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 warn missing prototype 6261539
229 -Wmissing-prototypes @gol
230 @c APPLE LOCAL -Wmost
231 -Wmost (APPLE ONLY) @gol
232 -Wno-multichar -Wnonnull -Wno-overflow @gol
233 -Woverlength-strings -Wpacked -Wpadded @gol
234 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
235 -Wredundant-decls @gol
236 -Wreturn-type -Wsequence-point -Wshadow @gol
237 -Wsign-compare -Wstack-protector @gol
238 -Wstrict-aliasing -Wstrict-aliasing=2 @gol
239 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
240 -Wswitch -Wswitch-default -Wswitch-enum @gol
241 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
242 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
243 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
244 -Wunused-value -Wunused-variable @gol
245 -Wvariadic-macros -Wvla @gol
246 -Wvolatile-register-var -Wwrite-strings}
248 @item C-only Warning Options
249 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
250 @c APPLE LOCAL warn missing prototype 6261539
251 -Wnested-externs -Wold-style-definition @gol
252 -Wstrict-prototypes -Wtraditional @gol
253 -Wdeclaration-after-statement -Wpointer-sign}
255 @item Debugging Options
256 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
257 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
258 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
259 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
260 -fdump-ipa-all -fdump-ipa-cgraph @gol
262 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
263 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
264 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
267 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
268 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
269 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
270 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
271 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
272 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
273 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
274 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
275 -fdump-tree-nrv -fdump-tree-vect @gol
276 -fdump-tree-sink @gol
277 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
278 -fdump-tree-salias @gol
279 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
280 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
281 -ftree-vectorizer-verbose=@var{n} @gol
282 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
283 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
284 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
285 -fmem-report -fprofile-arcs @gol
286 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
287 -ftest-coverage -ftime-report -fvar-tracking @gol
288 -g -g@var{level} -gcoff -gdwarf-2 @gol
289 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
290 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
291 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
292 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
293 -print-multi-directory -print-multi-lib @gol
294 -print-prog-name=@var{program} -print-search-dirs -Q @gol
297 @item Optimization Options
298 @xref{Optimize Options,,Options that Control Optimization}.
299 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
300 -falign-labels=@var{n} -falign-loops=@var{n} @gol
301 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
302 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
303 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
304 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
305 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
306 -fdelayed-branch -fdelete-null-pointer-checks -fearly-inlining @gol
307 -fexpensive-optimizations -ffast-math -ffloat-store @gol
308 -fforce-addr -ffunction-sections @gol
309 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
310 -fcrossjumping -fif-conversion -fif-conversion2 @gol
311 -finline-functions -finline-functions-called-once @gol
312 -finline-limit=@var{n} -fkeep-inline-functions @gol
313 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
314 -fmodulo-sched -fno-branch-count-reg @gol
315 -fno-default-inline -fno-defer-pop -fmove-loop-invariants @gol
316 -fno-function-cse -fno-guess-branch-probability @gol
317 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
318 -funsafe-math-optimizations -funsafe-loop-optimizations -ffinite-math-only @gol
319 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
320 -fomit-frame-pointer -foptimize-register-move @gol
321 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
322 -fprofile-generate -fprofile-use @gol
323 -fregmove -frename-registers @gol
324 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
325 -frerun-cse-after-loop @gol
326 -frounding-math -frtl-abstract-sequences @gol
327 -fschedule-insns -fschedule-insns2 @gol
328 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
329 -fsched-spec-load-dangerous @gol
330 -fsched-stalled-insns=@var{n} -fsched-stalled-insns-dep=@var{n} @gol
331 -fsched2-use-superblocks @gol
332 -fsched2-use-traces -fsee -freschedule-modulo-scheduled-loops @gol
333 -fsection-anchors -fsignaling-nans -fsingle-precision-constant @gol
334 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
335 -fstrict-aliasing -fstrict-overflow -ftracer -fthread-jumps @gol
336 -funroll-all-loops -funroll-loops -fpeel-loops @gol
337 -fsplit-ivs-in-unroller -funswitch-loops @gol
338 -fvariable-expansion-in-unroller @gol
339 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
340 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
341 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
342 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
343 -ftree-vect-loop-version -ftree-salias -fipa-pta -fweb @gol
344 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
345 --param @var{name}=@var{value}
346 -O -O0 -O1 -O2 -O3 -Os}
348 @item Preprocessor Options
349 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
350 @gccoptlist{-A@var{question}=@var{answer} @gol
351 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
352 -C -dD -dI -dM -dN @gol
353 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
354 -idirafter @var{dir} @gol
355 -include @var{file} -imacros @var{file} @gol
356 -iprefix @var{file} -iwithprefix @var{dir} @gol
357 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
358 -imultilib @var{dir} -isysroot @var{dir} @gol
359 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
360 -P -fworking-directory -remap @gol
361 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
362 -Xpreprocessor @var{option}}
364 @item Assembler Option
365 @xref{Assembler Options,,Passing Options to the Assembler}.
366 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
369 @xref{Link Options,,Options for Linking}.
370 @gccoptlist{@var{object-file-name} -l@var{library} @gol
371 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
372 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
373 -Wl,@var{option} -Xlinker @var{option} @gol
376 @item Directory Options
377 @xref{Directory Options,,Options for Directory Search}.
378 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
379 -specs=@var{file} -I- --sysroot=@var{dir}}
382 @c I wrote this xref this way to avoid overfull hbox. -- rms
383 @xref{Target Options}.
384 @gccoptlist{-V @var{version} -b @var{machine}}
386 @item Machine Dependent Options
387 @xref{Submodel Options,,Hardware Models and Configurations}.
388 @c This list is ordered alphanumerically by subsection name.
389 @c Try and put the significant identifier (CPU or system) first,
390 @c so users have a clue at guessing where the ones they want will be.
