1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
11 @c man begin COPYRIGHT
12 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
13 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
15 Permission is granted to copy, distribute and/or modify this document
16 under the terms of the GNU Free Documentation License, Version 1.2 or
17 any later version published by the Free Software Foundation; with the
18 Invariant Sections being ``GNU General Public License'' and ``Funding
19 Free Software'', the Front-Cover texts being (a) (see below), and with
20 the Back-Cover Texts being (b) (see below). A copy of the license is
21 included in the gfdl(7) man page.
23 (a) The FSF's Front-Cover Text is:
27 (b) The FSF's Back-Cover Text is:
29 You have freedom to copy and modify this GNU Manual, like GNU
30 software. Copies published by the Free Software Foundation raise
31 funds for GNU development.
33 @c Set file name and title for the man page.
35 @settitle GNU project C and C++ compiler
37 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
38 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
39 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
40 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
41 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
42 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
43 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
45 Only the most useful options are listed here; see below for the
46 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
49 gpl(7), gfdl(7), fsf-funding(7),
50 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
51 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
52 @file{ld}, @file{binutils} and @file{gdb}.
55 For instructions on reporting bugs, see
56 @w{@uref{http://gcc.gnu.org/bugs.html}}.
59 See the Info entry for @command{gcc}, or
60 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
61 for contributors to GCC@.
66 @chapter GCC Command Options
67 @cindex GCC command options
68 @cindex command options
69 @cindex options, GCC command
71 @c man begin DESCRIPTION
72 When you invoke GCC, it normally does preprocessing, compilation,
73 assembly and linking. The ``overall options'' allow you to stop this
74 process at an intermediate stage. For example, the @option{-c} option
75 says not to run the linker. Then the output consists of object files
76 output by the assembler.
78 Other options are passed on to one stage of processing. Some options
79 control the preprocessor and others the compiler itself. Yet other
80 options control the assembler and linker; most of these are not
81 documented here, since you rarely need to use any of them.
83 @cindex C compilation options
84 Most of the command line options that you can use with GCC are useful
85 for C programs; when an option is only useful with another language
86 (usually C++), the explanation says so explicitly. If the description
87 for a particular option does not mention a source language, you can use
88 that option with all supported languages.
90 @cindex C++ compilation options
91 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
92 options for compiling C++ programs.
94 @cindex grouping options
95 @cindex options, grouping
96 The @command{gcc} program accepts options and file names as operands. Many
97 options have multi-letter names; therefore multiple single-letter options
98 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
101 @cindex order of options
102 @cindex options, order
103 You can mix options and other arguments. For the most part, the order
104 you use doesn't matter. Order does matter when you use several options
105 of the same kind; for example, if you specify @option{-L} more than once,
106 the directories are searched in the order specified.
108 Many options have long names starting with @samp{-f} or with
109 @samp{-W}---for example,
110 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
111 these have both positive and negative forms; the negative form of
112 @option{-ffoo} would be @option{-fno-foo}. This manual documents
113 only one of these two forms, whichever one is not the default.
117 @xref{Option Index}, for an index to GCC's options.
120 * Option Summary:: Brief list of all options, without explanations.
121 * Overall Options:: Controlling the kind of output:
122 an executable, object files, assembler files,
123 or preprocessed source.
124 * Invoking G++:: Compiling C++ programs.
125 * C Dialect Options:: Controlling the variant of C language compiled.
126 * C++ Dialect Options:: Variations on C++.
127 * Language Independent Options:: Controlling how diagnostics should be
129 * Warning Options:: How picky should the compiler be?
130 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
131 * Optimize Options:: How much optimization?
132 * Preprocessor Options:: Controlling header files and macro definitions.
133 Also, getting dependency information for Make.
134 * Assembler Options:: Passing options to the assembler.
135 * Link Options:: Specifying libraries and so on.
136 * Directory Options:: Where to find header files and libraries.
137 Where to find the compiler executable files.
138 * Spec Files:: How to pass switches to sub-processes.
139 * Target Options:: Running a cross-compiler, or an old version of GCC.
140 * Submodel Options:: Specifying minor hardware or convention variations,
141 such as 68010 vs 68020.
142 * Code Gen Options:: Specifying conventions for function calls, data layout
144 * Environment Variables:: Env vars that affect GCC.
145 * Precompiled Headers:: Compiling a header once, and using it many times.
146 * Running Protoize:: Automatically adding or removing function prototypes.
152 @section Option Summary
154 Here is a summary of all the options, grouped by type. Explanations are
155 in the following sections.
158 @item Overall Options
159 @xref{Overall Options,,Options Controlling the Kind of Output}.
160 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
161 -x @var{language} -v -### --help --target-help --version @@@var{file}}
163 @item C Language Options
164 @xref{C Dialect Options,,Options Controlling C Dialect}.
165 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
166 -aux-info @var{filename} @gol
167 @c APPLE LOCAL blocks 7205047 5811887
168 -fno-asm -fno-blocks -fno-builtin -fno-builtin-@var{function} @gol
169 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
170 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
171 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
172 @c APPLE LOCAL nested functions 4357979 */
173 -fno-nested-functions @gol
174 -fsigned-bitfields -fsigned-char @gol
175 @c APPLE LOCAL -Wnewline-eof 2001-08-23 --sts **
176 -Wnewline-eof (Apple compatible) @gol
177 -funsigned-bitfields -funsigned-char}
179 @item C++ Language Options
180 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
181 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
182 -fconserve-space -ffriend-injection @gol
183 -fno-elide-constructors @gol
184 -fno-enforce-eh-specs @gol
185 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
186 -fno-implicit-templates @gol
187 -fno-implicit-inline-templates @gol
188 -fno-implement-inlines -fms-extensions @gol
189 -fno-nonansi-builtins -fno-operator-names @gol
190 -fno-optional-diags -fpermissive @gol
191 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
192 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
193 -fno-default-inline -fvisibility-inlines-hidden @gol
194 -fvisibility-ms-compat @gol
195 -Wabi -Wctor-dtor-privacy @gol
196 -Wnon-virtual-dtor -Wreorder @gol
197 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel @gol
198 -Wno-non-template-friend -Wold-style-cast @gol
199 -Woverloaded-virtual -Wno-pmf-conversions @gol
202 @item Language Independent Options
203 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
204 @gccoptlist{-fmessage-length=@var{n} @gol
205 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
206 -fdiagnostics-show-option}
208 @item Warning Options
209 @xref{Warning Options,,Options to Request or Suppress Warnings}.
210 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
211 -w -Wextra -Wall -Waddress -Waggregate-return -Wno-attributes @gol
212 -Wc++-compat -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
213 -Wconversion -Wno-deprecated-declarations @gol
214 -Wdisabled-optimization -Wno-div-by-zero -Wno-endif-labels @gol
215 -Werror -Werror=* -Werror-implicit-function-declaration @gol
216 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
217 -Wno-format-extra-args -Wformat-nonliteral @gol
218 -Wformat-security -Wformat-y2k @gol
219 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
220 -Wimport -Wno-import -Winit-self -Winline @gol
221 -Wno-int-to-pointer-cast @gol
222 -Wno-invalid-offsetof -Winvalid-pch @gol
223 -Wlarger-than-@var{len} -Wframe-larger-than-@var{len} @gol
224 -Wunsafe-loop-optimizations -Wlong-long @gol
225 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
226 -Wmissing-format-attribute -Wmissing-include-dirs @gol
227 -Wmissing-noreturn @gol
228 @c APPLE LOCAL -Wmost
229 -Wmost (APPLE ONLY) @gol
230 -Wno-multichar -Wnonnull -Wno-overflow @gol
231 -Woverlength-strings -Wpacked -Wpadded @gol
232 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
233 -Wredundant-decls @gol
234 -Wreturn-type -Wsequence-point -Wshadow @gol
235 -Wsign-compare -Wstack-protector @gol
236 -Wstrict-aliasing -Wstrict-aliasing=2 @gol
237 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
238 -Wswitch -Wswitch-default -Wswitch-enum @gol
239 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
240 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
241 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
242 -Wunused-value -Wunused-variable @gol
243 -Wvariadic-macros -Wvla @gol
244 -Wvolatile-register-var -Wwrite-strings}
246 @item C-only Warning Options
247 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
248 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
249 -Wstrict-prototypes -Wtraditional @gol
250 -Wdeclaration-after-statement -Wpointer-sign}
252 @item Debugging Options
253 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
254 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
255 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
256 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
257 -fdump-ipa-all -fdump-ipa-cgraph @gol
259 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
260 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
261 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
262 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
264 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
266 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
267 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
268 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
269 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
270 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
271 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
272 -fdump-tree-nrv -fdump-tree-vect @gol
273 -fdump-tree-sink @gol
274 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
275 -fdump-tree-salias @gol
276 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
277 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
278 -ftree-vectorizer-verbose=@var{n} @gol
279 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
280 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
281 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
282 -fmem-report -fprofile-arcs @gol
283 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
284 -ftest-coverage -ftime-report -fvar-tracking @gol
285 -g -g@var{level} -gcoff -gdwarf-2 @gol
286 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
287 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
288 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @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 -fcrossjumping -fif-conversion -fif-conversion2 @gol
308 -finline-functions -finline-functions-called-once @gol
309 -finline-limit=@var{n} -fkeep-inline-functions @gol
310 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
311 -fmodulo-sched -fno-branch-count-reg @gol
312 -fno-default-inline -fno-defer-pop -fmove-loop-invariants @gol
313 -fno-function-cse -fno-guess-branch-probability @gol
314 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
315 -funsafe-math-optimizations -funsafe-loop-optimizations -ffinite-math-only @gol
316 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
317 -fomit-frame-pointer -foptimize-register-move @gol
318 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
319 -fprofile-generate -fprofile-use @gol
320 -fregmove -frename-registers @gol
321 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
322 -frerun-cse-after-loop @gol
323 -frounding-math -frtl-abstract-sequences @gol
324 -fschedule-insns -fschedule-insns2 @gol
325 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
326 -fsched-spec-load-dangerous @gol
327 -fsched-stalled-insns=@var{n} -fsched-stalled-insns-dep=@var{n} @gol
328 -fsched2-use-superblocks @gol
329 -fsched2-use-traces -fsee -freschedule-modulo-scheduled-loops @gol
330 -fsection-anchors -fsignaling-nans -fsingle-precision-constant @gol
331 -fstack-protector -fstack-protector-all @gol
332 -fstrict-aliasing -fstrict-overflow -ftracer -fthread-jumps @gol
333 -funroll-all-loops -funroll-loops -fpeel-loops @gol
334 -fsplit-ivs-in-unroller -funswitch-loops @gol
335 -fvariable-expansion-in-unroller @gol
336 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
337 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
338 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
339 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
340 -ftree-vect-loop-version -ftree-salias -fipa-pta -fweb @gol
341 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
342 --param @var{name}=@var{value}
343 -O -O0 -O1 -O2 -O3 -Os}
345 @item Preprocessor Options
346 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
347 @gccoptlist{-A@var{question}=@var{answer} @gol
348 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
349 -C -dD -dI -dM -dN @gol
350 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
351 -idirafter @var{dir} @gol
352 -include @var{file} -imacros @var{file} @gol
353 -iprefix @var{file} -iwithprefix @var{dir} @gol
354 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
355 -imultilib @var{dir} -isysroot @var{dir} @gol
356 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
357 -P -fworking-directory -remap @gol
358 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
359 -Xpreprocessor @var{option}}
361 @item Assembler Option
362 @xref{Assembler Options,,Passing Options to the Assembler}.
363 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
366 @xref{Link Options,,Options for Linking}.
367 @gccoptlist{@var{object-file-name} -l@var{library} @gol
368 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
369 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
370 -Wl,@var{option} -Xlinker @var{option} @gol
373 @item Directory Options
374 @xref{Directory Options,,Options for Directory Search}.
375 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
376 -specs=@var{file} -I- --sysroot=@var{dir}}
379 @c I wrote this xref this way to avoid overfull hbox. -- rms
380 @xref{Target Options}.
381 @gccoptlist{-V @var{version} -b @var{machine}}
383 @item Machine Dependent Options
384 @xref{Submodel Options,,Hardware Models and Configurations}.
385 @c This list is ordered alphanumerically by subsection name.
386 @c Try and put the significant identifier (CPU or system) first,
387 @c so users have a clue at guessing where the ones they want will be.
390 @gccoptlist{-EB -EL @gol
391 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
392 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
395 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
396 -mabi=@var{name} @gol
397 -mapcs-stack-check -mno-apcs-stack-check @gol
398 -mapcs-float -mno-apcs-float @gol
399 -mapcs-reentrant -mno-apcs-reentrant @gol
400 -msched-prolog -mno-sched-prolog @gol
401 -mlittle-endian -mbig-endian -mwords-little-endian @gol
402 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
403 -mthumb-interwork -mno-thumb-interwork @gol
404 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
405 -mstructure-size-boundary=@var{n} @gol
406 -mabort-on-noreturn @gol
407 -mlong-calls -mno-long-calls @gol
408 -msingle-pic-base -mno-single-pic-base @gol
409 -mpic-register=@var{reg} @gol
410 -mnop-fun-dllimport @gol
411 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
412 -mpoke-function-name @gol
414 -mtpcs-frame -mtpcs-leaf-frame @gol
415 -mcaller-super-interworking -mcallee-super-interworking @gol
419 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
420 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
422 @emph{Blackfin Options}
423 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
424 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
425 -mlow-64k -mno-low64k -mid-shared-library @gol
426 -mno-id-shared-library -mshared-library-id=@var{n} @gol
427 -mlong-calls -mno-long-calls}
430 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
431 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
432 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
433 -mstack-align -mdata-align -mconst-align @gol
434 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
435 -melf -maout -melinux -mlinux -sim -sim2 @gol
436 -mmul-bug-workaround -mno-mul-bug-workaround}
439 @gccoptlist{-mmac -mpush-args}
441 @emph{Darwin Options}
442 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
443 -arch_only -bind_at_load -bundle -bundle_loader @gol
444 -client_name -compatibility_version -current_version @gol
446 -dependency-file -dylib_file -dylinker_install_name @gol
447 -dynamic -dynamiclib -exported_symbols_list @gol
448 -filelist -flat_namespace -force_cpusubtype_ALL @gol
449 -force_flat_namespace -headerpad_max_install_names @gol
450 -image_base -init -install_name -keep_private_externs @gol
451 -multi_module -multiply_defined -multiply_defined_unused @gol
452 -noall_load -no_dead_strip_inits_and_terms @gol
453 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
454 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
455 -private_bundle -read_only_relocs -sectalign @gol
456 -sectobjectsymbols -whyload -seg1addr @gol
457 -sectcreate -sectobjectsymbols -sectorder @gol
458 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
459 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
460 -segprot -segs_read_only_addr -segs_read_write_addr @gol
461 -single_module -static -sub_library -sub_umbrella @gol
462 -twolevel_namespace -umbrella -undefined @gol
463 -unexported_symbols_list -weak_reference_mismatches @gol
464 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
465 -mkernel -mone-byte-bool}
467 @emph{DEC Alpha Options}
468 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
469 -mieee -mieee-with-inexact -mieee-conformant @gol
470 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
471 -mtrap-precision=@var{mode} -mbuild-constants @gol
472 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
473 -mbwx -mmax -mfix -mcix @gol
474 -mfloat-vax -mfloat-ieee @gol
475 -mexplicit-relocs -msmall-data -mlarge-data @gol
476 -msmall-text -mlarge-text @gol
477 -mmemory-latency=@var{time}}
479 @emph{DEC Alpha/VMS Options}
480 @gccoptlist{-mvms-return-codes}
483 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
484 -mhard-float -msoft-float @gol
485 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
486 -mdouble -mno-double @gol
487 -mmedia -mno-media -mmuladd -mno-muladd @gol
488 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
489 -mlinked-fp -mlong-calls -malign-labels @gol
490 -mlibrary-pic -macc-4 -macc-8 @gol
491 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
492 -moptimize-membar -mno-optimize-membar @gol
493 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
494 -mvliw-branch -mno-vliw-branch @gol
495 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
496 -mno-nested-cond-exec -mtomcat-stats @gol
500 @emph{GNU/Linux Options}
501 @gccoptlist{-muclibc}
503 @emph{H8/300 Options}
504 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
507 @gccoptlist{-march=@var{architecture-type} @gol
508 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
509 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
510 -mfixed-range=@var{register-range} @gol
511 -mjump-in-delay -mlinker-opt -mlong-calls @gol
512 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
513 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
514 -mno-jump-in-delay -mno-long-load-store @gol
515 -mno-portable-runtime -mno-soft-float @gol
516 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
517 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
518 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
519 -munix=@var{unix-std} -nolibdld -static -threads}
521 @emph{i386 and x86-64 Options}
522 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
523 -mfpmath=@var{unit} @gol
524 -masm=@var{dialect} -mno-fancy-math-387 @gol
525 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
526 -mno-wide-multiply -mrtd -malign-double @gol
527 -mpreferred-stack-boundary=@var{num} @gol
528 -mmmx -msse -msse2 -msse3 -mssse3 -msse4a -m3dnow -mpopcnt -mabm -maes @gol
529 -mthreads -mno-align-stringops -minline-all-stringops @gol
530 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
531 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
533 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
534 -mcmodel=@var{code-model} @gol
535 -m32 -m64 -mlarge-data-threshold=@var{num}}
538 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
539 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
540 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
541 -minline-float-divide-max-throughput @gol
542 -minline-int-divide-min-latency @gol
543 -minline-int-divide-max-throughput @gol
544 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
545 -mno-dwarf2-asm -mearly-stop-bits @gol
546 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
547 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
548 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
549 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
550 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
551 -mno-sched-prefer-non-data-spec-insns @gol
552 -mno-sched-prefer-non-control-spec-insns @gol
553 -mno-sched-count-spec-in-critical-path}
555 @emph{M32R/D Options}
556 @gccoptlist{-m32r2 -m32rx -m32r @gol
558 -malign-loops -mno-align-loops @gol
559 -missue-rate=@var{number} @gol
560 -mbranch-cost=@var{number} @gol
561 -mmodel=@var{code-size-model-type} @gol
562 -msdata=@var{sdata-type} @gol
563 -mno-flush-func -mflush-func=@var{name} @gol
564 -mno-flush-trap -mflush-trap=@var{number} @gol
568 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
570 @emph{M680x0 Options}
571 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
572 -m68060 -mcpu32 -m5200 -mcfv4e -m68881 -mbitfield @gol
573 -mc68000 -mc68020 @gol
574 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
575 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
576 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
578 @emph{M68hc1x Options}
579 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
580 -mauto-incdec -minmax -mlong-calls -mshort @gol
581 -msoft-reg-count=@var{count}}
584 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
585 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
586 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
587 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
588 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
591 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
592 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
593 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
594 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
595 -mfp32 -mfp64 -mhard-float -msoft-float @gol
596 -msingle-float -mdouble-float -mdsp -mpaired-single -mips3d @gol
597 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
598 -G@var{num} -membedded-data -mno-embedded-data @gol
599 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
600 -msplit-addresses -mno-split-addresses @gol
601 -mexplicit-relocs -mno-explicit-relocs @gol
602 -mcheck-zero-division -mno-check-zero-division @gol
603 -mdivide-traps -mdivide-breaks @gol
604 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
605 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
606 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
607 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
608 -mfix-sb1 -mno-fix-sb1 @gol
609 -mflush-func=@var{func} -mno-flush-func @gol
610 -mbranch-likely -mno-branch-likely @gol
611 -mfp-exceptions -mno-fp-exceptions @gol
612 -mvr4130-align -mno-vr4130-align}
615 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
616 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
617 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
618 -mno-base-addresses -msingle-exit -mno-single-exit}
620 @emph{MN10300 Options}
621 @gccoptlist{-mmult-bug -mno-mult-bug @gol
622 -mam33 -mno-am33 @gol
623 -mam33-2 -mno-am33-2 @gol
624 -mreturn-pointer-on-d0 @gol
628 @gccoptlist{-mno-crt0 -mbacc -msim @gol
629 -march=@var{cpu-type} }
631 @emph{PDP-11 Options}
632 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
633 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
634 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
635 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
636 -mbranch-expensive -mbranch-cheap @gol
637 -msplit -mno-split -munix-asm -mdec-asm}
639 @emph{PowerPC Options}
640 See RS/6000 and PowerPC Options.
642 @emph{RS/6000 and PowerPC Options}
643 @gccoptlist{-mcpu=@var{cpu-type} @gol
644 -mtune=@var{cpu-type} @gol
645 -mpower -mno-power -mpower2 -mno-power2 @gol
646 -mpowerpc -mpowerpc64 -mno-powerpc @gol
647 -maltivec -mno-altivec @gol
648 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
649 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
650 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
651 -mnew-mnemonics -mold-mnemonics @gol
652 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
653 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
654 -malign-power -malign-natural @gol
655 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
656 -mstring -mno-string -mupdate -mno-update @gol
657 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
658 -mstrict-align -mno-strict-align -mrelocatable @gol
659 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
660 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
661 -mdynamic-no-pic -maltivec -mswdiv @gol
662 -mprioritize-restricted-insns=@var{priority} @gol
663 -msched-costly-dep=@var{dependence_type} @gol
664 -minsert-sched-nops=@var{scheme} @gol
665 -mcall-sysv -mcall-netbsd @gol
666 -maix-struct-return -msvr4-struct-return @gol
667 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
668 -misel -mno-isel @gol
669 -misel=yes -misel=no @gol
671 -mspe=yes -mspe=no @gol
672 -mvrsave -mno-vrsave @gol
673 -mmulhw -mno-mulhw @gol
674 -mdlmzb -mno-dlmzb @gol
675 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
676 -mprototype -mno-prototype @gol
677 -msim -mmvme -mads -myellowknife -memb -msdata @gol
678 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
680 @emph{S/390 and zSeries Options}
681 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
682 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
683 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
684 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
685 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
686 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
687 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
690 @gccoptlist{-meb -mel @gol
694 -mscore5 -mscore5u -mscore7 -mscore7d}
697 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
698 -m4-nofpu -m4-single-only -m4-single -m4 @gol
699 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
700 -m5-64media -m5-64media-nofpu @gol
701 -m5-32media -m5-32media-nofpu @gol
702 -m5-compact -m5-compact-nofpu @gol
703 -mb -ml -mdalign -mrelax @gol
704 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
705 -mieee -misize -mpadstruct -mspace @gol
706 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
707 -mdivsi3_libfunc=@var{name} @gol
708 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
712 @gccoptlist{-mcpu=@var{cpu-type} @gol
713 -mtune=@var{cpu-type} @gol
714 -mcmodel=@var{code-model} @gol
715 -m32 -m64 -mapp-regs -mno-app-regs @gol
716 -mfaster-structs -mno-faster-structs @gol
717 -mfpu -mno-fpu -mhard-float -msoft-float @gol
718 -mhard-quad-float -msoft-quad-float @gol
719 -mimpure-text -mno-impure-text -mlittle-endian @gol
720 -mstack-bias -mno-stack-bias @gol
721 -munaligned-doubles -mno-unaligned-doubles @gol
722 -mv8plus -mno-v8plus -mvis -mno-vis
723 -threads -pthreads -pthread}
725 @emph{System V Options}
726 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
728 @emph{TMS320C3x/C4x Options}
729 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
730 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
731 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
732 -mparallel-insns -mparallel-mpy -mpreserve-float}
735 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
736 -mprolog-function -mno-prolog-function -mspace @gol
737 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
738 -mapp-regs -mno-app-regs @gol
739 -mdisable-callt -mno-disable-callt @gol
745 @gccoptlist{-mg -mgnu -munix}
747 @emph{x86-64 Options}
748 See i386 and x86-64 Options.
750 @emph{Xstormy16 Options}
753 @emph{Xtensa Options}
754 @gccoptlist{-mconst16 -mno-const16 @gol
755 -mfused-madd -mno-fused-madd @gol
756 -mtext-section-literals -mno-text-section-literals @gol
757 -mtarget-align -mno-target-align @gol
758 -mlongcalls -mno-longcalls}
760 @emph{zSeries Options}
761 See S/390 and zSeries Options.
763 @item Code Generation Options
764 @xref{Code Gen Options,,Options for Code Generation Conventions}.
765 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
766 -ffixed-@var{reg} -fexceptions @gol
767 -fnon-call-exceptions -funwind-tables @gol
768 -fasynchronous-unwind-tables @gol
769 -finhibit-size-directive -finstrument-functions @gol
770 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
771 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
772 -fno-common -fno-ident @gol
773 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
774 -fno-jump-tables @gol
775 -freg-struct-return -fshort-enums @gol
776 -fshort-double -fshort-wchar @gol
777 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
778 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
779 -fargument-alias -fargument-noalias @gol
780 -fargument-noalias-global -fargument-noalias-anything
781 -fleading-underscore -ftls-model=@var{model} @gol
782 -ftrapv -fwrapv -fbounds-check @gol
787 * Overall Options:: Controlling the kind of output:
788 an executable, object files, assembler files,
789 or preprocessed source.
790 * C Dialect Options:: Controlling the variant of C language compiled.
791 * C++ Dialect Options:: Variations on C++.
792 * Language Independent Options:: Controlling how diagnostics should be
794 * Warning Options:: How picky should the compiler be?
795 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
796 * Optimize Options:: How much optimization?
797 * Preprocessor Options:: Controlling header files and macro definitions.
798 Also, getting dependency information for Make.
799 * Assembler Options:: Passing options to the assembler.
800 * Link Options:: Specifying libraries and so on.
801 * Directory Options:: Where to find header files and libraries.
802 Where to find the compiler executable files.
803 * Spec Files:: How to pass switches to sub-processes.
804 * Target Options:: Running a cross-compiler, or an old version of GCC.
807 @node Overall Options
808 @section Options Controlling the Kind of Output
810 Compilation can involve up to four stages: preprocessing, compilation
811 proper, assembly and linking, always in that order. GCC is capable of
812 preprocessing and compiling several files either into several
813 assembler input files, or into one assembler input file; then each
814 assembler input file produces an object file, and linking combines all
815 the object files (those newly compiled, and those specified as input)
816 into an executable file.
818 @cindex file name suffix
819 For any given input file, the file name suffix determines what kind of
824 C source code which must be preprocessed.
827 C source code which should not be preprocessed.
830 C++ source code which should not be preprocessed.
833 C, or C++ header file to be turned into a precompiled header.
837 @itemx @var{file}.cxx
838 @itemx @var{file}.cpp
839 @itemx @var{file}.CPP
840 @itemx @var{file}.c++
842 C++ source code which must be preprocessed. Note that in @samp{.cxx},
843 the last two letters must both be literally @samp{x}. Likewise,
844 @samp{.C} refers to a literal capital C@.
848 C++ header file to be turned into a precompiled header.
851 @itemx @var{file}.for
852 @itemx @var{file}.FOR
853 Fixed form Fortran source code which should not be preprocessed.
856 @itemx @var{file}.fpp
857 @itemx @var{file}.FPP
858 Fixed form Fortran source code which must be preprocessed (with the traditional
862 @itemx @var{file}.f95
863 Free form Fortran source code which should not be preprocessed.
866 @itemx @var{file}.F95
867 Free form Fortran source code which must be preprocessed (with the
868 traditional preprocessor).
870 @c FIXME: Descriptions of Java file types.
877 Ada source code file which contains a library unit declaration (a
878 declaration of a package, subprogram, or generic, or a generic
879 instantiation), or a library unit renaming declaration (a package,
880 generic, or subprogram renaming declaration). Such files are also
883 @itemx @var{file}.adb
884 Ada source code file containing a library unit body (a subprogram or
885 package body). Such files are also called @dfn{bodies}.
887 @c GCC also knows about some suffixes for languages not yet included:
898 Assembler code which must be preprocessed.
901 An object file to be fed straight into linking.
902 Any file name with no recognized suffix is treated this way.
906 You can specify the input language explicitly with the @option{-x} option:
909 @item -x @var{language}
910 Specify explicitly the @var{language} for the following input files
911 (rather than letting the compiler choose a default based on the file
912 name suffix). This option applies to all following input files until
913 the next @option{-x} option. Possible values for @var{language} are:
915 c c-header c-cpp-output
916 c++ c++-header c++-cpp-output
917 assembler assembler-with-cpp
925 Turn off any specification of a language, so that subsequent files are
926 handled according to their file name suffixes (as they are if @option{-x}
927 has not been used at all).
929 @item -pass-exit-codes
930 @opindex pass-exit-codes
931 Normally the @command{gcc} program will exit with the code of 1 if any
932 phase of the compiler returns a non-success return code. If you specify
933 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
934 numerically highest error produced by any phase that returned an error
935 indication. The C, C++, and Fortran frontends return 4, if an internal
936 compiler error is encountered.
939 If you only want some of the stages of compilation, you can use
940 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
941 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
942 @command{gcc} is to stop. Note that some combinations (for example,
943 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
948 Compile or assemble the source files, but do not link. The linking
949 stage simply is not done. The ultimate output is in the form of an
950 object file for each source file.
952 By default, the object file name for a source file is made by replacing
953 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
955 Unrecognized input files, not requiring compilation or assembly, are
960 Stop after the stage of compilation proper; do not assemble. The output
961 is in the form of an assembler code file for each non-assembler input
964 By default, the assembler file name for a source file is made by
965 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
967 Input files that don't require compilation are ignored.
971 Stop after the preprocessing stage; do not run the compiler proper. The
972 output is in the form of preprocessed source code, which is sent to the
975 Input files which don't require preprocessing are ignored.
977 @cindex output file option
980 Place output in file @var{file}. This applies regardless to whatever
981 sort of output is being produced, whether it be an executable file,
982 an object file, an assembler file or preprocessed C code.
984 If @option{-o} is not specified, the default is to put an executable
985 file in @file{a.out}, the object file for
986 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
987 assembler file in @file{@var{source}.s}, a precompiled header file in
988 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
993 Print (on standard error output) the commands executed to run the stages
994 of compilation. Also print the version number of the compiler driver
995 program and of the preprocessor and the compiler proper.
999 Like @option{-v} except the commands are not executed and all command
1000 arguments are quoted. This is useful for shell scripts to capture the
1001 driver-generated command lines.
1005 Use pipes rather than temporary files for communication between the
1006 various stages of compilation. This fails to work on some systems where
1007 the assembler is unable to read from a pipe; but the GNU assembler has
1012 If you are compiling multiple source files, this option tells the driver
1013 to pass all the source files to the compiler at once (for those
1014 languages for which the compiler can handle this). This will allow
1015 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1016 language for which this is supported is C@. If you pass source files for
1017 multiple languages to the driver, using this option, the driver will invoke
1018 the compiler(s) that support IMA once each, passing each compiler all the
1019 source files appropriate for it. For those languages that do not support
1020 IMA this option will be ignored, and the compiler will be invoked once for
1021 each source file in that language. If you use this option in conjunction
1022 with @option{-save-temps}, the compiler will generate multiple
1024 (one for each source file), but only one (combined) @file{.o} or
1029 Print (on the standard output) a description of the command line options
1030 understood by @command{gcc}. If the @option{-v} option is also specified
1031 then @option{--help} will also be passed on to the various processes
1032 invoked by @command{gcc}, so that they can display the command line options
1033 they accept. If the @option{-Wextra} option is also specified then command
1034 line options which have no documentation associated with them will also
1038 @opindex target-help
1039 Print (on the standard output) a description of target specific command
1040 line options for each tool.
1044 Display the version number and copyrights of the invoked GCC@.
1046 @include @value{srcdir}/../libiberty/at-file.texi
1050 @section Compiling C++ Programs
1052 @cindex suffixes for C++ source
1053 @cindex C++ source file suffixes
1054 C++ source files conventionally use one of the suffixes @samp{.C},
1055 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1056 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1057 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1058 files with these names and compiles them as C++ programs even if you
1059 call the compiler the same way as for compiling C programs (usually
1060 with the name @command{gcc}).
1064 However, the use of @command{gcc} does not add the C++ library.
1065 @command{g++} is a program that calls GCC and treats @samp{.c},
1066 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1067 files unless @option{-x} is used, and automatically specifies linking
1068 against the C++ library. This program is also useful when
1069 precompiling a C header file with a @samp{.h} extension for use in C++
1070 compilations. On many systems, @command{g++} is also installed with
1071 the name @command{c++}.
1073 @cindex invoking @command{g++}
1074 When you compile C++ programs, you may specify many of the same
1075 command-line options that you use for compiling programs in any
1076 language; or command-line options meaningful for C and related
1077 languages; or options that are meaningful only for C++ programs.
1078 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1079 explanations of options for languages related to C@.
1080 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1081 explanations of options that are meaningful only for C++ programs.
1083 @node C Dialect Options
1084 @section Options Controlling C Dialect
1085 @cindex dialect options
1086 @cindex language dialect options
1087 @cindex options, dialect
1089 The following options control the dialect of C (or languages derived
1090 from C, such as C++) that the compiler accepts:
1093 @cindex ANSI support
1097 In C mode, support all ISO C90 programs. In C++ mode,
1098 remove GNU extensions that conflict with ISO C++.
1100 This turns off certain features of GCC that are incompatible with ISO
1101 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1102 such as the @code{asm} and @code{typeof} keywords, and
1103 predefined macros such as @code{unix} and @code{vax} that identify the
1104 type of system you are using. It also enables the undesirable and
1105 rarely used ISO trigraph feature. For the C compiler,
1106 it disables recognition of C++ style @samp{//} comments as well as
1107 the @code{inline} keyword.
1109 The alternate keywords @code{__asm__}, @code{__extension__},
1110 @code{__inline__} and @code{__typeof__} continue to work despite
1111 @option{-ansi}. You would not want to use them in an ISO C program, of
1112 course, but it is useful to put them in header files that might be included
1113 in compilations done with @option{-ansi}. Alternate predefined macros
1114 such as @code{__unix__} and @code{__vax__} are also available, with or
1115 without @option{-ansi}.
1117 The @option{-ansi} option does not cause non-ISO programs to be
1118 rejected gratuitously. For that, @option{-pedantic} is required in
1119 addition to @option{-ansi}. @xref{Warning Options}.
1121 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1122 option is used. Some header files may notice this macro and refrain
1123 from declaring certain functions or defining certain macros that the
1124 ISO standard doesn't call for; this is to avoid interfering with any
1125 programs that might use these names for other things.
1127 Functions which would normally be built in but do not have semantics
1128 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1129 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1130 built-in functions provided by GCC}, for details of the functions
1135 Determine the language standard. This option is currently only
1136 supported when compiling C or C++. A value for this option must be
1137 provided; possible values are
1142 ISO C90 (same as @option{-ansi}).
1144 @item iso9899:199409
1145 ISO C90 as modified in amendment 1.
1151 ISO C99. Note that this standard is not yet fully supported; see
1152 @w{@uref{http://gcc.gnu.org/gcc-4.2/c99status.html}} for more information. The
1153 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1156 Default, ISO C90 plus GNU extensions (including some C99 features).
1160 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1161 this will become the default. The name @samp{gnu9x} is deprecated.
1164 The 1998 ISO C++ standard plus amendments.
1167 The same as @option{-std=c++98} plus GNU extensions. This is the
1168 default for C++ code.
1171 Even when this option is not specified, you can still use some of the
1172 features of newer standards in so far as they do not conflict with
1173 previous C standards. For example, you may use @code{__restrict__} even
1174 when @option{-std=c99} is not specified.
1176 The @option{-std} options specifying some version of ISO C have the same
1177 effects as @option{-ansi}, except that features that were not in ISO C90
1178 but are in the specified version (for example, @samp{//} comments and
1179 the @code{inline} keyword in ISO C99) are not disabled.
1181 @xref{Standards,,Language Standards Supported by GCC}, for details of
1182 these standard versions.
1184 @item -fgnu89-inline
1185 @opindex fgnu89-inline
1186 The option @option{-fgnu89-inline} tells GCC to use the traditional
1187 GNU semantics for @code{inline} functions when in C99 mode.
1188 @xref{Inline,,An Inline Function is As Fast As a Macro}. Using this
1189 option is roughly equivalent to adding the @code{gnu_inline} function
1190 attribute to all inline functions (@pxref{Function Attributes}).
1192 This option is accepted by GCC versions 4.1.3 and up. In GCC versions
1193 prior to 4.3, C99 inline semantics are not supported, and thus this
1194 option is effectively assumed to be present regardless of whether or not
1195 it is specified; the only effect of specifying it explicitly is to
1196 disable warnings about using inline functions in C99 mode. Likewise,
1197 the option @option{-fno-gnu89-inline} is not supported in versions of
1198 GCC before 4.3. It will be supported only in C99 or gnu99 mode, not in
1201 The preprocesor macros @code{__GNUC_GNU_INLINE__} and
1202 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1203 in effect for @code{inline} functions. @xref{Common Predefined
1204 Macros,,,cpp,The C Preprocessor}.
1206 @item -aux-info @var{filename}
1208 Output to the given filename prototyped declarations for all functions
1209 declared and/or defined in a translation unit, including those in header
1210 files. This option is silently ignored in any language other than C@.
1212 Besides declarations, the file indicates, in comments, the origin of
1213 each declaration (source file and line), whether the declaration was
1214 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1215 @samp{O} for old, respectively, in the first character after the line
1216 number and the colon), and whether it came from a declaration or a
1217 definition (@samp{C} or @samp{F}, respectively, in the following
1218 character). In the case of function definitions, a K&R-style list of
1219 arguments followed by their declarations is also provided, inside
1220 comments, after the declaration.
1224 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1225 keyword, so that code can use these words as identifiers. You can use
1226 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1227 instead. @option{-ansi} implies @option{-fno-asm}.
1229 In C++, this switch only affects the @code{typeof} keyword, since
1230 @code{asm} and @code{inline} are standard keywords. You may want to
1231 use the @option{-fno-gnu-keywords} flag instead, which has the same
1232 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1233 switch only affects the @code{asm} and @code{typeof} keywords, since
1234 @code{inline} is a standard keyword in ISO C99.
1236 @c APPLE LOCAL begin blocks 7205047 5811887
1239 Disable the use of blocks. In @option{-std=c99} mode, blocks are
1240 turned off by default. @option{-fblocks} can be used to re-enable the
1241 feature, if off. Runtime support for blocks first appeared in Mac OS
1242 X 10.6. When targeting 10.6 (see @option{-mmacosx-version-min}) and
1243 later, the extension is on by default.
1244 @c APPLE LOCAL end blocks 7205047 5811887
1247 @itemx -fno-builtin-@var{function}
1248 @opindex fno-builtin
1249 @cindex built-in functions
1250 Don't recognize built-in functions that do not begin with
1251 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1252 functions provided by GCC}, for details of the functions affected,
1253 including those which are not built-in functions when @option{-ansi} or
1254 @option{-std} options for strict ISO C conformance are used because they
1255 do not have an ISO standard meaning.
1257 GCC normally generates special code to handle certain built-in functions
1258 more efficiently; for instance, calls to @code{alloca} may become single
1259 instructions that adjust the stack directly, and calls to @code{memcpy}
1260 may become inline copy loops. The resulting code is often both smaller
1261 and faster, but since the function calls no longer appear as such, you
1262 cannot set a breakpoint on those calls, nor can you change the behavior
1263 of the functions by linking with a different library. In addition,
1264 when a function is recognized as a built-in function, GCC may use
1265 information about that function to warn about problems with calls to
1266 that function, or to generate more efficient code, even if the
1267 resulting code still contains calls to that function. For example,
1268 warnings are given with @option{-Wformat} for bad calls to
1269 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1270 known not to modify global memory.
1272 With the @option{-fno-builtin-@var{function}} option
1273 only the built-in function @var{function} is
1274 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1275 function is named this is not built-in in this version of GCC, this
1276 option is ignored. There is no corresponding
1277 @option{-fbuiltin-@var{function}} option; if you wish to enable
1278 built-in functions selectively when using @option{-fno-builtin} or
1279 @option{-ffreestanding}, you may define macros such as:
1282 #define abs(n) __builtin_abs ((n))
1283 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1288 @cindex hosted environment
1290 Assert that compilation takes place in a hosted environment. This implies
1291 @option{-fbuiltin}. A hosted environment is one in which the
1292 entire standard library is available, and in which @code{main} has a return
1293 type of @code{int}. Examples are nearly everything except a kernel.
1294 This is equivalent to @option{-fno-freestanding}.
1296 @item -ffreestanding
1297 @opindex ffreestanding
1298 @cindex hosted environment
1300 Assert that compilation takes place in a freestanding environment. This
1301 implies @option{-fno-builtin}. A freestanding environment
1302 is one in which the standard library may not exist, and program startup may
1303 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1304 This is equivalent to @option{-fno-hosted}.
1306 @xref{Standards,,Language Standards Supported by GCC}, for details of
1307 freestanding and hosted environments.
1311 @cindex openmp parallel
1312 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1313 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1314 compiler generates parallel code according to the OpenMP Application
1315 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}.
1317 @item -fms-extensions
1318 @opindex fms-extensions
1319 Accept some non-standard constructs used in Microsoft header files.
1321 Some cases of unnamed fields in structures and unions are only
1322 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1323 fields within structs/unions}, for details.
1327 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1328 options for strict ISO C conformance) implies @option{-trigraphs}.
1330 @item -no-integrated-cpp
1331 @opindex no-integrated-cpp
1332 Performs a compilation in two passes: preprocessing and compiling. This
1333 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1334 @option{-B} option. The user supplied compilation step can then add in
1335 an additional preprocessing step after normal preprocessing but before
1336 compiling. The default is to use the integrated cpp (internal cpp)
1338 The semantics of this option will change if "cc1", "cc1plus", and
1339 "cc1obj" are merged.
1341 @cindex traditional C language
1342 @cindex C language, traditional
1344 @itemx -traditional-cpp
1345 @opindex traditional-cpp
1346 @opindex traditional
1347 Formerly, these options caused GCC to attempt to emulate a pre-standard
1348 C compiler. They are now only supported with the @option{-E} switch.
1349 The preprocessor continues to support a pre-standard mode. See the GNU
1350 CPP manual for details.
1352 @c APPLE LOCAL begin nested functions 4357979
1353 @item -fno-nested-functions
1354 @opindex fno-nested-functions
1355 Disable nested functions. This option is not supported for C++ or
1356 Objective-C++. On FreeBSD, nested functions are disabled by default.
1357 @c APPLE LOCAL end nested functions 4357979
1359 @item -fcond-mismatch
1360 @opindex fcond-mismatch
1361 Allow conditional expressions with mismatched types in the second and
1362 third arguments. The value of such an expression is void. This option
1363 is not supported for C++.
1365 @item -flax-vector-conversions
1366 @opindex flax-vector-conversions
1367 Allow implicit conversions between vectors with differing numbers of
1368 elements and/or incompatible element types. This option should not be
1371 @item -funsigned-char
1372 @opindex funsigned-char
1373 Let the type @code{char} be unsigned, like @code{unsigned char}.
1375 Each kind of machine has a default for what @code{char} should
1376 be. It is either like @code{unsigned char} by default or like
1377 @code{signed char} by default.
1379 Ideally, a portable program should always use @code{signed char} or
1380 @code{unsigned char} when it depends on the signedness of an object.
1381 But many programs have been written to use plain @code{char} and
1382 expect it to be signed, or expect it to be unsigned, depending on the
1383 machines they were written for. This option, and its inverse, let you
1384 make such a program work with the opposite default.
1386 The type @code{char} is always a distinct type from each of
1387 @code{signed char} or @code{unsigned char}, even though its behavior
1388 is always just like one of those two.
1391 @opindex fsigned-char
1392 Let the type @code{char} be signed, like @code{signed char}.
1394 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1395 the negative form of @option{-funsigned-char}. Likewise, the option
1396 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1398 @item -fsigned-bitfields
1399 @itemx -funsigned-bitfields
1400 @itemx -fno-signed-bitfields
1401 @itemx -fno-unsigned-bitfields
1402 @opindex fsigned-bitfields
1403 @opindex funsigned-bitfields
1404 @opindex fno-signed-bitfields
1405 @opindex fno-unsigned-bitfields
1406 These options control whether a bit-field is signed or unsigned, when the
1407 declaration does not use either @code{signed} or @code{unsigned}. By
1408 default, such a bit-field is signed, because this is consistent: the
1409 basic integer types such as @code{int} are signed types.
1412 @node C++ Dialect Options
1413 @section Options Controlling C++ Dialect
1415 @cindex compiler options, C++
1416 @cindex C++ options, command line
1417 @cindex options, C++
1418 This section describes the command-line options that are only meaningful
1419 for C++ programs; but you can also use most of the GNU compiler options
1420 regardless of what language your program is in. For example, you
1421 might compile a file @code{firstClass.C} like this:
1424 g++ -g -frepo -O -c firstClass.C
1428 In this example, only @option{-frepo} is an option meant
1429 only for C++ programs; you can use the other options with any
1430 language supported by GCC@.
1432 Here is a list of options that are @emph{only} for compiling C++ programs:
1436 @item -fabi-version=@var{n}
1437 @opindex fabi-version
1438 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1439 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1440 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1441 the version that conforms most closely to the C++ ABI specification.
1442 Therefore, the ABI obtained using version 0 will change as ABI bugs
1445 The default is version 2.
1447 @item -fno-access-control
1448 @opindex fno-access-control
1449 Turn off all access checking. This switch is mainly useful for working
1450 around bugs in the access control code.
1454 Check that the pointer returned by @code{operator new} is non-null
1455 before attempting to modify the storage allocated. This check is
1456 normally unnecessary because the C++ standard specifies that
1457 @code{operator new} will only return @code{0} if it is declared
1458 @samp{throw()}, in which case the compiler will always check the
1459 return value even without this option. In all other cases, when
1460 @code{operator new} has a non-empty exception specification, memory
1461 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1462 @samp{new (nothrow)}.
1464 @item -fconserve-space
1465 @opindex fconserve-space
1466 Put uninitialized or runtime-initialized global variables into the
1467 common segment, as C does. This saves space in the executable at the
1468 cost of not diagnosing duplicate definitions. If you compile with this
1469 flag and your program mysteriously crashes after @code{main()} has
1470 completed, you may have an object that is being destroyed twice because
1471 two definitions were merged.
1473 This option is no longer useful on most targets, now that support has
1474 been added for putting variables into BSS without making them common.
1476 @item -ffriend-injection
1477 @opindex ffriend-injection
1478 Inject friend functions into the enclosing namespace, so that they are
1479 visible outside the scope of the class in which they are declared.
1480 Friend functions were documented to work this way in the old Annotated
1481 C++ Reference Manual, and versions of G++ before 4.1 always worked
1482 that way. However, in ISO C++ a friend function which is not declared
1483 in an enclosing scope can only be found using argument dependent
1484 lookup. This option causes friends to be injected as they were in
1487 This option is for compatibility, and may be removed in a future
1490 @item -fno-elide-constructors
1491 @opindex fno-elide-constructors
1492 The C++ standard allows an implementation to omit creating a temporary
1493 which is only used to initialize another object of the same type.
1494 Specifying this option disables that optimization, and forces G++ to
1495 call the copy constructor in all cases.
1497 @item -fno-enforce-eh-specs
1498 @opindex fno-enforce-eh-specs
1499 Don't generate code to check for violation of exception specifications
1500 at runtime. This option violates the C++ standard, but may be useful
1501 for reducing code size in production builds, much like defining
1502 @samp{NDEBUG}. This does not give user code permission to throw
1503 exceptions in violation of the exception specifications; the compiler
1504 will still optimize based on the specifications, so throwing an
1505 unexpected exception will result in undefined behavior.
1508 @itemx -fno-for-scope
1510 @opindex fno-for-scope
1511 If @option{-ffor-scope} is specified, the scope of variables declared in
1512 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1513 as specified by the C++ standard.
1514 If @option{-fno-for-scope} is specified, the scope of variables declared in
1515 a @i{for-init-statement} extends to the end of the enclosing scope,
1516 as was the case in old versions of G++, and other (traditional)
1517 implementations of C++.
1519 The default if neither flag is given to follow the standard,
1520 but to allow and give a warning for old-style code that would
1521 otherwise be invalid, or have different behavior.
1523 @item -fno-gnu-keywords
1524 @opindex fno-gnu-keywords
1525 Do not recognize @code{typeof} as a keyword, so that code can use this
1526 word as an identifier. You can use the keyword @code{__typeof__} instead.
1527 @option{-ansi} implies @option{-fno-gnu-keywords}.
1529 @item -fno-implicit-templates
1530 @opindex fno-implicit-templates
1531 Never emit code for non-inline templates which are instantiated
1532 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1533 @xref{Template Instantiation}, for more information.
1535 @item -fno-implicit-inline-templates
1536 @opindex fno-implicit-inline-templates
1537 Don't emit code for implicit instantiations of inline templates, either.
1538 The default is to handle inlines differently so that compiles with and
1539 without optimization will need the same set of explicit instantiations.
1541 @item -fno-implement-inlines
1542 @opindex fno-implement-inlines
1543 To save space, do not emit out-of-line copies of inline functions
1544 controlled by @samp{#pragma implementation}. This will cause linker
1545 errors if these functions are not inlined everywhere they are called.
1547 @item -fms-extensions
1548 @opindex fms-extensions
1549 Disable pedantic warnings about constructs used in MFC, such as implicit
1550 int and getting a pointer to member function via non-standard syntax.
1552 @item -fno-nonansi-builtins
1553 @opindex fno-nonansi-builtins
1554 Disable built-in declarations of functions that are not mandated by
1555 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1556 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1558 @item -fno-operator-names
1559 @opindex fno-operator-names
1560 Do not treat the operator name keywords @code{and}, @code{bitand},
1561 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1562 synonyms as keywords.
1564 @item -fno-optional-diags
1565 @opindex fno-optional-diags
1566 Disable diagnostics that the standard says a compiler does not need to
1567 issue. Currently, the only such diagnostic issued by G++ is the one for
1568 a name having multiple meanings within a class.
1571 @opindex fpermissive
1572 Downgrade some diagnostics about nonconformant code from errors to
1573 warnings. Thus, using @option{-fpermissive} will allow some
1574 nonconforming code to compile.
1578 Enable automatic template instantiation at link time. This option also
1579 implies @option{-fno-implicit-templates}. @xref{Template
1580 Instantiation}, for more information.
1584 Disable generation of information about every class with virtual
1585 functions for use by the C++ runtime type identification features
1586 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1587 of the language, you can save some space by using this flag. Note that
1588 exception handling uses the same information, but it will generate it as
1589 needed. The @samp{dynamic_cast} operator can still be used for casts that
1590 do not require runtime type information, i.e. casts to @code{void *} or to
1591 unambiguous base classes.
1595 Emit statistics about front-end processing at the end of the compilation.
1596 This information is generally only useful to the G++ development team.
1598 @item -ftemplate-depth-@var{n}
1599 @opindex ftemplate-depth
1600 Set the maximum instantiation depth for template classes to @var{n}.
1601 A limit on the template instantiation depth is needed to detect
1602 endless recursions during template class instantiation. ANSI/ISO C++
1603 conforming programs must not rely on a maximum depth greater than 17.
1605 @item -fno-threadsafe-statics
1606 @opindex fno-threadsafe-statics
1607 Do not emit the extra code to use the routines specified in the C++
1608 ABI for thread-safe initialization of local statics. You can use this
1609 option to reduce code size slightly in code that doesn't need to be
1612 @item -fuse-cxa-atexit
1613 @opindex fuse-cxa-atexit
1614 Register destructors for objects with static storage duration with the
1615 @code{__cxa_atexit} function rather than the @code{atexit} function.
1616 This option is required for fully standards-compliant handling of static
1617 destructors, but will only work if your C library supports
1618 @code{__cxa_atexit}.
1620 @item -fno-use-cxa-get-exception-ptr
1621 @opindex fno-use-cxa-get-exception-ptr
1622 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1623 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1624 if the runtime routine is not available.
1626 @item -fvisibility-inlines-hidden
1627 @opindex fvisibility-inlines-hidden
1628 This switch declares that the user does not attempt to compare
1629 pointers to inline methods where the addresses of the two functions
1630 were taken in different shared objects.
1632 The effect of this is that GCC may, effectively, mark inline methods with
1633 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1634 appear in the export table of a DSO and do not require a PLT indirection
1635 when used within the DSO@. Enabling this option can have a dramatic effect
1636 on load and link times of a DSO as it massively reduces the size of the
1637 dynamic export table when the library makes heavy use of templates.
1639 The behaviour of this switch is not quite the same as marking the
1640 methods as hidden directly, because it does not affect static variables
1641 local to the function or cause the compiler to deduce that
1642 the function is defined in only one shared object.
1644 You may mark a method as having a visibility explicitly to negate the
1645 effect of the switch for that method. For example, if you do want to
1646 compare pointers to a particular inline method, you might mark it as
1647 having default visibility. Marking the enclosing class with explicit
1648 visibility will have no effect.
1650 Explicitly instantiated inline methods are unaffected by this option
1651 as their linkage might otherwise cross a shared library boundary.
1652 @xref{Template Instantiation}.
1654 @item -fvisibility-ms-compat
1655 @opindex fvisibility-ms-compat
1656 This flag attempts to use visibility settings to make GCC's C++
1657 linkage model compatible with that of Microsoft Visual Studio.
1659 The flag makes these changes to GCC's linkage model:
1663 It sets the default visibility to @code{hidden}, like
1664 @option{-fvisibility=hidden}.
1667 Types, but not their members, are not hidden by default.
1670 The One Definition Rule is relaxed for types without explicit
1671 visibility specifications which are defined in more than one different
1672 shared object: those declarations are permitted if they would have
1673 been permitted when this option was not used.
1676 In new code it is better to use @option{-fvisibility=hidden} and
1677 export those classes which are intended to be externally visible.
1678 Unfortunately it is possible for code to rely, perhaps accidentally,
1679 on the Visual Studio behaviour.
1681 Among the consequences of these changes are that static data members
1682 of the same type with the same name but defined in different shared
1683 objects will be different, so changing one will not change the other;
1684 and that pointers to function members defined in different shared
1685 objects may not compare equal. When this flag is given, it is a
1686 violation of the ODR to define types with the same name differently.
1690 Do not use weak symbol support, even if it is provided by the linker.
1691 By default, G++ will use weak symbols if they are available. This
1692 option exists only for testing, and should not be used by end-users;
1693 it will result in inferior code and has no benefits. This option may
1694 be removed in a future release of G++.
1698 Do not search for header files in the standard directories specific to
1699 C++, but do still search the other standard directories. (This option
1700 is used when building the C++ library.)
