compiler supports with the given option, and there would be
2241 nothing to warn about.)
2243 @item -pedantic-errors
2244 @opindex pedantic-errors
2245 Like @option{-pedantic}, except that errors are produced rather than
2250 Inhibit all warning messages.
2254 Inhibit warning messages about the use of @samp{#import}.
2256 @item -Wchar-subscripts
2257 @opindex Wchar-subscripts
2258 Warn if an array subscript has type @code{char}. This is a common cause
2259 of error, as programmers often forget that this type is signed on some
2261 This warning is enabled by @option{-Wall}.
2265 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2266 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2267 This warning is enabled by @option{-Wall}.
2269 @item -Wfatal-errors
2270 @opindex Wfatal-errors
2271 This option causes the compiler to abort compilation on the first error
2272 occurred rather than trying to keep going and printing further error
2277 @opindex ffreestanding
2278 @opindex fno-builtin
2279 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2280 the arguments supplied have types appropriate to the format string
2281 specified, and that the conversions specified in the format string make
2282 sense. This includes standard functions, and others specified by format
2283 attributes (@pxref{Function Attributes}), in the @code{printf},
2284 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2285 not in the C standard) families (or other target-specific families).
2286 Which functions are checked without format attributes having been
2287 specified depends on the standard version selected, and such checks of
2288 functions without the attribute specified are disabled by
2289 @option{-ffreestanding} or @option{-fno-builtin}.
2291 The formats are checked against the format features supported by GNU
2292 libc version 2.2. These include all ISO C90 and C99 features, as well
2293 as features from the Single Unix Specification and some BSD and GNU
2294 extensions. Other library implementations may not support all these
2295 features; GCC does not support warning about features that go beyond a
2296 particular library's limitations. However, if @option{-pedantic} is used
2297 with @option{-Wformat}, warnings will be given about format features not
2298 in the selected standard version (but not for @code{strfmon} formats,
2299 since those are not in any version of the C standard). @xref{C Dialect
2300 Options,,Options Controlling C Dialect}.
2302 Since @option{-Wformat} also checks for null format arguments for
2303 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2305 @option{-Wformat} is included in @option{-Wall}. For more control over some
2306 aspects of format checking, the options @option{-Wformat-y2k},
2307 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2308 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2309 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2312 @opindex Wformat-y2k
2313 If @option{-Wformat} is specified, also warn about @code{strftime}
2314 formats which may yield only a two-digit year.
2316 @item -Wno-format-extra-args
2317 @opindex Wno-format-extra-args
2318 If @option{-Wformat} is specified, do not warn about excess arguments to a
2319 @code{printf} or @code{scanf} format function. The C standard specifies
2320 that such arguments are ignored.
2322 Where the unused arguments lie between used arguments that are
2323 specified with @samp{$} operand number specifications, normally
2324 warnings are still given, since the implementation could not know what
2325 type to pass to @code{va_arg} to skip the unused arguments. However,
2326 in the case of @code{scanf} formats, this option will suppress the
2327 warning if the unused arguments are all pointers, since the Single
2328 Unix Specification says that such unused arguments are allowed.
2330 @item -Wno-format-zero-length
2331 @opindex Wno-format-zero-length
2332 If @option{-Wformat} is specified, do not warn about zero-length formats.
2333 The C standard specifies that zero-length formats are allowed.
2335 @item -Wformat-nonliteral
2336 @opindex Wformat-nonliteral
2337 If @option{-Wformat} is specified, also warn if the format string is not a
2338 string literal and so cannot be checked, unless the format function
2339 takes its format arguments as a @code{va_list}.
2341 @item -Wformat-security
2342 @opindex Wformat-security
2343 If @option{-Wformat} is specified, also warn about uses of format
2344 functions that represent possible security problems. At present, this
2345 warns about calls to @code{printf} and @code{scanf} functions where the
2346 format string is not a string literal and there are no format arguments,
2347 as in @code{printf (foo);}. This may be a security hole if the format
2348 string came from untrusted input and contains @samp{%n}. (This is
2349 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2350 in future warnings may be added to @option{-Wformat-security} that are not
2351 included in @option{-Wformat-nonliteral}.)
2355 Enable @option{-Wformat} plus format checks not included in
2356 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2357 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2361 Warn about passing a null pointer for arguments marked as
2362 requiring a non-null value by the @code{nonnull} function attribute.
2364 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2365 can be disabled with the @option{-Wno-nonnull} option.
2367 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2369 Warn about uninitialized variables which are initialized with themselves.
2370 Note this option can only be used with the @option{-Wuninitialized} option,
2371 which in turn only works with @option{-O1} and above.
2373 For example, GCC will warn about @code{i} being uninitialized in the
2374 following snippet only when @option{-Winit-self} has been specified:
2385 @item -Wimplicit-int
2386 @opindex Wimplicit-int
2387 Warn when a declaration does not specify a type.
2388 This warning is enabled by @option{-Wall}.
2390 @item -Wimplicit-function-declaration
2391 @itemx -Werror-implicit-function-declaration
2392 @opindex Wimplicit-function-declaration
2393 @opindex Werror-implicit-function-declaration
2394 Give a warning (or error) whenever a function is used before being
2395 declared. The form @option{-Wno-error-implicit-function-declaration}
2397 This warning is enabled by @option{-Wall} (as a warning, not an error).
2401 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2402 This warning is enabled by @option{-Wall}.
2406 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2407 function with external linkage, returning int, taking either zero
2408 arguments, two, or three arguments of appropriate types.
2409 This warning is enabled by @option{-Wall}.
2411 @item -Wmissing-braces
2412 @opindex Wmissing-braces
2413 Warn if an aggregate or union initializer is not fully bracketed. In
2414 the following example, the initializer for @samp{a} is not fully
2415 bracketed, but that for @samp{b} is fully bracketed.
2418 int a[2][2] = @{ 0, 1, 2, 3 @};
2419 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2422 This warning is enabled by @option{-Wall}.
2424 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2425 @opindex Wmissing-include-dirs
2426 Warn if a user-supplied include directory does not exist.
2429 @opindex Wparentheses
2430 Warn if parentheses are omitted in certain contexts, such
2431 as when there is an assignment in a context where a truth value
2432 is expected, or when operators are nested whose precedence people
2433 often get confused about. Only the warning for an assignment used as
2434 a truth value is supported when compiling C++; the other warnings are
2435 only supported when compiling C@.
2437 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2438 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2439 interpretation from that of ordinary mathematical notation.
2441 Also warn about constructions where there may be confusion to which
2442 @code{if} statement an @code{else} branch belongs. Here is an example of
2457 In C, every @code{else} branch belongs to the innermost possible @code{if}
2458 statement, which in this example is @code{if (b)}. This is often not
2459 what the programmer expected, as illustrated in the above example by
2460 indentation the programmer chose. When there is the potential for this
2461 confusion, GCC will issue a warning when this flag is specified.
2462 To eliminate the warning, add explicit braces around the innermost
2463 @code{if} statement so there is no way the @code{else} could belong to
2464 the enclosing @code{if}. The resulting code would look like this:
2480 This warning is enabled by @option{-Wall}.
2482 @item -Wsequence-point
2483 @opindex Wsequence-point
2484 Warn about code that may have undefined semantics because of violations
2485 of sequence point rules in the C standard.
2487 The C standard defines the order in which expressions in a C program are
2488 evaluated in terms of @dfn{sequence points}, which represent a partial
2489 ordering between the execution of parts of the program: those executed
2490 before the sequence point, and those executed after it. These occur
2491 after the evaluation of a full expression (one which is not part of a
2492 larger expression), after the evaluation of the first operand of a
2493 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2494 function is called (but after the evaluation of its arguments and the
2495 expression denoting the called function), and in certain other places.
2496 Other than as expressed by the sequence point rules, the order of
2497 evaluation of subexpressions of an expression is not specified. All
2498 these rules describe only a partial order rather than a total order,
2499 since, for example, if two functions are called within one expression
2500 with no sequence point between them, the order in which the functions
2501 are called is not specified. However, the standards committee have
2502 ruled that function calls do not overlap.
2504 It is not specified when between sequence points modifications to the
2505 values of objects take effect. Programs whose behavior depends on this
2506 have undefined behavior; the C standard specifies that ``Between the
2507 previous and next sequence point an object shall have its stored value
2508 modified at most once by the evaluation of an expression. Furthermore,
2509 the prior value shall be read only to determine the value to be
2510 stored.''. If a program breaks these rules, the results on any
2511 particular implementation are entirely unpredictable.
2513 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2514 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2515 diagnosed by this option, and it may give an occasional false positive
2516 result, but in general it has been found fairly effective at detecting
2517 this sort of problem in programs.
2519 The present implementation of this option only works for C programs. A
2520 future implementation may also work for C++ programs.
2522 The C standard is worded confusingly, therefore there is some debate
2523 over the precise meaning of the sequence point rules in subtle cases.
2524 Links to discussions of the problem, including proposed formal
2525 definitions, may be found on the GCC readings page, at
2526 @w{@uref{http://gcc.gnu.org/readings.html}}.
2528 This warning is enabled by @option{-Wall}.
2531 @opindex Wreturn-type
2532 Warn whenever a function is defined with a return-type that defaults to
2533 @code{int}. Also warn about any @code{return} statement with no
2534 return-value in a function whose return-type is not @code{void}.
2536 For C, also warn if the return type of a function has a type qualifier
2537 such as @code{const}. Such a type qualifier has no effect, since the
2538 value returned by a function is not an lvalue. ISO C prohibits
2539 qualified @code{void} return types on function definitions, so such
2540 return types always receive a warning even without this option.
2542 For C++, a function without return type always produces a diagnostic
2543 message, even when @option{-Wno-return-type} is specified. The only
2544 exceptions are @samp{main} and functions defined in system headers.
2546 This warning is enabled by @option{-Wall}.
2550 Warn whenever a @code{switch} statement has an index of enumerated type
2551 and lacks a @code{case} for one or more of the named codes of that
2552 enumeration. (The presence of a @code{default} label prevents this
2553 warning.) @code{case} labels outside the enumeration range also
2554 provoke warnings when this option is used.
2555 This warning is enabled by @option{-Wall}.
2557 @item -Wswitch-default
2558 @opindex Wswitch-switch
2559 Warn whenever a @code{switch} statement does not have a @code{default}
2563 @opindex Wswitch-enum
2564 Warn whenever a @code{switch} statement has an index of enumerated type
2565 and lacks a @code{case} for one or more of the named codes of that
2566 enumeration. @code{case} labels outside the enumeration range also
2567 provoke warnings when this option is used.
2571 Warn if any trigraphs are encountered that might change the meaning of
2572 the program (trigraphs within comments are not warned about).
2573 This warning is enabled by @option{-Wall}.
2575 @item -Wunused-function
2576 @opindex Wunused-function
2577 Warn whenever a static function is declared but not defined or a
2578 non\-inline static function is unused.
2579 This warning is enabled by @option{-Wall}.
2581 @item -Wunused-label
2582 @opindex Wunused-label
2583 Warn whenever a label is declared but not used.
2584 This warning is enabled by @option{-Wall}.
2586 To suppress this warning use the @samp{unused} attribute
2587 (@pxref{Variable Attributes}).
2589 @item -Wunused-parameter
2590 @opindex Wunused-parameter
2591 Warn whenever a function parameter is unused aside from its declaration.
2593 To suppress this warning use the @samp{unused} attribute
2594 (@pxref{Variable Attributes}).
2596 @item -Wunused-variable
2597 @opindex Wunused-variable
2598 Warn whenever a local variable or non-constant static variable is unused
2599 aside from its declaration
2600 This warning is enabled by @option{-Wall}.
2602 To suppress this warning use the @samp{unused} attribute
2603 (@pxref{Variable Attributes}).
2605 @item -Wunused-value
2606 @opindex Wunused-value
2607 Warn whenever a statement computes a result that is explicitly not used.
2608 This warning is enabled by @option{-Wall}.
2610 To suppress this warning cast the expression to @samp{void}.
2614 All the above @option{-Wunused} options combined.
2616 In order to get a warning about an unused function parameter, you must
2617 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2618 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2620 @item -Wuninitialized
2621 @opindex Wuninitialized
2622 Warn if an automatic variable is used without first being initialized or
2623 if a variable may be clobbered by a @code{setjmp} call.
2625 These warnings are possible only in optimizing compilation,
2626 because they require data flow information that is computed only
2627 when optimizing. If you don't specify @option{-O}, you simply won't
2630 If you want to warn about code which uses the uninitialized value of the
2631 variable in its own initializer, use the @option{-Winit-self} option.
2633 These warnings occur for individual uninitialized or clobbered
2634 elements of structure, union or array variables as well as for
2635 variables which are uninitialized or clobbered as a whole. They do
2636 not occur for variables or elements declared @code{volatile}. Because
2637 these warnings depend on optimization, the exact variables or elements
2638 for which there are warnings will depend on the precise optimization
2639 options and version of GCC used.
2641 Note that there may be no warning about a variable that is used only
2642 to compute a value that itself is never used, because such
2643 computations may be deleted by data flow analysis before the warnings
2646 These warnings are made optional because GCC is not smart
2647 enough to see all the reasons why the code might be correct
2648 despite appearing to have an error. Here is one example of how
2669 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2670 always initialized, but GCC doesn't know this. Here is
2671 another common case:
2676 if (change_y) save_y = y, y = new_y;
2678 if (change_y) y = save_y;
2683 This has no bug because @code{save_y} is used only if it is set.
2685 @cindex @code{longjmp} warnings
2686 This option also warns when a non-volatile automatic variable might be
2687 changed by a call to @code{longjmp}. These warnings as well are possible
2688 only in optimizing compilation.
2690 The compiler sees only the calls to @code{setjmp}. It cannot know
2691 where @code{longjmp} will be called; in fact, a signal handler could
2692 call it at any point in the code. As a result, you may get a warning
2693 even when there is in fact no problem because @code{longjmp} cannot
2694 in fact be called at the place which would cause a problem.
2696 Some spurious warnings can be avoided if you declare all the functions
2697 you use that never return as @code{noreturn}. @xref{Function
2700 This warning is enabled by @option{-Wall}.
2702 @item -Wunknown-pragmas
2703 @opindex Wunknown-pragmas
2704 @cindex warning for unknown pragmas
2705 @cindex unknown pragmas, warning
2706 @cindex pragmas, warning of unknown
2707 Warn when a #pragma directive is encountered which is not understood by
2708 GCC@. If this command line option is used, warnings will even be issued
2709 for unknown pragmas in system header files. This is not the case if
2710 the warnings were only enabled by the @option{-Wall} command line option.
2713 @opindex Wno-pragmas
2715 Do not warn about misuses of pragmas, such as incorrect parameters,
2716 invalid syntax, or conflicts between pragmas. See also
2717 @samp{-Wunknown-pragmas}.
2719 @item -Wstrict-aliasing
2720 @opindex Wstrict-aliasing
2721 This option is only active when @option{-fstrict-aliasing} is active.
2722 It warns about code which might break the strict aliasing rules that the
2723 compiler is using for optimization. The warning does not catch all
2724 cases, but does attempt to catch the more common pitfalls. It is
2725 included in @option{-Wall}.
2727 @item -Wstrict-aliasing=2
2728 @opindex Wstrict-aliasing=2
2729 This option is only active when @option{-fstrict-aliasing} is active.
2730 It warns about code which might break the strict aliasing rules that the
2731 compiler is using for optimization. This warning catches more cases than
2732 @option{-Wstrict-aliasing}, but it will also give a warning for some ambiguous
2733 cases that are safe.
2737 All of the above @samp{-W} options combined. This enables all the
2738 warnings about constructions that some users consider questionable, and
2739 that are easy to avoid (or modify to prevent the warning), even in
2740 conjunction with macros. This also enables some language-specific
2741 warnings described in @ref{C++ Dialect Options} and
2742 @ref{Objective-C and Objective-C++ Dialect Options}.
2745 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2746 Some of them warn about constructions that users generally do not
2747 consider questionable, but which occasionally you might wish to check
2748 for; others warn about constructions that are necessary or hard to avoid
2749 in some cases, and there is no simple way to modify the code to suppress
2756 (This option used to be called @option{-W}. The older name is still
2757 supported, but the newer name is more descriptive.) Print extra warning
2758 messages for these events:
2762 A function can return either with or without a value. (Falling
2763 off the end of the function body is considered returning without
2764 a value.) For example, this function would evoke such a
2778 An expression-statement or the left-hand side of a comma expression
2779 contains no side effects.
2780 To suppress the warning, cast the unused expression to void.
2781 For example, an expression such as @samp{x[i,j]} will cause a warning,
2782 but @samp{x[(void)i,j]} will not.
2785 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2788 Storage-class specifiers like @code{static} are not the first things in
2789 a declaration. According to the C Standard, this usage is obsolescent.
2792 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2796 A comparison between signed and unsigned values could produce an
2797 incorrect result when the signed value is converted to unsigned.
2798 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2801 An aggregate has an initializer which does not initialize all members.
2802 This warning can be independently controlled by
2803 @option{-Wmissing-field-initializers}.
2806 A function parameter is declared without a type specifier in K&R-style
2814 An empty body occurs in an @samp{if} or @samp{else} statement.
2817 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2818 @samp{>}, or @samp{>=}.
2821 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2824 Any of several floating-point events that often indicate errors, such as
2825 overflow, underflow, loss of precision, etc.
2827 @item @r{(C++ only)}
2828 An enumerator and a non-enumerator both appear in a conditional expression.
2830 @item @r{(C++ only)}
2831 A non-static reference or non-static @samp{const} member appears in a
2832 class without constructors.
2834 @item @r{(C++ only)}
2835 Ambiguous virtual bases.
2837 @item @r{(C++ only)}
2838 Subscripting an array which has been declared @samp{register}.
2840 @item @r{(C++ only)}
2841 Taking the address of a variable which has been declared @samp{register}.
2843 @item @r{(C++ only)}
2844 A base class is not initialized in a derived class' copy constructor.
2847 @item -Wno-div-by-zero
2848 @opindex Wno-div-by-zero
2849 @opindex Wdiv-by-zero
2850 Do not warn about compile-time integer division by zero. Floating point
2851 division by zero is not warned about, as it can be a legitimate way of
2852 obtaining infinities and NaNs.
2854 @item -Wsystem-headers
2855 @opindex Wsystem-headers
2856 @cindex warnings from system headers
2857 @cindex system headers, warnings from
2858 Print warning messages for constructs found in system header files.
2859 Warnings from system headers are normally suppressed, on the assumption
2860 that they usually do not indicate real problems and would only make the
2861 compiler output harder to read. Using this command line option tells
2862 GCC to emit warnings from system headers as if they occurred in user
2863 code. However, note that using @option{-Wall} in conjunction with this
2864 option will @emph{not} warn about unknown pragmas in system
2865 headers---for that, @option{-Wunknown-pragmas} must also be used.
2868 @opindex Wfloat-equal
2869 Warn if floating point values are used in equality comparisons.
2871 The idea behind this is that sometimes it is convenient (for the
2872 programmer) to consider floating-point values as approximations to
2873 infinitely precise real numbers. If you are doing this, then you need
2874 to compute (by analyzing the code, or in some other way) the maximum or
2875 likely maximum error that the computation introduces, and allow for it
2876 when performing comparisons (and when producing output, but that's a
2877 different problem). In particular, instead of testing for equality, you
2878 would check to see whether the two values have ranges that overlap; and
2879 this is done with the relational operators, so equality comparisons are
2882 @item -Wtraditional @r{(C only)}
2883 @opindex Wtraditional
2884 Warn about certain constructs that behave differently in traditional and
2885 ISO C@. Also warn about ISO C constructs that have no traditional C
2886 equivalent, and/or problematic constructs which should be avoided.
2890 Macro parameters that appear within string literals in the macro body.
2891 In traditional C macro replacement takes place within string literals,
2892 but does not in ISO C@.
2895 In traditional C, some preprocessor directives did not exist.
2896 Traditional preprocessors would only consider a line to be a directive
2897 if the @samp{#} appeared in column 1 on the line. Therefore
2898 @option{-Wtraditional} warns about directives that traditional C
2899 understands but would ignore because the @samp{#} does not appear as the
2900 first character on the line. It also suggests you hide directives like
2901 @samp{#pragma} not understood by traditional C by indenting them. Some
2902 traditional implementations would not recognize @samp{#elif}, so it
2903 suggests avoiding it altogether.
2906 A function-like macro that appears without arguments.
2909 The unary plus operator.
2912 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2913 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2914 constants.) Note, these suffixes appear in macros defined in the system
2915 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2916 Use of these macros in user code might normally lead to spurious
2917 warnings, however GCC's integrated preprocessor has enough context to
2918 avoid warning in these cases.
2921 A function declared external in one block and then used after the end of
2925 A @code{switch} statement has an operand of type @code{long}.
2928 A non-@code{static} function declaration follows a @code{static} one.
2929 This construct is not accepted by some traditional C compilers.
2932 The ISO type of an integer constant has a different width or
2933 signedness from its traditional type. This warning is only issued if
2934 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2935 typically represent bit patterns, are not warned about.
2938 Usage of ISO string concatenation is detected.
2941 Initialization of automatic aggregates.
2944 Identifier conflicts with labels. Traditional C lacks a separate
2945 namespace for labels.
2948 Initialization of unions. If the initializer is zero, the warning is
2949 omitted. This is done under the assumption that the zero initializer in
2950 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2951 initializer warnings and relies on default initialization to zero in the
2955 Conversions by prototypes between fixed/floating point values and vice
2956 versa. The absence of these prototypes when compiling with traditional
2957 C would cause serious problems. This is a subset of the possible
2958 conversion warnings, for the full set use @option{-Wconversion}.
2961 Use of ISO C style function definitions. This warning intentionally is
2962 @emph{not} issued for prototype declarations or variadic functions
2963 because these ISO C features will appear in your code when using
2964 libiberty's traditional C compatibility macros, @code{PARAMS} and
2965 @code{VPARAMS}. This warning is also bypassed for nested functions
2966 because that feature is already a GCC extension and thus not relevant to
2967 traditional C compatibility.
2970 @item -Wdeclaration-after-statement @r{(C only)}
2971 @opindex Wdeclaration-after-statement
2972 Warn when a declaration is found after a statement in a block. This
2973 construct, known from C++, was introduced with ISO C99 and is by default
2974 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2975 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2979 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2981 @item -Wno-endif-labels
2982 @opindex Wno-endif-labels
2983 @opindex Wendif-labels
2984 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2988 Warn whenever a local variable shadows another local variable, parameter or
2989 global variable or whenever a built-in function is shadowed.
2991 @item -Wlarger-than-@var{len}
2992 @opindex Wlarger-than
2993 Warn whenever an object of larger than @var{len} bytes is defined.
2995 @item -Wunsafe-loop-optimizations
2996 @opindex Wunsafe-loop-optimizations
2997 Warn if the loop cannot be optimized because the compiler could not
2998 assume anything on the bounds of the loop indices. With
2999 @option{-funsafe-loop-optimizations} warn if the compiler made
3002 @item -Wpointer-arith
3003 @opindex Wpointer-arith
3004 Warn about anything that depends on the ``size of'' a function type or
3005 of @code{void}. GNU C assigns these types a size of 1, for
3006 convenience in calculations with @code{void *} pointers and pointers
3009 @item -Wbad-function-cast @r{(C only)}
3010 @opindex Wbad-function-cast
3011 Warn whenever a function call is cast to a non-matching type.
3012 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3015 Warn about ISO C constructs that are outside of the common subset of
3016 ISO C and ISO C++, e.g.@: request for implicit conversion from
3017 @code{void *} to a pointer to non-@code{void} type.
3021 Warn whenever a pointer is cast so as to remove a type qualifier from
3022 the target type. For example, warn if a @code{const char *} is cast
3023 to an ordinary @code{char *}.
3026 @opindex Wcast-align
3027 Warn whenever a pointer is cast such that the required alignment of the
3028 target is increased. For example, warn if a @code{char *} is cast to
3029 an @code{int *} on machines where integers can only be accessed at
3030 two- or four-byte boundaries.
3032 @item -Wwrite-strings
3033 @opindex Wwrite-strings
3034 When compiling C, give string constants the type @code{const
3035 char[@var{length}]} so that
3036 copying the address of one into a non-@code{const} @code{char *}
3037 pointer will get a warning; when compiling C++, warn about the
3038 deprecated conversion from string constants to @code{char *}.
3039 These warnings will help you find at
3040 compile time code that can try to write into a string constant, but
3041 only if you have been very careful about using @code{const} in
3042 declarations and prototypes. Otherwise, it will just be a nuisance;
3043 this is why we did not make @option{-Wall} request these warnings.
3046 @opindex Wconversion
3047 Warn if a prototype causes a type conversion that is different from what
3048 would happen to the same argument in the absence of a prototype. This
3049 includes conversions of fixed point to floating and vice versa, and
3050 conversions changing the width or signedness of a fixed point argument
3051 except when the same as the default promotion.
3053 Also, warn if a negative integer constant expression is implicitly
3054 converted to an unsigned type. For example, warn about the assignment
3055 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
3056 casts like @code{(unsigned) -1}.
3058 @item -Wsign-compare
3059 @opindex Wsign-compare
3060 @cindex warning for comparison of signed and unsigned values
3061 @cindex comparison of signed and unsigned values, warning
3062 @cindex signed and unsigned values, comparison warning
3063 Warn when a comparison between signed and unsigned values could produce
3064 an incorrect result when the signed value is converted to unsigned.
3065 This warning is also enabled by @option{-Wextra}; to get the other warnings
3066 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3068 @item -Waggregate-return
3069 @opindex Waggregate-return
3070 Warn if any functions that return structures or unions are defined or
3071 called. (In languages where you can return an array, this also elicits
3074 @item -Wno-attributes
3075 @opindex Wno-attributes
3076 @opindex Wattributes
3077 Do not warn if an unexpected @code{__attribute__} is used, such as
3078 unrecognized attributes, function attributes applied to variables,
3079 etc. This will not stop errors for incorrect use of supported
3082 @item -Wstrict-prototypes @r{(C only)}
3083 @opindex Wstrict-prototypes
3084 Warn if a function is declared or defined without specifying the
3085 argument types. (An old-style function definition is permitted without
3086 a warning if preceded by a declaration which specifies the argument
3089 @item -Wold-style-definition @r{(C only)}
3090 @opindex Wold-style-definition
3091 Warn if an old-style function definition is used. A warning is given
3092 even if there is a previous prototype.
3094 @item -Wmissing-prototypes @r{(C only)}
3095 @opindex Wmissing-prototypes
3096 Warn if a global function is defined without a previous prototype
3097 declaration. This warning is issued even if the definition itself
3098 provides a prototype. The aim is to detect global functions that fail
3099 to be declared in header files.
3101 @item -Wmissing-declarations @r{(C only)}
3102 @opindex Wmissing-declarations
3103 Warn if a global function is defined without a previous declaration.
3104 Do so even if the definition itself provides a prototype.
3105 Use this option to detect global functions that are not declared in
3108 @item -Wmissing-field-initializers
3109 @opindex Wmissing-field-initializers
3112 Warn if a structure's initializer has some fields missing. For
3113 example, the following code would cause such a warning, because
3114 @code{x.h} is implicitly zero:
3117 struct s @{ int f, g, h; @};
3118 struct s x = @{ 3, 4 @};
3121 This option does not warn about designated initializers, so the following
3122 modification would not trigger a warning:
3125 struct s @{ int f, g, h; @};
3126 struct s x = @{ .f = 3, .g = 4 @};
3129 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3130 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3132 @item -Wmissing-noreturn
3133 @opindex Wmissing-noreturn
3134 Warn about functions which might be candidates for attribute @code{noreturn}.
3135 Note these are only possible candidates, not absolute ones. Care should
3136 be taken to manually verify functions actually do not ever return before
3137 adding the @code{noreturn} attribute, otherwise subtle code generation
3138 bugs could be introduced. You will not get a warning for @code{main} in
3139 hosted C environments.
3141 @item -Wmissing-format-attribute
3142 @opindex Wmissing-format-attribute
3144 Warn about function pointers which might be candidates for @code{format}
3145 attributes. Note these are only possible candidates, not absolute ones.
3146 GCC will guess that function pointers with @code{format} attributes that
3147 are used in assignment, initialization, parameter passing or return
3148 statements should have a corresponding @code{format} attribute in the
3149 resulting type. I.e.@: the left-hand side of the assignment or
3150 initialization, the type of the parameter variable, or the return type
3151 of the containing function respectively should also have a @code{format}
3152 attribute to avoid the warning.
3154 GCC will also warn about function definitions which might be
3155 candidates for @code{format} attributes. Again, these are only
3156 possible candidates. GCC will guess that @code{format} attributes
3157 might be appropriate for any function that calls a function like
3158 @code{vprintf} or @code{vscanf}, but this might not always be the
3159 case, and some functions for which @code{format} attributes are
3160 appropriate may not be detected.
3162 @item -Wno-multichar
3163 @opindex Wno-multichar
3165 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3166 Usually they indicate a typo in the user's code, as they have
3167 implementation-defined values, and should not be used in portable code.
3169 @item -Wnormalized=<none|id|nfc|nfkc>
3170 @opindex Wnormalized
3173 @cindex character set, input normalization
3174 In ISO C and ISO C++, two identifiers are different if they are
3175 different sequences of characters. However, sometimes when characters
3176 outside the basic ASCII character set are used, you can have two
3177 different character sequences that look the same. To avoid confusion,
3178 the ISO 10646 standard sets out some @dfn{normalization rules} which
3179 when applied ensure that two sequences that look the same are turned into
3180 the same sequence. GCC can warn you if you are using identifiers which
3181 have not been normalized; this option controls that warning.
3183 There are four levels of warning that GCC supports. The default is
3184 @option{-Wnormalized=nfc}, which warns about any identifier which is
3185 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3186 recommended form for most uses.
3188 Unfortunately, there are some characters which ISO C and ISO C++ allow
3189 in identifiers that when turned into NFC aren't allowable as
3190 identifiers. That is, there's no way to use these symbols in portable
3191 ISO C or C++ and have all your identifiers in NFC.
3192 @option{-Wnormalized=id} suppresses the warning for these characters.
3193 It is hoped that future versions of the standards involved will correct
3194 this, which is why this option is not the default.
3196 You can switch the warning off for all characters by writing
3197 @option{-Wnormalized=none}. You would only want to do this if you
3198 were using some other normalization scheme (like ``D''), because
3199 otherwise you can easily create bugs that are literally impossible to see.
3201 Some characters in ISO 10646 have distinct meanings but look identical
3202 in some fonts or display methodologies, especially once formatting has
3203 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3204 LETTER N'', will display just like a regular @code{n} which has been
3205 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3206 normalisation scheme to convert all these into a standard form as
3207 well, and GCC will warn if your code is not in NFKC if you use
3208 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3209 about every identifier that contains the letter O because it might be
3210 confused with the digit 0, and so is not the default, but may be
3211 useful as a local coding convention if the programming environment is
3212 unable to be fixed to display these characters distinctly.
3214 @item -Wno-deprecated-declarations
3215 @opindex Wno-deprecated-declarations
3216 Do not warn about uses of functions, variables, and types marked as
3217 deprecated by using the @code{deprecated} attribute.
3218 (@pxref{Function Attributes}, @pxref{Variable Attributes},
3219 @pxref{Type Attributes}.)
3223 Warn if a structure is given the packed attribute, but the packed
3224 attribute has no effect on the layout or size of the structure.
3225 Such structures may be mis-aligned for little benefit. For
3226 instance, in this code, the variable @code{f.x} in @code{struct bar}
3227 will be misaligned even though @code{struct bar} does not itself
3228 have the packed attribute:
3235 @} __attribute__((packed));
3245 Warn if padding is included in a structure, either to align an element
3246 of the structure or to align the whole structure. Sometimes when this
3247 happens it is possible to rearrange the fields of the structure to
3248 reduce the padding and so make the structure smaller.
3250 @item -Wredundant-decls
3251 @opindex Wredundant-decls
3252 Warn if anything is declared more than once in the same scope, even in
3253 cases where multiple declaration is valid and changes nothing.
3255 @item -Wnested-externs @r{(C only)}
3256 @opindex Wnested-externs
3257 Warn if an @code{extern} declaration is encountered within a function.
3259 @item -Wunreachable-code
3260 @opindex Wunreachable-code
3261 Warn if the compiler detects that code will never be executed.
3263 This option is intended to warn when the compiler detects that at
3264 least a whole line of source code will never be executed, because
3265 some condition is never satisfied or because it is after a
3266 procedure that never returns.
3268 It is possible for this option to produce a warning even though there
3269 are circumstances under which part of the affected line can be executed,
3270 so care should be taken when removing apparently-unreachable code.
3272 For instance, when a function is inlined, a warning may mean that the
3273 line is unreachable in only one inlined copy of the function.
3275 This option is not made part of @option{-Wall} because in a debugging
3276 version of a program there is often substantial code which checks
3277 correct functioning of the program and is, hopefully, unreachable
3278 because the program does work. Another common use of unreachable
3279 code is to provide behavior which is selectable at compile-time.
3283 Warn if a function can not be inlined and it was declared as inline.
3284 Even with this option, the compiler will not warn about failures to
3285 inline functions declared in system headers.
3287 The compiler uses a variety of heuristics to determine whether or not
3288 to inline a function. For example, the compiler takes into account
3289 the size of the function being inlined and the amount of inlining
3290 that has already been done in the current function. Therefore,
3291 seemingly insignificant changes in the source program can cause the
3292 warnings produced by @option{-Winline} to appear or disappear.
3294 @item -Wno-invalid-offsetof @r{(C++ only)}
3295 @opindex Wno-invalid-offsetof
3296 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3297 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3298 to a non-POD type is undefined. In existing C++ implementations,
3299 however, @samp{offsetof} typically gives meaningful results even when
3300 applied to certain kinds of non-POD types. (Such as a simple
3301 @samp{struct} that fails to be a POD type only by virtue of having a
3302 constructor.) This flag is for users who are aware that they are
3303 writing nonportable code and who have deliberately chosen to ignore the
3306 The restrictions on @samp{offsetof} may be relaxed in a future version
3307 of the C++ standard.
3309 @item -Wno-int-to-pointer-cast @r{(C only)}
3310 @opindex Wno-int-to-pointer-cast
3311 Suppress warnings from casts to pointer type of an integer of a
3314 @item -Wno-pointer-to-int-cast @r{(C only)}
3315 @opindex Wno-pointer-to-int-cast
3316 Suppress warnings from casts from a pointer to an integer type of a
3320 @opindex Winvalid-pch
3321 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3322 the search path but can't be used.
3326 @opindex Wno-long-long
3327 Warn if @samp{long long} type is used. This is default. To inhibit
3328 the warning messages, use @option{-Wno-long-long}. Flags
3329 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3330 only when @option{-pedantic} flag is used.
3332 @item -Wvariadic-macros
3333 @opindex Wvariadic-macros
3334 @opindex Wno-variadic-macros
3335 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3336 alternate syntax when in pedantic ISO C99 mode. This is default.
3337 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3339 @item -Wdisabled-optimization
3340 @opindex Wdisabled-optimization
3341 Warn if a requested optimization pass is disabled. This warning does
3342 not generally indicate that there is anything wrong with your code; it
3343 merely indicates that GCC's optimizers were unable to handle the code
3344 effectively. Often, the problem is that your code is too big or too
3345 complex; GCC will refuse to optimize programs when the optimization
3346 itself is likely to take inordinate amounts of time.
3348 @item -Wno-pointer-sign
3349 @opindex Wno-pointer-sign
3350 Don't warn for pointer argument passing or assignment with different signedness.
3351 Only useful in the negative form since this warning is enabled by default.
3352 This option is only supported for C and Objective-C@.
3356 Make all warnings into errors.
3359 @node Debugging Options
3360 @section Options for Debugging Your Program or GCC
3361 @cindex options, debugging
3362 @cindex debugging information options
3364 GCC has various special options that are used for debugging
3365 either your program or GCC:
3370 Produce debugging information in the operating system's native format
3371 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3374 On most systems that use stabs format, @option{-g} enables use of extra
3375 debugging information that only GDB can use; this extra information
3376 makes debugging work better in GDB but will probably make other debuggers
3378 refuse to read the program. If you want to control for certain whether
3379 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3380 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3382 GCC allows you to use @option{-g} with
3383 @option{-O}. The shortcuts taken by optimized code may occasionally
3384 produce surprising results: some variables you declared may not exist
3385 at all; flow of control may briefly move where you did not expect it;
3386 some statements may not be executed because they compute constant
3387 results or their values were already at hand; some statements may
3388 execute in different places because they were moved out of loops.
3390 Nevertheless it proves possible to debug optimized output. This makes
3391 it reasonable to use the optimizer for programs that might have bugs.
3393 The following options are useful when GCC is generated with the
3394 capability for more than one debugging format.
3398 Produce debugging information for use by GDB@. This means to use the
3399 most expressive format available (DWARF 2, stabs, or the native format
3400 if neither of those are supported), including GDB extensions if at all
3405 Produce debugging information in stabs format (if that is supported),
3406 without GDB extensions. This is the format used by DBX on most BSD
3407 systems. On MIPS, Alpha and System V Release 4 systems this option
3408 produces stabs debugging output which is not understood by DBX or SDB@.
3409 On System V Release 4 systems this option requires the GNU assembler.
3411 @item -feliminate-unused-debug-symbols
3412 @opindex feliminate-unused-debug-symbols
3413 Produce debugging information in stabs format (if that is supported),
3414 for only symbols that are actually used.
3418 Produce debugging information in stabs format (if that is supported),
3419 using GNU extensions understood only by the GNU debugger (GDB)@. The
3420 use of these extensions is likely to make other debuggers crash or
3421 refuse to read the program.
3425 Produce debugging information in COFF format (if that is supported).
3426 This is the format used by SDB on most System V systems prior to
3431 Produce debugging information in XCOFF format (if that is supported).
3432 This is the format used by the DBX debugger on IBM RS/6000 systems.
3436 Produce debugging information in XCOFF format (if that is supported),
3437 using GNU extensions understood only by the GNU debugger (GDB)@. The
3438 use of these extensions is likely to make other debuggers crash or
3439 refuse to read the program, and may cause assemblers other than the GNU
3440 assembler (GAS) to fail with an error.
3444 Produce debugging information in DWARF version 2 format (if that is
3445 supported). This is the format used by DBX on IRIX 6. With this
3446 option, GCC uses features of DWARF version 3 when they are useful;
3447 version 3 is upward compatible with version 2, but may still cause
3448 problems for older debuggers.
3452 Produce debugging information in VMS debug format (if that is
3453 supported). This is the format used by DEBUG on VMS systems.
3456 @itemx -ggdb@var{level}
3457 @itemx -gstabs@var{level}
3458 @itemx -gcoff@var{level}
3459 @itemx -gxcoff@var{level}
3460 @itemx -gvms@var{level}
3461 Request debugging information and also use @var{level} to specify how
3462 much information. The default level is 2.
3464 Level 1 produces minimal information, enough for making backtraces in
3465 parts of the program that you don't plan to debug. This includes
3466 descriptions of functions and external variables, but no information
3467 about local variables and no line numbers.
3469 Level 3 includes extra information, such as all the macro definitions
3470 present in the program. Some debuggers support macro expansion when
3471 you use @option{-g3}.
3473 @option{-gdwarf-2} does not accept a concatenated debug level, because
3474 GCC used to support an option @option{-gdwarf} that meant to generate
3475 debug information in version 1 of the DWARF format (which is very
3476 different from version 2), and it would have been too confusing. That
3477 debug format is long obsolete, but the option cannot be changed now.
3478 Instead use an additional @option{-g@var{level}} option to change the
3479 debug level for DWARF2.
3481 @item -feliminate-dwarf2-dups
3482 @opindex feliminate-dwarf2-dups
3483 Compress DWARF2 debugging information by eliminating duplicated
3484 information about each symbol. This option only makes sense when
3485 generating DWARF2 debugging information with @option{-gdwarf-2}.
3487 @cindex @command{prof}
3490 Generate extra code to write profile information suitable for the
3491 analysis program @command{prof}. You must use this option when compiling
3492 the source files you want data about, and you must also use it when
3495 @cindex @command{gprof}
3498 Generate extra code to write profile information suitable for the
3499 analysis program @command{gprof}. You must use this option when compiling
3500 the source files you want data about, and you must also use it when
3505 Makes the compiler print out each function name as it is compiled, and
3506 print some statistics about each pass when it finishes.
3509 @opindex ftime-report
3510 Makes the compiler print some statistics about the time consumed by each
3511 pass when it finishes.
3514 @opindex fmem-report
3515 Makes the compiler print some statistics about permanent memory
3516 allocation when it finishes.
3518 @item -fprofile-arcs
3519 @opindex fprofile-arcs
3520 Add code so that program flow @dfn{arcs} are instrumented. During
3521 execution the program records how many times each branch and call is
3522 executed and how many times it is taken or returns. When the compiled
3523 program exits it saves this data to a file called
3524 @file{@var{auxname}.gcda} for each source file. The data may be used for
3525 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3526 test coverage analysis (@option{-ftest-coverage}). Each object file's
3527 @var{auxname} is generated from the name of the output file, if
3528 explicitly specified and it is not the final executable, otherwise it is
3529 the basename of the source file. In both cases any suffix is removed
3530 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3531 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3532 @xref{Cross-profiling}.
3534 @cindex @command{gcov}
3538 This option is used to compile and link code instrumented for coverage
3539 analysis. The option is a synonym for @option{-fprofile-arcs}
3540 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3541 linking). See the documentation for those options for more details.
3546 Compile the source files with @option{-fprofile-arcs} plus optimization
3547 and code generation options. For test coverage analysis, use the
3548 additional @option{-ftest-coverage} option. You do not need to profile
3549 every source file in a program.
3552 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3553 (the latter implies the former).
3556 Run the program on a representative workload to generate the arc profile
3557 information. This may be repeated any number of times. You can run
3558 concurrent instances of your program, and provided that the file system
3559 supports locking, the data files will be correctly updated. Also
3560 @code{fork} calls are detected and correctly handled (double counting
3564 For profile-directed optimizations, compile the source files again with
3565 the same optimization and code generation options plus
3566 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3567 Control Optimization}).
3570 For test coverage analysis, use @command{gcov} to produce human readable
3571 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3572 @command{gcov} documentation for further information.
3576 With @option{-fprofile-arcs}, for each function of your program GCC
3577 creates a program flow graph, then finds a spanning tree for the graph.