393 @gccoptlist{-EB -EL @gol
394 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
395 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
398 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
399 -mabi=@var{name} @gol
400 -mapcs-stack-check -mno-apcs-stack-check @gol
401 -mapcs-float -mno-apcs-float @gol
402 -mapcs-reentrant -mno-apcs-reentrant @gol
403 -msched-prolog -mno-sched-prolog @gol
404 -mlittle-endian -mbig-endian -mwords-little-endian @gol
405 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
406 -mthumb-interwork -mno-thumb-interwork @gol
407 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
408 -mstructure-size-boundary=@var{n} @gol
409 -mabort-on-noreturn @gol
410 -mlong-calls -mno-long-calls @gol
411 -msingle-pic-base -mno-single-pic-base @gol
412 -mpic-register=@var{reg} @gol
413 -mnop-fun-dllimport @gol
414 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
415 -mpoke-function-name @gol
417 -mtpcs-frame -mtpcs-leaf-frame @gol
418 -mcaller-super-interworking -mcallee-super-interworking @gol
422 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
423 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
425 @emph{Blackfin Options}
426 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
427 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
428 -mlow-64k -mno-low64k -mid-shared-library @gol
429 -mno-id-shared-library -mshared-library-id=@var{n} @gol
430 -mlong-calls -mno-long-calls}
433 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
434 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
435 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
436 -mstack-align -mdata-align -mconst-align @gol
437 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
438 -melf -maout -melinux -mlinux -sim -sim2 @gol
439 -mmul-bug-workaround -mno-mul-bug-workaround}
442 @gccoptlist{-mmac -mpush-args}
444 @emph{Darwin Options}
445 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
446 -arch_only -bind_at_load -bundle -bundle_loader @gol
447 -client_name -compatibility_version -current_version @gol
449 -dependency-file -dylib_file -dylinker_install_name @gol
450 -dynamic -dynamiclib -exported_symbols_list @gol
451 -filelist -flat_namespace -force_cpusubtype_ALL @gol
452 -force_flat_namespace -headerpad_max_install_names @gol
453 -image_base -init -install_name -keep_private_externs @gol
454 -multi_module -multiply_defined -multiply_defined_unused @gol
455 -noall_load -no_dead_strip_inits_and_terms @gol
456 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
457 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
458 -private_bundle -read_only_relocs -sectalign @gol
459 -sectobjectsymbols -whyload -seg1addr @gol
460 -sectcreate -sectobjectsymbols -sectorder @gol
461 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
462 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
463 -segprot -segs_read_only_addr -segs_read_write_addr @gol
464 -single_module -static -sub_library -sub_umbrella @gol
465 -twolevel_namespace -umbrella -undefined @gol
466 -unexported_symbols_list -weak_reference_mismatches @gol
467 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
468 -mkernel -mone-byte-bool}
470 @emph{DEC Alpha Options}
471 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
472 -mieee -mieee-with-inexact -mieee-conformant @gol
473 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
474 -mtrap-precision=@var{mode} -mbuild-constants @gol
475 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
476 -mbwx -mmax -mfix -mcix @gol
477 -mfloat-vax -mfloat-ieee @gol
478 -mexplicit-relocs -msmall-data -mlarge-data @gol
479 -msmall-text -mlarge-text @gol
480 -mmemory-latency=@var{time}}
482 @emph{DEC Alpha/VMS Options}
483 @gccoptlist{-mvms-return-codes}
486 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
487 -mhard-float -msoft-float @gol
488 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
489 -mdouble -mno-double @gol
490 -mmedia -mno-media -mmuladd -mno-muladd @gol
491 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
492 -mlinked-fp -mlong-calls -malign-labels @gol
493 -mlibrary-pic -macc-4 -macc-8 @gol
494 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
495 -moptimize-membar -mno-optimize-membar @gol
496 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
497 -mvliw-branch -mno-vliw-branch @gol
498 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
499 -mno-nested-cond-exec -mtomcat-stats @gol
503 @emph{GNU/Linux Options}
504 @gccoptlist{-muclibc}
506 @emph{H8/300 Options}
507 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
510 @gccoptlist{-march=@var{architecture-type} @gol
511 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
512 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
513 -mfixed-range=@var{register-range} @gol
514 -mjump-in-delay -mlinker-opt -mlong-calls @gol
515 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
516 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
517 -mno-jump-in-delay -mno-long-load-store @gol
518 -mno-portable-runtime -mno-soft-float @gol
519 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
520 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
521 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
522 -munix=@var{unix-std} -nolibdld -static -threads}
524 @emph{i386 and x86-64 Options}
525 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
526 -mfpmath=@var{unit} @gol
527 -masm=@var{dialect} -mno-fancy-math-387 @gol
528 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
529 -mno-wide-multiply -mrtd -malign-double @gol
530 -mpreferred-stack-boundary=@var{num} @gol
531 -mmmx -msse -msse2 -msse3 -mssse3 -msse4a -m3dnow -mpopcnt -mabm -maes @gol
532 -mthreads -mno-align-stringops -minline-all-stringops @gol
533 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
534 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
536 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
537 -mcmodel=@var{code-model} @gol
538 -m32 -m64 -mlarge-data-threshold=@var{num}}
541 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
542 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
543 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
544 -minline-float-divide-max-throughput @gol
545 -minline-int-divide-min-latency @gol
546 -minline-int-divide-max-throughput @gol
547 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
548 -mno-dwarf2-asm -mearly-stop-bits @gol
549 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
550 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
551 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
552 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
553 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
554 -mno-sched-prefer-non-data-spec-insns @gol
555 -mno-sched-prefer-non-control-spec-insns @gol
556 -mno-sched-count-spec-in-critical-path}
558 @emph{M32R/D Options}
559 @gccoptlist{-m32r2 -m32rx -m32r @gol
561 -malign-loops -mno-align-loops @gol
562 -missue-rate=@var{number} @gol
563 -mbranch-cost=@var{number} @gol
564 -mmodel=@var{code-size-model-type} @gol
565 -msdata=@var{sdata-type} @gol
566 -mno-flush-func -mflush-func=@var{name} @gol
567 -mno-flush-trap -mflush-trap=@var{number} @gol
571 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