1703 In addition, these optimization, warning, and code generation options
1704 have meanings only for C++ programs:
1707 @item -fno-default-inline
1708 @opindex fno-default-inline
1709 Do not assume @samp{inline} for functions defined inside a class scope.
1710 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1711 functions will have linkage like inline functions; they just won't be
1714 @item -Wabi @r{(C++ only)}
1716 Warn when G++ generates code that is probably not compatible with the
1717 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1718 all such cases, there are probably some cases that are not warned about,
1719 even though G++ is generating incompatible code. There may also be
1720 cases where warnings are emitted even though the code that is generated
1723 You should rewrite your code to avoid these warnings if you are
1724 concerned about the fact that code generated by G++ may not be binary
1725 compatible with code generated by other compilers.
1727 The known incompatibilities at this point include:
1732 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1733 pack data into the same byte as a base class. For example:
1736 struct A @{ virtual void f(); int f1 : 1; @};
1737 struct B : public A @{ int f2 : 1; @};
1741 In this case, G++ will place @code{B::f2} into the same byte
1742 as@code{A::f1}; other compilers will not. You can avoid this problem
1743 by explicitly padding @code{A} so that its size is a multiple of the
1744 byte size on your platform; that will cause G++ and other compilers to
1745 layout @code{B} identically.
1748 Incorrect handling of tail-padding for virtual bases. G++ does not use
1749 tail padding when laying out virtual bases. For example:
1752 struct A @{ virtual void f(); char c1; @};
1753 struct B @{ B(); char c2; @};
1754 struct C : public A, public virtual B @{@};
1758 In this case, G++ will not place @code{B} into the tail-padding for
1759 @code{A}; other compilers will. You can avoid this problem by
1760 explicitly padding @code{A} so that its size is a multiple of its
1761 alignment (ignoring virtual base classes); that will cause G++ and other
1762 compilers to layout @code{C} identically.
1765 Incorrect handling of bit-fields with declared widths greater than that
1766 of their underlying types, when the bit-fields appear in a union. For
1770 union U @{ int i : 4096; @};
1774 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1775 union too small by the number of bits in an @code{int}.
1778 Empty classes can be placed at incorrect offsets. For example:
1788 struct C : public B, public A @{@};
1792 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1793 it should be placed at offset zero. G++ mistakenly believes that the
1794 @code{A} data member of @code{B} is already at offset zero.
1797 Names of template functions whose types involve @code{typename} or
1798 template template parameters can be mangled incorrectly.
1801 template <typename Q>
1802 void f(typename Q::X) @{@}
1804 template <template <typename> class Q>
1805 void f(typename Q<int>::X) @{@}
1809 Instantiations of these templates may be mangled incorrectly.
1813 @item -Wctor-dtor-privacy @r{(C++ only)}
1814 @opindex Wctor-dtor-privacy
1815 Warn when a class seems unusable because all the constructors or
1816 destructors in that class are private, and it has neither friends nor
1817 public static member functions.
1819 @item -Wnon-virtual-dtor @r{(C++ only)}
1820 @opindex Wnon-virtual-dtor
1821 Warn when a class appears to be polymorphic, thereby requiring a virtual
1822 destructor, yet it declares a non-virtual one. This warning is also
1823 enabled if -Weffc++ is specified.
1825 @item -Wreorder @r{(C++ only)}
1827 @cindex reordering, warning
1828 @cindex warning for reordering of member initializers
1829 Warn when the order of member initializers given in the code does not
1830 match the order in which they must be executed. For instance:
1836 A(): j (0), i (1) @{ @}
1840 The compiler will rearrange the member initializers for @samp{i}
1841 and @samp{j} to match the declaration order of the members, emitting
1842 a warning to that effect. This warning is enabled by @option{-Wall}.
1845 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1848 @item -Weffc++ @r{(C++ only)}
1850 Warn about violations of the following style guidelines from Scott Meyers'
1851 @cite{Effective C++} book:
1855 Item 11: Define a copy constructor and an assignment operator for classes
1856 with dynamically allocated memory.
1859 Item 12: Prefer initialization to assignment in constructors.
1862 Item 14: Make destructors virtual in base classes.
1865 Item 15: Have @code{operator=} return a reference to @code{*this}.
1868 Item 23: Don't try to return a reference when you must return an object.
1872 Also warn about violations of the following style guidelines from
1873 Scott Meyers' @cite{More Effective C++} book:
1877 Item 6: Distinguish between prefix and postfix forms of increment and
1878 decrement operators.
1881 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1885 When selecting this option, be aware that the standard library
1886 headers do not obey all of these guidelines; use @samp{grep -v}
1887 to filter out those warnings.
1889 @item -Wno-deprecated @r{(C++ only)}
1890 @opindex Wno-deprecated
1891 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1893 @item -Wstrict-null-sentinel @r{(C++ only)}
1894 @opindex Wstrict-null-sentinel
1895 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1896 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1897 to @code{__null}. Although it is a null pointer constant not a null pointer,
1898 it is guaranteed to of the same size as a pointer. But this use is
1899 not portable across different compilers.
1901 @item -Wno-non-template-friend @r{(C++ only)}
1902 @opindex Wno-non-template-friend
1903 Disable warnings when non-templatized friend functions are declared
1904 within a template. Since the advent of explicit template specification
1905 support in G++, if the name of the friend is an unqualified-id (i.e.,
1906 @samp{friend foo(int)}), the C++ language specification demands that the
1907 friend declare or define an ordinary, nontemplate function. (Section
1908 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1909 could be interpreted as a particular specialization of a templatized
1910 function. Because this non-conforming behavior is no longer the default
1911 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1912 check existing code for potential trouble spots and is on by default.
1913 This new compiler behavior can be turned off with
1914 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1915 but disables the helpful warning.
1917 @item -Wold-style-cast @r{(C++ only)}
1918 @opindex Wold-style-cast
1919 Warn if an old-style (C-style) cast to a non-void type is used within
1920 a C++ program. The new-style casts (@samp{dynamic_cast},
1921 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
1922 less vulnerable to unintended effects and much easier to search for.
1924 @item -Woverloaded-virtual @r{(C++ only)}
1925 @opindex Woverloaded-virtual
1926 @cindex overloaded virtual fn, warning
1927 @cindex warning for overloaded virtual fn
1928 Warn when a function declaration hides virtual functions from a
1929 base class. For example, in:
1936 struct B: public A @{
1941 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1949 will fail to compile.
1951 @item -Wno-pmf-conversions @r{(C++ only)}
1952 @opindex Wno-pmf-conversions
1953 Disable the diagnostic for converting a bound pointer to member function
1956 @item -Wsign-promo @r{(C++ only)}
1957 @opindex Wsign-promo
1958 Warn when overload resolution chooses a promotion from unsigned or
1959 enumerated type to a signed type, over a conversion to an unsigned type of
1960 the same size. Previous versions of G++ would try to preserve
1961 unsignedness, but the standard mandates the current behavior.
1966 A& operator = (int);
1976 In this example, G++ will synthesize a default @samp{A& operator =
1977 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1980 @node Language Independent Options
1981 @section Options to Control Diagnostic Messages Formatting
1982 @cindex options to control diagnostics formatting
1983 @cindex diagnostic messages
1984 @cindex message formatting
1986 Traditionally, diagnostic messages have been formatted irrespective of
1987 the output device's aspect (e.g.@: its width, @dots{}). The options described
1988 below can be used to control the diagnostic messages formatting
1989 algorithm, e.g.@: how many characters per line, how often source location
1990 information should be reported. Right now, only the C++ front end can
1991 honor these options. However it is expected, in the near future, that
1992 the remaining front ends would be able to digest them correctly.
1995 @item -fmessage-length=@var{n}
1996 @opindex fmessage-length
1997 Try to format error messages so that they fit on lines of about @var{n}
1998 characters. The default is 72 characters for @command{g++} and 0 for the rest of
1999 the front ends supported by GCC@. If @var{n} is zero, then no
2000 line-wrapping will be done; each error message will appear on a single
2003 @opindex fdiagnostics-show-location
2004 @item -fdiagnostics-show-location=once
2005 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2006 reporter to emit @emph{once} source location information; that is, in
2007 case the message is too long to fit on a single physical line and has to
2008 be wrapped, the source location won't be emitted (as prefix) again,
2009 over and over, in subsequent continuation lines. This is the default
2012 @item -fdiagnostics-show-location=every-line
2013 Only meaningful in line-wrapping mode. Instructs the diagnostic
2014 messages reporter to emit the same source location information (as
2015 prefix) for physical lines that result from the process of breaking
2016 a message which is too long to fit on a single line.
2018 @item -fdiagnostics-show-option
2019 @opindex fdiagnostics-show-option
2020 This option instructs the diagnostic machinery to add text to each
2021 diagnostic emitted, which indicates which command line option directly
2022 controls that diagnostic, when such an option is known to the
2023 diagnostic machinery.
2027 @node Warning Options
2028 @section Options to Request or Suppress Warnings
2029 @cindex options to control warnings
2030 @cindex warning messages
2031 @cindex messages, warning
2032 @cindex suppressing warnings
2034 Warnings are diagnostic messages that report constructions which
2035 are not inherently erroneous but which are risky or suggest there
2036 may have been an error.
2038 You can request many specific warnings with options beginning @samp{-W},
2039 for example @option{-Wimplicit} to request warnings on implicit
2040 declarations. Each of these specific warning options also has a
2041 negative form beginning @samp{-Wno-} to turn off warnings;
2042 for example, @option{-Wno-implicit}. This manual lists only one of the
2043 two forms, whichever is not the default.
2045 The following options control the amount and kinds of warnings produced
2046 by GCC; for further, language-specific options also refer to
2047 @ref{C++ Dialect Options}.
2050 @cindex syntax checking
2052 @opindex fsyntax-only
2053 Check the code for syntax errors, but don't do anything beyond that.
2057 Issue all the warnings demanded by strict ISO C and ISO C++;
2058 reject all programs that use forbidden extensions, and some other
2059 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2060 version of the ISO C standard specified by any @option{-std} option used.
2062 Valid ISO C and ISO C++ programs should compile properly with or without
2063 this option (though a rare few will require @option{-ansi} or a
2064 @option{-std} option specifying the required version of ISO C)@. However,
2065 without this option, certain GNU extensions and traditional C and C++
2066 features are supported as well. With this option, they are rejected.
2068 @option{-pedantic} does not cause warning messages for use of the
2069 alternate keywords whose names begin and end with @samp{__}. Pedantic
2070 warnings are also disabled in the expression that follows
2071 @code{__extension__}. However, only system header files should use
2072 these escape routes; application programs should avoid them.
2073 @xref{Alternate Keywords}.
2075 Some users try to use @option{-pedantic} to check programs for strict ISO
2076 C conformance. They soon find that it does not do quite what they want:
2077 it finds some non-ISO practices, but not all---only those for which
2078 ISO C @emph{requires} a diagnostic, and some others for which
2079 diagnostics have been added.
2081 A feature to report any failure to conform to ISO C might be useful in
2082 some instances, but would require considerable additional work and would
2083 be quite different from @option{-pedantic}. We don't have plans to
2084 support such a feature in the near future.
2086 Where the standard specified with @option{-std} represents a GNU
2087 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2088 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2089 extended dialect is based. Warnings from @option{-pedantic} are given
2090 where they are required by the base standard. (It would not make sense
2091 for such warnings to be given only for features not in the specified GNU
2092 C dialect, since by definition the GNU dialects of C include all
2093 features the compiler supports with the given option, and there would be
2094 nothing to warn about.)
2096 @item -pedantic-errors
2097 @opindex pedantic-errors
2098 Like @option{-pedantic}, except that errors are produced rather than
2103 Inhibit all warning messages.
2107 Inhibit warning messages about the use of @samp{#import}.
2109 @c APPLE LOCAL begin -Wnewline-eof 2001-08-23 --sts **
2111 @opindex Wnewline-eof
2112 Warn about files missing a newline at the end of the file. (Apple compatible)
2113 @c APPLE LOCAL end -Wnewline-eof 2001-08-23 --sts **
2115 @item -Wchar-subscripts
2116 @opindex Wchar-subscripts
2117 Warn if an array subscript has type @code{char}. This is a common cause
2118 of error, as programmers often forget that this type is signed on some
2120 This warning is enabled by @option{-Wall}.
2124 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2125 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2126 This warning is enabled by @option{-Wall}.
2128 @item -Wfatal-errors
2129 @opindex Wfatal-errors
2130 This option causes the compiler to abort compilation on the first error
2131 occurred rather than trying to keep going and printing further error
2136 @opindex ffreestanding
2137 @opindex fno-builtin
2138 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2139 the arguments supplied have types appropriate to the format string
2140 specified, and that the conversions specified in the format string make
2141 sense. This includes standard functions, and others specified by format
2142 attributes (@pxref{Function Attributes}), in the @code{printf},
2143 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2144 not in the C standard) families (or other target-specific families).
2145 Which functions are checked without format attributes having been
2146 specified depends on the standard version selected, and such checks of
2147 functions without the attribute specified are disabled by
2148 @option{-ffreestanding} or @option{-fno-builtin}.
2150 The formats are checked against the format features supported by GNU
2151 libc version 2.2. These include all ISO C90 and C99 features, as well
2152 as features from the Single Unix Specification and some BSD and GNU
2153 extensions. Other library implementations may not support all these
2154 features; GCC does not support warning about features that go beyond a
2155 particular library's limitations. However, if @option{-pedantic} is used
2156 with @option{-Wformat}, warnings will be given about format features not
2157 in the selected standard version (but not for @code{strfmon} formats,
2158 since those are not in any version of the C standard). @xref{C Dialect
2159 Options,,Options Controlling C Dialect}.
2161 Since @option{-Wformat} also checks for null format arguments for
2162 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2164 @option{-Wformat} is included in @option{-Wall}. For more control over some
2165 aspects of format checking, the options @option{-Wformat-y2k},
2166 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2167 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2168 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2171 @opindex Wformat-y2k
2172 If @option{-Wformat} is specified, also warn about @code{strftime}
2173 formats which may yield only a two-digit year.
2175 @item -Wno-format-extra-args
2176 @opindex Wno-format-extra-args
2177 If @option{-Wformat} is specified, do not warn about excess arguments to a
2178 @code{printf} or @code{scanf} format function. The C standard specifies
2179 that such arguments are ignored.
2181 Where the unused arguments lie between used arguments that are
2182 specified with @samp{$} operand number specifications, normally
2183 warnings are still given, since the implementation could not know what
2184 type to pass to @code{va_arg} to skip the unused arguments. However,
2185 in the case of @code{scanf} formats, this option will suppress the
2186 warning if the unused arguments are all pointers, since the Single
2187 Unix Specification says that such unused arguments are allowed.
2189 @item -Wno-format-zero-length
2190 @opindex Wno-format-zero-length
2191 If @option{-Wformat} is specified, do not warn about zero-length formats.
2192 The C standard specifies that zero-length formats are allowed.
2194 @item -Wformat-nonliteral
2195 @opindex Wformat-nonliteral
2196 If @option{-Wformat} is specified, also warn if the format string is not a
2197 string literal and so cannot be checked, unless the format function
2198 takes its format arguments as a @code{va_list}.
2200 @item -Wformat-security
2201 @opindex Wformat-security
2202 If @option{-Wformat} is specified, also warn about uses of format
2203 functions that represent possible security problems. At present, this
2204 warns about calls to @code{printf} and @code{scanf} functions where the
2205 format string is not a string literal and there are no format arguments,
2206 as in @code{printf (foo);}. This may be a security hole if the format
2207 string came from untrusted input and contains @samp{%n}. (This is
2208 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2209 in future warnings may be added to @option{-Wformat-security} that are not
2210 included in @option{-Wformat-nonliteral}.)
2214 Enable @option{-Wformat} plus format checks not included in
2215 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2216 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2220 Warn about passing a null pointer for arguments marked as
2221 requiring a non-null value by the @code{nonnull} function attribute.
2223 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2224 can be disabled with the @option{-Wno-nonnull} option.
2226 @item -Winit-self @r{(C and C++ only)}
2228 Warn about uninitialized variables which are initialized with themselves.
2229 Note this option can only be used with the @option{-Wuninitialized} option,
2230 which in turn only works with @option{-O1} and above.
2232 For example, GCC will warn about @code{i} being uninitialized in the
2233 following snippet only when @option{-Winit-self} has been specified:
2244 @item -Wimplicit-int
2245 @opindex Wimplicit-int
2246 Warn when a declaration does not specify a type.
2247 This warning is enabled by @option{-Wall}.
2249 @item -Wimplicit-function-declaration
2250 @itemx -Werror-implicit-function-declaration
2251 @opindex Wimplicit-function-declaration
2252 @opindex Werror-implicit-function-declaration
2253 Give a warning (or error) whenever a function is used before being
2254 declared. The form @option{-Wno-error-implicit-function-declaration}
2256 This warning is enabled by @option{-Wall} (as a warning, not an error).
2260 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2261 This warning is enabled by @option{-Wall}.
2265 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2266 function with external linkage, returning int, taking either zero
2267 arguments, two, or three arguments of appropriate types.
2268 This warning is enabled by @option{-Wall}.
2270 @item -Wmissing-braces
2271 @opindex Wmissing-braces
2272 Warn if an aggregate or union initializer is not fully bracketed. In
2273 the following example, the initializer for @samp{a} is not fully
2274 bracketed, but that for @samp{b} is fully bracketed.
2277 int a[2][2] = @{ 0, 1, 2, 3 @};
2278 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2281 This warning is enabled by @option{-Wall}.
2283 @item -Wmissing-include-dirs @r{(C and C++ only)}
2284 @opindex Wmissing-include-dirs
2285 Warn if a user-supplied include directory does not exist.
2288 @opindex Wparentheses
2289 Warn if parentheses are omitted in certain contexts, such
2290 as when there is an assignment in a context where a truth value
2291 is expected, or when operators are nested whose precedence people
2292 often get confused about.
2294 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2295 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2296 interpretation from that of ordinary mathematical notation.
2298 Also warn about constructions where there may be confusion to which
2299 @code{if} statement an @code{else} branch belongs. Here is an example of
2314 In C/C++, every @code{else} branch belongs to the innermost possible
2315 @code{if} statement, which in this example is @code{if (b)}. This is
2316 often not what the programmer expected, as illustrated in the above
2317 example by indentation the programmer chose. When there is the
2318 potential for this confusion, GCC will issue a warning when this flag
2319 is specified. To eliminate the warning, add explicit braces around
2320 the innermost @code{if} statement so there is no way the @code{else}
2321 could belong to the enclosing @code{if}. The resulting code would
2338 This warning is enabled by @option{-Wall}.
2340 @item -Wsequence-point
2341 @opindex Wsequence-point
2342 Warn about code that may have undefined semantics because of violations
2343 of sequence point rules in the C and C++ standards.
2345 The C and C++ standards defines the order in which expressions in a C/C++
2346 program are evaluated in terms of @dfn{sequence points}, which represent
2347 a partial ordering between the execution of parts of the program: those
2348 executed before the sequence point, and those executed after it. These
2349 occur after the evaluation of a full expression (one which is not part
2350 of a larger expression), after the evaluation of the first operand of a
2351 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2352 function is called (but after the evaluation of its arguments and the
2353 expression denoting the called function), and in certain other places.
2354 Other than as expressed by the sequence point rules, the order of
2355 evaluation of subexpressions of an expression is not specified. All
2356 these rules describe only a partial order rather than a total order,
2357 since, for example, if two functions are called within one expression
2358 with no sequence point between them, the order in which the functions
2359 are called is not specified. However, the standards committee have
2360 ruled that function calls do not overlap.
2362 It is not specified when between sequence points modifications to the
2363 values of objects take effect. Programs whose behavior depends on this
2364 have undefined behavior; the C and C++ standards specify that ``Between
2365 the previous and next sequence point an object shall have its stored
2366 value modified at most once by the evaluation of an expression.
2367 Furthermore, the prior value shall be read only to determine the value
2368 to be stored.''. If a program breaks these rules, the results on any
2369 particular implementation are entirely unpredictable.
2371 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2372 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2373 diagnosed by this option, and it may give an occasional false positive
2374 result, but in general it has been found fairly effective at detecting
2375 this sort of problem in programs.
2377 The standard is worded confusingly, therefore there is some debate
2378 over the precise meaning of the sequence point rules in subtle cases.
2379 Links to discussions of the problem, including proposed formal
2380 definitions, may be found on the GCC readings page, at
2381 @w{@uref{http://gcc.gnu.org/readings.html}}.
2383 This warning is enabled by @option{-Wall} for C and C++.
2386 @opindex Wreturn-type
2387 Warn whenever a function is defined with a return-type that defaults to
2388 @code{int}. Also warn about any @code{return} statement with no
2389 return-value in a function whose return-type is not @code{void}.
2391 For C, also warn if the return type of a function has a type qualifier
2392 such as @code{const}. Such a type qualifier has no effect, since the
2393 value returned by a function is not an lvalue. ISO C prohibits
2394 qualified @code{void} return types on function definitions, so such
2395 return types always receive a warning even without this option.
2397 For C++, a function without return type always produces a diagnostic
2398 message, even when @option{-Wno-return-type} is specified. The only
2399 exceptions are @samp{main} and functions defined in system headers.
2401 This warning is enabled by @option{-Wall}.
2405 Warn whenever a @code{switch} statement has an index of enumerated type
2406 and lacks a @code{case} for one or more of the named codes of that
2407 enumeration. (The presence of a @code{default} label prevents this
2408 warning.) @code{case} labels outside the enumeration range also
2409 provoke warnings when this option is used.
2410 This warning is enabled by @option{-Wall}.
2412 @item -Wswitch-default
2413 @opindex Wswitch-switch
2414 Warn whenever a @code{switch} statement does not have a @code{default}
2418 @opindex Wswitch-enum
2419 Warn whenever a @code{switch} statement has an index of enumerated type
2420 and lacks a @code{case} for one or more of the named codes of that
2421 enumeration. @code{case} labels outside the enumeration range also
2422 provoke warnings when this option is used.
2426 Warn if any trigraphs are encountered that might change the meaning of
2427 the program (trigraphs within comments are not warned about).
2428 This warning is enabled by @option{-Wall}.
2430 @item -Wunused-function
2431 @opindex Wunused-function
2432 Warn whenever a static function is declared but not defined or a
2433 non-inline static function is unused.
2434 This warning is enabled by @option{-Wall}.
2436 @item -Wunused-label
2437 @opindex Wunused-label
2438 Warn whenever a label is declared but not used.
2439 This warning is enabled by @option{-Wall}.
2441 To suppress this warning use the @samp{unused} attribute
2442 (@pxref{Variable Attributes}).
2444 @item -Wunused-parameter
2445 @opindex Wunused-parameter
2446 Warn whenever a function parameter is unused aside from its declaration.
2448 To suppress this warning use the @samp{unused} attribute
2449 (@pxref{Variable Attributes}).
2451 @item -Wunused-variable
2452 @opindex Wunused-variable
2453 Warn whenever a local variable or non-constant static variable is unused
2454 aside from its declaration.
2455 This warning is enabled by @option{-Wall}.
2457 To suppress this warning use the @samp{unused} attribute
2458 (@pxref{Variable Attributes}).
2460 @item -Wunused-value
2461 @opindex Wunused-value
2462 Warn whenever a statement computes a result that is explicitly not used.
2463 This warning is enabled by @option{-Wall}.
2465 To suppress this warning cast the expression to @samp{void}.
2469 All the above @option{-Wunused} options combined.
2471 In order to get a warning about an unused function parameter, you must
2472 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2473 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2475 @item -Wuninitialized
2476 @opindex Wuninitialized
2477 Warn if an automatic variable is used without first being initialized or
2478 if a variable may be clobbered by a @code{setjmp} call.
2480 These warnings are possible only in optimizing compilation,
2481 because they require data flow information that is computed only
2482 when optimizing. If you do not specify @option{-O}, you will not get
2483 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
2484 requiring @option{-O}.
2486 If you want to warn about code which uses the uninitialized value of the
2487 variable in its own initializer, use the @option{-Winit-self} option.
2489 These warnings occur for individual uninitialized or clobbered
2490 elements of structure, union or array variables as well as for
2491 variables which are uninitialized or clobbered as a whole. They do
2492 not occur for variables or elements declared @code{volatile}. Because
2493 these warnings depend on optimization, the exact variables or elements
2494 for which there are warnings will depend on the precise optimization
2495 options and version of GCC used.
2497 Note that there may be no warning about a variable that is used only
2498 to compute a value that itself is never used, because such
2499 computations may be deleted by data flow analysis before the warnings
2502 These warnings are made optional because GCC is not smart
2503 enough to see all the reasons why the code might be correct
2504 despite appearing to have an error. Here is one example of how
2525 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2526 always initialized, but GCC doesn't know this. Here is
2527 another common case:
2532 if (change_y) save_y = y, y = new_y;
2534 if (change_y) y = save_y;
2539 This has no bug because @code{save_y} is used only if it is set.
2541 @cindex @code{longjmp} warnings
2542 This option also warns when a non-volatile automatic variable might be
2543 changed by a call to @code{longjmp}. These warnings as well are possible
2544 only in optimizing compilation.
2546 The compiler sees only the calls to @code{setjmp}. It cannot know
2547 where @code{longjmp} will be called; in fact, a signal handler could
2548 call it at any point in the code. As a result, you may get a warning
2549 even when there is in fact no problem because @code{longjmp} cannot
2550 in fact be called at the place which would cause a problem.
2552 Some spurious warnings can be avoided if you declare all the functions
2553 you use that never return as @code{noreturn}. @xref{Function
2556 This warning is enabled by @option{-Wall}.
2558 @item -Wunknown-pragmas
2559 @opindex Wunknown-pragmas
2560 @cindex warning for unknown pragmas
2561 @cindex unknown pragmas, warning
2562 @cindex pragmas, warning of unknown
2563 Warn when a #pragma directive is encountered which is not understood by
2564 GCC@. If this command line option is used, warnings will even be issued
2565 for unknown pragmas in system header files. This is not the case if
2566 the warnings were only enabled by the @option{-Wall} command line option.
2569 @opindex Wno-pragmas
2571 Do not warn about misuses of pragmas, such as incorrect parameters,
2572 invalid syntax, or conflicts between pragmas. See also
2573 @samp{-Wunknown-pragmas}.
2575 @item -Wstrict-aliasing
2576 @opindex Wstrict-aliasing
2577 This option is only active when @option{-fstrict-aliasing} is active.
2578 It warns about code which might break the strict aliasing rules that the
2579 compiler is using for optimization. The warning does not catch all
2580 cases, but does attempt to catch the more common pitfalls. It is
2581 included in @option{-Wall}.
2582 It is equivalent to -Wstrict-aliasing=3
2584 @item -Wstrict-aliasing=n
2585 @opindex Wstrict-aliasing=n
2586 This option is only active when @option{-fstrict-aliasing} is active.
2587 It warns about code which might break the strict aliasing rules that the
2588 compiler is using for optimization.
2589 Higher levels correspond to higher accuracy (fewer false positives).
2590 Higher levels also correspond to more effort, similar to the way -O works.
2591 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
2594 Level 1: Most aggressive, quick, least accurate.
2595 Possibly useful when higher levels
2596 do not warn but -fstrict-aliasing still breaks the code, as it has very few
2597 false negatives. However, it has many false positives.
2598 Warns for all pointer conversions between possibly incompatible types,
2599 even if never dereferenced. Runs in the frontend only.
2601 Level 2: Aggressive, quick, not too precise.
2602 May still have many false positives (not as many as level 1 though),
2603 and few false negatives (but possibly more than level 1).
2604 Unlike level 1, it only warns when an address is taken. Warns about
2605 incomplete types. Runs in the frontend only.
2607 Level 3 (default for @option{-Wstrict-aliasing}):
2608 Should have very few false positives and few false
2609 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
2610 Takes care of the common punn+dereference pattern in the frontend:
2611 @code{*(int*)&some_float}.
2612 If optimization is enabled, it also runs in the backend, where it deals
2613 with multiple statement cases using flow-sensitive points-to information.
2614 Only warns when the converted pointer is dereferenced.
2615 Does not warn about incomplete types.
2617 @item -Wstrict-overflow
2618 @item -Wstrict-overflow=@var{n}
2619 @opindex Wstrict-overflow
2620 This option is only active when @option{-fstrict-overflow} is active.
2621 It warns about cases where the compiler optimizes based on the
2622 assumption that signed overflow does not occur. Note that it does not
2623 warn about all cases where the code might overflow: it only warns
2624 about cases where the compiler implements some optimization. Thus
2625 this warning depends on the optimization level.
2627 An optimization which assumes that signed overflow does not occur is
2628 perfectly safe if the values of the variables involved are such that
2629 overflow never does, in fact, occur. Therefore this warning can
2630 easily give a false positive: a warning about code which is not
2631 actually a problem. To help focus on important issues, several
2632 warning levels are defined. No warnings are issued for the use of
2633 undefined signed overflow when estimating how many iterations a loop
2634 will require, in particular when determining whether a loop will be
2638 @item -Wstrict-overflow=1
2639 Warn about cases which are both questionable and easy to avoid. For
2640 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
2641 compiler will simplify this to @code{1}. This level of
2642 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
2643 are not, and must be explicitly requested.
2645 @item -Wstrict-overflow=2
2646 Also warn about other cases where a comparison is simplified to a
2647 constant. For example: @code{abs (x) >= 0}. This can only be
2648 simplified when @option{-fstrict-overflow} is in effect, because
2649 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
2650 zero. @option{-Wstrict-overflow} (with no level) is the same as
2651 @option{-Wstrict-overflow=2}.
2653 @item -Wstrict-overflow=3
2654 Also warn about other cases where a comparison is simplified. For
2655 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
2657 @item -Wstrict-overflow=4
2658 Also warn about other simplifications not covered by the above cases.
2659 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
2661 @item -Wstrict-overflow=5
2662 Also warn about cases where the compiler reduces the magnitude of a
2663 constant involved in a comparison. For example: @code{x + 2 > y} will
2664 be simplified to @code{x + 1 >= y}. This is reported only at the
2665 highest warning level because this simplification applies to many
2666 comparisons, so this warning level will give a very large number of
2672 All of the above @samp{-W} options combined. This enables all the
2673 warnings about constructions that some users consider questionable, and
2674 that are easy to avoid (or modify to prevent the warning), even in
2675 conjunction with macros. This also enables some language-specific
2676 warnings described in @ref{C++ Dialect Options}.
2677 @c APPLE LOCAL begin -Wmost
2680 This is equivalent to -Wall -Wno-parentheses. (Apple compatible)
2682 @c APPLE LOCAL end -Wmost
2684 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2685 Some of them warn about constructions that users generally do not
2686 consider questionable, but which occasionally you might wish to check
2687 for; others warn about constructions that are necessary or hard to avoid
2688 in some cases, and there is no simple way to modify the code to suppress
2695 (This option used to be called @option{-W}. The older name is still
2696 supported, but the newer name is more descriptive.) Print extra warning
2697 messages for these events:
2701 A function can return either with or without a value. (Falling
2702 off the end of the function body is considered returning without
2703 a value.) For example, this function would evoke such a
2717 An expression-statement or the left-hand side of a comma expression
2718 contains no side effects.
2719 To suppress the warning, cast the unused expression to void.
2720 For example, an expression such as @samp{x[i,j]} will cause a warning,
2721 but @samp{x[(void)i,j]} will not.
2724 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2727 Storage-class specifiers like @code{static} are not the first things in
2728 a declaration. According to the C Standard, this usage is obsolescent.
2731 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2735 A comparison between signed and unsigned values could produce an
2736 incorrect result when the signed value is converted to unsigned.
2737 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2740 An aggregate has an initializer which does not initialize all members.
2741 This warning can be independently controlled by
2742 @option{-Wmissing-field-initializers}.
2745 An initialized field without side effects is overridden when using
2746 designated initializers (@pxref{Designated Inits, , Designated
2747 Initializers}). This warning can be independently controlled by
2748 @option{-Woverride-init}.
2751 A function parameter is declared without a type specifier in K&R-style
2759 An empty body occurs in an @samp{if} or @samp{else} statement.
2762 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2763 @samp{>}, or @samp{>=}.
2766 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2768 @item @r{(C++ only)}
2769 An enumerator and a non-enumerator both appear in a conditional expression.
2771 @item @r{(C++ only)}
2772 A non-static reference or non-static @samp{const} member appears in a
2773 class without constructors.
2775 @item @r{(C++ only)}
2776 Ambiguous virtual bases.
2778 @item @r{(C++ only)}
2779 Subscripting an array which has been declared @samp{register}.
2781 @item @r{(C++ only)}
2782 Taking the address of a variable which has been declared @samp{register}.
2784 @item @r{(C++ only)}
2785 A base class is not initialized in a derived class' copy constructor.
2788 @item -Wno-div-by-zero
2789 @opindex Wno-div-by-zero
2790 @opindex Wdiv-by-zero
2791 Do not warn about compile-time integer division by zero. Floating point
2792 division by zero is not warned about, as it can be a legitimate way of
2793 obtaining infinities and NaNs.
2795 @item -Wsystem-headers
2796 @opindex Wsystem-headers
2797 @cindex warnings from system headers
2798 @cindex system headers, warnings from
2799 Print warning messages for constructs found in system header files.
2800 Warnings from system headers are normally suppressed, on the assumption
2801 that they usually do not indicate real problems and would only make the
2802 compiler output harder to read. Using this command line option tells
2803 GCC to emit warnings from system headers as if they occurred in user
2804 code. However, note that using @option{-Wall} in conjunction with this
2805 option will @emph{not} warn about unknown pragmas in system
2806 headers---for that, @option{-Wunknown-pragmas} must also be used.
2809 @opindex Wfloat-equal
2810 Warn if floating point values are used in equality comparisons.
2812 The idea behind this is that sometimes it is convenient (for the
2813 programmer) to consider floating-point values as approximations to
2814 infinitely precise real numbers. If you are doing this, then you need
2815 to compute (by analyzing the code, or in some other way) the maximum or
2816 likely maximum error that the computation introduces, and allow for it
2817 when performing comparisons (and when producing output, but that's a
2818 different problem). In particular, instead of testing for equality, you
2819 would check to see whether the two values have ranges that overlap; and
2820 this is done with the relational operators, so equality comparisons are
2823 @item -Wtraditional @r{(C only)}
2824 @opindex Wtraditional
2825 Warn about certain constructs that behave differently in traditional and
2826 ISO C@. Also warn about ISO C constructs that have no traditional C
2827 equivalent, and/or problematic constructs which should be avoided.
2831 Macro parameters that appear within string literals in the macro body.
2832 In traditional C macro replacement takes place within string literals,
2833 but does not in ISO C@.
2836 In traditional C, some preprocessor directives did not exist.
2837 Traditional preprocessors would only consider a line to be a directive
2838 if the @samp{#} appeared in column 1 on the line. Therefore
2839 @option{-Wtraditional} warns about directives that traditional C
2840 understands but would ignore because the @samp{#} does not appear as the
2841 first character on the line. It also suggests you hide directives like
2842 @samp{#pragma} not understood by traditional C by indenting them. Some
2843 traditional implementations would not recognize @samp{#elif}, so it
2844 suggests avoiding it altogether.
2847 A function-like macro that appears without arguments.
2850 The unary plus operator.
2853 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2854 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2855 constants.) Note, these suffixes appear in macros defined in the system
2856 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2857 Use of these macros in user code might normally lead to spurious
2858 warnings, however GCC's integrated preprocessor has enough context to
2859 avoid warning in these cases.
2862 A function declared external in one block and then used after the end of
2866 A @code{switch} statement has an operand of type @code{long}.
2869 A non-@code{static} function declaration follows a @code{static} one.
2870 This construct is not accepted by some traditional C compilers.
2873 The ISO type of an integer constant has a different width or
2874 signedness from its traditional type. This warning is only issued if
2875 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2876 typically represent bit patterns, are not warned about.
2879 Usage of ISO string concatenation is detected.
2882 Initialization of automatic aggregates.
2885 Identifier conflicts with labels. Traditional C lacks a separate
2886 namespace for labels.
2889 Initialization of unions. If the initializer is zero, the warning is
2890 omitted. This is done under the assumption that the zero initializer in
2891 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2892 initializer warnings and relies on default initialization to zero in the
2896 Conversions by prototypes between fixed/floating point values and vice
2897 versa. The absence of these prototypes when compiling with traditional
2898 C would cause serious problems. This is a subset of the possible
2899 conversion warnings, for the full set use @option{-Wconversion}.
2902 Use of ISO C style function definitions. This warning intentionally is
2903 @emph{not} issued for prototype declarations or variadic functions
2904 because these ISO C features will appear in your code when using
2905 libiberty's traditional C compatibility macros, @code{PARAMS} and
2906 @code{VPARAMS}. This warning is also bypassed for nested functions
2907 because that feature is already a GCC extension and thus not relevant to
2908 traditional C compatibility.
2911 @item -Wdeclaration-after-statement @r{(C only)}
2912 @opindex Wdeclaration-after-statement
2913 Warn when a declaration is found after a statement in a block. This
2914 construct, known from C++, was introduced with ISO C99 and is by default
2915 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2916 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2920 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2922 @item -Wno-endif-labels
2923 @opindex Wno-endif-labels
2924 @opindex Wendif-labels
2925 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2929 Warn whenever a local variable shadows another local variable, parameter or
2930 global variable or whenever a built-in function is shadowed.
2932 @item -Wlarger-than-@var{len}
2933 @opindex Wlarger-than
2934 Warn whenever an object of larger than @var{len} bytes is defined.
2936 @item -Wframe-larger-than-@var{len}
2937 @opindex Wframe-larger-than
2938 Warn whenever the frame size of a function is larger than @var{len} bytes.
2940 @item -Wunsafe-loop-optimizations
2941 @opindex Wunsafe-loop-optimizations
2942 Warn if the loop cannot be optimized because the compiler could not
2943 assume anything on the bounds of the loop indices. With
2944 @option{-funsafe-loop-optimizations} warn if the compiler made
2947 @item -Wpointer-arith
2948 @opindex Wpointer-arith
2949 Warn about anything that depends on the ``size of'' a function type or
2950 of @code{void}. GNU C assigns these types a size of 1, for
2951 convenience in calculations with @code{void *} pointers and pointers
2954 @item -Wbad-function-cast @r{(C only)}
2955 @opindex Wbad-function-cast
2956 Warn whenever a function call is cast to a non-matching type.
2957 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2960 Warn about ISO C constructs that are outside of the common subset of
2961 ISO C and ISO C++, e.g.@: request for implicit conversion from
2962 @code{void *} to a pointer to non-@code{void} type.
2966 Warn whenever a pointer is cast so as to remove a type qualifier from
2967 the target type. For example, warn if a @code{const char *} is cast
2968 to an ordinary @code{char *}.
2971 @opindex Wcast-align
2972 Warn whenever a pointer is cast such that the required alignment of the
2973 target is increased. For example, warn if a @code{char *} is cast to
2974 an @code{int *} on machines where integers can only be accessed at
2975 two- or four-byte boundaries.
2977 @item -Wwrite-strings
2978 @opindex Wwrite-strings
2979 When compiling C, give string constants the type @code{const
2980 char[@var{length}]} so that
2981 copying the address of one into a non-@code{const} @code{char *}
2982 pointer will get a warning; when compiling C++, warn about the
2983 deprecated conversion from string literals to @code{char *}. This
2984 warning, by default, is enabled for C++ programs.
2985 These warnings will help you find at
2986 compile time code that can try to write into a string constant, but
2987 only if you have been very careful about using @code{const} in
2988 declarations and prototypes. Otherwise, it will just be a nuisance;
2989 this is why we did not make @option{-Wall} request these warnings.
2992 @opindex Wconversion
2993 Warn if a prototype causes a type conversion that is different from what
2994 would happen to the same argument in the absence of a prototype. This
2995 includes conversions of fixed point to floating and vice versa, and
2996 conversions changing the width or signedness of a fixed point argument
2997 except when the same as the default promotion.
2999 Also, warn if a negative integer constant expression is implicitly
3000 converted to an unsigned type. For example, warn about the assignment
3001 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
3002 casts like @code{(unsigned) -1}.
3004 @item -Wsign-compare
3005 @opindex Wsign-compare
3006 @cindex warning for comparison of signed and unsigned values
3007 @cindex comparison of signed and unsigned values, warning
3008 @cindex signed and unsigned values, comparison warning
3009 Warn when a comparison between signed and unsigned values could produce
3010 an incorrect result when the signed value is converted to unsigned.
3011 This warning is also enabled by @option{-Wextra}; to get the other warnings
3012 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3016 @opindex Wno-address
3017 Warn about suspicious uses of memory addresses. These include using
3018 the address of a function in a conditional expression, such as
3019 @code{void func(void); if (func)}, and comparisons against the memory
3020 address of a string literal, such as @code{if (x == "abc")}. Such
3021 uses typically indicate a programmer error: the address of a function
3022 always evaluates to true, so their use in a conditional usually
3023 indicate that the programmer forgot the parentheses in a function
3024 call; and comparisons against string literals result in unspecified
3025 behavior and are not portable in C, so they usually indicate that the
3026 programmer intended to use @code{strcmp}. This warning is enabled by
3029 @item -Waggregate-return
3030 @opindex Waggregate-return
3031 Warn if any functions that return structures or unions are defined or
3032 called. (In languages where you can return an array, this also elicits
3035 @item -Wno-attributes
3036 @opindex Wno-attributes
3037 @opindex Wattributes
3038 Do not warn if an unexpected @code{__attribute__} is used, such as
3039 unrecognized attributes, function attributes applied to variables,
3040 etc. This will not stop errors for incorrect use of supported
3043 @item -Wstrict-prototypes @r{(C only)}
3044 @opindex Wstrict-prototypes
3045 Warn if a function is declared or defined without specifying the
3046 argument types. (An old-style function definition is permitted without
3047 a warning if preceded by a declaration which specifies the argument
3050 @item -Wold-style-definition @r{(C only)}
3051 @opindex Wold-style-definition
3052 Warn if an old-style function definition is used. A warning is given
3053 even if there is a previous prototype.
3055 @item -Wmissing-prototypes @r{(C only)}
3056 @opindex Wmissing-prototypes
3057 Warn if a global function is defined without a previous prototype
3058 declaration. This warning is issued even if the definition itself
3059 provides a prototype. The aim is to detect global functions that fail
3060 to be declared in header files.
3062 @item -Wmissing-declarations @r{(C only)}
3063 @opindex Wmissing-declarations
3064 Warn if a global function is defined without a previous declaration.
3065 Do so even if the definition itself provides a prototype.
3066 Use this option to detect global functions that are not declared in
3069 @item -Wmissing-field-initializers
3070 @opindex Wmissing-field-initializers
3073 Warn if a structure's initializer has some fields missing. For
3074 example, the following code would cause such a warning, because
3075 @code{x.h} is implicitly zero:
3078 struct s @{ int f, g, h; @};
3079 struct s x = @{ 3, 4 @};
3082 This option does not warn about designated initializers, so the following
3083 modification would not trigger a warning:
3086 struct s @{ int f, g, h; @};
3087 struct s x = @{ .f = 3, .g = 4 @};
3090 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3091 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3093 @item -Wmissing-noreturn
3094 @opindex Wmissing-noreturn
3095 Warn about functions which might be candidates for attribute @code{noreturn}.
3096 Note these are only possible candidates, not absolute ones. Care should
3097 be taken to manually verify functions actually do not ever return before
3098 adding the @code{noreturn} attribute, otherwise subtle code generation
3099 bugs could be introduced. You will not get a warning for @code{main} in
3100 hosted C environments.
3102 @item -Wmissing-format-attribute
3103 @opindex Wmissing-format-attribute
3105 Warn about function pointers which might be candidates for @code{format}
3106 attributes. Note these are only possible candidates, not absolute ones.
3107 GCC will guess that function pointers with @code{format} attributes that
3108 are used in assignment, initialization, parameter passing or return
3109 statements should have a corresponding @code{format} attribute in the
3110 resulting type. I.e.@: the left-hand side of the assignment or
3111 initialization, the type of the parameter variable, or the return type
3112 of the containing function respectively should also have a @code{format}
3113 attribute to avoid the warning.
3115 GCC will also warn about function definitions which might be
3116 candidates for @code{format} attributes. Again, these are only
3117 possible candidates. GCC will guess that @code{format} attributes
3118 might be appropriate for any function that calls a function like
3119 @code{vprintf} or @code{vscanf}, but this might not always be the
3120 case, and some functions for which @code{format} attributes are
3121 appropriate may not be detected.
3123 @item -Wno-multichar
3124 @opindex Wno-multichar
3126 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3127 Usually they indicate a typo in the user's code, as they have
3128 implementation-defined values, and should not be used in portable code.
3130 @item -Wnormalized=<none|id|nfc|nfkc>
3131 @opindex Wnormalized
3134 @cindex character set, input normalization
3135 In ISO C and ISO C++, two identifiers are different if they are
3136 different sequences of characters. However, sometimes when characters
3137 outside the basic ASCII character set are used, you can have two
3138 different character sequences that look the same. To avoid confusion,
3139 the ISO 10646 standard sets out some @dfn{normalization rules} which
3140 when applied ensure that two sequences that look the same are turned into
3141 the same sequence. GCC can warn you if you are using identifiers which
3142 have not been normalized; this option controls that warning.
3144 There are four levels of warning that GCC supports. The default is
3145 @option{-Wnormalized=nfc}, which warns about any identifier which is
3146 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3147 recommended form for most uses.
3149 Unfortunately, there are some characters which ISO C and ISO C++ allow
3150 in identifiers that when turned into NFC aren't allowable as
3151 identifiers. That is, there's no way to use these symbols in portable
3152 ISO C or C++ and have all your identifiers in NFC.
3153 @option{-Wnormalized=id} suppresses the warning for these characters.
3154 It is hoped that future versions of the standards involved will correct
3155 this, which is why this option is not the default.
3157 You can switch the warning off for all characters by writing
3158 @option{-Wnormalized=none}. You would only want to do this if you
3159 were using some other normalization scheme (like ``D''), because
3160 otherwise you can easily create bugs that are literally impossible to see.
3162 Some characters in ISO 10646 have distinct meanings but look identical
3163 in some fonts or display methodologies, especially once formatting has
3164 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3165 LETTER N'', will display just like a regular @code{n} which has been
3166 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3167 normalization scheme to convert all these into a standard form as
3168 well, and GCC will warn if your code is not in NFKC if you use
3169 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3170 about every identifier that contains the letter O because it might be
3171 confused with the digit 0, and so is not the default, but may be
3172 useful as a local coding convention if the programming environment is
3173 unable to be fixed to display these characters distinctly.
3175 @item -Wno-deprecated-declarations
3176 @opindex Wno-deprecated-declarations
3177 Do not warn about uses of functions (@pxref{Function Attributes}),
3178 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3179 Attributes}) marked as deprecated by using the @code{deprecated}
3183 @opindex Wno-overflow
3184 Do not warn about compile-time overflow in constant expressions.
3186 @item -Woverride-init
3187 @opindex Woverride-init
3190 Warn if an initialized field without side effects is overridden when
3191 using designated initializers (@pxref{Designated Inits, , Designated
3194 This warning is included in @option{-Wextra}. To get other
3195 @option{-Wextra} warnings without this one, use @samp{-Wextra
3196 -Wno-override-init}.
3200 Warn if a structure is given the packed attribute, but the packed
3201 attribute has no effect on the layout or size of the structure.
3202 Such structures may be mis-aligned for little benefit. For
3203 instance, in this code, the variable @code{f.x} in @code{struct bar}
3204 will be misaligned even though @code{struct bar} does not itself
3205 have the packed attribute:
3212 @} __attribute__((packed));
3222 Warn if padding is included in a structure, either to align an element
3223 of the structure or to align the whole structure. Sometimes when this
3224 happens it is possible to rearrange the fields of the structure to
3225 reduce the padding and so make the structure smaller.
3227 @item -Wredundant-decls
3228 @opindex Wredundant-decls
3229 Warn if anything is declared more than once in the same scope, even in
3230 cases where multiple declaration is valid and changes nothing.
3232 @item -Wnested-externs @r{(C only)}
3233 @opindex Wnested-externs
3234 Warn if an @code{extern} declaration is encountered within a function.
3236 @item -Wunreachable-code
3237 @opindex Wunreachable-code
3238 Warn if the compiler detects that code will never be executed.
3240 This option is intended to warn when the compiler detects that at
3241 least a whole line of source code will never be executed, because
3242 some condition is never satisfied or because it is after a
3243 procedure that never returns.
3245 It is possible for this option to produce a warning even though there
3246 are circumstances under which part of the affected line can be executed,
3247 so care should be taken when removing apparently-unreachable code.
3249 For instance, when a function is inlined, a warning may mean that the
3250 line is unreachable in only one inlined copy of the function.
3252 This option is not made part of @option{-Wall} because in a debugging
3253 version of a program there is often substantial code which checks
3254 correct functioning of the program and is, hopefully, unreachable
3255 because the program does work. Another common use of unreachable
3256 code is to provide behavior which is selectable at compile-time.
3260 Warn if a function can not be inlined and it was declared as inline.
3261 Even with this option, the compiler will not warn about failures to
3262 inline functions declared in system headers.
3264 The compiler uses a variety of heuristics to determine whether or not
3265 to inline a function. For example, the compiler takes into account
3266 the size of the function being inlined and the amount of inlining
3267 that has already been done in the current function. Therefore,
3268 seemingly insignificant changes in the source program can cause the
3269 warnings produced by @option{-Winline} to appear or disappear.
3271 @item -Wno-invalid-offsetof @r{(C++ only)}
3272 @opindex Wno-invalid-offsetof
3273 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3274 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3275 to a non-POD type is undefined. In existing C++ implementations,
3276 however, @samp{offsetof} typically gives meaningful results even when
3277 applied to certain kinds of non-POD types. (Such as a simple
3278 @samp{struct} that fails to be a POD type only by virtue of having a
3279 constructor.) This flag is for users who are aware that they are
3280 writing nonportable code and who have deliberately chosen to ignore the
3283 The restrictions on @samp{offsetof} may be relaxed in a future version
3284 of the C++ standard.
3286 @item -Wno-int-to-pointer-cast @r{(C only)}
3287 @opindex Wno-int-to-pointer-cast
3288 Suppress warnings from casts to pointer type of an integer of a
3291 @item -Wno-pointer-to-int-cast @r{(C only)}
3292 @opindex Wno-pointer-to-int-cast
3293 Suppress warnings from casts from a pointer to an integer type of a
3297 @opindex Winvalid-pch
3298 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3299 the search path but can't be used.
3303 @opindex Wno-long-long
3304 Warn if @samp{long long} type is used. This is default. To inhibit
3305 the warning messages, use @option{-Wno-long-long}. Flags
3306 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3307 only when @option{-pedantic} flag is used.
3309 @item -Wvariadic-macros
3310 @opindex Wvariadic-macros
3311 @opindex Wno-variadic-macros
3312 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3313 alternate syntax when in pedantic ISO C99 mode. This is default.
3314 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3319 Warn if variable length array is used in the code.
3320 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
3321 the variable length array.
3323 @item -Wvolatile-register-var
3324 @opindex Wvolatile-register-var
3325 @opindex Wno-volatile-register-var
3326 Warn if a register variable is declared volatile. The volatile
3327 modifier does not inhibit all optimizations that may eliminate reads
3328 and/or writes to register variables.
3330 @item -Wdisabled-optimization
3331 @opindex Wdisabled-optimization
3332 Warn if a requested optimization pass is disabled. This warning does
3333 not generally indicate that there is anything wrong with your code; it
3334 merely indicates that GCC's optimizers were unable to handle the code
3335 effectively. Often, the problem is that your code is too big or too
3336 complex; GCC will refuse to optimize programs when the optimization
3337 itself is likely to take inordinate amounts of time.
3339 @item -Wpointer-sign
3340 @opindex Wpointer-sign
3341 @opindex Wno-pointer-sign
3342 Warn for pointer argument passing or assignment with different signedness.
3343 This option is only supported for C. It is implied by @option{-Wall}
3344 and by @option{-pedantic}, which can be disabled with
3345 @option{-Wno-pointer-sign}.
3349 Make all warnings into errors.
3353 Make the specified warning into an errors. The specifier for a
3354 warning is appended, for example @option{-Werror=switch} turns the
3355 warnings controlled by @option{-Wswitch} into errors. This switch
3356 takes a negative form, to be used to negate @option{-Werror} for
3357 specific warnings, for example @option{-Wno-error=switch} makes
3358 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3359 is in effect. You can use the @option{-fdiagnostics-show-option}
3360 option to have each controllable warning amended with the option which
3361 controls it, to determine what to use with this option.
3363 Note that specifying @option{-Werror=}@var{foo} automatically implies
3364 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3367 @item -Wstack-protector
3368 @opindex Wstack-protector
3369 This option is only active when @option{-fstack-protector} is active. It
3370 warns about functions that will not be protected against stack smashing.
3372 @item -Woverlength-strings
3373 @opindex Woverlength-strings
3374 Warn about string constants which are longer than the ``minimum
3375 maximum'' length specified in the C standard. Modern compilers
3376 generally allow string constants which are much longer than the
3377 standard's minimum limit, but very portable programs should avoid
3378 using longer strings.
3380 The limit applies @emph{after} string constant concatenation, and does
3381 not count the trailing NUL@. In C89, the limit was 509 characters; in
3382 C99, it was raised to 4095. C++98 does not specify a normative
3383 minimum maximum, so we do not diagnose overlength strings in C++@.
3385 This option is implied by @option{-pedantic}, and can be disabled with
3386 @option{-Wno-overlength-strings}.
3389 @node Debugging Options
3390 @section Options for Debugging Your Program or GCC
3391 @cindex options, debugging
3392 @cindex debugging information options
3394 GCC has various special options that are used for debugging
3395 either your program or GCC:
3400 Produce debugging information in the operating system's native format
3401 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3404 On most systems that use stabs format, @option{-g} enables use of extra
3405 debugging information that only GDB can use; this extra information
3406 makes debugging work better in GDB but will probably make other debuggers
3408 refuse to read the program. If you want to control for certain whether
3409 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3410 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3412 GCC allows you to use @option{-g} with
3413 @option{-O}. The shortcuts taken by optimized code may occasionally
3414 produce surprising results: some variables you declared may not exist
3415 at all; flow of control may briefly move where you did not expect it;
3416 some statements may not be executed because they compute constant
3417 results or their values were already at hand; some statements may
3418 execute in different places because they were moved out of loops.
3420 Nevertheless it proves possible to debug optimized output. This makes
3421 it reasonable to use the optimizer for programs that might have bugs.
3423 The following options are useful when GCC is generated with the
3424 capability for more than one debugging format.
3428 Produce debugging information for use by GDB@. This means to use the
3429 most expressive format available (DWARF 2, stabs, or the native format
3430 if neither of those are supported), including GDB extensions if at all
3435 Produce debugging information in stabs format (if that is supported),
3436 without GDB extensions. This is the format used by DBX on most BSD
3437 systems. On MIPS, Alpha and System V Release 4 systems this option
3438 produces stabs debugging output which is not understood by DBX or SDB@.