3578 Only arcs that are not on the spanning tree have to be instrumented: the
3579 compiler adds code to count the number of times that these arcs are
3580 executed. When an arc is the only exit or only entrance to a block, the
3581 instrumentation code can be added to the block; otherwise, a new basic
3582 block must be created to hold the instrumentation code.
3585 @item -ftest-coverage
3586 @opindex ftest-coverage
3587 Produce a notes file that the @command{gcov} code-coverage utility
3588 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3589 show program coverage. Each source file's note file is called
3590 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3591 above for a description of @var{auxname} and instructions on how to
3592 generate test coverage data. Coverage data will match the source files
3593 more closely, if you do not optimize.
3595 @item -d@var{letters}
3596 @item -fdump-rtl-@var{pass}
3598 Says to make debugging dumps during compilation at times specified by
3599 @var{letters}. This is used for debugging the RTL-based passes of the
3600 compiler. The file names for most of the dumps are made by appending a
3601 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3602 from the name of the output file, if explicitly specified and it is not
3603 an executable, otherwise it is the basename of the source file.
3605 Most debug dumps can be enabled either passing a letter to the @option{-d}
3606 option, or with a long @option{-fdump-rtl} switch; here are the possible
3607 letters for use in @var{letters} and @var{pass}, and their meanings:
3612 Annotate the assembler output with miscellaneous debugging information.
3615 @itemx -fdump-rtl-bp
3617 @opindex fdump-rtl-bp
3618 Dump after computing branch probabilities, to @file{@var{file}.09.bp}.
3621 @itemx -fdump-rtl-bbro
3623 @opindex fdump-rtl-bbro
3624 Dump after block reordering, to @file{@var{file}.30.bbro}.
3627 @itemx -fdump-rtl-combine
3629 @opindex fdump-rtl-combine
3630 Dump after instruction combination, to the file @file{@var{file}.17.combine}.
3633 @itemx -fdump-rtl-ce1
3634 @itemx -fdump-rtl-ce2
3636 @opindex fdump-rtl-ce1
3637 @opindex fdump-rtl-ce2
3638 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3639 first if conversion, to the file @file{@var{file}.11.ce1}. @option{-dC}
3640 and @option{-fdump-rtl-ce2} enable dumping after the second if
3641 conversion, to the file @file{@var{file}.18.ce2}.
3644 @itemx -fdump-rtl-btl
3645 @itemx -fdump-rtl-dbr
3647 @opindex fdump-rtl-btl
3648 @opindex fdump-rtl-dbr
3649 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3650 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3651 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3652 scheduling, to @file{@var{file}.36.dbr}.
3656 Dump all macro definitions, at the end of preprocessing, in addition to
3660 @itemx -fdump-rtl-ce3
3662 @opindex fdump-rtl-ce3
3663 Dump after the third if conversion, to @file{@var{file}.28.ce3}.
3666 @itemx -fdump-rtl-cfg
3667 @itemx -fdump-rtl-life
3669 @opindex fdump-rtl-cfg
3670 @opindex fdump-rtl-life
3671 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3672 and data flow analysis, to @file{@var{file}.08.cfg}. @option{-df}
3673 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3674 to @file{@var{file}.16.life}.
3677 @itemx -fdump-rtl-greg
3679 @opindex fdump-rtl-greg
3680 Dump after global register allocation, to @file{@var{file}.23.greg}.
3683 @itemx -fdump-rtl-gcse
3684 @itemx -fdump-rtl-bypass
3686 @opindex fdump-rtl-gcse
3687 @opindex fdump-rtl-bypass
3688 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3689 @file{@var{file}.05.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3690 enable dumping after jump bypassing and control flow optimizations, to
3691 @file{@var{file}.07.bypass}.
3694 @itemx -fdump-rtl-eh
3696 @opindex fdump-rtl-eh
3697 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3700 @itemx -fdump-rtl-sibling
3702 @opindex fdump-rtl-sibling
3703 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
3706 @itemx -fdump-rtl-jump
3708 @opindex fdump-rtl-jump
3709 Dump after the first jump optimization, to @file{@var{file}.03.jump}.
3712 @itemx -fdump-rtl-stack
3714 @opindex fdump-rtl-stack
3715 Dump after conversion from registers to stack, to @file{@var{file}.33.stack}.
3718 @itemx -fdump-rtl-lreg
3720 @opindex fdump-rtl-lreg
3721 Dump after local register allocation, to @file{@var{file}.22.lreg}.
3724 @itemx -fdump-rtl-loop
3725 @itemx -fdump-rtl-loop2
3727 @opindex fdump-rtl-loop
3728 @opindex fdump-rtl-loop2
3729 @option{-dL} and @option{-fdump-rtl-loop} enable dumping after the first
3730 loop optimization pass, to @file{@var{file}.06.loop}. @option{-dL} and
3731 @option{-fdump-rtl-loop2} enable dumping after the second pass, to
3732 @file{@var{file}.13.loop2}.
3735 @itemx -fdump-rtl-sms
3737 @opindex fdump-rtl-sms
3738 Dump after modulo scheduling, to @file{@var{file}.20.sms}.
3741 @itemx -fdump-rtl-mach
3743 @opindex fdump-rtl-mach
3744 Dump after performing the machine dependent reorganization pass, to
3745 @file{@var{file}.35.mach}.
3748 @itemx -fdump-rtl-rnreg
3750 @opindex fdump-rtl-rnreg
3751 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3754 @itemx -fdump-rtl-regmove
3756 @opindex fdump-rtl-regmove
3757 Dump after the register move pass, to @file{@var{file}.19.regmove}.
3760 @itemx -fdump-rtl-postreload
3762 @opindex fdump-rtl-postreload
3763 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3766 @itemx -fdump-rtl-expand
3768 @opindex fdump-rtl-expand
3769 Dump after RTL generation, to @file{@var{file}.00.expand}.
3772 @itemx -fdump-rtl-sched2
3774 @opindex fdump-rtl-sched2
3775 Dump after the second scheduling pass, to @file{@var{file}.32.sched2}.
3778 @itemx -fdump-rtl-cse
3780 @opindex fdump-rtl-cse
3781 Dump after CSE (including the jump optimization that sometimes follows
3782 CSE), to @file{@var{file}.04.cse}.
3785 @itemx -fdump-rtl-sched
3787 @opindex fdump-rtl-sched
3788 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3791 @itemx -fdump-rtl-cse2
3793 @opindex fdump-rtl-cse2
3794 Dump after the second CSE pass (including the jump optimization that
3795 sometimes follows CSE), to @file{@var{file}.15.cse2}.
3798 @itemx -fdump-rtl-tracer
3800 @opindex fdump-rtl-tracer
3801 Dump after running tracer, to @file{@var{file}.12.tracer}.
3804 @itemx -fdump-rtl-vpt
3805 @itemx -fdump-rtl-vartrack
3807 @opindex fdump-rtl-vpt
3808 @opindex fdump-rtl-vartrack
3809 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3810 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3811 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3812 to @file{@var{file}.34.vartrack}.
3815 @itemx -fdump-rtl-flow2
3817 @opindex fdump-rtl-flow2
3818 Dump after the second flow pass, to @file{@var{file}.26.flow2}.
3821 @itemx -fdump-rtl-peephole2
3823 @opindex fdump-rtl-peephole2
3824 Dump after the peephole pass, to @file{@var{file}.27.peephole2}.
3827 @itemx -fdump-rtl-web
3829 @opindex fdump-rtl-web
3830 Dump after live range splitting, to @file{@var{file}.14.web}.
3833 @itemx -fdump-rtl-all
3835 @opindex fdump-rtl-all
3836 Produce all the dumps listed above.
3840 Produce a core dump whenever an error occurs.
3844 Print statistics on memory usage, at the end of the run, to
3849 Annotate the assembler output with a comment indicating which
3850 pattern and alternative was used. The length of each instruction is
3855 Dump the RTL in the assembler output as a comment before each instruction.
3856 Also turns on @option{-dp} annotation.
3860 For each of the other indicated dump files (either with @option{-d} or
3861 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3862 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3866 Just generate RTL for a function instead of compiling it. Usually used
3867 with @samp{r} (@option{-fdump-rtl-expand}).
3871 Dump debugging information during parsing, to standard error.
3874 @item -fdump-unnumbered
3875 @opindex fdump-unnumbered
3876 When doing debugging dumps (see @option{-d} option above), suppress instruction
3877 numbers and line number note output. This makes it more feasible to
3878 use diff on debugging dumps for compiler invocations with different
3879 options, in particular with and without @option{-g}.
3881 @item -fdump-translation-unit @r{(C++ only)}
3882 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
3883 @opindex fdump-translation-unit
3884 Dump a representation of the tree structure for the entire translation
3885 unit to a file. The file name is made by appending @file{.tu} to the
3886 source file name. If the @samp{-@var{options}} form is used, @var{options}
3887 controls the details of the dump as described for the
3888 @option{-fdump-tree} options.
3890 @item -fdump-class-hierarchy @r{(C++ only)}
3891 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3892 @opindex fdump-class-hierarchy
3893 Dump a representation of each class's hierarchy and virtual function
3894 table layout to a file. The file name is made by appending @file{.class}
3895 to the source file name. If the @samp{-@var{options}} form is used,
3896 @var{options} controls the details of the dump as described for the
3897 @option{-fdump-tree} options.
3899 @item -fdump-ipa-@var{switch}
3901 Control the dumping at various stages of inter-procedural analysis
3902 language tree to a file. The file name is generated by appending a switch
3903 specific suffix to the source file name. The following dumps are possible:
3907 Enables all inter-procedural analysis dumps; currently the only produced
3908 dump is the @samp{cgraph} dump.
3911 Dumps information about call-graph optimization, unused function removal,
3912 and inlining decisions.
3915 @item -fdump-tree-@var{switch} @r{(C and C++ only)}
3916 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C and C++ only)}
3918 Control the dumping at various stages of processing the intermediate
3919 language tree to a file. The file name is generated by appending a switch
3920 specific suffix to the source file name. If the @samp{-@var{options}}
3921 form is used, @var{options} is a list of @samp{-} separated options that
3922 control the details of the dump. Not all options are applicable to all
3923 dumps, those which are not meaningful will be ignored. The following
3924 options are available
3928 Print the address of each node. Usually this is not meaningful as it
3929 changes according to the environment and source file. Its primary use
3930 is for tying up a dump file with a debug environment.
3932 Inhibit dumping of members of a scope or body of a function merely
3933 because that scope has been reached. Only dump such items when they
3934 are directly reachable by some other path. When dumping pretty-printed
3935 trees, this option inhibits dumping the bodies of control structures.
3937 Print a raw representation of the tree. By default, trees are
3938 pretty-printed into a C-like representation.
3940 Enable more detailed dumps (not honored by every dump option).
3942 Enable dumping various statistics about the pass (not honored by every dump
3945 Enable showing basic block boundaries (disabled in raw dumps).
3947 Enable showing virtual operands for every statement.
3949 Enable showing line numbers for statements.
3951 Enable showing the unique ID (@code{DECL_UID}) for each variable.
3953 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
3956 The following tree dumps are possible:
3960 Dump before any tree based optimization, to @file{@var{file}.original}.
3963 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3966 Dump after function inlining, to @file{@var{file}.inlined}.
3969 @opindex fdump-tree-gimple
3970 Dump each function before and after the gimplification pass to a file. The
3971 file name is made by appending @file{.gimple} to the source file name.
3974 @opindex fdump-tree-cfg
3975 Dump the control flow graph of each function to a file. The file name is
3976 made by appending @file{.cfg} to the source file name.
3979 @opindex fdump-tree-vcg
3980 Dump the control flow graph of each function to a file in VCG format. The
3981 file name is made by appending @file{.vcg} to the source file name. Note
3982 that if the file contains more than one function, the generated file cannot
3983 be used directly by VCG@. You will need to cut and paste each function's
3984 graph into its own separate file first.
3987 @opindex fdump-tree-ch
3988 Dump each function after copying loop headers. The file name is made by
3989 appending @file{.ch} to the source file name.
3992 @opindex fdump-tree-ssa
3993 Dump SSA related information to a file. The file name is made by appending
3994 @file{.ssa} to the source file name.
3997 @opindex fdump-tree-salias
3998 Dump structure aliasing variable information to a file. This file name
3999 is made by appending @file{.salias} to the source file name.
4002 @opindex fdump-tree-alias
4003 Dump aliasing information for each function. The file name is made by
4004 appending @file{.alias} to the source file name.
4007 @opindex fdump-tree-ccp
4008 Dump each function after CCP@. The file name is made by appending
4009 @file{.ccp} to the source file name.
4012 @opindex fdump-tree-storeccp
4013 Dump each function after STORE-CCP. The file name is made by appending
4014 @file{.storeccp} to the source file name.
4017 @opindex fdump-tree-pre
4018 Dump trees after partial redundancy elimination. The file name is made
4019 by appending @file{.pre} to the source file name.
4022 @opindex fdump-tree-fre
4023 Dump trees after full redundancy elimination. The file name is made
4024 by appending @file{.fre} to the source file name.
4027 @opindex fdump-tree-copyprop
4028 Dump trees after copy propagation. The file name is made
4029 by appending @file{.copyprop} to the source file name.
4031 @item store_copyprop
4032 @opindex fdump-tree-store_copyprop
4033 Dump trees after store copy-propagation. The file name is made
4034 by appending @file{.store_copyprop} to the source file name.
4037 @opindex fdump-tree-dce
4038 Dump each function after dead code elimination. The file name is made by
4039 appending @file{.dce} to the source file name.
4042 @opindex fdump-tree-mudflap
4043 Dump each function after adding mudflap instrumentation. The file name is
4044 made by appending @file{.mudflap} to the source file name.
4047 @opindex fdump-tree-sra
4048 Dump each function after performing scalar replacement of aggregates. The
4049 file name is made by appending @file{.sra} to the source file name.
4052 @opindex fdump-tree-sink
4053 Dump each function after performing code sinking. The file name is made
4054 by appending @file{.sink} to the source file name.
4057 @opindex fdump-tree-dom
4058 Dump each function after applying dominator tree optimizations. The file
4059 name is made by appending @file{.dom} to the source file name.
4062 @opindex fdump-tree-dse
4063 Dump each function after applying dead store elimination. The file
4064 name is made by appending @file{.dse} to the source file name.
4067 @opindex fdump-tree-phiopt
4068 Dump each function after optimizing PHI nodes into straightline code. The file
4069 name is made by appending @file{.phiopt} to the source file name.
4072 @opindex fdump-tree-forwprop
4073 Dump each function after forward propagating single use variables. The file
4074 name is made by appending @file{.forwprop} to the source file name.
4077 @opindex fdump-tree-copyrename
4078 Dump each function after applying the copy rename optimization. The file
4079 name is made by appending @file{.copyrename} to the source file name.
4082 @opindex fdump-tree-nrv
4083 Dump each function after applying the named return value optimization on
4084 generic trees. The file name is made by appending @file{.nrv} to the source
4088 @opindex fdump-tree-vect
4089 Dump each function after applying vectorization of loops. The file name is
4090 made by appending @file{.vect} to the source file name.
4093 @opindex fdump-tree-vrp
4094 Dump each function after Value Range Propagation (VRP). The file name
4095 is made by appending @file{.vrp} to the source file name.
4098 @opindex fdump-tree-all
4099 Enable all the available tree dumps with the flags provided in this option.
4102 @item -ftree-vectorizer-verbose=@var{n}
4103 @opindex ftree-vectorizer-verbose
4104 This option controls the amount of debugging output the vectorizer prints.
4105 This information is written to standard error, unless @option{-fdump-tree-all}
4106 or @option{-fdump-tree-vect} is specified, in which case it is output to the
4107 usual dump listing file, @file{.vect}.
4109 @item -frandom-seed=@var{string}
4110 @opindex frandom-string
4111 This option provides a seed that GCC uses when it would otherwise use
4112 random numbers. It is used to generate certain symbol names
4113 that have to be different in every compiled file. It is also used to
4114 place unique stamps in coverage data files and the object files that
4115 produce them. You can use the @option{-frandom-seed} option to produce
4116 reproducibly identical object files.
4118 The @var{string} should be different for every file you compile.
4120 @item -fsched-verbose=@var{n}
4121 @opindex fsched-verbose
4122 On targets that use instruction scheduling, this option controls the
4123 amount of debugging output the scheduler prints. This information is
4124 written to standard error, unless @option{-dS} or @option{-dR} is
4125 specified, in which case it is output to the usual dump
4126 listing file, @file{.sched} or @file{.sched2} respectively. However
4127 for @var{n} greater than nine, the output is always printed to standard
4130 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4131 same information as @option{-dRS}. For @var{n} greater than one, it
4132 also output basic block probabilities, detailed ready list information
4133 and unit/insn info. For @var{n} greater than two, it includes RTL
4134 at abort point, control-flow and regions info. And for @var{n} over
4135 four, @option{-fsched-verbose} also includes dependence info.
4139 Store the usual ``temporary'' intermediate files permanently; place them
4140 in the current directory and name them based on the source file. Thus,
4141 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4142 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4143 preprocessed @file{foo.i} output file even though the compiler now
4144 normally uses an integrated preprocessor.
4146 When used in combination with the @option{-x} command line option,
4147 @option{-save-temps} is sensible enough to avoid over writing an
4148 input source file with the same extension as an intermediate file.
4149 The corresponding intermediate file may be obtained by renaming the
4150 source file before using @option{-save-temps}.
4154 Report the CPU time taken by each subprocess in the compilation
4155 sequence. For C source files, this is the compiler proper and assembler
4156 (plus the linker if linking is done). The output looks like this:
4163 The first number on each line is the ``user time'', that is time spent
4164 executing the program itself. The second number is ``system time'',
4165 time spent executing operating system routines on behalf of the program.
4166 Both numbers are in seconds.
4168 @item -fvar-tracking
4169 @opindex fvar-tracking
4170 Run variable tracking pass. It computes where variables are stored at each
4171 position in code. Better debugging information is then generated
4172 (if the debugging information format supports this information).
4174 It is enabled by default when compiling with optimization (@option{-Os},
4175 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4176 the debug info format supports it.
4178 @item -print-file-name=@var{library}
4179 @opindex print-file-name
4180 Print the full absolute name of the library file @var{library} that
4181 would be used when linking---and don't do anything else. With this
4182 option, GCC does not compile or link anything; it just prints the
4185 @item -print-multi-directory
4186 @opindex print-multi-directory
4187 Print the directory name corresponding to the multilib selected by any
4188 other switches present in the command line. This directory is supposed
4189 to exist in @env{GCC_EXEC_PREFIX}.
4191 @item -print-multi-lib
4192 @opindex print-multi-lib
4193 Print the mapping from multilib directory names to compiler switches
4194 that enable them. The directory name is separated from the switches by
4195 @samp{;}, and each switch starts with an @samp{@@} instead of the
4196 @samp{-}, without spaces between multiple switches. This is supposed to
4197 ease shell-processing.
4199 @item -print-prog-name=@var{program}
4200 @opindex print-prog-name
4201 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4203 @item -print-libgcc-file-name
4204 @opindex print-libgcc-file-name
4205 Same as @option{-print-file-name=libgcc.a}.
4207 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4208 but you do want to link with @file{libgcc.a}. You can do
4211 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4214 @item -print-search-dirs
4215 @opindex print-search-dirs
4216 Print the name of the configured installation directory and a list of
4217 program and library directories @command{gcc} will search---and don't do anything else.
4219 This is useful when @command{gcc} prints the error message
4220 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4221 To resolve this you either need to put @file{cpp0} and the other compiler
4222 components where @command{gcc} expects to find them, or you can set the environment
4223 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4224 Don't forget the trailing @samp{/}.
4225 @xref{Environment Variables}.
4228 @opindex dumpmachine
4229 Print the compiler's target machine (for example,
4230 @samp{i686-pc-linux-gnu})---and don't do anything else.
4233 @opindex dumpversion
4234 Print the compiler version (for example, @samp{3.0})---and don't do
4239 Print the compiler's built-in specs---and don't do anything else. (This
4240 is used when GCC itself is being built.) @xref{Spec Files}.
4242 @item -feliminate-unused-debug-types
4243 @opindex feliminate-unused-debug-types
4244 Normally, when producing DWARF2 output, GCC will emit debugging
4245 information for all types declared in a compilation
4246 unit, regardless of whether or not they are actually used
4247 in that compilation unit. Sometimes this is useful, such as
4248 if, in the debugger, you want to cast a value to a type that is
4249 not actually used in your program (but is declared). More often,
4250 however, this results in a significant amount of wasted space.
4251 With this option, GCC will avoid producing debug symbol output
4252 for types that are nowhere used in the source file being compiled.
4255 @node Optimize Options
4256 @section Options That Control Optimization
4257 @cindex optimize options
4258 @cindex options, optimization
4260 These options control various sorts of optimizations.
4262 Without any optimization option, the compiler's goal is to reduce the
4263 cost of compilation and to make debugging produce the expected
4264 results. Statements are independent: if you stop the program with a
4265 breakpoint between statements, you can then assign a new value to any
4266 variable or change the program counter to any other statement in the
4267 function and get exactly the results you would expect from the source
4270 Turning on optimization flags makes the compiler attempt to improve
4271 the performance and/or code size at the expense of compilation time
4272 and possibly the ability to debug the program.
4274 The compiler performs optimization based on the knowledge it has of
4275 the program. Optimization levels @option{-O2} and above, in
4276 particular, enable @emph{unit-at-a-time} mode, which allows the
4277 compiler to consider information gained from later functions in
4278 the file when compiling a function. Compiling multiple files at
4279 once to a single output file in @emph{unit-at-a-time} mode allows
4280 the compiler to use information gained from all of the files when
4281 compiling each of them.
4283 Not all optimizations are controlled directly by a flag. Only
4284 optimizations that have a flag are listed.
4291 Optimize. Optimizing compilation takes somewhat more time, and a lot
4292 more memory for a large function.
4294 With @option{-O}, the compiler tries to reduce code size and execution
4295 time, without performing any optimizations that take a great deal of
4298 @option{-O} turns on the following optimization flags:
4299 @gccoptlist{-fdefer-pop @gol
4300 -fdelayed-branch @gol
4301 -fguess-branch-probability @gol
4302 -fcprop-registers @gol
4303 -floop-optimize @gol
4304 -fif-conversion @gol
4305 -fif-conversion2 @gol
4308 -ftree-dominator-opts @gol
4313 -ftree-copyrename @gol
4318 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4319 where doing so does not interfere with debugging.
4323 Optimize even more. GCC performs nearly all supported optimizations
4324 that do not involve a space-speed tradeoff. The compiler does not
4325 perform loop unrolling or function inlining when you specify @option{-O2}.
4326 As compared to @option{-O}, this option increases both compilation time
4327 and the performance of the generated code.
4329 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4330 also turns on the following optimization flags:
4331 @gccoptlist{-fthread-jumps @gol
4333 -foptimize-sibling-calls @gol
4334 -fcse-follow-jumps -fcse-skip-blocks @gol
4335 -fgcse -fgcse-lm @gol
4336 -fexpensive-optimizations @gol
4337 -fstrength-reduce @gol
4338 -frerun-cse-after-loop -frerun-loop-opt @gol
4341 -fschedule-insns -fschedule-insns2 @gol
4342 -fsched-interblock -fsched-spec @gol
4344 -fstrict-aliasing @gol
4345 -fdelete-null-pointer-checks @gol
4346 -freorder-blocks -freorder-functions @gol
4347 -funit-at-a-time @gol
4348 -falign-functions -falign-jumps @gol
4349 -falign-loops -falign-labels @gol
4353 Please note the warning under @option{-fgcse} about
4354 invoking @option{-O2} on programs that use computed gotos.
4358 Optimize yet more. @option{-O3} turns on all optimizations specified by
4359 @option{-O2} and also turns on the @option{-finline-functions},
4360 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4364 Do not optimize. This is the default.
4368 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4369 do not typically increase code size. It also performs further
4370 optimizations designed to reduce code size.
4372 @option{-Os} disables the following optimization flags:
4373 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4374 -falign-labels -freorder-blocks -freorder-blocks-and-partition -fprefetch-loop-arrays}
4376 If you use multiple @option{-O} options, with or without level numbers,
4377 the last such option is the one that is effective.
4380 Options of the form @option{-f@var{flag}} specify machine-independent
4381 flags. Most flags have both positive and negative forms; the negative
4382 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4383 below, only one of the forms is listed---the one you typically will
4384 use. You can figure out the other form by either removing @samp{no-}
4387 The following options control specific optimizations. They are either
4388 activated by @option{-O} options or are related to ones that are. You
4389 can use the following flags in the rare cases when ``fine-tuning'' of
4390 optimizations to be performed is desired.
4393 @item -fno-default-inline
4394 @opindex fno-default-inline
4395 Do not make member functions inline by default merely because they are
4396 defined inside the class scope (C++ only). Otherwise, when you specify
4397 @w{@option{-O}}, member functions defined inside class scope are compiled
4398 inline by default; i.e., you don't need to add @samp{inline} in front of
4399 the member function name.
4401 @item -fno-defer-pop
4402 @opindex fno-defer-pop
4403 Always pop the arguments to each function call as soon as that function
4404 returns. For machines which must pop arguments after a function call,
4405 the compiler normally lets arguments accumulate on the stack for several
4406 function calls and pops them all at once.
4408 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4412 Force memory operands to be copied into registers before doing
4413 arithmetic on them. This produces better code by making all memory
4414 references potential common subexpressions. When they are not common
4415 subexpressions, instruction combination should eliminate the separate
4416 register-load. This option is now a nop and will be removed in 4.2.
4419 @opindex fforce-addr
4420 Force memory address constants to be copied into registers before
4421 doing arithmetic on them.
4423 @item -fomit-frame-pointer
4424 @opindex fomit-frame-pointer
4425 Don't keep the frame pointer in a register for functions that
4426 don't need one. This avoids the instructions to save, set up and
4427 restore frame pointers; it also makes an extra register available
4428 in many functions. @strong{It also makes debugging impossible on
4431 On some machines, such as the VAX, this flag has no effect, because
4432 the standard calling sequence automatically handles the frame pointer
4433 and nothing is saved by pretending it doesn't exist. The
4434 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4435 whether a target machine supports this flag. @xref{Registers,,Register
4436 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4438 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4440 @item -foptimize-sibling-calls
4441 @opindex foptimize-sibling-calls
4442 Optimize sibling and tail recursive calls.
4444 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4448 Don't pay attention to the @code{inline} keyword. Normally this option
4449 is used to keep the compiler from expanding any functions inline.
4450 Note that if you are not optimizing, no functions can be expanded inline.
4452 @item -finline-functions
4453 @opindex finline-functions
4454 Integrate all simple functions into their callers. The compiler
4455 heuristically decides which functions are simple enough to be worth
4456 integrating in this way.
4458 If all calls to a given function are integrated, and the function is
4459 declared @code{static}, then the function is normally not output as
4460 assembler code in its own right.
4462 Enabled at level @option{-O3}.
4464 @item -fearly-inlining
4465 @opindex fearly-inlining
4466 Inline functions marked by @code{always_inline} and functions whose body seems
4467 smaller than the function call overhead early before doing
4468 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
4469 makes profiling significantly cheaper and usually inlining faster on programs
4470 having large chains of nested wrapper functions.
4474 @item -finline-limit=@var{n}
4475 @opindex finline-limit
4476 By default, GCC limits the size of functions that can be inlined. This flag
4477 allows the control of this limit for functions that are explicitly marked as
4478 inline (i.e., marked with the inline keyword or defined within the class
4479 definition in c++). @var{n} is the size of functions that can be inlined in
4480 number of pseudo instructions (not counting parameter handling). The default
4481 value of @var{n} is 600.
4482 Increasing this value can result in more inlined code at
4483 the cost of compilation time and memory consumption. Decreasing usually makes
4484 the compilation faster and less code will be inlined (which presumably
4485 means slower programs). This option is particularly useful for programs that
4486 use inlining heavily such as those based on recursive templates with C++.
4488 Inlining is actually controlled by a number of parameters, which may be
4489 specified individually by using @option{--param @var{name}=@var{value}}.
4490 The @option{-finline-limit=@var{n}} option sets some of these parameters
4494 @item max-inline-insns-single
4495 is set to @var{n}/2.
4496 @item max-inline-insns-auto
4497 is set to @var{n}/2.
4498 @item min-inline-insns
4499 is set to 130 or @var{n}/4, whichever is smaller.
4500 @item max-inline-insns-rtl
4504 See below for a documentation of the individual
4505 parameters controlling inlining.
4507 @emph{Note:} pseudo instruction represents, in this particular context, an
4508 abstract measurement of function's size. In no way does it represent a count
4509 of assembly instructions and as such its exact meaning might change from one
4510 release to an another.
4512 @item -fkeep-inline-functions
4513 @opindex fkeep-inline-functions
4514 In C, emit @code{static} functions that are declared @code{inline}
4515 into the object file, even if the function has been inlined into all
4516 of its callers. This switch does not affect functions using the
4517 @code{extern inline} extension in GNU C@. In C++, emit any and all
4518 inline functions into the object file.
4520 @item -fkeep-static-consts
4521 @opindex fkeep-static-consts
4522 Emit variables declared @code{static const} when optimization isn't turned
4523 on, even if the variables aren't referenced.
4525 GCC enables this option by default. If you want to force the compiler to
4526 check if the variable was referenced, regardless of whether or not
4527 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4529 @item -fmerge-constants
4530 Attempt to merge identical constants (string constants and floating point
4531 constants) across compilation units.
4533 This option is the default for optimized compilation if the assembler and
4534 linker support it. Use @option{-fno-merge-constants} to inhibit this
4537 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4539 @item -fmerge-all-constants
4540 Attempt to merge identical constants and identical variables.
4542 This option implies @option{-fmerge-constants}. In addition to
4543 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4544 arrays or initialized constant variables with integral or floating point
4545 types. Languages like C or C++ require each non-automatic variable to
4546 have distinct location, so using this option will result in non-conforming
4549 @item -fmodulo-sched
4550 @opindex fmodulo-sched
4551 Perform swing modulo scheduling immediately before the first scheduling
4552 pass. This pass looks at innermost loops and reorders their
4553 instructions by overlapping different iterations.
4555 @item -fno-branch-count-reg
4556 @opindex fno-branch-count-reg
4557 Do not use ``decrement and branch'' instructions on a count register,
4558 but instead generate a sequence of instructions that decrement a
4559 register, compare it against zero, then branch based upon the result.
4560 This option is only meaningful on architectures that support such
4561 instructions, which include x86, PowerPC, IA-64 and S/390.
4563 The default is @option{-fbranch-count-reg}, enabled when
4564 @option{-fstrength-reduce} is enabled.
4566 @item -fno-function-cse
4567 @opindex fno-function-cse
4568 Do not put function addresses in registers; make each instruction that
4569 calls a constant function contain the function's address explicitly.
4571 This option results in less efficient code, but some strange hacks
4572 that alter the assembler output may be confused by the optimizations
4573 performed when this option is not used.
4575 The default is @option{-ffunction-cse}
4577 @item -fno-zero-initialized-in-bss
4578 @opindex fno-zero-initialized-in-bss
4579 If the target supports a BSS section, GCC by default puts variables that
4580 are initialized to zero into BSS@. This can save space in the resulting
4583 This option turns off this behavior because some programs explicitly
4584 rely on variables going to the data section. E.g., so that the
4585 resulting executable can find the beginning of that section and/or make
4586 assumptions based on that.
4588 The default is @option{-fzero-initialized-in-bss}.
4590 @item -fbounds-check
4591 @opindex fbounds-check
4592 For front-ends that support it, generate additional code to check that
4593 indices used to access arrays are within the declared range. This is
4594 currently only supported by the Java and Fortran front-ends, where
4595 this option defaults to true and false respectively.
4597 @item -fmudflap -fmudflapth -fmudflapir
4601 @cindex bounds checking
4603 For front-ends that support it (C and C++), instrument all risky
4604 pointer/array dereferencing operations, some standard library
4605 string/heap functions, and some other associated constructs with
4606 range/validity tests. Modules so instrumented should be immune to
4607 buffer overflows, invalid heap use, and some other classes of C/C++
4608 programming errors. The instrumentation relies on a separate runtime
4609 library (@file{libmudflap}), which will be linked into a program if
4610 @option{-fmudflap} is given at link time. Run-time behavior of the
4611 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4612 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4615 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4616 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4617 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4618 instrumentation should ignore pointer reads. This produces less
4619 instrumentation (and therefore faster execution) and still provides
4620 some protection against outright memory corrupting writes, but allows
4621 erroneously read data to propagate within a program.
4623 @item -fstrength-reduce
4624 @opindex fstrength-reduce
4625 Perform the optimizations of loop strength reduction and
4626 elimination of iteration variables.
4628 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4630 @item -fthread-jumps
4631 @opindex fthread-jumps
4632 Perform optimizations where we check to see if a jump branches to a
4633 location where another comparison subsumed by the first is found. If
4634 so, the first branch is redirected to either the destination of the
4635 second branch or a point immediately following it, depending on whether
4636 the condition is known to be true or false.
4638 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4640 @item -fcse-follow-jumps
4641 @opindex fcse-follow-jumps
4642 In common subexpression elimination, scan through jump instructions
4643 when the target of the jump is not reached by any other path. For
4644 example, when CSE encounters an @code{if} statement with an
4645 @code{else} clause, CSE will follow the jump when the condition
4648 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4650 @item -fcse-skip-blocks
4651 @opindex fcse-skip-blocks
4652 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4653 follow jumps which conditionally skip over blocks. When CSE
4654 encounters a simple @code{if} statement with no else clause,
4655 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4656 body of the @code{if}.
4658 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4660 @item -frerun-cse-after-loop
4661 @opindex frerun-cse-after-loop
4662 Re-run common subexpression elimination after loop optimizations has been
4665 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4667 @item -frerun-loop-opt
4668 @opindex frerun-loop-opt
4669 Run the loop optimizer twice.
4671 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4675 Perform a global common subexpression elimination pass.
4676 This pass also performs global constant and copy propagation.
4678 @emph{Note:} When compiling a program using computed gotos, a GCC
4679 extension, you may get better runtime performance if you disable
4680 the global common subexpression elimination pass by adding
4681 @option{-fno-gcse} to the command line.
4683 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4687 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4688 attempt to move loads which are only killed by stores into themselves. This
4689 allows a loop containing a load/store sequence to be changed to a load outside
4690 the loop, and a copy/store within the loop.
4692 Enabled by default when gcse is enabled.
4696 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4697 global common subexpression elimination. This pass will attempt to move
4698 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4699 loops containing a load/store sequence can be changed to a load before
4700 the loop and a store after the loop.
4702 Not enabled at any optimization level.
4706 When @option{-fgcse-las} is enabled, the global common subexpression
4707 elimination pass eliminates redundant loads that come after stores to the
4708 same memory location (both partial and full redundancies).
4710 Not enabled at any optimization level.
4712 @item -fgcse-after-reload
4713 @opindex fgcse-after-reload
4714 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4715 pass is performed after reload. The purpose of this pass is to cleanup
4718 @item -floop-optimize
4719 @opindex floop-optimize
4720 Perform loop optimizations: move constant expressions out of loops, simplify
4721 exit test conditions and optionally do strength-reduction as well.
4723 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4725 @item -floop-optimize2
4726 @opindex floop-optimize2
4727 Perform loop optimizations using the new loop optimizer. The optimizations
4728 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4731 @item -funsafe-loop-optimizations
4732 @opindex funsafe-loop-optimizations
4733 If given, the loop optimizer will assume that loop indices do not
4734 overflow, and that the loops with nontrivial exit condition are not
4735 infinite. This enables a wider range of loop optimizations even if
4736 the loop optimizer itself cannot prove that these assumptions are valid.
4737 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
4738 if it finds this kind of loop.
4740 @item -fcrossjumping
4741 @opindex crossjumping
4742 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4743 resulting code may or may not perform better than without cross-jumping.
4745 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4747 @item -fif-conversion
4748 @opindex if-conversion
4749 Attempt to transform conditional jumps into branch-less equivalents. This
4750 include use of conditional moves, min, max, set flags and abs instructions, and
4751 some tricks doable by standard arithmetics. The use of conditional execution
4752 on chips where it is available is controlled by @code{if-conversion2}.
4754 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4756 @item -fif-conversion2
4757 @opindex if-conversion2
4758 Use conditional execution (where available) to transform conditional jumps into
4759 branch-less equivalents.
4761 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4763 @item -fdelete-null-pointer-checks
4764 @opindex fdelete-null-pointer-checks
4765 Use global dataflow analysis to identify and eliminate useless checks
4766 for null pointers. The compiler assumes that dereferencing a null
4767 pointer would have halted the program. If a pointer is checked after
4768 it has already been dereferenced, it cannot be null.
4770 In some environments, this assumption is not true, and programs can
4771 safely dereference null pointers. Use
4772 @option{-fno-delete-null-pointer-checks} to disable this optimization
4773 for programs which depend on that behavior.
4775 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4777 @item -fexpensive-optimizations
4778 @opindex fexpensive-optimizations
4779 Perform a number of minor optimizations that are relatively expensive.
4781 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4783 @item -foptimize-register-move
4785 @opindex foptimize-register-move
4787 Attempt to reassign register numbers in move instructions and as
4788 operands of other simple instructions in order to maximize the amount of
4789 register tying. This is especially helpful on machines with two-operand
4792 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4795 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4797 @item -fdelayed-branch
4798 @opindex fdelayed-branch
4799 If supported for the target machine, attempt to reorder instructions
4800 to exploit instruction slots available after delayed branch
4803 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4805 @item -fschedule-insns
4806 @opindex fschedule-insns
4807 If supported for the target machine, attempt to reorder instructions to
4808 eliminate execution stalls due to required data being unavailable. This
4809 helps machines that have slow floating point or memory load instructions
4810 by allowing other instructions to be issued until the result of the load
4811 or floating point instruction is required.
4813 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4815 @item -fschedule-insns2
4816 @opindex fschedule-insns2
4817 Similar to @option{-fschedule-insns}, but requests an additional pass of
4818 instruction scheduling after register allocation has been done. This is
4819 especially useful on machines with a relatively small number of
4820 registers and where memory load instructions take more than one cycle.
4822 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4824 @item -fno-sched-interblock
4825 @opindex fno-sched-interblock
4826 Don't schedule instructions across basic blocks. This is normally
4827 enabled by default when scheduling before register allocation, i.e.@:
4828 with @option{-fschedule-insns} or at @option{-O2} or higher.
4830 @item -fno-sched-spec
4831 @opindex fno-sched-spec
4832 Don't allow speculative motion of non-load instructions. This is normally
4833 enabled by default when scheduling before register allocation, i.e.@:
4834 with @option{-fschedule-insns} or at @option{-O2} or higher.
4836 @item -fsched-spec-load
4837 @opindex fsched-spec-load
4838 Allow speculative motion of some load instructions. This only makes
4839 sense when scheduling before register allocation, i.e.@: with
4840 @option{-fschedule-insns} or at @option{-O2} or higher.
4842 @item -fsched-spec-load-dangerous
4843 @opindex fsched-spec-load-dangerous
4844 Allow speculative motion of more load instructions. This only makes
4845 sense when scheduling before register allocation, i.e.@: with
4846 @option{-fschedule-insns} or at @option{-O2} or higher.
4848 @item -fsched-stalled-insns=@var{n}
4849 @opindex fsched-stalled-insns
4850 Define how many insns (if any) can be moved prematurely from the queue
4851 of stalled insns into the ready list, during the second scheduling pass.
4853 @item -fsched-stalled-insns-dep=@var{n}
4854 @opindex fsched-stalled-insns-dep
4855 Define how many insn groups (cycles) will be examined for a dependency
4856 on a stalled insn that is candidate for premature removal from the queue
4857 of stalled insns. Has an effect only during the second scheduling pass,
4858 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4860 @item -fsched2-use-superblocks
4861 @opindex fsched2-use-superblocks
4862 When scheduling after register allocation, do use superblock scheduling
4863 algorithm. Superblock scheduling allows motion across basic block boundaries
4864 resulting on faster schedules. This option is experimental, as not all machine
4865 descriptions used by GCC model the CPU closely enough to avoid unreliable
4866 results from the algorithm.
4868 This only makes sense when scheduling after register allocation, i.e.@: with
4869 @option{-fschedule-insns2} or at @option{-O2} or higher.
4871 @item -fsched2-use-traces
4872 @opindex fsched2-use-traces
4873 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4874 allocation and additionally perform code duplication in order to increase the
4875 size of superblocks using tracer pass. See @option{-ftracer} for details on
4878 This mode should produce faster but significantly longer programs. Also
4879 without @option{-fbranch-probabilities} the traces constructed may not
4880 match the reality and hurt the performance. This only makes
4881 sense when scheduling after register allocation, i.e.@: with
4882 @option{-fschedule-insns2} or at @option{-O2} or higher.
4884 @item -freschedule-modulo-scheduled-loops
4885 @opindex fscheduling-in-modulo-scheduled-loops
4886 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4887 we may want to prevent the later scheduling passes from changing its schedule, we use this
4888 option to control that.
4890 @item -fcaller-saves
4891 @opindex fcaller-saves
4892 Enable values to be allocated in registers that will be clobbered by
4893 function calls, by emitting extra instructions to save and restore the
4894 registers around such calls. Such allocation is done only when it
4895 seems to result in better code than would otherwise be produced.
4897 This option is always enabled by default on certain machines, usually
4898 those which have no call-preserved registers to use instead.
4900 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4903 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4904 enabled by default at @option{-O2} and @option{-O3}.
4907 Perform Full Redundancy Elimination (FRE) on trees. The difference
4908 between FRE and PRE is that FRE only considers expressions
4909 that are computed on all paths leading to the redundant computation.
4910 This analysis faster than PRE, though it exposes fewer redundancies.
4911 This flag is enabled by default at @option{-O} and higher.
4913 @item -ftree-copy-prop
4914 Perform copy propagation on trees. This pass eliminates unnecessary
4915 copy operations. This flag is enabled by default at @option{-O} and
4918 @item -ftree-store-copy-prop
4919 Perform copy propagation of memory loads and stores. This pass
4920 eliminates unnecessary copy operations in memory references
4921 (structures, global variables, arrays, etc). This flag is enabled by
4922 default at @option{-O2} and higher.
4925 Perform structural alias analysis on trees. This flag
4926 is enabled by default at @option{-O} and higher.
4929 Perform forward store motion on trees. This flag is
4930 enabled by default at @option{-O} and higher.