573 @emph{M680x0 Options}
574 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
575 -m68060 -mcpu32 -m5200 -mcfv4e -m68881 -mbitfield @gol
576 -mc68000 -mc68020 @gol
577 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
578 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
579 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
581 @emph{M68hc1x Options}
582 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
583 -mauto-incdec -minmax -mlong-calls -mshort @gol
584 -msoft-reg-count=@var{count}}
587 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
588 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
589 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
590 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
591 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
594 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
595 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
596 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
597 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
598 -mfp32 -mfp64 -mhard-float -msoft-float @gol
599 -msingle-float -mdouble-float -mdsp -mpaired-single -mips3d @gol
600 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
601 -G@var{num} -membedded-data -mno-embedded-data @gol
602 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
603 -msplit-addresses -mno-split-addresses @gol
604 -mexplicit-relocs -mno-explicit-relocs @gol
605 -mcheck-zero-division -mno-check-zero-division @gol
606 -mdivide-traps -mdivide-breaks @gol
607 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
608 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
609 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
610 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
611 -mfix-sb1 -mno-fix-sb1 @gol
612 -mflush-func=@var{func} -mno-flush-func @gol
613 -mbranch-likely -mno-branch-likely @gol
614 -mfp-exceptions -mno-fp-exceptions @gol
615 -mvr4130-align -mno-vr4130-align}
618 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
619 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
620 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
621 -mno-base-addresses -msingle-exit -mno-single-exit}
623 @emph{MN10300 Options}
624 @gccoptlist{-mmult-bug -mno-mult-bug @gol
625 -mam33 -mno-am33 @gol
626 -mam33-2 -mno-am33-2 @gol
627 -mreturn-pointer-on-d0 @gol
631 @gccoptlist{-mno-crt0 -mbacc -msim @gol
632 -march=@var{cpu-type} }
634 @emph{PDP-11 Options}
635 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
636 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
637 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
638 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
639 -mbranch-expensive -mbranch-cheap @gol
640 -msplit -mno-split -munix-asm -mdec-asm}
642 @emph{PowerPC Options}
643 See RS/6000 and PowerPC Options.
645 @emph{RS/6000 and PowerPC Options}
646 @gccoptlist{-mcpu=@var{cpu-type} @gol
647 -mtune=@var{cpu-type} @gol
648 -mpower -mno-power -mpower2 -mno-power2 @gol
649 -mpowerpc -mpowerpc64 -mno-powerpc @gol
650 -maltivec -mno-altivec @gol
651 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
652 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
653 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
654 -mnew-mnemonics -mold-mnemonics @gol
655 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
656 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
657 -malign-power -malign-natural @gol
658 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
659 -mstring -mno-string -mupdate -mno-update @gol
660 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
661 -mstrict-align -mno-strict-align -mrelocatable @gol
662 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
663 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
664 -mdynamic-no-pic -maltivec -mswdiv @gol
665 -mprioritize-restricted-insns=@var{priority} @gol
666 -msched-costly-dep=@var{dependence_type} @gol
667 -minsert-sched-nops=@var{scheme} @gol
668 -mcall-sysv -mcall-netbsd @gol
669 -maix-struct-return -msvr4-struct-return @gol
670 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
671 -misel -mno-isel @gol
672 -misel=yes -misel=no @gol
674 -mspe=yes -mspe=no @gol
675 -mvrsave -mno-vrsave @gol
676 -mmulhw -mno-mulhw @gol
677 -mdlmzb -mno-dlmzb @gol
678 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
679 -mprototype -mno-prototype @gol
680 -msim -mmvme -mads -myellowknife -memb -msdata @gol
681 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
683 @emph{S/390 and zSeries Options}
684 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
685 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
686 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
687 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
688 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
689 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
690 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
693 @gccoptlist{-meb -mel @gol
697 -mscore5 -mscore5u -mscore7 -mscore7d}
700 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
701 -m4-nofpu -m4-single-only -m4-single -m4 @gol
702 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
703 -m5-64media -m5-64media-nofpu @gol
704 -m5-32media -m5-32media-nofpu @gol
705 -m5-compact -m5-compact-nofpu @gol
706 -mb -ml -mdalign -mrelax @gol
707 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
708 -mieee -misize -mpadstruct -mspace @gol
709 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
710 -mdivsi3_libfunc=@var{name} @gol
711 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
715 @gccoptlist{-mcpu=@var{cpu-type} @gol
716 -mtune=@var{cpu-type} @gol
717 -mcmodel=@var{code-model} @gol
718 -m32 -m64 -mapp-regs -mno-app-regs @gol
719 -mfaster-structs -mno-faster-structs @gol
720 -mfpu -mno-fpu -mhard-float -msoft-float @gol
721 -mhard-quad-float -msoft-quad-float @gol
722 -mimpure-text -mno-impure-text -mlittle-endian @gol
723 -mstack-bias -mno-stack-bias @gol
724 -munaligned-doubles -mno-unaligned-doubles @gol
725 -mv8plus -mno-v8plus -mvis -mno-vis
726 -threads -pthreads -pthread}
728 @emph{System V Options}
729 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
731 @emph{TMS320C3x/C4x Options}
732 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
733 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
734 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
735 -mparallel-insns -mparallel-mpy -mpreserve-float}
738 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
739 -mprolog-function -mno-prolog-function -mspace @gol
740 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
741 -mapp-regs -mno-app-regs @gol
742 -mdisable-callt -mno-disable-callt @gol
748 @gccoptlist{-mg -mgnu -munix}
750 @emph{x86-64 Options}
751 See i386 and x86-64 Options.
753 @emph{Xstormy16 Options}
756 @emph{Xtensa Options}
757 @gccoptlist{-mconst16 -mno-const16 @gol
758 -mfused-madd -mno-fused-madd @gol
759 -mtext-section-literals -mno-text-section-literals @gol
760 -mtarget-align -mno-target-align @gol
761 -mlongcalls -mno-longcalls}
763 @emph{zSeries Options}
764 See S/390 and zSeries Options.
766 @item Code Generation Options
767 @xref{Code Gen Options,,Options for Code Generation Conventions}.