3439 On System V Release 4 systems this option requires the GNU assembler.
3441 @item -feliminate-unused-debug-symbols
3442 @opindex feliminate-unused-debug-symbols
3443 Produce debugging information in stabs format (if that is supported),
3444 for only symbols that are actually used.
3446 @item -femit-class-debug-always
3447 Instead of emitting debugging information for a C++ class in only one
3448 object file, emit it in all object files using the class. This option
3449 should be used only with debuggers that are unable to handle the way GCC
3450 normally emits debugging information for classes because using this
3451 option will increase the size of debugging information by as much as a
3456 Produce debugging information in stabs format (if that is supported),
3457 using GNU extensions understood only by the GNU debugger (GDB)@. The
3458 use of these extensions is likely to make other debuggers crash or
3459 refuse to read the program.
3463 Produce debugging information in COFF format (if that is supported).
3464 This is the format used by SDB on most System V systems prior to
3469 Produce debugging information in XCOFF format (if that is supported).
3470 This is the format used by the DBX debugger on IBM RS/6000 systems.
3474 Produce debugging information in XCOFF format (if that is supported),
3475 using GNU extensions understood only by the GNU debugger (GDB)@. The
3476 use of these extensions is likely to make other debuggers crash or
3477 refuse to read the program, and may cause assemblers other than the GNU
3478 assembler (GAS) to fail with an error.
3482 Produce debugging information in DWARF version 2 format (if that is
3483 supported). This is the format used by DBX on IRIX 6. With this
3484 option, GCC uses features of DWARF version 3 when they are useful;
3485 version 3 is upward compatible with version 2, but may still cause
3486 problems for older debuggers.
3490 Produce debugging information in VMS debug format (if that is
3491 supported). This is the format used by DEBUG on VMS systems.
3494 @itemx -ggdb@var{level}
3495 @itemx -gstabs@var{level}
3496 @itemx -gcoff@var{level}
3497 @itemx -gxcoff@var{level}
3498 @itemx -gvms@var{level}
3499 Request debugging information and also use @var{level} to specify how
3500 much information. The default level is 2.
3502 Level 1 produces minimal information, enough for making backtraces in
3503 parts of the program that you don't plan to debug. This includes
3504 descriptions of functions and external variables, but no information
3505 about local variables and no line numbers.
3507 Level 3 includes extra information, such as all the macro definitions
3508 present in the program. Some debuggers support macro expansion when
3509 you use @option{-g3}.
3511 @option{-gdwarf-2} does not accept a concatenated debug level, because
3512 GCC used to support an option @option{-gdwarf} that meant to generate
3513 debug information in version 1 of the DWARF format (which is very
3514 different from version 2), and it would have been too confusing. That
3515 debug format is long obsolete, but the option cannot be changed now.
3516 Instead use an additional @option{-g@var{level}} option to change the
3517 debug level for DWARF2.
3519 @item -feliminate-dwarf2-dups
3520 @opindex feliminate-dwarf2-dups
3521 Compress DWARF2 debugging information by eliminating duplicated
3522 information about each symbol. This option only makes sense when
3523 generating DWARF2 debugging information with @option{-gdwarf-2}.
3525 @item -femit-struct-debug-baseonly
3526 Emit debug information for struct-like types
3527 only when the base name of the compilation source file
3528 matches the base name of file in which the struct was defined.
3530 This option substantially reduces the size of debugging information,
3531 but at significant potential loss in type information to the debugger.
3532 See @option{-femit-struct-debug-reduced} for a less aggressive option.
3533 See @option{-femit-struct-debug-detailed} for more detailed control.
3535 This option works only with DWARF 2.
3537 @item -femit-struct-debug-reduced
3538 Emit debug information for struct-like types
3539 only when the base name of the compilation source file
3540 matches the base name of file in which the type was defined,
3541 unless the struct is a template or defined in a system header.
3543 This option significantly reduces the size of debugging information,
3544 with some potential loss in type information to the debugger.
3545 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
3546 See @option{-femit-struct-debug-detailed} for more detailed control.
3548 This option works only with DWARF 2.
3550 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
3551 Specify the struct-like types
3552 for which the compiler will generate debug information.
3553 The intent is to reduce duplicate struct debug information
3554 between different object files within the same program.
3556 This option is a detailed version of
3557 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
3558 which will serve for most needs.
3560 A specification has the syntax
3561 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
3563 The optional first word limits the specification to
3564 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
3565 A struct type is used directly when it is the type of a variable, member.
3566 Indirect uses arise through pointers to structs.
3567 That is, when use of an incomplete struct would be legal, the use is indirect.
3569 @samp{struct one direct; struct two * indirect;}.
3571 The optional second word limits the specification to
3572 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
3573 Generic structs are a bit complicated to explain.
3574 For C++, these are non-explicit specializations of template classes,
3575 or non-template classes within the above.
3576 Other programming languages have generics,
3577 but @samp{-femit-struct-debug-detailed} does not yet implement them.
3579 The third word specifies the source files for those
3580 structs for which the compiler will emit debug information.
3581 The values @samp{none} and @samp{any} have the normal meaning.
3582 The value @samp{base} means that
3583 the base of name of the file in which the type declaration appears
3584 must match the base of the name of the main compilation file.
3585 In practice, this means that
3586 types declared in @file{foo.c} and @file{foo.h} will have debug information,
3587 but types declared in other header will not.
3588 The value @samp{sys} means those types satisfying @samp{base}
3589 or declared in system or compiler headers.
3591 You may need to experiment to determine the best settings for your application.
3593 The default is @samp{-femit-struct-debug-detailed=all}.
3595 This option works only with DWARF 2.
3597 @cindex @command{prof}
3600 Generate extra code to write profile information suitable for the
3601 analysis program @command{prof}. You must use this option when compiling
3602 the source files you want data about, and you must also use it when
3605 @cindex @command{gprof}
3608 Generate extra code to write profile information suitable for the
3609 analysis program @command{gprof}. You must use this option when compiling
3610 the source files you want data about, and you must also use it when
3615 Makes the compiler print out each function name as it is compiled, and
3616 print some statistics about each pass when it finishes.
3619 @opindex ftime-report
3620 Makes the compiler print some statistics about the time consumed by each
3621 pass when it finishes.
3624 @opindex fmem-report
3625 Makes the compiler print some statistics about permanent memory
3626 allocation when it finishes.
3628 @item -fprofile-arcs
3629 @opindex fprofile-arcs
3630 Add code so that program flow @dfn{arcs} are instrumented. During
3631 execution the program records how many times each branch and call is
3632 executed and how many times it is taken or returns. When the compiled
3633 program exits it saves this data to a file called
3634 @file{@var{auxname}.gcda} for each source file. The data may be used for
3635 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3636 test coverage analysis (@option{-ftest-coverage}). Each object file's
3637 @var{auxname} is generated from the name of the output file, if
3638 explicitly specified and it is not the final executable, otherwise it is
3639 the basename of the source file. In both cases any suffix is removed
3640 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3641 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3642 @xref{Cross-profiling}.
3644 @cindex @command{gcov}
3648 This option is used to compile and link code instrumented for coverage
3649 analysis. The option is a synonym for @option{-fprofile-arcs}
3650 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3651 linking). See the documentation for those options for more details.
3656 Compile the source files with @option{-fprofile-arcs} plus optimization
3657 and code generation options. For test coverage analysis, use the
3658 additional @option{-ftest-coverage} option. You do not need to profile
3659 every source file in a program.
3662 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3663 (the latter implies the former).
3666 Run the program on a representative workload to generate the arc profile
3667 information. This may be repeated any number of times. You can run
3668 concurrent instances of your program, and provided that the file system
3669 supports locking, the data files will be correctly updated. Also
3670 @code{fork} calls are detected and correctly handled (double counting
3674 For profile-directed optimizations, compile the source files again with
3675 the same optimization and code generation options plus
3676 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3677 Control Optimization}).
3680 For test coverage analysis, use @command{gcov} to produce human readable
3681 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3682 @command{gcov} documentation for further information.
3686 With @option{-fprofile-arcs}, for each function of your program GCC
3687 creates a program flow graph, then finds a spanning tree for the graph.
3688 Only arcs that are not on the spanning tree have to be instrumented: the
3689 compiler adds code to count the number of times that these arcs are
3690 executed. When an arc is the only exit or only entrance to a block, the
3691 instrumentation code can be added to the block; otherwise, a new basic
3692 block must be created to hold the instrumentation code.
3695 @item -ftest-coverage
3696 @opindex ftest-coverage
3697 Produce a notes file that the @command{gcov} code-coverage utility
3698 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3699 show program coverage. Each source file's note file is called
3700 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3701 above for a description of @var{auxname} and instructions on how to
3702 generate test coverage data. Coverage data will match the source files
3703 more closely, if you do not optimize.
3705 @item -d@var{letters}
3706 @item -fdump-rtl-@var{pass}
3708 Says to make debugging dumps during compilation at times specified by
3709 @var{letters}. This is used for debugging the RTL-based passes of the
3710 compiler. The file names for most of the dumps are made by appending a
3711 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3712 from the name of the output file, if explicitly specified and it is not
3713 an executable, otherwise it is the basename of the source file. These
3714 switches may have different effects when @option{-E} is used for
3717 Most debug dumps can be enabled either passing a letter to the @option{-d}
3718 option, or with a long @option{-fdump-rtl} switch; here are the possible
3719 letters for use in @var{letters} and @var{pass}, and their meanings:
3724 Annotate the assembler output with miscellaneous debugging information.
3727 @itemx -fdump-rtl-bbro
3729 @opindex fdump-rtl-bbro
3730 Dump after block reordering, to @file{@var{file}.148r.bbro}.
3733 @itemx -fdump-rtl-combine
3735 @opindex fdump-rtl-combine
3736 Dump after instruction combination, to the file @file{@var{file}.129r.combine}.
3739 @itemx -fdump-rtl-ce1
3740 @itemx -fdump-rtl-ce2
3742 @opindex fdump-rtl-ce1
3743 @opindex fdump-rtl-ce2
3744 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3745 first if conversion, to the file @file{@var{file}.117r.ce1}. @option{-dC}
3746 and @option{-fdump-rtl-ce2} enable dumping after the second if
3747 conversion, to the file @file{@var{file}.130r.ce2}.
3750 @itemx -fdump-rtl-btl
3751 @itemx -fdump-rtl-dbr
3753 @opindex fdump-rtl-btl
3754 @opindex fdump-rtl-dbr
3755 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3756 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3757 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3758 scheduling, to @file{@var{file}.36.dbr}.
3762 Dump all macro definitions, at the end of preprocessing, in addition to
3766 @itemx -fdump-rtl-ce3
3768 @opindex fdump-rtl-ce3
3769 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
3772 @itemx -fdump-rtl-cfg
3773 @itemx -fdump-rtl-life
3775 @opindex fdump-rtl-cfg
3776 @opindex fdump-rtl-life
3777 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3778 and data flow analysis, to @file{@var{file}.116r.cfg}. @option{-df}
3779 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3780 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
3783 @itemx -fdump-rtl-greg
3785 @opindex fdump-rtl-greg
3786 Dump after global register allocation, to @file{@var{file}.139r.greg}.
3789 @itemx -fdump-rtl-gcse
3790 @itemx -fdump-rtl-bypass
3792 @opindex fdump-rtl-gcse
3793 @opindex fdump-rtl-bypass
3794 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3795 @file{@var{file}.114r.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3796 enable dumping after jump bypassing and control flow optimizations, to
3797 @file{@var{file}.115r.bypass}.
3800 @itemx -fdump-rtl-eh
3802 @opindex fdump-rtl-eh
3803 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3806 @itemx -fdump-rtl-sibling
3808 @opindex fdump-rtl-sibling
3809 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
3812 @itemx -fdump-rtl-jump
3814 @opindex fdump-rtl-jump
3815 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
3818 @itemx -fdump-rtl-stack
3820 @opindex fdump-rtl-stack
3821 Dump after conversion from registers to stack, to @file{@var{file}.152r.stack}.
3824 @itemx -fdump-rtl-lreg
3826 @opindex fdump-rtl-lreg
3827 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
3830 @itemx -fdump-rtl-loop2
3832 @opindex fdump-rtl-loop2
3833 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
3834 loop optimization pass, to @file{@var{file}.119r.loop2},
3835 @file{@var{file}.120r.loop2_init},
3836 @file{@var{file}.121r.loop2_invariant}, and
3837 @file{@var{file}.125r.loop2_done}.
3840 @itemx -fdump-rtl-sms
3842 @opindex fdump-rtl-sms
3843 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
3846 @itemx -fdump-rtl-mach
3848 @opindex fdump-rtl-mach
3849 Dump after performing the machine dependent reorganization pass, to
3850 @file{@var{file}.155r.mach} if that pass exists.
3853 @itemx -fdump-rtl-rnreg
3855 @opindex fdump-rtl-rnreg
3856 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
3859 @itemx -fdump-rtl-regmove
3861 @opindex fdump-rtl-regmove
3862 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
3865 @itemx -fdump-rtl-postreload
3867 @opindex fdump-rtl-postreload
3868 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3871 @itemx -fdump-rtl-expand
3873 @opindex fdump-rtl-expand
3874 Dump after RTL generation, to @file{@var{file}.104r.expand}.
3877 @itemx -fdump-rtl-sched2
3879 @opindex fdump-rtl-sched2
3880 Dump after the second scheduling pass, to @file{@var{file}.150r.sched2}.
3883 @itemx -fdump-rtl-cse
3885 @opindex fdump-rtl-cse
3886 Dump after CSE (including the jump optimization that sometimes follows
3887 CSE), to @file{@var{file}.113r.cse}.
3890 @itemx -fdump-rtl-sched
3892 @opindex fdump-rtl-sched
3893 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3896 @itemx -fdump-rtl-cse2
3898 @opindex fdump-rtl-cse2
3899 Dump after the second CSE pass (including the jump optimization that
3900 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
3903 @itemx -fdump-rtl-tracer
3905 @opindex fdump-rtl-tracer
3906 Dump after running tracer, to @file{@var{file}.118r.tracer}.
3909 @itemx -fdump-rtl-vpt
3910 @itemx -fdump-rtl-vartrack
3912 @opindex fdump-rtl-vpt
3913 @opindex fdump-rtl-vartrack
3914 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3915 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3916 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3917 to @file{@var{file}.154r.vartrack}.
3920 @itemx -fdump-rtl-flow2
3922 @opindex fdump-rtl-flow2
3923 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
3926 @itemx -fdump-rtl-peephole2
3928 @opindex fdump-rtl-peephole2
3929 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
3932 @itemx -fdump-rtl-web
3934 @opindex fdump-rtl-web
3935 Dump after live range splitting, to @file{@var{file}.126r.web}.
3938 @itemx -fdump-rtl-all
3940 @opindex fdump-rtl-all
3941 Produce all the dumps listed above.
3945 Produce a core dump whenever an error occurs.
3949 Print statistics on memory usage, at the end of the run, to
3954 Annotate the assembler output with a comment indicating which
3955 pattern and alternative was used. The length of each instruction is
3960 Dump the RTL in the assembler output as a comment before each instruction.
3961 Also turns on @option{-dp} annotation.
3965 For each of the other indicated dump files (either with @option{-d} or
3966 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3967 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3971 Just generate RTL for a function instead of compiling it. Usually used
3972 with @samp{r} (@option{-fdump-rtl-expand}).
3976 Dump debugging information during parsing, to standard error.
3980 @opindex fdump-noaddr
3981 When doing debugging dumps (see @option{-d} option above), suppress
3982 address output. This makes it more feasible to use diff on debugging
3983 dumps for compiler invocations with different compiler binaries and/or
3984 different text / bss / data / heap / stack / dso start locations.
3986 @item -fdump-unnumbered
3987 @opindex fdump-unnumbered
3988 When doing debugging dumps (see @option{-d} option above), suppress instruction
3989 numbers, line number note and address output. This makes it more feasible to
3990 use diff on debugging dumps for compiler invocations with different
3991 options, in particular with and without @option{-g}.
3993 @item -fdump-translation-unit @r{(C++ only)}
3994 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
3995 @opindex fdump-translation-unit
3996 Dump a representation of the tree structure for the entire translation
3997 unit to a file. The file name is made by appending @file{.tu} to the
3998 source file name. If the @samp{-@var{options}} form is used, @var{options}
3999 controls the details of the dump as described for the
4000 @option{-fdump-tree} options.
4002 @item -fdump-class-hierarchy @r{(C++ only)}
4003 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4004 @opindex fdump-class-hierarchy
4005 Dump a representation of each class's hierarchy and virtual function
4006 table layout to a file. The file name is made by appending @file{.class}
4007 to the source file name. If the @samp{-@var{options}} form is used,
4008 @var{options} controls the details of the dump as described for the
4009 @option{-fdump-tree} options.
4011 @item -fdump-ipa-@var{switch}
4013 Control the dumping at various stages of inter-procedural analysis
4014 language tree to a file. The file name is generated by appending a switch
4015 specific suffix to the source file name. The following dumps are possible:
4019 Enables all inter-procedural analysis dumps; currently the only produced
4020 dump is the @samp{cgraph} dump.
4023 Dumps information about call-graph optimization, unused function removal,
4024 and inlining decisions.
4027 @item -fdump-tree-@var{switch}
4028 @itemx -fdump-tree-@var{switch}-@var{options}
4030 Control the dumping at various stages of processing the intermediate
4031 language tree to a file. The file name is generated by appending a switch
4032 specific suffix to the source file name. If the @samp{-@var{options}}
4033 form is used, @var{options} is a list of @samp{-} separated options that
4034 control the details of the dump. Not all options are applicable to all
4035 dumps, those which are not meaningful will be ignored. The following
4036 options are available
4040 Print the address of each node. Usually this is not meaningful as it
4041 changes according to the environment and source file. Its primary use
4042 is for tying up a dump file with a debug environment.
4044 Inhibit dumping of members of a scope or body of a function merely
4045 because that scope has been reached. Only dump such items when they
4046 are directly reachable by some other path. When dumping pretty-printed
4047 trees, this option inhibits dumping the bodies of control structures.
4049 Print a raw representation of the tree. By default, trees are
4050 pretty-printed into a C-like representation.
4052 Enable more detailed dumps (not honored by every dump option).
4054 Enable dumping various statistics about the pass (not honored by every dump
4057 Enable showing basic block boundaries (disabled in raw dumps).
4059 Enable showing virtual operands for every statement.
4061 Enable showing line numbers for statements.
4063 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4065 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4068 The following tree dumps are possible:
4072 Dump before any tree based optimization, to @file{@var{file}.original}.
4075 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4078 Dump after function inlining, to @file{@var{file}.inlined}.
4081 @opindex fdump-tree-gimple
4082 Dump each function before and after the gimplification pass to a file. The
4083 file name is made by appending @file{.gimple} to the source file name.
4086 @opindex fdump-tree-cfg
4087 Dump the control flow graph of each function to a file. The file name is
4088 made by appending @file{.cfg} to the source file name.
4091 @opindex fdump-tree-vcg
4092 Dump the control flow graph of each function to a file in VCG format. The
4093 file name is made by appending @file{.vcg} to the source file name. Note
4094 that if the file contains more than one function, the generated file cannot
4095 be used directly by VCG@. You will need to cut and paste each function's
4096 graph into its own separate file first.
4099 @opindex fdump-tree-ch
4100 Dump each function after copying loop headers. The file name is made by
4101 appending @file{.ch} to the source file name.
4104 @opindex fdump-tree-ssa
4105 Dump SSA related information to a file. The file name is made by appending
4106 @file{.ssa} to the source file name.
4109 @opindex fdump-tree-salias
4110 Dump structure aliasing variable information to a file. This file name
4111 is made by appending @file{.salias} to the source file name.
4114 @opindex fdump-tree-alias
4115 Dump aliasing information for each function. The file name is made by
4116 appending @file{.alias} to the source file name.
4119 @opindex fdump-tree-ccp
4120 Dump each function after CCP@. The file name is made by appending
4121 @file{.ccp} to the source file name.
4124 @opindex fdump-tree-storeccp
4125 Dump each function after STORE-CCP. The file name is made by appending
4126 @file{.storeccp} to the source file name.
4129 @opindex fdump-tree-pre
4130 Dump trees after partial redundancy elimination. The file name is made
4131 by appending @file{.pre} to the source file name.
4134 @opindex fdump-tree-fre
4135 Dump trees after full redundancy elimination. The file name is made
4136 by appending @file{.fre} to the source file name.
4139 @opindex fdump-tree-copyprop
4140 Dump trees after copy propagation. The file name is made
4141 by appending @file{.copyprop} to the source file name.
4143 @item store_copyprop
4144 @opindex fdump-tree-store_copyprop
4145 Dump trees after store copy-propagation. The file name is made
4146 by appending @file{.store_copyprop} to the source file name.
4149 @opindex fdump-tree-dce
4150 Dump each function after dead code elimination. The file name is made by
4151 appending @file{.dce} to the source file name.
4154 @opindex fdump-tree-mudflap
4155 Dump each function after adding mudflap instrumentation. The file name is
4156 made by appending @file{.mudflap} to the source file name.
4159 @opindex fdump-tree-sra
4160 Dump each function after performing scalar replacement of aggregates. The
4161 file name is made by appending @file{.sra} to the source file name.
4164 @opindex fdump-tree-sink
4165 Dump each function after performing code sinking. The file name is made
4166 by appending @file{.sink} to the source file name.
4169 @opindex fdump-tree-dom
4170 Dump each function after applying dominator tree optimizations. The file
4171 name is made by appending @file{.dom} to the source file name.
4174 @opindex fdump-tree-dse
4175 Dump each function after applying dead store elimination. The file
4176 name is made by appending @file{.dse} to the source file name.
4179 @opindex fdump-tree-phiopt
4180 Dump each function after optimizing PHI nodes into straightline code. The file
4181 name is made by appending @file{.phiopt} to the source file name.
4184 @opindex fdump-tree-forwprop
4185 Dump each function after forward propagating single use variables. The file
4186 name is made by appending @file{.forwprop} to the source file name.
4189 @opindex fdump-tree-copyrename
4190 Dump each function after applying the copy rename optimization. The file
4191 name is made by appending @file{.copyrename} to the source file name.
4194 @opindex fdump-tree-nrv
4195 Dump each function after applying the named return value optimization on
4196 generic trees. The file name is made by appending @file{.nrv} to the source
4200 @opindex fdump-tree-vect
4201 Dump each function after applying vectorization of loops. The file name is
4202 made by appending @file{.vect} to the source file name.
4205 @opindex fdump-tree-vrp
4206 Dump each function after Value Range Propagation (VRP). The file name
4207 is made by appending @file{.vrp} to the source file name.
4210 @opindex fdump-tree-all
4211 Enable all the available tree dumps with the flags provided in this option.
4214 @item -ftree-vectorizer-verbose=@var{n}
4215 @opindex ftree-vectorizer-verbose
4216 This option controls the amount of debugging output the vectorizer prints.
4217 This information is written to standard error, unless
4218 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4219 in which case it is output to the usual dump listing file, @file{.vect}.
4220 For @var{n}=0 no diagnostic information is reported.
4221 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4222 and the total number of loops that got vectorized.
4223 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4224 the first analysis phase (vect_analyze_loop_form) - i.e. countable,
4225 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4226 level that @option{-fdump-tree-vect-stats} uses.
4227 Higher verbosity levels mean either more information dumped for each
4228 reported loop, or same amount of information reported for more loops:
4229 If @var{n}=3, alignment related information is added to the reports.
4230 If @var{n}=4, data-references related information (e.g. memory dependences,
4231 memory access-patterns) is added to the reports.
4232 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4233 that did not pass the first analysis phase (i.e. may not be countable, or
4234 may have complicated control-flow).
4235 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4236 For @var{n}=7, all the information the vectorizer generates during its
4237 analysis and transformation is reported. This is the same verbosity level
4238 that @option{-fdump-tree-vect-details} uses.
4240 @item -frandom-seed=@var{string}
4241 @opindex frandom-string
4242 This option provides a seed that GCC uses when it would otherwise use
4243 random numbers. It is used to generate certain symbol names
4244 that have to be different in every compiled file. It is also used to
4245 place unique stamps in coverage data files and the object files that
4246 produce them. You can use the @option{-frandom-seed} option to produce
4247 reproducibly identical object files.
4249 The @var{string} should be different for every file you compile.
4251 @item -fsched-verbose=@var{n}
4252 @opindex fsched-verbose
4253 On targets that use instruction scheduling, this option controls the
4254 amount of debugging output the scheduler prints. This information is
4255 written to standard error, unless @option{-dS} or @option{-dR} is
4256 specified, in which case it is output to the usual dump
4257 listing file, @file{.sched} or @file{.sched2} respectively. However
4258 for @var{n} greater than nine, the output is always printed to standard
4261 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4262 same information as @option{-dRS}. For @var{n} greater than one, it
4263 also output basic block probabilities, detailed ready list information
4264 and unit/insn info. For @var{n} greater than two, it includes RTL
4265 at abort point, control-flow and regions info. And for @var{n} over
4266 four, @option{-fsched-verbose} also includes dependence info.
4270 Store the usual ``temporary'' intermediate files permanently; place them
4271 in the current directory and name them based on the source file. Thus,
4272 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4273 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4274 preprocessed @file{foo.i} output file even though the compiler now
4275 normally uses an integrated preprocessor.
4277 When used in combination with the @option{-x} command line option,
4278 @option{-save-temps} is sensible enough to avoid over writing an
4279 input source file with the same extension as an intermediate file.
4280 The corresponding intermediate file may be obtained by renaming the
4281 source file before using @option{-save-temps}.
4285 Report the CPU time taken by each subprocess in the compilation
4286 sequence. For C source files, this is the compiler proper and assembler
4287 (plus the linker if linking is done). The output looks like this:
4294 The first number on each line is the ``user time'', that is time spent
4295 executing the program itself. The second number is ``system time'',
4296 time spent executing operating system routines on behalf of the program.
4297 Both numbers are in seconds.
4299 @item -fvar-tracking
4300 @opindex fvar-tracking
4301 Run variable tracking pass. It computes where variables are stored at each
4302 position in code. Better debugging information is then generated
4303 (if the debugging information format supports this information).
4305 It is enabled by default when compiling with optimization (@option{-Os},
4306 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4307 the debug info format supports it.
4309 @item -print-file-name=@var{library}
4310 @opindex print-file-name
4311 Print the full absolute name of the library file @var{library} that
4312 would be used when linking---and don't do anything else. With this
4313 option, GCC does not compile or link anything; it just prints the
4316 @item -print-multi-directory
4317 @opindex print-multi-directory
4318 Print the directory name corresponding to the multilib selected by any
4319 other switches present in the command line. This directory is supposed
4320 to exist in @env{GCC_EXEC_PREFIX}.
4322 @item -print-multi-lib
4323 @opindex print-multi-lib
4324 Print the mapping from multilib directory names to compiler switches
4325 that enable them. The directory name is separated from the switches by
4326 @samp{;}, and each switch starts with an @samp{@@} instead of the
4327 @samp{-}, without spaces between multiple switches. This is supposed to
4328 ease shell-processing.
4330 @item -print-prog-name=@var{program}
4331 @opindex print-prog-name
4332 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4334 @item -print-libgcc-file-name
4335 @opindex print-libgcc-file-name
4336 Same as @option{-print-file-name=libgcc.a}.
4338 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4339 but you do want to link with @file{libgcc.a}. You can do
4342 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4345 @item -print-search-dirs
4346 @opindex print-search-dirs
4347 Print the name of the configured installation directory and a list of
4348 program and library directories @command{gcc} will search---and don't do anything else.
4350 This is useful when @command{gcc} prints the error message
4351 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4352 To resolve this you either need to put @file{cpp0} and the other compiler
4353 components where @command{gcc} expects to find them, or you can set the environment
4354 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4355 Don't forget the trailing @samp{/}.
4356 @xref{Environment Variables}.
4359 @opindex dumpmachine
4360 Print the compiler's target machine (for example,
4361 @samp{i686-pc-linux-gnu})---and don't do anything else.
4364 @opindex dumpversion
4365 Print the compiler version (for example, @samp{3.0})---and don't do
4370 Print the compiler's built-in specs---and don't do anything else. (This
4371 is used when GCC itself is being built.) @xref{Spec Files}.
4373 @item -feliminate-unused-debug-types
4374 @opindex feliminate-unused-debug-types
4375 Normally, when producing DWARF2 output, GCC will emit debugging
4376 information for all types declared in a compilation
4377 unit, regardless of whether or not they are actually used
4378 in that compilation unit. Sometimes this is useful, such as
4379 if, in the debugger, you want to cast a value to a type that is
4380 not actually used in your program (but is declared). More often,
4381 however, this results in a significant amount of wasted space.
4382 With this option, GCC will avoid producing debug symbol output
4383 for types that are nowhere used in the source file being compiled.
4386 @node Optimize Options
4387 @section Options That Control Optimization
4388 @cindex optimize options
4389 @cindex options, optimization
4391 These options control various sorts of optimizations.
4393 Without any optimization option, the compiler's goal is to reduce the
4394 cost of compilation and to make debugging produce the expected
4395 results. Statements are independent: if you stop the program with a
4396 breakpoint between statements, you can then assign a new value to any
4397 variable or change the program counter to any other statement in the
4398 function and get exactly the results you would expect from the source
4401 Turning on optimization flags makes the compiler attempt to improve
4402 the performance and/or code size at the expense of compilation time
4403 and possibly the ability to debug the program.
4405 The compiler performs optimization based on the knowledge it has of
4406 the program. Optimization levels @option{-O} and above, in
4407 particular, enable @emph{unit-at-a-time} mode, which allows the
4408 compiler to consider information gained from later functions in
4409 the file when compiling a function. Compiling multiple files at
4410 once to a single output file in @emph{unit-at-a-time} mode allows
4411 the compiler to use information gained from all of the files when
4412 compiling each of them.
4414 Not all optimizations are controlled directly by a flag. Only
4415 optimizations that have a flag are listed.
4422 Optimize. Optimizing compilation takes somewhat more time, and a lot
4423 more memory for a large function.
4425 With @option{-O}, the compiler tries to reduce code size and execution
4426 time, without performing any optimizations that take a great deal of
4429 @option{-O} turns on the following optimization flags:
4430 @gccoptlist{-fdefer-pop @gol
4431 -fdelayed-branch @gol
4432 -fguess-branch-probability @gol
4433 -fcprop-registers @gol
4434 -fif-conversion @gol
4435 -fif-conversion2 @gol
4438 -ftree-dominator-opts @gol
4443 -ftree-copyrename @gol
4446 -funit-at-a-time @gol
4449 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4450 where doing so does not interfere with debugging.
4454 Optimize even more. GCC performs nearly all supported optimizations
4455 that do not involve a space-speed tradeoff. The compiler does not
4456 perform loop unrolling or function inlining when you specify @option{-O2}.
4457 As compared to @option{-O}, this option increases both compilation time
4458 and the performance of the generated code.
4460 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4461 also turns on the following optimization flags:
4462 @gccoptlist{-fthread-jumps @gol
4464 -foptimize-sibling-calls @gol
4465 -fcse-follow-jumps -fcse-skip-blocks @gol
4466 -fgcse -fgcse-lm @gol
4467 -fexpensive-optimizations @gol
4468 -frerun-cse-after-loop @gol
4471 -fschedule-insns -fschedule-insns2 @gol
4472 -fsched-interblock -fsched-spec @gol
4474 -fstrict-aliasing -fstrict-overflow @gol
4475 -fdelete-null-pointer-checks @gol
4476 -freorder-blocks -freorder-functions @gol
4477 -falign-functions -falign-jumps @gol
4478 -falign-loops -falign-labels @gol
4482 Please note the warning under @option{-fgcse} about
4483 invoking @option{-O2} on programs that use computed gotos.
4485 @option{-O2} doesn't turn on @option{-ftree-vrp} for the Ada compiler.
4486 This option must be explicitly specified on the command line to be
4487 enabled for the Ada compiler.
4491 Optimize yet more. @option{-O3} turns on all optimizations specified by
4492 @option{-O2} and also turns on the @option{-finline-functions},
4493 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4497 Do not optimize. This is the default.
4501 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4502 do not typically increase code size. It also performs further
4503 optimizations designed to reduce code size.
4505 @option{-Os} disables the following optimization flags:
4506 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4507 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
4508 -fprefetch-loop-arrays -ftree-vect-loop-version}
4510 If you use multiple @option{-O} options, with or without level numbers,
4511 the last such option is the one that is effective.
4514 Options of the form @option{-f@var{flag}} specify machine-independent
4515 flags. Most flags have both positive and negative forms; the negative
4516 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4517 below, only one of the forms is listed---the one you typically will
4518 use. You can figure out the other form by either removing @samp{no-}
4521 The following options control specific optimizations. They are either
4522 activated by @option{-O} options or are related to ones that are. You
4523 can use the following flags in the rare cases when ``fine-tuning'' of
4524 optimizations to be performed is desired.
4527 @item -fno-default-inline
4528 @opindex fno-default-inline
4529 Do not make member functions inline by default merely because they are
4530 defined inside the class scope (C++ only). Otherwise, when you specify
4531 @w{@option{-O}}, member functions defined inside class scope are compiled
4532 inline by default; i.e., you don't need to add @samp{inline} in front of
4533 the member function name.
4535 @item -fno-defer-pop
4536 @opindex fno-defer-pop
4537 Always pop the arguments to each function call as soon as that function
4538 returns. For machines which must pop arguments after a function call,
4539 the compiler normally lets arguments accumulate on the stack for several
4540 function calls and pops them all at once.
4542 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4546 Force memory operands to be copied into registers before doing
4547 arithmetic on them. This produces better code by making all memory
4548 references potential common subexpressions. When they are not common
4549 subexpressions, instruction combination should eliminate the separate
4550 register-load. This option is now a nop and will be removed in 4.3.
4553 @opindex fforce-addr
4554 Force memory address constants to be copied into registers before
4555 doing arithmetic on them.
4557 @item -fomit-frame-pointer
4558 @opindex fomit-frame-pointer
4559 Don't keep the frame pointer in a register for functions that
4560 don't need one. This avoids the instructions to save, set up and
4561 restore frame pointers; it also makes an extra register available
4562 in many functions. @strong{It also makes debugging impossible on
4565 On some machines, such as the VAX, this flag has no effect, because
4566 the standard calling sequence automatically handles the frame pointer
4567 and nothing is saved by pretending it doesn't exist. The
4568 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4569 whether a target machine supports this flag. @xref{Registers,,Register
4570 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4572 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4574 @item -foptimize-sibling-calls
4575 @opindex foptimize-sibling-calls
4576 Optimize sibling and tail recursive calls.
4578 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4582 Don't pay attention to the @code{inline} keyword. Normally this option
4583 is used to keep the compiler from expanding any functions inline.
4584 Note that if you are not optimizing, no functions can be expanded inline.
4586 @item -finline-functions
4587 @opindex finline-functions
4588 Integrate all simple functions into their callers. The compiler
4589 heuristically decides which functions are simple enough to be worth
4590 integrating in this way.
4592 If all calls to a given function are integrated, and the function is
4593 declared @code{static}, then the function is normally not output as
4594 assembler code in its own right.
4596 Enabled at level @option{-O3}.
4598 @item -finline-functions-called-once
4599 @opindex finline-functions-called-once
4600 Consider all @code{static} functions called once for inlining into their
4601 caller even if they are not marked @code{inline}. If a call to a given
4602 function is integrated, then the function is not output as assembler code
4605 Enabled if @option{-funit-at-a-time} is enabled.
4607 @item -fearly-inlining
4608 @opindex fearly-inlining
4609 Inline functions marked by @code{always_inline} and functions whose body seems
4610 smaller than the function call overhead early before doing
4611 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
4612 makes profiling significantly cheaper and usually inlining faster on programs
4613 having large chains of nested wrapper functions.
4617 @item -finline-limit=@var{n}
4618 @opindex finline-limit
4619 By default, GCC limits the size of functions that can be inlined. This flag
4620 allows the control of this limit for functions that are explicitly marked as
4621 inline (i.e., marked with the inline keyword or defined within the class
4622 definition in c++). @var{n} is the size of functions that can be inlined in
4623 number of pseudo instructions (not counting parameter handling). The default
4624 value of @var{n} is 600.
4625 Increasing this value can result in more inlined code at
4626 the cost of compilation time and memory consumption. Decreasing usually makes
4627 the compilation faster and less code will be inlined (which presumably
4628 means slower programs). This option is particularly useful for programs that
4629 use inlining heavily such as those based on recursive templates with C++.
4631 Inlining is actually controlled by a number of parameters, which may be
4632 specified individually by using @option{--param @var{name}=@var{value}}.
4633 The @option{-finline-limit=@var{n}} option sets some of these parameters
4637 @item max-inline-insns-single
4638 is set to @var{n}/2.
4639 @item max-inline-insns-auto
4640 is set to @var{n}/2.
4641 @item min-inline-insns
4642 is set to 130 or @var{n}/4, whichever is smaller.
4643 @item max-inline-insns-rtl
4647 See below for a documentation of the individual
4648 parameters controlling inlining.
4650 @emph{Note:} pseudo instruction represents, in this particular context, an
4651 abstract measurement of function's size. In no way does it represent a count
4652 of assembly instructions and as such its exact meaning might change from one
4653 release to an another.
4655 @item -fkeep-inline-functions
4656 @opindex fkeep-inline-functions
4657 In C, emit @code{static} functions that are declared @code{inline}
4658 into the object file, even if the function has been inlined into all
4659 of its callers. This switch does not affect functions using the
4660 @code{extern inline} extension in GNU C@. In C++, emit any and all
4661 inline functions into the object file.
4663 @item -fkeep-static-consts
4664 @opindex fkeep-static-consts
4665 Emit variables declared @code{static const} when optimization isn't turned
4666 on, even if the variables aren't referenced.
4668 GCC enables this option by default. If you want to force the compiler to
4669 check if the variable was referenced, regardless of whether or not
4670 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4672 @item -fmerge-constants
4673 Attempt to merge identical constants (string constants and floating point
4674 constants) across compilation units.
4676 This option is the default for optimized compilation if the assembler and
4677 linker support it. Use @option{-fno-merge-constants} to inhibit this
4680 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4682 @item -fmerge-all-constants
4683 Attempt to merge identical constants and identical variables.
4685 This option implies @option{-fmerge-constants}. In addition to
4686 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4687 arrays or initialized constant variables with integral or floating point
4688 types. Languages like C or C++ require each non-automatic variable to
4689 have distinct location, so using this option will result in non-conforming
4692 @item -fmodulo-sched
4693 @opindex fmodulo-sched
4694 Perform swing modulo scheduling immediately before the first scheduling
4695 pass. This pass looks at innermost loops and reorders their
4696 instructions by overlapping different iterations.
4698 @item -fno-branch-count-reg
4699 @opindex fno-branch-count-reg
4700 Do not use ``decrement and branch'' instructions on a count register,
4701 but instead generate a sequence of instructions that decrement a
4702 register, compare it against zero, then branch based upon the result.
4703 This option is only meaningful on architectures that support such
4704 instructions, which include x86, PowerPC, IA-64 and S/390.
4706 The default is @option{-fbranch-count-reg}.
4708 @item -fno-function-cse
4709 @opindex fno-function-cse
4710 Do not put function addresses in registers; make each instruction that
4711 calls a constant function contain the function's address explicitly.
4713 This option results in less efficient code, but some strange hacks
4714 that alter the assembler output may be confused by the optimizations
4715 performed when this option is not used.
4717 The default is @option{-ffunction-cse}
4719 @item -fno-zero-initialized-in-bss
4720 @opindex fno-zero-initialized-in-bss
4721 If the target supports a BSS section, GCC by default puts variables that
4722 are initialized to zero into BSS@. This can save space in the resulting
4725 This option turns off this behavior because some programs explicitly
4726 rely on variables going to the data section. E.g., so that the
4727 resulting executable can find the beginning of that section and/or make
4728 assumptions based on that.
4730 The default is @option{-fzero-initialized-in-bss}.
4732 @item -fbounds-check
4733 @opindex fbounds-check
4734 For front-ends that support it, generate additional code to check that
4735 indices used to access arrays are within the declared range. This is
4736 currently only supported by the Java and Fortran front-ends, where
4737 this option defaults to true and false respectively.
4739 @item -fmudflap -fmudflapth -fmudflapir
4743 @cindex bounds checking
4745 For front-ends that support it (C and C++), instrument all risky
4746 pointer/array dereferencing operations, some standard library
4747 string/heap functions, and some other associated constructs with
4748 range/validity tests. Modules so instrumented should be immune to
4749 buffer overflows, invalid heap use, and some other classes of C/C++
4750 programming errors. The instrumentation relies on a separate runtime
4751 library (@file{libmudflap}), which will be linked into a program if
4752 @option{-fmudflap} is given at link time. Run-time behavior of the
4753 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4754 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4757 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4758 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4759 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4760 instrumentation should ignore pointer reads. This produces less
4761 instrumentation (and therefore faster execution) and still provides
4762 some protection against outright memory corrupting writes, but allows
4763 erroneously read data to propagate within a program.
4765 @item -fthread-jumps
4766 @opindex fthread-jumps
4767 Perform optimizations where we check to see if a jump branches to a
4768 location where another comparison subsumed by the first is found. If
4769 so, the first branch is redirected to either the destination of the
4770 second branch or a point immediately following it, depending on whether
4771 the condition is known to be true or false.
4773 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4775 @item -fcse-follow-jumps
4776 @opindex fcse-follow-jumps
4777 In common subexpression elimination, scan through jump instructions
4778 when the target of the jump is not reached by any other path. For
4779 example, when CSE encounters an @code{if} statement with an
4780 @code{else} clause, CSE will follow the jump when the condition
4783 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4785 @item -fcse-skip-blocks
4786 @opindex fcse-skip-blocks
4787 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4788 follow jumps which conditionally skip over blocks. When CSE
4789 encounters a simple @code{if} statement with no else clause,
4790 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4791 body of the @code{if}.
4793 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4795 @item -frerun-cse-after-loop
4796 @opindex frerun-cse-after-loop
4797 Re-run common subexpression elimination after loop optimizations has been
4800 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4804 Perform a global common subexpression elimination pass.
4805 This pass also performs global constant and copy propagation.
4807 @emph{Note:} When compiling a program using computed gotos, a GCC
4808 extension, you may get better runtime performance if you disable
4809 the global common subexpression elimination pass by adding
4810 @option{-fno-gcse} to the command line.
4812 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4816 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4817 attempt to move loads which are only killed by stores into themselves. This
4818 allows a loop containing a load/store sequence to be changed to a load outside
4819 the loop, and a copy/store within the loop.
4821 Enabled by default when gcse is enabled.
4825 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4826 global common subexpression elimination. This pass will attempt to move
4827 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4828 loops containing a load/store sequence can be changed to a load before
4829 the loop and a store after the loop.
4831 Not enabled at any optimization level.
4835 When @option{-fgcse-las} is enabled, the global common subexpression
4836 elimination pass eliminates redundant loads that come after stores to the
4837 same memory location (both partial and full redundancies).
4839 Not enabled at any optimization level.
4841 @item -fgcse-after-reload
4842 @opindex fgcse-after-reload
4843 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4844 pass is performed after reload. The purpose of this pass is to cleanup
4847 @item -funsafe-loop-optimizations
4848 @opindex funsafe-loop-optimizations
4849 If given, the loop optimizer will assume that loop indices do not
4850 overflow, and that the loops with nontrivial exit condition are not
4851 infinite. This enables a wider range of loop optimizations even if
4852 the loop optimizer itself cannot prove that these assumptions are valid.
4853 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
4854 if it finds this kind of loop.
4856 @item -fcrossjumping
4857 @opindex crossjumping
4858 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4859 resulting code may or may not perform better than without cross-jumping.
4861 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4863 @item -fif-conversion
4864 @opindex if-conversion
4865 Attempt to transform conditional jumps into branch-less equivalents. This
4866 include use of conditional moves, min, max, set flags and abs instructions, and
4867 some tricks doable by standard arithmetics. The use of conditional execution
4868 on chips where it is available is controlled by @code{if-conversion2}.
4870 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4872 @item -fif-conversion2
4873 @opindex if-conversion2
4874 Use conditional execution (where available) to transform conditional jumps into
4875 branch-less equivalents.
4877 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4879 @item -fdelete-null-pointer-checks
4880 @opindex fdelete-null-pointer-checks
4881 Use global dataflow analysis to identify and eliminate useless checks
4882 for null pointers. The compiler assumes that dereferencing a null
4883 pointer would have halted the program. If a pointer is checked after
4884 it has already been dereferenced, it cannot be null.
4886 In some environments, this assumption is not true, and programs can
4887 safely dereference null pointers. Use
4888 @option{-fno-delete-null-pointer-checks} to disable this optimization
4889 for programs which depend on that behavior.
4891 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4893 @item -fexpensive-optimizations
4894 @opindex fexpensive-optimizations
4895 Perform a number of minor optimizations that are relatively expensive.
4897 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4899 @item -foptimize-register-move
4901 @opindex foptimize-register-move
4903 Attempt to reassign register numbers in move instructions and as
4904 operands of other simple instructions in order to maximize the amount of
4905 register tying. This is especially helpful on machines with two-operand
4908 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4911 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4913 @item -fdelayed-branch
4914 @opindex fdelayed-branch
4915 If supported for the target machine, attempt to reorder instructions
4916 to exploit instruction slots available after delayed branch
4919 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4921 @item -fschedule-insns
4922 @opindex fschedule-insns
4923 If supported for the target machine, attempt to reorder instructions to
4924 eliminate execution stalls due to required data being unavailable. This
4925 helps machines that have slow floating point or memory load instructions
4926 by allowing other instructions to be issued until the result of the load
4927 or floating point instruction is required.
4929 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4931 @item -fschedule-insns2
4932 @opindex fschedule-insns2
4933 Similar to @option{-fschedule-insns}, but requests an additional pass of
4934 instruction scheduling after register allocation has been done. This is
4935 especially useful on machines with a relatively small number of
4936 registers and where memory load instructions take more than one cycle.
4938 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4940 @item -fno-sched-interblock
4941 @opindex fno-sched-interblock
4942 Don't schedule instructions across basic blocks. This is normally
4943 enabled by default when scheduling before register allocation, i.e.@:
4944 with @option{-fschedule-insns} or at @option{-O2} or higher.
4946 @item -fno-sched-spec
4947 @opindex fno-sched-spec
4948 Don't allow speculative motion of non-load instructions. This is normally
4949 enabled by default when scheduling before register allocation, i.e.@:
4950 with @option{-fschedule-insns} or at @option{-O2} or higher.
4952 @item -fsched-spec-load
4953 @opindex fsched-spec-load
4954 Allow speculative motion of some load instructions. This only makes
4955 sense when scheduling before register allocation, i.e.@: with
4956 @option{-fschedule-insns} or at @option{-O2} or higher.
4958 @item -fsched-spec-load-dangerous
4959 @opindex fsched-spec-load-dangerous
4960 Allow speculative motion of more load instructions. This only makes
4961 sense when scheduling before register allocation, i.e.@: with
4962 @option{-fschedule-insns} or at @option{-O2} or higher.
4964 @item -fsched-stalled-insns=@var{n}
4965 @opindex fsched-stalled-insns
4966 Define how many insns (if any) can be moved prematurely from the queue
4967 of stalled insns into the ready list, during the second scheduling pass.
4969 @item -fsched-stalled-insns-dep=@var{n}
4970 @opindex fsched-stalled-insns-dep
4971 Define how many insn groups (cycles) will be examined for a dependency
4972 on a stalled insn that is candidate for premature removal from the queue
4973 of stalled insns. Has an effect only during the second scheduling pass,
4974 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4976 @item -fsched2-use-superblocks
4977 @opindex fsched2-use-superblocks
4978 When scheduling after register allocation, do use superblock scheduling
4979 algorithm. Superblock scheduling allows motion across basic block boundaries
4980 resulting on faster schedules. This option is experimental, as not all machine
4981 descriptions used by GCC model the CPU closely enough to avoid unreliable
4982 results from the algorithm.
4984 This only makes sense when scheduling after register allocation, i.e.@: with
4985 @option{-fschedule-insns2} or at @option{-O2} or higher.
4987 @item -fsched2-use-traces
4988 @opindex fsched2-use-traces
4989 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4990 allocation and additionally perform code duplication in order to increase the
4991 size of superblocks using tracer pass. See @option{-ftracer} for details on
4994 This mode should produce faster but significantly longer programs. Also
4995 without @option{-fbranch-probabilities} the traces constructed may not
4996 match the reality and hurt the performance. This only makes
4997 sense when scheduling after register allocation, i.e.@: with
4998 @option{-fschedule-insns2} or at @option{-O2} or higher.
5002 Eliminates redundant extension instructions and move the non redundant
5003 ones to optimal placement using LCM.
5005 @item -freschedule-modulo-scheduled-loops
5006 @opindex fscheduling-in-modulo-scheduled-loops
5007 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
5008 we may want to prevent the later scheduling passes from changing its schedule, we use this
5009 option to control that.
5011 @item -fcaller-saves
5012 @opindex fcaller-saves
5013 Enable values to be allocated in registers that will be clobbered by
5014 function calls, by emitting extra instructions to save and restore the
5015 registers around such calls. Such allocation is done only when it
5016 seems to result in better code than would otherwise be produced.
5018 This option is always enabled by default on certain machines, usually
5019 those which have no call-preserved registers to use instead.
5021 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5024 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
5025 enabled by default at @option{-O2} and @option{-O3}.
5028 Perform Full Redundancy Elimination (FRE) on trees. The difference
5029 between FRE and PRE is that FRE only considers expressions
5030 that are computed on all paths leading to the redundant computation.
5031 This analysis faster than PRE, though it exposes fewer redundancies.
5032 This flag is enabled by default at @option{-O} and higher.
5034 @item -ftree-copy-prop
5035 Perform copy propagation on trees. This pass eliminates unnecessary
5036 copy operations. This flag is enabled by default at @option{-O} and
5039 @item -ftree-store-copy-prop
5040 Perform copy propagation of memory loads and stores. This pass
5041 eliminates unnecessary copy operations in memory references
5042 (structures, global variables, arrays, etc). This flag is enabled by
5043 default at @option{-O2} and higher.
5046 Perform structural alias analysis on trees. This flag
5047 is enabled by default at @option{-O} and higher.
5050 Perform interprocedural pointer analysis.
5053 Perform forward store motion on trees. This flag is
5054 enabled by default at @option{-O} and higher.
5057 Perform sparse conditional constant propagation (CCP) on trees. This
5058 pass only operates on local scalar variables and is enabled by default
5059 at @option{-O} and higher.
5061 @item -ftree-store-ccp
5062 Perform sparse conditional constant propagation (CCP) on trees. This
5063 pass operates on both local scalar variables and memory stores and
5064 loads (global variables, structures, arrays, etc). This flag is
5065 enabled by default at @option{-O2} and higher.
5068 Perform dead code elimination (DCE) on trees. This flag is enabled by
5069 default at @option{-O} and higher.
5071 @item -ftree-dominator-opts
5072 Perform a variety of simple scalar cleanups (constant/copy
5073 propagation, redundancy elimination, range propagation and expression
5074 simplification) based on a dominator tree traversal. This also
5075 performs jump threading (to reduce jumps to jumps). This flag is
5076 enabled by default at @option{-O} and higher.
5079 Perform loop header copying on trees. This is beneficial since it increases
5080 effectiveness of code motion optimizations. It also saves one jump. This flag
5081 is enabled by default at @option{-O} and higher. It is not enabled
5082 for @option{-Os}, since it usually increases code size.
5084 @item -ftree-loop-optimize
5085 Perform loop optimizations on trees. This flag is enabled by default
5086 at @option{-O} and higher.
5088 @item -ftree-loop-linear
5089 Perform linear loop transformations on tree. This flag can improve cache
5090 performance and allow further loop optimizations to take place.
5092 @item -ftree-loop-im
5093 Perform loop invariant motion on trees. This pass moves only invariants that
5094 would be hard to handle at RTL level (function calls, operations that expand to
5095 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5096 operands of conditions that are invariant out of the loop, so that we can use
5097 just trivial invariantness analysis in loop unswitching. The pass also includes
5100 @item -ftree-loop-ivcanon
5101 Create a canonical counter for number of iterations in the loop for that
5102 determining number of iterations requires complicated analysis. Later
5103 optimizations then may determine the number easily. Useful especially
5104 in connection with unrolling.
5107 Perform induction variable optimizations (strength reduction, induction
5108 variable merging and induction variable elimination) on trees.
5111 Perform scalar replacement of aggregates. This pass replaces structure
5112 references with scalars to prevent committing structures to memory too
5113 early. This flag is enabled by default at @option{-O} and higher.
5115 @item -ftree-copyrename
5116 Perform copy renaming on trees. This pass attempts to rename compiler
5117 temporaries to other variables at copy locations, usually resulting in
5118 variable names which more closely resemble the original variables. This flag
5119 is enabled by default at @option{-O} and higher.
5122 Perform temporary expression replacement during the SSA->normal phase. Single
5123 use/single def temporaries are replaced at their use location with their
5124 defining expression. This results in non-GIMPLE code, but gives the expanders
5125 much more complex trees to work on resulting in better RTL generation. This is
5126 enabled by default at @option{-O} and higher.
5129 Perform live range splitting during the SSA->normal phase. Distinct live
5130 ranges of a variable are split into unique variables, allowing for better
5131 optimization later. This is enabled by default at @option{-O} and higher.
5133 @item -ftree-vectorize
5134 Perform loop vectorization on trees.
5136 @item -ftree-vect-loop-version
5137 @opindex ftree-vect-loop-version
5138 Perform loop versioning when doing loop vectorization on trees. When a loop
5139 appears to be vectorizable except that data alignment or data dependence cannot
5140 be determined at compile time then vectorized and non-vectorized versions of
5141 the loop are generated along with runtime checks for alignment or dependence
5142 to control which version is executed. This option is enabled by default
5143 except at level @option{-Os} where it is disabled.
5146 Perform Value Range Propagation on trees. This is similar to the
5147 constant propagation pass, but instead of values, ranges of values are
5148 propagated. This allows the optimizers to remove unnecessary range
5149 checks like array bound checks and null pointer checks. This is
5150 enabled by default at @option{-O2} and higher. Null pointer check
5151 elimination is only done if @option{-fdelete-null-pointer-checks} is
5156 Perform tail duplication to enlarge superblock size. This transformation
5157 simplifies the control flow of the function allowing other optimizations to do
5160 @item -funroll-loops
5161 @opindex funroll-loops
5162 Unroll loops whose number of iterations can be determined at compile
5163 time or upon entry to the loop. @option{-funroll-loops} implies
5164 @option{-frerun-cse-after-loop}. This option makes code larger,
5165 and may or may not make it run faster.