4933 Perform sparse conditional constant propagation (CCP) on trees. This
4934 pass only operates on local scalar variables and is enabled by default
4935 at @option{-O} and higher.
4937 @item -ftree-store-ccp
4938 Perform sparse conditional constant propagation (CCP) on trees. This
4939 pass operates on both local scalar variables and memory stores and
4940 loads (global variables, structures, arrays, etc). This flag is
4941 enabled by default at @option{-O2} and higher.
4944 Perform dead code elimination (DCE) on trees. This flag is enabled by
4945 default at @option{-O} and higher.
4947 @item -ftree-dominator-opts
4948 Perform dead code elimination (DCE) on trees. This flag is enabled by
4949 default at @option{-O} and higher.
4952 Perform loop header copying on trees. This is beneficial since it increases
4953 effectiveness of code motion optimizations. It also saves one jump. This flag
4954 is enabled by default at @option{-O} and higher. It is not enabled
4955 for @option{-Os}, since it usually increases code size.
4957 @item -ftree-loop-optimize
4958 Perform loop optimizations on trees. This flag is enabled by default
4959 at @option{-O} and higher.
4961 @item -ftree-loop-linear
4962 Perform linear loop transformations on tree. This flag can improve cache
4963 performance and allow further loop optimizations to take place.
4965 @item -ftree-loop-im
4966 Perform loop invariant motion on trees. This pass moves only invariants that
4967 would be hard to handle at RTL level (function calls, operations that expand to
4968 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
4969 operands of conditions that are invariant out of the loop, so that we can use
4970 just trivial invariantness analysis in loop unswitching. The pass also includes
4973 @item -ftree-loop-ivcanon
4974 Create a canonical counter for number of iterations in the loop for that
4975 determining number of iterations requires complicated analysis. Later
4976 optimizations then may determine the number easily. Useful especially
4977 in connection with unrolling.
4980 Perform induction variable optimizations (strength reduction, induction
4981 variable merging and induction variable elimination) on trees.
4984 Perform scalar replacement of aggregates. This pass replaces structure
4985 references with scalars to prevent committing structures to memory too
4986 early. This flag is enabled by default at @option{-O} and higher.
4988 @item -ftree-copyrename
4989 Perform copy renaming on trees. This pass attempts to rename compiler
4990 temporaries to other variables at copy locations, usually resulting in
4991 variable names which more closely resemble the original variables. This flag
4992 is enabled by default at @option{-O} and higher.
4995 Perform temporary expression replacement during the SSA->normal phase. Single
4996 use/single def temporaries are replaced at their use location with their
4997 defining expression. This results in non-GIMPLE code, but gives the expanders
4998 much more complex trees to work on resulting in better RTL generation. This is
4999 enabled by default at @option{-O} and higher.
5002 Perform live range splitting during the SSA->normal phase. Distinct live
5003 ranges of a variable are split into unique variables, allowing for better
5004 optimization later. This is enabled by default at @option{-O} and higher.
5006 @item -ftree-vectorize
5007 Perform loop vectorization on trees.
5010 Perform Value Range Propagation on trees. This is similar to the
5011 constant propagation pass, but instead of values, ranges of values are
5012 propagated. This allows the optimizers to remove unnecessary range
5013 checks like array bound checks and null pointer checks. This is
5014 enabled by default at @option{-O2} and higher. Null pointer check
5015 elimination is only done if @option{-fdelete-null-pointer-checks} is
5020 Perform tail duplication to enlarge superblock size. This transformation
5021 simplifies the control flow of the function allowing other optimizations to do
5024 @item -funroll-loops
5025 @opindex funroll-loops
5026 Unroll loops whose number of iterations can be determined at compile
5027 time or upon entry to the loop. @option{-funroll-loops} implies both
5028 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
5029 option makes code larger, and may or may not make it run faster.
5031 @item -funroll-all-loops
5032 @opindex funroll-all-loops
5033 Unroll all loops, even if their number of iterations is uncertain when
5034 the loop is entered. This usually makes programs run more slowly.
5035 @option{-funroll-all-loops} implies the same options as
5036 @option{-funroll-loops},
5038 @item -fsplit-ivs-in-unroller
5039 @opindex -fsplit-ivs-in-unroller
5040 Enables expressing of values of induction variables in later iterations
5041 of the unrolled loop using the value in the first iteration. This breaks
5042 long dependency chains, thus improving efficiency of the scheduling passes.
5044 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5045 same effect. However in cases the loop body is more complicated than
5046 a single basic block, this is not reliable. It also does not work at all
5047 on some of the architectures due to restrictions in the CSE pass.
5049 This optimization is enabled by default.
5051 @item -fvariable-expansion-in-unroller
5052 @opindex -fvariable-expansion-in-unroller
5053 With this option, the compiler will create multiple copies of some
5054 local variables when unrolling a loop which can result in superior code.
5056 @item -fprefetch-loop-arrays
5057 @opindex fprefetch-loop-arrays
5058 If supported by the target machine, generate instructions to prefetch
5059 memory to improve the performance of loops that access large arrays.
5061 These options may generate better or worse code; results are highly
5062 dependent on the structure of loops within the source code.
5065 @itemx -fno-peephole2
5066 @opindex fno-peephole
5067 @opindex fno-peephole2
5068 Disable any machine-specific peephole optimizations. The difference
5069 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5070 are implemented in the compiler; some targets use one, some use the
5071 other, a few use both.
5073 @option{-fpeephole} is enabled by default.
5074 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5076 @item -fno-guess-branch-probability
5077 @opindex fno-guess-branch-probability
5078 Do not guess branch probabilities using heuristics.
5080 GCC will use heuristics to guess branch probabilities if they are
5081 not provided by profiling feedback (@option{-fprofile-arcs}). These
5082 heuristics are based on the control flow graph. If some branch probabilities
5083 are specified by @samp{__builtin_expect}, then the heuristics will be
5084 used to guess branch probabilities for the rest of the control flow graph,
5085 taking the @samp{__builtin_expect} info into account. The interactions
5086 between the heuristics and @samp{__builtin_expect} can be complex, and in
5087 some cases, it may be useful to disable the heuristics so that the effects
5088 of @samp{__builtin_expect} are easier to understand.
5090 The default is @option{-fguess-branch-probability} at levels
5091 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5093 @item -freorder-blocks
5094 @opindex freorder-blocks
5095 Reorder basic blocks in the compiled function in order to reduce number of
5096 taken branches and improve code locality.
5098 Enabled at levels @option{-O2}, @option{-O3}.
5100 @item -freorder-blocks-and-partition
5101 @opindex freorder-blocks-and-partition
5102 In addition to reordering basic blocks in the compiled function, in order
5103 to reduce number of taken branches, partitions hot and cold basic blocks
5104 into separate sections of the assembly and .o files, to improve
5105 paging and cache locality performance.
5107 This optimization is automatically turned off in the presence of
5108 exception handling, for linkonce sections, for functions with a user-defined
5109 section attribute and on any architecture that does not support named
5112 @item -freorder-functions
5113 @opindex freorder-functions
5114 Reorder functions in the object file in order to
5115 improve code locality. This is implemented by using special
5116 subsections @code{.text.hot} for most frequently executed functions and
5117 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5118 the linker so object file format must support named sections and linker must
5119 place them in a reasonable way.
5121 Also profile feedback must be available in to make this option effective. See
5122 @option{-fprofile-arcs} for details.
5124 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5126 @item -fstrict-aliasing
5127 @opindex fstrict-aliasing
5128 Allows the compiler to assume the strictest aliasing rules applicable to
5129 the language being compiled. For C (and C++), this activates
5130 optimizations based on the type of expressions. In particular, an
5131 object of one type is assumed never to reside at the same address as an
5132 object of a different type, unless the types are almost the same. For
5133 example, an @code{unsigned int} can alias an @code{int}, but not a
5134 @code{void*} or a @code{double}. A character type may alias any other
5137 Pay special attention to code like this:
5150 The practice of reading from a different union member than the one most
5151 recently written to (called ``type-punning'') is common. Even with
5152 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5153 is accessed through the union type. So, the code above will work as
5154 expected. However, this code might not:
5165 Every language that wishes to perform language-specific alias analysis
5166 should define a function that computes, given an @code{tree}
5167 node, an alias set for the node. Nodes in different alias sets are not
5168 allowed to alias. For an example, see the C front-end function
5169 @code{c_get_alias_set}.
5171 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5173 @item -falign-functions
5174 @itemx -falign-functions=@var{n}
5175 @opindex falign-functions
5176 Align the start of functions to the next power-of-two greater than
5177 @var{n}, skipping up to @var{n} bytes. For instance,
5178 @option{-falign-functions=32} aligns functions to the next 32-byte
5179 boundary, but @option{-falign-functions=24} would align to the next
5180 32-byte boundary only if this can be done by skipping 23 bytes or less.
5182 @option{-fno-align-functions} and @option{-falign-functions=1} are
5183 equivalent and mean that functions will not be aligned.
5185 Some assemblers only support this flag when @var{n} is a power of two;
5186 in that case, it is rounded up.
5188 If @var{n} is not specified or is zero, use a machine-dependent default.
5190 Enabled at levels @option{-O2}, @option{-O3}.
5192 @item -falign-labels
5193 @itemx -falign-labels=@var{n}
5194 @opindex falign-labels
5195 Align all branch targets to a power-of-two boundary, skipping up to
5196 @var{n} bytes like @option{-falign-functions}. This option can easily
5197 make code slower, because it must insert dummy operations for when the
5198 branch target is reached in the usual flow of the code.
5200 @option{-fno-align-labels} and @option{-falign-labels=1} are
5201 equivalent and mean that labels will not be aligned.
5203 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5204 are greater than this value, then their values are used instead.
5206 If @var{n} is not specified or is zero, use a machine-dependent default
5207 which is very likely to be @samp{1}, meaning no alignment.
5209 Enabled at levels @option{-O2}, @option{-O3}.
5212 @itemx -falign-loops=@var{n}
5213 @opindex falign-loops
5214 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5215 like @option{-falign-functions}. The hope is that the loop will be
5216 executed many times, which will make up for any execution of the dummy
5219 @option{-fno-align-loops} and @option{-falign-loops=1} are
5220 equivalent and mean that loops will not be aligned.
5222 If @var{n} is not specified or is zero, use a machine-dependent default.
5224 Enabled at levels @option{-O2}, @option{-O3}.
5227 @itemx -falign-jumps=@var{n}
5228 @opindex falign-jumps
5229 Align branch targets to a power-of-two boundary, for branch targets
5230 where the targets can only be reached by jumping, skipping up to @var{n}
5231 bytes like @option{-falign-functions}. In this case, no dummy operations
5234 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5235 equivalent and mean that loops will not be aligned.
5237 If @var{n} is not specified or is zero, use a machine-dependent default.
5239 Enabled at levels @option{-O2}, @option{-O3}.
5241 @item -funit-at-a-time
5242 @opindex funit-at-a-time
5243 Parse the whole compilation unit before starting to produce code.
5244 This allows some extra optimizations to take place but consumes
5245 more memory (in general). There are some compatibility issues
5246 with @emph{unit-at-at-time} mode:
5249 enabling @emph{unit-at-a-time} mode may change the order
5250 in which functions, variables, and top-level @code{asm} statements
5251 are emitted, and will likely break code relying on some particular
5252 ordering. The majority of such top-level @code{asm} statements,
5253 though, can be replaced by @code{section} attributes.
5256 @emph{unit-at-a-time} mode removes unreferenced static variables
5257 and functions. This may result in undefined references
5258 when an @code{asm} statement refers directly to variables or functions
5259 that are otherwise unused. In that case either the variable/function
5260 shall be listed as an operand of the @code{asm} statement operand or,
5261 in the case of top-level @code{asm} statements the attribute @code{used}
5262 shall be used on the declaration.
5265 Static functions now can use non-standard passing conventions that
5266 may break @code{asm} statements calling functions directly. Again,
5267 attribute @code{used} will prevent this behavior.
5270 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5271 but this scheme may not be supported by future releases of GCC@.
5273 Enabled at levels @option{-O2}, @option{-O3}.
5277 Constructs webs as commonly used for register allocation purposes and assign
5278 each web individual pseudo register. This allows the register allocation pass
5279 to operate on pseudos directly, but also strengthens several other optimization
5280 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5281 however, make debugging impossible, since variables will no longer stay in a
5284 Enabled by default with @option{-funroll-loops}.
5286 @item -fwhole-program
5287 @opindex fwhole-program
5288 Assume that the current compilation unit represents whole program being
5289 compiled. All public functions and variables with the exception of @code{main}
5290 and those merged by attribute @code{externally_visible} become static functions
5291 and in a affect gets more aggressively optimized by interprocedural optimizers.
5292 While this option is equivalent to proper use of @code{static} keyword for
5293 programs consisting of single file, in combination with option
5294 @option{--combine} this flag can be used to compile most of smaller scale C
5295 programs since the functions and variables become local for the whole combined
5296 compilation unit, not for the single source file itself.
5299 @item -fno-cprop-registers
5300 @opindex fno-cprop-registers
5301 After register allocation and post-register allocation instruction splitting,
5302 we perform a copy-propagation pass to try to reduce scheduling dependencies
5303 and occasionally eliminate the copy.
5305 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5307 @item -fprofile-generate
5308 @opindex fprofile-generate
5310 Enable options usually used for instrumenting application to produce
5311 profile useful for later recompilation with profile feedback based
5312 optimization. You must use @option{-fprofile-generate} both when
5313 compiling and when linking your program.
5315 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5318 @opindex fprofile-use
5319 Enable profile feedback directed optimizations, and optimizations
5320 generally profitable only with profile feedback available.
5322 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
5323 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer},
5324 @code{-fno-loop-optimize}.
5328 The following options control compiler behavior regarding floating
5329 point arithmetic. These options trade off between speed and
5330 correctness. All must be specifically enabled.
5334 @opindex ffloat-store
5335 Do not store floating point variables in registers, and inhibit other
5336 options that might change whether a floating point value is taken from a
5339 @cindex floating point precision
5340 This option prevents undesirable excess precision on machines such as
5341 the 68000 where the floating registers (of the 68881) keep more
5342 precision than a @code{double} is supposed to have. Similarly for the
5343 x86 architecture. For most programs, the excess precision does only
5344 good, but a few programs rely on the precise definition of IEEE floating
5345 point. Use @option{-ffloat-store} for such programs, after modifying
5346 them to store all pertinent intermediate computations into variables.
5350 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5351 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5352 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5353 and @option{fcx-limited-range}.
5355 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5357 This option should never be turned on by any @option{-O} option since
5358 it can result in incorrect output for programs which depend on
5359 an exact implementation of IEEE or ISO rules/specifications for
5362 @item -fno-math-errno
5363 @opindex fno-math-errno
5364 Do not set ERRNO after calling math functions that are executed
5365 with a single instruction, e.g., sqrt. A program that relies on
5366 IEEE exceptions for math error handling may want to use this flag
5367 for speed while maintaining IEEE arithmetic compatibility.
5369 This option should never be turned on by any @option{-O} option since
5370 it can result in incorrect output for programs which depend on
5371 an exact implementation of IEEE or ISO rules/specifications for
5374 The default is @option{-fmath-errno}.
5376 On Darwin systems, the math library never sets @code{errno}. There is therefore
5377 no reason for the compiler to consider the possibility that it might,
5378 and @option{-fno-math-errno} is the default.
5380 @item -funsafe-math-optimizations
5381 @opindex funsafe-math-optimizations
5382 Allow optimizations for floating-point arithmetic that (a) assume
5383 that arguments and results are valid and (b) may violate IEEE or
5384 ANSI standards. When used at link-time, it may include libraries
5385 or startup files that change the default FPU control word or other
5386 similar optimizations.
5388 This option should never be turned on by any @option{-O} option since
5389 it can result in incorrect output for programs which depend on
5390 an exact implementation of IEEE or ISO rules/specifications for
5393 The default is @option{-fno-unsafe-math-optimizations}.
5395 @item -ffinite-math-only
5396 @opindex ffinite-math-only
5397 Allow optimizations for floating-point arithmetic that assume
5398 that arguments and results are not NaNs or +-Infs.
5400 This option should never be turned on by any @option{-O} option since
5401 it can result in incorrect output for programs which depend on
5402 an exact implementation of IEEE or ISO rules/specifications.
5404 The default is @option{-fno-finite-math-only}.
5406 @item -fno-trapping-math
5407 @opindex fno-trapping-math
5408 Compile code assuming that floating-point operations cannot generate
5409 user-visible traps. These traps include division by zero, overflow,
5410 underflow, inexact result and invalid operation. This option implies
5411 @option{-fno-signaling-nans}. Setting this option may allow faster
5412 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5414 This option should never be turned on by any @option{-O} option since
5415 it can result in incorrect output for programs which depend on
5416 an exact implementation of IEEE or ISO rules/specifications for
5419 The default is @option{-ftrapping-math}.
5421 @item -frounding-math
5422 @opindex frounding-math
5423 Disable transformations and optimizations that assume default floating
5424 point rounding behavior. This is round-to-zero for all floating point
5425 to integer conversions, and round-to-nearest for all other arithmetic
5426 truncations. This option should be specified for programs that change
5427 the FP rounding mode dynamically, or that may be executed with a
5428 non-default rounding mode. This option disables constant folding of
5429 floating point expressions at compile-time (which may be affected by
5430 rounding mode) and arithmetic transformations that are unsafe in the
5431 presence of sign-dependent rounding modes.
5433 The default is @option{-fno-rounding-math}.
5435 This option is experimental and does not currently guarantee to
5436 disable all GCC optimizations that are affected by rounding mode.
5437 Future versions of GCC may provide finer control of this setting
5438 using C99's @code{FENV_ACCESS} pragma. This command line option
5439 will be used to specify the default state for @code{FENV_ACCESS}.
5441 @item -fsignaling-nans
5442 @opindex fsignaling-nans
5443 Compile code assuming that IEEE signaling NaNs may generate user-visible
5444 traps during floating-point operations. Setting this option disables
5445 optimizations that may change the number of exceptions visible with
5446 signaling NaNs. This option implies @option{-ftrapping-math}.
5448 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5451 The default is @option{-fno-signaling-nans}.
5453 This option is experimental and does not currently guarantee to
5454 disable all GCC optimizations that affect signaling NaN behavior.
5456 @item -fsingle-precision-constant
5457 @opindex fsingle-precision-constant
5458 Treat floating point constant as single precision constant instead of
5459 implicitly converting it to double precision constant.
5461 @item -fcx-limited-range
5462 @itemx -fno-cx-limited-range
5463 @opindex fcx-limited-range
5464 @opindex fno-cx-limited-range
5465 When enabled, this option states that a range reduction step is not
5466 needed when performing complex division. The default is
5467 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5469 This option controls the default setting of the ISO C99
5470 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5475 The following options control optimizations that may improve
5476 performance, but are not enabled by any @option{-O} options. This
5477 section includes experimental options that may produce broken code.
5480 @item -fbranch-probabilities
5481 @opindex fbranch-probabilities
5482 After running a program compiled with @option{-fprofile-arcs}
5483 (@pxref{Debugging Options,, Options for Debugging Your Program or
5484 @command{gcc}}), you can compile it a second time using
5485 @option{-fbranch-probabilities}, to improve optimizations based on
5486 the number of times each branch was taken. When the program
5487 compiled with @option{-fprofile-arcs} exits it saves arc execution
5488 counts to a file called @file{@var{sourcename}.gcda} for each source
5489 file The information in this data file is very dependent on the
5490 structure of the generated code, so you must use the same source code
5491 and the same optimization options for both compilations.
5493 With @option{-fbranch-probabilities}, GCC puts a
5494 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5495 These can be used to improve optimization. Currently, they are only
5496 used in one place: in @file{reorg.c}, instead of guessing which path a
5497 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5498 exactly determine which path is taken more often.
5500 @item -fprofile-values
5501 @opindex fprofile-values
5502 If combined with @option{-fprofile-arcs}, it adds code so that some
5503 data about values of expressions in the program is gathered.
5505 With @option{-fbranch-probabilities}, it reads back the data gathered
5506 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5507 notes to instructions for their later usage in optimizations.
5509 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5513 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5514 a code to gather information about values of expressions.
5516 With @option{-fbranch-probabilities}, it reads back the data gathered
5517 and actually performs the optimizations based on them.
5518 Currently the optimizations include specialization of division operation
5519 using the knowledge about the value of the denominator.
5521 @item -frename-registers
5522 @opindex frename-registers
5523 Attempt to avoid false dependencies in scheduled code by making use
5524 of registers left over after register allocation. This optimization
5525 will most benefit processors with lots of registers. Depending on the
5526 debug information format adopted by the target, however, it can
5527 make debugging impossible, since variables will no longer stay in
5528 a ``home register''.
5530 Enabled by default with @option{-funroll-loops}.
5534 Perform tail duplication to enlarge superblock size. This transformation
5535 simplifies the control flow of the function allowing other optimizations to do
5538 Enabled with @option{-fprofile-use}.
5540 @item -funroll-loops
5541 @opindex funroll-loops
5542 Unroll loops whose number of iterations can be determined at compile time or
5543 upon entry to the loop. @option{-funroll-loops} implies
5544 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
5545 It also turns on complete loop peeling (i.e.@: complete removal of loops with
5546 small constant number of iterations). This option makes code larger, and may
5547 or may not make it run faster.
5549 Enabled with @option{-fprofile-use}.
5551 @item -funroll-all-loops
5552 @opindex funroll-all-loops
5553 Unroll all loops, even if their number of iterations is uncertain when
5554 the loop is entered. This usually makes programs run more slowly.
5555 @option{-funroll-all-loops} implies the same options as
5556 @option{-funroll-loops}.
5559 @opindex fpeel-loops
5560 Peels the loops for that there is enough information that they do not
5561 roll much (from profile feedback). It also turns on complete loop peeling
5562 (i.e.@: complete removal of loops with small constant number of iterations).
5564 Enabled with @option{-fprofile-use}.
5566 @item -fmove-loop-invariants
5567 @opindex fmove-loop-invariants
5568 Enables the loop invariant motion pass in the new loop optimizer. Enabled
5569 at level @option{-O1}
5571 @item -funswitch-loops
5572 @opindex funswitch-loops
5573 Move branches with loop invariant conditions out of the loop, with duplicates
5574 of the loop on both branches (modified according to result of the condition).
5576 @item -fprefetch-loop-arrays
5577 @opindex fprefetch-loop-arrays
5578 If supported by the target machine, generate instructions to prefetch
5579 memory to improve the performance of loops that access large arrays.
5581 Disabled at level @option{-Os}.
5583 @item -ffunction-sections
5584 @itemx -fdata-sections
5585 @opindex ffunction-sections
5586 @opindex fdata-sections
5587 Place each function or data item into its own section in the output
5588 file if the target supports arbitrary sections. The name of the
5589 function or the name of the data item determines the section's name
5592 Use these options on systems where the linker can perform optimizations
5593 to improve locality of reference in the instruction space. Most systems
5594 using the ELF object format and SPARC processors running Solaris 2 have
5595 linkers with such optimizations. AIX may have these optimizations in
5598 Only use these options when there are significant benefits from doing
5599 so. When you specify these options, the assembler and linker will
5600 create larger object and executable files and will also be slower.
5601 You will not be able to use @code{gprof} on all systems if you
5602 specify this option and you may have problems with debugging if
5603 you specify both this option and @option{-g}.
5605 @item -fbranch-target-load-optimize
5606 @opindex fbranch-target-load-optimize
5607 Perform branch target register load optimization before prologue / epilogue
5609 The use of target registers can typically be exposed only during reload,
5610 thus hoisting loads out of loops and doing inter-block scheduling needs
5611 a separate optimization pass.
5613 @item -fbranch-target-load-optimize2
5614 @opindex fbranch-target-load-optimize2
5615 Perform branch target register load optimization after prologue / epilogue
5618 @item -fbtr-bb-exclusive
5619 @opindex fbtr-bb-exclusive
5620 When performing branch target register load optimization, don't reuse
5621 branch target registers in within any basic block.
5623 @item --param @var{name}=@var{value}
5625 In some places, GCC uses various constants to control the amount of
5626 optimization that is done. For example, GCC will not inline functions
5627 that contain more that a certain number of instructions. You can
5628 control some of these constants on the command-line using the
5629 @option{--param} option.
5631 The names of specific parameters, and the meaning of the values, are
5632 tied to the internals of the compiler, and are subject to change
5633 without notice in future releases.
5635 In each case, the @var{value} is an integer. The allowable choices for
5636 @var{name} are given in the following table:
5639 @item salias-max-implicit-fields
5640 The maximum number of fields in a variable without direct
5641 structure accesses for which structure aliasing will consider trying
5642 to track each field. The default is 5
5644 @item sra-max-structure-size
5645 The maximum structure size, in bytes, at which the scalar replacement
5646 of aggregates (SRA) optimization will perform block copies. The
5647 default value, 0, implies that GCC will select the most appropriate
5650 @item sra-field-structure-ratio
5651 The threshold ratio (as a percentage) between instantiated fields and
5652 the complete structure size. We say that if the ratio of the number
5653 of bytes in instantiated fields to the number of bytes in the complete
5654 structure exceeds this parameter, then block copies are not used. The
5657 @item max-crossjump-edges
5658 The maximum number of incoming edges to consider for crossjumping.
5659 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5660 the number of edges incoming to each block. Increasing values mean
5661 more aggressive optimization, making the compile time increase with
5662 probably small improvement in executable size.
5664 @item min-crossjump-insns
5665 The minimum number of instructions which must be matched at the end
5666 of two blocks before crossjumping will be performed on them. This
5667 value is ignored in the case where all instructions in the block being
5668 crossjumped from are matched. The default value is 5.
5670 @item max-goto-duplication-insns
5671 The maximum number of instructions to duplicate to a block that jumps
5672 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5673 passes, GCC factors computed gotos early in the compilation process,
5674 and unfactors them as late as possible. Only computed jumps at the
5675 end of a basic blocks with no more than max-goto-duplication-insns are
5676 unfactored. The default value is 8.
5678 @item max-delay-slot-insn-search
5679 The maximum number of instructions to consider when looking for an
5680 instruction to fill a delay slot. If more than this arbitrary number of
5681 instructions is searched, the time savings from filling the delay slot
5682 will be minimal so stop searching. Increasing values mean more
5683 aggressive optimization, making the compile time increase with probably
5684 small improvement in executable run time.
5686 @item max-delay-slot-live-search
5687 When trying to fill delay slots, the maximum number of instructions to
5688 consider when searching for a block with valid live register
5689 information. Increasing this arbitrarily chosen value means more
5690 aggressive optimization, increasing the compile time. This parameter
5691 should be removed when the delay slot code is rewritten to maintain the
5694 @item max-gcse-memory
5695 The approximate maximum amount of memory that will be allocated in
5696 order to perform the global common subexpression elimination
5697 optimization. If more memory than specified is required, the
5698 optimization will not be done.
5700 @item max-gcse-passes
5701 The maximum number of passes of GCSE to run. The default is 1.
5703 @item max-pending-list-length
5704 The maximum number of pending dependencies scheduling will allow
5705 before flushing the current state and starting over. Large functions
5706 with few branches or calls can create excessively large lists which
5707 needlessly consume memory and resources.
5709 @item max-inline-insns-single
5710 Several parameters control the tree inliner used in gcc.
5711 This number sets the maximum number of instructions (counted in GCC's
5712 internal representation) in a single function that the tree inliner
5713 will consider for inlining. This only affects functions declared
5714 inline and methods implemented in a class declaration (C++).
5715 The default value is 450.
5717 @item max-inline-insns-auto
5718 When you use @option{-finline-functions} (included in @option{-O3}),
5719 a lot of functions that would otherwise not be considered for inlining
5720 by the compiler will be investigated. To those functions, a different
5721 (more restrictive) limit compared to functions declared inline can
5723 The default value is 90.
5725 @item large-function-insns
5726 The limit specifying really large functions. For functions larger than this
5727 limit after inlining inlining is constrained by
5728 @option{--param large-function-growth}. This parameter is useful primarily
5729 to avoid extreme compilation time caused by non-linear algorithms used by the
5731 This parameter is ignored when @option{-funit-at-a-time} is not used.
5732 The default value is 2700.
5734 @item large-function-growth
5735 Specifies maximal growth of large function caused by inlining in percents.
5736 This parameter is ignored when @option{-funit-at-a-time} is not used.
5737 The default value is 100 which limits large function growth to 2.0 times
5740 @item inline-unit-growth
5741 Specifies maximal overall growth of the compilation unit caused by inlining.
5742 This parameter is ignored when @option{-funit-at-a-time} is not used.
5743 The default value is 50 which limits unit growth to 1.5 times the original
5746 @item max-inline-insns-recursive
5747 @itemx max-inline-insns-recursive-auto
5748 Specifies maximum number of instructions out-of-line copy of self recursive inline
5749 function can grow into by performing recursive inlining.
5751 For functions declared inline @option{--param max-inline-insns-recursive} is
5752 taken into acount. For function not declared inline, recursive inlining
5753 happens only when @option{-finline-functions} (included in @option{-O3}) is
5754 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5755 default value is 450.
5757 @item max-inline-recursive-depth
5758 @itemx max-inline-recursive-depth-auto
5759 Specifies maximum recursion depth used by the recursive inlining.
5761 For functions declared inline @option{--param max-inline-recursive-depth} is
5762 taken into acount. For function not declared inline, recursive inlining
5763 happens only when @option{-finline-functions} (included in @option{-O3}) is
5764 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5765 default value is 450.
5767 @item min-inline-recursive-probability
5768 Recursive inlining is profitable only for function having deep recursion
5769 in average and can hurt for function having little recursion depth by
5770 increasing the prologue size or complexity of function body to other
5773 When profile feedback is available (see @option{-fprofile-generate}) the actual
5774 recursion depth can be guessed from probability that function will recurse via
5775 given call expression. This parameter limits inlining only to call expression
5776 whose probability exceeds given threshold (in percents). The default value is
5779 @item inline-call-cost
5780 Specify cost of call instruction relative to simple arithmetics operations
5781 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
5782 functions and at the same time increases size of leaf function that is believed to
5783 reduce function size by being inlined. In effect it increases amount of
5784 inlining for code having large abstraction penalty (many functions that just
5785 pass the arguments to other functions) and decrease inlining for code with low
5786 abstraction penalty. The default value is 16.
5788 @item max-unrolled-insns
5789 The maximum number of instructions that a loop should have if that loop
5790 is unrolled, and if the loop is unrolled, it determines how many times
5791 the loop code is unrolled.
5793 @item max-average-unrolled-insns
5794 The maximum number of instructions biased by probabilities of their execution
5795 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5796 it determines how many times the loop code is unrolled.
5798 @item max-unroll-times
5799 The maximum number of unrollings of a single loop.
5801 @item max-peeled-insns
5802 The maximum number of instructions that a loop should have if that loop
5803 is peeled, and if the loop is peeled, it determines how many times
5804 the loop code is peeled.
5806 @item max-peel-times
5807 The maximum number of peelings of a single loop.
5809 @item max-completely-peeled-insns
5810 The maximum number of insns of a completely peeled loop.
5812 @item max-completely-peel-times
5813 The maximum number of iterations of a loop to be suitable for complete peeling.
5815 @item max-unswitch-insns
5816 The maximum number of insns of an unswitched loop.
5818 @item max-unswitch-level
5819 The maximum number of branches unswitched in a single loop.
5822 The minimum cost of an expensive expression in the loop invariant motion.
5824 @item iv-consider-all-candidates-bound
5825 Bound on number of candidates for induction variables below that
5826 all candidates are considered for each use in induction variable
5827 optimizations. Only the most relevant candidates are considered
5828 if there are more candidates, to avoid quadratic time complexity.
5830 @item iv-max-considered-uses
5831 The induction variable optimizations give up on loops that contain more
5832 induction variable uses.
5834 @item iv-always-prune-cand-set-bound
5835 If number of candidates in the set is smaller than this value,
5836 we always try to remove unnecessary ivs from the set during its
5837 optimization when a new iv is added to the set.
5839 @item scev-max-expr-size
5840 Bound on size of expressions used in the scalar evolutions analyzer.
5841 Large expressions slow the analyzer.
5843 @item max-iterations-to-track
5845 The maximum number of iterations of a loop the brute force algorithm
5846 for analysis of # of iterations of the loop tries to evaluate.
5848 @item hot-bb-count-fraction
5849 Select fraction of the maximal count of repetitions of basic block in program
5850 given basic block needs to have to be considered hot.
5852 @item hot-bb-frequency-fraction
5853 Select fraction of the maximal frequency of executions of basic block in
5854 function given basic block needs to have to be considered hot
5856 @item tracer-dynamic-coverage
5857 @itemx tracer-dynamic-coverage-feedback
5859 This value is used to limit superblock formation once the given percentage of
5860 executed instructions is covered. This limits unnecessary code size
5863 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5864 feedback is available. The real profiles (as opposed to statically estimated
5865 ones) are much less balanced allowing the threshold to be larger value.
5867 @item tracer-max-code-growth
5868 Stop tail duplication once code growth has reached given percentage. This is
5869 rather hokey argument, as most of the duplicates will be eliminated later in
5870 cross jumping, so it may be set to much higher values than is the desired code
5873 @item tracer-min-branch-ratio
5875 Stop reverse growth when the reverse probability of best edge is less than this
5876 threshold (in percent).
5878 @item tracer-min-branch-ratio
5879 @itemx tracer-min-branch-ratio-feedback
5881 Stop forward growth if the best edge do have probability lower than this
5884 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
5885 compilation for profile feedback and one for compilation without. The value
5886 for compilation with profile feedback needs to be more conservative (higher) in
5887 order to make tracer effective.
5889 @item max-cse-path-length
5891 Maximum number of basic blocks on path that cse considers. The default is 10.
5893 @item global-var-threshold
5895 Counts the number of function calls (@var{n}) and the number of
5896 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
5897 single artificial variable will be created to represent all the
5898 call-clobbered variables at function call sites. This artificial
5899 variable will then be made to alias every call-clobbered variable.
5900 (done as @code{int * size_t} on the host machine; beware overflow).
5902 @item max-aliased-vops
5904 Maximum number of virtual operands allowed to represent aliases
5905 before triggering the alias grouping heuristic. Alias grouping
5906 reduces compile times and memory consumption needed for aliasing at
5907 the expense of precision loss in alias information.
5909 @item ggc-min-expand
5911 GCC uses a garbage collector to manage its own memory allocation. This
5912 parameter specifies the minimum percentage by which the garbage
5913 collector's heap should be allowed to expand between collections.
5914 Tuning this may improve compilation speed; it has no effect on code
5917 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
5918 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
5919 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
5920 GCC is not able to calculate RAM on a particular platform, the lower
5921 bound of 30% is used. Setting this parameter and
5922 @option{ggc-min-heapsize} to zero causes a full collection to occur at
5923 every opportunity. This is extremely slow, but can be useful for
5926 @item ggc-min-heapsize
5928 Minimum size of the garbage collector's heap before it begins bothering
5929 to collect garbage. The first collection occurs after the heap expands
5930 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
5931 tuning this may improve compilation speed, and has no effect on code
5934 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
5935 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
5936 with a lower bound of 4096 (four megabytes) and an upper bound of
5937 131072 (128 megabytes). If GCC is not able to calculate RAM on a
5938 particular platform, the lower bound is used. Setting this parameter
5939 very large effectively disables garbage collection. Setting this
5940 parameter and @option{ggc-min-expand} to zero causes a full collection
5941 to occur at every opportunity.
5943 @item max-reload-search-insns
5944 The maximum number of instruction reload should look backward for equivalent
5945 register. Increasing values mean more aggressive optimization, making the
5946 compile time increase with probably slightly better performance. The default
5949 @item max-cselib-memory-location
5950 The maximum number of memory locations cselib should take into acount.
5951 Increasing values mean more aggressive optimization, making the compile time
5952 increase with probably slightly better performance. The default value is 500.
5954 @item reorder-blocks-duplicate
5955 @itemx reorder-blocks-duplicate-feedback
5957 Used by basic block reordering pass to decide whether to use unconditional
5958 branch or duplicate the code on its destination. Code is duplicated when its
5959 estimated size is smaller than this value multiplied by the estimated size of
5960 unconditional jump in the hot spots of the program.
5962 The @option{reorder-block-duplicate-feedback} is used only when profile
5963 feedback is available and may be set to higher values than
5964 @option{reorder-block-duplicate} since information about the hot spots is more
5967 @item max-sched-region-blocks
5968 The maximum number of blocks in a region to be considered for
5969 interblock scheduling. The default value is 10.
5971 @item max-sched-region-insns
5972 The maximum number of insns in a region to be considered for
5973 interblock scheduling. The default value is 100.
5975 @item max-last-value-rtl
5977 The maximum size measured as number of RTLs that can be recorded in an expression
5978 in combiner for a pseudo register as last known value of that register. The default
5981 @item integer-share-limit
5982 Small integer constants can use a shared data structure, reducing the
5983 compiler's memory usage and increasing its speed. This sets the maximum
5984 value of a shared integer constant's. The default value is 256.
5986 @item min-virtual-mappings
5987 Specifies the minimum number of virtual mappings in the incremental
5988 SSA updater that should be registered to trigger the virtual mappings
5989 heuristic defined by virtual-mappings-ratio. The default value is
5992 @item virtual-mappings-ratio
5993 If the number of virtual mappings is virtual-mappings-ratio bigger
5994 than the number of virtual symbols to be updated, then the incremental
5995 SSA updater switches to a full update for those symbols. The default
6001 @node Preprocessor Options
6002 @section Options Controlling the Preprocessor
6003 @cindex preprocessor options
6004 @cindex options, preprocessor
6006 These options control the C preprocessor, which is run on each C source
6007 file before actual compilation.
6009 If you use the @option{-E} option, nothing is done except preprocessing.
6010 Some of these options make sense only together with @option{-E} because
6011 they cause the preprocessor output to be unsuitable for actual
6016 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6017 and pass @var{option} directly through to the preprocessor. If
6018 @var{option} contains commas, it is split into multiple options at the
6019 commas. However, many options are modified, translated or interpreted
6020 by the compiler driver before being passed to the preprocessor, and
6021 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6022 interface is undocumented and subject to change, so whenever possible
6023 you should avoid using @option{-Wp} and let the driver handle the
6026 @item -Xpreprocessor @var{option}
6027 @opindex preprocessor
6028 Pass @var{option} as an option to the preprocessor. You can use this to
6029 supply system-specific preprocessor options which GCC does not know how to
6032 If you want to pass an option that takes an argument, you must use
6033 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6036 @include cppopts.texi
6038 @node Assembler Options
6039 @section Passing Options to the Assembler
6041 @c prevent bad page break with this line
6042 You can pass options to the assembler.
6045 @item -Wa,@var{option}
6047 Pass @var{option} as an option to the assembler. If @var{option}
6048 contains commas, it is split into multiple options at the commas.
6050 @item -Xassembler @var{option}
6052 Pass @var{option} as an option to the assembler. You can use this to
6053 supply system-specific assembler options which GCC does not know how to
6056 If you want to pass an option that takes an argument, you must use
6057 @option{-Xassembler} twice, once for the option and once for the argument.
6062 @section Options for Linking
6063 @cindex link options
6064 @cindex options, linking
6066 These options come into play when the compiler links object files into
6067 an executable output file. They are meaningless if the compiler is
6068 not doing a link step.
6072 @item @var{object-file-name}
6073 A file name that does not end in a special recognized suffix is
6074 considered to name an object file or library. (Object files are
6075 distinguished from libraries by the linker according to the file
6076 contents.) If linking is done, these object files are used as input
6085 If any of these options is used, then the linker is not run, and
6086 object file names should not be used as arguments. @xref{Overall
6090 @item -l@var{library}
6091 @itemx -l @var{library}
6093 Search the library named @var{library} when linking. (The second
6094 alternative with the library as a separate argument is only for
6095 POSIX compliance and is not recommended.)
6097 It makes a difference where in the command you write this option; the
6098 linker searches and processes libraries and object files in the order they
6099 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6100 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6101 to functions in @samp{z}, those functions may not be loaded.
6103 The linker searches a standard list of directories for the library,
6104 which is actually a file named @file{lib@var{library}.a}. The linker
6105 then uses this file as if it had been specified precisely by name.
6107 The directories searched include several standard system directories
6108 plus any that you specify with @option{-L}.
6110 Normally the files found this way are library files---archive files
6111 whose members are object files. The linker handles an archive file by
6112 scanning through it for members which define symbols that have so far
6113 been referenced but not defined. But if the file that is found is an
6114 ordinary object file, it is linked in the usual fashion. The only
6115 difference between using an @option{-l} option and specifying a file name
6116 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6117 and searches several directories.
6121 You need this special case of the @option{-l} option in order to
6122 link an Objective-C or Objective-C++ program.
6125 @opindex nostartfiles
6126 Do not use the standard system startup files when linking.
6127 The standard system libraries are used normally, unless @option{-nostdlib}
6128 or @option{-nodefaultlibs} is used.
6130 @item -nodefaultlibs
6131 @opindex nodefaultlibs
6132 Do not use the standard system libraries when linking.
6133 Only the libraries you specify will be passed to the linker.
6134 The standard startup files are used normally, unless @option{-nostartfiles}
6135 is used. The compiler may generate calls to @code{memcmp},
6136 @code{memset}, @code{memcpy} and @code{memmove}.
6137 These entries are usually resolved by entries in
6138 libc. These entry points should be supplied through some other
6139 mechanism when this option is specified.
6143 Do not use the standard system startup files or libraries when linking.
6144 No startup files and only the libraries you specify will be passed to
6145 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6146 @code{memcpy} and @code{memmove}.
6147 These entries are usually resolved by entries in
6148 libc. These entry points should be supplied through some other
6149 mechanism when this option is specified.
6151 @cindex @option{-lgcc}, use with @option{-nostdlib}
6152 @cindex @option{-nostdlib} and unresolved references
6153 @cindex unresolved references and @option{-nostdlib}
6154 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6155 @cindex @option{-nodefaultlibs} and unresolved references
6156 @cindex unresolved references and @option{-nodefaultlibs}
6157 One of the standard libraries bypassed by @option{-nostdlib} and
6158 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6159 that GCC uses to overcome shortcomings of particular machines, or special
6160 needs for some languages.
6161 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6162 Collection (GCC) Internals},
6163 for more discussion of @file{libgcc.a}.)
6164 In most cases, you need @file{libgcc.a} even when you want to avoid
6165 other standard libraries. In other words, when you specify @option{-nostdlib}
6166 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6167 This ensures that you have no unresolved references to internal GCC
6168 library subroutines. (For example, @samp{__main}, used to ensure C++
6169 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6170 GNU Compiler Collection (GCC) Internals}.)