768 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
769 -ffixed-@var{reg} -fexceptions @gol
770 -fnon-call-exceptions -funwind-tables @gol
771 -fasynchronous-unwind-tables @gol
772 -finhibit-size-directive -finstrument-functions @gol
773 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
774 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
775 -fno-common -fno-ident @gol
776 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
777 -fno-jump-tables @gol
778 -freg-struct-return -fshort-enums @gol
779 -fshort-double -fshort-wchar @gol
780 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
781 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
782 -fargument-alias -fargument-noalias @gol
783 -fargument-noalias-global -fargument-noalias-anything
784 -fleading-underscore -ftls-model=@var{model} @gol
785 -ftrapv -fwrapv -fbounds-check @gol
790 * Overall Options:: Controlling the kind of output:
791 an executable, object files, assembler files,
792 or preprocessed source.
793 * C Dialect Options:: Controlling the variant of C language compiled.
794 * C++ Dialect Options:: Variations on C++.
795 * Language Independent Options:: Controlling how diagnostics should be
797 * Warning Options:: How picky should the compiler be?
798 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
799 * Optimize Options:: How much optimization?
800 * Preprocessor Options:: Controlling header files and macro definitions.
801 Also, getting dependency information for Make.
802 * Assembler Options:: Passing options to the assembler.
803 * Link Options:: Specifying libraries and so on.
804 * Directory Options:: Where to find header files and libraries.
805 Where to find the compiler executable files.
806 * Spec Files:: How to pass switches to sub-processes.
807 * Target Options:: Running a cross-compiler, or an old version of GCC.
810 @node Overall Options
811 @section Options Controlling the Kind of Output
813 Compilation can involve up to four stages: preprocessing, compilation
814 proper, assembly and linking, always in that order. GCC is capable of
815 preprocessing and compiling several files either into several
816 assembler input files, or into one assembler input file; then each
817 assembler input file produces an object file, and linking combines all
818 the object files (those newly compiled, and those specified as input)
819 into an executable file.
821 @cindex file name suffix
822 For any given input file, the file name suffix determines what kind of
827 C source code which must be preprocessed.
830 C source code which should not be preprocessed.
833 C++ source code which should not be preprocessed.
836 C, or C++ header file to be turned into a precompiled header.
840 @itemx @var{file}.cxx
841 @itemx @var{file}.cpp
842 @itemx @var{file}.CPP
843 @itemx @var{file}.c++
845 C++ source code which must be preprocessed. Note that in @samp{.cxx},
846 the last two letters must both be literally @samp{x}. Likewise,
847 @samp{.C} refers to a literal capital C@.
851 C++ header file to be turned into a precompiled header.
854 @itemx @var{file}.for
855 @itemx @var{file}.FOR
856 Fixed form Fortran source code which should not be preprocessed.
859 @itemx @var{file}.fpp
860 @itemx @var{file}.FPP
861 Fixed form Fortran source code which must be preprocessed (with the traditional
865 @itemx @var{file}.f95
866 Free form Fortran source code which should not be preprocessed.
869 @itemx @var{file}.F95
870 Free form Fortran source code which must be preprocessed (with the
871 traditional preprocessor).
873 @c FIXME: Descriptions of Java file types.
880 Ada source code file which contains a library unit declaration (a
881 declaration of a package, subprogram, or generic, or a generic
882 instantiation), or a library unit renaming declaration (a package,
883 generic, or subprogram renaming declaration). Such files are also
886 @itemx @var{file}.adb
887 Ada source code file containing a library unit body (a subprogram or
888 package body). Such files are also called @dfn{bodies}.
890 @c GCC also knows about some suffixes for languages not yet included:
901 Assembler code which must be preprocessed.
904 An object file to be fed straight into linking.
905 Any file name with no recognized suffix is treated this way.
909 You can specify the input language explicitly with the @option{-x} option:
912 @item -x @var{language}
913 Specify explicitly the @var{language} for the following input files
914 (rather than letting the compiler choose a default based on the file
915 name suffix). This option applies to all following input files until
916 the next @option{-x} option. Possible values for @var{language} are:
918 c c-header c-cpp-output
919 c++ c++-header c++-cpp-output
920 assembler assembler-with-cpp
928 Turn off any specification of a language, so that subsequent files are
929 handled according to their file name suffixes (as they are if @option{-x}
930 has not been used at all).
932 @item -pass-exit-codes
933 @opindex pass-exit-codes
934 Normally the @command{gcc} program will exit with the code of 1 if any
935 phase of the compiler returns a non-success return code. If you specify
936 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
937 numerically highest error produced by any phase that returned an error
938 indication. The C, C++, and Fortran frontends return 4, if an internal
939 compiler error is encountered.
942 If you only want some of the stages of compilation, you can use
943 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
944 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
945 @command{gcc} is to stop. Note that some combinations (for example,
946 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
951 Compile or assemble the source files, but do not link. The linking
952 stage simply is not done. The ultimate output is in the form of an
953 object file for each source file.