5167 @item -funroll-all-loops
5168 @opindex funroll-all-loops
5169 Unroll all loops, even if their number of iterations is uncertain when
5170 the loop is entered. This usually makes programs run more slowly.
5171 @option{-funroll-all-loops} implies the same options as
5172 @option{-funroll-loops},
5174 @item -fsplit-ivs-in-unroller
5175 @opindex fsplit-ivs-in-unroller
5176 Enables expressing of values of induction variables in later iterations
5177 of the unrolled loop using the value in the first iteration. This breaks
5178 long dependency chains, thus improving efficiency of the scheduling passes.
5180 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5181 same effect. However in cases the loop body is more complicated than
5182 a single basic block, this is not reliable. It also does not work at all
5183 on some of the architectures due to restrictions in the CSE pass.
5185 This optimization is enabled by default.
5187 @item -fvariable-expansion-in-unroller
5188 @opindex fvariable-expansion-in-unroller
5189 With this option, the compiler will create multiple copies of some
5190 local variables when unrolling a loop which can result in superior code.
5192 @item -fprefetch-loop-arrays
5193 @opindex fprefetch-loop-arrays
5194 If supported by the target machine, generate instructions to prefetch
5195 memory to improve the performance of loops that access large arrays.
5197 This option may generate better or worse code; results are highly
5198 dependent on the structure of loops within the source code.
5200 Disabled at level @option{-Os}.
5203 @itemx -fno-peephole2
5204 @opindex fno-peephole
5205 @opindex fno-peephole2
5206 Disable any machine-specific peephole optimizations. The difference
5207 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5208 are implemented in the compiler; some targets use one, some use the
5209 other, a few use both.
5211 @option{-fpeephole} is enabled by default.
5212 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5214 @item -fno-guess-branch-probability
5215 @opindex fno-guess-branch-probability
5216 Do not guess branch probabilities using heuristics.
5218 GCC will use heuristics to guess branch probabilities if they are
5219 not provided by profiling feedback (@option{-fprofile-arcs}). These
5220 heuristics are based on the control flow graph. If some branch probabilities
5221 are specified by @samp{__builtin_expect}, then the heuristics will be
5222 used to guess branch probabilities for the rest of the control flow graph,
5223 taking the @samp{__builtin_expect} info into account. The interactions
5224 between the heuristics and @samp{__builtin_expect} can be complex, and in
5225 some cases, it may be useful to disable the heuristics so that the effects
5226 of @samp{__builtin_expect} are easier to understand.
5228 The default is @option{-fguess-branch-probability} at levels
5229 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5231 @item -freorder-blocks
5232 @opindex freorder-blocks
5233 Reorder basic blocks in the compiled function in order to reduce number of
5234 taken branches and improve code locality.
5236 Enabled at levels @option{-O2}, @option{-O3}.
5238 @item -freorder-blocks-and-partition
5239 @opindex freorder-blocks-and-partition
5240 In addition to reordering basic blocks in the compiled function, in order
5241 to reduce number of taken branches, partitions hot and cold basic blocks
5242 into separate sections of the assembly and .o files, to improve
5243 paging and cache locality performance.
5245 This optimization is automatically turned off in the presence of
5246 exception handling, for linkonce sections, for functions with a user-defined
5247 section attribute and on any architecture that does not support named
5250 @item -freorder-functions
5251 @opindex freorder-functions
5252 Reorder functions in the object file in order to
5253 improve code locality. This is implemented by using special
5254 subsections @code{.text.hot} for most frequently executed functions and
5255 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5256 the linker so object file format must support named sections and linker must
5257 place them in a reasonable way.
5259 Also profile feedback must be available in to make this option effective. See
5260 @option{-fprofile-arcs} for details.
5262 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5264 @item -fstrict-aliasing
5265 @opindex fstrict-aliasing
5266 Allows the compiler to assume the strictest aliasing rules applicable to
5267 the language being compiled. For C (and C++), this activates
5268 optimizations based on the type of expressions. In particular, an
5269 object of one type is assumed never to reside at the same address as an
5270 object of a different type, unless the types are almost the same. For
5271 example, an @code{unsigned int} can alias an @code{int}, but not a
5272 @code{void*} or a @code{double}. A character type may alias any other
5275 Pay special attention to code like this:
5288 The practice of reading from a different union member than the one most
5289 recently written to (called ``type-punning'') is common. Even with
5290 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5291 is accessed through the union type. So, the code above will work as
5292 expected. However, this code might not:
5303 Every language that wishes to perform language-specific alias analysis
5304 should define a function that computes, given an @code{tree}
5305 node, an alias set for the node. Nodes in different alias sets are not
5306 allowed to alias. For an example, see the C front-end function
5307 @code{c_get_alias_set}.
5309 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5311 @item -fstrict-overflow
5312 @opindex fstrict-overflow
5313 Allow the compiler to assume strict signed overflow rules, depending
5314 on the language being compiled. For C (and C++) this means that
5315 overflow when doing arithmetic with signed numbers is undefined, which
5316 means that the compiler may assume that it will not happen. This
5317 permits various optimizations. For example, the compiler will assume
5318 that an expression like @code{i + 10 > i} will always be true for
5319 signed @code{i}. This assumption is only valid if signed overflow is
5320 undefined, as the expression is false if @code{i + 10} overflows when
5321 using twos complement arithmetic. When this option is in effect any
5322 attempt to determine whether an operation on signed numbers will
5323 overflow must be written carefully to not actually involve overflow.
5325 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
5326 that signed overflow is fully defined: it wraps. When
5327 @option{-fwrapv} is used, there is no difference between
5328 @option{-fstrict-overflow} and @option{-fno-strict-overflow}. With
5329 @option{-fwrapv} certain types of overflow are permitted. For
5330 example, if the compiler gets an overflow when doing arithmetic on
5331 constants, the overflowed value can still be used with
5332 @option{-fwrapv}, but not otherwise.
5334 The @option{-fstrict-overflow} option is enabled at levels
5335 @option{-O2}, @option{-O3}, @option{-Os}.
5337 @item -falign-functions
5338 @itemx -falign-functions=@var{n}
5339 @opindex falign-functions
5340 Align the start of functions to the next power-of-two greater than
5341 @var{n}, skipping up to @var{n} bytes. For instance,
5342 @option{-falign-functions=32} aligns functions to the next 32-byte
5343 boundary, but @option{-falign-functions=24} would align to the next
5344 32-byte boundary only if this can be done by skipping 23 bytes or less.
5346 @option{-fno-align-functions} and @option{-falign-functions=1} are
5347 equivalent and mean that functions will not be aligned.
5349 Some assemblers only support this flag when @var{n} is a power of two;
5350 in that case, it is rounded up.
5352 If @var{n} is not specified or is zero, use a machine-dependent default.
5354 Enabled at levels @option{-O2}, @option{-O3}.
5356 @item -falign-labels
5357 @itemx -falign-labels=@var{n}
5358 @opindex falign-labels
5359 Align all branch targets to a power-of-two boundary, skipping up to
5360 @var{n} bytes like @option{-falign-functions}. This option can easily
5361 make code slower, because it must insert dummy operations for when the
5362 branch target is reached in the usual flow of the code.
5364 @option{-fno-align-labels} and @option{-falign-labels=1} are
5365 equivalent and mean that labels will not be aligned.
5367 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5368 are greater than this value, then their values are used instead.
5370 If @var{n} is not specified or is zero, use a machine-dependent default
5371 which is very likely to be @samp{1}, meaning no alignment.
5373 Enabled at levels @option{-O2}, @option{-O3}.
5376 @itemx -falign-loops=@var{n}
5377 @opindex falign-loops
5378 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5379 like @option{-falign-functions}. The hope is that the loop will be
5380 executed many times, which will make up for any execution of the dummy
5383 @option{-fno-align-loops} and @option{-falign-loops=1} are
5384 equivalent and mean that loops will not be aligned.
5386 If @var{n} is not specified or is zero, use a machine-dependent default.
5388 Enabled at levels @option{-O2}, @option{-O3}.
5391 @itemx -falign-jumps=@var{n}
5392 @opindex falign-jumps
5393 Align branch targets to a power-of-two boundary, for branch targets
5394 where the targets can only be reached by jumping, skipping up to @var{n}
5395 bytes like @option{-falign-functions}. In this case, no dummy operations
5398 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5399 equivalent and mean that loops will not be aligned.
5401 If @var{n} is not specified or is zero, use a machine-dependent default.
5403 Enabled at levels @option{-O2}, @option{-O3}.
5405 @item -funit-at-a-time
5406 @opindex funit-at-a-time
5407 Parse the whole compilation unit before starting to produce code.
5408 This allows some extra optimizations to take place but consumes
5409 more memory (in general). There are some compatibility issues
5410 with @emph{unit-at-a-time} mode:
5413 enabling @emph{unit-at-a-time} mode may change the order
5414 in which functions, variables, and top-level @code{asm} statements
5415 are emitted, and will likely break code relying on some particular
5416 ordering. The majority of such top-level @code{asm} statements,
5417 though, can be replaced by @code{section} attributes. The
5418 @option{fno-toplevel-reorder} option may be used to keep the ordering
5419 used in the input file, at the cost of some optimizations.
5422 @emph{unit-at-a-time} mode removes unreferenced static variables
5423 and functions. This may result in undefined references
5424 when an @code{asm} statement refers directly to variables or functions
5425 that are otherwise unused. In that case either the variable/function
5426 shall be listed as an operand of the @code{asm} statement operand or,
5427 in the case of top-level @code{asm} statements the attribute @code{used}
5428 shall be used on the declaration.
5431 Static functions now can use non-standard passing conventions that
5432 may break @code{asm} statements calling functions directly. Again,
5433 attribute @code{used} will prevent this behavior.
5436 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5437 but this scheme may not be supported by future releases of GCC@.
5439 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5441 @item -fno-toplevel-reorder
5442 Do not reorder top-level functions, variables, and @code{asm}
5443 statements. Output them in the same order that they appear in the
5444 input file. When this option is used, unreferenced static variables
5445 will not be removed. This option is intended to support existing code
5446 which relies on a particular ordering. For new code, it is better to
5451 Constructs webs as commonly used for register allocation purposes and assign
5452 each web individual pseudo register. This allows the register allocation pass
5453 to operate on pseudos directly, but also strengthens several other optimization
5454 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5455 however, make debugging impossible, since variables will no longer stay in a
5458 Enabled by default with @option{-funroll-loops}.
5460 @item -fwhole-program
5461 @opindex fwhole-program
5462 Assume that the current compilation unit represents whole program being
5463 compiled. All public functions and variables with the exception of @code{main}
5464 and those merged by attribute @code{externally_visible} become static functions
5465 and in a affect gets more aggressively optimized by interprocedural optimizers.
5466 While this option is equivalent to proper use of @code{static} keyword for
5467 programs consisting of single file, in combination with option
5468 @option{--combine} this flag can be used to compile most of smaller scale C
5469 programs since the functions and variables become local for the whole combined
5470 compilation unit, not for the single source file itself.
5473 @item -fno-cprop-registers
5474 @opindex fno-cprop-registers
5475 After register allocation and post-register allocation instruction splitting,
5476 we perform a copy-propagation pass to try to reduce scheduling dependencies
5477 and occasionally eliminate the copy.
5479 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5481 @item -fprofile-generate
5482 @opindex fprofile-generate
5484 Enable options usually used for instrumenting application to produce
5485 profile useful for later recompilation with profile feedback based
5486 optimization. You must use @option{-fprofile-generate} both when
5487 compiling and when linking your program.
5489 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5492 @opindex fprofile-use
5493 Enable profile feedback directed optimizations, and optimizations
5494 generally profitable only with profile feedback available.
5496 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
5497 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
5501 The following options control compiler behavior regarding floating
5502 point arithmetic. These options trade off between speed and
5503 correctness. All must be specifically enabled.
5507 @opindex ffloat-store
5508 Do not store floating point variables in registers, and inhibit other
5509 options that might change whether a floating point value is taken from a
5512 @cindex floating point precision
5513 This option prevents undesirable excess precision on machines such as
5514 the 68000 where the floating registers (of the 68881) keep more
5515 precision than a @code{double} is supposed to have. Similarly for the
5516 x86 architecture. For most programs, the excess precision does only
5517 good, but a few programs rely on the precise definition of IEEE floating
5518 point. Use @option{-ffloat-store} for such programs, after modifying
5519 them to store all pertinent intermediate computations into variables.
5523 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5524 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5525 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5526 and @option{fcx-limited-range}.
5528 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5530 This option should never be turned on by any @option{-O} option since
5531 it can result in incorrect output for programs which depend on
5532 an exact implementation of IEEE or ISO rules/specifications for
5535 @item -fno-math-errno
5536 @opindex fno-math-errno
5537 Do not set ERRNO after calling math functions that are executed
5538 with a single instruction, e.g., sqrt. A program that relies on
5539 IEEE exceptions for math error handling may want to use this flag
5540 for speed while maintaining IEEE arithmetic compatibility.
5542 This option should never be turned on by any @option{-O} option since
5543 it can result in incorrect output for programs which depend on
5544 an exact implementation of IEEE or ISO rules/specifications for
5547 The default is @option{-fmath-errno}.
5549 On Darwin and FreeBSD systems, the math library never sets @code{errno}.
5551 no reason for the compiler to consider the possibility that it might,
5552 and @option{-fno-math-errno} is the default.
5554 @item -funsafe-math-optimizations
5555 @opindex funsafe-math-optimizations
5556 Allow optimizations for floating-point arithmetic that (a) assume
5557 that arguments and results are valid and (b) may violate IEEE or
5558 ANSI standards. When used at link-time, it may include libraries
5559 or startup files that change the default FPU control word or other
5560 similar optimizations.
5562 This option should never be turned on by any @option{-O} option since
5563 it can result in incorrect output for programs which depend on
5564 an exact implementation of IEEE or ISO rules/specifications for
5567 The default is @option{-fno-unsafe-math-optimizations}.
5569 @item -ffinite-math-only
5570 @opindex ffinite-math-only
5571 Allow optimizations for floating-point arithmetic that assume
5572 that arguments and results are not NaNs or +-Infs.
5574 This option should never be turned on by any @option{-O} option since
5575 it can result in incorrect output for programs which depend on
5576 an exact implementation of IEEE or ISO rules/specifications.
5578 The default is @option{-fno-finite-math-only}.
5580 @item -fno-trapping-math
5581 @opindex fno-trapping-math
5582 Compile code assuming that floating-point operations cannot generate
5583 user-visible traps. These traps include division by zero, overflow,
5584 underflow, inexact result and invalid operation. This option implies
5585 @option{-fno-signaling-nans}. Setting this option may allow faster
5586 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5588 This option should never be turned on by any @option{-O} option since
5589 it can result in incorrect output for programs which depend on
5590 an exact implementation of IEEE or ISO rules/specifications for
5593 The default is @option{-ftrapping-math}.
5595 @item -frounding-math
5596 @opindex frounding-math
5597 Disable transformations and optimizations that assume default floating
5598 point rounding behavior. This is round-to-zero for all floating point
5599 to integer conversions, and round-to-nearest for all other arithmetic
5600 truncations. This option should be specified for programs that change
5601 the FP rounding mode dynamically, or that may be executed with a
5602 non-default rounding mode. This option disables constant folding of
5603 floating point expressions at compile-time (which may be affected by
5604 rounding mode) and arithmetic transformations that are unsafe in the
5605 presence of sign-dependent rounding modes.
5607 The default is @option{-fno-rounding-math}.
5609 This option is experimental and does not currently guarantee to
5610 disable all GCC optimizations that are affected by rounding mode.
5611 Future versions of GCC may provide finer control of this setting
5612 using C99's @code{FENV_ACCESS} pragma. This command line option
5613 will be used to specify the default state for @code{FENV_ACCESS}.
5615 @item -frtl-abstract-sequences
5616 @opindex frtl-abstract-sequences
5617 It is a size optimization method. This option is to find identical
5618 sequences of code, which can be turned into pseudo-procedures and
5619 then replace all occurrences with calls to the newly created
5620 subroutine. It is kind of an opposite of @option{-finline-functions}.
5621 This optimization runs at RTL level.
5623 @item -fsignaling-nans
5624 @opindex fsignaling-nans
5625 Compile code assuming that IEEE signaling NaNs may generate user-visible
5626 traps during floating-point operations. Setting this option disables
5627 optimizations that may change the number of exceptions visible with
5628 signaling NaNs. This option implies @option{-ftrapping-math}.
5630 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5633 The default is @option{-fno-signaling-nans}.
5635 This option is experimental and does not currently guarantee to
5636 disable all GCC optimizations that affect signaling NaN behavior.
5638 @item -fsingle-precision-constant
5639 @opindex fsingle-precision-constant
5640 Treat floating point constant as single precision constant instead of
5641 implicitly converting it to double precision constant.
5643 @item -fcx-limited-range
5644 @itemx -fno-cx-limited-range
5645 @opindex fcx-limited-range
5646 @opindex fno-cx-limited-range
5647 When enabled, this option states that a range reduction step is not
5648 needed when performing complex division. The default is
5649 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5651 This option controls the default setting of the ISO C99
5652 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5657 The following options control optimizations that may improve
5658 performance, but are not enabled by any @option{-O} options. This
5659 section includes experimental options that may produce broken code.
5662 @item -fbranch-probabilities
5663 @opindex fbranch-probabilities
5664 After running a program compiled with @option{-fprofile-arcs}
5665 (@pxref{Debugging Options,, Options for Debugging Your Program or
5666 @command{gcc}}), you can compile it a second time using
5667 @option{-fbranch-probabilities}, to improve optimizations based on
5668 the number of times each branch was taken. When the program
5669 compiled with @option{-fprofile-arcs} exits it saves arc execution
5670 counts to a file called @file{@var{sourcename}.gcda} for each source
5671 file The information in this data file is very dependent on the
5672 structure of the generated code, so you must use the same source code
5673 and the same optimization options for both compilations.
5675 With @option{-fbranch-probabilities}, GCC puts a
5676 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5677 These can be used to improve optimization. Currently, they are only
5678 used in one place: in @file{reorg.c}, instead of guessing which path a
5679 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5680 exactly determine which path is taken more often.
5682 @item -fprofile-values
5683 @opindex fprofile-values
5684 If combined with @option{-fprofile-arcs}, it adds code so that some
5685 data about values of expressions in the program is gathered.
5687 With @option{-fbranch-probabilities}, it reads back the data gathered
5688 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5689 notes to instructions for their later usage in optimizations.
5691 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5695 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5696 a code to gather information about values of expressions.
5698 With @option{-fbranch-probabilities}, it reads back the data gathered
5699 and actually performs the optimizations based on them.
5700 Currently the optimizations include specialization of division operation
5701 using the knowledge about the value of the denominator.
5703 @item -frename-registers
5704 @opindex frename-registers
5705 Attempt to avoid false dependencies in scheduled code by making use
5706 of registers left over after register allocation. This optimization
5707 will most benefit processors with lots of registers. Depending on the
5708 debug information format adopted by the target, however, it can
5709 make debugging impossible, since variables will no longer stay in
5710 a ``home register''.
5712 Enabled by default with @option{-funroll-loops}.
5716 Perform tail duplication to enlarge superblock size. This transformation
5717 simplifies the control flow of the function allowing other optimizations to do
5720 Enabled with @option{-fprofile-use}.
5722 @item -funroll-loops
5723 @opindex funroll-loops
5724 Unroll loops whose number of iterations can be determined at compile time or
5725 upon entry to the loop. @option{-funroll-loops} implies
5726 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
5727 It also turns on complete loop peeling (i.e.@: complete removal of loops with
5728 small constant number of iterations). This option makes code larger, and may
5729 or may not make it run faster.
5731 Enabled with @option{-fprofile-use}.
5733 @item -funroll-all-loops
5734 @opindex funroll-all-loops
5735 Unroll all loops, even if their number of iterations is uncertain when
5736 the loop is entered. This usually makes programs run more slowly.
5737 @option{-funroll-all-loops} implies the same options as
5738 @option{-funroll-loops}.
5741 @opindex fpeel-loops
5742 Peels the loops for that there is enough information that they do not
5743 roll much (from profile feedback). It also turns on complete loop peeling
5744 (i.e.@: complete removal of loops with small constant number of iterations).
5746 Enabled with @option{-fprofile-use}.
5748 @item -fmove-loop-invariants
5749 @opindex fmove-loop-invariants
5750 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
5751 at level @option{-O1}
5753 @item -funswitch-loops
5754 @opindex funswitch-loops
5755 Move branches with loop invariant conditions out of the loop, with duplicates
5756 of the loop on both branches (modified according to result of the condition).
5758 @item -ffunction-sections
5759 @itemx -fdata-sections
5760 @opindex ffunction-sections
5761 @opindex fdata-sections
5762 Place each function or data item into its own section in the output
5763 file if the target supports arbitrary sections. The name of the
5764 function or the name of the data item determines the section's name
5767 Use these options on systems where the linker can perform optimizations
5768 to improve locality of reference in the instruction space. Most systems
5769 using the ELF object format and SPARC processors running Solaris 2 have
5770 linkers with such optimizations. AIX may have these optimizations in
5773 Only use these options when there are significant benefits from doing
5774 so. When you specify these options, the assembler and linker will
5775 create larger object and executable files and will also be slower.
5776 You will not be able to use @code{gprof} on all systems if you
5777 specify this option and you may have problems with debugging if
5778 you specify both this option and @option{-g}.
5780 @item -fbranch-target-load-optimize
5781 @opindex fbranch-target-load-optimize
5782 Perform branch target register load optimization before prologue / epilogue
5784 The use of target registers can typically be exposed only during reload,
5785 thus hoisting loads out of loops and doing inter-block scheduling needs
5786 a separate optimization pass.
5788 @item -fbranch-target-load-optimize2
5789 @opindex fbranch-target-load-optimize2
5790 Perform branch target register load optimization after prologue / epilogue
5793 @item -fbtr-bb-exclusive
5794 @opindex fbtr-bb-exclusive
5795 When performing branch target register load optimization, don't reuse
5796 branch target registers in within any basic block.
5798 @item -fstack-protector
5799 Emit extra code to check for buffer overflows, such as stack smashing
5800 attacks. This is done by adding a guard variable to functions with
5801 vulnerable objects. This includes functions that call alloca, and
5802 functions with buffers larger than 8 bytes. The guards are initialized
5803 when a function is entered and then checked when the function exits.
5804 If a guard check fails, an error message is printed and the program exits.
5806 @item -fstack-protector-all
5807 Like @option{-fstack-protector} except that all functions are protected.
5809 @item -fsection-anchors
5810 @opindex fsection-anchors
5811 Try to reduce the number of symbolic address calculations by using
5812 shared ``anchor'' symbols to address nearby objects. This transformation
5813 can help to reduce the number of GOT entries and GOT accesses on some
5816 For example, the implementation of the following function @code{foo}:
5820 int foo (void) @{ return a + b + c; @}
5823 would usually calculate the addresses of all three variables, but if you
5824 compile it with @option{-fsection-anchors}, it will access the variables
5825 from a common anchor point instead. The effect is similar to the
5826 following pseudocode (which isn't valid C):
5831 register int *xr = &x;
5832 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5836 Not all targets support this option.
5838 @item --param @var{name}=@var{value}
5840 In some places, GCC uses various constants to control the amount of
5841 optimization that is done. For example, GCC will not inline functions
5842 that contain more that a certain number of instructions. You can
5843 control some of these constants on the command-line using the
5844 @option{--param} option.
5846 The names of specific parameters, and the meaning of the values, are
5847 tied to the internals of the compiler, and are subject to change
5848 without notice in future releases.
5850 In each case, the @var{value} is an integer. The allowable choices for
5851 @var{name} are given in the following table:
5854 @item salias-max-implicit-fields
5855 The maximum number of fields in a variable without direct
5856 structure accesses for which structure aliasing will consider trying
5857 to track each field. The default is 5
5859 @item salias-max-array-elements
5860 The maximum number of elements an array can have and its elements
5861 still be tracked individually by structure aliasing. The default is 4
5863 @item sra-max-structure-size
5864 The maximum structure size, in bytes, at which the scalar replacement
5865 of aggregates (SRA) optimization will perform block copies. The
5866 default value, 0, implies that GCC will select the most appropriate
5869 @item sra-field-structure-ratio
5870 The threshold ratio (as a percentage) between instantiated fields and
5871 the complete structure size. We say that if the ratio of the number
5872 of bytes in instantiated fields to the number of bytes in the complete
5873 structure exceeds this parameter, then block copies are not used. The
5876 @item max-crossjump-edges
5877 The maximum number of incoming edges to consider for crossjumping.
5878 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5879 the number of edges incoming to each block. Increasing values mean
5880 more aggressive optimization, making the compile time increase with
5881 probably small improvement in executable size.
5883 @item min-crossjump-insns
5884 The minimum number of instructions which must be matched at the end
5885 of two blocks before crossjumping will be performed on them. This
5886 value is ignored in the case where all instructions in the block being
5887 crossjumped from are matched. The default value is 5.
5889 @item max-grow-copy-bb-insns
5890 The maximum code size expansion factor when copying basic blocks
5891 instead of jumping. The expansion is relative to a jump instruction.
5892 The default value is 8.
5894 @item max-goto-duplication-insns
5895 The maximum number of instructions to duplicate to a block that jumps
5896 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5897 passes, GCC factors computed gotos early in the compilation process,
5898 and unfactors them as late as possible. Only computed jumps at the
5899 end of a basic blocks with no more than max-goto-duplication-insns are
5900 unfactored. The default value is 8.
5902 @item max-delay-slot-insn-search
5903 The maximum number of instructions to consider when looking for an
5904 instruction to fill a delay slot. If more than this arbitrary number of
5905 instructions is searched, the time savings from filling the delay slot
5906 will be minimal so stop searching. Increasing values mean more
5907 aggressive optimization, making the compile time increase with probably
5908 small improvement in executable run time.
5910 @item max-delay-slot-live-search
5911 When trying to fill delay slots, the maximum number of instructions to
5912 consider when searching for a block with valid live register
5913 information. Increasing this arbitrarily chosen value means more
5914 aggressive optimization, increasing the compile time. This parameter
5915 should be removed when the delay slot code is rewritten to maintain the
5918 @item max-gcse-memory
5919 The approximate maximum amount of memory that will be allocated in
5920 order to perform the global common subexpression elimination
5921 optimization. If more memory than specified is required, the
5922 optimization will not be done.
5924 @item max-gcse-passes
5925 The maximum number of passes of GCSE to run. The default is 1.
5927 @item max-pending-list-length
5928 The maximum number of pending dependencies scheduling will allow
5929 before flushing the current state and starting over. Large functions
5930 with few branches or calls can create excessively large lists which
5931 needlessly consume memory and resources.
5933 @item max-inline-insns-single
5934 Several parameters control the tree inliner used in gcc.
5935 This number sets the maximum number of instructions (counted in GCC's
5936 internal representation) in a single function that the tree inliner
5937 will consider for inlining. This only affects functions declared
5938 inline and methods implemented in a class declaration (C++).
5939 The default value is 450.
5941 @item max-inline-insns-auto
5942 When you use @option{-finline-functions} (included in @option{-O3}),
5943 a lot of functions that would otherwise not be considered for inlining
5944 by the compiler will be investigated. To those functions, a different
5945 (more restrictive) limit compared to functions declared inline can
5947 The default value is 90.
5949 @item large-function-insns
5950 The limit specifying really large functions. For functions larger than this
5951 limit after inlining inlining is constrained by
5952 @option{--param large-function-growth}. This parameter is useful primarily
5953 to avoid extreme compilation time caused by non-linear algorithms used by the
5955 This parameter is ignored when @option{-funit-at-a-time} is not used.
5956 The default value is 2700.
5958 @item large-function-growth
5959 Specifies maximal growth of large function caused by inlining in percents.
5960 This parameter is ignored when @option{-funit-at-a-time} is not used.
5961 The default value is 100 which limits large function growth to 2.0 times
5964 @item large-unit-insns
5965 The limit specifying large translation unit. Growth caused by inlining of
5966 units larger than this limit is limited by @option{--param inline-unit-growth}.
5967 For small units this might be too tight (consider unit consisting of function A
5968 that is inline and B that just calls A three time. If B is small relative to
5969 A, the growth of unit is 300\% and yet such inlining is very sane. For very
5970 large units consisting of small inlininable functions however the overall unit
5971 growth limit is needed to avoid exponential explosion of code size. Thus for
5972 smaller units, the size is increased to @option{--param large-unit-insns}
5973 before applying @option{--param inline-unit-growth}. The default is 10000
5975 @item inline-unit-growth
5976 Specifies maximal overall growth of the compilation unit caused by inlining.
5977 This parameter is ignored when @option{-funit-at-a-time} is not used.
5978 The default value is 50 which limits unit growth to 1.5 times the original
5981 @item max-inline-insns-recursive
5982 @itemx max-inline-insns-recursive-auto
5983 Specifies maximum number of instructions out-of-line copy of self recursive inline
5984 function can grow into by performing recursive inlining.
5986 For functions declared inline @option{--param max-inline-insns-recursive} is
5987 taken into account. For function not declared inline, recursive inlining
5988 happens only when @option{-finline-functions} (included in @option{-O3}) is
5989 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5990 default value is 450.
5992 @item max-inline-recursive-depth
5993 @itemx max-inline-recursive-depth-auto
5994 Specifies maximum recursion depth used by the recursive inlining.
5996 For functions declared inline @option{--param max-inline-recursive-depth} is
5997 taken into account. For function not declared inline, recursive inlining
5998 happens only when @option{-finline-functions} (included in @option{-O3}) is
5999 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
6000 default value is 450.
6002 @item min-inline-recursive-probability
6003 Recursive inlining is profitable only for function having deep recursion
6004 in average and can hurt for function having little recursion depth by
6005 increasing the prologue size or complexity of function body to other
6008 When profile feedback is available (see @option{-fprofile-generate}) the actual
6009 recursion depth can be guessed from probability that function will recurse via
6010 given call expression. This parameter limits inlining only to call expression
6011 whose probability exceeds given threshold (in percents). The default value is
6014 @item inline-call-cost
6015 Specify cost of call instruction relative to simple arithmetics operations
6016 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
6017 functions and at the same time increases size of leaf function that is believed to
6018 reduce function size by being inlined. In effect it increases amount of
6019 inlining for code having large abstraction penalty (many functions that just
6020 pass the arguments to other functions) and decrease inlining for code with low
6021 abstraction penalty. The default value is 16.
6023 @item max-unrolled-insns
6024 The maximum number of instructions that a loop should have if that loop
6025 is unrolled, and if the loop is unrolled, it determines how many times
6026 the loop code is unrolled.
6028 @item max-average-unrolled-insns
6029 The maximum number of instructions biased by probabilities of their execution
6030 that a loop should have if that loop is unrolled, and if the loop is unrolled,
6031 it determines how many times the loop code is unrolled.
6033 @item max-unroll-times
6034 The maximum number of unrollings of a single loop.
6036 @item max-peeled-insns
6037 The maximum number of instructions that a loop should have if that loop
6038 is peeled, and if the loop is peeled, it determines how many times
6039 the loop code is peeled.
6041 @item max-peel-times
6042 The maximum number of peelings of a single loop.
6044 @item max-completely-peeled-insns
6045 The maximum number of insns of a completely peeled loop.
6047 @item max-completely-peel-times
6048 The maximum number of iterations of a loop to be suitable for complete peeling.
6050 @item max-unswitch-insns
6051 The maximum number of insns of an unswitched loop.
6053 @item max-unswitch-level
6054 The maximum number of branches unswitched in a single loop.
6057 The minimum cost of an expensive expression in the loop invariant motion.
6059 @item iv-consider-all-candidates-bound
6060 Bound on number of candidates for induction variables below that
6061 all candidates are considered for each use in induction variable
6062 optimizations. Only the most relevant candidates are considered
6063 if there are more candidates, to avoid quadratic time complexity.
6065 @item iv-max-considered-uses
6066 The induction variable optimizations give up on loops that contain more
6067 induction variable uses.
6069 @item iv-always-prune-cand-set-bound
6070 If number of candidates in the set is smaller than this value,
6071 we always try to remove unnecessary ivs from the set during its
6072 optimization when a new iv is added to the set.
6074 @item scev-max-expr-size
6075 Bound on size of expressions used in the scalar evolutions analyzer.
6076 Large expressions slow the analyzer.
6078 @item vect-max-version-checks
6079 The maximum number of runtime checks that can be performed when doing
6080 loop versioning in the vectorizer. See option ftree-vect-loop-version
6081 for more information.
6083 @item max-iterations-to-track
6085 The maximum number of iterations of a loop the brute force algorithm
6086 for analysis of # of iterations of the loop tries to evaluate.
6088 @item hot-bb-count-fraction
6089 Select fraction of the maximal count of repetitions of basic block in program
6090 given basic block needs to have to be considered hot.
6092 @item hot-bb-frequency-fraction
6093 Select fraction of the maximal frequency of executions of basic block in
6094 function given basic block needs to have to be considered hot
6096 @item max-predicted-iterations
6097 The maximum number of loop iterations we predict statically. This is useful
6098 in cases where function contain single loop with known bound and other loop
6099 with unknown. We predict the known number of iterations correctly, while
6100 the unknown number of iterations average to roughly 10. This means that the
6101 loop without bounds would appear artificially cold relative to the other one.
6103 @item tracer-dynamic-coverage
6104 @itemx tracer-dynamic-coverage-feedback
6106 This value is used to limit superblock formation once the given percentage of
6107 executed instructions is covered. This limits unnecessary code size
6110 The @option{tracer-dynamic-coverage-feedback} is used only when profile
6111 feedback is available. The real profiles (as opposed to statically estimated
6112 ones) are much less balanced allowing the threshold to be larger value.
6114 @item tracer-max-code-growth
6115 Stop tail duplication once code growth has reached given percentage. This is
6116 rather hokey argument, as most of the duplicates will be eliminated later in
6117 cross jumping, so it may be set to much higher values than is the desired code
6120 @item tracer-min-branch-ratio
6122 Stop reverse growth when the reverse probability of best edge is less than this
6123 threshold (in percent).
6125 @item tracer-min-branch-ratio
6126 @itemx tracer-min-branch-ratio-feedback
6128 Stop forward growth if the best edge do have probability lower than this
6131 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
6132 compilation for profile feedback and one for compilation without. The value
6133 for compilation with profile feedback needs to be more conservative (higher) in
6134 order to make tracer effective.
6136 @item max-cse-path-length
6138 Maximum number of basic blocks on path that cse considers. The default is 10.
6141 The maximum instructions CSE process before flushing. The default is 1000.
6143 @item global-var-threshold
6145 Counts the number of function calls (@var{n}) and the number of
6146 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
6147 single artificial variable will be created to represent all the
6148 call-clobbered variables at function call sites. This artificial
6149 variable will then be made to alias every call-clobbered variable.
6150 (done as @code{int * size_t} on the host machine; beware overflow).
6152 @item max-aliased-vops
6154 Maximum number of virtual operands allowed to represent aliases
6155 before triggering the alias grouping heuristic. Alias grouping
6156 reduces compile times and memory consumption needed for aliasing at
6157 the expense of precision loss in alias information.
6159 @item ggc-min-expand
6161 GCC uses a garbage collector to manage its own memory allocation. This
6162 parameter specifies the minimum percentage by which the garbage
6163 collector's heap should be allowed to expand between collections.
6164 Tuning this may improve compilation speed; it has no effect on code
6167 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6168 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6169 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6170 GCC is not able to calculate RAM on a particular platform, the lower
6171 bound of 30% is used. Setting this parameter and
6172 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6173 every opportunity. This is extremely slow, but can be useful for
6176 @item ggc-min-heapsize
6178 Minimum size of the garbage collector's heap before it begins bothering
6179 to collect garbage. The first collection occurs after the heap expands
6180 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6181 tuning this may improve compilation speed, and has no effect on code
6184 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6185 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6186 with a lower bound of 4096 (four megabytes) and an upper bound of
6187 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6188 particular platform, the lower bound is used. Setting this parameter
6189 very large effectively disables garbage collection. Setting this
6190 parameter and @option{ggc-min-expand} to zero causes a full collection
6191 to occur at every opportunity.
6193 @item max-reload-search-insns
6194 The maximum number of instruction reload should look backward for equivalent
6195 register. Increasing values mean more aggressive optimization, making the
6196 compile time increase with probably slightly better performance. The default
6199 @item max-cselib-memory-locations
6200 The maximum number of memory locations cselib should take into account.
6201 Increasing values mean more aggressive optimization, making the compile time
6202 increase with probably slightly better performance. The default value is 500.
6204 @item max-flow-memory-locations
6205 Similar as @option{max-cselib-memory-locations} but for dataflow liveness.
6206 The default value is 100.
6208 @item reorder-blocks-duplicate
6209 @itemx reorder-blocks-duplicate-feedback
6211 Used by basic block reordering pass to decide whether to use unconditional
6212 branch or duplicate the code on its destination. Code is duplicated when its
6213 estimated size is smaller than this value multiplied by the estimated size of
6214 unconditional jump in the hot spots of the program.
6216 The @option{reorder-block-duplicate-feedback} is used only when profile
6217 feedback is available and may be set to higher values than
6218 @option{reorder-block-duplicate} since information about the hot spots is more
6221 @item max-sched-ready-insns
6222 The maximum number of instructions ready to be issued the scheduler should
6223 consider at any given time during the first scheduling pass. Increasing
6224 values mean more thorough searches, making the compilation time increase
6225 with probably little benefit. The default value is 100.
6227 @item max-sched-region-blocks
6228 The maximum number of blocks in a region to be considered for
6229 interblock scheduling. The default value is 10.
6231 @item max-sched-region-insns
6232 The maximum number of insns in a region to be considered for
6233 interblock scheduling. The default value is 100.
6236 The minimum probability (in percents) of reaching a source block
6237 for interblock speculative scheduling. The default value is 40.
6239 @item max-sched-extend-regions-iters
6240 The maximum number of iterations through CFG to extend regions.
6241 0 - disable region extension,
6242 N - do at most N iterations.
6243 The default value is 0.
6245 @item max-sched-insn-conflict-delay
6246 The maximum conflict delay for an insn to be considered for speculative motion.
6247 The default value is 3.
6249 @item sched-spec-prob-cutoff
6250 The minimal probability of speculation success (in percents), so that
6251 speculative insn will be scheduled.
6252 The default value is 40.
6254 @item max-last-value-rtl
6256 The maximum size measured as number of RTLs that can be recorded in an expression
6257 in combiner for a pseudo register as last known value of that register. The default
6260 @item integer-share-limit
6261 Small integer constants can use a shared data structure, reducing the
6262 compiler's memory usage and increasing its speed. This sets the maximum
6263 value of a shared integer constant's. The default value is 256.
6265 @item min-virtual-mappings
6266 Specifies the minimum number of virtual mappings in the incremental
6267 SSA updater that should be registered to trigger the virtual mappings
6268 heuristic defined by virtual-mappings-ratio. The default value is
6271 @item virtual-mappings-ratio
6272 If the number of virtual mappings is virtual-mappings-ratio bigger
6273 than the number of virtual symbols to be updated, then the incremental
6274 SSA updater switches to a full update for those symbols. The default
6277 @item ssp-buffer-size
6278 The minimum size of buffers (i.e. arrays) that will receive stack smashing
6279 protection when @option{-fstack-protection} is used.
6281 @item max-jump-thread-duplication-stmts
6282 Maximum number of statements allowed in a block that needs to be
6283 duplicated when threading jumps.
6285 @item max-fields-for-field-sensitive
6286 Maximum number of fields in a structure we will treat in
6287 a field sensitive manner during pointer analysis.
6292 @node Preprocessor Options
6293 @section Options Controlling the Preprocessor
6294 @cindex preprocessor options
6295 @cindex options, preprocessor
6297 These options control the C preprocessor, which is run on each C source
6298 file before actual compilation.
6300 If you use the @option{-E} option, nothing is done except preprocessing.
6301 Some of these options make sense only together with @option{-E} because
6302 they cause the preprocessor output to be unsuitable for actual
6307 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6308 and pass @var{option} directly through to the preprocessor. If
6309 @var{option} contains commas, it is split into multiple options at the
6310 commas. However, many options are modified, translated or interpreted
6311 by the compiler driver before being passed to the preprocessor, and
6312 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6313 interface is undocumented and subject to change, so whenever possible
6314 you should avoid using @option{-Wp} and let the driver handle the
6317 @item -Xpreprocessor @var{option}
6318 @opindex preprocessor
6319 Pass @var{option} as an option to the preprocessor. You can use this to
6320 supply system-specific preprocessor options which GCC does not know how to
6323 If you want to pass an option that takes an argument, you must use
6324 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6327 @include cppopts.texi
6329 @node Assembler Options
6330 @section Passing Options to the Assembler
6332 @c prevent bad page break with this line
6333 You can pass options to the assembler.
6336 @item -Wa,@var{option}
6338 Pass @var{option} as an option to the assembler. If @var{option}
6339 contains commas, it is split into multiple options at the commas.
6341 @item -Xassembler @var{option}
6343 Pass @var{option} as an option to the assembler. You can use this to
6344 supply system-specific assembler options which GCC does not know how to
6347 If you want to pass an option that takes an argument, you must use
6348 @option{-Xassembler} twice, once for the option and once for the argument.
6353 @section Options for Linking
6354 @cindex link options
6355 @cindex options, linking
6357 These options come into play when the compiler links object files into
6358 an executable output file. They are meaningless if the compiler is
6359 not doing a link step.
6363 @item @var{object-file-name}
6364 A file name that does not end in a special recognized suffix is
6365 considered to name an object file or library. (Object files are
6366 distinguished from libraries by the linker according to the file
6367 contents.) If linking is done, these object files are used as input
6376 If any of these options is used, then the linker is not run, and
6377 object file names should not be used as arguments. @xref{Overall
6381 @item -l@var{library}
6382 @itemx -l @var{library}
6384 Search the library named @var{library} when linking. (The second
6385 alternative with the library as a separate argument is only for
6386 POSIX compliance and is not recommended.)
6388 It makes a difference where in the command you write this option; the
6389 linker searches and processes libraries and object files in the order they
6390 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6391 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6392 to functions in @samp{z}, those functions may not be loaded.
6394 The linker searches a standard list of directories for the library,
6395 which is actually a file named @file{lib@var{library}.a}. The linker
6396 then uses this file as if it had been specified precisely by name.
6398 The directories searched include several standard system directories
6399 plus any that you specify with @option{-L}.
6401 Normally the files found this way are library files---archive files
6402 whose members are object files. The linker handles an archive file by
6403 scanning through it for members which define symbols that have so far
6404 been referenced but not defined. But if the file that is found is an
6405 ordinary object file, it is linked in the usual fashion. The only
6406 difference between using an @option{-l} option and specifying a file name
6407 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6408 and searches several directories.
6411 @opindex nostartfiles
6412 Do not use the standard system startup files when linking.
6413 The standard system libraries are used normally, unless @option{-nostdlib}
6414 or @option{-nodefaultlibs} is used.
6416 @item -nodefaultlibs
6417 @opindex nodefaultlibs
6418 Do not use the standard system libraries when linking.
6419 Only the libraries you specify will be passed to the linker.
6420 The standard startup files are used normally, unless @option{-nostartfiles}
6421 is used. The compiler may generate calls to @code{memcmp},
6422 @code{memset}, @code{memcpy} and @code{memmove}.
6423 These entries are usually resolved by entries in
6424 libc. These entry points should be supplied through some other
6425 mechanism when this option is specified.
6429 Do not use the standard system startup files or libraries when linking.
6430 No startup files and only the libraries you specify will be passed to
6431 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6432 @code{memcpy} and @code{memmove}.
6433 These entries are usually resolved by entries in
6434 libc. These entry points should be supplied through some other
6435 mechanism when this option is specified.
6437 @cindex @option{-lgcc}, use with @option{-nostdlib}
6438 @cindex @option{-nostdlib} and unresolved references
6439 @cindex unresolved references and @option{-nostdlib}
6440 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6441 @cindex @option{-nodefaultlibs} and unresolved references
6442 @cindex unresolved references and @option{-nodefaultlibs}
6443 One of the standard libraries bypassed by @option{-nostdlib} and
6444 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6445 that GCC uses to overcome shortcomings of particular machines, or special
6446 needs for some languages.
6447 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6448 Collection (GCC) Internals},
6449 for more discussion of @file{libgcc.a}.)
6450 In most cases, you need @file{libgcc.a} even when you want to avoid
6451 other standard libraries. In other words, when you specify @option{-nostdlib}
6452 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6453 This ensures that you have no unresolved references to internal GCC
6454 library subroutines. (For example, @samp{__main}, used to ensure C++
6455 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6456 GNU Compiler Collection (GCC) Internals}.)
6460 Produce a position independent executable on targets which support it.
6461 For predictable results, you must also specify the same set of options
6462 that were used to generate code (@option{-fpie}, @option{-fPIE},
6463 or model suboptions) when you specify this option.
6467 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
6468 that support it. This instructs the linker to add all symbols, not
6469 only used ones, to the dynamic symbol table. This option is needed
6470 for some uses of @code{dlopen} or to allow obtaining backtraces
6471 from within a program.
6475 Remove all symbol table and relocation information from the executable.
6479 On systems that support dynamic linking, this prevents linking with the shared
6480 libraries. On other systems, this option has no effect.
6484 Produce a shared object which can then be linked with other objects to
6485 form an executable. Not all systems support this option. For predictable
6486 results, you must also specify the same set of options that were used to
6487 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6488 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6489 needs to build supplementary stub code for constructors to work. On
6490 multi-libbed systems, @samp{gcc -shared} must select the correct support
6491 libraries to link against. Failing to supply the correct flags may lead
6492 to subtle defects. Supplying them in cases where they are not necessary
6495 @item -shared-libgcc
6496 @itemx -static-libgcc
6497 @opindex shared-libgcc
6498 @opindex static-libgcc
6499 On systems that provide @file{libgcc} as a shared library, these options
6500 force the use of either the shared or static version respectively.
6501 If no shared version of @file{libgcc} was built when the compiler was
6502 configured, these options have no effect.
6504 There are several situations in which an application should use the
6505 shared @file{libgcc} instead of the static version. The most common
6506 of these is when the application wishes to throw and catch exceptions
6507 across different shared libraries. In that case, each of the libraries
6508 as well as the application itself should use the shared @file{libgcc}.
6510 Therefore, the G++ and GCJ drivers automatically add
6511 @option{-shared-libgcc} whenever you build a shared library or a main
6512 executable, because C++ and Java programs typically use exceptions, so
6513 this is the right thing to do.
6515 If, instead, you use the GCC driver to create shared libraries, you may
6516 find that they will not always be linked with the shared @file{libgcc}.
6517 If GCC finds, at its configuration time, that you have a non-GNU linker
6518 or a GNU linker that does not support option @option{--eh-frame-hdr},
6519 it will link the shared version of @file{libgcc} into shared libraries
6520 by default. Otherwise, it will take advantage of the linker and optimize
6521 away the linking with the shared version of @file{libgcc}, linking with
6522 the static version of libgcc by default. This allows exceptions to
6523 propagate through such shared libraries, without incurring relocation
6524 costs at library load time.
6526 However, if a library or main executable is supposed to throw or catch
6527 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6528 for the languages used in the program, or using the option
6529 @option{-shared-libgcc}, such that it is linked with the shared
6534 Bind references to global symbols when building a shared object. Warn
6535 about any unresolved references (unless overridden by the link editor
6536 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6539 @item -Xlinker @var{option}
6541 Pass @var{option} as an option to the linker. You can use this to
6542 supply system-specific linker options which GCC does not know how to
6545 If you want to pass an option that takes an argument, you must use
6546 @option{-Xlinker} twice, once for the option and once for the argument.
6547 For example, to pass @option{-assert definitions}, you must write
6548 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6549 @option{-Xlinker "-assert definitions"}, because this passes the entire
6550 string as a single argument, which is not what the linker expects.
6552 @item -Wl,@var{option}
6554 Pass @var{option} as an option to the linker. If @var{option} contains
6555 commas, it is split into multiple options at the commas.
6557 @item -u @var{symbol}
6559 Pretend the symbol @var{symbol} is undefined, to force linking of
6560 library modules to define it. You can use @option{-u} multiple times with
6561 different symbols to force loading of additional library modules.
6564 @node Directory Options
6565 @section Options for Directory Search
6566 @cindex directory options
6567 @cindex options, directory search
6570 These options specify directories to search for header files, for
6571 libraries and for parts of the compiler:
6576 Add the directory @var{dir} to the head of the list of directories to be
6577 searched for header files. This can be used to override a system header
6578 file, substituting your own version, since these directories are
6579 searched before the system header file directories. However, you should
6580 not use this option to add directories that contain vendor-supplied
6581 system header files (use @option{-isystem} for that). If you use more than
6582 one @option{-I} option, the directories are scanned in left-to-right
6583 order; the standard system directories come after.
6585 If a standard system include directory, or a directory specified with
6586 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6587 option will be ignored. The directory will still be searched but as a
6588 system directory at its normal position in the system include chain.
6589 This is to ensure that GCC's procedure to fix buggy system headers and
6590 the ordering for the include_next directive are not inadvertently changed.
6591 If you really need to change the search order for system directories,
6592 use the @option{-nostdinc} and/or @option{-isystem} options.
6594 @item -iquote@var{dir}
6596 Add the directory @var{dir} to the head of the list of directories to
6597 be searched for header files only for the case of @samp{#include
6598 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6599 otherwise just like @option{-I}.
6603 Add directory @var{dir} to the list of directories to be searched
6606 @item -B@var{prefix}
6608 This option specifies where to find the executables, libraries,
6609 include files, and data files of the compiler itself.
6611 The compiler driver program runs one or more of the subprograms
6612 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6613 @var{prefix} as a prefix for each program it tries to run, both with and
6614 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6616 For each subprogram to be run, the compiler driver first tries the
6617 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6618 was not specified, the driver tries two standard prefixes, which are
6619 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6620 those results in a file name that is found, the unmodified program
6621 name is searched for using the directories specified in your
6622 @env{PATH} environment variable.
6624 The compiler will check to see if the path provided by the @option{-B}
6625 refers to a directory, and if necessary it will add a directory
6626 separator character at the end of the path.
6628 @option{-B} prefixes that effectively specify directory names also apply
6629 to libraries in the linker, because the compiler translates these
6630 options into @option{-L} options for the linker. They also apply to
6631 includes files in the preprocessor, because the compiler translates these
6632 options into @option{-isystem} options for the preprocessor. In this case,
6633 the compiler appends @samp{include} to the prefix.
6635 The run-time support file @file{libgcc.a} can also be searched for using
6636 the @option{-B} prefix, if needed. If it is not found there, the two
6637 standard prefixes above are tried, and that is all. The file is left
6638 out of the link if it is not found by those means.
6640 Another way to specify a prefix much like the @option{-B} prefix is to use
6641 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6644 As a special kludge, if the path provided by @option{-B} is
6645 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6646 9, then it will be replaced by @file{[dir/]include}. This is to help
6647 with boot-strapping the compiler.
6649 @item -specs=@var{file}
6651 Process @var{file} after the compiler reads in the standard @file{specs}
6652 file, in order to override the defaults that the @file{gcc} driver
6653 program uses when determining what switches to pass to @file{cc1},
6654 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6655 @option{-specs=@var{file}} can be specified on the command line, and they
6656 are processed in order, from left to right.
6658 @item --sysroot=@var{dir}
6660 Use @var{dir} as the logical root directory for headers and libraries.
6661 For example, if the compiler would normally search for headers in
6662 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
6663 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
6665 If you use both this option and the @option{-isysroot} option, then
6666 the @option{--sysroot} option will apply to libraries, but the
6667 @option{-isysroot} option will apply to header files.
6669 The GNU linker (beginning with version 2.16) has the necessary support
6670 for this option. If your linker does not support this option, the
6671 header file aspect of @option{--sysroot} will still work, but the
6672 library aspect will not.
6676 This option has been deprecated. Please use @option{-iquote} instead for
6677 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6678 Any directories you specify with @option{-I} options before the @option{-I-}
6679 option are searched only for the case of @samp{#include "@var{file}"};
6680 they are not searched for @samp{#include <@var{file}>}.
6682 If additional directories are specified with @option{-I} options after
6683 the @option{-I-}, these directories are searched for all @samp{#include}
6684 directives. (Ordinarily @emph{all} @option{-I} directories are used
6687 In addition, the @option{-I-} option inhibits the use of the current
6688 directory (where the current input file came from) as the first search
6689 directory for @samp{#include "@var{file}"}. There is no way to
6690 override this effect of @option{-I-}. With @option{-I.} you can specify
6691 searching the directory which was current when the compiler was
6692 invoked. That is not exactly the same as what the preprocessor does
6693 by default, but it is often satisfactory.
6695 @option{-I-} does not inhibit the use of the standard system directories
6696 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6703 @section Specifying subprocesses and the switches to pass to them
6706 @command{gcc} is a driver program. It performs its job by invoking a
6707 sequence of other programs to do the work of compiling, assembling and
6708 linking. GCC interprets its command-line parameters and uses these to
6709 deduce which programs it should invoke, and which command-line options
6710 it ought to place on their command lines. This behavior is controlled
6711 by @dfn{spec strings}. In most cases there is one spec string for each
6712 program that GCC can invoke, but a few programs have multiple spec
6713 strings to control their behavior. The spec strings built into GCC can
6714 be overridden by using the @option{-specs=} command-line switch to specify
6717 @dfn{Spec files} are plaintext files that are used to construct spec
6718 strings. They consist of a sequence of directives separated by blank
6719 lines. The type of directive is determined by the first non-whitespace
6720 character on the line and it can be one of the following:
6723 @item %@var{command}
6724 Issues a @var{command} to the spec file processor. The commands that can
6728 @item %include <@var{file}>
6730 Search for @var{file} and insert its text at the current point in the
6733 @item %include_noerr <@var{file}>
6734 @cindex %include_noerr
6735 Just like @samp{%include}, but do not generate an error message if the include
6736 file cannot be found.
6738 @item %rename @var{old_name} @var{new_name}
6740 Rename the spec string @var{old_name} to @var{new_name}.
6744 @item *[@var{spec_name}]:
6745 This tells the compiler to create, override or delete the named spec
6746 string. All lines after this directive up to the next directive or
6747 blank line are considered to be the text for the spec string. If this
6748 results in an empty string then the spec will be deleted. (Or, if the
6749 spec did not exist, then nothing will happened.) Otherwise, if the spec
6750 does not currently exist a new spec will be created. If the spec does
6751 exist then its contents will be overridden by the text of this
6752 directive, unless the first character of that text is the @samp{+}
6753 character, in which case the text will be appended to the spec.