6174 Produce a position independent executable on targets which support it.
6175 For predictable results, you must also specify the same set of options
6176 that were used to generate code (@option{-fpie}, @option{-fPIE},
6177 or model suboptions) when you specify this option.
6181 Remove all symbol table and relocation information from the executable.
6185 On systems that support dynamic linking, this prevents linking with the shared
6186 libraries. On other systems, this option has no effect.
6190 Produce a shared object which can then be linked with other objects to
6191 form an executable. Not all systems support this option. For predictable
6192 results, you must also specify the same set of options that were used to
6193 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6194 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6195 needs to build supplementary stub code for constructors to work. On
6196 multi-libbed systems, @samp{gcc -shared} must select the correct support
6197 libraries to link against. Failing to supply the correct flags may lead
6198 to subtle defects. Supplying them in cases where they are not necessary
6201 @item -shared-libgcc
6202 @itemx -static-libgcc
6203 @opindex shared-libgcc
6204 @opindex static-libgcc
6205 On systems that provide @file{libgcc} as a shared library, these options
6206 force the use of either the shared or static version respectively.
6207 If no shared version of @file{libgcc} was built when the compiler was
6208 configured, these options have no effect.
6210 There are several situations in which an application should use the
6211 shared @file{libgcc} instead of the static version. The most common
6212 of these is when the application wishes to throw and catch exceptions
6213 across different shared libraries. In that case, each of the libraries
6214 as well as the application itself should use the shared @file{libgcc}.
6216 Therefore, the G++ and GCJ drivers automatically add
6217 @option{-shared-libgcc} whenever you build a shared library or a main
6218 executable, because C++ and Java programs typically use exceptions, so
6219 this is the right thing to do.
6221 If, instead, you use the GCC driver to create shared libraries, you may
6222 find that they will not always be linked with the shared @file{libgcc}.
6223 If GCC finds, at its configuration time, that you have a non-GNU linker
6224 or a GNU linker that does not support option @option{--eh-frame-hdr},
6225 it will link the shared version of @file{libgcc} into shared libraries
6226 by default. Otherwise, it will take advantage of the linker and optimize
6227 away the linking with the shared version of @file{libgcc}, linking with
6228 the static version of libgcc by default. This allows exceptions to
6229 propagate through such shared libraries, without incurring relocation
6230 costs at library load time.
6232 However, if a library or main executable is supposed to throw or catch
6233 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6234 for the languages used in the program, or using the option
6235 @option{-shared-libgcc}, such that it is linked with the shared
6240 Bind references to global symbols when building a shared object. Warn
6241 about any unresolved references (unless overridden by the link editor
6242 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6245 @item -Xlinker @var{option}
6247 Pass @var{option} as an option to the linker. You can use this to
6248 supply system-specific linker options which GCC does not know how to
6251 If you want to pass an option that takes an argument, you must use
6252 @option{-Xlinker} twice, once for the option and once for the argument.
6253 For example, to pass @option{-assert definitions}, you must write
6254 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6255 @option{-Xlinker "-assert definitions"}, because this passes the entire
6256 string as a single argument, which is not what the linker expects.
6258 @item -Wl,@var{option}
6260 Pass @var{option} as an option to the linker. If @var{option} contains
6261 commas, it is split into multiple options at the commas.
6263 @item -u @var{symbol}
6265 Pretend the symbol @var{symbol} is undefined, to force linking of
6266 library modules to define it. You can use @option{-u} multiple times with
6267 different symbols to force loading of additional library modules.
6270 @node Directory Options
6271 @section Options for Directory Search
6272 @cindex directory options
6273 @cindex options, directory search
6276 These options specify directories to search for header files, for
6277 libraries and for parts of the compiler:
6282 Add the directory @var{dir} to the head of the list of directories to be
6283 searched for header files. This can be used to override a system header
6284 file, substituting your own version, since these directories are
6285 searched before the system header file directories. However, you should
6286 not use this option to add directories that contain vendor-supplied
6287 system header files (use @option{-isystem} for that). If you use more than
6288 one @option{-I} option, the directories are scanned in left-to-right
6289 order; the standard system directories come after.
6291 If a standard system include directory, or a directory specified with
6292 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6293 option will be ignored. The directory will still be searched but as a
6294 system directory at its normal position in the system include chain.
6295 This is to ensure that GCC's procedure to fix buggy system headers and
6296 the ordering for the include_next directive are not inadvertently changed.
6297 If you really need to change the search order for system directories,
6298 use the @option{-nostdinc} and/or @option{-isystem} options.
6300 @item -iquote@var{dir}
6302 Add the directory @var{dir} to the head of the list of directories to
6303 be searched for header files only for the case of @samp{#include
6304 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6305 otherwise just like @option{-I}.
6309 Add directory @var{dir} to the list of directories to be searched
6312 @item -B@var{prefix}
6314 This option specifies where to find the executables, libraries,
6315 include files, and data files of the compiler itself.
6317 The compiler driver program runs one or more of the subprograms
6318 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6319 @var{prefix} as a prefix for each program it tries to run, both with and
6320 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6322 For each subprogram to be run, the compiler driver first tries the
6323 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6324 was not specified, the driver tries two standard prefixes, which are
6325 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6326 those results in a file name that is found, the unmodified program
6327 name is searched for using the directories specified in your
6328 @env{PATH} environment variable.
6330 The compiler will check to see if the path provided by the @option{-B}
6331 refers to a directory, and if necessary it will add a directory
6332 separator character at the end of the path.
6334 @option{-B} prefixes that effectively specify directory names also apply
6335 to libraries in the linker, because the compiler translates these
6336 options into @option{-L} options for the linker. They also apply to
6337 includes files in the preprocessor, because the compiler translates these
6338 options into @option{-isystem} options for the preprocessor. In this case,
6339 the compiler appends @samp{include} to the prefix.
6341 The run-time support file @file{libgcc.a} can also be searched for using
6342 the @option{-B} prefix, if needed. If it is not found there, the two
6343 standard prefixes above are tried, and that is all. The file is left
6344 out of the link if it is not found by those means.
6346 Another way to specify a prefix much like the @option{-B} prefix is to use
6347 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6350 As a special kludge, if the path provided by @option{-B} is
6351 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6352 9, then it will be replaced by @file{[dir/]include}. This is to help
6353 with boot-strapping the compiler.
6355 @item -specs=@var{file}
6357 Process @var{file} after the compiler reads in the standard @file{specs}
6358 file, in order to override the defaults that the @file{gcc} driver
6359 program uses when determining what switches to pass to @file{cc1},
6360 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6361 @option{-specs=@var{file}} can be specified on the command line, and they
6362 are processed in order, from left to right.
6364 @item --sysroot=@var{dir}
6366 Use @var{dir} as the logical root directory for headers and libraries.
6367 For example, if the compiler would normally search for headers in
6368 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
6369 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
6371 If you use both this option and the @option{-isysroot} option, then
6372 the @option{--sysroot} option will apply to libraries, but the
6373 @option{-isysroot} option will apply to header files.
6375 The GNU linker (beginning with version 2.16) has the necessary support
6376 for this option. If your linker does not support this option, the
6377 header file aspect of @option{--sysroot} will still work, but the
6378 library aspect will not.
6382 This option has been deprecated. Please use @option{-iquote} instead for
6383 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6384 Any directories you specify with @option{-I} options before the @option{-I-}
6385 option are searched only for the case of @samp{#include "@var{file}"};
6386 they are not searched for @samp{#include <@var{file}>}.
6388 If additional directories are specified with @option{-I} options after
6389 the @option{-I-}, these directories are searched for all @samp{#include}
6390 directives. (Ordinarily @emph{all} @option{-I} directories are used
6393 In addition, the @option{-I-} option inhibits the use of the current
6394 directory (where the current input file came from) as the first search
6395 directory for @samp{#include "@var{file}"}. There is no way to
6396 override this effect of @option{-I-}. With @option{-I.} you can specify
6397 searching the directory which was current when the compiler was
6398 invoked. That is not exactly the same as what the preprocessor does
6399 by default, but it is often satisfactory.
6401 @option{-I-} does not inhibit the use of the standard system directories
6402 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6409 @section Specifying subprocesses and the switches to pass to them
6412 @command{gcc} is a driver program. It performs its job by invoking a
6413 sequence of other programs to do the work of compiling, assembling and
6414 linking. GCC interprets its command-line parameters and uses these to
6415 deduce which programs it should invoke, and which command-line options
6416 it ought to place on their command lines. This behavior is controlled
6417 by @dfn{spec strings}. In most cases there is one spec string for each
6418 program that GCC can invoke, but a few programs have multiple spec
6419 strings to control their behavior. The spec strings built into GCC can
6420 be overridden by using the @option{-specs=} command-line switch to specify
6423 @dfn{Spec files} are plaintext files that are used to construct spec
6424 strings. They consist of a sequence of directives separated by blank
6425 lines. The type of directive is determined by the first non-whitespace
6426 character on the line and it can be one of the following:
6429 @item %@var{command}
6430 Issues a @var{command} to the spec file processor. The commands that can
6434 @item %include <@var{file}>
6436 Search for @var{file} and insert its text at the current point in the
6439 @item %include_noerr <@var{file}>
6440 @cindex %include_noerr
6441 Just like @samp{%include}, but do not generate an error message if the include
6442 file cannot be found.
6444 @item %rename @var{old_name} @var{new_name}
6446 Rename the spec string @var{old_name} to @var{new_name}.
6450 @item *[@var{spec_name}]:
6451 This tells the compiler to create, override or delete the named spec
6452 string. All lines after this directive up to the next directive or
6453 blank line are considered to be the text for the spec string. If this
6454 results in an empty string then the spec will be deleted. (Or, if the
6455 spec did not exist, then nothing will happened.) Otherwise, if the spec
6456 does not currently exist a new spec will be created. If the spec does
6457 exist then its contents will be overridden by the text of this
6458 directive, unless the first character of that text is the @samp{+}
6459 character, in which case the text will be appended to the spec.
6461 @item [@var{suffix}]:
6462 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6463 and up to the next directive or blank line are considered to make up the
6464 spec string for the indicated suffix. When the compiler encounters an
6465 input file with the named suffix, it will processes the spec string in
6466 order to work out how to compile that file. For example:
6473 This says that any input file whose name ends in @samp{.ZZ} should be
6474 passed to the program @samp{z-compile}, which should be invoked with the
6475 command-line switch @option{-input} and with the result of performing the
6476 @samp{%i} substitution. (See below.)
6478 As an alternative to providing a spec string, the text that follows a
6479 suffix directive can be one of the following:
6482 @item @@@var{language}
6483 This says that the suffix is an alias for a known @var{language}. This is
6484 similar to using the @option{-x} command-line switch to GCC to specify a
6485 language explicitly. For example:
6492 Says that .ZZ files are, in fact, C++ source files.
6495 This causes an error messages saying:
6498 @var{name} compiler not installed on this system.
6502 GCC already has an extensive list of suffixes built into it.
6503 This directive will add an entry to the end of the list of suffixes, but
6504 since the list is searched from the end backwards, it is effectively
6505 possible to override earlier entries using this technique.
6509 GCC has the following spec strings built into it. Spec files can
6510 override these strings or create their own. Note that individual
6511 targets can also add their own spec strings to this list.
6514 asm Options to pass to the assembler
6515 asm_final Options to pass to the assembler post-processor
6516 cpp Options to pass to the C preprocessor
6517 cc1 Options to pass to the C compiler
6518 cc1plus Options to pass to the C++ compiler
6519 endfile Object files to include at the end of the link
6520 link Options to pass to the linker
6521 lib Libraries to include on the command line to the linker
6522 libgcc Decides which GCC support library to pass to the linker
6523 linker Sets the name of the linker
6524 predefines Defines to be passed to the C preprocessor
6525 signed_char Defines to pass to CPP to say whether @code{char} is signed
6527 startfile Object files to include at the start of the link
6530 Here is a small example of a spec file:
6536 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6539 This example renames the spec called @samp{lib} to @samp{old_lib} and
6540 then overrides the previous definition of @samp{lib} with a new one.
6541 The new definition adds in some extra command-line options before
6542 including the text of the old definition.
6544 @dfn{Spec strings} are a list of command-line options to be passed to their
6545 corresponding program. In addition, the spec strings can contain
6546 @samp{%}-prefixed sequences to substitute variable text or to
6547 conditionally insert text into the command line. Using these constructs
6548 it is possible to generate quite complex command lines.
6550 Here is a table of all defined @samp{%}-sequences for spec
6551 strings. Note that spaces are not generated automatically around the
6552 results of expanding these sequences. Therefore you can concatenate them
6553 together or combine them with constant text in a single argument.
6557 Substitute one @samp{%} into the program name or argument.
6560 Substitute the name of the input file being processed.
6563 Substitute the basename of the input file being processed.
6564 This is the substring up to (and not including) the last period
6565 and not including the directory.
6568 This is the same as @samp{%b}, but include the file suffix (text after
6572 Marks the argument containing or following the @samp{%d} as a
6573 temporary file name, so that that file will be deleted if GCC exits
6574 successfully. Unlike @samp{%g}, this contributes no text to the
6577 @item %g@var{suffix}
6578 Substitute a file name that has suffix @var{suffix} and is chosen
6579 once per compilation, and mark the argument in the same way as
6580 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6581 name is now chosen in a way that is hard to predict even when previously
6582 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6583 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6584 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6585 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6586 was simply substituted with a file name chosen once per compilation,
6587 without regard to any appended suffix (which was therefore treated
6588 just like ordinary text), making such attacks more likely to succeed.
6590 @item %u@var{suffix}
6591 Like @samp{%g}, but generates a new temporary file name even if
6592 @samp{%u@var{suffix}} was already seen.
6594 @item %U@var{suffix}
6595 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6596 new one if there is no such last file name. In the absence of any
6597 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6598 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6599 would involve the generation of two distinct file names, one
6600 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6601 simply substituted with a file name chosen for the previous @samp{%u},
6602 without regard to any appended suffix.
6604 @item %j@var{suffix}
6605 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6606 writable, and if save-temps is off; otherwise, substitute the name
6607 of a temporary file, just like @samp{%u}. This temporary file is not
6608 meant for communication between processes, but rather as a junk
6611 @item %|@var{suffix}
6612 @itemx %m@var{suffix}
6613 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6614 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6615 all. These are the two most common ways to instruct a program that it
6616 should read from standard input or write to standard output. If you
6617 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6618 construct: see for example @file{f/lang-specs.h}.
6620 @item %.@var{SUFFIX}
6621 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6622 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6623 terminated by the next space or %.
6626 Marks the argument containing or following the @samp{%w} as the
6627 designated output file of this compilation. This puts the argument
6628 into the sequence of arguments that @samp{%o} will substitute later.
6631 Substitutes the names of all the output files, with spaces
6632 automatically placed around them. You should write spaces
6633 around the @samp{%o} as well or the results are undefined.
6634 @samp{%o} is for use in the specs for running the linker.
6635 Input files whose names have no recognized suffix are not compiled
6636 at all, but they are included among the output files, so they will
6640 Substitutes the suffix for object files. Note that this is
6641 handled specially when it immediately follows @samp{%g, %u, or %U},
6642 because of the need for those to form complete file names. The
6643 handling is such that @samp{%O} is treated exactly as if it had already
6644 been substituted, except that @samp{%g, %u, and %U} do not currently
6645 support additional @var{suffix} characters following @samp{%O} as they would
6646 following, for example, @samp{.o}.
6649 Substitutes the standard macro predefinitions for the
6650 current target machine. Use this when running @code{cpp}.
6653 Like @samp{%p}, but puts @samp{__} before and after the name of each
6654 predefined macro, except for macros that start with @samp{__} or with
6655 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6659 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6660 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
6661 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6665 Current argument is the name of a library or startup file of some sort.
6666 Search for that file in a standard list of directories and substitute
6667 the full name found.
6670 Print @var{str} as an error message. @var{str} is terminated by a newline.
6671 Use this when inconsistent options are detected.
6674 Substitute the contents of spec string @var{name} at this point.
6677 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6679 @item %x@{@var{option}@}
6680 Accumulate an option for @samp{%X}.
6683 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6687 Output the accumulated assembler options specified by @option{-Wa}.
6690 Output the accumulated preprocessor options specified by @option{-Wp}.
6693 Process the @code{asm} spec. This is used to compute the
6694 switches to be passed to the assembler.
6697 Process the @code{asm_final} spec. This is a spec string for
6698 passing switches to an assembler post-processor, if such a program is
6702 Process the @code{link} spec. This is the spec for computing the
6703 command line passed to the linker. Typically it will make use of the
6704 @samp{%L %G %S %D and %E} sequences.
6707 Dump out a @option{-L} option for each directory that GCC believes might
6708 contain startup files. If the target supports multilibs then the
6709 current multilib directory will be prepended to each of these paths.
6712 Process the @code{lib} spec. This is a spec string for deciding which
6713 libraries should be included on the command line to the linker.
6716 Process the @code{libgcc} spec. This is a spec string for deciding
6717 which GCC support library should be included on the command line to the linker.
6720 Process the @code{startfile} spec. This is a spec for deciding which
6721 object files should be the first ones passed to the linker. Typically
6722 this might be a file named @file{crt0.o}.
6725 Process the @code{endfile} spec. This is a spec string that specifies
6726 the last object files that will be passed to the linker.
6729 Process the @code{cpp} spec. This is used to construct the arguments
6730 to be passed to the C preprocessor.
6733 Process the @code{cc1} spec. This is used to construct the options to be
6734 passed to the actual C compiler (@samp{cc1}).
6737 Process the @code{cc1plus} spec. This is used to construct the options to be
6738 passed to the actual C++ compiler (@samp{cc1plus}).
6741 Substitute the variable part of a matched option. See below.
6742 Note that each comma in the substituted string is replaced by
6746 Remove all occurrences of @code{-S} from the command line. Note---this
6747 command is position dependent. @samp{%} commands in the spec string
6748 before this one will see @code{-S}, @samp{%} commands in the spec string
6749 after this one will not.
6751 @item %:@var{function}(@var{args})
6752 Call the named function @var{function}, passing it @var{args}.
6753 @var{args} is first processed as a nested spec string, then split
6754 into an argument vector in the usual fashion. The function returns
6755 a string which is processed as if it had appeared literally as part
6756 of the current spec.
6758 The following built-in spec functions are provided:
6761 @item @code{if-exists}
6762 The @code{if-exists} spec function takes one argument, an absolute
6763 pathname to a file. If the file exists, @code{if-exists} returns the
6764 pathname. Here is a small example of its usage:
6768 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6771 @item @code{if-exists-else}
6772 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6773 spec function, except that it takes two arguments. The first argument is
6774 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6775 returns the pathname. If it does not exist, it returns the second argument.
6776 This way, @code{if-exists-else} can be used to select one file or another,
6777 based on the existence of the first. Here is a small example of its usage:
6781 crt0%O%s %:if-exists(crti%O%s) \
6782 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6785 @item @code{replace-outfile}
6786 The @code{replace-outfile} spec function takes two arguments. It looks for the
6787 first argument in the outfiles array and replaces it with the second argument. Here
6788 is a small example of its usage:
6791 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
6797 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6798 If that switch was not specified, this substitutes nothing. Note that
6799 the leading dash is omitted when specifying this option, and it is
6800 automatically inserted if the substitution is performed. Thus the spec
6801 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6802 and would output the command line option @option{-foo}.
6804 @item %W@{@code{S}@}
6805 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6808 @item %@{@code{S}*@}
6809 Substitutes all the switches specified to GCC whose names start
6810 with @code{-S}, but which also take an argument. This is used for
6811 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6812 GCC considers @option{-o foo} as being
6813 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6814 text, including the space. Thus two arguments would be generated.
6816 @item %@{@code{S}*&@code{T}*@}
6817 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6818 (the order of @code{S} and @code{T} in the spec is not significant).
6819 There can be any number of ampersand-separated variables; for each the
6820 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6822 @item %@{@code{S}:@code{X}@}
6823 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6825 @item %@{!@code{S}:@code{X}@}
6826 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6828 @item %@{@code{S}*:@code{X}@}
6829 Substitutes @code{X} if one or more switches whose names start with
6830 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6831 once, no matter how many such switches appeared. However, if @code{%*}
6832 appears somewhere in @code{X}, then @code{X} will be substituted once
6833 for each matching switch, with the @code{%*} replaced by the part of
6834 that switch that matched the @code{*}.
6836 @item %@{.@code{S}:@code{X}@}
6837 Substitutes @code{X}, if processing a file with suffix @code{S}.
6839 @item %@{!.@code{S}:@code{X}@}
6840 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6842 @item %@{@code{S}|@code{P}:@code{X}@}
6843 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6844 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
6845 although they have a stronger binding than the @samp{|}. If @code{%*}
6846 appears in @code{X}, all of the alternatives must be starred, and only
6847 the first matching alternative is substituted.
6849 For example, a spec string like this:
6852 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
6855 will output the following command-line options from the following input
6856 command-line options:
6861 -d fred.c -foo -baz -boggle
6862 -d jim.d -bar -baz -boggle
6865 @item %@{S:X; T:Y; :D@}
6867 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
6868 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
6869 be as many clauses as you need. This may be combined with @code{.},
6870 @code{!}, @code{|}, and @code{*} as needed.
6875 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
6876 construct may contain other nested @samp{%} constructs or spaces, or
6877 even newlines. They are processed as usual, as described above.
6878 Trailing white space in @code{X} is ignored. White space may also
6879 appear anywhere on the left side of the colon in these constructs,
6880 except between @code{.} or @code{*} and the corresponding word.
6882 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
6883 handled specifically in these constructs. If another value of
6884 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
6885 @option{-W} switch is found later in the command line, the earlier
6886 switch value is ignored, except with @{@code{S}*@} where @code{S} is
6887 just one letter, which passes all matching options.
6889 The character @samp{|} at the beginning of the predicate text is used to
6890 indicate that a command should be piped to the following command, but
6891 only if @option{-pipe} is specified.
6893 It is built into GCC which switches take arguments and which do not.
6894 (You might think it would be useful to generalize this to allow each
6895 compiler's spec to say which switches take arguments. But this cannot
6896 be done in a consistent fashion. GCC cannot even decide which input
6897 files have been specified without knowing which switches take arguments,
6898 and it must know which input files to compile in order to tell which
6901 GCC also knows implicitly that arguments starting in @option{-l} are to be
6902 treated as compiler output files, and passed to the linker in their
6903 proper position among the other output files.
6905 @c man begin OPTIONS
6907 @node Target Options
6908 @section Specifying Target Machine and Compiler Version
6909 @cindex target options
6910 @cindex cross compiling
6911 @cindex specifying machine version
6912 @cindex specifying compiler version and target machine
6913 @cindex compiler version, specifying
6914 @cindex target machine, specifying
6916 The usual way to run GCC is to run the executable called @file{gcc}, or
6917 @file{<machine>-gcc} when cross-compiling, or
6918 @file{<machine>-gcc-<version>} to run a version other than the one that
6919 was installed last. Sometimes this is inconvenient, so GCC provides
6920 options that will switch to another cross-compiler or version.
6923 @item -b @var{machine}
6925 The argument @var{machine} specifies the target machine for compilation.
6927 The value to use for @var{machine} is the same as was specified as the
6928 machine type when configuring GCC as a cross-compiler. For
6929 example, if a cross-compiler was configured with @samp{configure
6930 i386v}, meaning to compile for an 80386 running System V, then you
6931 would specify @option{-b i386v} to run that cross compiler.
6933 @item -V @var{version}
6935 The argument @var{version} specifies which version of GCC to run.
6936 This is useful when multiple versions are installed. For example,
6937 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
6940 The @option{-V} and @option{-b} options work by running the
6941 @file{<machine>-gcc-<version>} executable, so there's no real reason to
6942 use them if you can just run that directly.
6944 @node Submodel Options
6945 @section Hardware Models and Configurations
6946 @cindex submodel options
6947 @cindex specifying hardware config
6948 @cindex hardware models and configurations, specifying
6949 @cindex machine dependent options
6951 Earlier we discussed the standard option @option{-b} which chooses among
6952 different installed compilers for completely different target
6953 machines, such as VAX vs.@: 68000 vs.@: 80386.
6955 In addition, each of these target machine types can have its own
6956 special options, starting with @samp{-m}, to choose among various
6957 hardware models or configurations---for example, 68010 vs 68020,
6958 floating coprocessor or none. A single installed version of the
6959 compiler can compile for any model or configuration, according to the
6962 Some configurations of the compiler also support additional special
6963 options, usually for compatibility with other compilers on the same
6966 @c This list is ordered alphanumerically by subsection name.
6967 @c It should be the same order and spelling as these options are listed
6968 @c in Machine Dependent Options
6974 * Blackfin Options::
6977 * DEC Alpha Options::
6978 * DEC Alpha/VMS Options::
6982 * i386 and x86-64 Options::
6994 * RS/6000 and PowerPC Options::
6995 * S/390 and zSeries Options::
6998 * System V Options::
6999 * TMS320C3x/C4x Options::
7003 * Xstormy16 Options::
7009 @subsection ARC Options
7012 These options are defined for ARC implementations:
7017 Compile code for little endian mode. This is the default.
7021 Compile code for big endian mode.
7024 @opindex mmangle-cpu
7025 Prepend the name of the cpu to all public symbol names.
7026 In multiple-processor systems, there are many ARC variants with different
7027 instruction and register set characteristics. This flag prevents code
7028 compiled for one cpu to be linked with code compiled for another.
7029 No facility exists for handling variants that are ``almost identical''.
7030 This is an all or nothing option.
7032 @item -mcpu=@var{cpu}
7034 Compile code for ARC variant @var{cpu}.
7035 Which variants are supported depend on the configuration.
7036 All variants support @option{-mcpu=base}, this is the default.
7038 @item -mtext=@var{text-section}
7039 @itemx -mdata=@var{data-section}
7040 @itemx -mrodata=@var{readonly-data-section}
7044 Put functions, data, and readonly data in @var{text-section},
7045 @var{data-section}, and @var{readonly-data-section} respectively
7046 by default. This can be overridden with the @code{section} attribute.
7047 @xref{Variable Attributes}.
7052 @subsection ARM Options
7055 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7059 @item -mabi=@var{name}
7061 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7062 @samp{atpcs}, @samp{aapcs} and @samp{iwmmxt}.
7065 @opindex mapcs-frame
7066 Generate a stack frame that is compliant with the ARM Procedure Call
7067 Standard for all functions, even if this is not strictly necessary for
7068 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7069 with this option will cause the stack frames not to be generated for
7070 leaf functions. The default is @option{-mno-apcs-frame}.
7074 This is a synonym for @option{-mapcs-frame}.
7077 @c not currently implemented
7078 @item -mapcs-stack-check
7079 @opindex mapcs-stack-check
7080 Generate code to check the amount of stack space available upon entry to
7081 every function (that actually uses some stack space). If there is
7082 insufficient space available then either the function
7083 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7084 called, depending upon the amount of stack space required. The run time
7085 system is required to provide these functions. The default is
7086 @option{-mno-apcs-stack-check}, since this produces smaller code.
7088 @c not currently implemented
7090 @opindex mapcs-float
7091 Pass floating point arguments using the float point registers. This is
7092 one of the variants of the APCS@. This option is recommended if the
7093 target hardware has a floating point unit or if a lot of floating point
7094 arithmetic is going to be performed by the code. The default is
7095 @option{-mno-apcs-float}, since integer only code is slightly increased in
7096 size if @option{-mapcs-float} is used.
7098 @c not currently implemented
7099 @item -mapcs-reentrant
7100 @opindex mapcs-reentrant
7101 Generate reentrant, position independent code. The default is
7102 @option{-mno-apcs-reentrant}.
7105 @item -mthumb-interwork
7106 @opindex mthumb-interwork
7107 Generate code which supports calling between the ARM and Thumb
7108 instruction sets. Without this option the two instruction sets cannot
7109 be reliably used inside one program. The default is
7110 @option{-mno-thumb-interwork}, since slightly larger code is generated
7111 when @option{-mthumb-interwork} is specified.
7113 @item -mno-sched-prolog
7114 @opindex mno-sched-prolog
7115 Prevent the reordering of instructions in the function prolog, or the
7116 merging of those instruction with the instructions in the function's
7117 body. This means that all functions will start with a recognizable set
7118 of instructions (or in fact one of a choice from a small set of
7119 different function prologues), and this information can be used to
7120 locate the start if functions inside an executable piece of code. The
7121 default is @option{-msched-prolog}.
7124 @opindex mhard-float
7125 Generate output containing floating point instructions. This is the
7129 @opindex msoft-float
7130 Generate output containing library calls for floating point.
7131 @strong{Warning:} the requisite libraries are not available for all ARM
7132 targets. Normally the facilities of the machine's usual C compiler are
7133 used, but this cannot be done directly in cross-compilation. You must make
7134 your own arrangements to provide suitable library functions for
7137 @option{-msoft-float} changes the calling convention in the output file;
7138 therefore, it is only useful if you compile @emph{all} of a program with
7139 this option. In particular, you need to compile @file{libgcc.a}, the
7140 library that comes with GCC, with @option{-msoft-float} in order for
7143 @item -mfloat-abi=@var{name}
7145 Specifies which ABI to use for floating point values. Permissible values
7146 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7148 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7149 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7150 of floating point instructions, but still uses the soft-float calling
7153 @item -mlittle-endian
7154 @opindex mlittle-endian
7155 Generate code for a processor running in little-endian mode. This is
7156 the default for all standard configurations.
7159 @opindex mbig-endian
7160 Generate code for a processor running in big-endian mode; the default is
7161 to compile code for a little-endian processor.
7163 @item -mwords-little-endian
7164 @opindex mwords-little-endian
7165 This option only applies when generating code for big-endian processors.
7166 Generate code for a little-endian word order but a big-endian byte
7167 order. That is, a byte order of the form @samp{32107654}. Note: this
7168 option should only be used if you require compatibility with code for
7169 big-endian ARM processors generated by versions of the compiler prior to
7172 @item -mcpu=@var{name}
7174 This specifies the name of the target ARM processor. GCC uses this name
7175 to determine what kind of instructions it can emit when generating
7176 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7177 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7178 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7179 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7180 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7181 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7182 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7183 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7184 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7185 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7186 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7187 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7188 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7189 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
7192 @itemx -mtune=@var{name}
7194 This option is very similar to the @option{-mcpu=} option, except that
7195 instead of specifying the actual target processor type, and hence
7196 restricting which instructions can be used, it specifies that GCC should
7197 tune the performance of the code as if the target were of the type
7198 specified in this option, but still choosing the instructions that it
7199 will generate based on the cpu specified by a @option{-mcpu=} option.
7200 For some ARM implementations better performance can be obtained by using
7203 @item -march=@var{name}
7205 This specifies the name of the target ARM architecture. GCC uses this
7206 name to determine what kind of instructions it can emit when generating
7207 assembly code. This option can be used in conjunction with or instead
7208 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7209 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7210 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7211 @samp{iwmmxt}, @samp{ep9312}.
7213 @item -mfpu=@var{name}
7214 @itemx -mfpe=@var{number}
7215 @itemx -mfp=@var{number}
7219 This specifies what floating point hardware (or hardware emulation) is
7220 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7221 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7222 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7223 with older versions of GCC@.
7225 If @option{-msoft-float} is specified this specifies the format of
7226 floating point values.
7228 @item -mstructure-size-boundary=@var{n}
7229 @opindex mstructure-size-boundary
7230 The size of all structures and unions will be rounded up to a multiple
7231 of the number of bits set by this option. Permissible values are 8, 32
7232 and 64. The default value varies for different toolchains. For the COFF
7233 targeted toolchain the default value is 8. A value of 64 is only allowed
7234 if the underlying ABI supports it.
7236 Specifying the larger number can produce faster, more efficient code, but
7237 can also increase the size of the program. Different values are potentially
7238 incompatible. Code compiled with one value cannot necessarily expect to
7239 work with code or libraries compiled with another value, if they exchange
7240 information using structures or unions.
7242 @item -mabort-on-noreturn
7243 @opindex mabort-on-noreturn
7244 Generate a call to the function @code{abort} at the end of a
7245 @code{noreturn} function. It will be executed if the function tries to
7249 @itemx -mno-long-calls
7250 @opindex mlong-calls
7251 @opindex mno-long-calls
7252 Tells the compiler to perform function calls by first loading the
7253 address of the function into a register and then performing a subroutine
7254 call on this register. This switch is needed if the target function
7255 will lie outside of the 64 megabyte addressing range of the offset based
7256 version of subroutine call instruction.
7258 Even if this switch is enabled, not all function calls will be turned
7259 into long calls. The heuristic is that static functions, functions
7260 which have the @samp{short-call} attribute, functions that are inside
7261 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7262 definitions have already been compiled within the current compilation
7263 unit, will not be turned into long calls. The exception to this rule is
7264 that weak function definitions, functions with the @samp{long-call}
7265 attribute or the @samp{section} attribute, and functions that are within
7266 the scope of a @samp{#pragma long_calls} directive, will always be
7267 turned into long calls.
7269 This feature is not enabled by default. Specifying
7270 @option{-mno-long-calls} will restore the default behavior, as will
7271 placing the function calls within the scope of a @samp{#pragma
7272 long_calls_off} directive. Note these switches have no effect on how
7273 the compiler generates code to handle function calls via function
7276 @item -mnop-fun-dllimport
7277 @opindex mnop-fun-dllimport
7278 Disable support for the @code{dllimport} attribute.
7280 @item -msingle-pic-base
7281 @opindex msingle-pic-base
7282 Treat the register used for PIC addressing as read-only, rather than
7283 loading it in the prologue for each function. The run-time system is
7284 responsible for initializing this register with an appropriate value
7285 before execution begins.
7287 @item -mpic-register=@var{reg}
7288 @opindex mpic-register
7289 Specify the register to be used for PIC addressing. The default is R10
7290 unless stack-checking is enabled, when R9 is used.
7292 @item -mcirrus-fix-invalid-insns
7293 @opindex mcirrus-fix-invalid-insns
7294 @opindex mno-cirrus-fix-invalid-insns
7295 Insert NOPs into the instruction stream to in order to work around
7296 problems with invalid Maverick instruction combinations. This option
7297 is only valid if the @option{-mcpu=ep9312} option has been used to
7298 enable generation of instructions for the Cirrus Maverick floating
7299 point co-processor. This option is not enabled by default, since the
7300 problem is only present in older Maverick implementations. The default
7301 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7304 @item -mpoke-function-name
7305 @opindex mpoke-function-name
7306 Write the name of each function into the text section, directly
7307 preceding the function prologue. The generated code is similar to this:
7311 .ascii "arm_poke_function_name", 0
7314 .word 0xff000000 + (t1 - t0)
7315 arm_poke_function_name
7317 stmfd sp!, @{fp, ip, lr, pc@}
7321 When performing a stack backtrace, code can inspect the value of
7322 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7323 location @code{pc - 12} and the top 8 bits are set, then we know that
7324 there is a function name embedded immediately preceding this location
7325 and has length @code{((pc[-3]) & 0xff000000)}.
7329 Generate code for the 16-bit Thumb instruction set. The default is to
7330 use the 32-bit ARM instruction set.
7333 @opindex mtpcs-frame
7334 Generate a stack frame that is compliant with the Thumb Procedure Call
7335 Standard for all non-leaf functions. (A leaf function is one that does
7336 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7338 @item -mtpcs-leaf-frame
7339 @opindex mtpcs-leaf-frame
7340 Generate a stack frame that is compliant with the Thumb Procedure Call
7341 Standard for all leaf functions. (A leaf function is one that does
7342 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7344 @item -mcallee-super-interworking
7345 @opindex mcallee-super-interworking
7346 Gives all externally visible functions in the file being compiled an ARM
7347 instruction set header which switches to Thumb mode before executing the
7348 rest of the function. This allows these functions to be called from
7349 non-interworking code.
7351 @item -mcaller-super-interworking
7352 @opindex mcaller-super-interworking
7353 Allows calls via function pointers (including virtual functions) to
7354 execute correctly regardless of whether the target code has been
7355 compiled for interworking or not. There is a small overhead in the cost
7356 of executing a function pointer if this option is enabled.
7361 @subsection AVR Options
7364 These options are defined for AVR implementations:
7367 @item -mmcu=@var{mcu}
7369 Specify ATMEL AVR instruction set or MCU type.
7371 Instruction set avr1 is for the minimal AVR core, not supported by the C
7372 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7373 attiny11, attiny12, attiny15, attiny28).
7375 Instruction set avr2 (default) is for the classic AVR core with up to
7376 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7377 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7378 at90c8534, at90s8535).
7380 Instruction set avr3 is for the classic AVR core with up to 128K program
7381 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7383 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7384 memory space (MCU types: atmega8, atmega83, atmega85).
7386 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7387 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7388 atmega64, atmega128, at43usb355, at94k).
7392 Output instruction sizes to the asm file.
7394 @item -minit-stack=@var{N}
7395 @opindex minit-stack
7396 Specify the initial stack address, which may be a symbol or numeric value,
7397 @samp{__stack} is the default.
7399 @item -mno-interrupts
7400 @opindex mno-interrupts
7401 Generated code is not compatible with hardware interrupts.
7402 Code size will be smaller.
7404 @item -mcall-prologues
7405 @opindex mcall-prologues
7406 Functions prologues/epilogues expanded as call to appropriate
7407 subroutines. Code size will be smaller.
7409 @item -mno-tablejump
7410 @opindex mno-tablejump
7411 Do not generate tablejump insns which sometimes increase code size.
7414 @opindex mtiny-stack
7415 Change only the low 8 bits of the stack pointer.
7419 Assume int to be 8 bit integer. This affects the sizes of all types: A
7420 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7421 and long long will be 4 bytes. Please note that this option does not
7422 comply to the C standards, but it will provide you with smaller code
7426 @node Blackfin Options
7427 @subsection Blackfin Options
7428 @cindex Blackfin Options
7431 @item -momit-leaf-frame-pointer
7432 @opindex momit-leaf-frame-pointer
7433 Don't keep the frame pointer in a register for leaf functions. This
7434 avoids the instructions to save, set up and restore frame pointers and
7435 makes an extra register available in leaf functions. The option
7436 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7437 which might make debugging harder.
7439 @item -mspecld-anomaly
7440 @opindex mspecld-anomaly
7441 When enabled, the compiler will ensure that the generated code does not
7442 contain speculative loads after jump instructions. This option is enabled
7445 @item -mno-specld-anomaly
7446 @opindex mno-specld-anomaly
7447 Don't generate extra code to prevent speculative loads from occurring.
7449 @item -mcsync-anomaly
7450 @opindex mcsync-anomaly
7451 When enabled, the compiler will ensure that the generated code does not
7452 contain CSYNC or SSYNC instructions too soon after conditional branches.
7453 This option is enabled by default.
7455 @item -mno-csync-anomaly
7456 @opindex mno-csync-anomaly
7457 Don't generate extra code to prevent CSYNC or SSYNC instructions from
7458 occurring too soon after a conditional branch.
7462 When enabled, the compiler is free to take advantage of the knowledge that
7463 the entire program fits into the low 64k of memory.
7466 @opindex mno-low-64k
7467 Assume that the program is arbitrarily large. This is the default.
7469 @item -mid-shared-library
7470 @opindex mid-shared-library
7471 Generate code that supports shared libraries via the library ID method.
7472 This allows for execute in place and shared libraries in an environment
7473 without virtual memory management. This option implies @option{-fPIC}.
7475 @item -mno-id-shared-library
7476 @opindex mno-id-shared-library
7477 Generate code that doesn't assume ID based shared libraries are being used.
7478 This is the default.
7480 @item -mshared-library-id=n
7481 @opindex mshared-library-id
7482 Specified the identification number of the ID based shared library being
7483 compiled. Specifying a value of 0 will generate more compact code, specifying
7484 other values will force the allocation of that number to the current
7485 library but is no more space or time efficient than omitting this option.
7488 @itemx -mno-long-calls
7489 @opindex mlong-calls
7490 @opindex mno-long-calls
7491 Tells the compiler to perform function calls by first loading the
7492 address of the function into a register and then performing a subroutine
7493 call on this register. This switch is needed if the target function
7494 will lie outside of the 24 bit addressing range of the offset based
7495 version of subroutine call instruction.
7497 This feature is not enabled by default. Specifying
7498 @option{-mno-long-calls} will restore the default behavior. Note these
7499 switches have no effect on how the compiler generates code to handle
7500 function calls via function pointers.
7504 @subsection CRIS Options
7505 @cindex CRIS Options
7507 These options are defined specifically for the CRIS ports.
7510 @item -march=@var{architecture-type}
7511 @itemx -mcpu=@var{architecture-type}
7514 Generate code for the specified architecture. The choices for
7515 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7516 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7517 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7520 @item -mtune=@var{architecture-type}
7522 Tune to @var{architecture-type} everything applicable about the generated
7523 code, except for the ABI and the set of available instructions. The
7524 choices for @var{architecture-type} are the same as for
7525 @option{-march=@var{architecture-type}}.
7527 @item -mmax-stack-frame=@var{n}
7528 @opindex mmax-stack-frame
7529 Warn when the stack frame of a function exceeds @var{n} bytes.
7531 @item -melinux-stacksize=@var{n}
7532 @opindex melinux-stacksize
7533 Only available with the @samp{cris-axis-aout} target. Arranges for
7534 indications in the program to the kernel loader that the stack of the
7535 program should be set to @var{n} bytes.
7541 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7542 @option{-march=v3} and @option{-march=v8} respectively.
7544 @item -mmul-bug-workaround
7545 @itemx -mno-mul-bug-workaround
7546 @opindex mmul-bug-workaround
7547 @opindex mno-mul-bug-workaround
7548 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7549 models where it applies. This option is active by default.
7553 Enable CRIS-specific verbose debug-related information in the assembly
7554 code. This option also has the effect to turn off the @samp{#NO_APP}
7555 formatted-code indicator to the assembler at the beginning of the
7560 Do not use condition-code results from previous instruction; always emit
7561 compare and test instructions before use of condition codes.