955 By default, the object file name for a source file is made by replacing
956 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
958 Unrecognized input files, not requiring compilation or assembly, are
963 Stop after the stage of compilation proper; do not assemble. The output
964 is in the form of an assembler code file for each non-assembler input
967 By default, the assembler file name for a source file is made by
968 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
970 Input files that don't require compilation are ignored.
974 Stop after the preprocessing stage; do not run the compiler proper. The
975 output is in the form of preprocessed source code, which is sent to the
978 Input files which don't require preprocessing are ignored.
980 @cindex output file option
983 Place output in file @var{file}. This applies regardless to whatever
984 sort of output is being produced, whether it be an executable file,
985 an object file, an assembler file or preprocessed C code.
987 If @option{-o} is not specified, the default is to put an executable
988 file in @file{a.out}, the object file for
989 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
990 assembler file in @file{@var{source}.s}, a precompiled header file in
991 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
996 Print (on standard error output) the commands executed to run the stages
997 of compilation. Also print the version number of the compiler driver
998 program and of the preprocessor and the compiler proper.
1002 Like @option{-v} except the commands are not executed and all command
1003 arguments are quoted. This is useful for shell scripts to capture the
1004 driver-generated command lines.
1008 Use pipes rather than temporary files for communication between the
1009 various stages of compilation. This fails to work on some systems where
1010 the assembler is unable to read from a pipe; but the GNU assembler has
1015 If you are compiling multiple source files, this option tells the driver
1016 to pass all the source files to the compiler at once (for those
1017 languages for which the compiler can handle this). This will allow
1018 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1019 language for which this is supported is C@. If you pass source files for
1020 multiple languages to the driver, using this option, the driver will invoke
1021 the compiler(s) that support IMA once each, passing each compiler all the
1022 source files appropriate for it. For those languages that do not support
1023 IMA this option will be ignored, and the compiler will be invoked once for
1024 each source file in that language. If you use this option in conjunction
1025 with @option{-save-temps}, the compiler will generate multiple
1027 (one for each source file), but only one (combined) @file{.o} or
1032 Print (on the standard output) a description of the command line options
1033 understood by @command{gcc}. If the @option{-v} option is also specified
1034 then @option{--help} will also be passed on to the various processes
1035 invoked by @command{gcc}, so that they can display the command line options
1036 they accept. If the @option{-Wextra} option is also specified then command
1037 line options which have no documentation associated with them will also
1041 @opindex target-help
1042 Print (on the standard output) a description of target specific command
1043 line options for each tool.
1047 Display the version number and copyrights of the invoked GCC@.
1049 @include @value{srcdir}/../libiberty/at-file.texi
1053 @section Compiling C++ Programs
1055 @cindex suffixes for C++ source
1056 @cindex C++ source file suffixes
1057 C++ source files conventionally use one of the suffixes @samp{.C},
1058 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1059 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1060 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1061 files with these names and compiles them as C++ programs even if you
1062 call the compiler the same way as for compiling C programs (usually
1063 with the name @command{gcc}).
1067 However, the use of @command{gcc} does not add the C++ library.
1068 @command{g++} is a program that calls GCC and treats @samp{.c},
1069 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1070 files unless @option{-x} is used, and automatically specifies linking
1071 against the C++ library. This program is also useful when
1072 precompiling a C header file with a @samp{.h} extension for use in C++
1073 compilations. On many systems, @command{g++} is also installed with
1074 the name @command{c++}.
1076 @cindex invoking @command{g++}
1077 When you compile C++ programs, you may specify many of the same
1078 command-line options that you use for compiling programs in any
1079 language; or command-line options meaningful for C and related
1080 languages; or options that are meaningful only for C++ programs.
1081 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1082 explanations of options for languages related to C@.
1083 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1084 explanations of options that are meaningful only for C++ programs.
1086 @node C Dialect Options
1087 @section Options Controlling C Dialect
1088 @cindex dialect options
1089 @cindex language dialect options
1090 @cindex options, dialect
1092 The following options control the dialect of C (or languages derived
1093 from C, such as C++) that the compiler accepts:
1096 @cindex ANSI support
1100 In C mode, support all ISO C90 programs. In C++ mode,
1101 remove GNU extensions that conflict with ISO C++.
1103 This turns off certain features of GCC that are incompatible with ISO
1104 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1105 such as the @code{asm} and @code{typeof} keywords, and
1106 predefined macros such as @code{unix} and @code{vax} that identify the
1107 type of system you are using. It also enables the undesirable and
1108 rarely used ISO trigraph feature. For the C compiler,
1109 it disables recognition of C++ style @samp{//} comments as well as
1110 the @code{inline} keyword.
1112 The alternate keywords @code{__asm__}, @code{__extension__},
1113 @code{__inline__} and @code{__typeof__} continue to work despite
1114 @option{-ansi}. You would not want to use them in an ISO C program, of
1115 course, but it is useful to put them in header files that might be included
1116 in compilations done with @option{-ansi}. Alternate predefined macros
1117 such as @code{__unix__} and @code{__vax__} are also available, with or
1118 without @option{-ansi}.
1120 The @option{-ansi} option does not cause non-ISO programs to be
1121 rejected gratuitously. For that, @option{-pedantic} is required in
1122 addition to @option{-ansi}. @xref{Warning Options}.
1124 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1125 option is used. Some header files may notice this macro and refrain
1126 from declaring certain functions or defining certain macros that the
1127 ISO standard doesn't call for; this is to avoid interfering with any
1128 programs that might use these names for other things.
1130 Functions which would normally be built in but do not have semantics
1131 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1132 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1133 built-in functions provided by GCC}, for details of the functions
1138 Determine the language standard. This option is currently only
1139 supported when compiling C or C++. A value for this option must be
1140 provided; possible values are
1145 ISO C90 (same as @option{-ansi}).
1147 @item iso9899:199409
1148 ISO C90 as modified in amendment 1.
1154 ISO C99. Note that this standard is not yet fully supported; see
1155 @w{@uref{http://gcc.gnu.org/gcc-4.2/c99status.html}} for more information. The
1156 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1159 Default, ISO C90 plus GNU extensions (including some C99 features).