6755 @item [@var{suffix}]:
6756 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6757 and up to the next directive or blank line are considered to make up the
6758 spec string for the indicated suffix. When the compiler encounters an
6759 input file with the named suffix, it will processes the spec string in
6760 order to work out how to compile that file. For example:
6767 This says that any input file whose name ends in @samp{.ZZ} should be
6768 passed to the program @samp{z-compile}, which should be invoked with the
6769 command-line switch @option{-input} and with the result of performing the
6770 @samp{%i} substitution. (See below.)
6772 As an alternative to providing a spec string, the text that follows a
6773 suffix directive can be one of the following:
6776 @item @@@var{language}
6777 This says that the suffix is an alias for a known @var{language}. This is
6778 similar to using the @option{-x} command-line switch to GCC to specify a
6779 language explicitly. For example:
6786 Says that .ZZ files are, in fact, C++ source files.
6789 This causes an error messages saying:
6792 @var{name} compiler not installed on this system.
6796 GCC already has an extensive list of suffixes built into it.
6797 This directive will add an entry to the end of the list of suffixes, but
6798 since the list is searched from the end backwards, it is effectively
6799 possible to override earlier entries using this technique.
6803 GCC has the following spec strings built into it. Spec files can
6804 override these strings or create their own. Note that individual
6805 targets can also add their own spec strings to this list.
6808 asm Options to pass to the assembler
6809 asm_final Options to pass to the assembler post-processor
6810 cpp Options to pass to the C preprocessor
6811 cc1 Options to pass to the C compiler
6812 cc1plus Options to pass to the C++ compiler
6813 endfile Object files to include at the end of the link
6814 link Options to pass to the linker
6815 lib Libraries to include on the command line to the linker
6816 libgcc Decides which GCC support library to pass to the linker
6817 linker Sets the name of the linker
6818 predefines Defines to be passed to the C preprocessor
6819 signed_char Defines to pass to CPP to say whether @code{char} is signed
6821 startfile Object files to include at the start of the link
6824 Here is a small example of a spec file:
6830 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6833 This example renames the spec called @samp{lib} to @samp{old_lib} and
6834 then overrides the previous definition of @samp{lib} with a new one.
6835 The new definition adds in some extra command-line options before
6836 including the text of the old definition.
6838 @dfn{Spec strings} are a list of command-line options to be passed to their
6839 corresponding program. In addition, the spec strings can contain
6840 @samp{%}-prefixed sequences to substitute variable text or to
6841 conditionally insert text into the command line. Using these constructs
6842 it is possible to generate quite complex command lines.
6844 Here is a table of all defined @samp{%}-sequences for spec
6845 strings. Note that spaces are not generated automatically around the
6846 results of expanding these sequences. Therefore you can concatenate them
6847 together or combine them with constant text in a single argument.
6851 Substitute one @samp{%} into the program name or argument.
6854 Substitute the name of the input file being processed.
6857 Substitute the basename of the input file being processed.
6858 This is the substring up to (and not including) the last period
6859 and not including the directory.
6862 This is the same as @samp{%b}, but include the file suffix (text after
6866 Marks the argument containing or following the @samp{%d} as a
6867 temporary file name, so that that file will be deleted if GCC exits
6868 successfully. Unlike @samp{%g}, this contributes no text to the
6871 @item %g@var{suffix}
6872 Substitute a file name that has suffix @var{suffix} and is chosen
6873 once per compilation, and mark the argument in the same way as
6874 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6875 name is now chosen in a way that is hard to predict even when previously
6876 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6877 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6878 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6879 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6880 was simply substituted with a file name chosen once per compilation,
6881 without regard to any appended suffix (which was therefore treated
6882 just like ordinary text), making such attacks more likely to succeed.
6884 @item %u@var{suffix}
6885 Like @samp{%g}, but generates a new temporary file name even if
6886 @samp{%u@var{suffix}} was already seen.
6888 @item %U@var{suffix}
6889 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6890 new one if there is no such last file name. In the absence of any
6891 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6892 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6893 would involve the generation of two distinct file names, one
6894 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6895 simply substituted with a file name chosen for the previous @samp{%u},
6896 without regard to any appended suffix.
6898 @item %j@var{suffix}
6899 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6900 writable, and if save-temps is off; otherwise, substitute the name
6901 of a temporary file, just like @samp{%u}. This temporary file is not
6902 meant for communication between processes, but rather as a junk
6905 @item %|@var{suffix}
6906 @itemx %m@var{suffix}
6907 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6908 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6909 all. These are the two most common ways to instruct a program that it
6910 should read from standard input or write to standard output. If you
6911 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6912 construct: see for example @file{f/lang-specs.h}.
6914 @item %.@var{SUFFIX}
6915 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6916 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6917 terminated by the next space or %.
6920 Marks the argument containing or following the @samp{%w} as the
6921 designated output file of this compilation. This puts the argument
6922 into the sequence of arguments that @samp{%o} will substitute later.
6925 Substitutes the names of all the output files, with spaces
6926 automatically placed around them. You should write spaces
6927 around the @samp{%o} as well or the results are undefined.
6928 @samp{%o} is for use in the specs for running the linker.
6929 Input files whose names have no recognized suffix are not compiled
6930 at all, but they are included among the output files, so they will
6934 Substitutes the suffix for object files. Note that this is
6935 handled specially when it immediately follows @samp{%g, %u, or %U},
6936 because of the need for those to form complete file names. The
6937 handling is such that @samp{%O} is treated exactly as if it had already
6938 been substituted, except that @samp{%g, %u, and %U} do not currently
6939 support additional @var{suffix} characters following @samp{%O} as they would
6940 following, for example, @samp{.o}.
6943 Substitutes the standard macro predefinitions for the
6944 current target machine. Use this when running @code{cpp}.
6947 Like @samp{%p}, but puts @samp{__} before and after the name of each
6948 predefined macro, except for macros that start with @samp{__} or with
6949 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6953 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6954 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
6955 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6956 and @option{-imultilib} as necessary.
6959 Current argument is the name of a library or startup file of some sort.
6960 Search for that file in a standard list of directories and substitute
6961 the full name found.
6964 Print @var{str} as an error message. @var{str} is terminated by a newline.
6965 Use this when inconsistent options are detected.
6968 Substitute the contents of spec string @var{name} at this point.
6971 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6973 @item %x@{@var{option}@}
6974 Accumulate an option for @samp{%X}.
6977 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6981 Output the accumulated assembler options specified by @option{-Wa}.
6984 Output the accumulated preprocessor options specified by @option{-Wp}.
6987 Process the @code{asm} spec. This is used to compute the
6988 switches to be passed to the assembler.
6991 Process the @code{asm_final} spec. This is a spec string for
6992 passing switches to an assembler post-processor, if such a program is
6996 Process the @code{link} spec. This is the spec for computing the
6997 command line passed to the linker. Typically it will make use of the
6998 @samp{%L %G %S %D and %E} sequences.
7001 Dump out a @option{-L} option for each directory that GCC believes might
7002 contain startup files. If the target supports multilibs then the
7003 current multilib directory will be prepended to each of these paths.
7006 Process the @code{lib} spec. This is a spec string for deciding which
7007 libraries should be included on the command line to the linker.
7010 Process the @code{libgcc} spec. This is a spec string for deciding
7011 which GCC support library should be included on the command line to the linker.
7014 Process the @code{startfile} spec. This is a spec for deciding which
7015 object files should be the first ones passed to the linker. Typically
7016 this might be a file named @file{crt0.o}.
7019 Process the @code{endfile} spec. This is a spec string that specifies
7020 the last object files that will be passed to the linker.
7023 Process the @code{cpp} spec. This is used to construct the arguments
7024 to be passed to the C preprocessor.
7027 Process the @code{cc1} spec. This is used to construct the options to be
7028 passed to the actual C compiler (@samp{cc1}).
7031 Process the @code{cc1plus} spec. This is used to construct the options to be
7032 passed to the actual C++ compiler (@samp{cc1plus}).
7035 Substitute the variable part of a matched option. See below.
7036 Note that each comma in the substituted string is replaced by
7040 Remove all occurrences of @code{-S} from the command line. Note---this
7041 command is position dependent. @samp{%} commands in the spec string
7042 before this one will see @code{-S}, @samp{%} commands in the spec string
7043 after this one will not.
7045 @item %:@var{function}(@var{args})
7046 Call the named function @var{function}, passing it @var{args}.
7047 @var{args} is first processed as a nested spec string, then split
7048 into an argument vector in the usual fashion. The function returns
7049 a string which is processed as if it had appeared literally as part
7050 of the current spec.
7052 The following built-in spec functions are provided:
7055 @item @code{if-exists}
7056 The @code{if-exists} spec function takes one argument, an absolute
7057 pathname to a file. If the file exists, @code{if-exists} returns the
7058 pathname. Here is a small example of its usage:
7062 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
7065 @item @code{if-exists-else}
7066 The @code{if-exists-else} spec function is similar to the @code{if-exists}
7067 spec function, except that it takes two arguments. The first argument is
7068 an absolute pathname to a file. If the file exists, @code{if-exists-else}
7069 returns the pathname. If it does not exist, it returns the second argument.
7070 This way, @code{if-exists-else} can be used to select one file or another,
7071 based on the existence of the first. Here is a small example of its usage:
7075 crt0%O%s %:if-exists(crti%O%s) \
7076 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
7079 @item @code{replace-outfile}
7080 The @code{replace-outfile} spec function takes two arguments. It looks for the
7081 first argument in the outfiles array and replaces it with the second argument. Here
7082 is a small example of its usage:
7085 %@{static|static-libgcc|static-libstdc++:%:replace-outfile(-lstdc++ \
7093 Substitutes the @code{-S} switch, if that switch was given to GCC@.
7094 If that switch was not specified, this substitutes nothing. Note that
7095 the leading dash is omitted when specifying this option, and it is
7096 automatically inserted if the substitution is performed. Thus the spec
7097 string @samp{%@{foo@}} would match the command-line option @option{-foo}
7098 and would output the command line option @option{-foo}.
7100 @item %W@{@code{S}@}
7101 Like %@{@code{S}@} but mark last argument supplied within as a file to be
7104 @item %@{@code{S}*@}
7105 Substitutes all the switches specified to GCC whose names start
7106 with @code{-S}, but which also take an argument. This is used for
7107 switches like @option{-o}, @option{-D}, @option{-I}, etc.
7108 GCC considers @option{-o foo} as being
7109 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
7110 text, including the space. Thus two arguments would be generated.
7112 @item %@{@code{S}*&@code{T}*@}
7113 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
7114 (the order of @code{S} and @code{T} in the spec is not significant).
7115 There can be any number of ampersand-separated variables; for each the
7116 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
7118 @item %@{@code{S}:@code{X}@}
7119 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
7121 @item %@{!@code{S}:@code{X}@}
7122 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
7124 @item %@{@code{S}*:@code{X}@}
7125 Substitutes @code{X} if one or more switches whose names start with
7126 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
7127 once, no matter how many such switches appeared. However, if @code{%*}
7128 appears somewhere in @code{X}, then @code{X} will be substituted once
7129 for each matching switch, with the @code{%*} replaced by the part of
7130 that switch that matched the @code{*}.
7132 @item %@{.@code{S}:@code{X}@}
7133 Substitutes @code{X}, if processing a file with suffix @code{S}.
7135 @item %@{!.@code{S}:@code{X}@}
7136 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
7138 @item %@{@code{S}|@code{P}:@code{X}@}
7139 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
7140 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
7141 although they have a stronger binding than the @samp{|}. If @code{%*}
7142 appears in @code{X}, all of the alternatives must be starred, and only
7143 the first matching alternative is substituted.
7145 For example, a spec string like this:
7148 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
7151 will output the following command-line options from the following input
7152 command-line options:
7157 -d fred.c -foo -baz -boggle
7158 -d jim.d -bar -baz -boggle
7161 @item %@{S:X; T:Y; :D@}
7163 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
7164 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
7165 be as many clauses as you need. This may be combined with @code{.},
7166 @code{!}, @code{|}, and @code{*} as needed.
7171 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
7172 construct may contain other nested @samp{%} constructs or spaces, or
7173 even newlines. They are processed as usual, as described above.
7174 Trailing white space in @code{X} is ignored. White space may also
7175 appear anywhere on the left side of the colon in these constructs,
7176 except between @code{.} or @code{*} and the corresponding word.
7178 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
7179 handled specifically in these constructs. If another value of
7180 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
7181 @option{-W} switch is found later in the command line, the earlier
7182 switch value is ignored, except with @{@code{S}*@} where @code{S} is
7183 just one letter, which passes all matching options.
7185 The character @samp{|} at the beginning of the predicate text is used to
7186 indicate that a command should be piped to the following command, but
7187 only if @option{-pipe} is specified.
7189 It is built into GCC which switches take arguments and which do not.
7190 (You might think it would be useful to generalize this to allow each
7191 compiler's spec to say which switches take arguments. But this cannot
7192 be done in a consistent fashion. GCC cannot even decide which input
7193 files have been specified without knowing which switches take arguments,
7194 and it must know which input files to compile in order to tell which
7197 GCC also knows implicitly that arguments starting in @option{-l} are to be
7198 treated as compiler output files, and passed to the linker in their
7199 proper position among the other output files.
7201 @c man begin OPTIONS
7203 @node Target Options
7204 @section Specifying Target Machine and Compiler Version
7205 @cindex target options
7206 @cindex cross compiling
7207 @cindex specifying machine version
7208 @cindex specifying compiler version and target machine
7209 @cindex compiler version, specifying
7210 @cindex target machine, specifying
7212 The usual way to run GCC is to run the executable called @file{gcc}, or
7213 @file{<machine>-gcc} when cross-compiling, or
7214 @file{<machine>-gcc-<version>} to run a version other than the one that
7215 was installed last. Sometimes this is inconvenient, so GCC provides
7216 options that will switch to another cross-compiler or version.
7219 @item -b @var{machine}
7221 The argument @var{machine} specifies the target machine for compilation.
7223 The value to use for @var{machine} is the same as was specified as the
7224 machine type when configuring GCC as a cross-compiler. For
7225 example, if a cross-compiler was configured with @samp{configure
7226 arm-elf}, meaning to compile for an arm processor with elf binaries,
7227 then you would specify @option{-b arm-elf} to run that cross compiler.
7228 Because there are other options beginning with @option{-b}, the
7229 configuration must contain a hyphen.
7231 @item -V @var{version}
7233 The argument @var{version} specifies which version of GCC to run.
7234 This is useful when multiple versions are installed. For example,
7235 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
7238 The @option{-V} and @option{-b} options work by running the
7239 @file{<machine>-gcc-<version>} executable, so there's no real reason to
7240 use them if you can just run that directly.
7242 @node Submodel Options
7243 @section Hardware Models and Configurations
7244 @cindex submodel options
7245 @cindex specifying hardware config
7246 @cindex hardware models and configurations, specifying
7247 @cindex machine dependent options
7249 Earlier we discussed the standard option @option{-b} which chooses among
7250 different installed compilers for completely different target
7251 machines, such as VAX vs.@: 68000 vs.@: 80386.
7253 In addition, each of these target machine types can have its own
7254 special options, starting with @samp{-m}, to choose among various
7255 hardware models or configurations---for example, 68010 vs 68020,
7256 floating coprocessor or none. A single installed version of the
7257 compiler can compile for any model or configuration, according to the
7260 Some configurations of the compiler also support additional special
7261 options, usually for compatibility with other compilers on the same
7264 @c This list is ordered alphanumerically by subsection name.
7265 @c It should be the same order and spelling as these options are listed
7266 @c in Machine Dependent Options
7272 * Blackfin Options::
7276 * DEC Alpha Options::
7277 * DEC Alpha/VMS Options::
7279 * GNU/Linux Options::
7282 * i386 and x86-64 Options::
7295 * RS/6000 and PowerPC Options::
7296 * S/390 and zSeries Options::
7300 * System V Options::
7301 * TMS320C3x/C4x Options::
7305 * Xstormy16 Options::
7311 @subsection ARC Options
7314 These options are defined for ARC implementations:
7319 Compile code for little endian mode. This is the default.
7323 Compile code for big endian mode.
7326 @opindex mmangle-cpu
7327 Prepend the name of the cpu to all public symbol names.
7328 In multiple-processor systems, there are many ARC variants with different
7329 instruction and register set characteristics. This flag prevents code
7330 compiled for one cpu to be linked with code compiled for another.
7331 No facility exists for handling variants that are ``almost identical''.
7332 This is an all or nothing option.
7334 @item -mcpu=@var{cpu}
7336 Compile code for ARC variant @var{cpu}.
7337 Which variants are supported depend on the configuration.
7338 All variants support @option{-mcpu=base}, this is the default.
7340 @item -mtext=@var{text-section}
7341 @itemx -mdata=@var{data-section}
7342 @itemx -mrodata=@var{readonly-data-section}
7346 Put functions, data, and readonly data in @var{text-section},
7347 @var{data-section}, and @var{readonly-data-section} respectively
7348 by default. This can be overridden with the @code{section} attribute.
7349 @xref{Variable Attributes}.
7354 @subsection ARM Options
7357 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7361 @item -mabi=@var{name}
7363 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7364 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
7367 @opindex mapcs-frame
7368 Generate a stack frame that is compliant with the ARM Procedure Call
7369 Standard for all functions, even if this is not strictly necessary for
7370 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7371 with this option will cause the stack frames not to be generated for
7372 leaf functions. The default is @option{-mno-apcs-frame}.
7376 This is a synonym for @option{-mapcs-frame}.
7379 @c not currently implemented
7380 @item -mapcs-stack-check
7381 @opindex mapcs-stack-check
7382 Generate code to check the amount of stack space available upon entry to
7383 every function (that actually uses some stack space). If there is
7384 insufficient space available then either the function
7385 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7386 called, depending upon the amount of stack space required. The run time
7387 system is required to provide these functions. The default is
7388 @option{-mno-apcs-stack-check}, since this produces smaller code.
7390 @c not currently implemented
7392 @opindex mapcs-float
7393 Pass floating point arguments using the float point registers. This is
7394 one of the variants of the APCS@. This option is recommended if the
7395 target hardware has a floating point unit or if a lot of floating point
7396 arithmetic is going to be performed by the code. The default is
7397 @option{-mno-apcs-float}, since integer only code is slightly increased in
7398 size if @option{-mapcs-float} is used.
7400 @c not currently implemented
7401 @item -mapcs-reentrant
7402 @opindex mapcs-reentrant
7403 Generate reentrant, position independent code. The default is
7404 @option{-mno-apcs-reentrant}.
7407 @item -mthumb-interwork
7408 @opindex mthumb-interwork
7409 Generate code which supports calling between the ARM and Thumb
7410 instruction sets. Without this option the two instruction sets cannot
7411 be reliably used inside one program. The default is
7412 @option{-mno-thumb-interwork}, since slightly larger code is generated
7413 when @option{-mthumb-interwork} is specified.
7415 @item -mno-sched-prolog
7416 @opindex mno-sched-prolog
7417 Prevent the reordering of instructions in the function prolog, or the
7418 merging of those instruction with the instructions in the function's
7419 body. This means that all functions will start with a recognizable set
7420 of instructions (or in fact one of a choice from a small set of
7421 different function prologues), and this information can be used to
7422 locate the start if functions inside an executable piece of code. The
7423 default is @option{-msched-prolog}.
7426 @opindex mhard-float
7427 Generate output containing floating point instructions. This is the
7431 @opindex msoft-float
7432 Generate output containing library calls for floating point.
7433 @strong{Warning:} the requisite libraries are not available for all ARM
7434 targets. Normally the facilities of the machine's usual C compiler are
7435 used, but this cannot be done directly in cross-compilation. You must make
7436 your own arrangements to provide suitable library functions for
7439 @option{-msoft-float} changes the calling convention in the output file;
7440 therefore, it is only useful if you compile @emph{all} of a program with
7441 this option. In particular, you need to compile @file{libgcc.a}, the
7442 library that comes with GCC, with @option{-msoft-float} in order for
7445 @item -mfloat-abi=@var{name}
7447 Specifies which ABI to use for floating point values. Permissible values
7448 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7450 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7451 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7452 of floating point instructions, but still uses the soft-float calling
7455 @item -mlittle-endian
7456 @opindex mlittle-endian
7457 Generate code for a processor running in little-endian mode. This is
7458 the default for all standard configurations.
7461 @opindex mbig-endian
7462 Generate code for a processor running in big-endian mode; the default is
7463 to compile code for a little-endian processor.
7465 @item -mwords-little-endian
7466 @opindex mwords-little-endian
7467 This option only applies when generating code for big-endian processors.
7468 Generate code for a little-endian word order but a big-endian byte
7469 order. That is, a byte order of the form @samp{32107654}. Note: this
7470 option should only be used if you require compatibility with code for
7471 big-endian ARM processors generated by versions of the compiler prior to
7474 @item -mcpu=@var{name}
7476 This specifies the name of the target ARM processor. GCC uses this name
7477 to determine what kind of instructions it can emit when generating
7478 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7479 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7480 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7481 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7482 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7483 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7484 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7485 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7486 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7487 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7488 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7489 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7490 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7491 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
7494 @itemx -mtune=@var{name}
7496 This option is very similar to the @option{-mcpu=} option, except that
7497 instead of specifying the actual target processor type, and hence
7498 restricting which instructions can be used, it specifies that GCC should
7499 tune the performance of the code as if the target were of the type
7500 specified in this option, but still choosing the instructions that it
7501 will generate based on the cpu specified by a @option{-mcpu=} option.
7502 For some ARM implementations better performance can be obtained by using
7505 @item -march=@var{name}
7507 This specifies the name of the target ARM architecture. GCC uses this
7508 name to determine what kind of instructions it can emit when generating
7509 assembly code. This option can be used in conjunction with or instead
7510 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7511 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7512 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7513 @samp{iwmmxt}, @samp{ep9312}.
7515 @item -mfpu=@var{name}
7516 @itemx -mfpe=@var{number}
7517 @itemx -mfp=@var{number}
7521 This specifies what floating point hardware (or hardware emulation) is
7522 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7523 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7524 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7525 with older versions of GCC@.
7527 If @option{-msoft-float} is specified this specifies the format of
7528 floating point values.
7530 @item -mstructure-size-boundary=@var{n}
7531 @opindex mstructure-size-boundary
7532 The size of all structures and unions will be rounded up to a multiple
7533 of the number of bits set by this option. Permissible values are 8, 32
7534 and 64. The default value varies for different toolchains. For the COFF
7535 targeted toolchain the default value is 8. A value of 64 is only allowed
7536 if the underlying ABI supports it.
7538 Specifying the larger number can produce faster, more efficient code, but
7539 can also increase the size of the program. Different values are potentially
7540 incompatible. Code compiled with one value cannot necessarily expect to
7541 work with code or libraries compiled with another value, if they exchange
7542 information using structures or unions.
7544 @item -mabort-on-noreturn
7545 @opindex mabort-on-noreturn
7546 Generate a call to the function @code{abort} at the end of a
7547 @code{noreturn} function. It will be executed if the function tries to
7551 @itemx -mno-long-calls
7552 @opindex mlong-calls
7553 @opindex mno-long-calls
7554 Tells the compiler to perform function calls by first loading the
7555 address of the function into a register and then performing a subroutine
7556 call on this register. This switch is needed if the target function
7557 will lie outside of the 64 megabyte addressing range of the offset based
7558 version of subroutine call instruction.
7560 Even if this switch is enabled, not all function calls will be turned
7561 into long calls. The heuristic is that static functions, functions
7562 which have the @samp{short-call} attribute, functions that are inside
7563 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7564 definitions have already been compiled within the current compilation
7565 unit, will not be turned into long calls. The exception to this rule is
7566 that weak function definitions, functions with the @samp{long-call}
7567 attribute or the @samp{section} attribute, and functions that are within
7568 the scope of a @samp{#pragma long_calls} directive, will always be
7569 turned into long calls.
7571 This feature is not enabled by default. Specifying
7572 @option{-mno-long-calls} will restore the default behavior, as will
7573 placing the function calls within the scope of a @samp{#pragma
7574 long_calls_off} directive. Note these switches have no effect on how
7575 the compiler generates code to handle function calls via function
7578 @item -mnop-fun-dllimport
7579 @opindex mnop-fun-dllimport
7580 Disable support for the @code{dllimport} attribute.
7582 @item -msingle-pic-base
7583 @opindex msingle-pic-base
7584 Treat the register used for PIC addressing as read-only, rather than
7585 loading it in the prologue for each function. The run-time system is
7586 responsible for initializing this register with an appropriate value
7587 before execution begins.
7589 @item -mpic-register=@var{reg}
7590 @opindex mpic-register
7591 Specify the register to be used for PIC addressing. The default is R10
7592 unless stack-checking is enabled, when R9 is used.
7594 @item -mcirrus-fix-invalid-insns
7595 @opindex mcirrus-fix-invalid-insns
7596 @opindex mno-cirrus-fix-invalid-insns
7597 Insert NOPs into the instruction stream to in order to work around
7598 problems with invalid Maverick instruction combinations. This option
7599 is only valid if the @option{-mcpu=ep9312} option has been used to
7600 enable generation of instructions for the Cirrus Maverick floating
7601 point co-processor. This option is not enabled by default, since the
7602 problem is only present in older Maverick implementations. The default
7603 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7606 @item -mpoke-function-name
7607 @opindex mpoke-function-name
7608 Write the name of each function into the text section, directly
7609 preceding the function prologue. The generated code is similar to this:
7613 .ascii "arm_poke_function_name", 0
7616 .word 0xff000000 + (t1 - t0)
7617 arm_poke_function_name
7619 stmfd sp!, @{fp, ip, lr, pc@}
7623 When performing a stack backtrace, code can inspect the value of
7624 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7625 location @code{pc - 12} and the top 8 bits are set, then we know that
7626 there is a function name embedded immediately preceding this location
7627 and has length @code{((pc[-3]) & 0xff000000)}.
7631 Generate code for the 16-bit Thumb instruction set. The default is to
7632 use the 32-bit ARM instruction set.
7635 @opindex mtpcs-frame
7636 Generate a stack frame that is compliant with the Thumb Procedure Call
7637 Standard for all non-leaf functions. (A leaf function is one that does
7638 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7640 @item -mtpcs-leaf-frame
7641 @opindex mtpcs-leaf-frame
7642 Generate a stack frame that is compliant with the Thumb Procedure Call
7643 Standard for all leaf functions. (A leaf function is one that does
7644 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7646 @item -mcallee-super-interworking
7647 @opindex mcallee-super-interworking
7648 Gives all externally visible functions in the file being compiled an ARM
7649 instruction set header which switches to Thumb mode before executing the
7650 rest of the function. This allows these functions to be called from
7651 non-interworking code.
7653 @item -mcaller-super-interworking
7654 @opindex mcaller-super-interworking
7655 Allows calls via function pointers (including virtual functions) to
7656 execute correctly regardless of whether the target code has been
7657 compiled for interworking or not. There is a small overhead in the cost
7658 of executing a function pointer if this option is enabled.
7660 @item -mtp=@var{name}
7662 Specify the access model for the thread local storage pointer. The valid
7663 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
7664 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
7665 (supported in the arm6k architecture), and @option{auto}, which uses the
7666 best available method for the selected processor. The default setting is
7672 @subsection AVR Options
7675 These options are defined for AVR implementations:
7678 @item -mmcu=@var{mcu}
7680 Specify ATMEL AVR instruction set or MCU type.
7682 Instruction set avr1 is for the minimal AVR core, not supported by the C
7683 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7684 attiny11, attiny12, attiny15, attiny28).
7686 Instruction set avr2 (default) is for the classic AVR core with up to
7687 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7688 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7689 at90c8534, at90s8535).
7691 Instruction set avr3 is for the classic AVR core with up to 128K program
7692 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7694 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7695 memory space (MCU types: atmega8, atmega83, atmega85).
7697 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7698 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7699 atmega64, atmega128, at43usb355, at94k).
7703 Output instruction sizes to the asm file.
7705 @item -minit-stack=@var{N}
7706 @opindex minit-stack
7707 Specify the initial stack address, which may be a symbol or numeric value,
7708 @samp{__stack} is the default.
7710 @item -mno-interrupts
7711 @opindex mno-interrupts
7712 Generated code is not compatible with hardware interrupts.
7713 Code size will be smaller.
7715 @item -mcall-prologues
7716 @opindex mcall-prologues
7717 Functions prologues/epilogues expanded as call to appropriate
7718 subroutines. Code size will be smaller.
7720 @item -mno-tablejump
7721 @opindex mno-tablejump
7722 Do not generate tablejump insns which sometimes increase code size.
7725 @opindex mtiny-stack
7726 Change only the low 8 bits of the stack pointer.
7730 Assume int to be 8 bit integer. This affects the sizes of all types: A
7731 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7732 and long long will be 4 bytes. Please note that this option does not
7733 comply to the C standards, but it will provide you with smaller code
7737 @node Blackfin Options
7738 @subsection Blackfin Options
7739 @cindex Blackfin Options
7742 @item -momit-leaf-frame-pointer
7743 @opindex momit-leaf-frame-pointer
7744 Don't keep the frame pointer in a register for leaf functions. This
7745 avoids the instructions to save, set up and restore frame pointers and
7746 makes an extra register available in leaf functions. The option
7747 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7748 which might make debugging harder.
7750 @item -mspecld-anomaly
7751 @opindex mspecld-anomaly
7752 When enabled, the compiler will ensure that the generated code does not
7753 contain speculative loads after jump instructions. This option is enabled
7756 @item -mno-specld-anomaly
7757 @opindex mno-specld-anomaly
7758 Don't generate extra code to prevent speculative loads from occurring.
7760 @item -mcsync-anomaly
7761 @opindex mcsync-anomaly
7762 When enabled, the compiler will ensure that the generated code does not
7763 contain CSYNC or SSYNC instructions too soon after conditional branches.
7764 This option is enabled by default.
7766 @item -mno-csync-anomaly
7767 @opindex mno-csync-anomaly
7768 Don't generate extra code to prevent CSYNC or SSYNC instructions from
7769 occurring too soon after a conditional branch.
7773 When enabled, the compiler is free to take advantage of the knowledge that
7774 the entire program fits into the low 64k of memory.
7777 @opindex mno-low-64k
7778 Assume that the program is arbitrarily large. This is the default.
7780 @item -mid-shared-library
7781 @opindex mid-shared-library
7782 Generate code that supports shared libraries via the library ID method.
7783 This allows for execute in place and shared libraries in an environment
7784 without virtual memory management. This option implies @option{-fPIC}.
7786 @item -mno-id-shared-library
7787 @opindex mno-id-shared-library
7788 Generate code that doesn't assume ID based shared libraries are being used.
7789 This is the default.
7791 @item -mshared-library-id=n
7792 @opindex mshared-library-id
7793 Specified the identification number of the ID based shared library being
7794 compiled. Specifying a value of 0 will generate more compact code, specifying
7795 other values will force the allocation of that number to the current
7796 library but is no more space or time efficient than omitting this option.
7799 @itemx -mno-long-calls
7800 @opindex mlong-calls
7801 @opindex mno-long-calls
7802 Tells the compiler to perform function calls by first loading the
7803 address of the function into a register and then performing a subroutine
7804 call on this register. This switch is needed if the target function
7805 will lie outside of the 24 bit addressing range of the offset based
7806 version of subroutine call instruction.
7808 This feature is not enabled by default. Specifying
7809 @option{-mno-long-calls} will restore the default behavior. Note these
7810 switches have no effect on how the compiler generates code to handle
7811 function calls via function pointers.
7815 @subsection CRIS Options
7816 @cindex CRIS Options
7818 These options are defined specifically for the CRIS ports.
7821 @item -march=@var{architecture-type}
7822 @itemx -mcpu=@var{architecture-type}
7825 Generate code for the specified architecture. The choices for
7826 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7827 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7828 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7831 @item -mtune=@var{architecture-type}
7833 Tune to @var{architecture-type} everything applicable about the generated
7834 code, except for the ABI and the set of available instructions. The
7835 choices for @var{architecture-type} are the same as for
7836 @option{-march=@var{architecture-type}}.
7838 @item -mmax-stack-frame=@var{n}
7839 @opindex mmax-stack-frame
7840 Warn when the stack frame of a function exceeds @var{n} bytes.
7842 @item -melinux-stacksize=@var{n}
7843 @opindex melinux-stacksize
7844 Only available with the @samp{cris-axis-aout} target. Arranges for
7845 indications in the program to the kernel loader that the stack of the
7846 program should be set to @var{n} bytes.
7852 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7853 @option{-march=v3} and @option{-march=v8} respectively.
7855 @item -mmul-bug-workaround
7856 @itemx -mno-mul-bug-workaround
7857 @opindex mmul-bug-workaround
7858 @opindex mno-mul-bug-workaround
7859 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7860 models where it applies. This option is active by default.
7864 Enable CRIS-specific verbose debug-related information in the assembly
7865 code. This option also has the effect to turn off the @samp{#NO_APP}
7866 formatted-code indicator to the assembler at the beginning of the
7871 Do not use condition-code results from previous instruction; always emit
7872 compare and test instructions before use of condition codes.
7874 @item -mno-side-effects
7875 @opindex mno-side-effects
7876 Do not emit instructions with side-effects in addressing modes other than
7880 @itemx -mno-stack-align
7882 @itemx -mno-data-align
7883 @itemx -mconst-align
7884 @itemx -mno-const-align
7885 @opindex mstack-align
7886 @opindex mno-stack-align
7887 @opindex mdata-align
7888 @opindex mno-data-align
7889 @opindex mconst-align
7890 @opindex mno-const-align
7891 These options (no-options) arranges (eliminate arrangements) for the
7892 stack-frame, individual data and constants to be aligned for the maximum
7893 single data access size for the chosen CPU model. The default is to
7894 arrange for 32-bit alignment. ABI details such as structure layout are
7895 not affected by these options.
7903 Similar to the stack- data- and const-align options above, these options
7904 arrange for stack-frame, writable data and constants to all be 32-bit,
7905 16-bit or 8-bit aligned. The default is 32-bit alignment.
7907 @item -mno-prologue-epilogue
7908 @itemx -mprologue-epilogue
7909 @opindex mno-prologue-epilogue
7910 @opindex mprologue-epilogue
7911 With @option{-mno-prologue-epilogue}, the normal function prologue and
7912 epilogue that sets up the stack-frame are omitted and no return
7913 instructions or return sequences are generated in the code. Use this
7914 option only together with visual inspection of the compiled code: no
7915 warnings or errors are generated when call-saved registers must be saved,
7916 or storage for local variable needs to be allocated.
7922 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7923 instruction sequences that load addresses for functions from the PLT part
7924 of the GOT rather than (traditional on other architectures) calls to the
7925 PLT@. The default is @option{-mgotplt}.
7929 Legacy no-op option only recognized with the cris-axis-aout target.
7933 Legacy no-op option only recognized with the cris-axis-elf and
7934 cris-axis-linux-gnu targets.
7938 Only recognized with the cris-axis-aout target, where it selects a
7939 GNU/linux-like multilib, include files and instruction set for
7944 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7948 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7949 to link with input-output functions from a simulator library. Code,
7950 initialized data and zero-initialized data are allocated consecutively.
7954 Like @option{-sim}, but pass linker options to locate initialized data at
7955 0x40000000 and zero-initialized data at 0x80000000.
7959 @subsection CRX Options
7962 These options are defined specifically for the CRX ports.
7968 Enable the use of multiply-accumulate instructions. Disabled by default.
7972 Push instructions will be used to pass outgoing arguments when functions
7973 are called. Enabled by default.
7976 @node Darwin Options
7977 @subsection Darwin Options
7978 @cindex Darwin options
7980 These options are defined for all architectures running the Darwin operating
7983 FSF GCC on Darwin does not create ``fat'' object files; it will create
7984 an object file for the single architecture that it was built to
7985 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7986 @option{-arch} options are used; it does so by running the compiler or
7987 linker multiple times and joining the results together with
7990 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7991 @samp{i686}) is determined by the flags that specify the ISA
7992 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7993 @option{-force_cpusubtype_ALL} option can be used to override this.
7995 The Darwin tools vary in their behavior when presented with an ISA
7996 mismatch. The assembler, @file{as}, will only permit instructions to
7997 be used that are valid for the subtype of the file it is generating,
7998 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7999 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
8000 and print an error if asked to create a shared library with a less
8001 restrictive subtype than its input files (for instance, trying to put
8002 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
8003 for executables, @file{ld}, will quietly give the executable the most
8004 restrictive subtype of any of its input files.
8009 Add the framework directory @var{dir} to the head of the list of
8010 directories to be searched for header files. These directories are
8011 interleaved with those specified by @option{-I} options and are
8012 scanned in a left-to-right order.
8014 A framework directory is a directory with frameworks in it. A
8015 framework is a directory with a @samp{"Headers"} and/or
8016 @samp{"PrivateHeaders"} directory contained directly in it that ends
8017 in @samp{".framework"}. The name of a framework is the name of this
8018 directory excluding the @samp{".framework"}. Headers associated with
8019 the framework are found in one of those two directories, with
8020 @samp{"Headers"} being searched first. A subframework is a framework
8021 directory that is in a framework's @samp{"Frameworks"} directory.
8022 Includes of subframework headers can only appear in a header of a
8023 framework that contains the subframework, or in a sibling subframework
8024 header. Two subframeworks are siblings if they occur in the same
8025 framework. A subframework should not have the same name as a
8026 framework, a warning will be issued if this is violated. Currently a
8027 subframework cannot have subframeworks, in the future, the mechanism
8028 may be extended to support this. The standard frameworks can be found
8029 in @samp{"/System/Library/Frameworks"} and
8030 @samp{"/Library/Frameworks"}. An example include looks like
8031 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
8032 the name of the framework and header.h is found in the
8033 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
8037 Emit debugging information for symbols that are used. For STABS
8038 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
8039 This is by default ON@.
8043 Emit debugging information for all symbols and types.
8045 @item -mmacosx-version-min=@var{version}
8046 The earliest version of MacOS X that this executable will run on
8047 is @var{version}. Typical values of @var{version} include @code{10.1},
8048 @code{10.2}, and @code{10.3.9}.
8050 The default for this option is to make choices that seem to be most
8055 Enable kernel development mode. The @option{-mkernel} option sets
8056 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
8057 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
8058 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
8059 applicable. This mode also sets @option{-mno-altivec},
8060 @option{-msoft-float}, @option{-fno-builtin} and
8061 @option{-mlong-branch} for PowerPC targets.
8063 @item -mone-byte-bool
8064 @opindex mone-byte-bool
8065 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
8066 By default @samp{sizeof(bool)} is @samp{4} when compiling for
8067 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
8068 option has no effect on x86.
8070 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
8071 to generate code that is not binary compatible with code generated
8072 without that switch. Using this switch may require recompiling all
8073 other modules in a program, including system libraries. Use this
8074 switch to conform to a non-default data model.
8076 @item -mfix-and-continue
8077 @itemx -ffix-and-continue
8078 @itemx -findirect-data
8079 @opindex mfix-and-continue
8080 @opindex ffix-and-continue
8081 @opindex findirect-data
8082 Generate code suitable for fast turn around development. Needed to
8083 enable gdb to dynamically load @code{.o} files into already running
8084 programs. @option{-findirect-data} and @option{-ffix-and-continue}
8085 are provided for backwards compatibility.
8089 Loads all members of static archive libraries.
8090 See man ld(1) for more information.
8092 @item -arch_errors_fatal
8093 @opindex arch_errors_fatal
8094 Cause the errors having to do with files that have the wrong architecture
8098 @opindex bind_at_load
8099 Causes the output file to be marked such that the dynamic linker will
8100 bind all undefined references when the file is loaded or launched.
8104 Produce a Mach-o bundle format file.
8105 See man ld(1) for more information.
8107 @item -bundle_loader @var{executable}
8108 @opindex bundle_loader
8109 This option specifies the @var{executable} that will be loading the build
8110 output file being linked. See man ld(1) for more information.
8114 When passed this option, GCC will produce a dynamic library instead of
8115 an executable when linking, using the Darwin @file{libtool} command.
8117 @item -force_cpusubtype_ALL
8118 @opindex force_cpusubtype_ALL
8119 This causes GCC's output file to have the @var{ALL} subtype, instead of
8120 one controlled by the @option{-mcpu} or @option{-march} option.
8122 @item -allowable_client @var{client_name}
8124 @itemx -compatibility_version
8125 @itemx -current_version
8127 @itemx -dependency-file
8129 @itemx -dylinker_install_name
8131 @itemx -exported_symbols_list
8133 @itemx -flat_namespace
8134 @itemx -force_flat_namespace
8135 @itemx -headerpad_max_install_names
8138 @itemx -install_name
8139 @itemx -keep_private_externs
8140 @itemx -multi_module
8141 @itemx -multiply_defined
8142 @itemx -multiply_defined_unused
8144 @itemx -no_dead_strip_inits_and_terms
8145 @itemx -nofixprebinding
8148 @itemx -noseglinkedit
8149 @itemx -pagezero_size
8151 @itemx -prebind_all_twolevel_modules
8152 @itemx -private_bundle
8153 @itemx -read_only_relocs
8155 @itemx -sectobjectsymbols
8159 @itemx -sectobjectsymbols
8162 @itemx -segs_read_only_addr
8163 @itemx -segs_read_write_addr
8164 @itemx -seg_addr_table
8165 @itemx -seg_addr_table_filename
8168 @itemx -segs_read_only_addr
8169 @itemx -segs_read_write_addr
8170 @itemx -single_module
8173 @itemx -sub_umbrella
8174 @itemx -twolevel_namespace
8177 @itemx -unexported_symbols_list
8178 @itemx -weak_reference_mismatches
8181 @opindex allowable_client
8182 @opindex client_name
8183 @opindex compatibility_version
8184 @opindex current_version
8186 @opindex dependency-file
8188 @opindex dylinker_install_name
8190 @opindex exported_symbols_list
8192 @opindex flat_namespace
8193 @opindex force_flat_namespace
8194 @opindex headerpad_max_install_names
8197 @opindex install_name
8198 @opindex keep_private_externs
8199 @opindex multi_module
8200 @opindex multiply_defined
8201 @opindex multiply_defined_unused
8203 @opindex no_dead_strip_inits_and_terms
8204 @opindex nofixprebinding
8205 @opindex nomultidefs
8207 @opindex noseglinkedit
8208 @opindex pagezero_size
8210 @opindex prebind_all_twolevel_modules
8211 @opindex private_bundle
8212 @opindex read_only_relocs
8214 @opindex sectobjectsymbols
8218 @opindex sectobjectsymbols
8221 @opindex segs_read_only_addr
8222 @opindex segs_read_write_addr
8223 @opindex seg_addr_table
8224 @opindex seg_addr_table_filename
8225 @opindex seglinkedit
8227 @opindex segs_read_only_addr
8228 @opindex segs_read_write_addr
8229 @opindex single_module
8231 @opindex sub_library
8232 @opindex sub_umbrella
8233 @opindex twolevel_namespace
8236 @opindex unexported_symbols_list
8237 @opindex weak_reference_mismatches
8238 @opindex whatsloaded
8240 These options are passed to the Darwin linker. The Darwin linker man page
8241 describes them in detail.
8244 @node DEC Alpha Options
8245 @subsection DEC Alpha Options
8247 These @samp{-m} options are defined for the DEC Alpha implementations:
8250 @item -mno-soft-float
8252 @opindex mno-soft-float
8253 @opindex msoft-float
8254 Use (do not use) the hardware floating-point instructions for
8255 floating-point operations. When @option{-msoft-float} is specified,
8256 functions in @file{libgcc.a} will be used to perform floating-point
8257 operations. Unless they are replaced by routines that emulate the
8258 floating-point operations, or compiled in such a way as to call such
8259 emulations routines, these routines will issue floating-point
8260 operations. If you are compiling for an Alpha without floating-point
8261 operations, you must ensure that the library is built so as not to call
8264 Note that Alpha implementations without floating-point operations are
8265 required to have floating-point registers.
8270 @opindex mno-fp-regs
8271 Generate code that uses (does not use) the floating-point register set.
8272 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8273 register set is not used, floating point operands are passed in integer
8274 registers as if they were integers and floating-point results are passed
8275 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8276 so any function with a floating-point argument or return value called by code
8277 compiled with @option{-mno-fp-regs} must also be compiled with that
8280 A typical use of this option is building a kernel that does not use,
8281 and hence need not save and restore, any floating-point registers.
8285 The Alpha architecture implements floating-point hardware optimized for
8286 maximum performance. It is mostly compliant with the IEEE floating
8287 point standard. However, for full compliance, software assistance is
8288 required. This option generates code fully IEEE compliant code
8289 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8290 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8291 defined during compilation. The resulting code is less efficient but is
8292 able to correctly support denormalized numbers and exceptional IEEE
8293 values such as not-a-number and plus/minus infinity. Other Alpha
8294 compilers call this option @option{-ieee_with_no_inexact}.
8296 @item -mieee-with-inexact
8297 @opindex mieee-with-inexact
8298 This is like @option{-mieee} except the generated code also maintains
8299 the IEEE @var{inexact-flag}. Turning on this option causes the
8300 generated code to implement fully-compliant IEEE math. In addition to
8301 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8302 macro. On some Alpha implementations the resulting code may execute
8303 significantly slower than the code generated by default. Since there is
8304 very little code that depends on the @var{inexact-flag}, you should
8305 normally not specify this option. Other Alpha compilers call this
8306 option @option{-ieee_with_inexact}.
8308 @item -mfp-trap-mode=@var{trap-mode}
8309 @opindex mfp-trap-mode
8310 This option controls what floating-point related traps are enabled.
8311 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8312 The trap mode can be set to one of four values:
8316 This is the default (normal) setting. The only traps that are enabled
8317 are the ones that cannot be disabled in software (e.g., division by zero
8321 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8325 Like @samp{u}, but the instructions are marked to be safe for software
8326 completion (see Alpha architecture manual for details).
8329 Like @samp{su}, but inexact traps are enabled as well.
8332 @item -mfp-rounding-mode=@var{rounding-mode}
8333 @opindex mfp-rounding-mode
8334 Selects the IEEE rounding mode. Other Alpha compilers call this option
8335 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8340 Normal IEEE rounding mode. Floating point numbers are rounded towards
8341 the nearest machine number or towards the even machine number in case
8345 Round towards minus infinity.
8348 Chopped rounding mode. Floating point numbers are rounded towards zero.
8351 Dynamic rounding mode. A field in the floating point control register
8352 (@var{fpcr}, see Alpha architecture reference manual) controls the
8353 rounding mode in effect. The C library initializes this register for
8354 rounding towards plus infinity. Thus, unless your program modifies the
8355 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8358 @item -mtrap-precision=@var{trap-precision}
8359 @opindex mtrap-precision
8360 In the Alpha architecture, floating point traps are imprecise. This
8361 means without software assistance it is impossible to recover from a
8362 floating trap and program execution normally needs to be terminated.
8363 GCC can generate code that can assist operating system trap handlers
8364 in determining the exact location that caused a floating point trap.
8365 Depending on the requirements of an application, different levels of
8366 precisions can be selected:
8370 Program precision. This option is the default and means a trap handler
8371 can only identify which program caused a floating point exception.
8374 Function precision. The trap handler can determine the function that
8375 caused a floating point exception.
8378 Instruction precision. The trap handler can determine the exact
8379 instruction that caused a floating point exception.
8382 Other Alpha compilers provide the equivalent options called
8383 @option{-scope_safe} and @option{-resumption_safe}.
8385 @item -mieee-conformant
8386 @opindex mieee-conformant
8387 This option marks the generated code as IEEE conformant. You must not
8388 use this option unless you also specify @option{-mtrap-precision=i} and either
8389 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8390 is to emit the line @samp{.eflag 48} in the function prologue of the
8391 generated assembly file. Under DEC Unix, this has the effect that
8392 IEEE-conformant math library routines will be linked in.
8394 @item -mbuild-constants
8395 @opindex mbuild-constants
8396 Normally GCC examines a 32- or 64-bit integer constant to
8397 see if it can construct it from smaller constants in two or three
8398 instructions. If it cannot, it will output the constant as a literal and
8399 generate code to load it from the data segment at runtime.
8401 Use this option to require GCC to construct @emph{all} integer constants
8402 using code, even if it takes more instructions (the maximum is six).
8404 You would typically use this option to build a shared library dynamic
8405 loader. Itself a shared library, it must relocate itself in memory
8406 before it can find the variables and constants in its own data segment.
8412 Select whether to generate code to be assembled by the vendor-supplied
8413 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8431 Indicate whether GCC should generate code to use the optional BWX,
8432 CIX, FIX and MAX instruction sets. The default is to use the instruction
8433 sets supported by the CPU type specified via @option{-mcpu=} option or that
8434 of the CPU on which GCC was built if none was specified.
8439 @opindex mfloat-ieee
8440 Generate code that uses (does not use) VAX F and G floating point
8441 arithmetic instead of IEEE single and double precision.
8443 @item -mexplicit-relocs
8444 @itemx -mno-explicit-relocs
8445 @opindex mexplicit-relocs
8446 @opindex mno-explicit-relocs
8447 Older Alpha assemblers provided no way to generate symbol relocations
8448 except via assembler macros. Use of these macros does not allow
8449 optimal instruction scheduling. GNU binutils as of version 2.12
8450 supports a new syntax that allows the compiler to explicitly mark
8451 which relocations should apply to which instructions. This option
8452 is mostly useful for debugging, as GCC detects the capabilities of
8453 the assembler when it is built and sets the default accordingly.
8457 @opindex msmall-data
8458 @opindex mlarge-data
8459 When @option{-mexplicit-relocs} is in effect, static data is
8460 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8461 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8462 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8463 16-bit relocations off of the @code{$gp} register. This limits the
8464 size of the small data area to 64KB, but allows the variables to be
8465 directly accessed via a single instruction.
8467 The default is @option{-mlarge-data}. With this option the data area
8468 is limited to just below 2GB@. Programs that require more than 2GB of
8469 data must use @code{malloc} or @code{mmap} to allocate the data in the
8470 heap instead of in the program's data segment.
8472 When generating code for shared libraries, @option{-fpic} implies
8473 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8477 @opindex msmall-text
8478 @opindex mlarge-text
8479 When @option{-msmall-text} is used, the compiler assumes that the
8480 code of the entire program (or shared library) fits in 4MB, and is
8481 thus reachable with a branch instruction. When @option{-msmall-data}
8482 is used, the compiler can assume that all local symbols share the
8483 same @code{$gp} value, and thus reduce the number of instructions
8484 required for a function call from 4 to 1.
8486 The default is @option{-mlarge-text}.
8488 @item -mcpu=@var{cpu_type}
8490 Set the instruction set and instruction scheduling parameters for
8491 machine type @var{cpu_type}. You can specify either the @samp{EV}
8492 style name or the corresponding chip number. GCC supports scheduling
8493 parameters for the EV4, EV5 and EV6 family of processors and will
8494 choose the default values for the instruction set from the processor
8495 you specify. If you do not specify a processor type, GCC will default
8496 to the processor on which the compiler was built.
8498 Supported values for @var{cpu_type} are
8504 Schedules as an EV4 and has no instruction set extensions.
8508 Schedules as an EV5 and has no instruction set extensions.
8512 Schedules as an EV5 and supports the BWX extension.
8517 Schedules as an EV5 and supports the BWX and MAX extensions.
8521 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8525 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8528 @item -mtune=@var{cpu_type}
8530 Set only the instruction scheduling parameters for machine type
8531 @var{cpu_type}. The instruction set is not changed.
8533 @item -mmemory-latency=@var{time}
8534 @opindex mmemory-latency
8535 Sets the latency the scheduler should assume for typical memory
8536 references as seen by the application. This number is highly
8537 dependent on the memory access patterns used by the application
8538 and the size of the external cache on the machine.
8540 Valid options for @var{time} are
8544 A decimal number representing clock cycles.
8550 The compiler contains estimates of the number of clock cycles for
8551 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8552 (also called Dcache, Scache, and Bcache), as well as to main memory.
8553 Note that L3 is only valid for EV5.
8558 @node DEC Alpha/VMS Options
8559 @subsection DEC Alpha/VMS Options
8561 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8564 @item -mvms-return-codes
8565 @opindex mvms-return-codes
8566 Return VMS condition codes from main. The default is to return POSIX
8567 style condition (e.g.@ error) codes.
8571 @subsection FRV Options
8578 Only use the first 32 general purpose registers.
8583 Use all 64 general purpose registers.
8588 Use only the first 32 floating point registers.
8593 Use all 64 floating point registers
8596 @opindex mhard-float
8598 Use hardware instructions for floating point operations.
8601 @opindex msoft-float
8603 Use library routines for floating point operations.
8608 Dynamically allocate condition code registers.
8613 Do not try to dynamically allocate condition code registers, only
8614 use @code{icc0} and @code{fcc0}.
8619 Change ABI to use double word insns.
8624 Do not use double word instructions.
8629 Use floating point double instructions.
8634 Do not use floating point double instructions.
8639 Use media instructions.
8644 Do not use media instructions.
8649 Use multiply and add/subtract instructions.
8654 Do not use multiply and add/subtract instructions.
8659 Select the FDPIC ABI, that uses function descriptors to represent
8660 pointers to functions. Without any PIC/PIE-related options, it
8661 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8662 assumes GOT entries and small data are within a 12-bit range from the
8663 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8664 are computed with 32 bits.
8667 @opindex minline-plt
8669 Enable inlining of PLT entries in function calls to functions that are
8670 not known to bind locally. It has no effect without @option{-mfdpic}.
8671 It's enabled by default if optimizing for speed and compiling for
8672 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8673 optimization option such as @option{-O3} or above is present in the
8679 Assume a large TLS segment when generating thread-local code.
8684 Do not assume a large TLS segment when generating thread-local code.
8689 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8690 that is known to be in read-only sections. It's enabled by default,
8691 except for @option{-fpic} or @option{-fpie}: even though it may help
8692 make the global offset table smaller, it trades 1 instruction for 4.
8693 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8694 one of which may be shared by multiple symbols, and it avoids the need
8695 for a GOT entry for the referenced symbol, so it's more likely to be a
8696 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8698 @item -multilib-library-pic
8699 @opindex multilib-library-pic
8701 Link with the (library, not FD) pic libraries. It's implied by
8702 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8703 @option{-fpic} without @option{-mfdpic}. You should never have to use
8709 Follow the EABI requirement of always creating a frame pointer whenever
8710 a stack frame is allocated. This option is enabled by default and can
8711 be disabled with @option{-mno-linked-fp}.
8714 @opindex mlong-calls
8716 Use indirect addressing to call functions outside the current
8717 compilation unit. This allows the functions to be placed anywhere
8718 within the 32-bit address space.