7563 @item -mno-side-effects
7564 @opindex mno-side-effects
7565 Do not emit instructions with side-effects in addressing modes other than
7569 @itemx -mno-stack-align
7571 @itemx -mno-data-align
7572 @itemx -mconst-align
7573 @itemx -mno-const-align
7574 @opindex mstack-align
7575 @opindex mno-stack-align
7576 @opindex mdata-align
7577 @opindex mno-data-align
7578 @opindex mconst-align
7579 @opindex mno-const-align
7580 These options (no-options) arranges (eliminate arrangements) for the
7581 stack-frame, individual data and constants to be aligned for the maximum
7582 single data access size for the chosen CPU model. The default is to
7583 arrange for 32-bit alignment. ABI details such as structure layout are
7584 not affected by these options.
7592 Similar to the stack- data- and const-align options above, these options
7593 arrange for stack-frame, writable data and constants to all be 32-bit,
7594 16-bit or 8-bit aligned. The default is 32-bit alignment.
7596 @item -mno-prologue-epilogue
7597 @itemx -mprologue-epilogue
7598 @opindex mno-prologue-epilogue
7599 @opindex mprologue-epilogue
7600 With @option{-mno-prologue-epilogue}, the normal function prologue and
7601 epilogue that sets up the stack-frame are omitted and no return
7602 instructions or return sequences are generated in the code. Use this
7603 option only together with visual inspection of the compiled code: no
7604 warnings or errors are generated when call-saved registers must be saved,
7605 or storage for local variable needs to be allocated.
7611 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7612 instruction sequences that load addresses for functions from the PLT part
7613 of the GOT rather than (traditional on other architectures) calls to the
7614 PLT@. The default is @option{-mgotplt}.
7618 Legacy no-op option only recognized with the cris-axis-aout target.
7622 Legacy no-op option only recognized with the cris-axis-elf and
7623 cris-axis-linux-gnu targets.
7627 Only recognized with the cris-axis-aout target, where it selects a
7628 GNU/linux-like multilib, include files and instruction set for
7633 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7637 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7638 to link with input-output functions from a simulator library. Code,
7639 initialized data and zero-initialized data are allocated consecutively.
7643 Like @option{-sim}, but pass linker options to locate initialized data at
7644 0x40000000 and zero-initialized data at 0x80000000.
7647 @node Darwin Options
7648 @subsection Darwin Options
7649 @cindex Darwin options
7651 These options are defined for all architectures running the Darwin operating
7654 FSF GCC on Darwin does not create ``fat'' object files; it will create
7655 an object file for the single architecture that it was built to
7656 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7657 @option{-arch} options are used; it does so by running the compiler or
7658 linker multiple times and joining the results together with
7661 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7662 @samp{i686}) is determined by the flags that specify the ISA
7663 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7664 @option{-force_cpusubtype_ALL} option can be used to override this.
7666 The Darwin tools vary in their behavior when presented with an ISA
7667 mismatch. The assembler, @file{as}, will only permit instructions to
7668 be used that are valid for the subtype of the file it is generating,
7669 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7670 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
7671 and print an error if asked to create a shared library with a less
7672 restrictive subtype than its input files (for instance, trying to put
7673 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
7674 for executables, @file{ld}, will quietly give the executable the most
7675 restrictive subtype of any of its input files.
7680 Add the framework directory @var{dir} to the head of the list of
7681 directories to be searched for header files. These directories are
7682 interleaved with those specified by @option{-I} options and are
7683 scanned in a left-to-right order.
7685 A framework directory is a directory with frameworks in it. A
7686 framework is a directory with a @samp{"Headers"} and/or
7687 @samp{"PrivateHeaders"} directory contained directly in it that ends
7688 in @samp{".framework"}. The name of a framework is the name of this
7689 directory excluding the @samp{".framework"}. Headers associated with
7690 the framework are found in one of those two directories, with
7691 @samp{"Headers"} being searched first. A subframework is a framework
7692 directory that is in a framework's @samp{"Frameworks"} directory.
7693 Includes of subframework headers can only appear in a header of a
7694 framework that contains the subframework, or in a sibling subframework
7695 header. Two subframeworks are siblings if they occur in the same
7696 framework. A subframework should not have the same name as a
7697 framework, a warning will be issued if this is violated. Currently a
7698 subframework cannot have subframeworks, in the future, the mechanism
7699 may be extended to support this. The standard frameworks can be found
7700 in @samp{"/System/Library/Frameworks"} and
7701 @samp{"/Library/Frameworks"}. An example include looks like
7702 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
7703 the name of the framework and header.h is found in the
7704 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
7708 Emit debugging information for symbols that are used. For STABS
7709 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
7710 This is by default ON@.
7714 Emit debugging information for all symbols and types.
7716 @item -mmacosx-version-min=@var{version}
7717 The earliest version of MacOS X that this executable will run on
7718 is @var{version}. Typical values of @var{version} include @code{10.1},
7719 @code{10.2}, and @code{10.3.9}.
7721 The default for this option is to make choices that seem to be most
7724 @item -mone-byte-bool
7725 @opindex -mone-byte-bool
7726 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7727 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7728 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7729 option has no effect on x86.
7731 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
7732 to generate code that is not binary compatible with code generated
7733 without that switch. Using this switch may require recompiling all
7734 other modules in a program, including system libraries. Use this
7735 switch to conform to a non-default data model.
7737 @item -mfix-and-continue
7738 @itemx -ffix-and-continue
7739 @itemx -findirect-data
7740 @opindex mfix-and-continue
7741 @opindex ffix-and-continue
7742 @opindex findirect-data
7743 Generate code suitable for fast turn around development. Needed to
7744 enable gdb to dynamically load @code{.o} files into already running
7745 programs. @option{-findirect-data} and @option{-ffix-and-continue}
7746 are provided for backwards compatibility.
7750 Loads all members of static archive libraries.
7751 See man ld(1) for more information.
7753 @item -arch_errors_fatal
7754 @opindex arch_errors_fatal
7755 Cause the errors having to do with files that have the wrong architecture
7759 @opindex bind_at_load
7760 Causes the output file to be marked such that the dynamic linker will
7761 bind all undefined references when the file is loaded or launched.
7765 Produce a Mach-o bundle format file.
7766 See man ld(1) for more information.
7768 @item -bundle_loader @var{executable}
7769 @opindex bundle_loader
7770 This option specifies the @var{executable} that will be loading the build
7771 output file being linked. See man ld(1) for more information.
7774 @opindex -dynamiclib
7775 When passed this option, GCC will produce a dynamic library instead of
7776 an executable when linking, using the Darwin @file{libtool} command.
7778 @item -force_cpusubtype_ALL
7779 @opindex -force_cpusubtype_ALL
7780 This causes GCC's output file to have the @var{ALL} subtype, instead of
7781 one controlled by the @option{-mcpu} or @option{-march} option.
7783 @item -allowable_client @var{client_name}
7785 @itemx -compatibility_version
7786 @itemx -current_version
7788 @itemx -dependency-file
7790 @itemx -dylinker_install_name
7792 @itemx -exported_symbols_list
7794 @itemx -flat_namespace
7795 @itemx -force_flat_namespace
7796 @itemx -headerpad_max_install_names
7799 @itemx -install_name
7800 @itemx -keep_private_externs
7801 @itemx -multi_module
7802 @itemx -multiply_defined
7803 @itemx -multiply_defined_unused
7805 @itemx -no_dead_strip_inits_and_terms
7806 @itemx -nofixprebinding
7809 @itemx -noseglinkedit
7810 @itemx -pagezero_size
7812 @itemx -prebind_all_twolevel_modules
7813 @itemx -private_bundle
7814 @itemx -read_only_relocs
7816 @itemx -sectobjectsymbols
7820 @itemx -sectobjectsymbols
7823 @itemx -segs_read_only_addr
7824 @itemx -segs_read_write_addr
7825 @itemx -seg_addr_table
7826 @itemx -seg_addr_table_filename
7829 @itemx -segs_read_only_addr
7830 @itemx -segs_read_write_addr
7831 @itemx -single_module
7834 @itemx -sub_umbrella
7835 @itemx -twolevel_namespace
7838 @itemx -unexported_symbols_list
7839 @itemx -weak_reference_mismatches
7842 @opindex allowable_client
7843 @opindex client_name
7844 @opindex compatibility_version
7845 @opindex current_version
7847 @opindex dependency-file
7849 @opindex dylinker_install_name
7851 @opindex exported_symbols_list
7853 @opindex flat_namespace
7854 @opindex force_flat_namespace
7855 @opindex headerpad_max_install_names
7858 @opindex install_name
7859 @opindex keep_private_externs
7860 @opindex multi_module
7861 @opindex multiply_defined
7862 @opindex multiply_defined_unused
7864 @opindex no_dead_strip_inits_and_terms
7865 @opindex nofixprebinding
7866 @opindex nomultidefs
7868 @opindex noseglinkedit
7869 @opindex pagezero_size
7871 @opindex prebind_all_twolevel_modules
7872 @opindex private_bundle
7873 @opindex read_only_relocs
7875 @opindex sectobjectsymbols
7879 @opindex sectobjectsymbols
7882 @opindex segs_read_only_addr
7883 @opindex segs_read_write_addr
7884 @opindex seg_addr_table
7885 @opindex seg_addr_table_filename
7886 @opindex seglinkedit
7888 @opindex segs_read_only_addr
7889 @opindex segs_read_write_addr
7890 @opindex single_module
7892 @opindex sub_library
7893 @opindex sub_umbrella
7894 @opindex twolevel_namespace
7897 @opindex unexported_symbols_list
7898 @opindex weak_reference_mismatches
7899 @opindex whatsloaded
7901 These options are passed to the Darwin linker. The Darwin linker man page
7902 describes them in detail.
7905 @node DEC Alpha Options
7906 @subsection DEC Alpha Options
7908 These @samp{-m} options are defined for the DEC Alpha implementations:
7911 @item -mno-soft-float
7913 @opindex mno-soft-float
7914 @opindex msoft-float
7915 Use (do not use) the hardware floating-point instructions for
7916 floating-point operations. When @option{-msoft-float} is specified,
7917 functions in @file{libgcc.a} will be used to perform floating-point
7918 operations. Unless they are replaced by routines that emulate the
7919 floating-point operations, or compiled in such a way as to call such
7920 emulations routines, these routines will issue floating-point
7921 operations. If you are compiling for an Alpha without floating-point
7922 operations, you must ensure that the library is built so as not to call
7925 Note that Alpha implementations without floating-point operations are
7926 required to have floating-point registers.
7931 @opindex mno-fp-regs
7932 Generate code that uses (does not use) the floating-point register set.
7933 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
7934 register set is not used, floating point operands are passed in integer
7935 registers as if they were integers and floating-point results are passed
7936 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
7937 so any function with a floating-point argument or return value called by code
7938 compiled with @option{-mno-fp-regs} must also be compiled with that
7941 A typical use of this option is building a kernel that does not use,
7942 and hence need not save and restore, any floating-point registers.
7946 The Alpha architecture implements floating-point hardware optimized for
7947 maximum performance. It is mostly compliant with the IEEE floating
7948 point standard. However, for full compliance, software assistance is
7949 required. This option generates code fully IEEE compliant code
7950 @emph{except} that the @var{inexact-flag} is not maintained (see below).
7951 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
7952 defined during compilation. The resulting code is less efficient but is
7953 able to correctly support denormalized numbers and exceptional IEEE
7954 values such as not-a-number and plus/minus infinity. Other Alpha
7955 compilers call this option @option{-ieee_with_no_inexact}.
7957 @item -mieee-with-inexact
7958 @opindex mieee-with-inexact
7959 This is like @option{-mieee} except the generated code also maintains
7960 the IEEE @var{inexact-flag}. Turning on this option causes the
7961 generated code to implement fully-compliant IEEE math. In addition to
7962 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
7963 macro. On some Alpha implementations the resulting code may execute
7964 significantly slower than the code generated by default. Since there is
7965 very little code that depends on the @var{inexact-flag}, you should
7966 normally not specify this option. Other Alpha compilers call this
7967 option @option{-ieee_with_inexact}.
7969 @item -mfp-trap-mode=@var{trap-mode}
7970 @opindex mfp-trap-mode
7971 This option controls what floating-point related traps are enabled.
7972 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
7973 The trap mode can be set to one of four values:
7977 This is the default (normal) setting. The only traps that are enabled
7978 are the ones that cannot be disabled in software (e.g., division by zero
7982 In addition to the traps enabled by @samp{n}, underflow traps are enabled
7986 Like @samp{su}, but the instructions are marked to be safe for software
7987 completion (see Alpha architecture manual for details).
7990 Like @samp{su}, but inexact traps are enabled as well.
7993 @item -mfp-rounding-mode=@var{rounding-mode}
7994 @opindex mfp-rounding-mode
7995 Selects the IEEE rounding mode. Other Alpha compilers call this option
7996 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8001 Normal IEEE rounding mode. Floating point numbers are rounded towards
8002 the nearest machine number or towards the even machine number in case
8006 Round towards minus infinity.
8009 Chopped rounding mode. Floating point numbers are rounded towards zero.
8012 Dynamic rounding mode. A field in the floating point control register
8013 (@var{fpcr}, see Alpha architecture reference manual) controls the
8014 rounding mode in effect. The C library initializes this register for
8015 rounding towards plus infinity. Thus, unless your program modifies the
8016 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8019 @item -mtrap-precision=@var{trap-precision}
8020 @opindex mtrap-precision
8021 In the Alpha architecture, floating point traps are imprecise. This
8022 means without software assistance it is impossible to recover from a
8023 floating trap and program execution normally needs to be terminated.
8024 GCC can generate code that can assist operating system trap handlers
8025 in determining the exact location that caused a floating point trap.
8026 Depending on the requirements of an application, different levels of
8027 precisions can be selected:
8031 Program precision. This option is the default and means a trap handler
8032 can only identify which program caused a floating point exception.
8035 Function precision. The trap handler can determine the function that
8036 caused a floating point exception.
8039 Instruction precision. The trap handler can determine the exact
8040 instruction that caused a floating point exception.
8043 Other Alpha compilers provide the equivalent options called
8044 @option{-scope_safe} and @option{-resumption_safe}.
8046 @item -mieee-conformant
8047 @opindex mieee-conformant
8048 This option marks the generated code as IEEE conformant. You must not
8049 use this option unless you also specify @option{-mtrap-precision=i} and either
8050 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8051 is to emit the line @samp{.eflag 48} in the function prologue of the
8052 generated assembly file. Under DEC Unix, this has the effect that
8053 IEEE-conformant math library routines will be linked in.
8055 @item -mbuild-constants
8056 @opindex mbuild-constants
8057 Normally GCC examines a 32- or 64-bit integer constant to
8058 see if it can construct it from smaller constants in two or three
8059 instructions. If it cannot, it will output the constant as a literal and
8060 generate code to load it from the data segment at runtime.
8062 Use this option to require GCC to construct @emph{all} integer constants
8063 using code, even if it takes more instructions (the maximum is six).
8065 You would typically use this option to build a shared library dynamic
8066 loader. Itself a shared library, it must relocate itself in memory
8067 before it can find the variables and constants in its own data segment.
8073 Select whether to generate code to be assembled by the vendor-supplied
8074 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8092 Indicate whether GCC should generate code to use the optional BWX,
8093 CIX, FIX and MAX instruction sets. The default is to use the instruction
8094 sets supported by the CPU type specified via @option{-mcpu=} option or that
8095 of the CPU on which GCC was built if none was specified.
8100 @opindex mfloat-ieee
8101 Generate code that uses (does not use) VAX F and G floating point
8102 arithmetic instead of IEEE single and double precision.
8104 @item -mexplicit-relocs
8105 @itemx -mno-explicit-relocs
8106 @opindex mexplicit-relocs
8107 @opindex mno-explicit-relocs
8108 Older Alpha assemblers provided no way to generate symbol relocations
8109 except via assembler macros. Use of these macros does not allow
8110 optimal instruction scheduling. GNU binutils as of version 2.12
8111 supports a new syntax that allows the compiler to explicitly mark
8112 which relocations should apply to which instructions. This option
8113 is mostly useful for debugging, as GCC detects the capabilities of
8114 the assembler when it is built and sets the default accordingly.
8118 @opindex msmall-data
8119 @opindex mlarge-data
8120 When @option{-mexplicit-relocs} is in effect, static data is
8121 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8122 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8123 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8124 16-bit relocations off of the @code{$gp} register. This limits the
8125 size of the small data area to 64KB, but allows the variables to be
8126 directly accessed via a single instruction.
8128 The default is @option{-mlarge-data}. With this option the data area
8129 is limited to just below 2GB@. Programs that require more than 2GB of
8130 data must use @code{malloc} or @code{mmap} to allocate the data in the
8131 heap instead of in the program's data segment.
8133 When generating code for shared libraries, @option{-fpic} implies
8134 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8138 @opindex msmall-text
8139 @opindex mlarge-text
8140 When @option{-msmall-text} is used, the compiler assumes that the
8141 code of the entire program (or shared library) fits in 4MB, and is
8142 thus reachable with a branch instruction. When @option{-msmall-data}
8143 is used, the compiler can assume that all local symbols share the
8144 same @code{$gp} value, and thus reduce the number of instructions
8145 required for a function call from 4 to 1.
8147 The default is @option{-mlarge-text}.
8149 @item -mcpu=@var{cpu_type}
8151 Set the instruction set and instruction scheduling parameters for
8152 machine type @var{cpu_type}. You can specify either the @samp{EV}
8153 style name or the corresponding chip number. GCC supports scheduling
8154 parameters for the EV4, EV5 and EV6 family of processors and will
8155 choose the default values for the instruction set from the processor
8156 you specify. If you do not specify a processor type, GCC will default
8157 to the processor on which the compiler was built.
8159 Supported values for @var{cpu_type} are
8165 Schedules as an EV4 and has no instruction set extensions.
8169 Schedules as an EV5 and has no instruction set extensions.
8173 Schedules as an EV5 and supports the BWX extension.
8178 Schedules as an EV5 and supports the BWX and MAX extensions.
8182 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8186 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8189 @item -mtune=@var{cpu_type}
8191 Set only the instruction scheduling parameters for machine type
8192 @var{cpu_type}. The instruction set is not changed.
8194 @item -mmemory-latency=@var{time}
8195 @opindex mmemory-latency
8196 Sets the latency the scheduler should assume for typical memory
8197 references as seen by the application. This number is highly
8198 dependent on the memory access patterns used by the application
8199 and the size of the external cache on the machine.
8201 Valid options for @var{time} are
8205 A decimal number representing clock cycles.
8211 The compiler contains estimates of the number of clock cycles for
8212 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8213 (also called Dcache, Scache, and Bcache), as well as to main memory.
8214 Note that L3 is only valid for EV5.
8219 @node DEC Alpha/VMS Options
8220 @subsection DEC Alpha/VMS Options
8222 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8225 @item -mvms-return-codes
8226 @opindex mvms-return-codes
8227 Return VMS condition codes from main. The default is to return POSIX
8228 style condition (e.g.@ error) codes.
8232 @subsection FRV Options
8239 Only use the first 32 general purpose registers.
8244 Use all 64 general purpose registers.
8249 Use only the first 32 floating point registers.
8254 Use all 64 floating point registers
8257 @opindex mhard-float
8259 Use hardware instructions for floating point operations.
8262 @opindex msoft-float
8264 Use library routines for floating point operations.
8269 Dynamically allocate condition code registers.
8274 Do not try to dynamically allocate condition code registers, only
8275 use @code{icc0} and @code{fcc0}.
8280 Change ABI to use double word insns.
8285 Do not use double word instructions.
8290 Use floating point double instructions.
8295 Do not use floating point double instructions.
8300 Use media instructions.
8305 Do not use media instructions.
8310 Use multiply and add/subtract instructions.
8315 Do not use multiply and add/subtract instructions.
8320 Select the FDPIC ABI, that uses function descriptors to represent
8321 pointers to functions. Without any PIC/PIE-related options, it
8322 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8323 assumes GOT entries and small data are within a 12-bit range from the
8324 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8325 are computed with 32 bits.
8328 @opindex minline-plt
8330 Enable inlining of PLT entries in function calls to functions that are
8331 not known to bind locally. It has no effect without @option{-mfdpic}.
8332 It's enabled by default if optimizing for speed and compiling for
8333 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8334 optimization option such as @option{-O3} or above is present in the
8340 Assume a large TLS segment when generating thread-local code.
8345 Do not assume a large TLS segment when generating thread-local code.
8350 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8351 that is known to be in read-only sections. It's enabled by default,
8352 except for @option{-fpic} or @option{-fpie}: even though it may help
8353 make the global offset table smaller, it trades 1 instruction for 4.
8354 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8355 one of which may be shared by multiple symbols, and it avoids the need
8356 for a GOT entry for the referenced symbol, so it's more likely to be a
8357 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8359 @item -multilib-library-pic
8360 @opindex multilib-library-pic
8362 Link with the (library, not FD) pic libraries. It's implied by
8363 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8364 @option{-fpic} without @option{-mfdpic}. You should never have to use
8370 Follow the EABI requirement of always creating a frame pointer whenever
8371 a stack frame is allocated. This option is enabled by default and can
8372 be disabled with @option{-mno-linked-fp}.
8375 @opindex mlong-calls
8377 Use indirect addressing to call functions outside the current
8378 compilation unit. This allows the functions to be placed anywhere
8379 within the 32-bit address space.
8381 @item -malign-labels
8382 @opindex malign-labels
8384 Try to align labels to an 8-byte boundary by inserting nops into the
8385 previous packet. This option only has an effect when VLIW packing
8386 is enabled. It doesn't create new packets; it merely adds nops to
8390 @opindex mlibrary-pic
8392 Generate position-independent EABI code.
8397 Use only the first four media accumulator registers.
8402 Use all eight media accumulator registers.
8407 Pack VLIW instructions.
8412 Do not pack VLIW instructions.
8417 Do not mark ABI switches in e_flags.
8422 Enable the use of conditional-move instructions (default).
8424 This switch is mainly for debugging the compiler and will likely be removed
8425 in a future version.
8427 @item -mno-cond-move
8428 @opindex mno-cond-move
8430 Disable the use of conditional-move instructions.
8432 This switch is mainly for debugging the compiler and will likely be removed
8433 in a future version.
8438 Enable the use of conditional set instructions (default).
8440 This switch is mainly for debugging the compiler and will likely be removed
8441 in a future version.
8446 Disable the use of conditional set instructions.
8448 This switch is mainly for debugging the compiler and will likely be removed
8449 in a future version.
8454 Enable the use of conditional execution (default).
8456 This switch is mainly for debugging the compiler and will likely be removed
8457 in a future version.
8459 @item -mno-cond-exec
8460 @opindex mno-cond-exec
8462 Disable the use of conditional execution.
8464 This switch is mainly for debugging the compiler and will likely be removed
8465 in a future version.
8468 @opindex mvliw-branch
8470 Run a pass to pack branches into VLIW instructions (default).
8472 This switch is mainly for debugging the compiler and will likely be removed
8473 in a future version.
8475 @item -mno-vliw-branch
8476 @opindex mno-vliw-branch
8478 Do not run a pass to pack branches into VLIW instructions.
8480 This switch is mainly for debugging the compiler and will likely be removed
8481 in a future version.
8483 @item -mmulti-cond-exec
8484 @opindex mmulti-cond-exec
8486 Enable optimization of @code{&&} and @code{||} in conditional execution
8489 This switch is mainly for debugging the compiler and will likely be removed
8490 in a future version.
8492 @item -mno-multi-cond-exec
8493 @opindex mno-multi-cond-exec
8495 Disable optimization of @code{&&} and @code{||} in conditional execution.
8497 This switch is mainly for debugging the compiler and will likely be removed
8498 in a future version.
8500 @item -mnested-cond-exec
8501 @opindex mnested-cond-exec
8503 Enable nested conditional execution optimizations (default).
8505 This switch is mainly for debugging the compiler and will likely be removed
8506 in a future version.
8508 @item -mno-nested-cond-exec
8509 @opindex mno-nested-cond-exec
8511 Disable nested conditional execution optimizations.
8513 This switch is mainly for debugging the compiler and will likely be removed
8514 in a future version.
8516 @item -moptimize-membar
8517 @opindex moptimize-membar
8519 This switch removes redundant @code{membar} instructions from the
8520 compiler generated code. It is enabled by default.
8522 @item -mno-optimize-membar
8523 @opindex mno-optimize-membar
8525 This switch disables the automatic removal of redundant @code{membar}
8526 instructions from the generated code.
8528 @item -mtomcat-stats
8529 @opindex mtomcat-stats
8531 Cause gas to print out tomcat statistics.
8533 @item -mcpu=@var{cpu}
8536 Select the processor type for which to generate code. Possible values are
8537 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8538 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8542 @node H8/300 Options
8543 @subsection H8/300 Options
8545 These @samp{-m} options are defined for the H8/300 implementations:
8550 Shorten some address references at link time, when possible; uses the
8551 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8552 ld, Using ld}, for a fuller description.
8556 Generate code for the H8/300H@.
8560 Generate code for the H8S@.
8564 Generate code for the H8S and H8/300H in the normal mode. This switch
8565 must be used either with @option{-mh} or @option{-ms}.
8569 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8573 Make @code{int} data 32 bits by default.
8577 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8578 The default for the H8/300H and H8S is to align longs and floats on 4
8580 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8581 This option has no effect on the H8/300.
8585 @subsection HPPA Options
8586 @cindex HPPA Options
8588 These @samp{-m} options are defined for the HPPA family of computers:
8591 @item -march=@var{architecture-type}
8593 Generate code for the specified architecture. The choices for
8594 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8595 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8596 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8597 architecture option for your machine. Code compiled for lower numbered
8598 architectures will run on higher numbered architectures, but not the
8602 @itemx -mpa-risc-1-1
8603 @itemx -mpa-risc-2-0
8604 @opindex mpa-risc-1-0
8605 @opindex mpa-risc-1-1
8606 @opindex mpa-risc-2-0
8607 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8610 @opindex mbig-switch
8611 Generate code suitable for big switch tables. Use this option only if
8612 the assembler/linker complain about out of range branches within a switch
8615 @item -mjump-in-delay
8616 @opindex mjump-in-delay
8617 Fill delay slots of function calls with unconditional jump instructions
8618 by modifying the return pointer for the function call to be the target
8619 of the conditional jump.
8621 @item -mdisable-fpregs
8622 @opindex mdisable-fpregs
8623 Prevent floating point registers from being used in any manner. This is
8624 necessary for compiling kernels which perform lazy context switching of
8625 floating point registers. If you use this option and attempt to perform
8626 floating point operations, the compiler will abort.
8628 @item -mdisable-indexing
8629 @opindex mdisable-indexing
8630 Prevent the compiler from using indexing address modes. This avoids some
8631 rather obscure problems when compiling MIG generated code under MACH@.
8633 @item -mno-space-regs
8634 @opindex mno-space-regs
8635 Generate code that assumes the target has no space registers. This allows
8636 GCC to generate faster indirect calls and use unscaled index address modes.
8638 Such code is suitable for level 0 PA systems and kernels.
8640 @item -mfast-indirect-calls
8641 @opindex mfast-indirect-calls
8642 Generate code that assumes calls never cross space boundaries. This
8643 allows GCC to emit code which performs faster indirect calls.
8645 This option will not work in the presence of shared libraries or nested
8648 @item -mfixed-range=@var{register-range}
8649 @opindex mfixed-range
8650 Generate code treating the given register range as fixed registers.
8651 A fixed register is one that the register allocator can not use. This is
8652 useful when compiling kernel code. A register range is specified as
8653 two registers separated by a dash. Multiple register ranges can be
8654 specified separated by a comma.
8656 @item -mlong-load-store
8657 @opindex mlong-load-store
8658 Generate 3-instruction load and store sequences as sometimes required by
8659 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8662 @item -mportable-runtime
8663 @opindex mportable-runtime
8664 Use the portable calling conventions proposed by HP for ELF systems.
8668 Enable the use of assembler directives only GAS understands.
8670 @item -mschedule=@var{cpu-type}
8672 Schedule code according to the constraints for the machine type
8673 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8674 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8675 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8676 proper scheduling option for your machine. The default scheduling is
8680 @opindex mlinker-opt
8681 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8682 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8683 linkers in which they give bogus error messages when linking some programs.
8686 @opindex msoft-float
8687 Generate output containing library calls for floating point.
8688 @strong{Warning:} the requisite libraries are not available for all HPPA
8689 targets. Normally the facilities of the machine's usual C compiler are
8690 used, but this cannot be done directly in cross-compilation. You must make
8691 your own arrangements to provide suitable library functions for
8692 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8693 does provide software floating point support.
8695 @option{-msoft-float} changes the calling convention in the output file;
8696 therefore, it is only useful if you compile @emph{all} of a program with
8697 this option. In particular, you need to compile @file{libgcc.a}, the
8698 library that comes with GCC, with @option{-msoft-float} in order for
8703 Generate the predefine, @code{_SIO}, for server IO@. The default is
8704 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8705 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
8706 options are available under HP-UX and HI-UX@.
8710 Use GNU ld specific options. This passes @option{-shared} to ld when
8711 building a shared library. It is the default when GCC is configured,
8712 explicitly or implicitly, with the GNU linker. This option does not
8713 have any affect on which ld is called, it only changes what parameters
8714 are passed to that ld. The ld that is called is determined by the
8715 @option{--with-ld} configure option, GCC's program search path, and
8716 finally by the user's @env{PATH}. The linker used by GCC can be printed
8717 using @samp{which `gcc -print-prog-name=ld`}.
8721 Use HP ld specific options. This passes @option{-b} to ld when building
8722 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8723 links. It is the default when GCC is configured, explicitly or
8724 implicitly, with the HP linker. This option does not have any affect on
8725 which ld is called, it only changes what parameters are passed to that
8726 ld. The ld that is called is determined by the @option{--with-ld}
8727 configure option, GCC's program search path, and finally by the user's
8728 @env{PATH}. The linker used by GCC can be printed using @samp{which
8729 `gcc -print-prog-name=ld`}.
8732 @opindex mno-long-calls
8733 Generate code that uses long call sequences. This ensures that a call
8734 is always able to reach linker generated stubs. The default is to generate
8735 long calls only when the distance from the call site to the beginning
8736 of the function or translation unit, as the case may be, exceeds a
8737 predefined limit set by the branch type being used. The limits for
8738 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8739 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8742 Distances are measured from the beginning of functions when using the
8743 @option{-ffunction-sections} option, or when using the @option{-mgas}
8744 and @option{-mno-portable-runtime} options together under HP-UX with
8747 It is normally not desirable to use this option as it will degrade
8748 performance. However, it may be useful in large applications,
8749 particularly when partial linking is used to build the application.
8751 The types of long calls used depends on the capabilities of the
8752 assembler and linker, and the type of code being generated. The
8753 impact on systems that support long absolute calls, and long pic
8754 symbol-difference or pc-relative calls should be relatively small.
8755 However, an indirect call is used on 32-bit ELF systems in pic code
8756 and it is quite long.
8758 @item -munix=@var{unix-std}
8760 Generate compiler predefines and select a startfile for the specified
8761 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
8762 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
8763 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
8764 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
8765 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
8768 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
8769 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
8770 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
8771 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
8772 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
8773 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
8775 It is @emph{important} to note that this option changes the interfaces
8776 for various library routines. It also affects the operational behavior
8777 of the C library. Thus, @emph{extreme} care is needed in using this
8780 Library code that is intended to operate with more than one UNIX
8781 standard must test, set and restore the variable @var{__xpg4_extended_mask}
8782 as appropriate. Most GNU software doesn't provide this capability.
8786 Suppress the generation of link options to search libdld.sl when the
8787 @option{-static} option is specified on HP-UX 10 and later.
8791 The HP-UX implementation of setlocale in libc has a dependency on
8792 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8793 when the @option{-static} option is specified, special link options
8794 are needed to resolve this dependency.
8796 On HP-UX 10 and later, the GCC driver adds the necessary options to
8797 link with libdld.sl when the @option{-static} option is specified.
8798 This causes the resulting binary to be dynamic. On the 64-bit port,
8799 the linkers generate dynamic binaries by default in any case. The
8800 @option{-nolibdld} option can be used to prevent the GCC driver from
8801 adding these link options.
8805 Add support for multithreading with the @dfn{dce thread} library
8806 under HP-UX@. This option sets flags for both the preprocessor and
8810 @node i386 and x86-64 Options
8811 @subsection Intel 386 and AMD x86-64 Options
8812 @cindex i386 Options
8813 @cindex x86-64 Options
8814 @cindex Intel 386 Options
8815 @cindex AMD x86-64 Options
8817 These @samp{-m} options are defined for the i386 and x86-64 family of
8821 @item -mtune=@var{cpu-type}
8823 Tune to @var{cpu-type} everything applicable about the generated code, except
8824 for the ABI and the set of available instructions. The choices for
8828 Original Intel's i386 CPU@.
8830 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
8832 Intel Pentium CPU with no MMX support.
8834 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
8835 @item i686, pentiumpro
8836 Intel PentiumPro CPU@.
8838 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
8839 @item pentium3, pentium3m
8840 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
8843 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
8844 support. Used by Centrino notebooks.
8845 @item pentium4, pentium4m
8846 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
8848 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
8851 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
8852 SSE2 and SSE3 instruction set support.
8854 AMD K6 CPU with MMX instruction set support.
8856 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
8857 @item athlon, athlon-tbird
8858 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
8860 @item athlon-4, athlon-xp, athlon-mp
8861 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
8862 instruction set support.
8863 @item k8, opteron, athlon64, athlon-fx
8864 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
8865 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
8867 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
8870 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
8871 instruction set support.
8873 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
8874 implemented for this chip.)
8876 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
8877 implemented for this chip.)
8880 While picking a specific @var{cpu-type} will schedule things appropriately
8881 for that particular chip, the compiler will not generate any code that
8882 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8885 @item -march=@var{cpu-type}
8887 Generate instructions for the machine type @var{cpu-type}. The choices
8888 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8889 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8891 @item -mcpu=@var{cpu-type}
8893 A deprecated synonym for @option{-mtune}.
8902 @opindex mpentiumpro
8903 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8904 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8905 These synonyms are deprecated.
8907 @item -mfpmath=@var{unit}
8909 Generate floating point arithmetics for selected unit @var{unit}. The choices
8914 Use the standard 387 floating point coprocessor present majority of chips and
8915 emulated otherwise. Code compiled with this option will run almost everywhere.
8916 The temporary results are computed in 80bit precision instead of precision
8917 specified by the type resulting in slightly different results compared to most
8918 of other chips. See @option{-ffloat-store} for more detailed description.
8920 This is the default choice for i386 compiler.
8923 Use scalar floating point instructions present in the SSE instruction set.
8924 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8925 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8926 instruction set supports only single precision arithmetics, thus the double and
8927 extended precision arithmetics is still done using 387. Later version, present
8928 only in Pentium4 and the future AMD x86-64 chips supports double precision
8931 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
8932 or @option{-msse2} switches to enable SSE extensions and make this option
8933 effective. For the x86-64 compiler, these extensions are enabled by default.
8935 The resulting code should be considerably faster in the majority of cases and avoid
8936 the numerical instability problems of 387 code, but may break some existing
8937 code that expects temporaries to be 80bit.
8939 This is the default choice for the x86-64 compiler.
8942 Attempt to utilize both instruction sets at once. This effectively double the
8943 amount of available registers and on chips with separate execution units for
8944 387 and SSE the execution resources too. Use this option with care, as it is
8945 still experimental, because the GCC register allocator does not model separate
8946 functional units well resulting in instable performance.
8949 @item -masm=@var{dialect}
8950 @opindex masm=@var{dialect}
8951 Output asm instructions using selected @var{dialect}. Supported choices are
8952 @samp{intel} or @samp{att} (the default one).
8957 @opindex mno-ieee-fp
8958 Control whether or not the compiler uses IEEE floating point
8959 comparisons. These handle correctly the case where the result of a
8960 comparison is unordered.
8963 @opindex msoft-float
8964 Generate output containing library calls for floating point.
8965 @strong{Warning:} the requisite libraries are not part of GCC@.
8966 Normally the facilities of the machine's usual C compiler are used, but
8967 this can't be done directly in cross-compilation. You must make your
8968 own arrangements to provide suitable library functions for
8971 On machines where a function returns floating point results in the 80387
8972 register stack, some floating point opcodes may be emitted even if
8973 @option{-msoft-float} is used.
8975 @item -mno-fp-ret-in-387
8976 @opindex mno-fp-ret-in-387
8977 Do not use the FPU registers for return values of functions.
8979 The usual calling convention has functions return values of types
8980 @code{float} and @code{double} in an FPU register, even if there
8981 is no FPU@. The idea is that the operating system should emulate
8984 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8985 in ordinary CPU registers instead.
8987 @item -mno-fancy-math-387
8988 @opindex mno-fancy-math-387
8989 Some 387 emulators do not support the @code{sin}, @code{cos} and
8990 @code{sqrt} instructions for the 387. Specify this option to avoid
8991 generating those instructions. This option is the default on FreeBSD,
8992 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8993 indicates that the target cpu will always have an FPU and so the
8994 instruction will not need emulation. As of revision 2.6.1, these
8995 instructions are not generated unless you also use the
8996 @option{-funsafe-math-optimizations} switch.
8998 @item -malign-double
8999 @itemx -mno-align-double
9000 @opindex malign-double
9001 @opindex mno-align-double
9002 Control whether GCC aligns @code{double}, @code{long double}, and
9003 @code{long long} variables on a two word boundary or a one word
9004 boundary. Aligning @code{double} variables on a two word boundary will
9005 produce code that runs somewhat faster on a @samp{Pentium} at the
9006 expense of more memory.
9008 @strong{Warning:} if you use the @option{-malign-double} switch,
9009 structures containing the above types will be aligned differently than
9010 the published application binary interface specifications for the 386
9011 and will not be binary compatible with structures in code compiled
9012 without that switch.
9014 @item -m96bit-long-double
9015 @itemx -m128bit-long-double
9016 @opindex m96bit-long-double
9017 @opindex m128bit-long-double
9018 These switches control the size of @code{long double} type. The i386
9019 application binary interface specifies the size to be 96 bits,
9020 so @option{-m96bit-long-double} is the default in 32 bit mode.
9022 Modern architectures (Pentium and newer) would prefer @code{long double}
9023 to be aligned to an 8 or 16 byte boundary. In arrays or structures
9024 conforming to the ABI, this would not be possible. So specifying a
9025 @option{-m128bit-long-double} will align @code{long double}
9026 to a 16 byte boundary by padding the @code{long double} with an additional
9029 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
9030 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
9032 Notice that neither of these options enable any extra precision over the x87
9033 standard of 80 bits for a @code{long double}.
9035 @strong{Warning:} if you override the default value for your target ABI, the
9036 structures and arrays containing @code{long double} variables will change
9037 their size as well as function calling convention for function taking
9038 @code{long double} will be modified. Hence they will not be binary
9039 compatible with arrays or structures in code compiled without that switch.
9043 @itemx -mno-svr3-shlib
9044 @opindex msvr3-shlib
9045 @opindex mno-svr3-shlib
9046 Control whether GCC places uninitialized local variables into the
9047 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
9048 into @code{bss}. These options are meaningful only on System V Release 3.
9052 Use a different function-calling convention, in which functions that
9053 take a fixed number of arguments return with the @code{ret} @var{num}
9054 instruction, which pops their arguments while returning. This saves one
9055 instruction in the caller since there is no need to pop the arguments
9058 You can specify that an individual function is called with this calling
9059 sequence with the function attribute @samp{stdcall}. You can also
9060 override the @option{-mrtd} option by using the function attribute
9061 @samp{cdecl}. @xref{Function Attributes}.
9063 @strong{Warning:} this calling convention is incompatible with the one
9064 normally used on Unix, so you cannot use it if you need to call
9065 libraries compiled with the Unix compiler.
9067 Also, you must provide function prototypes for all functions that
9068 take variable numbers of arguments (including @code{printf});
9069 otherwise incorrect code will be generated for calls to those
9072 In addition, seriously incorrect code will result if you call a
9073 function with too many arguments. (Normally, extra arguments are
9074 harmlessly ignored.)
9076 @item -mregparm=@var{num}
9078 Control how many registers are used to pass integer arguments. By
9079 default, no registers are used to pass arguments, and at most 3
9080 registers can be used. You can control this behavior for a specific
9081 function by using the function attribute @samp{regparm}.
9082 @xref{Function Attributes}.
9084 @strong{Warning:} if you use this switch, and
9085 @var{num} is nonzero, then you must build all modules with the same
9086 value, including any libraries. This includes the system libraries and
9090 @opindex msseregparm
9091 Use SSE register passing conventions for float and double arguments
9092 and return values. You can control this behavior for a specific
9093 function by using the function attribute @samp{sseregparm}.
9094 @xref{Function Attributes}.
9096 @strong{Warning:} if you use this switch then you must build all
9097 modules with the same value, including any libraries. This includes
9098 the system libraries and startup modules.
9100 @item -mpreferred-stack-boundary=@var{num}
9101 @opindex mpreferred-stack-boundary
9102 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
9103 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
9104 the default is 4 (16 bytes or 128 bits), except when optimizing for code
9105 size (@option{-Os}), in which case the default is the minimum correct
9106 alignment (4 bytes for x86, and 8 bytes for x86-64).
9108 On Pentium and PentiumPro, @code{double} and @code{long double} values
9109 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
9110 suffer significant run time performance penalties. On Pentium III, the
9111 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
9112 penalties if it is not 16 byte aligned.
9114 To ensure proper alignment of this values on the stack, the stack boundary
9115 must be as aligned as that required by any value stored on the stack.
9116 Further, every function must be generated such that it keeps the stack
9117 aligned. Thus calling a function compiled with a higher preferred
9118 stack boundary from a function compiled with a lower preferred stack
9119 boundary will most likely misalign the stack. It is recommended that
9120 libraries that use callbacks always use the default setting.
9122 This extra alignment does consume extra stack space, and generally
9123 increases code size. Code that is sensitive to stack space usage, such
9124 as embedded systems and operating system kernels, may want to reduce the
9125 preferred alignment to @option{-mpreferred-stack-boundary=2}.
9143 These switches enable or disable the use of built-in functions that allow
9144 direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow extensions of the
9147 @xref{X86 Built-in Functions}, for details of the functions enabled
9148 and disabled by these switches.
9150 To have SSE/SSE2 instructions generated automatically from floating-point
9151 code, see @option{-mfpmath=sse}.
9154 @itemx -mno-push-args
9156 @opindex mno-push-args
9157 Use PUSH operations to store outgoing parameters. This method is shorter
9158 and usually equally fast as method using SUB/MOV operations and is enabled
9159 by default. In some cases disabling it may improve performance because of
9160 improved scheduling and reduced dependencies.