1163 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1164 this will become the default. The name @samp{gnu9x} is deprecated.
1167 The 1998 ISO C++ standard plus amendments.
1170 The same as @option{-std=c++98} plus GNU extensions. This is the
1171 default for C++ code.
1174 Even when this option is not specified, you can still use some of the
1175 features of newer standards in so far as they do not conflict with
1176 previous C standards. For example, you may use @code{__restrict__} even
1177 when @option{-std=c99} is not specified.
1179 The @option{-std} options specifying some version of ISO C have the same
1180 effects as @option{-ansi}, except that features that were not in ISO C90
1181 but are in the specified version (for example, @samp{//} comments and
1182 the @code{inline} keyword in ISO C99) are not disabled.
1184 @xref{Standards,,Language Standards Supported by GCC}, for details of
1185 these standard versions.
1187 @item -fgnu89-inline
1188 @opindex fgnu89-inline
1189 The option @option{-fgnu89-inline} tells GCC to use the traditional
1190 GNU semantics for @code{inline} functions when in C99 mode.
1191 @xref{Inline,,An Inline Function is As Fast As a Macro}. Using this
1192 option is roughly equivalent to adding the @code{gnu_inline} function
1193 attribute to all inline functions (@pxref{Function Attributes}).
1195 This option is accepted by GCC versions 4.1.3 and up. In GCC versions
1196 prior to 4.3, C99 inline semantics are not supported, and thus this
1197 option is effectively assumed to be present regardless of whether or not
1198 it is specified; the only effect of specifying it explicitly is to
1199 disable warnings about using inline functions in C99 mode. Likewise,
1200 the option @option{-fno-gnu89-inline} is not supported in versions of
1201 GCC before 4.3. It will be supported only in C99 or gnu99 mode, not in
1204 The preprocesor macros @code{__GNUC_GNU_INLINE__} and
1205 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1206 in effect for @code{inline} functions. @xref{Common Predefined
1207 Macros,,,cpp,The C Preprocessor}.
1209 @item -aux-info @var{filename}
1211 Output to the given filename prototyped declarations for all functions
1212 declared and/or defined in a translation unit, including those in header
1213 files. This option is silently ignored in any language other than C@.
1215 Besides declarations, the file indicates, in comments, the origin of
1216 each declaration (source file and line), whether the declaration was
1217 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1218 @samp{O} for old, respectively, in the first character after the line
1219 number and the colon), and whether it came from a declaration or a
1220 definition (@samp{C} or @samp{F}, respectively, in the following
1221 character). In the case of function definitions, a K&R-style list of
1222 arguments followed by their declarations is also provided, inside
1223 comments, after the declaration.
1227 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1228 keyword, so that code can use these words as identifiers. You can use
1229 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1230 instead. @option{-ansi} implies @option{-fno-asm}.
1232 In C++, this switch only affects the @code{typeof} keyword, since
1233 @code{asm} and @code{inline} are standard keywords. You may want to
1234 use the @option{-fno-gnu-keywords} flag instead, which has the same
1235 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1236 switch only affects the @code{asm} and @code{typeof} keywords, since
1237 @code{inline} is a standard keyword in ISO C99.
1239 @c APPLE LOCAL begin blocks 7205047 5811887
1242 Disable the use of blocks. In @option{-std=c99} mode, blocks are
1243 turned off by default. @option{-fblocks} can be used to re-enable the
1244 feature, if off. Runtime support for blocks first appeared in Mac OS
1245 X 10.6. When targeting 10.6 (see @option{-mmacosx-version-min}) and
1246 later, the extension is on by default.
1247 @c APPLE LOCAL end blocks 7205047 5811887
1250 @itemx -fno-builtin-@var{function}
1251 @opindex fno-builtin
1252 @cindex built-in functions
1253 Don't recognize built-in functions that do not begin with
1254 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1255 functions provided by GCC}, for details of the functions affected,
1256 including those which are not built-in functions when @option{-ansi} or
1257 @option{-std} options for strict ISO C conformance are used because they
1258 do not have an ISO standard meaning.
1260 GCC normally generates special code to handle certain built-in functions
1261 more efficiently; for instance, calls to @code{alloca} may become single
1262 instructions that adjust the stack directly, and calls to @code{memcpy}
1263 may become inline copy loops. The resulting code is often both smaller
1264 and faster, but since the function calls no longer appear as such, you
1265 cannot set a breakpoint on those calls, nor can you change the behavior
1266 of the functions by linking with a different library. In addition,
1267 when a function is recognized as a built-in function, GCC may use
1268 information about that function to warn about problems with calls to
1269 that function, or to generate more efficient code, even if the
1270 resulting code still contains calls to that function. For example,
1271 warnings are given with @option{-Wformat} for bad calls to
1272 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1273 known not to modify global memory.
1275 With the @option{-fno-builtin-@var{function}} option
1276 only the built-in function @var{function} is
1277 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1278 function is named this is not built-in in this version of GCC, this
1279 option is ignored. There is no corresponding
1280 @option{-fbuiltin-@var{function}} option; if you wish to enable
1281 built-in functions selectively when using @option{-fno-builtin} or
1282 @option{-ffreestanding}, you may define macros such as:
1285 #define abs(n) __builtin_abs ((n))
1286 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1291 @cindex hosted environment
1293 Assert that compilation takes place in a hosted environment. This implies
1294 @option{-fbuiltin}. A hosted environment is one in which the
1295 entire standard library is available, and in which @code{main} has a return
1296 type of @code{int}. Examples are nearly everything except a kernel.