8720 @item -malign-labels
8721 @opindex malign-labels
8723 Try to align labels to an 8-byte boundary by inserting nops into the
8724 previous packet. This option only has an effect when VLIW packing
8725 is enabled. It doesn't create new packets; it merely adds nops to
8729 @opindex mlibrary-pic
8731 Generate position-independent EABI code.
8736 Use only the first four media accumulator registers.
8741 Use all eight media accumulator registers.
8746 Pack VLIW instructions.
8751 Do not pack VLIW instructions.
8756 Do not mark ABI switches in e_flags.
8761 Enable the use of conditional-move instructions (default).
8763 This switch is mainly for debugging the compiler and will likely be removed
8764 in a future version.
8766 @item -mno-cond-move
8767 @opindex mno-cond-move
8769 Disable the use of conditional-move instructions.
8771 This switch is mainly for debugging the compiler and will likely be removed
8772 in a future version.
8777 Enable the use of conditional set instructions (default).
8779 This switch is mainly for debugging the compiler and will likely be removed
8780 in a future version.
8785 Disable the use of conditional set instructions.
8787 This switch is mainly for debugging the compiler and will likely be removed
8788 in a future version.
8793 Enable the use of conditional execution (default).
8795 This switch is mainly for debugging the compiler and will likely be removed
8796 in a future version.
8798 @item -mno-cond-exec
8799 @opindex mno-cond-exec
8801 Disable the use of conditional execution.
8803 This switch is mainly for debugging the compiler and will likely be removed
8804 in a future version.
8807 @opindex mvliw-branch
8809 Run a pass to pack branches into VLIW instructions (default).
8811 This switch is mainly for debugging the compiler and will likely be removed
8812 in a future version.
8814 @item -mno-vliw-branch
8815 @opindex mno-vliw-branch
8817 Do not run a pass to pack branches into VLIW instructions.
8819 This switch is mainly for debugging the compiler and will likely be removed
8820 in a future version.
8822 @item -mmulti-cond-exec
8823 @opindex mmulti-cond-exec
8825 Enable optimization of @code{&&} and @code{||} in conditional execution
8828 This switch is mainly for debugging the compiler and will likely be removed
8829 in a future version.
8831 @item -mno-multi-cond-exec
8832 @opindex mno-multi-cond-exec
8834 Disable optimization of @code{&&} and @code{||} in conditional execution.
8836 This switch is mainly for debugging the compiler and will likely be removed
8837 in a future version.
8839 @item -mnested-cond-exec
8840 @opindex mnested-cond-exec
8842 Enable nested conditional execution optimizations (default).
8844 This switch is mainly for debugging the compiler and will likely be removed
8845 in a future version.
8847 @item -mno-nested-cond-exec
8848 @opindex mno-nested-cond-exec
8850 Disable nested conditional execution optimizations.
8852 This switch is mainly for debugging the compiler and will likely be removed
8853 in a future version.
8855 @item -moptimize-membar
8856 @opindex moptimize-membar
8858 This switch removes redundant @code{membar} instructions from the
8859 compiler generated code. It is enabled by default.
8861 @item -mno-optimize-membar
8862 @opindex mno-optimize-membar
8864 This switch disables the automatic removal of redundant @code{membar}
8865 instructions from the generated code.
8867 @item -mtomcat-stats
8868 @opindex mtomcat-stats
8870 Cause gas to print out tomcat statistics.
8872 @item -mcpu=@var{cpu}
8875 Select the processor type for which to generate code. Possible values are
8876 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8877 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8881 @node GNU/Linux Options
8882 @subsection GNU/Linux Options
8884 These @samp{-m} options are defined for GNU/Linux targets:
8889 Use the GNU C library instead of uClibc. This is the default except
8890 on @samp{*-*-linux-*uclibc*} targets.
8894 Use uClibc instead of the GNU C library. This is the default on
8895 @samp{*-*-linux-*uclibc*} targets.
8898 @node H8/300 Options
8899 @subsection H8/300 Options
8901 These @samp{-m} options are defined for the H8/300 implementations:
8906 Shorten some address references at link time, when possible; uses the
8907 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8908 ld, Using ld}, for a fuller description.
8912 Generate code for the H8/300H@.
8916 Generate code for the H8S@.
8920 Generate code for the H8S and H8/300H in the normal mode. This switch
8921 must be used either with @option{-mh} or @option{-ms}.
8925 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8929 Make @code{int} data 32 bits by default.
8933 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8934 The default for the H8/300H and H8S is to align longs and floats on 4
8936 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8937 This option has no effect on the H8/300.
8941 @subsection HPPA Options
8942 @cindex HPPA Options
8944 These @samp{-m} options are defined for the HPPA family of computers:
8947 @item -march=@var{architecture-type}
8949 Generate code for the specified architecture. The choices for
8950 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8951 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8952 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8953 architecture option for your machine. Code compiled for lower numbered
8954 architectures will run on higher numbered architectures, but not the
8958 @itemx -mpa-risc-1-1
8959 @itemx -mpa-risc-2-0
8960 @opindex mpa-risc-1-0
8961 @opindex mpa-risc-1-1
8962 @opindex mpa-risc-2-0
8963 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8966 @opindex mbig-switch
8967 Generate code suitable for big switch tables. Use this option only if
8968 the assembler/linker complain about out of range branches within a switch
8971 @item -mjump-in-delay
8972 @opindex mjump-in-delay
8973 Fill delay slots of function calls with unconditional jump instructions
8974 by modifying the return pointer for the function call to be the target
8975 of the conditional jump.
8977 @item -mdisable-fpregs
8978 @opindex mdisable-fpregs
8979 Prevent floating point registers from being used in any manner. This is
8980 necessary for compiling kernels which perform lazy context switching of
8981 floating point registers. If you use this option and attempt to perform
8982 floating point operations, the compiler will abort.
8984 @item -mdisable-indexing
8985 @opindex mdisable-indexing
8986 Prevent the compiler from using indexing address modes. This avoids some
8987 rather obscure problems when compiling MIG generated code under MACH@.
8989 @item -mno-space-regs
8990 @opindex mno-space-regs
8991 Generate code that assumes the target has no space registers. This allows
8992 GCC to generate faster indirect calls and use unscaled index address modes.
8994 Such code is suitable for level 0 PA systems and kernels.
8996 @item -mfast-indirect-calls
8997 @opindex mfast-indirect-calls
8998 Generate code that assumes calls never cross space boundaries. This
8999 allows GCC to emit code which performs faster indirect calls.
9001 This option will not work in the presence of shared libraries or nested
9004 @item -mfixed-range=@var{register-range}
9005 @opindex mfixed-range
9006 Generate code treating the given register range as fixed registers.
9007 A fixed register is one that the register allocator can not use. This is
9008 useful when compiling kernel code. A register range is specified as
9009 two registers separated by a dash. Multiple register ranges can be
9010 specified separated by a comma.
9012 @item -mlong-load-store
9013 @opindex mlong-load-store
9014 Generate 3-instruction load and store sequences as sometimes required by
9015 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
9018 @item -mportable-runtime
9019 @opindex mportable-runtime
9020 Use the portable calling conventions proposed by HP for ELF systems.
9024 Enable the use of assembler directives only GAS understands.
9026 @item -mschedule=@var{cpu-type}
9028 Schedule code according to the constraints for the machine type
9029 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
9030 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
9031 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
9032 proper scheduling option for your machine. The default scheduling is
9036 @opindex mlinker-opt
9037 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
9038 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
9039 linkers in which they give bogus error messages when linking some programs.
9042 @opindex msoft-float
9043 Generate output containing library calls for floating point.
9044 @strong{Warning:} the requisite libraries are not available for all HPPA
9045 targets. Normally the facilities of the machine's usual C compiler are
9046 used, but this cannot be done directly in cross-compilation. You must make
9047 your own arrangements to provide suitable library functions for
9048 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
9049 does provide software floating point support.
9051 @option{-msoft-float} changes the calling convention in the output file;
9052 therefore, it is only useful if you compile @emph{all} of a program with
9053 this option. In particular, you need to compile @file{libgcc.a}, the
9054 library that comes with GCC, with @option{-msoft-float} in order for
9059 Generate the predefine, @code{_SIO}, for server IO@. The default is
9060 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
9061 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
9062 options are available under HP-UX and HI-UX@.
9066 Use GNU ld specific options. This passes @option{-shared} to ld when
9067 building a shared library. It is the default when GCC is configured,
9068 explicitly or implicitly, with the GNU linker. This option does not
9069 have any affect on which ld is called, it only changes what parameters
9070 are passed to that ld. The ld that is called is determined by the
9071 @option{--with-ld} configure option, GCC's program search path, and
9072 finally by the user's @env{PATH}. The linker used by GCC can be printed
9073 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
9074 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9078 Use HP ld specific options. This passes @option{-b} to ld when building
9079 a shared library and passes @option{+Accept TypeMismatch} to ld on all
9080 links. It is the default when GCC is configured, explicitly or
9081 implicitly, with the HP linker. This option does not have any affect on
9082 which ld is called, it only changes what parameters are passed to that
9083 ld. The ld that is called is determined by the @option{--with-ld}
9084 configure option, GCC's program search path, and finally by the user's
9085 @env{PATH}. The linker used by GCC can be printed using @samp{which
9086 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
9087 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9090 @opindex mno-long-calls
9091 Generate code that uses long call sequences. This ensures that a call
9092 is always able to reach linker generated stubs. The default is to generate
9093 long calls only when the distance from the call site to the beginning
9094 of the function or translation unit, as the case may be, exceeds a
9095 predefined limit set by the branch type being used. The limits for
9096 normal calls are 7,600,000 and 240,000 bytes, respectively for the
9097 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
9100 Distances are measured from the beginning of functions when using the
9101 @option{-ffunction-sections} option, or when using the @option{-mgas}
9102 and @option{-mno-portable-runtime} options together under HP-UX with
9105 It is normally not desirable to use this option as it will degrade
9106 performance. However, it may be useful in large applications,
9107 particularly when partial linking is used to build the application.
9109 The types of long calls used depends on the capabilities of the
9110 assembler and linker, and the type of code being generated. The
9111 impact on systems that support long absolute calls, and long pic
9112 symbol-difference or pc-relative calls should be relatively small.
9113 However, an indirect call is used on 32-bit ELF systems in pic code
9114 and it is quite long.
9116 @item -munix=@var{unix-std}
9118 Generate compiler predefines and select a startfile for the specified
9119 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
9120 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
9121 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
9122 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
9123 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
9126 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
9127 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
9128 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
9129 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
9130 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
9131 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
9133 It is @emph{important} to note that this option changes the interfaces
9134 for various library routines. It also affects the operational behavior
9135 of the C library. Thus, @emph{extreme} care is needed in using this
9138 Library code that is intended to operate with more than one UNIX
9139 standard must test, set and restore the variable @var{__xpg4_extended_mask}
9140 as appropriate. Most GNU software doesn't provide this capability.
9144 Suppress the generation of link options to search libdld.sl when the
9145 @option{-static} option is specified on HP-UX 10 and later.
9149 The HP-UX implementation of setlocale in libc has a dependency on
9150 libdld.sl. There isn't an archive version of libdld.sl. Thus,
9151 when the @option{-static} option is specified, special link options
9152 are needed to resolve this dependency.
9154 On HP-UX 10 and later, the GCC driver adds the necessary options to
9155 link with libdld.sl when the @option{-static} option is specified.
9156 This causes the resulting binary to be dynamic. On the 64-bit port,
9157 the linkers generate dynamic binaries by default in any case. The
9158 @option{-nolibdld} option can be used to prevent the GCC driver from
9159 adding these link options.
9163 Add support for multithreading with the @dfn{dce thread} library
9164 under HP-UX@. This option sets flags for both the preprocessor and
9168 @node i386 and x86-64 Options
9169 @subsection Intel 386 and AMD x86-64 Options
9170 @cindex i386 Options
9171 @cindex x86-64 Options
9172 @cindex Intel 386 Options
9173 @cindex AMD x86-64 Options
9175 These @samp{-m} options are defined for the i386 and x86-64 family of
9179 @item -mtune=@var{cpu-type}
9181 Tune to @var{cpu-type} everything applicable about the generated code, except
9182 for the ABI and the set of available instructions. The choices for
9186 Produce code optimized for the most common IA32/AMD64/EM64T processors.
9187 If you know the CPU on which your code will run, then you should use
9188 the corresponding @option{-mtune} option instead of
9189 @option{-mtune=generic}. But, if you do not know exactly what CPU users
9190 of your application will have, then you should use this option.
9192 As new processors are deployed in the marketplace, the behavior of this
9193 option will change. Therefore, if you upgrade to a newer version of
9194 GCC, the code generated option will change to reflect the processors
9195 that were most common when that version of GCC was released.
9197 There is no @option{-march=generic} option because @option{-march}
9198 indicates the instruction set the compiler can use, and there is no
9199 generic instruction set applicable to all processors. In contrast,
9200 @option{-mtune} indicates the processor (or, in this case, collection of
9201 processors) for which the code is optimized.
9203 This selects the CPU to tune for at compilation time by determining
9204 the processor type of the compiling machine. Using @option{-mtune=native}
9205 will produce code optimized for the local machine under the constraints
9206 of the selected instruction set. Using @option{-march=native} will
9207 enable all instruction subsets supported by the local machine (hence
9208 the result might not run on different machines).
9210 Original Intel's i386 CPU@.
9212 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
9214 Intel Pentium CPU with no MMX support.
9216 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
9218 Intel PentiumPro CPU@.
9220 Same as @code{generic}, but when used as @code{march} option, PentiumPro
9221 instruction set will be used, so the code will run on all i686 family chips.
9223 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
9224 @item pentium3, pentium3m
9225 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
9228 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
9229 support. Used by Centrino notebooks.
9230 @item pentium4, pentium4m
9231 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
9233 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
9236 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
9237 SSE2 and SSE3 instruction set support.
9239 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
9240 instruction set support.
9242 AMD K6 CPU with MMX instruction set support.
9244 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
9245 @item athlon, athlon-tbird
9246 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
9248 @item athlon-4, athlon-xp, athlon-mp
9249 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
9250 instruction set support.
9251 @item k8, opteron, athlon64, athlon-fx
9252 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
9253 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
9254 @item k8-sse3, opteron-sse3, athlon64-sse3
9255 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
9256 @item amdfam10, barcelona
9257 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
9258 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
9259 instruction set extensions.)
9261 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
9264 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
9265 instruction set support.
9267 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
9268 implemented for this chip.)
9270 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
9271 implemented for this chip.)
9273 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
9276 While picking a specific @var{cpu-type} will schedule things appropriately
9277 for that particular chip, the compiler will not generate any code that
9278 does not run on the i386 without the @option{-march=@var{cpu-type}} option
9281 @item -march=@var{cpu-type}
9283 Generate instructions for the machine type @var{cpu-type}. The choices
9284 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
9285 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
9287 @item -mcpu=@var{cpu-type}
9289 A deprecated synonym for @option{-mtune}.
9298 @opindex mpentiumpro
9299 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
9300 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
9301 These synonyms are deprecated.
9303 @item -mfpmath=@var{unit}
9305 Generate floating point arithmetics for selected unit @var{unit}. The choices
9310 Use the standard 387 floating point coprocessor present majority of chips and
9311 emulated otherwise. Code compiled with this option will run almost everywhere.
9312 The temporary results are computed in 80bit precision instead of precision
9313 specified by the type resulting in slightly different results compared to most
9314 of other chips. See @option{-ffloat-store} for more detailed description.
9316 This is the default choice for i386 compiler.
9319 Use scalar floating point instructions present in the SSE instruction set.
9320 This instruction set is supported by Pentium3 and newer chips, in the AMD line
9321 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
9322 instruction set supports only single precision arithmetics, thus the double and
9323 extended precision arithmetics is still done using 387. Later version, present
9324 only in Pentium4 and the future AMD x86-64 chips supports double precision
9327 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
9328 or @option{-msse2} switches to enable SSE extensions and make this option
9329 effective. For the x86-64 compiler, these extensions are enabled by default.
9331 The resulting code should be considerably faster in the majority of cases and avoid
9332 the numerical instability problems of 387 code, but may break some existing
9333 code that expects temporaries to be 80bit.
9335 This is the default choice for the x86-64 compiler.
9338 Attempt to utilize both instruction sets at once. This effectively double the
9339 amount of available registers and on chips with separate execution units for
9340 387 and SSE the execution resources too. Use this option with care, as it is
9341 still experimental, because the GCC register allocator does not model separate
9342 functional units well resulting in instable performance.
9345 @item -masm=@var{dialect}
9346 @opindex masm=@var{dialect}
9347 Output asm instructions using selected @var{dialect}. Supported
9348 choices are @samp{intel} or @samp{att} (the default one). Darwin does
9349 not support @samp{intel}.
9354 @opindex mno-ieee-fp
9355 Control whether or not the compiler uses IEEE floating point
9356 comparisons. These handle correctly the case where the result of a
9357 comparison is unordered.
9360 @opindex msoft-float
9361 Generate output containing library calls for floating point.
9362 @strong{Warning:} the requisite libraries are not part of GCC@.
9363 Normally the facilities of the machine's usual C compiler are used, but
9364 this can't be done directly in cross-compilation. You must make your
9365 own arrangements to provide suitable library functions for
9368 On machines where a function returns floating point results in the 80387
9369 register stack, some floating point opcodes may be emitted even if
9370 @option{-msoft-float} is used.
9372 @item -mno-fp-ret-in-387
9373 @opindex mno-fp-ret-in-387
9374 Do not use the FPU registers for return values of functions.
9376 The usual calling convention has functions return values of types
9377 @code{float} and @code{double} in an FPU register, even if there
9378 is no FPU@. The idea is that the operating system should emulate
9381 The option @option{-mno-fp-ret-in-387} causes such values to be returned
9382 in ordinary CPU registers instead.
9384 @item -mno-fancy-math-387
9385 @opindex mno-fancy-math-387
9386 Some 387 emulators do not support the @code{sin}, @code{cos} and
9387 @code{sqrt} instructions for the 387. Specify this option to avoid
9388 generating those instructions. This option is the default on
9389 OpenBSD and NetBSD@. This option is overridden when @option{-march}
9390 indicates that the target cpu will always have an FPU and so the
9391 instruction will not need emulation. As of revision 2.6.1, these
9392 instructions are not generated unless you also use the
9393 @option{-funsafe-math-optimizations} switch.
9395 @item -malign-double
9396 @itemx -mno-align-double
9397 @opindex malign-double
9398 @opindex mno-align-double
9399 Control whether GCC aligns @code{double}, @code{long double}, and
9400 @code{long long} variables on a two word boundary or a one word
9401 boundary. Aligning @code{double} variables on a two word boundary will
9402 produce code that runs somewhat faster on a @samp{Pentium} at the
9403 expense of more memory.
9405 On x86-64, @option{-malign-double} is enabled by default.
9407 @strong{Warning:} if you use the @option{-malign-double} switch,
9408 structures containing the above types will be aligned differently than
9409 the published application binary interface specifications for the 386
9410 and will not be binary compatible with structures in code compiled
9411 without that switch.
9413 @item -m96bit-long-double
9414 @itemx -m128bit-long-double
9415 @opindex m96bit-long-double
9416 @opindex m128bit-long-double
9417 These switches control the size of @code{long double} type. The i386
9418 application binary interface specifies the size to be 96 bits,
9419 so @option{-m96bit-long-double} is the default in 32 bit mode.
9421 Modern architectures (Pentium and newer) would prefer @code{long double}
9422 to be aligned to an 8 or 16 byte boundary. In arrays or structures
9423 conforming to the ABI, this would not be possible. So specifying a
9424 @option{-m128bit-long-double} will align @code{long double}
9425 to a 16 byte boundary by padding the @code{long double} with an additional
9428 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
9429 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
9431 Notice that neither of these options enable any extra precision over the x87
9432 standard of 80 bits for a @code{long double}.
9434 @strong{Warning:} if you override the default value for your target ABI, the
9435 structures and arrays containing @code{long double} variables will change
9436 their size as well as function calling convention for function taking
9437 @code{long double} will be modified. Hence they will not be binary
9438 compatible with arrays or structures in code compiled without that switch.
9440 @item -mmlarge-data-threshold=@var{number}
9441 @opindex mlarge-data-threshold=@var{number}
9442 When @option{-mcmodel=medium} is specified, the data greater than
9443 @var{threshold} are placed in large data section. This value must be the
9444 same across all object linked into the binary and defaults to 65535.
9447 @itemx -mno-svr3-shlib
9448 @opindex msvr3-shlib
9449 @opindex mno-svr3-shlib
9450 Control whether GCC places uninitialized local variables into the
9451 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
9452 into @code{bss}. These options are meaningful only on System V Release 3.
9456 Use a different function-calling convention, in which functions that
9457 take a fixed number of arguments return with the @code{ret} @var{num}
9458 instruction, which pops their arguments while returning. This saves one
9459 instruction in the caller since there is no need to pop the arguments
9462 You can specify that an individual function is called with this calling
9463 sequence with the function attribute @samp{stdcall}. You can also
9464 override the @option{-mrtd} option by using the function attribute
9465 @samp{cdecl}. @xref{Function Attributes}.
9467 @strong{Warning:} this calling convention is incompatible with the one
9468 normally used on Unix, so you cannot use it if you need to call
9469 libraries compiled with the Unix compiler.
9471 Also, you must provide function prototypes for all functions that
9472 take variable numbers of arguments (including @code{printf});
9473 otherwise incorrect code will be generated for calls to those
9476 In addition, seriously incorrect code will result if you call a
9477 function with too many arguments. (Normally, extra arguments are
9478 harmlessly ignored.)
9480 @item -mregparm=@var{num}
9482 Control how many registers are used to pass integer arguments. By
9483 default, no registers are used to pass arguments, and at most 3
9484 registers can be used. You can control this behavior for a specific
9485 function by using the function attribute @samp{regparm}.
9486 @xref{Function Attributes}.
9488 @strong{Warning:} if you use this switch, and
9489 @var{num} is nonzero, then you must build all modules with the same
9490 value, including any libraries. This includes the system libraries and
9494 @opindex msseregparm
9495 Use SSE register passing conventions for float and double arguments
9496 and return values. You can control this behavior for a specific
9497 function by using the function attribute @samp{sseregparm}.
9498 @xref{Function Attributes}.
9500 @strong{Warning:} if you use this switch then you must build all
9501 modules with the same value, including any libraries. This includes
9502 the system libraries and startup modules.
9504 @item -mstackrealign
9505 @opindex mstackrealign
9506 Realign the stack at entry. On the Intel x86, the
9507 @option{-mstackrealign} option will generate an alternate prologue and
9508 epilogue that realigns the runtime stack. This supports mixing legacy
9509 codes that keep a 4-byte aligned stack with modern codes that keep a
9510 16-byte stack for SSE compatibility. The alternate prologue and
9511 epilogue are slower and bigger than the regular ones, and the
9512 alternate prologue requires an extra scratch register; this lowers the
9513 number of registers available if used in conjunction with the
9514 @code{regparm} attribute. The @option{-mstackrealign} option is
9515 incompatible with the nested function prologue; this is considered a
9516 hard error. See also the attribute @code{force_align_arg_pointer},
9517 applicable to individual functions.
9519 @item -mpreferred-stack-boundary=@var{num}
9520 @opindex mpreferred-stack-boundary
9521 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
9522 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
9523 the default is 4 (16 bytes or 128 bits).
9525 On Pentium and PentiumPro, @code{double} and @code{long double} values
9526 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
9527 suffer significant run time performance penalties. On Pentium III, the
9528 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
9529 properly if it is not 16 byte aligned.
9531 To ensure proper alignment of this values on the stack, the stack boundary
9532 must be as aligned as that required by any value stored on the stack.
9533 Further, every function must be generated such that it keeps the stack
9534 aligned. Thus calling a function compiled with a higher preferred
9535 stack boundary from a function compiled with a lower preferred stack
9536 boundary will most likely misalign the stack. It is recommended that
9537 libraries that use callbacks always use the default setting.
9539 This extra alignment does consume extra stack space, and generally
9540 increases code size. Code that is sensitive to stack space usage, such
9541 as embedded systems and operating system kernels, may want to reduce the
9542 preferred alignment to @option{-mpreferred-stack-boundary=2}.
9570 These switches enable or disable the use of instructions in the MMX,
9571 SSE, SSE2, SSE3, SSSE3, SSE4A, ABM, AES or 3DNow! extended
9572 instruction sets. These extensions are also available as built-in
9573 functions: see @ref{X86 Built-in Functions}, for details of the functions
9574 enabled and disabled by these switches.
9576 To have SSE/SSE2 instructions generated automatically from floating-point
9577 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
9579 These options will enable GCC to use these extended instructions in
9580 generated code, even without @option{-mfpmath=sse}. Applications which
9581 perform runtime CPU detection must compile separate files for each
9582 supported architecture, using the appropriate flags. In particular,
9583 the file containing the CPU detection code should be compiled without
9587 @itemx -mno-push-args
9589 @opindex mno-push-args
9590 Use PUSH operations to store outgoing parameters. This method is shorter
9591 and usually equally fast as method using SUB/MOV operations and is enabled
9592 by default. In some cases disabling it may improve performance because of
9593 improved scheduling and reduced dependencies.
9595 @item -maccumulate-outgoing-args
9596 @opindex maccumulate-outgoing-args
9597 If enabled, the maximum amount of space required for outgoing arguments will be
9598 computed in the function prologue. This is faster on most modern CPUs
9599 because of reduced dependencies, improved scheduling and reduced stack usage
9600 when preferred stack boundary is not equal to 2. The drawback is a notable
9601 increase in code size. This switch implies @option{-mno-push-args}.
9605 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
9606 on thread-safe exception handling must compile and link all code with the
9607 @option{-mthreads} option. When compiling, @option{-mthreads} defines
9608 @option{-D_MT}; when linking, it links in a special thread helper library
9609 @option{-lmingwthrd} which cleans up per thread exception handling data.
9611 @item -mno-align-stringops
9612 @opindex mno-align-stringops
9613 Do not align destination of inlined string operations. This switch reduces
9614 code size and improves performance in case the destination is already aligned,
9615 but GCC doesn't know about it.
9617 @item -minline-all-stringops
9618 @opindex minline-all-stringops
9619 By default GCC inlines string operations only when destination is known to be
9620 aligned at least to 4 byte boundary. This enables more inlining, increase code
9621 size, but may improve performance of code that depends on fast memcpy, strlen
9622 and memset for short lengths.
9624 @item -momit-leaf-frame-pointer
9625 @opindex momit-leaf-frame-pointer
9626 Don't keep the frame pointer in a register for leaf functions. This
9627 avoids the instructions to save, set up and restore frame pointers and
9628 makes an extra register available in leaf functions. The option
9629 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9630 which might make debugging harder.
9632 @item -mtls-direct-seg-refs
9633 @itemx -mno-tls-direct-seg-refs
9634 @opindex mtls-direct-seg-refs
9635 Controls whether TLS variables may be accessed with offsets from the
9636 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
9637 or whether the thread base pointer must be added. Whether or not this
9638 is legal depends on the operating system, and whether it maps the
9639 segment to cover the entire TLS area.
9641 For systems that use GNU libc, the default is on.
9644 These @samp{-m} switches are supported in addition to the above
9645 on AMD x86-64 processors in 64-bit environments.
9652 Generate code for a 32-bit or 64-bit environment.
9653 The 32-bit environment sets int, long and pointer to 32 bits and
9654 generates code that runs on any i386 system.
9655 The 64-bit environment sets int to 32 bits and long and pointer
9656 to 64 bits and generates code for AMD's x86-64 architecture. For
9657 darwin only the -m64 option turns off the @option{-fno-pic} and
9658 @option{-mdynamic-no-pic} options.
9661 @opindex no-red-zone
9662 Do not use a so called red zone for x86-64 code. The red zone is mandated
9663 by the x86-64 ABI, it is a 128-byte area beyond the location of the
9664 stack pointer that will not be modified by signal or interrupt handlers
9665 and therefore can be used for temporary data without adjusting the stack
9666 pointer. The flag @option{-mno-red-zone} disables this red zone.
9668 @item -mcmodel=small
9669 @opindex mcmodel=small
9670 Generate code for the small code model: the program and its symbols must
9671 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
9672 Programs can be statically or dynamically linked. This is the default
9675 @item -mcmodel=kernel
9676 @opindex mcmodel=kernel
9677 Generate code for the kernel code model. The kernel runs in the
9678 negative 2 GB of the address space.
9679 This model has to be used for Linux kernel code.
9681 @item -mcmodel=medium
9682 @opindex mcmodel=medium
9683 Generate code for the medium model: The program is linked in the lower 2
9684 GB of the address space but symbols can be located anywhere in the
9685 address space. Programs can be statically or dynamically linked, but
9686 building of shared libraries are not supported with the medium model.
9688 @item -mcmodel=large
9689 @opindex mcmodel=large
9690 Generate code for the large model: This model makes no assumptions
9691 about addresses and sizes of sections. Currently GCC does not implement
9696 @subsection IA-64 Options
9697 @cindex IA-64 Options
9699 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9703 @opindex mbig-endian
9704 Generate code for a big endian target. This is the default for HP-UX@.
9706 @item -mlittle-endian
9707 @opindex mlittle-endian
9708 Generate code for a little endian target. This is the default for AIX5
9715 Generate (or don't) code for the GNU assembler. This is the default.
9716 @c Also, this is the default if the configure option @option{--with-gnu-as}
9723 Generate (or don't) code for the GNU linker. This is the default.
9724 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9729 Generate code that does not use a global pointer register. The result
9730 is not position independent code, and violates the IA-64 ABI@.
9732 @item -mvolatile-asm-stop
9733 @itemx -mno-volatile-asm-stop
9734 @opindex mvolatile-asm-stop
9735 @opindex mno-volatile-asm-stop
9736 Generate (or don't) a stop bit immediately before and after volatile asm
9739 @item -mregister-names
9740 @itemx -mno-register-names
9741 @opindex mregister-names
9742 @opindex mno-register-names
9743 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9744 the stacked registers. This may make assembler output more readable.
9750 Disable (or enable) optimizations that use the small data section. This may
9751 be useful for working around optimizer bugs.
9754 @opindex mconstant-gp
9755 Generate code that uses a single constant global pointer value. This is
9756 useful when compiling kernel code.
9760 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9761 This is useful when compiling firmware code.
9763 @item -minline-float-divide-min-latency
9764 @opindex minline-float-divide-min-latency
9765 Generate code for inline divides of floating point values
9766 using the minimum latency algorithm.
9768 @item -minline-float-divide-max-throughput
9769 @opindex minline-float-divide-max-throughput
9770 Generate code for inline divides of floating point values
9771 using the maximum throughput algorithm.
9773 @item -minline-int-divide-min-latency
9774 @opindex minline-int-divide-min-latency
9775 Generate code for inline divides of integer values
9776 using the minimum latency algorithm.
9778 @item -minline-int-divide-max-throughput
9779 @opindex minline-int-divide-max-throughput
9780 Generate code for inline divides of integer values
9781 using the maximum throughput algorithm.
9783 @item -minline-sqrt-min-latency
9784 @opindex minline-sqrt-min-latency
9785 Generate code for inline square roots
9786 using the minimum latency algorithm.
9788 @item -minline-sqrt-max-throughput
9789 @opindex minline-sqrt-max-throughput
9790 Generate code for inline square roots
9791 using the maximum throughput algorithm.
9793 @item -mno-dwarf2-asm
9795 @opindex mno-dwarf2-asm
9796 @opindex mdwarf2-asm
9797 Don't (or do) generate assembler code for the DWARF2 line number debugging
9798 info. This may be useful when not using the GNU assembler.
9800 @item -mearly-stop-bits
9801 @itemx -mno-early-stop-bits
9802 @opindex mearly-stop-bits
9803 @opindex mno-early-stop-bits
9804 Allow stop bits to be placed earlier than immediately preceding the
9805 instruction that triggered the stop bit. This can improve instruction
9806 scheduling, but does not always do so.
9808 @item -mfixed-range=@var{register-range}
9809 @opindex mfixed-range
9810 Generate code treating the given register range as fixed registers.
9811 A fixed register is one that the register allocator can not use. This is
9812 useful when compiling kernel code. A register range is specified as
9813 two registers separated by a dash. Multiple register ranges can be
9814 specified separated by a comma.
9816 @item -mtls-size=@var{tls-size}
9818 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9821 @item -mtune=@var{cpu-type}
9823 Tune the instruction scheduling for a particular CPU, Valid values are
9824 itanium, itanium1, merced, itanium2, and mckinley.
9830 Add support for multithreading using the POSIX threads library. This
9831 option sets flags for both the preprocessor and linker. It does
9832 not affect the thread safety of object code produced by the compiler or
9833 that of libraries supplied with it. These are HP-UX specific flags.
9839 Generate code for a 32-bit or 64-bit environment.
9840 The 32-bit environment sets int, long and pointer to 32 bits.
9841 The 64-bit environment sets int to 32 bits and long and pointer
9842 to 64 bits. These are HP-UX specific flags.
9844 @item -mno-sched-br-data-spec
9845 @itemx -msched-br-data-spec
9846 @opindex mno-sched-br-data-spec
9847 @opindex msched-br-data-spec
9848 (Dis/En)able data speculative scheduling before reload.
9849 This will result in generation of the ld.a instructions and
9850 the corresponding check instructions (ld.c / chk.a).
9851 The default is 'disable'.
9853 @item -msched-ar-data-spec
9854 @itemx -mno-sched-ar-data-spec
9855 @opindex msched-ar-data-spec
9856 @opindex mno-sched-ar-data-spec
9857 (En/Dis)able data speculative scheduling after reload.
9858 This will result in generation of the ld.a instructions and
9859 the corresponding check instructions (ld.c / chk.a).
9860 The default is 'enable'.
9862 @item -mno-sched-control-spec
9863 @itemx -msched-control-spec
9864 @opindex mno-sched-control-spec
9865 @opindex msched-control-spec
9866 (Dis/En)able control speculative scheduling. This feature is
9867 available only during region scheduling (i.e. before reload).
9868 This will result in generation of the ld.s instructions and
9869 the corresponding check instructions chk.s .
9870 The default is 'disable'.
9872 @item -msched-br-in-data-spec
9873 @itemx -mno-sched-br-in-data-spec
9874 @opindex msched-br-in-data-spec
9875 @opindex mno-sched-br-in-data-spec
9876 (En/Dis)able speculative scheduling of the instructions that
9877 are dependent on the data speculative loads before reload.
9878 This is effective only with @option{-msched-br-data-spec} enabled.
9879 The default is 'enable'.
9881 @item -msched-ar-in-data-spec
9882 @itemx -mno-sched-ar-in-data-spec
9883 @opindex msched-ar-in-data-spec
9884 @opindex mno-sched-ar-in-data-spec
9885 (En/Dis)able speculative scheduling of the instructions that
9886 are dependent on the data speculative loads after reload.
9887 This is effective only with @option{-msched-ar-data-spec} enabled.
9888 The default is 'enable'.
9890 @item -msched-in-control-spec
9891 @itemx -mno-sched-in-control-spec
9892 @opindex msched-in-control-spec
9893 @opindex mno-sched-in-control-spec
9894 (En/Dis)able speculative scheduling of the instructions that
9895 are dependent on the control speculative loads.
9896 This is effective only with @option{-msched-control-spec} enabled.
9897 The default is 'enable'.
9900 @itemx -mno-sched-ldc
9902 @opindex mno-sched-ldc
9903 (En/Dis)able use of simple data speculation checks ld.c .
9904 If disabled, only chk.a instructions will be emitted to check
9905 data speculative loads.
9906 The default is 'enable'.
9908 @item -mno-sched-control-ldc
9909 @itemx -msched-control-ldc
9910 @opindex mno-sched-control-ldc
9911 @opindex msched-control-ldc
9912 (Dis/En)able use of ld.c instructions to check control speculative loads.
9913 If enabled, in case of control speculative load with no speculatively
9914 scheduled dependent instructions this load will be emitted as ld.sa and
9915 ld.c will be used to check it.
9916 The default is 'disable'.
9918 @item -mno-sched-spec-verbose
9919 @itemx -msched-spec-verbose
9920 @opindex mno-sched-spec-verbose
9921 @opindex msched-spec-verbose
9922 (Dis/En)able printing of the information about speculative motions.
9924 @item -mno-sched-prefer-non-data-spec-insns
9925 @itemx -msched-prefer-non-data-spec-insns
9926 @opindex mno-sched-prefer-non-data-spec-insns
9927 @opindex msched-prefer-non-data-spec-insns
9928 If enabled, data speculative instructions will be chosen for schedule
9929 only if there are no other choices at the moment. This will make
9930 the use of the data speculation much more conservative.
9931 The default is 'disable'.
9933 @item -mno-sched-prefer-non-control-spec-insns
9934 @itemx -msched-prefer-non-control-spec-insns
9935 @opindex mno-sched-prefer-non-control-spec-insns
9936 @opindex msched-prefer-non-control-spec-insns
9937 If enabled, control speculative instructions will be chosen for schedule
9938 only if there are no other choices at the moment. This will make
9939 the use of the control speculation much more conservative.
9940 The default is 'disable'.
9942 @item -mno-sched-count-spec-in-critical-path
9943 @itemx -msched-count-spec-in-critical-path
9944 @opindex mno-sched-count-spec-in-critical-path
9945 @opindex msched-count-spec-in-critical-path
9946 If enabled, speculative dependencies will be considered during
9947 computation of the instructions priorities. This will make the use of the
9948 speculation a bit more conservative.
9949 The default is 'disable'.
9954 @subsection M32C Options
9955 @cindex M32C options
9958 @item -mcpu=@var{name}
9960 Select the CPU for which code is generated. @var{name} may be one of
9961 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
9962 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
9967 Specifies that the program will be run on the simulator. This causes
9968 an alternate runtime library to be linked in which supports, for
9969 example, file I/O. You must not use this option when generating
9970 programs that will run on real hardware; you must provide your own
9971 runtime library for whatever I/O functions are needed.
9973 @item -memregs=@var{number}
9975 Specifies the number of memory-based pseudo-registers GCC will use
9976 during code generation. These pseudo-registers will be used like real
9977 registers, so there is a tradeoff between GCC's ability to fit the
9978 code into available registers, and the performance penalty of using
9979 memory instead of registers. Note that all modules in a program must
9980 be compiled with the same value for this option. Because of that, you
9981 must not use this option with the default runtime libraries gcc
9986 @node M32R/D Options
9987 @subsection M32R/D Options
9988 @cindex M32R/D options
9990 These @option{-m} options are defined for Renesas M32R/D architectures:
9995 Generate code for the M32R/2@.
9999 Generate code for the M32R/X@.
10003 Generate code for the M32R@. This is the default.
10005 @item -mmodel=small
10006 @opindex mmodel=small
10007 Assume all objects live in the lower 16MB of memory (so that their addresses
10008 can be loaded with the @code{ld24} instruction), and assume all subroutines
10009 are reachable with the @code{bl} instruction.
10010 This is the default.
10012 The addressability of a particular object can be set with the
10013 @code{model} attribute.
10015 @item -mmodel=medium
10016 @opindex mmodel=medium
10017 Assume objects may be anywhere in the 32-bit address space (the compiler
10018 will generate @code{seth/add3} instructions to load their addresses), and
10019 assume all subroutines are reachable with the @code{bl} instruction.
10021 @item -mmodel=large
10022 @opindex mmodel=large
10023 Assume objects may be anywhere in the 32-bit address space (the compiler
10024 will generate @code{seth/add3} instructions to load their addresses), and
10025 assume subroutines may not be reachable with the @code{bl} instruction
10026 (the compiler will generate the much slower @code{seth/add3/jl}
10027 instruction sequence).
10030 @opindex msdata=none
10031 Disable use of the small data area. Variables will be put into
10032 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
10033 @code{section} attribute has been specified).
10034 This is the default.
10036 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
10037 Objects may be explicitly put in the small data area with the
10038 @code{section} attribute using one of these sections.
10040 @item -msdata=sdata
10041 @opindex msdata=sdata
10042 Put small global and static data in the small data area, but do not
10043 generate special code to reference them.
10046 @opindex msdata=use
10047 Put small global and static data in the small data area, and generate
10048 special instructions to reference them.
10052 @cindex smaller data references
10053 Put global and static objects less than or equal to @var{num} bytes
10054 into the small data or bss sections instead of the normal data or bss
10055 sections. The default value of @var{num} is 8.
10056 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
10057 for this option to have any effect.
10059 All modules should be compiled with the same @option{-G @var{num}} value.
10060 Compiling with different values of @var{num} may or may not work; if it
10061 doesn't the linker will give an error message---incorrect code will not be
10066 Makes the M32R specific code in the compiler display some statistics
10067 that might help in debugging programs.
10069 @item -malign-loops
10070 @opindex malign-loops
10071 Align all loops to a 32-byte boundary.
10073 @item -mno-align-loops
10074 @opindex mno-align-loops
10075 Do not enforce a 32-byte alignment for loops. This is the default.
10077 @item -missue-rate=@var{number}
10078 @opindex missue-rate=@var{number}
10079 Issue @var{number} instructions per cycle. @var{number} can only be 1
10082 @item -mbranch-cost=@var{number}
10083 @opindex mbranch-cost=@var{number}
10084 @var{number} can only be 1 or 2. If it is 1 then branches will be
10085 preferred over conditional code, if it is 2, then the opposite will
10088 @item -mflush-trap=@var{number}
10089 @opindex mflush-trap=@var{number}
10090 Specifies the trap number to use to flush the cache. The default is
10091 12. Valid numbers are between 0 and 15 inclusive.
10093 @item -mno-flush-trap
10094 @opindex mno-flush-trap
10095 Specifies that the cache cannot be flushed by using a trap.
10097 @item -mflush-func=@var{name}
10098 @opindex mflush-func=@var{name}
10099 Specifies the name of the operating system function to call to flush
10100 the cache. The default is @emph{_flush_cache}, but a function call
10101 will only be used if a trap is not available.
10103 @item -mno-flush-func
10104 @opindex mno-flush-func
10105 Indicates that there is no OS function for flushing the cache.
10109 @node M680x0 Options
10110 @subsection M680x0 Options
10111 @cindex M680x0 options
10113 These are the @samp{-m} options defined for the 68000 series. The default
10114 values for these options depends on which style of 68000 was selected when
10115 the compiler was configured; the defaults for the most common choices are
10123 Generate output for a 68000. This is the default
10124 when the compiler is configured for 68000-based systems.
10126 Use this option for microcontrollers with a 68000 or EC000 core,
10127 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
10133 Generate output for a 68020. This is the default
10134 when the compiler is configured for 68020-based systems.
10138 Generate output containing 68881 instructions for floating point.
10139 This is the default for most 68020 systems unless @option{--nfp} was
10140 specified when the compiler was configured.
10144 Generate output for a 68030. This is the default when the compiler is
10145 configured for 68030-based systems.
10149 Generate output for a 68040. This is the default when the compiler is
10150 configured for 68040-based systems.
10152 This option inhibits the use of 68881/68882 instructions that have to be
10153 emulated by software on the 68040. Use this option if your 68040 does not
10154 have code to emulate those instructions.
10158 Generate output for a 68060. This is the default when the compiler is
10159 configured for 68060-based systems.
10161 This option inhibits the use of 68020 and 68881/68882 instructions that
10162 have to be emulated by software on the 68060. Use this option if your 68060
10163 does not have code to emulate those instructions.
10167 Generate output for a CPU32. This is the default
10168 when the compiler is configured for CPU32-based systems.
10170 Use this option for microcontrollers with a
10171 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
10172 68336, 68340, 68341, 68349 and 68360.
10176 Generate output for a 520X ``coldfire'' family cpu. This is the default
10177 when the compiler is configured for 520X-based systems.
10179 Use this option for microcontroller with a 5200 core, including
10180 the MCF5202, MCF5203, MCF5204 and MCF5202.
10184 Generate output for a ColdFire V4e family cpu (e.g.@: 547x/548x).
10185 This includes use of hardware floating point instructions.
10189 Generate output for a 68040, without using any of the new instructions.
10190 This results in code which can run relatively efficiently on either a
10191 68020/68881 or a 68030 or a 68040. The generated code does use the
10192 68881 instructions that are emulated on the 68040.
10196 Generate output for a 68060, without using any of the new instructions.
10197 This results in code which can run relatively efficiently on either a
10198 68020/68881 or a 68030 or a 68040. The generated code does use the
10199 68881 instructions that are emulated on the 68060.
10202 @opindex msoft-float
10203 Generate output containing library calls for floating point.
10204 @strong{Warning:} the requisite libraries are not available for all m68k
10205 targets. Normally the facilities of the machine's usual C compiler are
10206 used, but this can't be done directly in cross-compilation. You must
10207 make your own arrangements to provide suitable library functions for
10208 cross-compilation. The embedded targets @samp{m68k-*-aout} and
10209 @samp{m68k-*-coff} do provide software floating point support.
10213 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10214 Additionally, parameters passed on the stack are also aligned to a
10215 16-bit boundary even on targets whose API mandates promotion to 32-bit.
10218 @opindex mnobitfield
10219 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
10220 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
10224 Do use the bit-field instructions. The @option{-m68020} option implies
10225 @option{-mbitfield}. This is the default if you use a configuration
10226 designed for a 68020.
10230 Use a different function-calling convention, in which functions
10231 that take a fixed number of arguments return with the @code{rtd}
10232 instruction, which pops their arguments while returning. This
10233 saves one instruction in the caller since there is no need to pop
10234 the arguments there.
10236 This calling convention is incompatible with the one normally
10237 used on Unix, so you cannot use it if you need to call libraries
10238 compiled with the Unix compiler.
10240 Also, you must provide function prototypes for all functions that
10241 take variable numbers of arguments (including @code{printf});
10242 otherwise incorrect code will be generated for calls to those
10245 In addition, seriously incorrect code will result if you call a
10246 function with too many arguments. (Normally, extra arguments are
10247 harmlessly ignored.)
10249 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
10250 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
10253 @itemx -mno-align-int
10254 @opindex malign-int
10255 @opindex mno-align-int
10256 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
10257 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
10258 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
10259 Aligning variables on 32-bit boundaries produces code that runs somewhat
10260 faster on processors with 32-bit busses at the expense of more memory.
10262 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
10263 align structures containing the above types differently than
10264 most published application binary interface specifications for the m68k.
10268 Use the pc-relative addressing mode of the 68000 directly, instead of
10269 using a global offset table. At present, this option implies @option{-fpic},
10270 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
10271 not presently supported with @option{-mpcrel}, though this could be supported for
10272 68020 and higher processors.
10274 @item -mno-strict-align
10275 @itemx -mstrict-align
10276 @opindex mno-strict-align
10277 @opindex mstrict-align
10278 Do not (do) assume that unaligned memory references will be handled by
10282 Generate code that allows the data segment to be located in a different
10283 area of memory from the text segment. This allows for execute in place in
10284 an environment without virtual memory management. This option implies
10287 @item -mno-sep-data
10288 Generate code that assumes that the data segment follows the text segment.
10289 This is the default.
10291 @item -mid-shared-library
10292 Generate code that supports shared libraries via the library ID method.
10293 This allows for execute in place and shared libraries in an environment
10294 without virtual memory management. This option implies @option{-fPIC}.
10296 @item -mno-id-shared-library
10297 Generate code that doesn't assume ID based shared libraries are being used.
10298 This is the default.
10300 @item -mshared-library-id=n
10301 Specified the identification number of the ID based shared library being
10302 compiled. Specifying a value of 0 will generate more compact code, specifying
10303 other values will force the allocation of that number to the current
10304 library but is no more space or time efficient than omitting this option.
10308 @node M68hc1x Options
10309 @subsection M68hc1x Options
10310 @cindex M68hc1x options
10312 These are the @samp{-m} options defined for the 68hc11 and 68hc12
10313 microcontrollers. The default values for these options depends on
10314 which style of microcontroller was selected when the compiler was configured;
10315 the defaults for the most common choices are given below.
10322 Generate output for a 68HC11. This is the default
10323 when the compiler is configured for 68HC11-based systems.
10329 Generate output for a 68HC12. This is the default
10330 when the compiler is configured for 68HC12-based systems.
10336 Generate output for a 68HCS12.
10338 @item -mauto-incdec
10339 @opindex mauto-incdec
10340 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
10347 Enable the use of 68HC12 min and max instructions.
10350 @itemx -mno-long-calls
10351 @opindex mlong-calls
10352 @opindex mno-long-calls
10353 Treat all calls as being far away (near). If calls are assumed to be
10354 far away, the compiler will use the @code{call} instruction to
10355 call a function and the @code{rtc} instruction for returning.
10359 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10361 @item -msoft-reg-count=@var{count}
10362 @opindex msoft-reg-count
10363 Specify the number of pseudo-soft registers which are used for the
10364 code generation. The maximum number is 32. Using more pseudo-soft
10365 register may or may not result in better code depending on the program.
10366 The default is 4 for 68HC11 and 2 for 68HC12.
10370 @node MCore Options
10371 @subsection MCore Options
10372 @cindex MCore options
10374 These are the @samp{-m} options defined for the Motorola M*Core
10380 @itemx -mno-hardlit
10382 @opindex mno-hardlit
10383 Inline constants into the code stream if it can be done in two
10384 instructions or less.
10390 Use the divide instruction. (Enabled by default).
10392 @item -mrelax-immediate
10393 @itemx -mno-relax-immediate
10394 @opindex mrelax-immediate
10395 @opindex mno-relax-immediate
10396 Allow arbitrary sized immediates in bit operations.
10398 @item -mwide-bitfields
10399 @itemx -mno-wide-bitfields
10400 @opindex mwide-bitfields
10401 @opindex mno-wide-bitfields
10402 Always treat bit-fields as int-sized.
10404 @item -m4byte-functions
10405 @itemx -mno-4byte-functions
10406 @opindex m4byte-functions
10407 @opindex mno-4byte-functions
10408 Force all functions to be aligned to a four byte boundary.
10410 @item -mcallgraph-data
10411 @itemx -mno-callgraph-data
10412 @opindex mcallgraph-data
10413 @opindex mno-callgraph-data
10414 Emit callgraph information.
10417 @itemx -mno-slow-bytes
10418 @opindex mslow-bytes
10419 @opindex mno-slow-bytes
10420 Prefer word access when reading byte quantities.
10422 @item -mlittle-endian
10423 @itemx -mbig-endian
10424 @opindex mlittle-endian
10425 @opindex mbig-endian
10426 Generate code for a little endian target.
10432 Generate code for the 210 processor.
10436 @subsection MIPS Options
10437 @cindex MIPS options
10443 Generate big-endian code.
10447 Generate little-endian code. This is the default for @samp{mips*el-*-*}
10450 @item -march=@var{arch}
10452 Generate code that will run on @var{arch}, which can be the name of a
10453 generic MIPS ISA, or the name of a particular processor.
10455 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
10456 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
10457 The processor names are:
10458 @samp{4kc}, @samp{4km}, @samp{4kp},
10459 @samp{5kc}, @samp{5kf},
10461 @samp{24k}, @samp{24kc}, @samp{24kf}, @samp{24kx},
10464 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
10465 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
10466 @samp{rm7000}, @samp{rm9000},
10469 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
10470 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
10471 The special value @samp{from-abi} selects the
10472 most compatible architecture for the selected ABI (that is,
10473 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
10475 In processor names, a final @samp{000} can be abbreviated as @samp{k}
10476 (for example, @samp{-march=r2k}). Prefixes are optional, and
10477 @samp{vr} may be written @samp{r}.
10479 GCC defines two macros based on the value of this option. The first
10480 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
10481 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
10482 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
10483 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
10484 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
10486 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
10487 above. In other words, it will have the full prefix and will not
10488 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
10489 the macro names the resolved architecture (either @samp{"mips1"} or
10490 @samp{"mips3"}). It names the default architecture when no
10491 @option{-march} option is given.
10493 @item -mtune=@var{arch}
10495 Optimize for @var{arch}. Among other things, this option controls
10496 the way instructions are scheduled, and the perceived cost of arithmetic
10497 operations. The list of @var{arch} values is the same as for
10500 When this option is not used, GCC will optimize for the processor
10501 specified by @option{-march}. By using @option{-march} and
10502 @option{-mtune} together, it is possible to generate code that will
10503 run on a family of processors, but optimize the code for one
10504 particular member of that family.
10506 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
10507 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
10508 @samp{-march} ones described above.
10512 Equivalent to @samp{-march=mips1}.
10516 Equivalent to @samp{-march=mips2}.
10520 Equivalent to @samp{-march=mips3}.
10524 Equivalent to @samp{-march=mips4}.
10528 Equivalent to @samp{-march=mips32}.
10532 Equivalent to @samp{-march=mips32r2}.
10536 Equivalent to @samp{-march=mips64}.
10541 @opindex mno-mips16
10542 Generate (do not generate) MIPS16 code. If GCC is targetting a
10543 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
10555 Generate code for the given ABI@.
10557 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
10558 generates 64-bit code when you select a 64-bit architecture, but you
10559 can use @option{-mgp32} to get 32-bit code instead.
10561 For information about the O64 ABI, see
10562 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
10565 @itemx -mno-abicalls
10567 @opindex mno-abicalls
10568 Generate (do not generate) code that is suitable for SVR4-style
10569 dynamic objects. @option{-mabicalls} is the default for SVR4-based
10574 Generate (do not generate) code that is fully position-independent,
10575 and that can therefore be linked into shared libraries. This option
10576 only affects @option{-mabicalls}.
10578 All @option{-mabicalls} code has traditionally been position-independent,
10579 regardless of options like @option{-fPIC} and @option{-fpic}. However,
10580 as an extension, the GNU toolchain allows executables to use absolute
10581 accesses for locally-binding symbols. It can also use shorter GP
10582 initialization sequences and generate direct calls to locally-defined
10583 functions. This mode is selected by @option{-mno-shared}.
10585 @option{-mno-shared} depends on binutils 2.16 or higher and generates
10586 objects that can only be linked by the GNU linker. However, the option
10587 does not affect the ABI of the final executable; it only affects the ABI
10588 of relocatable objects. Using @option{-mno-shared} will generally make
10589 executables both smaller and quicker.
10591 @option{-mshared} is the default.
10597 Lift (do not lift) the usual restrictions on the size of the global
10600 GCC normally uses a single instruction to load values from the GOT@.
10601 While this is relatively efficient, it will only work if the GOT
10602 is smaller than about 64k. Anything larger will cause the linker
10603 to report an error such as:
10605 @cindex relocation truncated to fit (MIPS)
10607 relocation truncated to fit: R_MIPS_GOT16 foobar
10610 If this happens, you should recompile your code with @option{-mxgot}.
10611 It should then work with very large GOTs, although it will also be
10612 less efficient, since it will take three instructions to fetch the
10613 value of a global symbol.
10615 Note that some linkers can create multiple GOTs. If you have such a
10616 linker, you should only need to use @option{-mxgot} when a single object
10617 file accesses more than 64k's worth of GOT entries. Very few do.