9162 @item -maccumulate-outgoing-args
9163 @opindex maccumulate-outgoing-args
9164 If enabled, the maximum amount of space required for outgoing arguments will be
9165 computed in the function prologue. This is faster on most modern CPUs
9166 because of reduced dependencies, improved scheduling and reduced stack usage
9167 when preferred stack boundary is not equal to 2. The drawback is a notable
9168 increase in code size. This switch implies @option{-mno-push-args}.
9172 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
9173 on thread-safe exception handling must compile and link all code with the
9174 @option{-mthreads} option. When compiling, @option{-mthreads} defines
9175 @option{-D_MT}; when linking, it links in a special thread helper library
9176 @option{-lmingwthrd} which cleans up per thread exception handling data.
9178 @item -mno-align-stringops
9179 @opindex mno-align-stringops
9180 Do not align destination of inlined string operations. This switch reduces
9181 code size and improves performance in case the destination is already aligned,
9182 but GCC doesn't know about it.
9184 @item -minline-all-stringops
9185 @opindex minline-all-stringops
9186 By default GCC inlines string operations only when destination is known to be
9187 aligned at least to 4 byte boundary. This enables more inlining, increase code
9188 size, but may improve performance of code that depends on fast memcpy, strlen
9189 and memset for short lengths.
9191 @item -momit-leaf-frame-pointer
9192 @opindex momit-leaf-frame-pointer
9193 Don't keep the frame pointer in a register for leaf functions. This
9194 avoids the instructions to save, set up and restore frame pointers and
9195 makes an extra register available in leaf functions. The option
9196 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9197 which might make debugging harder.
9199 @item -mtls-direct-seg-refs
9200 @itemx -mno-tls-direct-seg-refs
9201 @opindex mtls-direct-seg-refs
9202 Controls whether TLS variables may be accessed with offsets from the
9203 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
9204 or whether the thread base pointer must be added. Whether or not this
9205 is legal depends on the operating system, and whether it maps the
9206 segment to cover the entire TLS area.
9208 For systems that use GNU libc, the default is on.
9211 These @samp{-m} switches are supported in addition to the above
9212 on AMD x86-64 processors in 64-bit environments.
9219 Generate code for a 32-bit or 64-bit environment.
9220 The 32-bit environment sets int, long and pointer to 32 bits and
9221 generates code that runs on any i386 system.
9222 The 64-bit environment sets int to 32 bits and long and pointer
9223 to 64 bits and generates code for AMD's x86-64 architecture.
9226 @opindex no-red-zone
9227 Do not use a so called red zone for x86-64 code. The red zone is mandated
9228 by the x86-64 ABI, it is a 128-byte area beyond the location of the
9229 stack pointer that will not be modified by signal or interrupt handlers
9230 and therefore can be used for temporary data without adjusting the stack
9231 pointer. The flag @option{-mno-red-zone} disables this red zone.
9233 @item -mcmodel=small
9234 @opindex mcmodel=small
9235 Generate code for the small code model: the program and its symbols must
9236 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
9237 Programs can be statically or dynamically linked. This is the default
9240 @item -mcmodel=kernel
9241 @opindex mcmodel=kernel
9242 Generate code for the kernel code model. The kernel runs in the
9243 negative 2 GB of the address space.
9244 This model has to be used for Linux kernel code.
9246 @item -mcmodel=medium
9247 @opindex mcmodel=medium
9248 Generate code for the medium model: The program is linked in the lower 2
9249 GB of the address space but symbols can be located anywhere in the
9250 address space. Programs can be statically or dynamically linked, but
9251 building of shared libraries are not supported with the medium model.
9253 @item -mcmodel=large
9254 @opindex mcmodel=large
9255 Generate code for the large model: This model makes no assumptions
9256 about addresses and sizes of sections. Currently GCC does not implement
9261 @subsection IA-64 Options
9262 @cindex IA-64 Options
9264 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9268 @opindex mbig-endian
9269 Generate code for a big endian target. This is the default for HP-UX@.
9271 @item -mlittle-endian
9272 @opindex mlittle-endian
9273 Generate code for a little endian target. This is the default for AIX5
9280 Generate (or don't) code for the GNU assembler. This is the default.
9281 @c Also, this is the default if the configure option @option{--with-gnu-as}
9288 Generate (or don't) code for the GNU linker. This is the default.
9289 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9294 Generate code that does not use a global pointer register. The result
9295 is not position independent code, and violates the IA-64 ABI@.
9297 @item -mvolatile-asm-stop
9298 @itemx -mno-volatile-asm-stop
9299 @opindex mvolatile-asm-stop
9300 @opindex mno-volatile-asm-stop
9301 Generate (or don't) a stop bit immediately before and after volatile asm
9304 @item -mregister-names
9305 @itemx -mno-register-names
9306 @opindex mregister-names
9307 @opindex mno-register-names
9308 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9309 the stacked registers. This may make assembler output more readable.
9315 Disable (or enable) optimizations that use the small data section. This may
9316 be useful for working around optimizer bugs.
9319 @opindex mconstant-gp
9320 Generate code that uses a single constant global pointer value. This is
9321 useful when compiling kernel code.
9325 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9326 This is useful when compiling firmware code.
9328 @item -minline-float-divide-min-latency
9329 @opindex minline-float-divide-min-latency
9330 Generate code for inline divides of floating point values
9331 using the minimum latency algorithm.
9333 @item -minline-float-divide-max-throughput
9334 @opindex minline-float-divide-max-throughput
9335 Generate code for inline divides of floating point values
9336 using the maximum throughput algorithm.
9338 @item -minline-int-divide-min-latency
9339 @opindex minline-int-divide-min-latency
9340 Generate code for inline divides of integer values
9341 using the minimum latency algorithm.
9343 @item -minline-int-divide-max-throughput
9344 @opindex minline-int-divide-max-throughput
9345 Generate code for inline divides of integer values
9346 using the maximum throughput algorithm.
9348 @item -minline-sqrt-min-latency
9349 @opindex minline-sqrt-min-latency
9350 Generate code for inline square roots
9351 using the minimum latency algorithm.
9353 @item -minline-sqrt-max-throughput
9354 @opindex minline-sqrt-max-throughput
9355 Generate code for inline square roots
9356 using the maximum throughput algorithm.
9358 @item -mno-dwarf2-asm
9360 @opindex mno-dwarf2-asm
9361 @opindex mdwarf2-asm
9362 Don't (or do) generate assembler code for the DWARF2 line number debugging
9363 info. This may be useful when not using the GNU assembler.
9365 @item -mearly-stop-bits
9366 @itemx -mno-early-stop-bits
9367 @opindex mearly-stop-bits
9368 @opindex mno-early-stop-bits
9369 Allow stop bits to be placed earlier than immediately preceding the
9370 instruction that triggered the stop bit. This can improve instruction
9371 scheduling, but does not always do so.
9373 @item -mfixed-range=@var{register-range}
9374 @opindex mfixed-range
9375 Generate code treating the given register range as fixed registers.
9376 A fixed register is one that the register allocator can not use. This is
9377 useful when compiling kernel code. A register range is specified as
9378 two registers separated by a dash. Multiple register ranges can be
9379 specified separated by a comma.
9381 @item -mtls-size=@var{tls-size}
9383 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9386 @item -mtune-arch=@var{cpu-type}
9388 Tune the instruction scheduling for a particular CPU, Valid values are
9389 itanium, itanium1, merced, itanium2, and mckinley.
9395 Add support for multithreading using the POSIX threads library. This
9396 option sets flags for both the preprocessor and linker. It does
9397 not affect the thread safety of object code produced by the compiler or
9398 that of libraries supplied with it. These are HP-UX specific flags.
9404 Generate code for a 32-bit or 64-bit environment.
9405 The 32-bit environment sets int, long and pointer to 32 bits.
9406 The 64-bit environment sets int to 32 bits and long and pointer
9407 to 64 bits. These are HP-UX specific flags.
9412 @subsection M32C Options
9413 @cindex M32C options
9416 @item -mcpu=@var{name}
9418 Select the CPU for which code is generated. @var{name} may be one of
9419 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
9420 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
9425 Specifies that the program will be run on the simulator. This causes
9426 an alternate runtime library to be linked in which supports, for
9427 example, file I/O. You must not use this option when generating
9428 programs that will run on real hardware; you must provide your own
9429 runtime library for whatever I/O functions are needed.
9431 @item -memregs=@var{number}
9433 Specifies the number of memory-based pseudo-registers GCC will use
9434 during code generation. These pseudo-registers will be used like real
9435 registers, so there is a tradeoff between GCC's ability to fit the
9436 code into available registers, and the performance penalty of using
9437 memory instead of registers. Note that all modules in a program must
9438 be compiled with the same value for this option. Because of that, you
9439 must not use this option with the default runtime libraries gcc
9444 @node M32R/D Options
9445 @subsection M32R/D Options
9446 @cindex M32R/D options
9448 These @option{-m} options are defined for Renesas M32R/D architectures:
9453 Generate code for the M32R/2@.
9457 Generate code for the M32R/X@.
9461 Generate code for the M32R@. This is the default.
9464 @opindex mmodel=small
9465 Assume all objects live in the lower 16MB of memory (so that their addresses
9466 can be loaded with the @code{ld24} instruction), and assume all subroutines
9467 are reachable with the @code{bl} instruction.
9468 This is the default.
9470 The addressability of a particular object can be set with the
9471 @code{model} attribute.
9473 @item -mmodel=medium
9474 @opindex mmodel=medium
9475 Assume objects may be anywhere in the 32-bit address space (the compiler
9476 will generate @code{seth/add3} instructions to load their addresses), and
9477 assume all subroutines are reachable with the @code{bl} instruction.
9480 @opindex mmodel=large
9481 Assume objects may be anywhere in the 32-bit address space (the compiler
9482 will generate @code{seth/add3} instructions to load their addresses), and
9483 assume subroutines may not be reachable with the @code{bl} instruction
9484 (the compiler will generate the much slower @code{seth/add3/jl}
9485 instruction sequence).
9488 @opindex msdata=none
9489 Disable use of the small data area. Variables will be put into
9490 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
9491 @code{section} attribute has been specified).
9492 This is the default.
9494 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
9495 Objects may be explicitly put in the small data area with the
9496 @code{section} attribute using one of these sections.
9499 @opindex msdata=sdata
9500 Put small global and static data in the small data area, but do not
9501 generate special code to reference them.
9505 Put small global and static data in the small data area, and generate
9506 special instructions to reference them.
9510 @cindex smaller data references
9511 Put global and static objects less than or equal to @var{num} bytes
9512 into the small data or bss sections instead of the normal data or bss
9513 sections. The default value of @var{num} is 8.
9514 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
9515 for this option to have any effect.
9517 All modules should be compiled with the same @option{-G @var{num}} value.
9518 Compiling with different values of @var{num} may or may not work; if it
9519 doesn't the linker will give an error message---incorrect code will not be
9524 Makes the M32R specific code in the compiler display some statistics
9525 that might help in debugging programs.
9528 @opindex malign-loops
9529 Align all loops to a 32-byte boundary.
9531 @item -mno-align-loops
9532 @opindex mno-align-loops
9533 Do not enforce a 32-byte alignment for loops. This is the default.
9535 @item -missue-rate=@var{number}
9536 @opindex missue-rate=@var{number}
9537 Issue @var{number} instructions per cycle. @var{number} can only be 1
9540 @item -mbranch-cost=@var{number}
9541 @opindex mbranch-cost=@var{number}
9542 @var{number} can only be 1 or 2. If it is 1 then branches will be
9543 preferred over conditional code, if it is 2, then the opposite will
9546 @item -mflush-trap=@var{number}
9547 @opindex mflush-trap=@var{number}
9548 Specifies the trap number to use to flush the cache. The default is
9549 12. Valid numbers are between 0 and 15 inclusive.
9551 @item -mno-flush-trap
9552 @opindex mno-flush-trap
9553 Specifies that the cache cannot be flushed by using a trap.
9555 @item -mflush-func=@var{name}
9556 @opindex mflush-func=@var{name}
9557 Specifies the name of the operating system function to call to flush
9558 the cache. The default is @emph{_flush_cache}, but a function call
9559 will only be used if a trap is not available.
9561 @item -mno-flush-func
9562 @opindex mno-flush-func
9563 Indicates that there is no OS function for flushing the cache.
9567 @node M680x0 Options
9568 @subsection M680x0 Options
9569 @cindex M680x0 options
9571 These are the @samp{-m} options defined for the 68000 series. The default
9572 values for these options depends on which style of 68000 was selected when
9573 the compiler was configured; the defaults for the most common choices are
9581 Generate output for a 68000. This is the default
9582 when the compiler is configured for 68000-based systems.
9584 Use this option for microcontrollers with a 68000 or EC000 core,
9585 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
9591 Generate output for a 68020. This is the default
9592 when the compiler is configured for 68020-based systems.
9596 Generate output containing 68881 instructions for floating point.
9597 This is the default for most 68020 systems unless @option{--nfp} was
9598 specified when the compiler was configured.
9602 Generate output for a 68030. This is the default when the compiler is
9603 configured for 68030-based systems.
9607 Generate output for a 68040. This is the default when the compiler is
9608 configured for 68040-based systems.
9610 This option inhibits the use of 68881/68882 instructions that have to be
9611 emulated by software on the 68040. Use this option if your 68040 does not
9612 have code to emulate those instructions.
9616 Generate output for a 68060. This is the default when the compiler is
9617 configured for 68060-based systems.
9619 This option inhibits the use of 68020 and 68881/68882 instructions that
9620 have to be emulated by software on the 68060. Use this option if your 68060
9621 does not have code to emulate those instructions.
9625 Generate output for a CPU32. This is the default
9626 when the compiler is configured for CPU32-based systems.
9628 Use this option for microcontrollers with a
9629 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
9630 68336, 68340, 68341, 68349 and 68360.
9634 Generate output for a 520X ``coldfire'' family cpu. This is the default
9635 when the compiler is configured for 520X-based systems.
9637 Use this option for microcontroller with a 5200 core, including
9638 the MCF5202, MCF5203, MCF5204 and MCF5202.
9643 Generate output for a 68040, without using any of the new instructions.
9644 This results in code which can run relatively efficiently on either a
9645 68020/68881 or a 68030 or a 68040. The generated code does use the
9646 68881 instructions that are emulated on the 68040.
9650 Generate output for a 68060, without using any of the new instructions.
9651 This results in code which can run relatively efficiently on either a
9652 68020/68881 or a 68030 or a 68040. The generated code does use the
9653 68881 instructions that are emulated on the 68060.
9656 @opindex msoft-float
9657 Generate output containing library calls for floating point.
9658 @strong{Warning:} the requisite libraries are not available for all m68k
9659 targets. Normally the facilities of the machine's usual C compiler are
9660 used, but this can't be done directly in cross-compilation. You must
9661 make your own arrangements to provide suitable library functions for
9662 cross-compilation. The embedded targets @samp{m68k-*-aout} and
9663 @samp{m68k-*-coff} do provide software floating point support.
9667 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9668 Additionally, parameters passed on the stack are also aligned to a
9669 16-bit boundary even on targets whose API mandates promotion to 32-bit.
9672 @opindex mnobitfield
9673 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
9674 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
9678 Do use the bit-field instructions. The @option{-m68020} option implies
9679 @option{-mbitfield}. This is the default if you use a configuration
9680 designed for a 68020.
9684 Use a different function-calling convention, in which functions
9685 that take a fixed number of arguments return with the @code{rtd}
9686 instruction, which pops their arguments while returning. This
9687 saves one instruction in the caller since there is no need to pop
9688 the arguments there.
9690 This calling convention is incompatible with the one normally
9691 used on Unix, so you cannot use it if you need to call libraries
9692 compiled with the Unix compiler.
9694 Also, you must provide function prototypes for all functions that
9695 take variable numbers of arguments (including @code{printf});
9696 otherwise incorrect code will be generated for calls to those
9699 In addition, seriously incorrect code will result if you call a
9700 function with too many arguments. (Normally, extra arguments are
9701 harmlessly ignored.)
9703 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
9704 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
9707 @itemx -mno-align-int
9709 @opindex mno-align-int
9710 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
9711 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
9712 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
9713 Aligning variables on 32-bit boundaries produces code that runs somewhat
9714 faster on processors with 32-bit busses at the expense of more memory.
9716 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
9717 align structures containing the above types differently than
9718 most published application binary interface specifications for the m68k.
9722 Use the pc-relative addressing mode of the 68000 directly, instead of
9723 using a global offset table. At present, this option implies @option{-fpic},
9724 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
9725 not presently supported with @option{-mpcrel}, though this could be supported for
9726 68020 and higher processors.
9728 @item -mno-strict-align
9729 @itemx -mstrict-align
9730 @opindex mno-strict-align
9731 @opindex mstrict-align
9732 Do not (do) assume that unaligned memory references will be handled by
9736 Generate code that allows the data segment to be located in a different
9737 area of memory from the text segment. This allows for execute in place in
9738 an environment without virtual memory management. This option implies
9742 Generate code that assumes that the data segment follows the text segment.
9743 This is the default.
9745 @item -mid-shared-library
9746 Generate code that supports shared libraries via the library ID method.
9747 This allows for execute in place and shared libraries in an environment
9748 without virtual memory management. This option implies @option{-fPIC}.
9750 @item -mno-id-shared-library
9751 Generate code that doesn't assume ID based shared libraries are being used.
9752 This is the default.
9754 @item -mshared-library-id=n
9755 Specified the identification number of the ID based shared library being
9756 compiled. Specifying a value of 0 will generate more compact code, specifying
9757 other values will force the allocation of that number to the current
9758 library but is no more space or time efficient than omitting this option.
9762 @node M68hc1x Options
9763 @subsection M68hc1x Options
9764 @cindex M68hc1x options
9766 These are the @samp{-m} options defined for the 68hc11 and 68hc12
9767 microcontrollers. The default values for these options depends on
9768 which style of microcontroller was selected when the compiler was configured;
9769 the defaults for the most common choices are given below.
9776 Generate output for a 68HC11. This is the default
9777 when the compiler is configured for 68HC11-based systems.
9783 Generate output for a 68HC12. This is the default
9784 when the compiler is configured for 68HC12-based systems.
9790 Generate output for a 68HCS12.
9793 @opindex mauto-incdec
9794 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
9801 Enable the use of 68HC12 min and max instructions.
9804 @itemx -mno-long-calls
9805 @opindex mlong-calls
9806 @opindex mno-long-calls
9807 Treat all calls as being far away (near). If calls are assumed to be
9808 far away, the compiler will use the @code{call} instruction to
9809 call a function and the @code{rtc} instruction for returning.
9813 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9815 @item -msoft-reg-count=@var{count}
9816 @opindex msoft-reg-count
9817 Specify the number of pseudo-soft registers which are used for the
9818 code generation. The maximum number is 32. Using more pseudo-soft
9819 register may or may not result in better code depending on the program.
9820 The default is 4 for 68HC11 and 2 for 68HC12.
9825 @subsection MCore Options
9826 @cindex MCore options
9828 These are the @samp{-m} options defined for the Motorola M*Core
9836 @opindex mno-hardlit
9837 Inline constants into the code stream if it can be done in two
9838 instructions or less.
9844 Use the divide instruction. (Enabled by default).
9846 @item -mrelax-immediate
9847 @itemx -mno-relax-immediate
9848 @opindex mrelax-immediate
9849 @opindex mno-relax-immediate
9850 Allow arbitrary sized immediates in bit operations.
9852 @item -mwide-bitfields
9853 @itemx -mno-wide-bitfields
9854 @opindex mwide-bitfields
9855 @opindex mno-wide-bitfields
9856 Always treat bit-fields as int-sized.
9858 @item -m4byte-functions
9859 @itemx -mno-4byte-functions
9860 @opindex m4byte-functions
9861 @opindex mno-4byte-functions
9862 Force all functions to be aligned to a four byte boundary.
9864 @item -mcallgraph-data
9865 @itemx -mno-callgraph-data
9866 @opindex mcallgraph-data
9867 @opindex mno-callgraph-data
9868 Emit callgraph information.
9871 @itemx -mno-slow-bytes
9872 @opindex mslow-bytes
9873 @opindex mno-slow-bytes
9874 Prefer word access when reading byte quantities.
9876 @item -mlittle-endian
9878 @opindex mlittle-endian
9879 @opindex mbig-endian
9880 Generate code for a little endian target.
9886 Generate code for the 210 processor.
9890 @subsection MIPS Options
9891 @cindex MIPS options
9897 Generate big-endian code.
9901 Generate little-endian code. This is the default for @samp{mips*el-*-*}
9904 @item -march=@var{arch}
9906 Generate code that will run on @var{arch}, which can be the name of a
9907 generic MIPS ISA, or the name of a particular processor.
9909 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
9910 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
9911 The processor names are:
9912 @samp{4kc}, @samp{4km}, @samp{4kp},
9913 @samp{5kc}, @samp{5kf},
9915 @samp{24k}, @samp{24kc}, @samp{24kf}, @samp{24kx},
9918 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
9919 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
9920 @samp{rm7000}, @samp{rm9000},
9923 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
9924 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
9925 The special value @samp{from-abi} selects the
9926 most compatible architecture for the selected ABI (that is,
9927 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
9929 In processor names, a final @samp{000} can be abbreviated as @samp{k}
9930 (for example, @samp{-march=r2k}). Prefixes are optional, and
9931 @samp{vr} may be written @samp{r}.
9933 GCC defines two macros based on the value of this option. The first
9934 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
9935 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
9936 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
9937 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
9938 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
9940 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
9941 above. In other words, it will have the full prefix and will not
9942 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
9943 the macro names the resolved architecture (either @samp{"mips1"} or
9944 @samp{"mips3"}). It names the default architecture when no
9945 @option{-march} option is given.
9947 @item -mtune=@var{arch}
9949 Optimize for @var{arch}. Among other things, this option controls
9950 the way instructions are scheduled, and the perceived cost of arithmetic
9951 operations. The list of @var{arch} values is the same as for
9954 When this option is not used, GCC will optimize for the processor
9955 specified by @option{-march}. By using @option{-march} and
9956 @option{-mtune} together, it is possible to generate code that will
9957 run on a family of processors, but optimize the code for one
9958 particular member of that family.
9960 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
9961 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
9962 @samp{-march} ones described above.
9966 Equivalent to @samp{-march=mips1}.
9970 Equivalent to @samp{-march=mips2}.
9974 Equivalent to @samp{-march=mips3}.
9978 Equivalent to @samp{-march=mips4}.
9982 Equivalent to @samp{-march=mips32}.
9986 Equivalent to @samp{-march=mips32r2}.
9990 Equivalent to @samp{-march=mips64}.
9996 Generate (do not generate) MIPS16 code. If GCC is targetting a
9997 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
10009 Generate code for the given ABI@.
10011 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
10012 generates 64-bit code when you select a 64-bit architecture, but you
10013 can use @option{-mgp32} to get 32-bit code instead.
10015 For information about the O64 ABI, see
10016 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
10019 @itemx -mno-abicalls
10021 @opindex mno-abicalls
10022 Generate (do not generate) SVR4-style position-independent code.
10023 @option{-mabicalls} is the default for SVR4-based systems.
10029 Lift (do not lift) the usual restrictions on the size of the global
10032 GCC normally uses a single instruction to load values from the GOT@.
10033 While this is relatively efficient, it will only work if the GOT
10034 is smaller than about 64k. Anything larger will cause the linker
10035 to report an error such as:
10037 @cindex relocation truncated to fit (MIPS)
10039 relocation truncated to fit: R_MIPS_GOT16 foobar
10042 If this happens, you should recompile your code with @option{-mxgot}.
10043 It should then work with very large GOTs, although it will also be
10044 less efficient, since it will take three instructions to fetch the
10045 value of a global symbol.
10047 Note that some linkers can create multiple GOTs. If you have such a
10048 linker, you should only need to use @option{-mxgot} when a single object
10049 file accesses more than 64k's worth of GOT entries. Very few do.
10051 These options have no effect unless GCC is generating position
10056 Assume that general-purpose registers are 32 bits wide.
10060 Assume that general-purpose registers are 64 bits wide.
10064 Assume that floating-point registers are 32 bits wide.
10068 Assume that floating-point registers are 64 bits wide.
10071 @opindex mhard-float
10072 Use floating-point coprocessor instructions.
10075 @opindex msoft-float
10076 Do not use floating-point coprocessor instructions. Implement
10077 floating-point calculations using library calls instead.
10079 @item -msingle-float
10080 @opindex msingle-float
10081 Assume that the floating-point coprocessor only supports single-precision
10084 @itemx -mdouble-float
10085 @opindex mdouble-float
10086 Assume that the floating-point coprocessor supports double-precision
10087 operations. This is the default.
10093 Use (do not use) the MIPS DSP ASE. @xref{MIPS DSP Built-in Functions}.
10095 @itemx -mpaired-single
10096 @itemx -mno-paired-single
10097 @opindex mpaired-single
10098 @opindex mno-paired-single
10099 Use (do not use) paired-single floating-point instructions.
10100 @xref{MIPS Paired-Single Support}. This option can only be used
10101 when generating 64-bit code and requires hardware floating-point
10102 support to be enabled.
10107 @opindex mno-mips3d
10108 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
10109 The option @option{-mips3d} implies @option{-mpaired-single}.
10113 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
10114 an explanation of the default and the way that the pointer size is
10119 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
10121 The default size of @code{int}s, @code{long}s and pointers depends on
10122 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
10123 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
10124 32-bit @code{long}s. Pointers are the same size as @code{long}s,
10125 or the same size as integer registers, whichever is smaller.
10131 Assume (do not assume) that all symbols have 32-bit values, regardless
10132 of the selected ABI@. This option is useful in combination with
10133 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
10134 to generate shorter and faster references to symbolic addresses.
10138 @cindex smaller data references (MIPS)
10139 @cindex gp-relative references (MIPS)
10140 Put global and static items less than or equal to @var{num} bytes into
10141 the small data or bss section instead of the normal data or bss section.
10142 This allows the data to be accessed using a single instruction.
10144 All modules should be compiled with the same @option{-G @var{num}}
10147 @item -membedded-data
10148 @itemx -mno-embedded-data
10149 @opindex membedded-data
10150 @opindex mno-embedded-data
10151 Allocate variables to the read-only data section first if possible, then
10152 next in the small data section if possible, otherwise in data. This gives
10153 slightly slower code than the default, but reduces the amount of RAM required
10154 when executing, and thus may be preferred for some embedded systems.
10156 @item -muninit-const-in-rodata
10157 @itemx -mno-uninit-const-in-rodata
10158 @opindex muninit-const-in-rodata
10159 @opindex mno-uninit-const-in-rodata
10160 Put uninitialized @code{const} variables in the read-only data section.
10161 This option is only meaningful in conjunction with @option{-membedded-data}.
10163 @item -msplit-addresses
10164 @itemx -mno-split-addresses
10165 @opindex msplit-addresses
10166 @opindex mno-split-addresses
10167 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
10168 relocation operators. This option has been superseded by
10169 @option{-mexplicit-relocs} but is retained for backwards compatibility.
10171 @item -mexplicit-relocs
10172 @itemx -mno-explicit-relocs
10173 @opindex mexplicit-relocs
10174 @opindex mno-explicit-relocs
10175 Use (do not use) assembler relocation operators when dealing with symbolic
10176 addresses. The alternative, selected by @option{-mno-explicit-relocs},
10177 is to use assembler macros instead.
10179 @option{-mexplicit-relocs} is the default if GCC was configured
10180 to use an assembler that supports relocation operators.
10182 @item -mcheck-zero-division
10183 @itemx -mno-check-zero-division
10184 @opindex mcheck-zero-division
10185 @opindex mno-check-zero-division
10186 Trap (do not trap) on integer division by zero. The default is
10187 @option{-mcheck-zero-division}.
10189 @item -mdivide-traps
10190 @itemx -mdivide-breaks
10191 @opindex mdivide-traps
10192 @opindex mdivide-breaks
10193 MIPS systems check for division by zero by generating either a
10194 conditional trap or a break instruction. Using traps results in
10195 smaller code, but is only supported on MIPS II and later. Also, some
10196 versions of the Linux kernel have a bug that prevents trap from
10197 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
10198 allow conditional traps on architectures that support them and
10199 @option{-mdivide-breaks} to force the use of breaks.
10201 The default is usually @option{-mdivide-traps}, but this can be
10202 overridden at configure time using @option{--with-divide=breaks}.
10203 Divide-by-zero checks can be completely disabled using
10204 @option{-mno-check-zero-division}.
10209 @opindex mno-memcpy
10210 Force (do not force) the use of @code{memcpy()} for non-trivial block
10211 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
10212 most constant-sized copies.
10215 @itemx -mno-long-calls
10216 @opindex mlong-calls
10217 @opindex mno-long-calls
10218 Disable (do not disable) use of the @code{jal} instruction. Calling
10219 functions using @code{jal} is more efficient but requires the caller
10220 and callee to be in the same 256 megabyte segment.
10222 This option has no effect on abicalls code. The default is
10223 @option{-mno-long-calls}.
10229 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
10230 instructions, as provided by the R4650 ISA@.
10233 @itemx -mno-fused-madd
10234 @opindex mfused-madd
10235 @opindex mno-fused-madd
10236 Enable (disable) use of the floating point multiply-accumulate
10237 instructions, when they are available. The default is
10238 @option{-mfused-madd}.
10240 When multiply-accumulate instructions are used, the intermediate
10241 product is calculated to infinite precision and is not subject to
10242 the FCSR Flush to Zero bit. This may be undesirable in some
10247 Tell the MIPS assembler to not run its preprocessor over user
10248 assembler files (with a @samp{.s} suffix) when assembling them.
10251 @itemx -mno-fix-r4000
10252 @opindex mfix-r4000
10253 @opindex mno-fix-r4000
10254 Work around certain R4000 CPU errata:
10257 A double-word or a variable shift may give an incorrect result if executed
10258 immediately after starting an integer division.
10260 A double-word or a variable shift may give an incorrect result if executed
10261 while an integer multiplication is in progress.
10263 An integer division may give an incorrect result if started in a delay slot
10264 of a taken branch or a jump.
10268 @itemx -mno-fix-r4400
10269 @opindex mfix-r4400
10270 @opindex mno-fix-r4400
10271 Work around certain R4400 CPU errata:
10274 A double-word or a variable shift may give an incorrect result if executed
10275 immediately after starting an integer division.
10279 @itemx -mno-fix-vr4120
10280 @opindex mfix-vr4120
10281 Work around certain VR4120 errata:
10284 @code{dmultu} does not always produce the correct result.
10286 @code{div} and @code{ddiv} do not always produce the correct result if one
10287 of the operands is negative.
10289 The workarounds for the division errata rely on special functions in
10290 @file{libgcc.a}. At present, these functions are only provided by
10291 the @code{mips64vr*-elf} configurations.
10293 Other VR4120 errata require a nop to be inserted between certain pairs of
10294 instructions. These errata are handled by the assembler, not by GCC itself.
10297 @opindex mfix-vr4130
10298 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
10299 workarounds are implemented by the assembler rather than by GCC,
10300 although GCC will avoid using @code{mflo} and @code{mfhi} if the
10301 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
10302 instructions are available instead.
10305 @itemx -mno-fix-sb1
10307 Work around certain SB-1 CPU core errata.
10308 (This flag currently works around the SB-1 revision 2
10309 ``F1'' and ``F2'' floating point errata.)
10311 @item -mflush-func=@var{func}
10312 @itemx -mno-flush-func
10313 @opindex mflush-func
10314 Specifies the function to call to flush the I and D caches, or to not
10315 call any such function. If called, the function must take the same
10316 arguments as the common @code{_flush_func()}, that is, the address of the
10317 memory range for which the cache is being flushed, the size of the
10318 memory range, and the number 3 (to flush both caches). The default
10319 depends on the target GCC was configured for, but commonly is either
10320 @samp{_flush_func} or @samp{__cpu_flush}.
10322 @item -mbranch-likely
10323 @itemx -mno-branch-likely
10324 @opindex mbranch-likely
10325 @opindex mno-branch-likely
10326 Enable or disable use of Branch Likely instructions, regardless of the
10327 default for the selected architecture. By default, Branch Likely
10328 instructions may be generated if they are supported by the selected
10329 architecture. An exception is for the MIPS32 and MIPS64 architectures
10330 and processors which implement those architectures; for those, Branch
10331 Likely instructions will not be generated by default because the MIPS32
10332 and MIPS64 architectures specifically deprecate their use.
10334 @item -mfp-exceptions
10335 @itemx -mno-fp-exceptions
10336 @opindex mfp-exceptions
10337 Specifies whether FP exceptions are enabled. This affects how we schedule
10338 FP instructions for some processors. The default is that FP exceptions are
10341 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
10342 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
10345 @item -mvr4130-align
10346 @itemx -mno-vr4130-align
10347 @opindex mvr4130-align
10348 The VR4130 pipeline is two-way superscalar, but can only issue two
10349 instructions together if the first one is 8-byte aligned. When this
10350 option is enabled, GCC will align pairs of instructions that it
10351 thinks should execute in parallel.
10353 This option only has an effect when optimizing for the VR4130.
10354 It normally makes code faster, but at the expense of making it bigger.
10355 It is enabled by default at optimization level @option{-O3}.
10359 @subsection MMIX Options
10360 @cindex MMIX Options
10362 These options are defined for the MMIX:
10366 @itemx -mno-libfuncs
10368 @opindex mno-libfuncs
10369 Specify that intrinsic library functions are being compiled, passing all
10370 values in registers, no matter the size.
10373 @itemx -mno-epsilon
10375 @opindex mno-epsilon
10376 Generate floating-point comparison instructions that compare with respect
10377 to the @code{rE} epsilon register.
10379 @item -mabi=mmixware
10381 @opindex mabi-mmixware
10383 Generate code that passes function parameters and return values that (in
10384 the called function) are seen as registers @code{$0} and up, as opposed to
10385 the GNU ABI which uses global registers @code{$231} and up.
10387 @item -mzero-extend
10388 @itemx -mno-zero-extend
10389 @opindex mzero-extend
10390 @opindex mno-zero-extend
10391 When reading data from memory in sizes shorter than 64 bits, use (do not
10392 use) zero-extending load instructions by default, rather than
10393 sign-extending ones.
10396 @itemx -mno-knuthdiv
10398 @opindex mno-knuthdiv
10399 Make the result of a division yielding a remainder have the same sign as
10400 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10401 remainder follows the sign of the dividend. Both methods are
10402 arithmetically valid, the latter being almost exclusively used.
10404 @item -mtoplevel-symbols
10405 @itemx -mno-toplevel-symbols
10406 @opindex mtoplevel-symbols
10407 @opindex mno-toplevel-symbols
10408 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10409 code can be used with the @code{PREFIX} assembly directive.
10413 Generate an executable in the ELF format, rather than the default
10414 @samp{mmo} format used by the @command{mmix} simulator.
10416 @item -mbranch-predict
10417 @itemx -mno-branch-predict
10418 @opindex mbranch-predict
10419 @opindex mno-branch-predict
10420 Use (do not use) the probable-branch instructions, when static branch
10421 prediction indicates a probable branch.
10423 @item -mbase-addresses
10424 @itemx -mno-base-addresses
10425 @opindex mbase-addresses
10426 @opindex mno-base-addresses
10427 Generate (do not generate) code that uses @emph{base addresses}. Using a
10428 base address automatically generates a request (handled by the assembler
10429 and the linker) for a constant to be set up in a global register. The
10430 register is used for one or more base address requests within the range 0
10431 to 255 from the value held in the register. The generally leads to short
10432 and fast code, but the number of different data items that can be
10433 addressed is limited. This means that a program that uses lots of static
10434 data may require @option{-mno-base-addresses}.
10436 @item -msingle-exit
10437 @itemx -mno-single-exit
10438 @opindex msingle-exit
10439 @opindex mno-single-exit
10440 Force (do not force) generated code to have a single exit point in each
10444 @node MN10300 Options
10445 @subsection MN10300 Options
10446 @cindex MN10300 options
10448 These @option{-m} options are defined for Matsushita MN10300 architectures:
10453 Generate code to avoid bugs in the multiply instructions for the MN10300
10454 processors. This is the default.
10456 @item -mno-mult-bug
10457 @opindex mno-mult-bug
10458 Do not generate code to avoid bugs in the multiply instructions for the
10459 MN10300 processors.
10463 Generate code which uses features specific to the AM33 processor.
10467 Do not generate code which uses features specific to the AM33 processor. This
10472 Do not link in the C run-time initialization object file.
10476 Indicate to the linker that it should perform a relaxation optimization pass
10477 to shorten branches, calls and absolute memory addresses. This option only
10478 has an effect when used on the command line for the final link step.
10480 This option makes symbolic debugging impossible.
10483 @node PDP-11 Options
10484 @subsection PDP-11 Options
10485 @cindex PDP-11 Options
10487 These options are defined for the PDP-11:
10492 Use hardware FPP floating point. This is the default. (FIS floating
10493 point on the PDP-11/40 is not supported.)
10496 @opindex msoft-float
10497 Do not use hardware floating point.
10501 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10505 Return floating-point results in memory. This is the default.
10509 Generate code for a PDP-11/40.
10513 Generate code for a PDP-11/45. This is the default.
10517 Generate code for a PDP-11/10.
10519 @item -mbcopy-builtin
10520 @opindex bcopy-builtin
10521 Use inline @code{movmemhi} patterns for copying memory. This is the
10526 Do not use inline @code{movmemhi} patterns for copying memory.
10532 Use 16-bit @code{int}. This is the default.
10538 Use 32-bit @code{int}.
10541 @itemx -mno-float32
10543 @opindex mno-float32
10544 Use 64-bit @code{float}. This is the default.
10547 @itemx -mno-float64
10549 @opindex mno-float64
10550 Use 32-bit @code{float}.
10554 Use @code{abshi2} pattern. This is the default.
10558 Do not use @code{abshi2} pattern.
10560 @item -mbranch-expensive
10561 @opindex mbranch-expensive
10562 Pretend that branches are expensive. This is for experimenting with
10563 code generation only.
10565 @item -mbranch-cheap
10566 @opindex mbranch-cheap
10567 Do not pretend that branches are expensive. This is the default.
10571 Generate code for a system with split I&D@.
10575 Generate code for a system without split I&D@. This is the default.
10579 Use Unix assembler syntax. This is the default when configured for
10580 @samp{pdp11-*-bsd}.
10584 Use DEC assembler syntax. This is the default when configured for any
10585 PDP-11 target other than @samp{pdp11-*-bsd}.
10588 @node PowerPC Options
10589 @subsection PowerPC Options
10590 @cindex PowerPC options
10592 These are listed under @xref{RS/6000 and PowerPC Options}.
10594 @node RS/6000 and PowerPC Options
10595 @subsection IBM RS/6000 and PowerPC Options
10596 @cindex RS/6000 and PowerPC Options
10597 @cindex IBM RS/6000 and PowerPC Options
10599 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
10606 @itemx -mno-powerpc
10607 @itemx -mpowerpc-gpopt
10608 @itemx -mno-powerpc-gpopt
10609 @itemx -mpowerpc-gfxopt
10610 @itemx -mno-powerpc-gfxopt
10612 @itemx -mno-powerpc64
10616 @opindex mno-power2
10618 @opindex mno-powerpc
10619 @opindex mpowerpc-gpopt
10620 @opindex mno-powerpc-gpopt
10621 @opindex mpowerpc-gfxopt
10622 @opindex mno-powerpc-gfxopt
10623 @opindex mpowerpc64
10624 @opindex mno-powerpc64
10625 GCC supports two related instruction set architectures for the
10626 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
10627 instructions supported by the @samp{rios} chip set used in the original
10628 RS/6000 systems and the @dfn{PowerPC} instruction set is the
10629 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
10630 the IBM 4xx microprocessors.
10632 Neither architecture is a subset of the other. However there is a
10633 large common subset of instructions supported by both. An MQ
10634 register is included in processors supporting the POWER architecture.
10636 You use these options to specify which instructions are available on the
10637 processor you are using. The default value of these options is
10638 determined when configuring GCC@. Specifying the
10639 @option{-mcpu=@var{cpu_type}} overrides the specification of these
10640 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
10641 rather than the options listed above.
10643 The @option{-mpower} option allows GCC to generate instructions that
10644 are found only in the POWER architecture and to use the MQ register.
10645 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
10646 to generate instructions that are present in the POWER2 architecture but
10647 not the original POWER architecture.
10649 The @option{-mpowerpc} option allows GCC to generate instructions that
10650 are found only in the 32-bit subset of the PowerPC architecture.
10651 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
10652 GCC to use the optional PowerPC architecture instructions in the
10653 General Purpose group, including floating-point square root. Specifying
10654 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
10655 use the optional PowerPC architecture instructions in the Graphics
10656 group, including floating-point select.
10658 The @option{-mpowerpc64} option allows GCC to generate the additional
10659 64-bit instructions that are found in the full PowerPC64 architecture
10660 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
10661 @option{-mno-powerpc64}.
10663 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
10664 will use only the instructions in the common subset of both
10665 architectures plus some special AIX common-mode calls, and will not use
10666 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
10667 permits GCC to use any instruction from either architecture and to
10668 allow use of the MQ register; specify this for the Motorola MPC601.
10670 @item -mnew-mnemonics
10671 @itemx -mold-mnemonics
10672 @opindex mnew-mnemonics
10673 @opindex mold-mnemonics
10674 Select which mnemonics to use in the generated assembler code. With
10675 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
10676 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
10677 assembler mnemonics defined for the POWER architecture. Instructions
10678 defined in only one architecture have only one mnemonic; GCC uses that
10679 mnemonic irrespective of which of these options is specified.
10681 GCC defaults to the mnemonics appropriate for the architecture in
10682 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
10683 value of these option. Unless you are building a cross-compiler, you
10684 should normally not specify either @option{-mnew-mnemonics} or
10685 @option{-mold-mnemonics}, but should instead accept the default.
10687 @item -mcpu=@var{cpu_type}
10689 Set architecture type, register usage, choice of mnemonics, and
10690 instruction scheduling parameters for machine type @var{cpu_type}.
10691 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
10692 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
10693 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
10694 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
10695 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
10696 @samp{860}, @samp{970}, @samp{8540}, @samp{common}, @samp{ec603e}, @samp{G3},
10697 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
10698 @samp{power4}, @samp{power5}, @samp{powerpc}, @samp{powerpc64},
10699 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
10701 @option{-mcpu=common} selects a completely generic processor. Code
10702 generated under this option will run on any POWER or PowerPC processor.