1297 This is equivalent to @option{-fno-freestanding}.
1299 @item -ffreestanding
1300 @opindex ffreestanding
1301 @cindex hosted environment
1303 Assert that compilation takes place in a freestanding environment. This
1304 implies @option{-fno-builtin}. A freestanding environment
1305 is one in which the standard library may not exist, and program startup may
1306 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1307 This is equivalent to @option{-fno-hosted}.
1309 @xref{Standards,,Language Standards Supported by GCC}, for details of
1310 freestanding and hosted environments.
1314 @cindex openmp parallel
1315 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1316 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1317 compiler generates parallel code according to the OpenMP Application
1318 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}.
1320 @item -fms-extensions
1321 @opindex fms-extensions
1322 Accept some non-standard constructs used in Microsoft header files.
1324 Some cases of unnamed fields in structures and unions are only
1325 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1326 fields within structs/unions}, for details.
1330 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1331 options for strict ISO C conformance) implies @option{-trigraphs}.
1333 @item -no-integrated-cpp
1334 @opindex no-integrated-cpp
1335 Performs a compilation in two passes: preprocessing and compiling. This
1336 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1337 @option{-B} option. The user supplied compilation step can then add in
1338 an additional preprocessing step after normal preprocessing but before
1339 compiling. The default is to use the integrated cpp (internal cpp)
1341 The semantics of this option will change if "cc1", "cc1plus", and
1342 "cc1obj" are merged.
1344 @cindex traditional C language
1345 @cindex C language, traditional
1347 @itemx -traditional-cpp
1348 @opindex traditional-cpp
1349 @opindex traditional
1350 Formerly, these options caused GCC to attempt to emulate a pre-standard
1351 C compiler. They are now only supported with the @option{-E} switch.
1352 The preprocessor continues to support a pre-standard mode. See the GNU
1353 CPP manual for details.
1355 @c APPLE LOCAL begin nested functions 4357979
1356 @item -fno-nested-functions
1357 @opindex fno-nested-functions
1358 Disable nested functions. This option is not supported for C++ or
1359 Objective-C++. On FreeBSD, nested functions are disabled by default.
1360 @c APPLE LOCAL end nested functions 4357979
1362 @item -fcond-mismatch
1363 @opindex fcond-mismatch
1364 Allow conditional expressions with mismatched types in the second and
1365 third arguments. The value of such an expression is void. This option
1366 is not supported for C++.
1368 @item -flax-vector-conversions
1369 @opindex flax-vector-conversions
1370 Allow implicit conversions between vectors with differing numbers of
1371 elements and/or incompatible element types. This option should not be
1374 @item -funsigned-char
1375 @opindex funsigned-char
1376 Let the type @code{char} be unsigned, like @code{unsigned char}.
1378 Each kind of machine has a default for what @code{char} should
1379 be. It is either like @code{unsigned char} by default or like
1380 @code{signed char} by default.
1382 Ideally, a portable program should always use @code{signed char} or
1383 @code{unsigned char} when it depends on the signedness of an object.
1384 But many programs have been written to use plain @code{char} and
1385 expect it to be signed, or expect it to be unsigned, depending on the
1386 machines they were written for. This option, and its inverse, let you
1387 make such a program work with the opposite default.
1389 The type @code{char} is always a distinct type from each of
1390 @code{signed char} or @code{unsigned char}, even though its behavior
1391 is always just like one of those two.
1394 @opindex fsigned-char
1395 Let the type @code{char} be signed, like @code{signed char}.
1397 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1398 the negative form of @option{-funsigned-char}. Likewise, the option
1399 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1401 @item -fsigned-bitfields
1402 @itemx -funsigned-bitfields
1403 @itemx -fno-signed-bitfields
1404 @itemx -fno-unsigned-bitfields
1405 @opindex fsigned-bitfields
1406 @opindex funsigned-bitfields
1407 @opindex fno-signed-bitfields
1408 @opindex fno-unsigned-bitfields
1409 These options control whether a bit-field is signed or unsigned, when the
1410 declaration does not use either @code{signed} or @code{unsigned}. By
1411 default, such a bit-field is signed, because this is consistent: the
1412 basic integer types such as @code{int} are signed types.
1415 @node C++ Dialect Options
1416 @section Options Controlling C++ Dialect
1418 @cindex compiler options, C++
1419 @cindex C++ options, command line
1420 @cindex options, C++
1421 This section describes the command-line options that are only meaningful
1422 for C++ programs; but you can also use most of the GNU compiler options
1423 regardless of what language your program is in. For example, you
1424 might compile a file @code{firstClass.C} like this:
1427 g++ -g -frepo -O -c firstClass.C
1431 In this example, only @option{-frepo} is an option meant
1432 only for C++ programs; you can use the other options with any
1433 language supported by GCC@.
1435 Here is a list of options that are @emph{only} for compiling C++ programs:
1439 @item -fabi-version=@var{n}
1440 @opindex fabi-version
1441 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1442 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1443 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1444 the version that conforms most closely to the C++ ABI specification.
1445 Therefore, the ABI obtained using version 0 will change as ABI bugs
1448 The default is version 2.
1450 @item -fno-access-control
1451 @opindex fno-access-control
1452 Turn off all access checking. This switch is mainly useful for working
1453 around bugs in the access control code.
1457 Check that the pointer returned by @code{operator new} is non-null
1458 before attempting to modify the storage allocated. This check is
1459 normally unnecessary because the C++ standard specifies that
1460 @code{operator new} will only return @code{0} if it is declared
1461 @samp{throw()}, in which case the compiler will always check the
1462 return value even without this option. In all other cases, when
1463 @code{operator new} has a non-empty exception specification, memory
1464 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1465 @samp{new (nothrow)}.
1467 @item -fconserve-space
1468 @opindex fconserve-space
1469 Put uninitialized or runtime-initialized global variables into the
1470 common segment, as C does. This saves space in the executable at the
1471 cost of not diagnosing duplicate definitions. If you compile with this
1472 flag and your program mysteriously crashes after @code{main()} has
1473 completed, you may have an object that is being destroyed twice because
1474 two definitions were merged.