10619 These options have no effect unless GCC is generating position
10624 Assume that general-purpose registers are 32 bits wide.
10628 Assume that general-purpose registers are 64 bits wide.
10632 Assume that floating-point registers are 32 bits wide.
10636 Assume that floating-point registers are 64 bits wide.
10639 @opindex mhard-float
10640 Use floating-point coprocessor instructions.
10643 @opindex msoft-float
10644 Do not use floating-point coprocessor instructions. Implement
10645 floating-point calculations using library calls instead.
10647 @item -msingle-float
10648 @opindex msingle-float
10649 Assume that the floating-point coprocessor only supports single-precision
10652 @itemx -mdouble-float
10653 @opindex mdouble-float
10654 Assume that the floating-point coprocessor supports double-precision
10655 operations. This is the default.
10661 Use (do not use) the MIPS DSP ASE. @xref{MIPS DSP Built-in Functions}.
10663 @itemx -mpaired-single
10664 @itemx -mno-paired-single
10665 @opindex mpaired-single
10666 @opindex mno-paired-single
10667 Use (do not use) paired-single floating-point instructions.
10668 @xref{MIPS Paired-Single Support}. This option can only be used
10669 when generating 64-bit code and requires hardware floating-point
10670 support to be enabled.
10675 @opindex mno-mips3d
10676 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
10677 The option @option{-mips3d} implies @option{-mpaired-single}.
10681 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
10682 an explanation of the default and the way that the pointer size is
10687 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
10689 The default size of @code{int}s, @code{long}s and pointers depends on
10690 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
10691 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
10692 32-bit @code{long}s. Pointers are the same size as @code{long}s,
10693 or the same size as integer registers, whichever is smaller.
10699 Assume (do not assume) that all symbols have 32-bit values, regardless
10700 of the selected ABI@. This option is useful in combination with
10701 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
10702 to generate shorter and faster references to symbolic addresses.
10706 @cindex smaller data references (MIPS)
10707 @cindex gp-relative references (MIPS)
10708 Put global and static items less than or equal to @var{num} bytes into
10709 the small data or bss section instead of the normal data or bss section.
10710 This allows the data to be accessed using a single instruction.
10712 All modules should be compiled with the same @option{-G @var{num}}
10715 @item -membedded-data
10716 @itemx -mno-embedded-data
10717 @opindex membedded-data
10718 @opindex mno-embedded-data
10719 Allocate variables to the read-only data section first if possible, then
10720 next in the small data section if possible, otherwise in data. This gives
10721 slightly slower code than the default, but reduces the amount of RAM required
10722 when executing, and thus may be preferred for some embedded systems.
10724 @item -muninit-const-in-rodata
10725 @itemx -mno-uninit-const-in-rodata
10726 @opindex muninit-const-in-rodata
10727 @opindex mno-uninit-const-in-rodata
10728 Put uninitialized @code{const} variables in the read-only data section.
10729 This option is only meaningful in conjunction with @option{-membedded-data}.
10731 @item -msplit-addresses
10732 @itemx -mno-split-addresses
10733 @opindex msplit-addresses
10734 @opindex mno-split-addresses
10735 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
10736 relocation operators. This option has been superseded by
10737 @option{-mexplicit-relocs} but is retained for backwards compatibility.
10739 @item -mexplicit-relocs
10740 @itemx -mno-explicit-relocs
10741 @opindex mexplicit-relocs
10742 @opindex mno-explicit-relocs
10743 Use (do not use) assembler relocation operators when dealing with symbolic
10744 addresses. The alternative, selected by @option{-mno-explicit-relocs},
10745 is to use assembler macros instead.
10747 @option{-mexplicit-relocs} is the default if GCC was configured
10748 to use an assembler that supports relocation operators.
10750 @item -mcheck-zero-division
10751 @itemx -mno-check-zero-division
10752 @opindex mcheck-zero-division
10753 @opindex mno-check-zero-division
10754 Trap (do not trap) on integer division by zero. The default is
10755 @option{-mcheck-zero-division}.
10757 @item -mdivide-traps
10758 @itemx -mdivide-breaks
10759 @opindex mdivide-traps
10760 @opindex mdivide-breaks
10761 MIPS systems check for division by zero by generating either a
10762 conditional trap or a break instruction. Using traps results in
10763 smaller code, but is only supported on MIPS II and later. Also, some
10764 versions of the Linux kernel have a bug that prevents trap from
10765 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
10766 allow conditional traps on architectures that support them and
10767 @option{-mdivide-breaks} to force the use of breaks.
10769 The default is usually @option{-mdivide-traps}, but this can be
10770 overridden at configure time using @option{--with-divide=breaks}.
10771 Divide-by-zero checks can be completely disabled using
10772 @option{-mno-check-zero-division}.
10777 @opindex mno-memcpy
10778 Force (do not force) the use of @code{memcpy()} for non-trivial block
10779 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
10780 most constant-sized copies.
10783 @itemx -mno-long-calls
10784 @opindex mlong-calls
10785 @opindex mno-long-calls
10786 Disable (do not disable) use of the @code{jal} instruction. Calling
10787 functions using @code{jal} is more efficient but requires the caller
10788 and callee to be in the same 256 megabyte segment.
10790 This option has no effect on abicalls code. The default is
10791 @option{-mno-long-calls}.
10797 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
10798 instructions, as provided by the R4650 ISA@.
10801 @itemx -mno-fused-madd
10802 @opindex mfused-madd
10803 @opindex mno-fused-madd
10804 Enable (disable) use of the floating point multiply-accumulate
10805 instructions, when they are available. The default is
10806 @option{-mfused-madd}.
10808 When multiply-accumulate instructions are used, the intermediate
10809 product is calculated to infinite precision and is not subject to
10810 the FCSR Flush to Zero bit. This may be undesirable in some
10815 Tell the MIPS assembler to not run its preprocessor over user
10816 assembler files (with a @samp{.s} suffix) when assembling them.
10819 @itemx -mno-fix-r4000
10820 @opindex mfix-r4000
10821 @opindex mno-fix-r4000
10822 Work around certain R4000 CPU errata:
10825 A double-word or a variable shift may give an incorrect result if executed
10826 immediately after starting an integer division.
10828 A double-word or a variable shift may give an incorrect result if executed
10829 while an integer multiplication is in progress.
10831 An integer division may give an incorrect result if started in a delay slot
10832 of a taken branch or a jump.
10836 @itemx -mno-fix-r4400
10837 @opindex mfix-r4400
10838 @opindex mno-fix-r4400
10839 Work around certain R4400 CPU errata:
10842 A double-word or a variable shift may give an incorrect result if executed
10843 immediately after starting an integer division.
10847 @itemx -mno-fix-vr4120
10848 @opindex mfix-vr4120
10849 Work around certain VR4120 errata:
10852 @code{dmultu} does not always produce the correct result.
10854 @code{div} and @code{ddiv} do not always produce the correct result if one
10855 of the operands is negative.
10857 The workarounds for the division errata rely on special functions in
10858 @file{libgcc.a}. At present, these functions are only provided by
10859 the @code{mips64vr*-elf} configurations.
10861 Other VR4120 errata require a nop to be inserted between certain pairs of
10862 instructions. These errata are handled by the assembler, not by GCC itself.
10865 @opindex mfix-vr4130
10866 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
10867 workarounds are implemented by the assembler rather than by GCC,
10868 although GCC will avoid using @code{mflo} and @code{mfhi} if the
10869 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
10870 instructions are available instead.
10873 @itemx -mno-fix-sb1
10875 Work around certain SB-1 CPU core errata.
10876 (This flag currently works around the SB-1 revision 2
10877 ``F1'' and ``F2'' floating point errata.)
10879 @item -mflush-func=@var{func}
10880 @itemx -mno-flush-func
10881 @opindex mflush-func
10882 Specifies the function to call to flush the I and D caches, or to not
10883 call any such function. If called, the function must take the same
10884 arguments as the common @code{_flush_func()}, that is, the address of the
10885 memory range for which the cache is being flushed, the size of the
10886 memory range, and the number 3 (to flush both caches). The default
10887 depends on the target GCC was configured for, but commonly is either
10888 @samp{_flush_func} or @samp{__cpu_flush}.
10890 @item -mbranch-likely
10891 @itemx -mno-branch-likely
10892 @opindex mbranch-likely
10893 @opindex mno-branch-likely
10894 Enable or disable use of Branch Likely instructions, regardless of the
10895 default for the selected architecture. By default, Branch Likely
10896 instructions may be generated if they are supported by the selected
10897 architecture. An exception is for the MIPS32 and MIPS64 architectures
10898 and processors which implement those architectures; for those, Branch
10899 Likely instructions will not be generated by default because the MIPS32
10900 and MIPS64 architectures specifically deprecate their use.
10902 @item -mfp-exceptions
10903 @itemx -mno-fp-exceptions
10904 @opindex mfp-exceptions
10905 Specifies whether FP exceptions are enabled. This affects how we schedule
10906 FP instructions for some processors. The default is that FP exceptions are
10909 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
10910 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
10913 @item -mvr4130-align
10914 @itemx -mno-vr4130-align
10915 @opindex mvr4130-align
10916 The VR4130 pipeline is two-way superscalar, but can only issue two
10917 instructions together if the first one is 8-byte aligned. When this
10918 option is enabled, GCC will align pairs of instructions that it
10919 thinks should execute in parallel.
10921 This option only has an effect when optimizing for the VR4130.
10922 It normally makes code faster, but at the expense of making it bigger.
10923 It is enabled by default at optimization level @option{-O3}.
10927 @subsection MMIX Options
10928 @cindex MMIX Options
10930 These options are defined for the MMIX:
10934 @itemx -mno-libfuncs
10936 @opindex mno-libfuncs
10937 Specify that intrinsic library functions are being compiled, passing all
10938 values in registers, no matter the size.
10941 @itemx -mno-epsilon
10943 @opindex mno-epsilon
10944 Generate floating-point comparison instructions that compare with respect
10945 to the @code{rE} epsilon register.
10947 @item -mabi=mmixware
10949 @opindex mabi-mmixware
10951 Generate code that passes function parameters and return values that (in
10952 the called function) are seen as registers @code{$0} and up, as opposed to
10953 the GNU ABI which uses global registers @code{$231} and up.
10955 @item -mzero-extend
10956 @itemx -mno-zero-extend
10957 @opindex mzero-extend
10958 @opindex mno-zero-extend
10959 When reading data from memory in sizes shorter than 64 bits, use (do not
10960 use) zero-extending load instructions by default, rather than
10961 sign-extending ones.
10964 @itemx -mno-knuthdiv
10966 @opindex mno-knuthdiv
10967 Make the result of a division yielding a remainder have the same sign as
10968 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10969 remainder follows the sign of the dividend. Both methods are
10970 arithmetically valid, the latter being almost exclusively used.
10972 @item -mtoplevel-symbols
10973 @itemx -mno-toplevel-symbols
10974 @opindex mtoplevel-symbols
10975 @opindex mno-toplevel-symbols
10976 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10977 code can be used with the @code{PREFIX} assembly directive.
10981 Generate an executable in the ELF format, rather than the default
10982 @samp{mmo} format used by the @command{mmix} simulator.
10984 @item -mbranch-predict
10985 @itemx -mno-branch-predict
10986 @opindex mbranch-predict
10987 @opindex mno-branch-predict
10988 Use (do not use) the probable-branch instructions, when static branch
10989 prediction indicates a probable branch.
10991 @item -mbase-addresses
10992 @itemx -mno-base-addresses
10993 @opindex mbase-addresses
10994 @opindex mno-base-addresses
10995 Generate (do not generate) code that uses @emph{base addresses}. Using a
10996 base address automatically generates a request (handled by the assembler
10997 and the linker) for a constant to be set up in a global register. The
10998 register is used for one or more base address requests within the range 0
10999 to 255 from the value held in the register. The generally leads to short
11000 and fast code, but the number of different data items that can be
11001 addressed is limited. This means that a program that uses lots of static
11002 data may require @option{-mno-base-addresses}.
11004 @item -msingle-exit
11005 @itemx -mno-single-exit
11006 @opindex msingle-exit
11007 @opindex mno-single-exit
11008 Force (do not force) generated code to have a single exit point in each
11012 @node MN10300 Options
11013 @subsection MN10300 Options
11014 @cindex MN10300 options
11016 These @option{-m} options are defined for Matsushita MN10300 architectures:
11021 Generate code to avoid bugs in the multiply instructions for the MN10300
11022 processors. This is the default.
11024 @item -mno-mult-bug
11025 @opindex mno-mult-bug
11026 Do not generate code to avoid bugs in the multiply instructions for the
11027 MN10300 processors.
11031 Generate code which uses features specific to the AM33 processor.
11035 Do not generate code which uses features specific to the AM33 processor. This
11038 @item -mreturn-pointer-on-d0
11039 @opindex mreturn-pointer-on-d0
11040 When generating a function which returns a pointer, return the pointer
11041 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
11042 only in a0, and attempts to call such functions without a prototype
11043 would result in errors. Note that this option is on by default; use
11044 @option{-mno-return-pointer-on-d0} to disable it.
11048 Do not link in the C run-time initialization object file.
11052 Indicate to the linker that it should perform a relaxation optimization pass
11053 to shorten branches, calls and absolute memory addresses. This option only
11054 has an effect when used on the command line for the final link step.
11056 This option makes symbolic debugging impossible.
11060 @subsection MT Options
11063 These @option{-m} options are defined for Morpho MT architectures:
11067 @item -march=@var{cpu-type}
11069 Generate code that will run on @var{cpu-type}, which is the name of a system
11070 representing a certain processor type. Possible values for
11071 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
11072 @samp{ms1-16-003} and @samp{ms2}.
11074 When this option is not used, the default is @option{-march=ms1-16-002}.
11078 Use byte loads and stores when generating code.
11082 Do not use byte loads and stores when generating code.
11086 Use simulator runtime
11090 Do not link in the C run-time initialization object file
11091 @file{crti.o}. Other run-time initialization and termination files
11092 such as @file{startup.o} and @file{exit.o} are still included on the
11093 linker command line.
11097 @node PDP-11 Options
11098 @subsection PDP-11 Options
11099 @cindex PDP-11 Options
11101 These options are defined for the PDP-11:
11106 Use hardware FPP floating point. This is the default. (FIS floating
11107 point on the PDP-11/40 is not supported.)
11110 @opindex msoft-float
11111 Do not use hardware floating point.
11115 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
11119 Return floating-point results in memory. This is the default.
11123 Generate code for a PDP-11/40.
11127 Generate code for a PDP-11/45. This is the default.
11131 Generate code for a PDP-11/10.
11133 @item -mbcopy-builtin
11134 @opindex bcopy-builtin
11135 Use inline @code{movmemhi} patterns for copying memory. This is the
11140 Do not use inline @code{movmemhi} patterns for copying memory.
11146 Use 16-bit @code{int}. This is the default.
11152 Use 32-bit @code{int}.
11155 @itemx -mno-float32
11157 @opindex mno-float32
11158 Use 64-bit @code{float}. This is the default.
11161 @itemx -mno-float64
11163 @opindex mno-float64
11164 Use 32-bit @code{float}.
11168 Use @code{abshi2} pattern. This is the default.
11172 Do not use @code{abshi2} pattern.
11174 @item -mbranch-expensive
11175 @opindex mbranch-expensive
11176 Pretend that branches are expensive. This is for experimenting with
11177 code generation only.
11179 @item -mbranch-cheap
11180 @opindex mbranch-cheap
11181 Do not pretend that branches are expensive. This is the default.
11185 Generate code for a system with split I&D@.
11189 Generate code for a system without split I&D@. This is the default.
11193 Use Unix assembler syntax. This is the default when configured for
11194 @samp{pdp11-*-bsd}.
11198 Use DEC assembler syntax. This is the default when configured for any
11199 PDP-11 target other than @samp{pdp11-*-bsd}.
11202 @node PowerPC Options
11203 @subsection PowerPC Options
11204 @cindex PowerPC options
11206 These are listed under @xref{RS/6000 and PowerPC Options}.
11208 @node RS/6000 and PowerPC Options
11209 @subsection IBM RS/6000 and PowerPC Options
11210 @cindex RS/6000 and PowerPC Options
11211 @cindex IBM RS/6000 and PowerPC Options
11213 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
11220 @itemx -mno-powerpc
11221 @itemx -mpowerpc-gpopt
11222 @itemx -mno-powerpc-gpopt
11223 @itemx -mpowerpc-gfxopt
11224 @itemx -mno-powerpc-gfxopt
11226 @itemx -mno-powerpc64
11230 @itemx -mno-popcntb
11236 @opindex mno-power2
11238 @opindex mno-powerpc
11239 @opindex mpowerpc-gpopt
11240 @opindex mno-powerpc-gpopt
11241 @opindex mpowerpc-gfxopt
11242 @opindex mno-powerpc-gfxopt
11243 @opindex mpowerpc64
11244 @opindex mno-powerpc64
11248 @opindex mno-popcntb
11251 GCC supports two related instruction set architectures for the
11252 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
11253 instructions supported by the @samp{rios} chip set used in the original
11254 RS/6000 systems and the @dfn{PowerPC} instruction set is the
11255 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
11256 the IBM 4xx, 6xx, and follow-on microprocessors.
11258 Neither architecture is a subset of the other. However there is a
11259 large common subset of instructions supported by both. An MQ
11260 register is included in processors supporting the POWER architecture.
11262 You use these options to specify which instructions are available on the
11263 processor you are using. The default value of these options is
11264 determined when configuring GCC@. Specifying the
11265 @option{-mcpu=@var{cpu_type}} overrides the specification of these
11266 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
11267 rather than the options listed above.
11269 The @option{-mpower} option allows GCC to generate instructions that
11270 are found only in the POWER architecture and to use the MQ register.
11271 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
11272 to generate instructions that are present in the POWER2 architecture but
11273 not the original POWER architecture.
11275 The @option{-mpowerpc} option allows GCC to generate instructions that
11276 are found only in the 32-bit subset of the PowerPC architecture.
11277 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
11278 GCC to use the optional PowerPC architecture instructions in the
11279 General Purpose group, including floating-point square root. Specifying
11280 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
11281 use the optional PowerPC architecture instructions in the Graphics
11282 group, including floating-point select.
11284 The @option{-mmfcrf} option allows GCC to generate the move from
11285 condition register field instruction implemented on the POWER4
11286 processor and other processors that support the PowerPC V2.01
11288 The @option{-mpopcntb} option allows GCC to generate the popcount and
11289 double precision FP reciprocal estimate instruction implemented on the
11290 POWER5 processor and other processors that support the PowerPC V2.02
11292 The @option{-mfprnd} option allows GCC to generate the FP round to
11293 integer instructions implemented on the POWER5+ processor and other
11294 processors that support the PowerPC V2.03 architecture.
11296 The @option{-mpowerpc64} option allows GCC to generate the additional
11297 64-bit instructions that are found in the full PowerPC64 architecture
11298 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
11299 @option{-mno-powerpc64}.
11301 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
11302 will use only the instructions in the common subset of both
11303 architectures plus some special AIX common-mode calls, and will not use
11304 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
11305 permits GCC to use any instruction from either architecture and to
11306 allow use of the MQ register; specify this for the Motorola MPC601.
11308 @item -mnew-mnemonics
11309 @itemx -mold-mnemonics
11310 @opindex mnew-mnemonics
11311 @opindex mold-mnemonics
11312 Select which mnemonics to use in the generated assembler code. With
11313 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
11314 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
11315 assembler mnemonics defined for the POWER architecture. Instructions
11316 defined in only one architecture have only one mnemonic; GCC uses that
11317 mnemonic irrespective of which of these options is specified.
11319 GCC defaults to the mnemonics appropriate for the architecture in
11320 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
11321 value of these option. Unless you are building a cross-compiler, you
11322 should normally not specify either @option{-mnew-mnemonics} or
11323 @option{-mold-mnemonics}, but should instead accept the default.
11325 @item -mcpu=@var{cpu_type}
11327 Set architecture type, register usage, choice of mnemonics, and
11328 instruction scheduling parameters for machine type @var{cpu_type}.
11329 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
11330 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
11331 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
11332 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
11333 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
11334 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
11335 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
11336 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6},
11337 @samp{common}, @samp{powerpc}, @samp{powerpc64},
11338 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
11340 @option{-mcpu=common} selects a completely generic processor. Code
11341 generated under this option will run on any POWER or PowerPC processor.
11342 GCC will use only the instructions in the common subset of both
11343 architectures, and will not use the MQ register. GCC assumes a generic
11344 processor model for scheduling purposes.
11346 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
11347 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
11348 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
11349 types, with an appropriate, generic processor model assumed for
11350 scheduling purposes.
11352 The other options specify a specific processor. Code generated under
11353 those options will run best on that processor, and may not run at all on
11356 The @option{-mcpu} options automatically enable or disable the
11357 following options: @option{-maltivec}, @option{-mfprnd},
11358 @option{-mhard-float}, @option{-mmfcrf}, @option{-mmultiple},
11359 @option{-mnew-mnemonics}, @option{-mpopcntb}, @option{-mpower},
11360 @option{-mpower2}, @option{-mpowerpc64}, @option{-mpowerpc-gpopt},
11361 @option{-mpowerpc-gfxopt}, @option{-mstring}, @option{-mmulhw}, @option{-mdlmzb}.
11362 The particular options
11363 set for any particular CPU will vary between compiler versions,
11364 depending on what setting seems to produce optimal code for that CPU;
11365 it doesn't necessarily reflect the actual hardware's capabilities. If
11366 you wish to set an individual option to a particular value, you may
11367 specify it after the @option{-mcpu} option, like @samp{-mcpu=970
11370 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
11371 not enabled or disabled by the @option{-mcpu} option at present because
11372 AIX does not have full support for these options. You may still
11373 enable or disable them individually if you're sure it'll work in your
11376 @item -mtune=@var{cpu_type}
11378 Set the instruction scheduling parameters for machine type
11379 @var{cpu_type}, but do not set the architecture type, register usage, or
11380 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
11381 values for @var{cpu_type} are used for @option{-mtune} as for
11382 @option{-mcpu}. If both are specified, the code generated will use the
11383 architecture, registers, and mnemonics set by @option{-mcpu}, but the
11384 scheduling parameters set by @option{-mtune}.
11390 Generate code to compute division as reciprocal estimate and iterative
11391 refinement, creating opportunities for increased throughput. This
11392 feature requires: optional PowerPC Graphics instruction set for single
11393 precision and FRE instruction for double precision, assuming divides
11394 cannot generate user-visible traps, and the domain values not include
11395 Infinities, denormals or zero denominator.
11398 @itemx -mno-altivec
11400 @opindex mno-altivec
11401 Generate code that uses (does not use) AltiVec instructions, and also
11402 enable the use of built-in functions that allow more direct access to
11403 the AltiVec instruction set. You may also need to set
11404 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
11410 @opindex mno-vrsave
11411 Generate VRSAVE instructions when generating AltiVec code.
11414 @opindex msecure-plt
11415 Generate code that allows ld and ld.so to build executables and shared
11416 libraries with non-exec .plt and .got sections. This is a PowerPC
11417 32-bit SYSV ABI option.
11421 Generate code that uses a BSS .plt section that ld.so fills in, and
11422 requires .plt and .got sections that are both writable and executable.
11423 This is a PowerPC 32-bit SYSV ABI option.
11429 This switch enables or disables the generation of ISEL instructions.
11431 @item -misel=@var{yes/no}
11432 This switch has been deprecated. Use @option{-misel} and
11433 @option{-mno-isel} instead.
11439 This switch enables or disables the generation of SPE simd
11442 @item -mspe=@var{yes/no}
11443 This option has been deprecated. Use @option{-mspe} and
11444 @option{-mno-spe} instead.
11446 @item -mfloat-gprs=@var{yes/single/double/no}
11447 @itemx -mfloat-gprs
11448 @opindex mfloat-gprs
11449 This switch enables or disables the generation of floating point
11450 operations on the general purpose registers for architectures that
11453 The argument @var{yes} or @var{single} enables the use of
11454 single-precision floating point operations.
11456 The argument @var{double} enables the use of single and
11457 double-precision floating point operations.
11459 The argument @var{no} disables floating point operations on the
11460 general purpose registers.
11462 This option is currently only available on the MPC854x.
11468 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
11469 targets (including GNU/Linux). The 32-bit environment sets int, long
11470 and pointer to 32 bits and generates code that runs on any PowerPC
11471 variant. The 64-bit environment sets int to 32 bits and long and
11472 pointer to 64 bits, and generates code for PowerPC64, as for
11473 @option{-mpowerpc64}.
11476 @itemx -mno-fp-in-toc
11477 @itemx -mno-sum-in-toc
11478 @itemx -mminimal-toc
11480 @opindex mno-fp-in-toc
11481 @opindex mno-sum-in-toc
11482 @opindex mminimal-toc
11483 Modify generation of the TOC (Table Of Contents), which is created for
11484 every executable file. The @option{-mfull-toc} option is selected by
11485 default. In that case, GCC will allocate at least one TOC entry for
11486 each unique non-automatic variable reference in your program. GCC
11487 will also place floating-point constants in the TOC@. However, only
11488 16,384 entries are available in the TOC@.
11490 If you receive a linker error message that saying you have overflowed
11491 the available TOC space, you can reduce the amount of TOC space used
11492 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
11493 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
11494 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
11495 generate code to calculate the sum of an address and a constant at
11496 run-time instead of putting that sum into the TOC@. You may specify one
11497 or both of these options. Each causes GCC to produce very slightly
11498 slower and larger code at the expense of conserving TOC space.
11500 If you still run out of space in the TOC even when you specify both of
11501 these options, specify @option{-mminimal-toc} instead. This option causes
11502 GCC to make only one TOC entry for every file. When you specify this
11503 option, GCC will produce code that is slower and larger but which
11504 uses extremely little TOC space. You may wish to use this option
11505 only on files that contain less frequently executed code.
11511 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
11512 @code{long} type, and the infrastructure needed to support them.
11513 Specifying @option{-maix64} implies @option{-mpowerpc64} and
11514 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
11515 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
11518 @itemx -mno-xl-compat
11519 @opindex mxl-compat
11520 @opindex mno-xl-compat
11521 Produce code that conforms more closely to IBM XL compiler semantics
11522 when using AIX-compatible ABI. Pass floating-point arguments to
11523 prototyped functions beyond the register save area (RSA) on the stack
11524 in addition to argument FPRs. Do not assume that most significant
11525 double in 128-bit long double value is properly rounded when comparing
11526 values and converting to double. Use XL symbol names for long double
11529 The AIX calling convention was extended but not initially documented to
11530 handle an obscure K&R C case of calling a function that takes the
11531 address of its arguments with fewer arguments than declared. IBM XL
11532 compilers access floating point arguments which do not fit in the
11533 RSA from the stack when a subroutine is compiled without
11534 optimization. Because always storing floating-point arguments on the
11535 stack is inefficient and rarely needed, this option is not enabled by
11536 default and only is necessary when calling subroutines compiled by IBM
11537 XL compilers without optimization.
11541 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
11542 application written to use message passing with special startup code to
11543 enable the application to run. The system must have PE installed in the
11544 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
11545 must be overridden with the @option{-specs=} option to specify the
11546 appropriate directory location. The Parallel Environment does not
11547 support threads, so the @option{-mpe} option and the @option{-pthread}
11548 option are incompatible.
11550 @item -malign-natural
11551 @itemx -malign-power
11552 @opindex malign-natural
11553 @opindex malign-power
11554 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
11555 @option{-malign-natural} overrides the ABI-defined alignment of larger
11556 types, such as floating-point doubles, on their natural size-based boundary.
11557 The option @option{-malign-power} instructs GCC to follow the ABI-specified
11558 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
11560 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
11564 @itemx -mhard-float
11565 @opindex msoft-float
11566 @opindex mhard-float
11567 Generate code that does not use (uses) the floating-point register set.
11568 Software floating point emulation is provided if you use the
11569 @option{-msoft-float} option, and pass the option to GCC when linking.
11572 @itemx -mno-multiple
11574 @opindex mno-multiple
11575 Generate code that uses (does not use) the load multiple word
11576 instructions and the store multiple word instructions. These
11577 instructions are generated by default on POWER systems, and not
11578 generated on PowerPC systems. Do not use @option{-mmultiple} on little
11579 endian PowerPC systems, since those instructions do not work when the
11580 processor is in little endian mode. The exceptions are PPC740 and
11581 PPC750 which permit the instructions usage in little endian mode.
11586 @opindex mno-string
11587 Generate code that uses (does not use) the load string instructions
11588 and the store string word instructions to save multiple registers and
11589 do small block moves. These instructions are generated by default on
11590 POWER systems, and not generated on PowerPC systems. Do not use
11591 @option{-mstring} on little endian PowerPC systems, since those
11592 instructions do not work when the processor is in little endian mode.
11593 The exceptions are PPC740 and PPC750 which permit the instructions
11594 usage in little endian mode.
11599 @opindex mno-update
11600 Generate code that uses (does not use) the load or store instructions
11601 that update the base register to the address of the calculated memory
11602 location. These instructions are generated by default. If you use
11603 @option{-mno-update}, there is a small window between the time that the
11604 stack pointer is updated and the address of the previous frame is
11605 stored, which means code that walks the stack frame across interrupts or
11606 signals may get corrupted data.
11609 @itemx -mno-fused-madd
11610 @opindex mfused-madd
11611 @opindex mno-fused-madd
11612 Generate code that uses (does not use) the floating point multiply and
11613 accumulate instructions. These instructions are generated by default if
11614 hardware floating is used.
11620 Generate code that uses (does not use) the half-word multiply and
11621 multiply-accumulate instructions on the IBM 405 and 440 processors.
11622 These instructions are generated by default when targetting those
11629 Generate code that uses (does not use) the string-search @samp{dlmzb}
11630 instruction on the IBM 405 and 440 processors. This instruction is
11631 generated by default when targetting those processors.
11633 @item -mno-bit-align
11635 @opindex mno-bit-align
11636 @opindex mbit-align
11637 On System V.4 and embedded PowerPC systems do not (do) force structures
11638 and unions that contain bit-fields to be aligned to the base type of the
11641 For example, by default a structure containing nothing but 8
11642 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
11643 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
11644 the structure would be aligned to a 1 byte boundary and be one byte in
11647 @item -mno-strict-align
11648 @itemx -mstrict-align
11649 @opindex mno-strict-align
11650 @opindex mstrict-align
11651 On System V.4 and embedded PowerPC systems do not (do) assume that
11652 unaligned memory references will be handled by the system.
11654 @item -mrelocatable
11655 @itemx -mno-relocatable
11656 @opindex mrelocatable
11657 @opindex mno-relocatable
11658 On embedded PowerPC systems generate code that allows (does not allow)
11659 the program to be relocated to a different address at runtime. If you
11660 use @option{-mrelocatable} on any module, all objects linked together must
11661 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
11663 @item -mrelocatable-lib
11664 @itemx -mno-relocatable-lib
11665 @opindex mrelocatable-lib
11666 @opindex mno-relocatable-lib
11667 On embedded PowerPC systems generate code that allows (does not allow)
11668 the program to be relocated to a different address at runtime. Modules
11669 compiled with @option{-mrelocatable-lib} can be linked with either modules
11670 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
11671 with modules compiled with the @option{-mrelocatable} options.
11677 On System V.4 and embedded PowerPC systems do not (do) assume that
11678 register 2 contains a pointer to a global area pointing to the addresses
11679 used in the program.
11682 @itemx -mlittle-endian
11684 @opindex mlittle-endian
11685 On System V.4 and embedded PowerPC systems compile code for the
11686 processor in little endian mode. The @option{-mlittle-endian} option is
11687 the same as @option{-mlittle}.
11690 @itemx -mbig-endian
11692 @opindex mbig-endian
11693 On System V.4 and embedded PowerPC systems compile code for the
11694 processor in big endian mode. The @option{-mbig-endian} option is
11695 the same as @option{-mbig}.
11697 @item -mdynamic-no-pic
11698 @opindex mdynamic-no-pic
11699 On Darwin and Mac OS X systems, compile code so that it is not
11700 relocatable, but that its external references are relocatable. The
11701 resulting code is suitable for applications, but not shared
11704 @item -mprioritize-restricted-insns=@var{priority}
11705 @opindex mprioritize-restricted-insns
11706 This option controls the priority that is assigned to
11707 dispatch-slot restricted instructions during the second scheduling
11708 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
11709 @var{no/highest/second-highest} priority to dispatch slot restricted
11712 @item -msched-costly-dep=@var{dependence_type}
11713 @opindex msched-costly-dep
11714 This option controls which dependences are considered costly
11715 by the target during instruction scheduling. The argument
11716 @var{dependence_type} takes one of the following values:
11717 @var{no}: no dependence is costly,
11718 @var{all}: all dependences are costly,
11719 @var{true_store_to_load}: a true dependence from store to load is costly,
11720 @var{store_to_load}: any dependence from store to load is costly,
11721 @var{number}: any dependence which latency >= @var{number} is costly.
11723 @item -minsert-sched-nops=@var{scheme}
11724 @opindex minsert-sched-nops
11725 This option controls which nop insertion scheme will be used during
11726 the second scheduling pass. The argument @var{scheme} takes one of the
11728 @var{no}: Don't insert nops.
11729 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
11730 according to the scheduler's grouping.
11731 @var{regroup_exact}: Insert nops to force costly dependent insns into
11732 separate groups. Insert exactly as many nops as needed to force an insn
11733 to a new group, according to the estimated processor grouping.
11734 @var{number}: Insert nops to force costly dependent insns into
11735 separate groups. Insert @var{number} nops to force an insn to a new group.
11738 @opindex mcall-sysv
11739 On System V.4 and embedded PowerPC systems compile code using calling
11740 conventions that adheres to the March 1995 draft of the System V
11741 Application Binary Interface, PowerPC processor supplement. This is the
11742 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
11744 @item -mcall-sysv-eabi
11745 @opindex mcall-sysv-eabi
11746 Specify both @option{-mcall-sysv} and @option{-meabi} options.
11748 @item -mcall-sysv-noeabi
11749 @opindex mcall-sysv-noeabi
11750 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
11752 @item -mcall-solaris
11753 @opindex mcall-solaris
11754 On System V.4 and embedded PowerPC systems compile code for the Solaris
11758 @opindex mcall-linux
11759 On System V.4 and embedded PowerPC systems compile code for the
11760 Linux-based GNU system.
11764 On System V.4 and embedded PowerPC systems compile code for the
11765 Hurd-based GNU system.
11767 @item -mcall-netbsd
11768 @opindex mcall-netbsd
11769 On System V.4 and embedded PowerPC systems compile code for the
11770 NetBSD operating system.
11772 @item -maix-struct-return
11773 @opindex maix-struct-return
11774 Return all structures in memory (as specified by the AIX ABI)@.
11776 @item -msvr4-struct-return
11777 @opindex msvr4-struct-return
11778 Return structures smaller than 8 bytes in registers (as specified by the
11781 @item -mabi=@var{abi-type}
11783 Extend the current ABI with a particular extension, or remove such extension.
11784 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
11785 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
11789 Extend the current ABI with SPE ABI extensions. This does not change
11790 the default ABI, instead it adds the SPE ABI extensions to the current
11794 @opindex mabi=no-spe
11795 Disable Booke SPE ABI extensions for the current ABI@.
11797 @item -mabi=ibmlongdouble
11798 @opindex mabi=ibmlongdouble
11799 Change the current ABI to use IBM extended precision long double.
11800 This is a PowerPC 32-bit SYSV ABI option.
11802 @item -mabi=ieeelongdouble
11803 @opindex mabi=ieeelongdouble
11804 Change the current ABI to use IEEE extended precision long double.
11805 This is a PowerPC 32-bit Linux ABI option.
11808 @itemx -mno-prototype
11809 @opindex mprototype
11810 @opindex mno-prototype
11811 On System V.4 and embedded PowerPC systems assume that all calls to
11812 variable argument functions are properly prototyped. Otherwise, the
11813 compiler must insert an instruction before every non prototyped call to
11814 set or clear bit 6 of the condition code register (@var{CR}) to
11815 indicate whether floating point values were passed in the floating point
11816 registers in case the function takes a variable arguments. With
11817 @option{-mprototype}, only calls to prototyped variable argument functions
11818 will set or clear the bit.
11822 On embedded PowerPC systems, assume that the startup module is called
11823 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
11824 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
11829 On embedded PowerPC systems, assume that the startup module is called
11830 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
11835 On embedded PowerPC systems, assume that the startup module is called
11836 @file{crt0.o} and the standard C libraries are @file{libads.a} and
11839 @item -myellowknife
11840 @opindex myellowknife
11841 On embedded PowerPC systems, assume that the startup module is called
11842 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
11847 On System V.4 and embedded PowerPC systems, specify that you are
11848 compiling for a VxWorks system.
11852 Specify that you are compiling for the WindISS simulation environment.
11856 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
11857 header to indicate that @samp{eabi} extended relocations are used.
11863 On System V.4 and embedded PowerPC systems do (do not) adhere to the
11864 Embedded Applications Binary Interface (eabi) which is a set of
11865 modifications to the System V.4 specifications. Selecting @option{-meabi}
11866 means that the stack is aligned to an 8 byte boundary, a function
11867 @code{__eabi} is called to from @code{main} to set up the eabi
11868 environment, and the @option{-msdata} option can use both @code{r2} and
11869 @code{r13} to point to two separate small data areas. Selecting
11870 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
11871 do not call an initialization function from @code{main}, and the
11872 @option{-msdata} option will only use @code{r13} to point to a single
11873 small data area. The @option{-meabi} option is on by default if you
11874 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
11877 @opindex msdata=eabi
11878 On System V.4 and embedded PowerPC systems, put small initialized
11879 @code{const} global and static data in the @samp{.sdata2} section, which
11880 is pointed to by register @code{r2}. Put small initialized
11881 non-@code{const} global and static data in the @samp{.sdata} section,
11882 which is pointed to by register @code{r13}. Put small uninitialized
11883 global and static data in the @samp{.sbss} section, which is adjacent to
11884 the @samp{.sdata} section. The @option{-msdata=eabi} option is
11885 incompatible with the @option{-mrelocatable} option. The
11886 @option{-msdata=eabi} option also sets the @option{-memb} option.
11889 @opindex msdata=sysv
11890 On System V.4 and embedded PowerPC systems, put small global and static
11891 data in the @samp{.sdata} section, which is pointed to by register
11892 @code{r13}. Put small uninitialized global and static data in the
11893 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11894 The @option{-msdata=sysv} option is incompatible with the
11895 @option{-mrelocatable} option.
11897 @item -msdata=default
11899 @opindex msdata=default
11901 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
11902 compile code the same as @option{-msdata=eabi}, otherwise compile code the
11903 same as @option{-msdata=sysv}.
11906 @opindex msdata-data
11907 On System V.4 and embedded PowerPC systems, put small global
11908 data in the @samp{.sdata} section. Put small uninitialized global
11909 data in the @samp{.sbss} section. Do not use register @code{r13}
11910 to address small data however. This is the default behavior unless
11911 other @option{-msdata} options are used.
11915 @opindex msdata=none
11917 On embedded PowerPC systems, put all initialized global and static data
11918 in the @samp{.data} section, and all uninitialized data in the
11919 @samp{.bss} section.
11923 @cindex smaller data references (PowerPC)
11924 @cindex .sdata/.sdata2 references (PowerPC)
11925 On embedded PowerPC systems, put global and static items less than or
11926 equal to @var{num} bytes into the small data or bss sections instead of
11927 the normal data or bss section. By default, @var{num} is 8. The
11928 @option{-G @var{num}} switch is also passed to the linker.
11929 All modules should be compiled with the same @option{-G @var{num}} value.
11932 @itemx -mno-regnames
11934 @opindex mno-regnames
11935 On System V.4 and embedded PowerPC systems do (do not) emit register
11936 names in the assembly language output using symbolic forms.
11939 @itemx -mno-longcall
11941 @opindex mno-longcall
11942 By default assume that all calls are far away so that a longer more
11943 expensive calling sequence is required. This is required for calls
11944 further than 32 megabytes (33,554,432 bytes) from the current location.
11945 A short call will be generated if the compiler knows
11946 the call cannot be that far away. This setting can be overridden by
11947 the @code{shortcall} function attribute, or by @code{#pragma
11950 Some linkers are capable of detecting out-of-range calls and generating
11951 glue code on the fly. On these systems, long calls are unnecessary and
11952 generate slower code. As of this writing, the AIX linker can do this,
11953 as can the GNU linker for PowerPC/64. It is planned to add this feature
11954 to the GNU linker for 32-bit PowerPC systems as well.
11956 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11957 callee, L42'', plus a ``branch island'' (glue code). The two target
11958 addresses represent the callee and the ``branch island''. The
11959 Darwin/PPC linker will prefer the first address and generate a ``bl
11960 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11961 otherwise, the linker will generate ``bl L42'' to call the ``branch
11962 island''. The ``branch island'' is appended to the body of the
11963 calling function; it computes the full 32-bit address of the callee
11966 On Mach-O (Darwin) systems, this option directs the compiler emit to
11967 the glue for every direct call, and the Darwin linker decides whether
11968 to use or discard it.
11970 In the future, we may cause GCC to ignore all longcall specifications
11971 when the linker is known to generate glue.
11975 Adds support for multithreading with the @dfn{pthreads} library.
11976 This option sets flags for both the preprocessor and linker.
11980 @node S/390 and zSeries Options
11981 @subsection S/390 and zSeries Options
11982 @cindex S/390 and zSeries Options
11984 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11988 @itemx -msoft-float
11989 @opindex mhard-float
11990 @opindex msoft-float
11991 Use (do not use) the hardware floating-point instructions and registers
11992 for floating-point operations. When @option{-msoft-float} is specified,
11993 functions in @file{libgcc.a} will be used to perform floating-point
11994 operations. When @option{-mhard-float} is specified, the compiler
11995 generates IEEE floating-point instructions. This is the default.
11997 @item -mlong-double-64
11998 @itemx -mlong-double-128
11999 @opindex mlong-double-64
12000 @opindex mlong-double-128
12001 These switches control the size of @code{long double} type. A size
12002 of 64bit makes the @code{long double} type equivalent to the @code{double}
12003 type. This is the default.
12006 @itemx -mno-backchain
12007 @opindex mbackchain
12008 @opindex mno-backchain
12009 Store (do not store) the address of the caller's frame as backchain pointer
12010 into the callee's stack frame.
12011 A backchain may be needed to allow debugging using tools that do not understand
12012 DWARF-2 call frame information.
12013 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
12014 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
12015 the backchain is placed into the topmost word of the 96/160 byte register
12018 In general, code compiled with @option{-mbackchain} is call-compatible with
12019 code compiled with @option{-mmo-backchain}; however, use of the backchain
12020 for debugging purposes usually requires that the whole binary is built with
12021 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
12022 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12023 to build a linux kernel use @option{-msoft-float}.
12025 The default is to not maintain the backchain.
12027 @item -mpacked-stack
12028 @item -mno-packed-stack
12029 @opindex mpacked-stack
12030 @opindex mno-packed-stack
12031 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
12032 specified, the compiler uses the all fields of the 96/160 byte register save
12033 area only for their default purpose; unused fields still take up stack space.
12034 When @option{-mpacked-stack} is specified, register save slots are densely
12035 packed at the top of the register save area; unused space is reused for other
12036 purposes, allowing for more efficient use of the available stack space.
12037 However, when @option{-mbackchain} is also in effect, the topmost word of
12038 the save area is always used to store the backchain, and the return address
12039 register is always saved two words below the backchain.
12041 As long as the stack frame backchain is not used, code generated with
12042 @option{-mpacked-stack} is call-compatible with code generated with
12043 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
12044 S/390 or zSeries generated code that uses the stack frame backchain at run
12045 time, not just for debugging purposes. Such code is not call-compatible
12046 with code compiled with @option{-mpacked-stack}. Also, note that the
12047 combination of @option{-mbackchain},
12048 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12049 to build a linux kernel use @option{-msoft-float}.
12051 The default is to not use the packed stack layout.
12054 @itemx -mno-small-exec
12055 @opindex msmall-exec
12056 @opindex mno-small-exec
12057 Generate (or do not generate) code using the @code{bras} instruction
12058 to do subroutine calls.
12059 This only works reliably if the total executable size does not
12060 exceed 64k. The default is to use the @code{basr} instruction instead,
12061 which does not have this limitation.
12067 When @option{-m31} is specified, generate code compliant to the
12068 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
12069 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
12070 particular to generate 64-bit instructions. For the @samp{s390}
12071 targets, the default is @option{-m31}, while the @samp{s390x}
12072 targets default to @option{-m64}.
12078 When @option{-mzarch} is specified, generate code using the
12079 instructions available on z/Architecture.
12080 When @option{-mesa} is specified, generate code using the
12081 instructions available on ESA/390. Note that @option{-mesa} is
12082 not possible with @option{-m64}.
12083 When generating code compliant to the GNU/Linux for S/390 ABI,
12084 the default is @option{-mesa}. When generating code compliant
12085 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
12091 Generate (or do not generate) code using the @code{mvcle} instruction
12092 to perform block moves. When @option{-mno-mvcle} is specified,
12093 use a @code{mvc} loop instead. This is the default unless optimizing for
12100 Print (or do not print) additional debug information when compiling.
12101 The default is to not print debug information.
12103 @item -march=@var{cpu-type}
12105 Generate code that will run on @var{cpu-type}, which is the name of a system
12106 representing a certain processor type. Possible values for
12107 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
12108 When generating code using the instructions available on z/Architecture,
12109 the default is @option{-march=z900}. Otherwise, the default is
12110 @option{-march=g5}.
12112 @item -mtune=@var{cpu-type}
12114 Tune to @var{cpu-type} everything applicable about the generated code,
12115 except for the ABI and the set of available instructions.
12116 The list of @var{cpu-type} values is the same as for @option{-march}.
12117 The default is the value used for @option{-march}.
12120 @itemx -mno-tpf-trace
12121 @opindex mtpf-trace
12122 @opindex mno-tpf-trace
12123 Generate code that adds (does not add) in TPF OS specific branches to trace
12124 routines in the operating system. This option is off by default, even
12125 when compiling for the TPF OS@.
12128 @itemx -mno-fused-madd
12129 @opindex mfused-madd
12130 @opindex mno-fused-madd
12131 Generate code that uses (does not use) the floating point multiply and
12132 accumulate instructions. These instructions are generated by default if
12133 hardware floating point is used.
12135 @item -mwarn-framesize=@var{framesize}
12136 @opindex mwarn-framesize
12137 Emit a warning if the current function exceeds the given frame size. Because
12138 this is a compile time check it doesn't need to be a real problem when the program
12139 runs. It is intended to identify functions which most probably cause
12140 a stack overflow. It is useful to be used in an environment with limited stack
12141 size e.g.@: the linux kernel.
12143 @item -mwarn-dynamicstack
12144 @opindex mwarn-dynamicstack
12145 Emit a warning if the function calls alloca or uses dynamically
12146 sized arrays. This is generally a bad idea with a limited stack size.
12148 @item -mstack-guard=@var{stack-guard}
12149 @item -mstack-size=@var{stack-size}
12150 @opindex mstack-guard
12151 @opindex mstack-size
12152 These arguments always have to be used in conjunction. If they are present the s390
12153 back end emits additional instructions in the function prologue which trigger a trap
12154 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
12155 (remember that the stack on s390 grows downward). These options are intended to
12156 be used to help debugging stack overflow problems. The additionally emitted code
12157 causes only little overhead and hence can also be used in production like systems
12158 without greater performance degradation. The given values have to be exact
12159 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard} without
12161 In order to be efficient the extra code makes the assumption that the stack starts
12162 at an address aligned to the value given by @var{stack-size}.
12165 @node Score Options
12166 @subsection Score Options
12167 @cindex Score Options
12169 These options are defined for Score implementations:
12174 Compile code for big endian mode. This is the default.
12178 Compile code for little endian mode.
12182 Disable generate bcnz instruction.
12186 Enable generate unaligned load and store instruction.
12190 Enable the use of multiply-accumulate instructions. Disabled by default.
12194 Specify the SCORE5 as the target architecture.
12198 Specify the SCORE5U of the target architecture.
12202 Specify the SCORE7 as the target architecture. This is the default.
12206 Specify the SCORE7D as the target architecture.
12210 @subsection SH Options
12212 These @samp{-m} options are defined for the SH implementations:
12217 Generate code for the SH1.
12221 Generate code for the SH2.
12224 Generate code for the SH2e.
12228 Generate code for the SH3.
12232 Generate code for the SH3e.
12236 Generate code for the SH4 without a floating-point unit.
12238 @item -m4-single-only
12239 @opindex m4-single-only
12240 Generate code for the SH4 with a floating-point unit that only
12241 supports single-precision arithmetic.
12245 Generate code for the SH4 assuming the floating-point unit is in
12246 single-precision mode by default.
12250 Generate code for the SH4.
12254 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
12255 floating-point unit is not used.
12257 @item -m4a-single-only
12258 @opindex m4a-single-only
12259 Generate code for the SH4a, in such a way that no double-precision
12260 floating point operations are used.
12263 @opindex m4a-single
12264 Generate code for the SH4a assuming the floating-point unit is in
12265 single-precision mode by default.
12269 Generate code for the SH4a.
12273 Same as @option{-m4a-nofpu}, except that it implicitly passes
12274 @option{-dsp} to the assembler. GCC doesn't generate any DSP
12275 instructions at the moment.
12279 Compile code for the processor in big endian mode.
12283 Compile code for the processor in little endian mode.
12287 Align doubles at 64-bit boundaries. Note that this changes the calling
12288 conventions, and thus some functions from the standard C library will
12289 not work unless you recompile it first with @option{-mdalign}.
12293 Shorten some address references at link time, when possible; uses the
12294 linker option @option{-relax}.
12298 Use 32-bit offsets in @code{switch} tables. The default is to use
12303 Enable the use of the instruction @code{fmovd}.
12307 Comply with the calling conventions defined by Renesas.
12311 Comply with the calling conventions defined by Renesas.
12315 Comply with the calling conventions defined for GCC before the Renesas
12316 conventions were available. This option is the default for all
12317 targets of the SH toolchain except for @samp{sh-symbianelf}.
12320 @opindex mnomacsave
12321 Mark the @code{MAC} register as call-clobbered, even if
12322 @option{-mhitachi} is given.
12326 Increase IEEE-compliance of floating-point code.
12327 At the moment, this is equivalent to @option{-fno-finite-math-only}.
12328 When generating 16 bit SH opcodes, getting IEEE-conforming results for
12329 comparisons of NANs / infinities incurs extra overhead in every
12330 floating point comparison, therefore the default is set to
12331 @option{-ffinite-math-only}.
12335 Dump instruction size and location in the assembly code.
12338 @opindex mpadstruct
12339 This option is deprecated. It pads structures to multiple of 4 bytes,
12340 which is incompatible with the SH ABI@.
12344 Optimize for space instead of speed. Implied by @option{-Os}.
12347 @opindex mprefergot
12348 When generating position-independent code, emit function calls using
12349 the Global Offset Table instead of the Procedure Linkage Table.
12353 Generate a library function call to invalidate instruction cache
12354 entries, after fixing up a trampoline. This library function call
12355 doesn't assume it can write to the whole memory address space. This
12356 is the default when the target is @code{sh-*-linux*}.
12358 @item -multcost=@var{number}
12359 @opindex multcost=@var{number}
12360 Set the cost to assume for a multiply insn.
12362 @item -mdiv=@var{strategy}
12363 @opindex mdiv=@var{strategy}
12364 Set the division strategy to use for SHmedia code. @var{strategy} must be
12365 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
12366 inv:call2, inv:fp .
12367 "fp" performs the operation in floating point. This has a very high latency,
12368 but needs only a few instructions, so it might be a good choice if
12369 your code has enough easily exploitable ILP to allow the compiler to
12370 schedule the floating point instructions together with other instructions.
12371 Division by zero causes a floating point exception.
12372 "inv" uses integer operations to calculate the inverse of the divisor,
12373 and then multiplies the dividend with the inverse. This strategy allows
12374 cse and hoisting of the inverse calculation. Division by zero calculates
12375 an unspecified result, but does not trap.
12376 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
12377 have been found, or if the entire operation has been hoisted to the same
12378 place, the last stages of the inverse calculation are intertwined with the
12379 final multiply to reduce the overall latency, at the expense of using a few
12380 more instructions, and thus offering fewer scheduling opportunities with
12382 "call" calls a library function that usually implements the inv:minlat
12384 This gives high code density for m5-*media-nofpu compilations.
12385 "call2" uses a different entry point of the same library function, where it
12386 assumes that a pointer to a lookup table has already been set up, which
12387 exposes the pointer load to cse / code hoisting optimizations.
12388 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
12389 code generation, but if the code stays unoptimized, revert to the "call",
12390 "call2", or "fp" strategies, respectively. Note that the
12391 potentially-trapping side effect of division by zero is carried by a
12392 separate instruction, so it is possible that all the integer instructions
12393 are hoisted out, but the marker for the side effect stays where it is.
12394 A recombination to fp operations or a call is not possible in that case.
12395 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
12396 that the inverse calculation was nor separated from the multiply, they speed
12397 up division where the dividend fits into 20 bits (plus sign where applicable),
12398 by inserting a test to skip a number of operations in this case; this test
12399 slows down the case of larger dividends. inv20u assumes the case of a such
12400 a small dividend to be unlikely, and inv20l assumes it to be likely.
12402 @item -mdivsi3_libfunc=@var{name}
12403 @opindex mdivsi3_libfunc=@var{name}
12404 Set the name of the library function used for 32 bit signed division to
12405 @var{name}. This only affect the name used in the call and inv:call
12406 division strategies, and the compiler will still expect the same
12407 sets of input/output/clobbered registers as if this option was not present.
12409 @item -madjust-unroll
12410 @opindex madjust-unroll
12411 Throttle unrolling to avoid thrashing target registers.
12412 This option only has an effect if the gcc code base supports the
12413 TARGET_ADJUST_UNROLL_MAX target hook.
12415 @item -mindexed-addressing
12416 @opindex mindexed-addressing
12417 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
12418 This is only safe if the hardware and/or OS implement 32 bit wrap-around
12419 semantics for the indexed addressing mode. The architecture allows the
12420 implementation of processors with 64 bit MMU, which the OS could use to
12421 get 32 bit addressing, but since no current hardware implementation supports
12422 this or any other way to make the indexed addressing mode safe to use in
12423 the 32 bit ABI, the default is -mno-indexed-addressing.