10703 GCC will use only the instructions in the common subset of both
10704 architectures, and will not use the MQ register. GCC assumes a generic
10705 processor model for scheduling purposes.
10707 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
10708 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
10709 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
10710 types, with an appropriate, generic processor model assumed for
10711 scheduling purposes.
10713 The other options specify a specific processor. Code generated under
10714 those options will run best on that processor, and may not run at all on
10717 The @option{-mcpu} options automatically enable or disable the
10718 following options: @option{-maltivec}, @option{-mhard-float},
10719 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
10720 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
10721 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
10722 @option{-mstring}. The particular options set for any particular CPU
10723 will vary between compiler versions, depending on what setting seems
10724 to produce optimal code for that CPU; it doesn't necessarily reflect
10725 the actual hardware's capabilities. If you wish to set an individual
10726 option to a particular value, you may specify it after the
10727 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
10729 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
10730 not enabled or disabled by the @option{-mcpu} option at present, since
10731 AIX does not have full support for these options. You may still
10732 enable or disable them individually if you're sure it'll work in your
10735 @item -mtune=@var{cpu_type}
10737 Set the instruction scheduling parameters for machine type
10738 @var{cpu_type}, but do not set the architecture type, register usage, or
10739 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
10740 values for @var{cpu_type} are used for @option{-mtune} as for
10741 @option{-mcpu}. If both are specified, the code generated will use the
10742 architecture, registers, and mnemonics set by @option{-mcpu}, but the
10743 scheduling parameters set by @option{-mtune}.
10749 Generate code to compute division as reciprocal estimate and iterative
10750 refinement, creating opportunities for increased throughput. This
10751 feature requires: optional PowerPC Graphics instruction set for single
10752 precision and FRE instruction for double precision, assuming divides
10753 cannot generate user-visible traps, and the domain values not include
10754 Infinities, denormals or zero denominator.
10757 @itemx -mno-altivec
10759 @opindex mno-altivec
10760 Generate code that uses (does not use) AltiVec instructions, and also
10761 enable the use of built-in functions that allow more direct access to
10762 the AltiVec instruction set. You may also need to set
10763 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
10769 @opindex mno-vrsave
10770 Generate VRSAVE instructions when generating AltiVec code.
10774 Extend the current ABI with SPE ABI extensions. This does not change
10775 the default ABI, instead it adds the SPE ABI extensions to the current
10779 @opindex mabi=no-spe
10780 Disable Booke SPE ABI extensions for the current ABI@.
10783 @opindex msecure-plt
10784 Generate code that allows ld and ld.so to build executables and shared
10785 libraries with non-exec .plt and .got sections. This is a PowerPC
10786 32-bit SYSV ABI option.
10790 Generate code that uses a BSS .plt section that ld.so fills in, and
10791 requires .plt and .got sections that are both writable and executable.
10792 This is a PowerPC 32-bit SYSV ABI option.
10798 This switch enables or disables the generation of ISEL instructions.
10800 @item -misel=@var{yes/no}
10801 This switch has been deprecated. Use @option{-misel} and
10802 @option{-mno-isel} instead.
10808 This switch enables or disables the generation of SPE simd
10811 @item -mspe=@var{yes/no}
10812 This option has been deprecated. Use @option{-mspe} and
10813 @option{-mno-spe} instead.
10815 @item -mfloat-gprs=@var{yes/single/double/no}
10816 @itemx -mfloat-gprs
10817 @opindex mfloat-gprs
10818 This switch enables or disables the generation of floating point
10819 operations on the general purpose registers for architectures that
10822 The argument @var{yes} or @var{single} enables the use of
10823 single-precision floating point operations.
10825 The argument @var{double} enables the use of single and
10826 double-precision floating point operations.
10828 The argument @var{no} disables floating point operations on the
10829 general purpose registers.
10831 This option is currently only available on the MPC854x.
10837 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
10838 targets (including GNU/Linux). The 32-bit environment sets int, long
10839 and pointer to 32 bits and generates code that runs on any PowerPC
10840 variant. The 64-bit environment sets int to 32 bits and long and
10841 pointer to 64 bits, and generates code for PowerPC64, as for
10842 @option{-mpowerpc64}.
10845 @itemx -mno-fp-in-toc
10846 @itemx -mno-sum-in-toc
10847 @itemx -mminimal-toc
10849 @opindex mno-fp-in-toc
10850 @opindex mno-sum-in-toc
10851 @opindex mminimal-toc
10852 Modify generation of the TOC (Table Of Contents), which is created for
10853 every executable file. The @option{-mfull-toc} option is selected by
10854 default. In that case, GCC will allocate at least one TOC entry for
10855 each unique non-automatic variable reference in your program. GCC
10856 will also place floating-point constants in the TOC@. However, only
10857 16,384 entries are available in the TOC@.
10859 If you receive a linker error message that saying you have overflowed
10860 the available TOC space, you can reduce the amount of TOC space used
10861 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
10862 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
10863 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
10864 generate code to calculate the sum of an address and a constant at
10865 run-time instead of putting that sum into the TOC@. You may specify one
10866 or both of these options. Each causes GCC to produce very slightly
10867 slower and larger code at the expense of conserving TOC space.
10869 If you still run out of space in the TOC even when you specify both of
10870 these options, specify @option{-mminimal-toc} instead. This option causes
10871 GCC to make only one TOC entry for every file. When you specify this
10872 option, GCC will produce code that is slower and larger but which
10873 uses extremely little TOC space. You may wish to use this option
10874 only on files that contain less frequently executed code.
10880 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
10881 @code{long} type, and the infrastructure needed to support them.
10882 Specifying @option{-maix64} implies @option{-mpowerpc64} and
10883 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
10884 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
10887 @itemx -mno-xl-compat
10888 @opindex mxl-compat
10889 @opindex mno-xl-compat
10890 Produce code that conforms more closely to IBM XLC semantics when using
10891 AIX-compatible ABI. Pass floating-point arguments to prototyped
10892 functions beyond the register save area (RSA) on the stack in addition
10893 to argument FPRs. Do not assume that most significant double in 128
10894 bit long double value is properly rounded when comparing values.
10896 The AIX calling convention was extended but not initially documented to
10897 handle an obscure K&R C case of calling a function that takes the
10898 address of its arguments with fewer arguments than declared. AIX XL
10899 compilers access floating point arguments which do not fit in the
10900 RSA from the stack when a subroutine is compiled without
10901 optimization. Because always storing floating-point arguments on the
10902 stack is inefficient and rarely needed, this option is not enabled by
10903 default and only is necessary when calling subroutines compiled by AIX
10904 XL compilers without optimization.
10908 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
10909 application written to use message passing with special startup code to
10910 enable the application to run. The system must have PE installed in the
10911 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
10912 must be overridden with the @option{-specs=} option to specify the
10913 appropriate directory location. The Parallel Environment does not
10914 support threads, so the @option{-mpe} option and the @option{-pthread}
10915 option are incompatible.
10917 @item -malign-natural
10918 @itemx -malign-power
10919 @opindex malign-natural
10920 @opindex malign-power
10921 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
10922 @option{-malign-natural} overrides the ABI-defined alignment of larger
10923 types, such as floating-point doubles, on their natural size-based boundary.
10924 The option @option{-malign-power} instructs GCC to follow the ABI-specified
10925 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
10927 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
10931 @itemx -mhard-float
10932 @opindex msoft-float
10933 @opindex mhard-float
10934 Generate code that does not use (uses) the floating-point register set.
10935 Software floating point emulation is provided if you use the
10936 @option{-msoft-float} option, and pass the option to GCC when linking.
10939 @itemx -mno-multiple
10941 @opindex mno-multiple
10942 Generate code that uses (does not use) the load multiple word
10943 instructions and the store multiple word instructions. These
10944 instructions are generated by default on POWER systems, and not
10945 generated on PowerPC systems. Do not use @option{-mmultiple} on little
10946 endian PowerPC systems, since those instructions do not work when the
10947 processor is in little endian mode. The exceptions are PPC740 and
10948 PPC750 which permit the instructions usage in little endian mode.
10953 @opindex mno-string
10954 Generate code that uses (does not use) the load string instructions
10955 and the store string word instructions to save multiple registers and
10956 do small block moves. These instructions are generated by default on
10957 POWER systems, and not generated on PowerPC systems. Do not use
10958 @option{-mstring} on little endian PowerPC systems, since those
10959 instructions do not work when the processor is in little endian mode.
10960 The exceptions are PPC740 and PPC750 which permit the instructions
10961 usage in little endian mode.
10966 @opindex mno-update
10967 Generate code that uses (does not use) the load or store instructions
10968 that update the base register to the address of the calculated memory
10969 location. These instructions are generated by default. If you use
10970 @option{-mno-update}, there is a small window between the time that the
10971 stack pointer is updated and the address of the previous frame is
10972 stored, which means code that walks the stack frame across interrupts or
10973 signals may get corrupted data.
10976 @itemx -mno-fused-madd
10977 @opindex mfused-madd
10978 @opindex mno-fused-madd
10979 Generate code that uses (does not use) the floating point multiply and
10980 accumulate instructions. These instructions are generated by default if
10981 hardware floating is used.
10983 @item -mno-bit-align
10985 @opindex mno-bit-align
10986 @opindex mbit-align
10987 On System V.4 and embedded PowerPC systems do not (do) force structures
10988 and unions that contain bit-fields to be aligned to the base type of the
10991 For example, by default a structure containing nothing but 8
10992 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
10993 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
10994 the structure would be aligned to a 1 byte boundary and be one byte in
10997 @item -mno-strict-align
10998 @itemx -mstrict-align
10999 @opindex mno-strict-align
11000 @opindex mstrict-align
11001 On System V.4 and embedded PowerPC systems do not (do) assume that
11002 unaligned memory references will be handled by the system.
11004 @item -mrelocatable
11005 @itemx -mno-relocatable
11006 @opindex mrelocatable
11007 @opindex mno-relocatable
11008 On embedded PowerPC systems generate code that allows (does not allow)
11009 the program to be relocated to a different address at runtime. If you
11010 use @option{-mrelocatable} on any module, all objects linked together must
11011 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
11013 @item -mrelocatable-lib
11014 @itemx -mno-relocatable-lib
11015 @opindex mrelocatable-lib
11016 @opindex mno-relocatable-lib
11017 On embedded PowerPC systems generate code that allows (does not allow)
11018 the program to be relocated to a different address at runtime. Modules
11019 compiled with @option{-mrelocatable-lib} can be linked with either modules
11020 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
11021 with modules compiled with the @option{-mrelocatable} options.
11027 On System V.4 and embedded PowerPC systems do not (do) assume that
11028 register 2 contains a pointer to a global area pointing to the addresses
11029 used in the program.
11032 @itemx -mlittle-endian
11034 @opindex mlittle-endian
11035 On System V.4 and embedded PowerPC systems compile code for the
11036 processor in little endian mode. The @option{-mlittle-endian} option is
11037 the same as @option{-mlittle}.
11040 @itemx -mbig-endian
11042 @opindex mbig-endian
11043 On System V.4 and embedded PowerPC systems compile code for the
11044 processor in big endian mode. The @option{-mbig-endian} option is
11045 the same as @option{-mbig}.
11047 @item -mdynamic-no-pic
11048 @opindex mdynamic-no-pic
11049 On Darwin and Mac OS X systems, compile code so that it is not
11050 relocatable, but that its external references are relocatable. The
11051 resulting code is suitable for applications, but not shared
11054 @item -mprioritize-restricted-insns=@var{priority}
11055 @opindex mprioritize-restricted-insns
11056 This option controls the priority that is assigned to
11057 dispatch-slot restricted instructions during the second scheduling
11058 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
11059 @var{no/highest/second-highest} priority to dispatch slot restricted
11062 @item -msched-costly-dep=@var{dependence_type}
11063 @opindex msched-costly-dep
11064 This option controls which dependences are considered costly
11065 by the target during instruction scheduling. The argument
11066 @var{dependence_type} takes one of the following values:
11067 @var{no}: no dependence is costly,
11068 @var{all}: all dependences are costly,
11069 @var{true_store_to_load}: a true dependence from store to load is costly,
11070 @var{store_to_load}: any dependence from store to load is costly,
11071 @var{number}: any dependence which latency >= @var{number} is costly.
11073 @item -minsert-sched-nops=@var{scheme}
11074 @opindex minsert-sched-nops
11075 This option controls which nop insertion scheme will be used during
11076 the second scheduling pass. The argument @var{scheme} takes one of the
11078 @var{no}: Don't insert nops.
11079 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
11080 according to the scheduler's grouping.
11081 @var{regroup_exact}: Insert nops to force costly dependent insns into
11082 separate groups. Insert exactly as many nops as needed to force an insn
11083 to a new group, according to the estimated processor grouping.
11084 @var{number}: Insert nops to force costly dependent insns into
11085 separate groups. Insert @var{number} nops to force an insn to a new group.
11088 @opindex mcall-sysv
11089 On System V.4 and embedded PowerPC systems compile code using calling
11090 conventions that adheres to the March 1995 draft of the System V
11091 Application Binary Interface, PowerPC processor supplement. This is the
11092 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
11094 @item -mcall-sysv-eabi
11095 @opindex mcall-sysv-eabi
11096 Specify both @option{-mcall-sysv} and @option{-meabi} options.
11098 @item -mcall-sysv-noeabi
11099 @opindex mcall-sysv-noeabi
11100 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
11102 @item -mcall-solaris
11103 @opindex mcall-solaris
11104 On System V.4 and embedded PowerPC systems compile code for the Solaris
11108 @opindex mcall-linux
11109 On System V.4 and embedded PowerPC systems compile code for the
11110 Linux-based GNU system.
11114 On System V.4 and embedded PowerPC systems compile code for the
11115 Hurd-based GNU system.
11117 @item -mcall-netbsd
11118 @opindex mcall-netbsd
11119 On System V.4 and embedded PowerPC systems compile code for the
11120 NetBSD operating system.
11122 @item -maix-struct-return
11123 @opindex maix-struct-return
11124 Return all structures in memory (as specified by the AIX ABI)@.
11126 @item -msvr4-struct-return
11127 @opindex msvr4-struct-return
11128 Return structures smaller than 8 bytes in registers (as specified by the
11131 @item -mabi=@var{abi-type}
11133 Extend the current ABI with a particular extension, or remove such extension.
11134 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
11138 @itemx -mno-prototype
11139 @opindex mprototype
11140 @opindex mno-prototype
11141 On System V.4 and embedded PowerPC systems assume that all calls to
11142 variable argument functions are properly prototyped. Otherwise, the
11143 compiler must insert an instruction before every non prototyped call to
11144 set or clear bit 6 of the condition code register (@var{CR}) to
11145 indicate whether floating point values were passed in the floating point
11146 registers in case the function takes a variable arguments. With
11147 @option{-mprototype}, only calls to prototyped variable argument functions
11148 will set or clear the bit.
11152 On embedded PowerPC systems, assume that the startup module is called
11153 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
11154 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
11159 On embedded PowerPC systems, assume that the startup module is called
11160 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
11165 On embedded PowerPC systems, assume that the startup module is called
11166 @file{crt0.o} and the standard C libraries are @file{libads.a} and
11169 @item -myellowknife
11170 @opindex myellowknife
11171 On embedded PowerPC systems, assume that the startup module is called
11172 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
11177 On System V.4 and embedded PowerPC systems, specify that you are
11178 compiling for a VxWorks system.
11182 Specify that you are compiling for the WindISS simulation environment.
11186 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
11187 header to indicate that @samp{eabi} extended relocations are used.
11193 On System V.4 and embedded PowerPC systems do (do not) adhere to the
11194 Embedded Applications Binary Interface (eabi) which is a set of
11195 modifications to the System V.4 specifications. Selecting @option{-meabi}
11196 means that the stack is aligned to an 8 byte boundary, a function
11197 @code{__eabi} is called to from @code{main} to set up the eabi
11198 environment, and the @option{-msdata} option can use both @code{r2} and
11199 @code{r13} to point to two separate small data areas. Selecting
11200 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
11201 do not call an initialization function from @code{main}, and the
11202 @option{-msdata} option will only use @code{r13} to point to a single
11203 small data area. The @option{-meabi} option is on by default if you
11204 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
11207 @opindex msdata=eabi
11208 On System V.4 and embedded PowerPC systems, put small initialized
11209 @code{const} global and static data in the @samp{.sdata2} section, which
11210 is pointed to by register @code{r2}. Put small initialized
11211 non-@code{const} global and static data in the @samp{.sdata} section,
11212 which is pointed to by register @code{r13}. Put small uninitialized
11213 global and static data in the @samp{.sbss} section, which is adjacent to
11214 the @samp{.sdata} section. The @option{-msdata=eabi} option is
11215 incompatible with the @option{-mrelocatable} option. The
11216 @option{-msdata=eabi} option also sets the @option{-memb} option.
11219 @opindex msdata=sysv
11220 On System V.4 and embedded PowerPC systems, put small global and static
11221 data in the @samp{.sdata} section, which is pointed to by register
11222 @code{r13}. Put small uninitialized global and static data in the
11223 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11224 The @option{-msdata=sysv} option is incompatible with the
11225 @option{-mrelocatable} option.
11227 @item -msdata=default
11229 @opindex msdata=default
11231 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
11232 compile code the same as @option{-msdata=eabi}, otherwise compile code the
11233 same as @option{-msdata=sysv}.
11236 @opindex msdata-data
11237 On System V.4 and embedded PowerPC systems, put small global and static
11238 data in the @samp{.sdata} section. Put small uninitialized global and
11239 static data in the @samp{.sbss} section. Do not use register @code{r13}
11240 to address small data however. This is the default behavior unless
11241 other @option{-msdata} options are used.
11245 @opindex msdata=none
11247 On embedded PowerPC systems, put all initialized global and static data
11248 in the @samp{.data} section, and all uninitialized data in the
11249 @samp{.bss} section.
11253 @cindex smaller data references (PowerPC)
11254 @cindex .sdata/.sdata2 references (PowerPC)
11255 On embedded PowerPC systems, put global and static items less than or
11256 equal to @var{num} bytes into the small data or bss sections instead of
11257 the normal data or bss section. By default, @var{num} is 8. The
11258 @option{-G @var{num}} switch is also passed to the linker.
11259 All modules should be compiled with the same @option{-G @var{num}} value.
11262 @itemx -mno-regnames
11264 @opindex mno-regnames
11265 On System V.4 and embedded PowerPC systems do (do not) emit register
11266 names in the assembly language output using symbolic forms.
11269 @itemx -mno-longcall
11271 @opindex mno-longcall
11272 Default to making all function calls indirectly, using a register, so
11273 that functions which reside further than 32 megabytes (33,554,432
11274 bytes) from the current location can be called. This setting can be
11275 overridden by the @code{shortcall} function attribute, or by
11276 @code{#pragma longcall(0)}.
11278 Some linkers are capable of detecting out-of-range calls and generating
11279 glue code on the fly. On these systems, long calls are unnecessary and
11280 generate slower code. As of this writing, the AIX linker can do this,
11281 as can the GNU linker for PowerPC/64. It is planned to add this feature
11282 to the GNU linker for 32-bit PowerPC systems as well.
11284 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11285 callee, L42'', plus a ``branch island'' (glue code). The two target
11286 addresses represent the callee and the ``branch island''. The
11287 Darwin/PPC linker will prefer the first address and generate a ``bl
11288 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11289 otherwise, the linker will generate ``bl L42'' to call the ``branch
11290 island''. The ``branch island'' is appended to the body of the
11291 calling function; it computes the full 32-bit address of the callee
11294 On Mach-O (Darwin) systems, this option directs the compiler emit to
11295 the glue for every direct call, and the Darwin linker decides whether
11296 to use or discard it.
11298 In the future, we may cause GCC to ignore all longcall specifications
11299 when the linker is known to generate glue.
11303 Adds support for multithreading with the @dfn{pthreads} library.
11304 This option sets flags for both the preprocessor and linker.
11308 @node S/390 and zSeries Options
11309 @subsection S/390 and zSeries Options
11310 @cindex S/390 and zSeries Options
11312 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11316 @itemx -msoft-float
11317 @opindex mhard-float
11318 @opindex msoft-float
11319 Use (do not use) the hardware floating-point instructions and registers
11320 for floating-point operations. When @option{-msoft-float} is specified,
11321 functions in @file{libgcc.a} will be used to perform floating-point
11322 operations. When @option{-mhard-float} is specified, the compiler
11323 generates IEEE floating-point instructions. This is the default.
11326 @itemx -mno-backchain
11327 @opindex mbackchain
11328 @opindex mno-backchain
11329 Store (do not store) the address of the caller's frame as backchain pointer
11330 into the callee's stack frame.
11331 A backchain may be needed to allow debugging using tools that do not understand
11332 DWARF-2 call frame information.
11333 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
11334 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
11335 the backchain is placed into the topmost word of the 96/160 byte register
11338 In general, code compiled with @option{-mbackchain} is call-compatible with
11339 code compiled with @option{-mmo-backchain}; however, use of the backchain
11340 for debugging purposes usually requires that the whole binary is built with
11341 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
11342 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11343 to build a linux kernel use @option{-msoft-float}.
11345 The default is to not maintain the backchain.
11347 @item -mpacked-stack
11348 @item -mno-packed-stack
11349 @opindex mpacked-stack
11350 @opindex mno-packed-stack
11351 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
11352 specified, the compiler uses the all fields of the 96/160 byte register save
11353 area only for their default purpose; unused fields still take up stack space.
11354 When @option{-mpacked-stack} is specified, register save slots are densely
11355 packed at the top of the register save area; unused space is reused for other
11356 purposes, allowing for more efficient use of the available stack space.
11357 However, when @option{-mbackchain} is also in effect, the topmost word of
11358 the save area is always used to store the backchain, and the return address
11359 register is always saved two words below the backchain.
11361 As long as the stack frame backchain is not used, code generated with
11362 @option{-mpacked-stack} is call-compatible with code generated with
11363 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
11364 S/390 or zSeries generated code that uses the stack frame backchain at run
11365 time, not just for debugging purposes. Such code is not call-compatible
11366 with code compiled with @option{-mpacked-stack}. Also, note that the
11367 combination of @option{-mbackchain},
11368 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11369 to build a linux kernel use @option{-msoft-float}.
11371 The default is to not use the packed stack layout.
11374 @itemx -mno-small-exec
11375 @opindex msmall-exec
11376 @opindex mno-small-exec
11377 Generate (or do not generate) code using the @code{bras} instruction
11378 to do subroutine calls.
11379 This only works reliably if the total executable size does not
11380 exceed 64k. The default is to use the @code{basr} instruction instead,
11381 which does not have this limitation.
11387 When @option{-m31} is specified, generate code compliant to the
11388 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
11389 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
11390 particular to generate 64-bit instructions. For the @samp{s390}
11391 targets, the default is @option{-m31}, while the @samp{s390x}
11392 targets default to @option{-m64}.
11398 When @option{-mzarch} is specified, generate code using the
11399 instructions available on z/Architecture.
11400 When @option{-mesa} is specified, generate code using the
11401 instructions available on ESA/390. Note that @option{-mesa} is
11402 not possible with @option{-m64}.
11403 When generating code compliant to the GNU/Linux for S/390 ABI,
11404 the default is @option{-mesa}. When generating code compliant
11405 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
11411 Generate (or do not generate) code using the @code{mvcle} instruction
11412 to perform block moves. When @option{-mno-mvcle} is specified,
11413 use a @code{mvc} loop instead. This is the default unless optimizing for
11420 Print (or do not print) additional debug information when compiling.
11421 The default is to not print debug information.
11423 @item -march=@var{cpu-type}
11425 Generate code that will run on @var{cpu-type}, which is the name of a system
11426 representing a certain processor type. Possible values for
11427 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
11428 When generating code using the instructions available on z/Architecture,
11429 the default is @option{-march=z900}. Otherwise, the default is
11430 @option{-march=g5}.
11432 @item -mtune=@var{cpu-type}
11434 Tune to @var{cpu-type} everything applicable about the generated code,
11435 except for the ABI and the set of available instructions.
11436 The list of @var{cpu-type} values is the same as for @option{-march}.
11437 The default is the value used for @option{-march}.
11440 @itemx -mno-tpf-trace
11441 @opindex mtpf-trace
11442 @opindex mno-tpf-trace
11443 Generate code that adds (does not add) in TPF OS specific branches to trace
11444 routines in the operating system. This option is off by default, even
11445 when compiling for the TPF OS@.
11448 @itemx -mno-fused-madd
11449 @opindex mfused-madd
11450 @opindex mno-fused-madd
11451 Generate code that uses (does not use) the floating point multiply and
11452 accumulate instructions. These instructions are generated by default if
11453 hardware floating point is used.
11455 @item -mwarn-framesize=@var{framesize}
11456 @opindex mwarn-framesize
11457 Emit a warning if the current function exceeds the given frame size. Because
11458 this is a compile time check it doesn't need to be a real problem when the program
11459 runs. It is intended to identify functions which most probably cause
11460 a stack overflow. It is useful to be used in an environment with limited stack
11461 size e.g.@: the linux kernel.
11463 @item -mwarn-dynamicstack
11464 @opindex mwarn-dynamicstack
11465 Emit a warning if the function calls alloca or uses dynamically
11466 sized arrays. This is generally a bad idea with a limited stack size.
11468 @item -mstack-guard=@var{stack-guard}
11469 @item -mstack-size=@var{stack-size}
11470 @opindex mstack-guard
11471 @opindex mstack-size
11472 These arguments always have to be used in conjunction. If they are present the s390
11473 back end emits additional instructions in the function prologue which trigger a trap
11474 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
11475 (remember that the stack on s390 grows downward). These options are intended to
11476 be used to help debugging stack overflow problems. The additionally emitted code
11477 cause only little overhead and hence can also be used in production like systems
11478 without greater performance degradation. The given values have to be exact
11479 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard}.
11480 In order to be efficient the extra code makes the assumption that the stack starts
11481 at an address aligned to the value given by @var{stack-size}.
11485 @subsection SH Options
11487 These @samp{-m} options are defined for the SH implementations:
11492 Generate code for the SH1.
11496 Generate code for the SH2.
11499 Generate code for the SH2e.
11503 Generate code for the SH3.
11507 Generate code for the SH3e.
11511 Generate code for the SH4 without a floating-point unit.
11513 @item -m4-single-only
11514 @opindex m4-single-only
11515 Generate code for the SH4 with a floating-point unit that only
11516 supports single-precision arithmetic.
11520 Generate code for the SH4 assuming the floating-point unit is in
11521 single-precision mode by default.
11525 Generate code for the SH4.
11529 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
11530 floating-point unit is not used.
11532 @item -m4a-single-only
11533 @opindex m4a-single-only
11534 Generate code for the SH4a, in such a way that no double-precision
11535 floating point operations are used.
11538 @opindex m4a-single
11539 Generate code for the SH4a assuming the floating-point unit is in
11540 single-precision mode by default.
11544 Generate code for the SH4a.
11548 Same as @option{-m4a-nofpu}, except that it implicitly passes
11549 @option{-dsp} to the assembler. GCC doesn't generate any DSP
11550 instructions at the moment.
11554 Compile code for the processor in big endian mode.
11558 Compile code for the processor in little endian mode.
11562 Align doubles at 64-bit boundaries. Note that this changes the calling
11563 conventions, and thus some functions from the standard C library will
11564 not work unless you recompile it first with @option{-mdalign}.
11568 Shorten some address references at link time, when possible; uses the
11569 linker option @option{-relax}.
11573 Use 32-bit offsets in @code{switch} tables. The default is to use
11578 Enable the use of the instruction @code{fmovd}.
11582 Comply with the calling conventions defined by Renesas.
11586 Comply with the calling conventions defined by Renesas.
11590 Comply with the calling conventions defined for GCC before the Renesas
11591 conventions were available. This option is the default for all
11592 targets of the SH toolchain except for @samp{sh-symbianelf}.
11595 @opindex mnomacsave
11596 Mark the @code{MAC} register as call-clobbered, even if
11597 @option{-mhitachi} is given.
11601 Increase IEEE-compliance of floating-point code.
11602 At the moment, this is equivalent to @option{-fno-finite-math-only}.
11603 When generating 16 bit SH opcodes, getting IEEE-conforming results for
11604 comparisons of NANs / infinities incurs extra overhead in every
11605 floating point comparison, therefore the default is set to
11606 @option{-ffinite-math-only}.
11610 Dump instruction size and location in the assembly code.
11613 @opindex mpadstruct
11614 This option is deprecated. It pads structures to multiple of 4 bytes,
11615 which is incompatible with the SH ABI@.
11619 Optimize for space instead of speed. Implied by @option{-Os}.
11622 @opindex mprefergot
11623 When generating position-independent code, emit function calls using
11624 the Global Offset Table instead of the Procedure Linkage Table.
11628 Generate a library function call to invalidate instruction cache
11629 entries, after fixing up a trampoline. This library function call
11630 doesn't assume it can write to the whole memory address space. This
11631 is the default when the target is @code{sh-*-linux*}.
11633 @item -multcost=@var{number}
11634 @opindex multcost=@var{number}
11635 Set the cost to assume for a multiply insn.
11637 @item -mdiv=@var{strategy}
11638 @opindex mdiv=@var{strategy}
11639 Set the division strategy to use for SHmedia code. @var{strategy} must be
11640 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
11641 inv:call2, inv:fp .
11642 "fp" performs the operation in floating point. This has a very high latency,
11643 but needs only a few instructions, so it might be a good choice if
11644 your code has enough easily exploitable ILP to allow the compiler to
11645 schedule the floating point instructions together with other instructions.
11646 Division by zero causes a floating point exception.
11647 "inv" uses integer operations to calculate the inverse of the divisor,
11648 and then multiplies the dividend with the inverse. This strategy allows
11649 cse and hoisting of the inverse calculation. Division by zero calculates
11650 an unspecified result, but does not trap.
11651 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
11652 have been found, or if the entire operation has been hoisted to the same
11653 place, the last stages of the inverse calculation are intertwined with the
11654 final multiply to reduce the overall latency, at the expense of using a few
11655 more instructions, and thus offering fewer scheduling opportunities with
11657 "call" calls a library function that usually implements the inv:minlat
11659 This gives high code density for m5-*media-nofpu compilations.
11660 "call2" uses a different entry point of the same library function, where it
11661 assumes that a pointer to a lookup table has already been set up, which
11662 exposes the pointer load to cse / code hoisting optimizations.
11663 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
11664 code generation, but if the code stays unoptimized, revert to the "call",
11665 "call2", or "fp" strategies, respectively. Note that the
11666 potentially-trapping side effect of division by zero is carried by a
11667 separate instruction, so it is possible that all the integer instructions
11668 are hoisted out, but the marker for the side effect stays where it is.
11669 A recombination to fp operations or a call is not possible in that case.
11670 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
11671 that the inverse calculation was nor separated from the multiply, they speed
11672 up division where the dividend fits into 20 bits (plus sign where applicable),
11673 by inserting a test to skip a number of operations in this case; this test
11674 slows down the case of larger dividends. inv20u assumes the case of a such
11675 a small dividend to be unlikely, and inv20l assumes it to be likely.
11677 @item -mdivsi3_libfunc=@var{name}
11678 @opindex mdivsi3_libfunc=@var{name}
11679 Set the name of the library function used for 32 bit signed division to
11680 @var{name}. This only affect the name used in the call and inv:call
11681 division strategies, and the compiler will still expect the same
11682 sets of input/output/clobbered registers as if this option was not present.
11684 @item -madjust-unroll
11685 @opindex madjust-unroll
11686 Throttle unrolling to avoid thrashing target registers.
11687 This option only has an effect if the gcc code base supports the
11688 TARGET_ADJUST_UNROLL_MAX target hook.
11690 @item -mindexed-addressing
11691 @opindex mindexed-addressing
11692 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
11693 This is only safe if the hardware and/or OS implement 32 bit wrap-around
11694 semantics for the indexed addressing mode. The architecture allows the
11695 implementation of processors with 64 bit MMU, which the OS could use to
11696 get 32 bit addressing, but since no current hardware implementation supports
11697 this or any other way to make the indexed addressing mode safe to use in
11698 the 32 bit ABI, the default is -mno-indexed-addressing.
11700 @item -mgettrcost=@var{number}
11701 @opindex mgettrcost=@var{number}
11702 Set the cost assumed for the gettr instruction to @var{number}.
11703 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
11707 Assume pt* instructions won't trap. This will generally generate better
11708 scheduled code, but is unsafe on current hardware. The current architecture
11709 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
11710 This has the unintentional effect of making it unsafe to schedule ptabs /
11711 ptrel before a branch, or hoist it out of a loop. For example,
11712 __do_global_ctors, a part of libgcc that runs constructors at program
11713 startup, calls functions in a list which is delimited by -1. With the
11714 -mpt-fixed option, the ptabs will be done before testing against -1.
11715 That means that all the constructors will be run a bit quicker, but when
11716 the loop comes to the end of the list, the program crashes because ptabs
11717 loads -1 into a target register. Since this option is unsafe for any
11718 hardware implementing the current architecture specification, the default
11719 is -mno-pt-fixed. Unless the user specifies a specific cost with
11720 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
11721 this deters register allocation using target registers for storing
11724 @item -minvalid-symbols
11725 @opindex minvalid-symbols
11726 Assume symbols might be invalid. Ordinary function symbols generated by
11727 the compiler will always be valid to load with movi/shori/ptabs or
11728 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
11729 to generate symbols that will cause ptabs / ptrel to trap.
11730 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
11731 It will then prevent cross-basic-block cse, hoisting and most scheduling
11732 of symbol loads. The default is @option{-mno-invalid-symbols}.
11735 @node SPARC Options
11736 @subsection SPARC Options
11737 @cindex SPARC options
11739 These @samp{-m} options are supported on the SPARC:
11742 @item -mno-app-regs
11744 @opindex mno-app-regs
11746 Specify @option{-mapp-regs} to generate output using the global registers
11747 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
11750 To be fully SVR4 ABI compliant at the cost of some performance loss,
11751 specify @option{-mno-app-regs}. You should compile libraries and system
11752 software with this option.
11755 @itemx -mhard-float
11757 @opindex mhard-float
11758 Generate output containing floating point instructions. This is the
11762 @itemx -msoft-float
11764 @opindex msoft-float
11765 Generate output containing library calls for floating point.
11766 @strong{Warning:} the requisite libraries are not available for all SPARC
11767 targets. Normally the facilities of the machine's usual C compiler are
11768 used, but this cannot be done directly in cross-compilation. You must make
11769 your own arrangements to provide suitable library functions for
11770 cross-compilation. The embedded targets @samp{sparc-*-aout} and
11771 @samp{sparclite-*-*} do provide software floating point support.
11773 @option{-msoft-float} changes the calling convention in the output file;
11774 therefore, it is only useful if you compile @emph{all} of a program with
11775 this option. In particular, you need to compile @file{libgcc.a}, the
11776 library that comes with GCC, with @option{-msoft-float} in order for
11779 @item -mhard-quad-float
11780 @opindex mhard-quad-float
11781 Generate output containing quad-word (long double) floating point
11784 @item -msoft-quad-float
11785 @opindex msoft-quad-float
11786 Generate output containing library calls for quad-word (long double)
11787 floating point instructions. The functions called are those specified
11788 in the SPARC ABI@. This is the default.
11790 As of this writing, there are no SPARC implementations that have hardware
11791 support for the quad-word floating point instructions. They all invoke
11792 a trap handler for one of these instructions, and then the trap handler
11793 emulates the effect of the instruction. Because of the trap handler overhead,
11794 this is much slower than calling the ABI library routines. Thus the
11795 @option{-msoft-quad-float} option is the default.
11797 @item -mno-unaligned-doubles
11798 @itemx -munaligned-doubles
11799 @opindex mno-unaligned-doubles
11800 @opindex munaligned-doubles
11801 Assume that doubles have 8 byte alignment. This is the default.
11803 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
11804 alignment only if they are contained in another type, or if they have an
11805 absolute address. Otherwise, it assumes they have 4 byte alignment.
11806 Specifying this option avoids some rare compatibility problems with code
11807 generated by other compilers. It is not the default because it results
11808 in a performance loss, especially for floating point code.
11810 @item -mno-faster-structs
11811 @itemx -mfaster-structs
11812 @opindex mno-faster-structs
11813 @opindex mfaster-structs
11814 With @option{-mfaster-structs}, the compiler assumes that structures
11815 should have 8 byte alignment. This enables the use of pairs of
11816 @code{ldd} and @code{std} instructions for copies in structure
11817 assignment, in place of twice as many @code{ld} and @code{st} pairs.
11818 However, the use of this changed alignment directly violates the SPARC
11819 ABI@. Thus, it's intended only for use on targets where the developer
11820 acknowledges that their resulting code will not be directly in line with
11821 the rules of the ABI@.
11823 @item -mimpure-text
11824 @opindex mimpure-text
11825 @option{-mimpure-text}, used in addition to @option{-shared}, tells
11826 the compiler to not pass @option{-z text} to the linker when linking a
11827 shared object. Using this option, you can link position-dependent
11828 code into a shared object.
11830 @option{-mimpure-text} suppresses the ``relocations remain against
11831 allocatable but non-writable sections'' linker error message.
11832 However, the necessary relocations will trigger copy-on-write, and the
11833 shared object is not actually shared across processes. Instead of
11834 using @option{-mimpure-text}, you should compile all source code with
11835 @option{-fpic} or @option{-fPIC}.
11837 This option is only available on SunOS and Solaris.
11839 @item -mcpu=@var{cpu_type}
11841 Set the instruction set, register set, and instruction scheduling parameters
11842 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
11843 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
11844 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
11845 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
11846 @samp{ultrasparc3}.
11848 Default instruction scheduling parameters are used for values that select
11849 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
11850 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
11852 Here is a list of each supported architecture and their supported
11857 v8: supersparc, hypersparc
11858 sparclite: f930, f934, sparclite86x
11860 v9: ultrasparc, ultrasparc3
11863 By default (unless configured otherwise), GCC generates code for the V7
11864 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
11865 additionally optimizes it for the Cypress CY7C602 chip, as used in the
11866 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
11867 SPARCStation 1, 2, IPX etc.
11869 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
11870 architecture. The only difference from V7 code is that the compiler emits
11871 the integer multiply and integer divide instructions which exist in SPARC-V8
11872 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
11873 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
11876 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
11877 the SPARC architecture. This adds the integer multiply, integer divide step
11878 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
11879 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
11880 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
11881 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
11882 MB86934 chip, which is the more recent SPARClite with FPU@.
11884 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
11885 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
11886 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
11887 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
11888 optimizes it for the TEMIC SPARClet chip.
11890 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
11891 architecture. This adds 64-bit integer and floating-point move instructions,
11892 3 additional floating-point condition code registers and conditional move
11893 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
11894 optimizes it for the Sun UltraSPARC I/II chips. With
11895 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
11896 Sun UltraSPARC III chip.
11898 @item -mtune=@var{cpu_type}
11900 Set the instruction scheduling parameters for machine type
11901 @var{cpu_type}, but do not set the instruction set or register set that the
11902 option @option{-mcpu=@var{cpu_type}} would.
11904 The same values for @option{-mcpu=@var{cpu_type}} can be used for
11905 @option{-mtune=@var{cpu_type}}, but the only useful values are those
11906 that select a particular cpu implementation. Those are @samp{cypress},
11907 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
11908 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
11909 @samp{ultrasparc3}.
11914 @opindex mno-v8plus
11915 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
11916 difference from the V8 ABI is that the global and out registers are
11917 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
11918 mode for all SPARC-V9 processors.
11924 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
11925 Visual Instruction Set extensions. The default is @option{-mno-vis}.
11928 These @samp{-m} options are supported in addition to the above
11929 on SPARC-V9 processors in 64-bit environments:
11932 @item -mlittle-endian
11933 @opindex mlittle-endian
11934 Generate code for a processor running in little-endian mode. It is only
11935 available for a few configurations and most notably not on Solaris and Linux.
11941 Generate code for a 32-bit or 64-bit environment.
11942 The 32-bit environment sets int, long and pointer to 32 bits.
11943 The 64-bit environment sets int to 32 bits and long and pointer
11946 @item -mcmodel=medlow
11947 @opindex mcmodel=medlow
11948 Generate code for the Medium/Low code model: 64-bit addresses, programs
11949 must be linked in the low 32 bits of memory. Programs can be statically
11950 or dynamically linked.
11952 @item -mcmodel=medmid
11953 @opindex mcmodel=medmid
11954 Generate code for the Medium/Middle code model: 64-bit addresses, programs
11955 must be linked in the low 44 bits of memory, the text and data segments must
11956 be less than 2GB in size and the data segment must be located within 2GB of
11959 @item -mcmodel=medany
11960 @opindex mcmodel=medany
11961 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
11962 may be linked anywhere in memory, the text and data segments must be less
11963 than 2GB in size and the data segment must be located within 2GB of the
11966 @item -mcmodel=embmedany
11967 @opindex mcmodel=embmedany
11968 Generate code for the Medium/Anywhere code model for embedded systems:
11969 64-bit addresses, the text and data segments must be less than 2GB in
11970 size, both starting anywhere in memory (determined at link time). The
11971 global register %g4 points to the base of the data segment. Programs
11972 are statically linked and PIC is not supported.
11975 @itemx -mno-stack-bias
11976 @opindex mstack-bias
11977 @opindex mno-stack-bias
11978 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
11979 frame pointer if present, are offset by @minus{}2047 which must be added back
11980 when making stack frame references. This is the default in 64-bit mode.
11981 Otherwise, assume no such offset is present.
11984 These switches are supported in addition to the above on Solaris:
11989 Add support for multithreading using the Solaris threads library. This
11990 option sets flags for both the preprocessor and linker. This option does
11991 not affect the thread safety of object code produced by the compiler or
11992 that of libraries supplied with it.
11996 Add support for multithreading using the POSIX threads library. This
11997 option sets flags for both the preprocessor and linker. This option does
11998 not affect the thread safety of object code produced by the compiler or
11999 that of libraries supplied with it.
12002 @node System V Options
12003 @subsection Options for System V
12005 These additional options are available on System V Release 4 for
12006 compatibility with other compilers on those systems:
12011 Create a shared object.
12012 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
12016 Identify the versions of each tool used by the compiler, in a
12017 @code{.ident} assembler directive in the output.
12021 Refrain from adding @code{.ident} directives to the output file (this is
12024 @item -YP,@var{dirs}
12026 Search the directories @var{dirs}, and no others, for libraries
12027 specified with @option{-l}.
12029 @item -Ym,@var{dir}
12031 Look in the directory @var{dir} to find the M4 preprocessor.
12032 The assembler uses this option.
12033 @c This is supposed to go with a -Yd for predefined M4 macro files, but
12034 @c the generic assembler that comes with Solaris takes just -Ym.
12037 @node TMS320C3x/C4x Options
12038 @subsection TMS320C3x/C4x Options
12039 @cindex TMS320C3x/C4x Options
12041 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
12045 @item -mcpu=@var{cpu_type}
12047 Set the instruction set, register set, and instruction scheduling
12048 parameters for machine type @var{cpu_type}. Supported values for
12049 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
12050 @samp{c44}. The default is @samp{c40} to generate code for the
12055 @itemx -msmall-memory
12057 @opindex mbig-memory
12059 @opindex msmall-memory
12061 Generates code for the big or small memory model. The small memory
12062 model assumed that all data fits into one 64K word page. At run-time
12063 the data page (DP) register must be set to point to the 64K page
12064 containing the .bss and .data program sections. The big memory model is
12065 the default and requires reloading of the DP register for every direct
12072 Allow (disallow) allocation of general integer operands into the block
12073 count register BK@.
12079 Enable (disable) generation of code using decrement and branch,
12080 DBcond(D), instructions. This is enabled by default for the C4x. To be
12081 on the safe side, this is disabled for the C3x, since the maximum
12082 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
12083 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
12084 that it can utilize the decrement and branch instruction, but will give
12085 up if there is more than one memory reference in the loop. Thus a loop
12086 where the loop counter is decremented can generate slightly more
12087 efficient code, in cases where the RPTB instruction cannot be utilized.
12089 @item -mdp-isr-reload
12091 @opindex mdp-isr-reload
12093 Force the DP register to be saved on entry to an interrupt service
12094 routine (ISR), reloaded to point to the data section, and restored on
12095 exit from the ISR@. This should not be required unless someone has
12096 violated the small memory model by modifying the DP register, say within
12103 For the C3x use the 24-bit MPYI instruction for integer multiplies
12104 instead of a library call to guarantee 32-bit results. Note that if one
12105 of the operands is a constant, then the multiplication will be performed
12106 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
12107 then squaring operations are performed inline instead of a library call.
12110 @itemx -mno-fast-fix
12112 @opindex mno-fast-fix
12113 The C3x/C4x FIX instruction to convert a floating point value to an
12114 integer value chooses the nearest integer less than or equal to the
12115 floating point value rather than to the nearest integer. Thus if the
12116 floating point number is negative, the result will be incorrectly
12117 truncated an additional code is necessary to detect and correct this
12118 case. This option can be used to disable generation of the additional
12119 code required to correct the result.
12125 Enable (disable) generation of repeat block sequences using the RPTB
12126 instruction for zero overhead looping. The RPTB construct is only used
12127 for innermost loops that do not call functions or jump across the loop
12128 boundaries. There is no advantage having nested RPTB loops due to the
12129 overhead required to save and restore the RC, RS, and RE registers.
12130 This is enabled by default with @option{-O2}.
12132 @item -mrpts=@var{count}
12136 Enable (disable) the use of the single instruction repeat instruction
12137 RPTS@. If a repeat block contains a single instruction, and the loop
12138 count can be guaranteed to be less than the value @var{count}, GCC will
12139 emit a RPTS instruction instead of a RPTB@. If no value is specified,
12140 then a RPTS will be emitted even if the loop count cannot be determined
12141 at compile time. Note that the repeated instruction following RPTS does
12142 not have to be reloaded from memory each iteration, thus freeing up the
12143 CPU buses for operands. However, since interrupts are blocked by this
12144 instruction, it is disabled by default.
12146 @item -mloop-unsigned
12147 @itemx -mno-loop-unsigned
12148 @opindex mloop-unsigned
12149 @opindex mno-loop-unsigned
12150 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
12151 is @math{2^{31} + 1} since these instructions test if the iteration count is
12152 negative to terminate the loop. If the iteration count is unsigned
12153 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
12154 exceeded. This switch allows an unsigned iteration count.
12158 Try to emit an assembler syntax that the TI assembler (asm30) is happy
12159 with. This also enforces compatibility with the API employed by the TI
12160 C3x C compiler. For example, long doubles are passed as structures
12161 rather than in floating point registers.
12167 Generate code that uses registers (stack) for passing arguments to functions.
12168 By default, arguments are passed in registers where possible rather
12169 than by pushing arguments on to the stack.
12171 @item -mparallel-insns
12172 @itemx -mno-parallel-insns
12173 @opindex mparallel-insns
12174 @opindex mno-parallel-insns
12175 Allow the generation of parallel instructions. This is enabled by
12176 default with @option{-O2}.
12178 @item -mparallel-mpy
12179 @itemx -mno-parallel-mpy
12180 @opindex mparallel-mpy
12181 @opindex mno-parallel-mpy
12182 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
12183 provided @option{-mparallel-insns} is also specified. These instructions have
12184 tight register constraints which can pessimize the code generation
12185 of large functions.
12190 @subsection V850 Options
12191 @cindex V850 Options
12193 These @samp{-m} options are defined for V850 implementations:
12197 @itemx -mno-long-calls
12198 @opindex mlong-calls
12199 @opindex mno-long-calls
12200 Treat all calls as being far away (near). If calls are assumed to be
12201 far away, the compiler will always load the functions address up into a
12202 register, and call indirect through the pointer.
12208 Do not optimize (do optimize) basic blocks that use the same index
12209 pointer 4 or more times to copy pointer into the @code{ep} register, and
12210 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
12211 option is on by default if you optimize.
12213 @item -mno-prolog-function
12214 @itemx -mprolog-function
12215 @opindex mno-prolog-function
12216 @opindex mprolog-function
12217 Do not use (do use) external functions to save and restore registers
12218 at the prologue and epilogue of a function. The external functions
12219 are slower, but use less code space if more than one function saves
12220 the same number of registers. The @option{-mprolog-function} option
12221 is on by default if you optimize.
12225 Try to make the code as small as possible. At present, this just turns
12226 on the @option{-mep} and @option{-mprolog-function} options.
12228 @item -mtda=@var{n}
12230 Put static or global variables whose size is @var{n} bytes or less into
12231 the tiny data area that register @code{ep} points to. The tiny data
12232 area can hold up to 256 bytes in total (128 bytes for byte references).
12234 @item -msda=@var{n}
12236 Put static or global variables whose size is @var{n} bytes or less into
12237 the small data area that register @code{gp} points to. The small data
12238 area can hold up to 64 kilobytes.
12240 @item -mzda=@var{n}
12242 Put static or global variables whose size is @var{n} bytes or less into
12243 the first 32 kilobytes of memory.
12247 Specify that the target processor is the V850.
12250 @opindex mbig-switch
12251 Generate code suitable for big switch tables. Use this option only if
12252 the assembler/linker complain about out of range branches within a switch
12257 This option will cause r2 and r5 to be used in the code generated by
12258 the compiler. This setting is the default.
12260 @item -mno-app-regs
12261 @opindex mno-app-regs
12262 This option will cause r2 and r5 to be treated as fixed registers.
12266 Specify that the target processor is the V850E1. The preprocessor
12267 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
12268 this option is used.
12272 Specify that the target processor is the V850E@. The preprocessor
12273 constant @samp{__v850e__} will be defined if this option is used.
12275 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
12276 are defined then a default target processor will be chosen and the
12277 relevant @samp{__v850*__} preprocessor constant will be defined.
12279 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
12280 defined, regardless of which processor variant is the target.
12282 @item -mdisable-callt
12283 @opindex mdisable-callt
12284 This option will suppress generation of the CALLT instruction for the
12285 v850e and v850e1 flavors of the v850 architecture. The default is
12286 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
12291 @subsection VAX Options
12292 @cindex VAX options
12294 These @samp{-m} options are defined for the VAX:
12299 Do not output certain jump instructions (@code{aobleq} and so on)
12300 that the Unix assembler for the VAX cannot handle across long
12305 Do output those jump instructions, on the assumption that you
12306 will assemble with the GNU assembler.
12310 Output code for g-format floating point numbers instead of d-format.
12313 @node x86-64 Options
12314 @subsection x86-64 Options
12315 @cindex x86-64 options
12317 These are listed under @xref{i386 and x86-64 Options}.
12319 @node Xstormy16 Options
12320 @subsection Xstormy16 Options
12321 @cindex Xstormy16 Options
12323 These options are defined for Xstormy16:
12328 Choose startup files and linker script suitable for the simulator.
12331 @node Xtensa Options
12332 @subsection Xtensa Options
12333 @cindex Xtensa Options
12335 These options are supported for Xtensa targets:
12339 @itemx -mno-const16
12341 @opindex mno-const16
12342 Enable or disable use of @code{CONST16} instructions for loading
12343 constant values. The @code{CONST16} instruction is currently not a
12344 standard option from Tensilica. When enabled, @code{CONST16}
12345 instructions are always used in place of the standard @code{L32R}
12346 instructions. The use of @code{CONST16} is enabled by default only if
12347 the @code{L32R} instruction is not available.
12350 @itemx -mno-fused-madd
12351 @opindex mfused-madd
12352 @opindex mno-fused-madd
12353 Enable or disable use of fused multiply/add and multiply/subtract
12354 instructions in the floating-point option. This has no effect if the
12355 floating-point option is not also enabled. Disabling fused multiply/add
12356 and multiply/subtract instructions forces the compiler to use separate
12357 instructions for the multiply and add/subtract operations. This may be
12358 desirable in some cases where strict IEEE 754-compliant results are
12359 required: the fused multiply add/subtract instructions do not round the
12360 intermediate result, thereby producing results with @emph{more} bits of
12361 precision than specified by the IEEE standard. Disabling fused multiply
12362 add/subtract instructions also ensures that the program output is not
12363 sensitive to the compiler's ability to combine multiply and add/subtract
12366 @item -mtext-section-literals
12367 @itemx -mno-text-section-literals
12368 @opindex mtext-section-literals
12369 @opindex mno-text-section-literals
12370 Control the treatment of literal pools. The default is
12371 @option{-mno-text-section-literals}, which places literals in a separate
12372 section in the output file. This allows the literal pool to be placed
12373 in a data RAM/ROM, and it also allows the linker to combine literal
12374 pools from separate object files to remove redundant literals and
12375 improve code size. With @option{-mtext-section-literals}, the literals
12376 are interspersed in the text section in order to keep them as close as
12377 possible to their references. This may be necessary for large assembly
12380 @item -mtarget-align
12381 @itemx -mno-target-align
12382 @opindex mtarget-align
12383 @opindex mno-target-align
12384 When this option is enabled, GCC instructs the assembler to
12385 automatically align instructions to reduce branch penalties at the
12386 expense of some code density. The assembler attempts to widen density
12387 instructions to align branch targets and the instructions following call
12388 instructions. If there are not enough preceding safe density
12389 instructions to align a target, no widening will be performed. The
12390 default is @option{-mtarget-align}. These options do not affect the
12391 treatment of auto-aligned instructions like @code{LOOP}, which the
12392 assembler will always align, either by widening density instructions or
12393 by inserting no-op instructions.
12396 @itemx -mno-longcalls
12397 @opindex mlongcalls
12398 @opindex mno-longcalls
12399 When this option is enabled, GCC instructs the assembler to translate
12400 direct calls to indirect calls unless it can determine that the target
12401 of a direct call is in the range allowed by the call instruction. This
12402 translation typically occurs for calls to functions in other source
12403 files. Specifically, the assembler translates a direct @code{CALL}
12404 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
12405 The default is @option{-mno-longcalls}. This option should be used in
12406 programs where the call target can potentially be out of range. This
12407 option is implemented in the assembler, not the compiler, so the
12408 assembly code generated by GCC will still show direct call
12409 instructions---look at the disassembled object code to see the actual
12410 instructions. Note that the assembler will use an indirect call for
12411 every cross-file call, not just those that really will be out of range.
12414 @node zSeries Options
12415 @subsection zSeries Options
12416 @cindex zSeries options
12418 These are listed under @xref{S/390 and zSeries Options}.
12420 @node Code Gen Options
12421 @section Options for Code Generation Conventions
12422 @cindex code generation conventions
12423 @cindex options, code generation
12424 @cindex run-time options
12426 These machine-independent options control the interface conventions
12427 used in code generation.
12429 Most of them have both positive and negative forms; the negative form
12430 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
12431 one of the forms is listed---the one which is not the default. You
12432 can figure out the other form by either removing @samp{no-} or adding
12436 @item -fbounds-check
12437 @opindex fbounds-check
12438 For front-ends that support it, generate additional code to check that
12439 indices used to access arrays are within the declared range. This is
12440 currently only supported by the Java and Fortran 77 front-ends, where
12441 this option defaults to true and false respectively.
12445 This option generates traps for signed overflow on addition, subtraction,
12446 multiplication operations.
12450 This option instructs the compiler to assume that signed arithmetic
12451 overflow of addition, subtraction and multiplication wraps around
12452 using twos-complement representation. This flag enables some optimizations
12453 and disables other. This option is enabled by default for the Java
12454 front-end, as required by the Java language specification.
12457 @opindex fexceptions
12458 Enable exception handling. Generates extra code needed to propagate
12459 exceptions. For some targets, this implies GCC will generate frame
12460 unwind information for all functions, which can produce significant data
12461 size overhead, although it does not affect execution. If you do not
12462 specify this option, GCC will enable it by default for languages like
12463 C++ which normally require exception handling, and disable it for
12464 languages like C that do not normally require it. However, you may need
12465 to enable this option when compiling C code that needs to interoperate
12466 properly with exception handlers written in C++. You may also wish to
12467 disable this option if you are compiling older C++ programs that don't
12468 use exception handling.
12470 @item -fnon-call-exceptions
12471 @opindex fnon-call-exceptions
12472 Generate code that allows trapping instructions to throw exceptions.
12473 Note that this requires platform-specific runtime support that does
12474 not exist everywhere. Moreover, it only allows @emph{trapping}
12475 instructions to throw exceptions, i.e.@: memory references or floating
12476 point instructions. It does not allow exceptions to be thrown from
12477 arbitrary signal handlers such as @code{SIGALRM}.
12479 @item -funwind-tables
12480 @opindex funwind-tables
12481 Similar to @option{-fexceptions}, except that it will just generate any needed
12482 static data, but will not affect the generated code in any other way.
12483 You will normally not enable this option; instead, a language processor
12484 that needs this handling would enable it on your behalf.
12486 @item -fasynchronous-unwind-tables
12487 @opindex fasynchronous-unwind-tables
12488 Generate unwind table in dwarf2 format, if supported by target machine. The
12489 table is exact at each instruction boundary, so it can be used for stack
12490 unwinding from asynchronous events (such as debugger or garbage collector).
12492 @item -fpcc-struct-return
12493 @opindex fpcc-struct-return
12494 Return ``short'' @code{struct} and @code{union} values in memory like
12495 longer ones, rather than in registers. This convention is less
12496 efficient, but it has the advantage of allowing intercallability between
12497 GCC-compiled files and files compiled with other compilers, particularly
12498 the Portable C Compiler (pcc).
12500 The precise convention for returning structures in memory depends
12501 on the target configuration macros.
12503 Short structures and unions are those whose size and alignment match
12504 that of some integer type.
12506 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12507 switch is not binary compatible with code compiled with the
12508 @option{-freg-struct-return} switch.
12509 Use it to conform to a non-default application binary interface.
12511 @item -freg-struct-return
12512 @opindex freg-struct-return
12513 Return @code{struct} and @code{union} values in registers when possible.
12514 This is more efficient for small structures than
12515 @option{-fpcc-struct-return}.
12517 If you specify neither @option{-fpcc-struct-return} nor
12518 @option{-freg-struct-return}, GCC defaults to whichever convention is
12519 standard for the target. If there is no standard convention, GCC
12520 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12521 the principal compiler. In those cases, we can choose the standard, and
12522 we chose the more efficient register return alternative.
12524 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12525 switch is not binary compatible with code compiled with the
12526 @option{-fpcc-struct-return} switch.
12527 Use it to conform to a non-default application binary interface.
12529 @item -fshort-enums
12530 @opindex fshort-enums
12531 Allocate to an @code{enum} type only as many bytes as it needs for the
12532 declared range of possible values. Specifically, the @code{enum} type
12533 will be equivalent to the smallest integer type which has enough room.
12535 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12536 code that is not binary compatible with code generated without that switch.
12537 Use it to conform to a non-default application binary interface.
12539 @item -fshort-double
12540 @opindex fshort-double
12541 Use the same size for @code{double} as for @code{float}.
12543 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
12544 code that is not binary compatible with code generated without that switch.
12545 Use it to conform to a non-default application binary interface.
12547 @item -fshort-wchar
12548 @opindex fshort-wchar
12549 Override the underlying type for @samp{wchar_t} to be @samp{short
12550 unsigned int} instead of the default for the target. This option is
12551 useful for building programs to run under WINE@.
12553 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12554 code that is not binary compatible with code generated without that switch.
12555 Use it to conform to a non-default application binary interface.
12557 @item -fshared-data
12558 @opindex fshared-data
12559 Requests that the data and non-@code{const} variables of this
12560 compilation be shared data rather than private data. The distinction
12561 makes sense only on certain operating systems, where shared data is
12562 shared between processes running the same program, while private data
12563 exists in one copy per process.
12566 @opindex fno-common
12567 In C, allocate even uninitialized global variables in the data section of the
12568 object file, rather than generating them as common blocks. This has the
12569 effect that if the same variable is declared (without @code{extern}) in
12570 two different compilations, you will get an error when you link them.
12571 The only reason this might be useful is if you wish to verify that the
12572 program will work on other systems which always work this way.
12576 Ignore the @samp{#ident} directive.
12578 @item -finhibit-size-directive
12579 @opindex finhibit-size-directive
12580 Don't output a @code{.size} assembler directive, or anything else that
12581 would cause trouble if the function is split in the middle, and the
12582 two halves are placed at locations far apart in memory. This option is
12583 used when compiling @file{crtstuff.c}; you should not need to use it
12586 @item -fverbose-asm
12587 @opindex fverbose-asm
12588 Put extra commentary information in the generated assembly code to
12589 make it more readable. This option is generally only of use to those
12590 who actually need to read the generated assembly code (perhaps while
12591 debugging the compiler itself).
12593 @option{-fno-verbose-asm}, the default, causes the
12594 extra information to be omitted and is useful when comparing two assembler
12599 @cindex global offset table
12601 Generate position-independent code (PIC) suitable for use in a shared
12602 library, if supported for the target machine. Such code accesses all
12603 constant addresses through a global offset table (GOT)@. The dynamic
12604 loader resolves the GOT entries when the program starts (the dynamic
12605 loader is not part of GCC; it is part of the operating system). If
12606 the GOT size for the linked executable exceeds a machine-specific
12607 maximum size, you get an error message from the linker indicating that
12608 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12609 instead. (These maximums are 8k on the SPARC and 32k
12610 on the m68k and RS/6000. The 386 has no such limit.)
12612 Position-independent code requires special support, and therefore works
12613 only on certain machines. For the 386, GCC supports PIC for System V
12614 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12615 position-independent.
12619 If supported for the target machine, emit position-independent code,
12620 suitable for dynamic linking and avoiding any limit on the size of the
12621 global offset table. This option makes a difference on the m68k,
12622 PowerPC and SPARC@.
12624 Position-independent code requires special support, and therefore works
12625 only on certain machines.
12631 These options are similar to @option{-fpic} and @option{-fPIC}, but
12632 generated position independent code can be only linked into executables.
12633 Usually these options are used when @option{-pie} GCC option will be
12634 used during linking.
12636 @item -fno-jump-tables
12637 @opindex fno-jump-tables
12638 Do not use jump tables for switch statements even where it would be
12639 more efficient than other code generation strategies. This option is
12640 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12641 building code which forms part of a dynamic linker and cannot
12642 reference the address of a jump table. On some targets, jump tables
12643 do not require a GOT and this option is not needed.
12645 @item -ffixed-@var{reg}
12647 Treat the register named @var{reg} as a fixed register; generated code
12648 should never refer to it (except perhaps as a stack pointer, frame
12649 pointer or in some other fixed role).
12651 @var{reg} must be the name of a register. The register names accepted
12652 are machine-specific and are defined in the @code{REGISTER_NAMES}
12653 macro in the machine description macro file.
12655 This flag does not have a negative form, because it specifies a
12658 @item -fcall-used-@var{reg}
12659 @opindex fcall-used
12660 Treat the register named @var{reg} as an allocable register that is
12661 clobbered by function calls. It may be allocated for temporaries or
12662 variables that do not live across a call. Functions compiled this way
12663 will not save and restore the register @var{reg}.
12665 It is an error to used this flag with the frame pointer or stack pointer.
12666 Use of this flag for other registers that have fixed pervasive roles in
12667 the machine's execution model will produce disastrous results.
12669 This flag does not have a negative form, because it specifies a
12672 @item -fcall-saved-@var{reg}
12673 @opindex fcall-saved
12674 Treat the register named @var{reg} as an allocable register saved by
12675 functions. It may be allocated even for temporaries or variables that
12676 live across a call. Functions compiled this way will save and restore
12677 the register @var{reg} if they use it.
12679 It is an error to used this flag with the frame pointer or stack pointer.
12680 Use of this flag for other registers that have fixed pervasive roles in
12681 the machine's execution model will produce disastrous results.
12683 A different sort of disaster will result from the use of this flag for
12684 a register in which function values may be returned.
12686 This flag does not have a negative form, because it specifies a
12689 @item -fpack-struct[=@var{n}]
12690 @opindex fpack-struct
12691 Without a value specified, pack all structure members together without
12692 holes. When a value is specified (which must be a small power of two), pack
12693 structure members according to this value, representing the maximum
12694 alignment (that is, objects with default alignment requirements larger than
12695 this will be output potentially unaligned at the next fitting location.
12697 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12698 code that is not binary compatible with code generated without that switch.
12699 Additionally, it makes the code suboptimal.
12700 Use it to conform to a non-default application binary interface.
12702 @item -finstrument-functions
12703 @opindex finstrument-functions
12704 Generate instrumentation calls for entry and exit to functions. Just
12705 after function entry and just before function exit, the following
12706 profiling functions will be called with the address of the current
12707 function and its call site. (On some platforms,
12708 @code{__builtin_return_address} does not work beyond the current
12709 function, so the call site information may not be available to the
12710 profiling functions otherwise.)
12713 void __cyg_profile_func_enter (void *this_fn,
12715 void __cyg_profile_func_exit (void *this_fn,
12719 The first argument is the address of the start of the current function,
12720 which may be looked up exactly in the symbol table.
12722 This instrumentation is also done for functions expanded inline in other
12723 functions. The profiling calls will indicate where, conceptually, the
12724 inline function is entered and exited. This means that addressable
12725 versions of such functions must be available. If all your uses of a
12726 function are expanded inline, this may mean an additional expansion of
12727 code size. If you use @samp{extern inline} in your C code, an
12728 addressable version of such functions must be provided. (This is
12729 normally the case anyways, but if you get lucky and the optimizer always
12730 expands the functions inline, you might have gotten away without
12731 providing static copies.)
12733 A function may be given the attribute @code{no_instrument_function}, in
12734 which case this instrumentation will not be done. This can be used, for
12735 example, for the profiling functions listed above, high-priority
12736 interrupt routines, and any functions from which the profiling functions
12737 cannot safely be called (perhaps signal handlers, if the profiling
12738 routines generate output or allocate memory).
12740 @item -fstack-check
12741 @opindex fstack-check
12742 Generate code to verify that you do not go beyond the boundary of the
12743 stack. You should specify this flag if you are running in an
12744 environment with multiple threads, but only rarely need to specify it in
12745 a single-threaded environment since stack overflow is automatically
12746 detected on nearly all systems if there is only one stack.
12748 Note that this switch does not actually cause checking to be done; the
12749 operating system must do that. The switch causes generation of code
12750 to ensure that the operating system sees the stack being extended.
12752 @item -fstack-limit-register=@var{reg}
12753 @itemx -fstack-limit-symbol=@var{sym}
12754 @itemx -fno-stack-limit
12755 @opindex fstack-limit-register
12756 @opindex fstack-limit-symbol
12757 @opindex fno-stack-limit
12758 Generate code to ensure that the stack does not grow beyond a certain value,
12759 either the value of a register or the address of a symbol. If the stack
12760 would grow beyond the value, a signal is raised. For most targets,
12761 the signal is raised before the stack overruns the boundary, so
12762 it is possible to catch the signal without taking special precautions.
12764 For instance, if the stack starts at absolute address @samp{0x80000000}
12765 and grows downwards, you can use the flags
12766 @option{-fstack-limit-symbol=__stack_limit} and
12767 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12768 of 128KB@. Note that this may only work with the GNU linker.
12770 @cindex aliasing of parameters
12771 @cindex parameters, aliased
12772 @item -fargument-alias
12773 @itemx -fargument-noalias
12774 @itemx -fargument-noalias-global
12775 @opindex fargument-alias
12776 @opindex fargument-noalias
12777 @opindex fargument-noalias-global
12778 Specify the possible relationships among parameters and between
12779 parameters and global data.
12781 @option{-fargument-alias} specifies that arguments (parameters) may
12782 alias each other and may alias global storage.@*
12783 @option{-fargument-noalias} specifies that arguments do not alias
12784 each other, but may alias global storage.@*
12785 @option{-fargument-noalias-global} specifies that arguments do not
12786 alias each other and do not alias global storage.
12788 Each language will automatically use whatever option is required by
12789 the language standard. You should not need to use these options yourself.
12791 @item -fleading-underscore
12792 @opindex fleading-underscore
12793 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12794 change the way C symbols are represented in the object file. One use
12795 is to help link with legacy assembly code.
12797 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12798 generate code that is not binary compatible with code generated without that
12799 switch. Use it to conform to a non-default application binary interface.
12800 Not all targets provide complete support for this switch.
12802 @item -ftls-model=@var{model}
12803 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12804 The @var{model} argument should be one of @code{global-dynamic},
12805 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
12807 The default without @option{-fpic} is @code{initial-exec}; with
12808 @option{-fpic} the default is @code{global-dynamic}.
12810 @item -fvisibility=@var{default|internal|hidden|protected}
12811 @opindex fvisibility
12812 Set the default ELF image symbol visibility to the specified option---all
12813 symbols will be marked with this unless overridden within the code.
12814 Using this feature can very substantially improve linking and
12815 load times of shared object libraries, produce more optimized
12816 code, provide near-perfect API export and prevent symbol clashes.
12817 It is @strong{strongly} recommended that you use this in any shared objects
12820 Despite the nomenclature, @code{default} always means public ie;
12821 available to be linked against from outside the shared object.
12822 @code{protected} and @code{internal} are pretty useless in real-world
12823 usage so the only other commonly used option will be @code{hidden}.
12824 The default if @option{-fvisibility} isn't specified is
12825 @code{default}, i.e., make every
12826 symbol public---this causes the same behavior as previous versions of
12829 A good explanation of the benefits offered by ensuring ELF
12830 symbols have the correct visibility is given by ``How To Write
12831 Shared Libraries'' by Ulrich Drepper (which can be found at
12832 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
12833 solution made possible by this option to marking things hidden when
12834 the default is public is to make the default hidden and mark things
12835 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
12836 and @code{__attribute__ ((visibility("default")))} instead of
12837 @code{__declspec(dllexport)} you get almost identical semantics with
12838 identical syntax. This is a great boon to those working with
12839 cross-platform projects.
12841 For those adding visibility support to existing code, you may find
12842 @samp{#pragma GCC visibility} of use. This works by you enclosing
12843 the declarations you wish to set visibility for with (for example)
12844 @samp{#pragma GCC visibility push(hidden)} and
12845 @samp{#pragma GCC visibility pop}.
12846 Bear in mind that symbol visibility should be viewed @strong{as
12847 part of the API interface contract} and thus all new code should
12848 always specify visibility when it is not the default ie; declarations
12849 only for use within the local DSO should @strong{always} be marked explicitly
12850 as hidden as so to avoid PLT indirection overheads---making this
12851 abundantly clear also aids readability and self-documentation of the code.
12852 Note that due to ISO C++ specification requirements, operator new and
12853 operator delete must always be of default visibility.
12855 An overview of these techniques, their benefits and how to use them
12856 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
12862 @node Environment Variables
12863 @section Environment Variables Affecting GCC
12864 @cindex environment variables
12866 @c man begin ENVIRONMENT
12867 This section describes several environment variables that affect how GCC
12868 operates. Some of them work by specifying directories or prefixes to use
12869 when searching for various kinds of files. Some are used to specify other
12870 aspects of the compilation environment.
12872 Note that you can also specify places to search using options such as
12873 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
12874 take precedence over places specified using environment variables, which
12875 in turn take precedence over those specified by the configuration of GCC@.
12876 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
12877 GNU Compiler Collection (GCC) Internals}.
12882 @c @itemx LC_COLLATE
12884 @c @itemx LC_MONETARY
12885 @c @itemx LC_NUMERIC
12890 @c @findex LC_COLLATE
12891 @findex LC_MESSAGES
12892 @c @findex LC_MONETARY
12893 @c @findex LC_NUMERIC
12897 These environment variables control the way that GCC uses
12898 localization information that allow GCC to work with different
12899 national conventions. GCC inspects the locale categories
12900 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
12901 so. These locale categories can be set to any value supported by your
12902 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
12903 Kingdom encoded in UTF-8.
12905 The @env{LC_CTYPE} environment variable specifies character
12906 classification. GCC uses it to determine the character boundaries in
12907 a string; this is needed for some multibyte encodings that contain quote
12908 and escape characters that would otherwise be interpreted as a string
12911 The @env{LC_MESSAGES} environment variable specifies the language to
12912 use in diagnostic messages.
12914 If the @env{LC_ALL} environment variable is set, it overrides the value
12915 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
12916 and @env{LC_MESSAGES} default to the value of the @env{LANG}
12917 environment variable. If none of these variables are set, GCC
12918 defaults to traditional C English behavior.
12922 If @env{TMPDIR} is set, it specifies the directory to use for temporary
12923 files. GCC uses temporary files to hold the output of one stage of
12924 compilation which is to be used as input to the next stage: for example,
12925 the output of the preprocessor, which is the input to the compiler
12928 @item GCC_EXEC_PREFIX
12929 @findex GCC_EXEC_PREFIX
12930 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
12931 names of the subprograms executed by the compiler. No slash is added
12932 when this prefix is combined with the name of a subprogram, but you can
12933 specify a prefix that ends with a slash if you wish.
12935 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
12936 an appropriate prefix to use based on the pathname it was invoked with.
12938 If GCC cannot find the subprogram using the specified prefix, it
12939 tries looking in the usual places for the subprogram.
12941 The default value of @env{GCC_EXEC_PREFIX} is
12942 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
12943 of @code{prefix} when you ran the @file{configure} script.
12945 Other prefixes specified with @option{-B} take precedence over this prefix.
12947 This prefix is also used for finding files such as @file{crt0.o} that are
12950 In addition, the prefix is used in an unusual way in finding the
12951 directories to search for header files. For each of the standard
12952 directories whose name normally begins with @samp{/usr/local/lib/gcc}
12953 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
12954 replacing that beginning with the specified prefix to produce an
12955 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
12956 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
12957 These alternate directories are searched first; the standard directories
12960 @item COMPILER_PATH
12961 @findex COMPILER_PATH
12962 The value of @env{COMPILER_PATH} is a colon-separated list of
12963 directories, much like @env{PATH}. GCC tries the directories thus
12964 specified when searching for subprograms, if it can't find the
12965 subprograms using @env{GCC_EXEC_PREFIX}.
12968 @findex LIBRARY_PATH
12969 The value of @env{LIBRARY_PATH} is a colon-separated list of
12970 directories, much like @env{PATH}. When configured as a native compiler,
12971 GCC tries the directories thus specified when searching for special
12972 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
12973 using GCC also uses these directories when searching for ordinary
12974 libraries for the @option{-l} option (but directories specified with
12975 @option{-L} come first).
12979 @cindex locale definition
12980 This variable is used to pass locale information to the compiler. One way in
12981 which this information is used is to determine the character set to be used
12982 when character literals, string literals and comments are parsed in C and C++.
12983 When the compiler is configured to allow multibyte characters,
12984 the following values for @env{LANG} are recognized:
12988 Recognize JIS characters.
12990 Recognize SJIS characters.
12992 Recognize EUCJP characters.
12995 If @env{LANG} is not defined, or if it has some other value, then the
12996 compiler will use mblen and mbtowc as defined by the default locale to
12997 recognize and translate multibyte characters.
13001 Some additional environments variables affect the behavior of the
13004 @include cppenv.texi
13008 @node Precompiled Headers
13009 @section Using Precompiled Headers
13010 @cindex precompiled headers
13011 @cindex speed of compilation
13013 Often large projects have many header files that are included in every
13014 source file. The time the compiler takes to process these header files
13015 over and over again can account for nearly all of the time required to
13016 build the project. To make builds faster, GCC allows users to
13017 `precompile' a header file; then, if builds can use the precompiled
13018 header file they will be much faster.
13020 To create a precompiled header file, simply compile it as you would any
13021 other file, if necessary using the @option{-x} option to make the driver
13022 treat it as a C or C++ header file. You will probably want to use a
13023 tool like @command{make} to keep the precompiled header up-to-date when
13024 the headers it contains change.
13026 A precompiled header file will be searched for when @code{#include} is
13027 seen in the compilation. As it searches for the included file
13028 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
13029 compiler looks for a precompiled header in each directory just before it
13030 looks for the include file in that directory. The name searched for is
13031 the name specified in the @code{#include} with @samp{.gch} appended. If
13032 the precompiled header file can't be used, it is ignored.
13034 For instance, if you have @code{#include "all.h"}, and you have
13035 @file{all.h.gch} in the same directory as @file{all.h}, then the
13036 precompiled header file will be used if possible, and the original
13037 header will be used otherwise.
13039 Alternatively, you might decide to put the precompiled header file in a
13040 directory and use @option{-I} to ensure that directory is searched
13041 before (or instead of) the directory containing the original header.
13042 Then, if you want to check that the precompiled header file is always
13043 used, you can put a file of the same name as the original header in this
13044 directory containing an @code{#error} command.
13046 This also works with @option{-include}. So yet another way to use
13047 precompiled headers, good for projects not designed with precompiled
13048 header files in mind, is to simply take most of the header files used by
13049 a project, include them from another header file, precompile that header
13050 file, and @option{-include} the precompiled header. If the header files
13051 have guards against multiple inclusion, they will be skipped because
13052 they've already been included (in the precompiled header).
13054 If you need to precompile the same header file for different
13055 languages, targets, or compiler options, you can instead make a
13056 @emph{directory} named like @file{all.h.gch}, and put each precompiled
13057 header in the directory, perhaps using @option{-o}. It doesn't matter
13058 what you call the files in the directory, every precompiled header in
13059 the directory will be considered. The first precompiled header
13060 encountered in the directory that is valid for this compilation will
13061 be used; they're searched in no particular order.
13063 There are many other possibilities, limited only by your imagination,
13064 good sense, and the constraints of your build system.
13066 A precompiled header file can be used only when these conditions apply:
13070 Only one precompiled header can be used in a particular compilation.
13073 A precompiled header can't be used once the first C token is seen. You
13074 can have preprocessor directives before a precompiled header; you can
13075 even include a precompiled header from inside another header, so long as
13076 there are no C tokens before the @code{#include}.
13079 The precompiled header file must be produced for the same language as
13080 the current compilation. You can't use a C precompiled header for a C++
13084 The precompiled header file must have been produced by the same compiler
13085 binary as the current compilation is using.
13088 Any macros defined before the precompiled header is included must
13089 either be defined in the same way as when the precompiled header was
13090 generated, or must not affect the precompiled header, which usually
13091 means that they don't appear in the precompiled header at all.
13093 The @option{-D} option is one way to define a macro before a
13094 precompiled header is included; using a @code{#define} can also do it.
13095 There are also some options that define macros implicitly, like
13096 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
13099 @item If debugging information is output when using the precompiled
13100 header, using @option{-g} or similar, the same kind of debugging information
13101 must have been output when building the precompiled header. However,
13102 a precompiled header built using @option{-g} can be used in a compilation
13103 when no debugging information is being output.
13105 @item The same @option{-m} options must generally be used when building
13106 and using the precompiled header. @xref{Submodel Options},
13107 for any cases where this rule is relaxed.
13109 @item Each of the following options must be the same when building and using
13110 the precompiled header:
13112 @gccoptlist{-fexceptions -funit-at-a-time}
13115 Some other command-line options starting with @option{-f},
13116 @option{-p}, or @option{-O} must be defined in the same way as when
13117 the precompiled header was generated. At present, it's not clear
13118 which options are safe to change and which are not; the safest choice
13119 is to use exactly the same options when generating and using the
13120 precompiled header. The following are known to be safe:
13122 @gccoptlist{-fpreprocessed
13123 -fsched-interblock -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
13124 -fsched-verbose=<number> -fschedule-insns
13129 For all of these except the last, the compiler will automatically
13130 ignore the precompiled header if the conditions aren't met. If you
13131 find an option combination that doesn't work and doesn't cause the
13132 precompiled header to be ignored, please consider filing a bug report,
13135 If you do use differing options when generating and using the
13136 precompiled header, the actual behavior will be a mixture of the
13137 behavior for the options. For instance, if you use @option{-g} to
13138 generate the precompiled header but not when using it, you may or may
13139 not get debugging information for routines in the precompiled header.
13141 @node Running Protoize
13142 @section Running Protoize
13144 The program @code{protoize} is an optional part of GCC@. You can use
13145 it to add prototypes to a program, thus converting the program to ISO
13146 C in one respect. The companion program @code{unprotoize} does the
13147 reverse: it removes argument types from any prototypes that are found.
13149 When you run these programs, you must specify a set of source files as
13150 command line arguments. The conversion programs start out by compiling
13151 these files to see what functions they define. The information gathered
13152 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
13154 After scanning comes actual conversion. The specified files are all
13155 eligible to be converted; any files they include (whether sources or
13156 just headers) are eligible as well.
13158 But not all the eligible files are converted. By default,
13159 @code{protoize} and @code{unprotoize} convert only source and header
13160 files in the current directory. You can specify additional directories
13161 whose files should be converted with the @option{-d @var{directory}}
13162 option. You can also specify particular files to exclude with the
13163 @option{-x @var{file}} option. A file is converted if it is eligible, its
13164 directory name matches one of the specified directory names, and its
13165 name within the directory has not been excluded.
13167 Basic conversion with @code{protoize} consists of rewriting most
13168 function definitions and function declarations to specify the types of
13169 the arguments. The only ones not rewritten are those for varargs
13172 @code{protoize} optionally inserts prototype declarations at the
13173 beginning of the source file, to make them available for any calls that
13174 precede the function's definition. Or it can insert prototype
13175 declarations with block scope in the blocks where undeclared functions
13178 Basic conversion with @code{unprotoize} consists of rewriting most
13179 function declarations to remove any argument types, and rewriting
13180 function definitions to the old-style pre-ISO form.
13182 Both conversion programs print a warning for any function declaration or
13183 definition that they can't convert. You can suppress these warnings
13186 The output from @code{protoize} or @code{unprotoize} replaces the
13187 original source file. The original file is renamed to a name ending
13188 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
13189 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
13190 for DOS) file already exists, then the source file is simply discarded.
13192 @code{protoize} and @code{unprotoize} both depend on GCC itself to
13193 scan the program and collect information about the functions it uses.
13194 So neither of these programs will work until GCC is installed.
13196 Here is a table of the options you can use with @code{protoize} and
13197 @code{unprotoize}. Each option works with both programs unless
13201 @item -B @var{directory}
13202 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
13203 usual directory (normally @file{/usr/local/lib}). This file contains
13204 prototype information about standard system functions. This option
13205 applies only to @code{protoize}.
13207 @item -c @var{compilation-options}
13208 Use @var{compilation-options} as the options when running @command{gcc} to
13209 produce the @samp{.X} files. The special option @option{-aux-info} is
13210 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
13212 Note that the compilation options must be given as a single argument to
13213 @code{protoize} or @code{unprotoize}. If you want to specify several
13214 @command{gcc} options, you must quote the entire set of compilation options
13215 to make them a single word in the shell.
13217 There are certain @command{gcc} arguments that you cannot use, because they
13218 would produce the wrong kind of output. These include @option{-g},
13219 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
13220 the @var{compilation-options}, they are ignored.
13223 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
13224 systems) instead of @samp{.c}. This is convenient if you are converting
13225 a C program to C++. This option applies only to @code{protoize}.
13228 Add explicit global declarations. This means inserting explicit
13229 declarations at the beginning of each source file for each function
13230 that is called in the file and was not declared. These declarations
13231 precede the first function definition that contains a call to an
13232 undeclared function. This option applies only to @code{protoize}.
13234 @item -i @var{string}
13235 Indent old-style parameter declarations with the string @var{string}.
13236 This option applies only to @code{protoize}.
13238 @code{unprotoize} converts prototyped function definitions to old-style
13239 function definitions, where the arguments are declared between the
13240 argument list and the initial @samp{@{}. By default, @code{unprotoize}
13241 uses five spaces as the indentation. If you want to indent with just
13242 one space instead, use @option{-i " "}.
13245 Keep the @samp{.X} files. Normally, they are deleted after conversion
13249 Add explicit local declarations. @code{protoize} with @option{-l} inserts
13250 a prototype declaration for each function in each block which calls the
13251 function without any declaration. This option applies only to
13255 Make no real changes. This mode just prints information about the conversions
13256 that would have been done without @option{-n}.
13259 Make no @samp{.save} files. The original files are simply deleted.
13260 Use this option with caution.
13262 @item -p @var{program}
13263 Use the program @var{program} as the compiler. Normally, the name
13264 @file{gcc} is used.
13267 Work quietly. Most warnings are suppressed.
13270 Print the version number, just like @option{-v} for @command{gcc}.
13273 If you need special compiler options to compile one of your program's
13274 source files, then you should generate that file's @samp{.X} file
13275 specially, by running @command{gcc} on that source file with the
13276 appropriate options and the option @option{-aux-info}. Then run
13277 @code{protoize} on the entire set of files. @code{protoize} will use
13278 the existing @samp{.X} file because it is newer than the source file.
13282 gcc -Dfoo=bar file1.c -aux-info file1.X
13287 You need to include the special files along with the rest in the
13288 @code{protoize} command, even though their @samp{.X} files already
13289 exist, because otherwise they won't get converted.
13291 @xref{Protoize Caveats}, for more information on how to use
13292 @code{protoize} successfully.