12425 @item -mgettrcost=@var{number}
12426 @opindex mgettrcost=@var{number}
12427 Set the cost assumed for the gettr instruction to @var{number}.
12428 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
12432 Assume pt* instructions won't trap. This will generally generate better
12433 scheduled code, but is unsafe on current hardware. The current architecture
12434 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
12435 This has the unintentional effect of making it unsafe to schedule ptabs /
12436 ptrel before a branch, or hoist it out of a loop. For example,
12437 __do_global_ctors, a part of libgcc that runs constructors at program
12438 startup, calls functions in a list which is delimited by -1. With the
12439 -mpt-fixed option, the ptabs will be done before testing against -1.
12440 That means that all the constructors will be run a bit quicker, but when
12441 the loop comes to the end of the list, the program crashes because ptabs
12442 loads -1 into a target register. Since this option is unsafe for any
12443 hardware implementing the current architecture specification, the default
12444 is -mno-pt-fixed. Unless the user specifies a specific cost with
12445 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
12446 this deters register allocation using target registers for storing
12449 @item -minvalid-symbols
12450 @opindex minvalid-symbols
12451 Assume symbols might be invalid. Ordinary function symbols generated by
12452 the compiler will always be valid to load with movi/shori/ptabs or
12453 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
12454 to generate symbols that will cause ptabs / ptrel to trap.
12455 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
12456 It will then prevent cross-basic-block cse, hoisting and most scheduling
12457 of symbol loads. The default is @option{-mno-invalid-symbols}.
12460 @node SPARC Options
12461 @subsection SPARC Options
12462 @cindex SPARC options
12464 These @samp{-m} options are supported on the SPARC:
12467 @item -mno-app-regs
12469 @opindex mno-app-regs
12471 Specify @option{-mapp-regs} to generate output using the global registers
12472 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
12475 To be fully SVR4 ABI compliant at the cost of some performance loss,
12476 specify @option{-mno-app-regs}. You should compile libraries and system
12477 software with this option.
12480 @itemx -mhard-float
12482 @opindex mhard-float
12483 Generate output containing floating point instructions. This is the
12487 @itemx -msoft-float
12489 @opindex msoft-float
12490 Generate output containing library calls for floating point.
12491 @strong{Warning:} the requisite libraries are not available for all SPARC
12492 targets. Normally the facilities of the machine's usual C compiler are
12493 used, but this cannot be done directly in cross-compilation. You must make
12494 your own arrangements to provide suitable library functions for
12495 cross-compilation. The embedded targets @samp{sparc-*-aout} and
12496 @samp{sparclite-*-*} do provide software floating point support.
12498 @option{-msoft-float} changes the calling convention in the output file;
12499 therefore, it is only useful if you compile @emph{all} of a program with
12500 this option. In particular, you need to compile @file{libgcc.a}, the
12501 library that comes with GCC, with @option{-msoft-float} in order for
12504 @item -mhard-quad-float
12505 @opindex mhard-quad-float
12506 Generate output containing quad-word (long double) floating point
12509 @item -msoft-quad-float
12510 @opindex msoft-quad-float
12511 Generate output containing library calls for quad-word (long double)
12512 floating point instructions. The functions called are those specified
12513 in the SPARC ABI@. This is the default.
12515 As of this writing, there are no SPARC implementations that have hardware
12516 support for the quad-word floating point instructions. They all invoke
12517 a trap handler for one of these instructions, and then the trap handler
12518 emulates the effect of the instruction. Because of the trap handler overhead,
12519 this is much slower than calling the ABI library routines. Thus the
12520 @option{-msoft-quad-float} option is the default.
12522 @item -mno-unaligned-doubles
12523 @itemx -munaligned-doubles
12524 @opindex mno-unaligned-doubles
12525 @opindex munaligned-doubles
12526 Assume that doubles have 8 byte alignment. This is the default.
12528 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
12529 alignment only if they are contained in another type, or if they have an
12530 absolute address. Otherwise, it assumes they have 4 byte alignment.
12531 Specifying this option avoids some rare compatibility problems with code
12532 generated by other compilers. It is not the default because it results
12533 in a performance loss, especially for floating point code.
12535 @item -mno-faster-structs
12536 @itemx -mfaster-structs
12537 @opindex mno-faster-structs
12538 @opindex mfaster-structs
12539 With @option{-mfaster-structs}, the compiler assumes that structures
12540 should have 8 byte alignment. This enables the use of pairs of
12541 @code{ldd} and @code{std} instructions for copies in structure
12542 assignment, in place of twice as many @code{ld} and @code{st} pairs.
12543 However, the use of this changed alignment directly violates the SPARC
12544 ABI@. Thus, it's intended only for use on targets where the developer
12545 acknowledges that their resulting code will not be directly in line with
12546 the rules of the ABI@.
12548 @item -mimpure-text
12549 @opindex mimpure-text
12550 @option{-mimpure-text}, used in addition to @option{-shared}, tells
12551 the compiler to not pass @option{-z text} to the linker when linking a
12552 shared object. Using this option, you can link position-dependent
12553 code into a shared object.
12555 @option{-mimpure-text} suppresses the ``relocations remain against
12556 allocatable but non-writable sections'' linker error message.
12557 However, the necessary relocations will trigger copy-on-write, and the
12558 shared object is not actually shared across processes. Instead of
12559 using @option{-mimpure-text}, you should compile all source code with
12560 @option{-fpic} or @option{-fPIC}.
12562 This option is only available on SunOS and Solaris.
12564 @item -mcpu=@var{cpu_type}
12566 Set the instruction set, register set, and instruction scheduling parameters
12567 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
12568 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
12569 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
12570 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
12571 @samp{ultrasparc3}, and @samp{niagara}.
12573 Default instruction scheduling parameters are used for values that select
12574 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
12575 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
12577 Here is a list of each supported architecture and their supported
12582 v8: supersparc, hypersparc
12583 sparclite: f930, f934, sparclite86x
12585 v9: ultrasparc, ultrasparc3, niagara
12588 By default (unless configured otherwise), GCC generates code for the V7
12589 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
12590 additionally optimizes it for the Cypress CY7C602 chip, as used in the
12591 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
12592 SPARCStation 1, 2, IPX etc.
12594 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
12595 architecture. The only difference from V7 code is that the compiler emits
12596 the integer multiply and integer divide instructions which exist in SPARC-V8
12597 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
12598 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
12601 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
12602 the SPARC architecture. This adds the integer multiply, integer divide step
12603 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
12604 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
12605 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
12606 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
12607 MB86934 chip, which is the more recent SPARClite with FPU@.
12609 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
12610 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
12611 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
12612 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
12613 optimizes it for the TEMIC SPARClet chip.
12615 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
12616 architecture. This adds 64-bit integer and floating-point move instructions,
12617 3 additional floating-point condition code registers and conditional move
12618 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
12619 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
12620 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
12621 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
12622 @option{-mcpu=niagara}, the compiler additionally optimizes it for
12623 Sun UltraSPARC T1 chips.
12625 @item -mtune=@var{cpu_type}
12627 Set the instruction scheduling parameters for machine type
12628 @var{cpu_type}, but do not set the instruction set or register set that the
12629 option @option{-mcpu=@var{cpu_type}} would.
12631 The same values for @option{-mcpu=@var{cpu_type}} can be used for
12632 @option{-mtune=@var{cpu_type}}, but the only useful values are those
12633 that select a particular cpu implementation. Those are @samp{cypress},
12634 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
12635 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
12636 @samp{ultrasparc3}, and @samp{niagara}.
12641 @opindex mno-v8plus
12642 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
12643 difference from the V8 ABI is that the global and out registers are
12644 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
12645 mode for all SPARC-V9 processors.
12651 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
12652 Visual Instruction Set extensions. The default is @option{-mno-vis}.
12655 These @samp{-m} options are supported in addition to the above
12656 on SPARC-V9 processors in 64-bit environments:
12659 @item -mlittle-endian
12660 @opindex mlittle-endian
12661 Generate code for a processor running in little-endian mode. It is only
12662 available for a few configurations and most notably not on Solaris and Linux.
12668 Generate code for a 32-bit or 64-bit environment.
12669 The 32-bit environment sets int, long and pointer to 32 bits.
12670 The 64-bit environment sets int to 32 bits and long and pointer
12673 @item -mcmodel=medlow
12674 @opindex mcmodel=medlow
12675 Generate code for the Medium/Low code model: 64-bit addresses, programs
12676 must be linked in the low 32 bits of memory. Programs can be statically
12677 or dynamically linked.
12679 @item -mcmodel=medmid
12680 @opindex mcmodel=medmid
12681 Generate code for the Medium/Middle code model: 64-bit addresses, programs
12682 must be linked in the low 44 bits of memory, the text and data segments must
12683 be less than 2GB in size and the data segment must be located within 2GB of
12686 @item -mcmodel=medany
12687 @opindex mcmodel=medany
12688 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
12689 may be linked anywhere in memory, the text and data segments must be less
12690 than 2GB in size and the data segment must be located within 2GB of the
12693 @item -mcmodel=embmedany
12694 @opindex mcmodel=embmedany
12695 Generate code for the Medium/Anywhere code model for embedded systems:
12696 64-bit addresses, the text and data segments must be less than 2GB in
12697 size, both starting anywhere in memory (determined at link time). The
12698 global register %g4 points to the base of the data segment. Programs
12699 are statically linked and PIC is not supported.
12702 @itemx -mno-stack-bias
12703 @opindex mstack-bias
12704 @opindex mno-stack-bias
12705 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
12706 frame pointer if present, are offset by @minus{}2047 which must be added back
12707 when making stack frame references. This is the default in 64-bit mode.
12708 Otherwise, assume no such offset is present.
12711 These switches are supported in addition to the above on Solaris:
12716 Add support for multithreading using the Solaris threads library. This
12717 option sets flags for both the preprocessor and linker. This option does
12718 not affect the thread safety of object code produced by the compiler or
12719 that of libraries supplied with it.
12723 Add support for multithreading using the POSIX threads library. This
12724 option sets flags for both the preprocessor and linker. This option does
12725 not affect the thread safety of object code produced by the compiler or
12726 that of libraries supplied with it.
12730 This is a synonym for @option{-pthreads}.
12733 @node System V Options
12734 @subsection Options for System V
12736 These additional options are available on System V Release 4 for
12737 compatibility with other compilers on those systems:
12742 Create a shared object.
12743 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
12747 Identify the versions of each tool used by the compiler, in a
12748 @code{.ident} assembler directive in the output.
12752 Refrain from adding @code{.ident} directives to the output file (this is
12755 @item -YP,@var{dirs}
12757 Search the directories @var{dirs}, and no others, for libraries
12758 specified with @option{-l}.
12760 @item -Ym,@var{dir}
12762 Look in the directory @var{dir} to find the M4 preprocessor.
12763 The assembler uses this option.
12764 @c This is supposed to go with a -Yd for predefined M4 macro files, but
12765 @c the generic assembler that comes with Solaris takes just -Ym.
12768 @node TMS320C3x/C4x Options
12769 @subsection TMS320C3x/C4x Options
12770 @cindex TMS320C3x/C4x Options
12772 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
12776 @item -mcpu=@var{cpu_type}
12778 Set the instruction set, register set, and instruction scheduling
12779 parameters for machine type @var{cpu_type}. Supported values for
12780 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
12781 @samp{c44}. The default is @samp{c40} to generate code for the
12786 @itemx -msmall-memory
12788 @opindex mbig-memory
12790 @opindex msmall-memory
12792 Generates code for the big or small memory model. The small memory
12793 model assumed that all data fits into one 64K word page. At run-time
12794 the data page (DP) register must be set to point to the 64K page
12795 containing the .bss and .data program sections. The big memory model is
12796 the default and requires reloading of the DP register for every direct
12803 Allow (disallow) allocation of general integer operands into the block
12804 count register BK@.
12810 Enable (disable) generation of code using decrement and branch,
12811 DBcond(D), instructions. This is enabled by default for the C4x. To be
12812 on the safe side, this is disabled for the C3x, since the maximum
12813 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
12814 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
12815 that it can utilize the decrement and branch instruction, but will give
12816 up if there is more than one memory reference in the loop. Thus a loop
12817 where the loop counter is decremented can generate slightly more
12818 efficient code, in cases where the RPTB instruction cannot be utilized.
12820 @item -mdp-isr-reload
12822 @opindex mdp-isr-reload
12824 Force the DP register to be saved on entry to an interrupt service
12825 routine (ISR), reloaded to point to the data section, and restored on
12826 exit from the ISR@. This should not be required unless someone has
12827 violated the small memory model by modifying the DP register, say within
12834 For the C3x use the 24-bit MPYI instruction for integer multiplies
12835 instead of a library call to guarantee 32-bit results. Note that if one
12836 of the operands is a constant, then the multiplication will be performed
12837 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
12838 then squaring operations are performed inline instead of a library call.
12841 @itemx -mno-fast-fix
12843 @opindex mno-fast-fix
12844 The C3x/C4x FIX instruction to convert a floating point value to an
12845 integer value chooses the nearest integer less than or equal to the
12846 floating point value rather than to the nearest integer. Thus if the
12847 floating point number is negative, the result will be incorrectly
12848 truncated an additional code is necessary to detect and correct this
12849 case. This option can be used to disable generation of the additional
12850 code required to correct the result.
12856 Enable (disable) generation of repeat block sequences using the RPTB
12857 instruction for zero overhead looping. The RPTB construct is only used
12858 for innermost loops that do not call functions or jump across the loop
12859 boundaries. There is no advantage having nested RPTB loops due to the
12860 overhead required to save and restore the RC, RS, and RE registers.
12861 This is enabled by default with @option{-O2}.
12863 @item -mrpts=@var{count}
12867 Enable (disable) the use of the single instruction repeat instruction
12868 RPTS@. If a repeat block contains a single instruction, and the loop
12869 count can be guaranteed to be less than the value @var{count}, GCC will
12870 emit a RPTS instruction instead of a RPTB@. If no value is specified,
12871 then a RPTS will be emitted even if the loop count cannot be determined
12872 at compile time. Note that the repeated instruction following RPTS does
12873 not have to be reloaded from memory each iteration, thus freeing up the
12874 CPU buses for operands. However, since interrupts are blocked by this
12875 instruction, it is disabled by default.
12877 @item -mloop-unsigned
12878 @itemx -mno-loop-unsigned
12879 @opindex mloop-unsigned
12880 @opindex mno-loop-unsigned
12881 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
12882 is @math{2^{31} + 1} since these instructions test if the iteration count is
12883 negative to terminate the loop. If the iteration count is unsigned
12884 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
12885 exceeded. This switch allows an unsigned iteration count.
12889 Try to emit an assembler syntax that the TI assembler (asm30) is happy
12890 with. This also enforces compatibility with the API employed by the TI
12891 C3x C compiler. For example, long doubles are passed as structures
12892 rather than in floating point registers.
12898 Generate code that uses registers (stack) for passing arguments to functions.
12899 By default, arguments are passed in registers where possible rather
12900 than by pushing arguments on to the stack.
12902 @item -mparallel-insns
12903 @itemx -mno-parallel-insns
12904 @opindex mparallel-insns
12905 @opindex mno-parallel-insns
12906 Allow the generation of parallel instructions. This is enabled by
12907 default with @option{-O2}.
12909 @item -mparallel-mpy
12910 @itemx -mno-parallel-mpy
12911 @opindex mparallel-mpy
12912 @opindex mno-parallel-mpy
12913 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
12914 provided @option{-mparallel-insns} is also specified. These instructions have
12915 tight register constraints which can pessimize the code generation
12916 of large functions.
12921 @subsection V850 Options
12922 @cindex V850 Options
12924 These @samp{-m} options are defined for V850 implementations:
12928 @itemx -mno-long-calls
12929 @opindex mlong-calls
12930 @opindex mno-long-calls
12931 Treat all calls as being far away (near). If calls are assumed to be
12932 far away, the compiler will always load the functions address up into a
12933 register, and call indirect through the pointer.
12939 Do not optimize (do optimize) basic blocks that use the same index
12940 pointer 4 or more times to copy pointer into the @code{ep} register, and
12941 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
12942 option is on by default if you optimize.
12944 @item -mno-prolog-function
12945 @itemx -mprolog-function
12946 @opindex mno-prolog-function
12947 @opindex mprolog-function
12948 Do not use (do use) external functions to save and restore registers
12949 at the prologue and epilogue of a function. The external functions
12950 are slower, but use less code space if more than one function saves
12951 the same number of registers. The @option{-mprolog-function} option
12952 is on by default if you optimize.
12956 Try to make the code as small as possible. At present, this just turns
12957 on the @option{-mep} and @option{-mprolog-function} options.
12959 @item -mtda=@var{n}
12961 Put static or global variables whose size is @var{n} bytes or less into
12962 the tiny data area that register @code{ep} points to. The tiny data
12963 area can hold up to 256 bytes in total (128 bytes for byte references).
12965 @item -msda=@var{n}
12967 Put static or global variables whose size is @var{n} bytes or less into
12968 the small data area that register @code{gp} points to. The small data
12969 area can hold up to 64 kilobytes.
12971 @item -mzda=@var{n}
12973 Put static or global variables whose size is @var{n} bytes or less into
12974 the first 32 kilobytes of memory.
12978 Specify that the target processor is the V850.
12981 @opindex mbig-switch
12982 Generate code suitable for big switch tables. Use this option only if
12983 the assembler/linker complain about out of range branches within a switch
12988 This option will cause r2 and r5 to be used in the code generated by
12989 the compiler. This setting is the default.
12991 @item -mno-app-regs
12992 @opindex mno-app-regs
12993 This option will cause r2 and r5 to be treated as fixed registers.
12997 Specify that the target processor is the V850E1. The preprocessor
12998 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
12999 this option is used.
13003 Specify that the target processor is the V850E@. The preprocessor
13004 constant @samp{__v850e__} will be defined if this option is used.
13006 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
13007 are defined then a default target processor will be chosen and the
13008 relevant @samp{__v850*__} preprocessor constant will be defined.
13010 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
13011 defined, regardless of which processor variant is the target.
13013 @item -mdisable-callt
13014 @opindex mdisable-callt
13015 This option will suppress generation of the CALLT instruction for the
13016 v850e and v850e1 flavors of the v850 architecture. The default is
13017 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
13022 @subsection VAX Options
13023 @cindex VAX options
13025 These @samp{-m} options are defined for the VAX:
13030 Do not output certain jump instructions (@code{aobleq} and so on)
13031 that the Unix assembler for the VAX cannot handle across long
13036 Do output those jump instructions, on the assumption that you
13037 will assemble with the GNU assembler.
13041 Output code for g-format floating point numbers instead of d-format.
13044 @node x86-64 Options
13045 @subsection x86-64 Options
13046 @cindex x86-64 options
13048 These are listed under @xref{i386 and x86-64 Options}.
13050 @node Xstormy16 Options
13051 @subsection Xstormy16 Options
13052 @cindex Xstormy16 Options
13054 These options are defined for Xstormy16:
13059 Choose startup files and linker script suitable for the simulator.
13062 @node Xtensa Options
13063 @subsection Xtensa Options
13064 @cindex Xtensa Options
13066 These options are supported for Xtensa targets:
13070 @itemx -mno-const16
13072 @opindex mno-const16
13073 Enable or disable use of @code{CONST16} instructions for loading
13074 constant values. The @code{CONST16} instruction is currently not a
13075 standard option from Tensilica. When enabled, @code{CONST16}
13076 instructions are always used in place of the standard @code{L32R}
13077 instructions. The use of @code{CONST16} is enabled by default only if
13078 the @code{L32R} instruction is not available.
13081 @itemx -mno-fused-madd
13082 @opindex mfused-madd
13083 @opindex mno-fused-madd
13084 Enable or disable use of fused multiply/add and multiply/subtract
13085 instructions in the floating-point option. This has no effect if the
13086 floating-point option is not also enabled. Disabling fused multiply/add
13087 and multiply/subtract instructions forces the compiler to use separate
13088 instructions for the multiply and add/subtract operations. This may be
13089 desirable in some cases where strict IEEE 754-compliant results are
13090 required: the fused multiply add/subtract instructions do not round the
13091 intermediate result, thereby producing results with @emph{more} bits of
13092 precision than specified by the IEEE standard. Disabling fused multiply
13093 add/subtract instructions also ensures that the program output is not
13094 sensitive to the compiler's ability to combine multiply and add/subtract
13097 @item -mtext-section-literals
13098 @itemx -mno-text-section-literals
13099 @opindex mtext-section-literals
13100 @opindex mno-text-section-literals
13101 Control the treatment of literal pools. The default is
13102 @option{-mno-text-section-literals}, which places literals in a separate
13103 section in the output file. This allows the literal pool to be placed
13104 in a data RAM/ROM, and it also allows the linker to combine literal
13105 pools from separate object files to remove redundant literals and
13106 improve code size. With @option{-mtext-section-literals}, the literals
13107 are interspersed in the text section in order to keep them as close as
13108 possible to their references. This may be necessary for large assembly
13111 @item -mtarget-align
13112 @itemx -mno-target-align
13113 @opindex mtarget-align
13114 @opindex mno-target-align
13115 When this option is enabled, GCC instructs the assembler to
13116 automatically align instructions to reduce branch penalties at the
13117 expense of some code density. The assembler attempts to widen density
13118 instructions to align branch targets and the instructions following call
13119 instructions. If there are not enough preceding safe density
13120 instructions to align a target, no widening will be performed. The
13121 default is @option{-mtarget-align}. These options do not affect the
13122 treatment of auto-aligned instructions like @code{LOOP}, which the
13123 assembler will always align, either by widening density instructions or
13124 by inserting no-op instructions.
13127 @itemx -mno-longcalls
13128 @opindex mlongcalls
13129 @opindex mno-longcalls
13130 When this option is enabled, GCC instructs the assembler to translate
13131 direct calls to indirect calls unless it can determine that the target
13132 of a direct call is in the range allowed by the call instruction. This
13133 translation typically occurs for calls to functions in other source
13134 files. Specifically, the assembler translates a direct @code{CALL}
13135 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
13136 The default is @option{-mno-longcalls}. This option should be used in
13137 programs where the call target can potentially be out of range. This
13138 option is implemented in the assembler, not the compiler, so the
13139 assembly code generated by GCC will still show direct call
13140 instructions---look at the disassembled object code to see the actual
13141 instructions. Note that the assembler will use an indirect call for
13142 every cross-file call, not just those that really will be out of range.
13145 @node zSeries Options
13146 @subsection zSeries Options
13147 @cindex zSeries options
13149 These are listed under @xref{S/390 and zSeries Options}.
13151 @node Code Gen Options
13152 @section Options for Code Generation Conventions
13153 @cindex code generation conventions
13154 @cindex options, code generation
13155 @cindex run-time options
13157 These machine-independent options control the interface conventions
13158 used in code generation.
13160 Most of them have both positive and negative forms; the negative form
13161 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
13162 one of the forms is listed---the one which is not the default. You
13163 can figure out the other form by either removing @samp{no-} or adding
13167 @item -fbounds-check
13168 @opindex fbounds-check
13169 For front-ends that support it, generate additional code to check that
13170 indices used to access arrays are within the declared range. This is
13171 currently only supported by the Java and Fortran front-ends, where
13172 this option defaults to true and false respectively.
13176 This option generates traps for signed overflow on addition, subtraction,
13177 multiplication operations.
13181 This option instructs the compiler to assume that signed arithmetic
13182 overflow of addition, subtraction and multiplication wraps around
13183 using twos-complement representation. This flag enables some optimizations
13184 and disables others. This option is enabled by default for the Java
13185 front-end, as required by the Java language specification.
13188 @opindex fexceptions
13189 Enable exception handling. Generates extra code needed to propagate
13190 exceptions. For some targets, this implies GCC will generate frame
13191 unwind information for all functions, which can produce significant data
13192 size overhead, although it does not affect execution. If you do not
13193 specify this option, GCC will enable it by default for languages like
13194 C++ which normally require exception handling, and disable it for
13195 languages like C that do not normally require it. However, you may need
13196 to enable this option when compiling C code that needs to interoperate
13197 properly with exception handlers written in C++. You may also wish to
13198 disable this option if you are compiling older C++ programs that don't
13199 use exception handling.
13201 @item -fnon-call-exceptions
13202 @opindex fnon-call-exceptions
13203 Generate code that allows trapping instructions to throw exceptions.
13204 Note that this requires platform-specific runtime support that does
13205 not exist everywhere. Moreover, it only allows @emph{trapping}
13206 instructions to throw exceptions, i.e.@: memory references or floating
13207 point instructions. It does not allow exceptions to be thrown from
13208 arbitrary signal handlers such as @code{SIGALRM}.
13210 @item -funwind-tables
13211 @opindex funwind-tables
13212 Similar to @option{-fexceptions}, except that it will just generate any needed
13213 static data, but will not affect the generated code in any other way.
13214 You will normally not enable this option; instead, a language processor
13215 that needs this handling would enable it on your behalf.
13217 @item -fasynchronous-unwind-tables
13218 @opindex fasynchronous-unwind-tables
13219 Generate unwind table in dwarf2 format, if supported by target machine. The
13220 table is exact at each instruction boundary, so it can be used for stack
13221 unwinding from asynchronous events (such as debugger or garbage collector).
13223 @item -fpcc-struct-return
13224 @opindex fpcc-struct-return
13225 Return ``short'' @code{struct} and @code{union} values in memory like
13226 longer ones, rather than in registers. This convention is less
13227 efficient, but it has the advantage of allowing intercallability between
13228 GCC-compiled files and files compiled with other compilers, particularly
13229 the Portable C Compiler (pcc).
13231 The precise convention for returning structures in memory depends
13232 on the target configuration macros.
13234 Short structures and unions are those whose size and alignment match
13235 that of some integer type.
13237 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
13238 switch is not binary compatible with code compiled with the
13239 @option{-freg-struct-return} switch.
13240 Use it to conform to a non-default application binary interface.
13242 @item -freg-struct-return
13243 @opindex freg-struct-return
13244 Return @code{struct} and @code{union} values in registers when possible.
13245 This is more efficient for small structures than
13246 @option{-fpcc-struct-return}.
13248 If you specify neither @option{-fpcc-struct-return} nor
13249 @option{-freg-struct-return}, GCC defaults to whichever convention is
13250 standard for the target. If there is no standard convention, GCC
13251 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
13252 the principal compiler. In those cases, we can choose the standard, and
13253 we chose the more efficient register return alternative.
13255 @strong{Warning:} code compiled with the @option{-freg-struct-return}
13256 switch is not binary compatible with code compiled with the
13257 @option{-fpcc-struct-return} switch.
13258 Use it to conform to a non-default application binary interface.
13260 @item -fshort-enums
13261 @opindex fshort-enums
13262 Allocate to an @code{enum} type only as many bytes as it needs for the
13263 declared range of possible values. Specifically, the @code{enum} type
13264 will be equivalent to the smallest integer type which has enough room.
13266 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
13267 code that is not binary compatible with code generated without that switch.
13268 Use it to conform to a non-default application binary interface.
13270 @item -fshort-double
13271 @opindex fshort-double
13272 Use the same size for @code{double} as for @code{float}.
13274 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
13275 code that is not binary compatible with code generated without that switch.
13276 Use it to conform to a non-default application binary interface.
13278 @item -fshort-wchar
13279 @opindex fshort-wchar
13280 Override the underlying type for @samp{wchar_t} to be @samp{short
13281 unsigned int} instead of the default for the target. This option is
13282 useful for building programs to run under WINE@.
13284 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
13285 code that is not binary compatible with code generated without that switch.
13286 Use it to conform to a non-default application binary interface.
13289 @opindex fno-common
13290 In C, allocate even uninitialized global variables in the data section of the
13291 object file, rather than generating them as common blocks. This has the
13292 effect that if the same variable is declared (without @code{extern}) in
13293 two different compilations, you will get an error when you link them.
13294 The only reason this might be useful is if you wish to verify that the
13295 program will work on other systems which always work this way.
13299 Ignore the @samp{#ident} directive.
13301 @item -finhibit-size-directive
13302 @opindex finhibit-size-directive
13303 Don't output a @code{.size} assembler directive, or anything else that
13304 would cause trouble if the function is split in the middle, and the
13305 two halves are placed at locations far apart in memory. This option is
13306 used when compiling @file{crtstuff.c}; you should not need to use it
13309 @item -fverbose-asm
13310 @opindex fverbose-asm
13311 Put extra commentary information in the generated assembly code to
13312 make it more readable. This option is generally only of use to those
13313 who actually need to read the generated assembly code (perhaps while
13314 debugging the compiler itself).
13316 @option{-fno-verbose-asm}, the default, causes the
13317 extra information to be omitted and is useful when comparing two assembler
13322 @cindex global offset table
13324 Generate position-independent code (PIC) suitable for use in a shared
13325 library, if supported for the target machine. Such code accesses all
13326 constant addresses through a global offset table (GOT)@. The dynamic
13327 loader resolves the GOT entries when the program starts (the dynamic
13328 loader is not part of GCC; it is part of the operating system). If
13329 the GOT size for the linked executable exceeds a machine-specific
13330 maximum size, you get an error message from the linker indicating that
13331 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
13332 instead. (These maximums are 8k on the SPARC and 32k
13333 on the m68k and RS/6000. The 386 has no such limit.)
13335 Position-independent code requires special support, and therefore works
13336 only on certain machines. For the 386, GCC supports PIC for System V
13337 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
13338 position-independent.
13340 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13345 If supported for the target machine, emit position-independent code,
13346 suitable for dynamic linking and avoiding any limit on the size of the
13347 global offset table. This option makes a difference on the m68k,
13348 PowerPC and SPARC@.
13350 Position-independent code requires special support, and therefore works
13351 only on certain machines.
13353 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13360 These options are similar to @option{-fpic} and @option{-fPIC}, but
13361 generated position independent code can be only linked into executables.
13362 Usually these options are used when @option{-pie} GCC option will be
13363 used during linking.
13365 @item -fno-jump-tables
13366 @opindex fno-jump-tables
13367 Do not use jump tables for switch statements even where it would be
13368 more efficient than other code generation strategies. This option is
13369 of use in conjunction with @option{-fpic} or @option{-fPIC} for
13370 building code which forms part of a dynamic linker and cannot
13371 reference the address of a jump table. On some targets, jump tables
13372 do not require a GOT and this option is not needed.
13374 @item -ffixed-@var{reg}
13376 Treat the register named @var{reg} as a fixed register; generated code
13377 should never refer to it (except perhaps as a stack pointer, frame
13378 pointer or in some other fixed role).
13380 @var{reg} must be the name of a register. The register names accepted
13381 are machine-specific and are defined in the @code{REGISTER_NAMES}
13382 macro in the machine description macro file.
13384 This flag does not have a negative form, because it specifies a
13387 @item -fcall-used-@var{reg}
13388 @opindex fcall-used
13389 Treat the register named @var{reg} as an allocable register that is
13390 clobbered by function calls. It may be allocated for temporaries or
13391 variables that do not live across a call. Functions compiled this way
13392 will not save and restore the register @var{reg}.
13394 It is an error to used this flag with the frame pointer or stack pointer.
13395 Use of this flag for other registers that have fixed pervasive roles in
13396 the machine's execution model will produce disastrous results.
13398 This flag does not have a negative form, because it specifies a
13401 @item -fcall-saved-@var{reg}
13402 @opindex fcall-saved
13403 Treat the register named @var{reg} as an allocable register saved by
13404 functions. It may be allocated even for temporaries or variables that
13405 live across a call. Functions compiled this way will save and restore
13406 the register @var{reg} if they use it.
13408 It is an error to used this flag with the frame pointer or stack pointer.
13409 Use of this flag for other registers that have fixed pervasive roles in
13410 the machine's execution model will produce disastrous results.
13412 A different sort of disaster will result from the use of this flag for
13413 a register in which function values may be returned.
13415 This flag does not have a negative form, because it specifies a
13418 @item -fpack-struct[=@var{n}]
13419 @opindex fpack-struct
13420 Without a value specified, pack all structure members together without
13421 holes. When a value is specified (which must be a small power of two), pack
13422 structure members according to this value, representing the maximum
13423 alignment (that is, objects with default alignment requirements larger than
13424 this will be output potentially unaligned at the next fitting location.
13426 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
13427 code that is not binary compatible with code generated without that switch.
13428 Additionally, it makes the code suboptimal.
13429 Use it to conform to a non-default application binary interface.
13431 @item -finstrument-functions
13432 @opindex finstrument-functions
13433 Generate instrumentation calls for entry and exit to functions. Just
13434 after function entry and just before function exit, the following
13435 profiling functions will be called with the address of the current
13436 function and its call site. (On some platforms,
13437 @code{__builtin_return_address} does not work beyond the current
13438 function, so the call site information may not be available to the
13439 profiling functions otherwise.)
13442 void __cyg_profile_func_enter (void *this_fn,
13444 void __cyg_profile_func_exit (void *this_fn,
13448 The first argument is the address of the start of the current function,
13449 which may be looked up exactly in the symbol table.
13451 This instrumentation is also done for functions expanded inline in other
13452 functions. The profiling calls will indicate where, conceptually, the
13453 inline function is entered and exited. This means that addressable
13454 versions of such functions must be available. If all your uses of a
13455 function are expanded inline, this may mean an additional expansion of
13456 code size. If you use @samp{extern inline} in your C code, an
13457 addressable version of such functions must be provided. (This is
13458 normally the case anyways, but if you get lucky and the optimizer always
13459 expands the functions inline, you might have gotten away without
13460 providing static copies.)
13462 A function may be given the attribute @code{no_instrument_function}, in
13463 which case this instrumentation will not be done. This can be used, for
13464 example, for the profiling functions listed above, high-priority
13465 interrupt routines, and any functions from which the profiling functions
13466 cannot safely be called (perhaps signal handlers, if the profiling
13467 routines generate output or allocate memory).
13469 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
13470 @opindex finstrument-functions-exclude-file-list
13472 Set the list of functions that are excluded from instrumentation (see
13473 the description of @code{-finstrument-functions}). If the file that
13474 contains a function definition matches with one of @var{file}, then
13475 that function is not instrumented. The match is done on substrings:
13476 if the @var{file} parameter is a substring of the file name, it is
13477 considered to be a match.
13480 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
13481 will exclude any inline function defined in files whose pathnames
13482 contain @code{/bits/stl} or @code{include/sys}.
13484 If, for some reason, you want to include letter @code{','} in one of
13485 @var{sym}, write @code{'\,'}. For example,
13486 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
13487 (note the single quote surrounding the option).
13489 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
13490 @opindex finstrument-functions-exclude-function-list
13492 This is similar to @code{-finstrument-functions-exclude-file-list},
13493 but this option sets the list of function names to be excluded from
13494 instrumentation. The function name to be matched is its user-visible
13495 name, such as @code{vector<int> blah(const vector<int> &)}, not the
13496 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
13497 match is done on substrings: if the @var{sym} parameter is a substring
13498 of the function name, it is considered to be a match.
13500 @item -fstack-check
13501 @opindex fstack-check
13502 Generate code to verify that you do not go beyond the boundary of the
13503 stack. You should specify this flag if you are running in an
13504 environment with multiple threads, but only rarely need to specify it in
13505 a single-threaded environment since stack overflow is automatically
13506 detected on nearly all systems if there is only one stack.
13508 Note that this switch does not actually cause checking to be done; the
13509 operating system must do that. The switch causes generation of code
13510 to ensure that the operating system sees the stack being extended.
13512 @item -fstack-limit-register=@var{reg}
13513 @itemx -fstack-limit-symbol=@var{sym}
13514 @itemx -fno-stack-limit
13515 @opindex fstack-limit-register
13516 @opindex fstack-limit-symbol
13517 @opindex fno-stack-limit
13518 Generate code to ensure that the stack does not grow beyond a certain value,
13519 either the value of a register or the address of a symbol. If the stack
13520 would grow beyond the value, a signal is raised. For most targets,
13521 the signal is raised before the stack overruns the boundary, so
13522 it is possible to catch the signal without taking special precautions.
13524 For instance, if the stack starts at absolute address @samp{0x80000000}
13525 and grows downwards, you can use the flags
13526 @option{-fstack-limit-symbol=__stack_limit} and
13527 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
13528 of 128KB@. Note that this may only work with the GNU linker.
13530 @cindex aliasing of parameters
13531 @cindex parameters, aliased
13532 @item -fargument-alias
13533 @itemx -fargument-noalias
13534 @itemx -fargument-noalias-global
13535 @itemx -fargument-noalias-anything
13536 @opindex fargument-alias
13537 @opindex fargument-noalias
13538 @opindex fargument-noalias-global
13539 @opindex fargument-noalias-anything
13540 Specify the possible relationships among parameters and between
13541 parameters and global data.
13543 @option{-fargument-alias} specifies that arguments (parameters) may
13544 alias each other and may alias global storage.@*
13545 @option{-fargument-noalias} specifies that arguments do not alias
13546 each other, but may alias global storage.@*
13547 @option{-fargument-noalias-global} specifies that arguments do not
13548 alias each other and do not alias global storage.
13549 @option{-fargument-noalias-anything} specifies that arguments do not
13550 alias any other storage.
13552 Each language will automatically use whatever option is required by
13553 the language standard. You should not need to use these options yourself.
13555 @item -fleading-underscore
13556 @opindex fleading-underscore
13557 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
13558 change the way C symbols are represented in the object file. One use
13559 is to help link with legacy assembly code.
13561 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
13562 generate code that is not binary compatible with code generated without that
13563 switch. Use it to conform to a non-default application binary interface.
13564 Not all targets provide complete support for this switch.
13566 @item -ftls-model=@var{model}
13567 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
13568 The @var{model} argument should be one of @code{global-dynamic},
13569 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
13571 The default without @option{-fpic} is @code{initial-exec}; with
13572 @option{-fpic} the default is @code{global-dynamic}.
13574 @item -fvisibility=@var{default|internal|hidden|protected}
13575 @opindex fvisibility
13576 Set the default ELF image symbol visibility to the specified option---all
13577 symbols will be marked with this unless overridden within the code.
13578 Using this feature can very substantially improve linking and
13579 load times of shared object libraries, produce more optimized
13580 code, provide near-perfect API export and prevent symbol clashes.
13581 It is @strong{strongly} recommended that you use this in any shared objects
13584 Despite the nomenclature, @code{default} always means public ie;
13585 available to be linked against from outside the shared object.
13586 @code{protected} and @code{internal} are pretty useless in real-world
13587 usage so the only other commonly used option will be @code{hidden}.
13588 The default if @option{-fvisibility} isn't specified is
13589 @code{default}, i.e., make every
13590 symbol public---this causes the same behavior as previous versions of
13593 A good explanation of the benefits offered by ensuring ELF
13594 symbols have the correct visibility is given by ``How To Write
13595 Shared Libraries'' by Ulrich Drepper (which can be found at
13596 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
13597 solution made possible by this option to marking things hidden when
13598 the default is public is to make the default hidden and mark things
13599 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
13600 and @code{__attribute__ ((visibility("default")))} instead of
13601 @code{__declspec(dllexport)} you get almost identical semantics with
13602 identical syntax. This is a great boon to those working with
13603 cross-platform projects.
13605 For those adding visibility support to existing code, you may find
13606 @samp{#pragma GCC visibility} of use. This works by you enclosing
13607 the declarations you wish to set visibility for with (for example)
13608 @samp{#pragma GCC visibility push(hidden)} and
13609 @samp{#pragma GCC visibility pop}.
13610 Bear in mind that symbol visibility should be viewed @strong{as
13611 part of the API interface contract} and thus all new code should
13612 always specify visibility when it is not the default ie; declarations
13613 only for use within the local DSO should @strong{always} be marked explicitly
13614 as hidden as so to avoid PLT indirection overheads---making this
13615 abundantly clear also aids readability and self-documentation of the code.
13616 Note that due to ISO C++ specification requirements, operator new and
13617 operator delete must always be of default visibility.
13619 Be aware that headers from outside your project, in particular system
13620 headers and headers from any other library you use, may not be
13621 expecting to be compiled with visibility other than the default. You
13622 may need to explicitly say @samp{#pragma GCC visibility push(default)}
13623 before including any such headers.
13625 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
13626 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
13627 no modifications. However, this means that calls to @samp{extern}
13628 functions with no explicit visibility will use the PLT, so it is more
13629 effective to use @samp{__attribute ((visibility))} and/or
13630 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
13631 declarations should be treated as hidden.
13633 Note that @samp{-fvisibility} does affect C++ vague linkage
13634 entities. This means that, for instance, an exception class that will
13635 be thrown between DSOs must be explicitly marked with default
13636 visibility so that the @samp{type_info} nodes will be unified between
13639 An overview of these techniques, their benefits and how to use them
13640 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
13646 @node Environment Variables
13647 @section Environment Variables Affecting GCC
13648 @cindex environment variables
13650 @c man begin ENVIRONMENT
13651 This section describes several environment variables that affect how GCC
13652 operates. Some of them work by specifying directories or prefixes to use
13653 when searching for various kinds of files. Some are used to specify other
13654 aspects of the compilation environment.
13656 Note that you can also specify places to search using options such as
13657 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
13658 take precedence over places specified using environment variables, which
13659 in turn take precedence over those specified by the configuration of GCC@.
13660 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
13661 GNU Compiler Collection (GCC) Internals}.
13666 @c @itemx LC_COLLATE
13668 @c @itemx LC_MONETARY
13669 @c @itemx LC_NUMERIC
13674 @c @findex LC_COLLATE
13675 @findex LC_MESSAGES
13676 @c @findex LC_MONETARY
13677 @c @findex LC_NUMERIC
13681 These environment variables control the way that GCC uses
13682 localization information that allow GCC to work with different
13683 national conventions. GCC inspects the locale categories
13684 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
13685 so. These locale categories can be set to any value supported by your
13686 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
13687 Kingdom encoded in UTF-8.
13689 The @env{LC_CTYPE} environment variable specifies character
13690 classification. GCC uses it to determine the character boundaries in
13691 a string; this is needed for some multibyte encodings that contain quote
13692 and escape characters that would otherwise be interpreted as a string
13695 The @env{LC_MESSAGES} environment variable specifies the language to
13696 use in diagnostic messages.
13698 If the @env{LC_ALL} environment variable is set, it overrides the value
13699 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
13700 and @env{LC_MESSAGES} default to the value of the @env{LANG}
13701 environment variable. If none of these variables are set, GCC
13702 defaults to traditional C English behavior.
13706 If @env{TMPDIR} is set, it specifies the directory to use for temporary
13707 files. GCC uses temporary files to hold the output of one stage of
13708 compilation which is to be used as input to the next stage: for example,
13709 the output of the preprocessor, which is the input to the compiler
13712 @item GCC_EXEC_PREFIX
13713 @findex GCC_EXEC_PREFIX
13714 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
13715 names of the subprograms executed by the compiler. No slash is added
13716 when this prefix is combined with the name of a subprogram, but you can
13717 specify a prefix that ends with a slash if you wish.
13719 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
13720 an appropriate prefix to use based on the pathname it was invoked with.
13722 If GCC cannot find the subprogram using the specified prefix, it
13723 tries looking in the usual places for the subprogram.
13725 The default value of @env{GCC_EXEC_PREFIX} is
13726 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
13727 of @code{prefix} when you ran the @file{configure} script.
13729 Other prefixes specified with @option{-B} take precedence over this prefix.
13731 This prefix is also used for finding files such as @file{crt0.o} that are
13734 In addition, the prefix is used in an unusual way in finding the
13735 directories to search for header files. For each of the standard
13736 directories whose name normally begins with @samp{/usr/local/lib/gcc}
13737 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
13738 replacing that beginning with the specified prefix to produce an
13739 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
13740 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
13741 These alternate directories are searched first; the standard directories
13744 @item COMPILER_PATH
13745 @findex COMPILER_PATH
13746 The value of @env{COMPILER_PATH} is a colon-separated list of
13747 directories, much like @env{PATH}. GCC tries the directories thus
13748 specified when searching for subprograms, if it can't find the
13749 subprograms using @env{GCC_EXEC_PREFIX}.
13752 @findex LIBRARY_PATH
13753 The value of @env{LIBRARY_PATH} is a colon-separated list of
13754 directories, much like @env{PATH}. When configured as a native compiler,
13755 GCC tries the directories thus specified when searching for special
13756 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
13757 using GCC also uses these directories when searching for ordinary
13758 libraries for the @option{-l} option (but directories specified with
13759 @option{-L} come first).
13763 @cindex locale definition
13764 This variable is used to pass locale information to the compiler. One way in
13765 which this information is used is to determine the character set to be used
13766 when character literals, string literals and comments are parsed in C and C++.
13767 When the compiler is configured to allow multibyte characters,
13768 the following values for @env{LANG} are recognized:
13772 Recognize JIS characters.
13774 Recognize SJIS characters.
13776 Recognize EUCJP characters.
13779 If @env{LANG} is not defined, or if it has some other value, then the
13780 compiler will use mblen and mbtowc as defined by the default locale to
13781 recognize and translate multibyte characters.
13785 Some additional environments variables affect the behavior of the
13788 @include cppenv.texi
13792 @node Precompiled Headers
13793 @section Using Precompiled Headers
13794 @cindex precompiled headers
13795 @cindex speed of compilation
13797 Often large projects have many header files that are included in every
13798 source file. The time the compiler takes to process these header files
13799 over and over again can account for nearly all of the time required to
13800 build the project. To make builds faster, GCC allows users to
13801 `precompile' a header file; then, if builds can use the precompiled
13802 header file they will be much faster.
13804 To create a precompiled header file, simply compile it as you would any
13805 other file, if necessary using the @option{-x} option to make the driver
13806 treat it as a C or C++ header file. You will probably want to use a
13807 tool like @command{make} to keep the precompiled header up-to-date when
13808 the headers it contains change.
13810 A precompiled header file will be searched for when @code{#include} is
13811 seen in the compilation. As it searches for the included file
13812 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
13813 compiler looks for a precompiled header in each directory just before it
13814 looks for the include file in that directory. The name searched for is
13815 the name specified in the @code{#include} with @samp{.gch} appended. If
13816 the precompiled header file can't be used, it is ignored.
13818 For instance, if you have @code{#include "all.h"}, and you have
13819 @file{all.h.gch} in the same directory as @file{all.h}, then the
13820 precompiled header file will be used if possible, and the original
13821 header will be used otherwise.
13823 Alternatively, you might decide to put the precompiled header file in a
13824 directory and use @option{-I} to ensure that directory is searched
13825 before (or instead of) the directory containing the original header.
13826 Then, if you want to check that the precompiled header file is always
13827 used, you can put a file of the same name as the original header in this
13828 directory containing an @code{#error} command.
13830 This also works with @option{-include}. So yet another way to use
13831 precompiled headers, good for projects not designed with precompiled
13832 header files in mind, is to simply take most of the header files used by
13833 a project, include them from another header file, precompile that header
13834 file, and @option{-include} the precompiled header. If the header files
13835 have guards against multiple inclusion, they will be skipped because
13836 they've already been included (in the precompiled header).
13838 If you need to precompile the same header file for different
13839 languages, targets, or compiler options, you can instead make a
13840 @emph{directory} named like @file{all.h.gch}, and put each precompiled
13841 header in the directory, perhaps using @option{-o}. It doesn't matter
13842 what you call the files in the directory, every precompiled header in
13843 the directory will be considered. The first precompiled header
13844 encountered in the directory that is valid for this compilation will
13845 be used; they're searched in no particular order.
13847 There are many other possibilities, limited only by your imagination,
13848 good sense, and the constraints of your build system.
13850 A precompiled header file can be used only when these conditions apply:
13854 Only one precompiled header can be used in a particular compilation.
13857 A precompiled header can't be used once the first C token is seen. You
13858 can have preprocessor directives before a precompiled header; you can
13859 even include a precompiled header from inside another header, so long as
13860 there are no C tokens before the @code{#include}.
13863 The precompiled header file must be produced for the same language as
13864 the current compilation. You can't use a C precompiled header for a C++
13868 The precompiled header file must have been produced by the same compiler
13869 binary as the current compilation is using.
13872 Any macros defined before the precompiled header is included must
13873 either be defined in the same way as when the precompiled header was
13874 generated, or must not affect the precompiled header, which usually
13875 means that they don't appear in the precompiled header at all.
13877 The @option{-D} option is one way to define a macro before a
13878 precompiled header is included; using a @code{#define} can also do it.
13879 There are also some options that define macros implicitly, like
13880 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
13883 @item If debugging information is output when using the precompiled
13884 header, using @option{-g} or similar, the same kind of debugging information
13885 must have been output when building the precompiled header. However,
13886 a precompiled header built using @option{-g} can be used in a compilation
13887 when no debugging information is being output.
13889 @item The same @option{-m} options must generally be used when building
13890 and using the precompiled header. @xref{Submodel Options},
13891 for any cases where this rule is relaxed.
13893 @item Each of the following options must be the same when building and using
13894 the precompiled header:
13896 @gccoptlist{-fexceptions -funit-at-a-time}
13899 Some other command-line options starting with @option{-f},
13900 @option{-p}, or @option{-O} must be defined in the same way as when
13901 the precompiled header was generated. At present, it's not clear
13902 which options are safe to change and which are not; the safest choice
13903 is to use exactly the same options when generating and using the
13904 precompiled header. The following are known to be safe:
13906 @gccoptlist{-fmessage-length= -fpreprocessed
13907 -fsched-interblock -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
13908 -fsched-verbose=<number> -fschedule-insns -fvisibility=
13913 For all of these except the last, the compiler will automatically
13914 ignore the precompiled header if the conditions aren't met. If you
13915 find an option combination that doesn't work and doesn't cause the
13916 precompiled header to be ignored, please consider filing a bug report,
13919 If you do use differing options when generating and using the
13920 precompiled header, the actual behavior will be a mixture of the
13921 behavior for the options. For instance, if you use @option{-g} to
13922 generate the precompiled header but not when using it, you may or may
13923 not get debugging information for routines in the precompiled header.
13925 @node Running Protoize
13926 @section Running Protoize
13928 The program @code{protoize} is an optional part of GCC@. You can use
13929 it to add prototypes to a program, thus converting the program to ISO
13930 C in one respect. The companion program @code{unprotoize} does the
13931 reverse: it removes argument types from any prototypes that are found.
13933 When you run these programs, you must specify a set of source files as
13934 command line arguments. The conversion programs start out by compiling
13935 these files to see what functions they define. The information gathered
13936 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
13938 After scanning comes actual conversion. The specified files are all
13939 eligible to be converted; any files they include (whether sources or
13940 just headers) are eligible as well.
13942 But not all the eligible files are converted. By default,
13943 @code{protoize} and @code{unprotoize} convert only source and header
13944 files in the current directory. You can specify additional directories
13945 whose files should be converted with the @option{-d @var{directory}}
13946 option. You can also specify particular files to exclude with the
13947 @option{-x @var{file}} option. A file is converted if it is eligible, its
13948 directory name matches one of the specified directory names, and its
13949 name within the directory has not been excluded.
13951 Basic conversion with @code{protoize} consists of rewriting most
13952 function definitions and function declarations to specify the types of
13953 the arguments. The only ones not rewritten are those for varargs
13956 @code{protoize} optionally inserts prototype declarations at the
13957 beginning of the source file, to make them available for any calls that
13958 precede the function's definition. Or it can insert prototype
13959 declarations with block scope in the blocks where undeclared functions
13962 Basic conversion with @code{unprotoize} consists of rewriting most
13963 function declarations to remove any argument types, and rewriting
13964 function definitions to the old-style pre-ISO form.
13966 Both conversion programs print a warning for any function declaration or
13967 definition that they can't convert. You can suppress these warnings
13970 The output from @code{protoize} or @code{unprotoize} replaces the
13971 original source file. The original file is renamed to a name ending
13972 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
13973 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
13974 for DOS) file already exists, then the source file is simply discarded.
13976 @code{protoize} and @code{unprotoize} both depend on GCC itself to
13977 scan the program and collect information about the functions it uses.
13978 So neither of these programs will work until GCC is installed.
13980 Here is a table of the options you can use with @code{protoize} and
13981 @code{unprotoize}. Each option works with both programs unless
13985 @item -B @var{directory}
13986 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
13987 usual directory (normally @file{/usr/local/lib}). This file contains
13988 prototype information about standard system functions. This option
13989 applies only to @code{protoize}.
13991 @item -c @var{compilation-options}
13992 Use @var{compilation-options} as the options when running @command{gcc} to
13993 produce the @samp{.X} files. The special option @option{-aux-info} is
13994 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
13996 Note that the compilation options must be given as a single argument to
13997 @code{protoize} or @code{unprotoize}. If you want to specify several
13998 @command{gcc} options, you must quote the entire set of compilation options
13999 to make them a single word in the shell.
14001 There are certain @command{gcc} arguments that you cannot use, because they
14002 would produce the wrong kind of output. These include @option{-g},
14003 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
14004 the @var{compilation-options}, they are ignored.
14007 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
14008 systems) instead of @samp{.c}. This is convenient if you are converting
14009 a C program to C++. This option applies only to @code{protoize}.
14012 Add explicit global declarations. This means inserting explicit
14013 declarations at the beginning of each source file for each function
14014 that is called in the file and was not declared. These declarations
14015 precede the first function definition that contains a call to an
14016 undeclared function. This option applies only to @code{protoize}.
14018 @item -i @var{string}
14019 Indent old-style parameter declarations with the string @var{string}.
14020 This option applies only to @code{protoize}.
14022 @code{unprotoize} converts prototyped function definitions to old-style
14023 function definitions, where the arguments are declared between the
14024 argument list and the initial @samp{@{}. By default, @code{unprotoize}
14025 uses five spaces as the indentation. If you want to indent with just
14026 one space instead, use @option{-i " "}.
14029 Keep the @samp{.X} files. Normally, they are deleted after conversion
14033 Add explicit local declarations. @code{protoize} with @option{-l} inserts
14034 a prototype declaration for each function in each block which calls the
14035 function without any declaration. This option applies only to
14039 Make no real changes. This mode just prints information about the conversions
14040 that would have been done without @option{-n}.
14043 Make no @samp{.save} files. The original files are simply deleted.
14044 Use this option with caution.
14046 @item -p @var{program}
14047 Use the program @var{program} as the compiler. Normally, the name
14048 @file{gcc} is used.
14051 Work quietly. Most warnings are suppressed.
14054 Print the version number, just like @option{-v} for @command{gcc}.
14057 If you need special compiler options to compile one of your program's
14058 source files, then you should generate that file's @samp{.X} file
14059 specially, by running @command{gcc} on that source file with the
14060 appropriate options and the option @option{-aux-info}. Then run
14061 @code{protoize} on the entire set of files. @code{protoize} will use
14062 the existing @samp{.X} file because it is newer than the source file.
14066 gcc -Dfoo=bar file1.c -aux-info file1.X
14071 You need to include the special files along with the rest in the
14072 @code{protoize} command, even though their @samp{.X} files already
14073 exist, because otherwise they won't get converted.
14075 @xref{Protoize Caveats}, for more information on how to use
14076 @code{protoize} successfully.