1476 This option is no longer useful on most targets, now that support has
1477 been added for putting variables into BSS without making them common.
1479 @item -ffriend-injection
1480 @opindex ffriend-injection
1481 Inject friend functions into the enclosing namespace, so that they are
1482 visible outside the scope of the class in which they are declared.
1483 Friend functions were documented to work this way in the old Annotated
1484 C++ Reference Manual, and versions of G++ before 4.1 always worked
1485 that way. However, in ISO C++ a friend function which is not declared
1486 in an enclosing scope can only be found using argument dependent
1487 lookup. This option causes friends to be injected as they were in
1490 This option is for compatibility, and may be removed in a future
1493 @item -fno-elide-constructors
1494 @opindex fno-elide-constructors
1495 The C++ standard allows an implementation to omit creating a temporary
1496 which is only used to initialize another object of the same type.
1497 Specifying this option disables that optimization, and forces G++ to
1498 call the copy constructor in all cases.
1500 @item -fno-enforce-eh-specs
1501 @opindex fno-enforce-eh-specs
1502 Don't generate code to check for violation of exception specifications
1503 at runtime. This option violates the C++ standard, but may be useful
1504 for reducing code size in production builds, much like defining
1505 @samp{NDEBUG}. This does not give user code permission to throw
1506 exceptions in violation of the exception specifications; the compiler
1507 will still optimize based on the specifications, so throwing an
1508 unexpected exception will result in undefined behavior.
1511 @itemx -fno-for-scope
1513 @opindex fno-for-scope
1514 If @option{-ffor-scope} is specified, the scope of variables declared in
1515 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1516 as specified by the C++ standard.
1517 If @option{-fno-for-scope} is specified, the scope of variables declared in
1518 a @i{for-init-statement} extends to the end of the enclosing scope,
1519 as was the case in old versions of G++, and other (traditional)
1520 implementations of C++.
1522 The default if neither flag is given to follow the standard,
1523 but to allow and give a warning for old-style code that would
1524 otherwise be invalid, or have different behavior.
1526 @item -fno-gnu-keywords
1527 @opindex fno-gnu-keywords
1528 Do not recognize @code{typeof} as a keyword, so that code can use this
1529 word as an identifier. You can use the keyword @code{__typeof__} instead.
1530 @option{-ansi} implies @option{-fno-gnu-keywords}.
1532 @item -fno-implicit-templates
1533 @opindex fno-implicit-templates
1534 Never emit code for non-inline templates which are instantiated
1535 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1536 @xref{Template Instantiation}, for more information.
1538 @item -fno-implicit-inline-templates
1539 @opindex fno-implicit-inline-templates
1540 Don't emit code for implicit instantiations of inline templates, either.
1541 The default is to handle inlines differently so that compiles with and
1542 without optimization will need the same set of explicit instantiations.
1544 @item -fno-implement-inlines
1545 @opindex fno-implement-inlines
1546 To save space, do not emit out-of-line copies of inline functions
1547 controlled by @samp{#pragma implementation}. This will cause linker
1548 errors if these functions are not inlined everywhere they are called.
1550 @item -fms-extensions
1551 @opindex fms-extensions
1552 Disable pedantic warnings about constructs used in MFC, such as implicit
1553 int and getting a pointer to member function via non-standard syntax.
1555 @item -fno-nonansi-builtins
1556 @opindex fno-nonansi-builtins
1557 Disable built-in declarations of functions that are not mandated by
1558 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1559 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1561 @item -fno-operator-names
1562 @opindex fno-operator-names
1563 Do not treat the operator name keywords @code{and}, @code{bitand},
1564 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1565 synonyms as keywords.
1567 @item -fno-optional-diags
1568 @opindex fno-optional-diags
1569 Disable diagnostics that the standard says a compiler does not need to
1570 issue. Currently, the only such diagnostic issued by G++ is the one for
1571 a name having multiple meanings within a class.
1574 @opindex fpermissive
1575 Downgrade some diagnostics about nonconformant code from errors to
1576 warnings. Thus, using @option{-fpermissive} will allow some
1577 nonconforming code to compile.
1581 Enable automatic template instantiation at link time. This option also
1582 implies @option{-fno-implicit-templates}. @xref{Template
1583 Instantiation}, for more information.
1587 Disable generation of information about every class with virtual
1588 functions for use by the C++ runtime type identification features
1589 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1590 of the language, you can save some space by using this flag. Note that
1591 exception handling uses the same information, but it will generate it as
1592 needed. The @samp{dynamic_cast} operator can still be used for casts that
1593 do not require runtime type information, i.e. casts to @code{void *} or to
1594 unambiguous base classes.
1598 Emit statistics about front-end processing at the end of the compilation.
1599 This information is generally only useful to the G++ development team.
1601 @item -ftemplate-depth-@var{n}
1602 @opindex ftemplate-depth
1603 Set the maximum instantiation depth for template classes to @var{n}.
1604 A limit on the template instantiation depth is needed to detect
1605 endless recursions during template class instantiation. ANSI/ISO C++
1606 conforming programs must not rely on a maximum depth greater than 17.
1608 @item -fno-threadsafe-statics
1609 @opindex fno-threadsafe-statics
1610 Do not emit the extra code to use the routines specified in the C++
1611 ABI for thread-safe initialization of local statics. You can use this
1612 option to reduce code size slightly in code that doesn't need to be
1615 @item -fuse-cxa-atexit
1616 @opindex fuse-cxa-atexit
1617 Register destructors for objects with static storage duration with the
1618 @code{__cxa_atexit} function rather than the @code{atexit} function.
1619 This option is required for fully standards-compliant handling of static
1620 destructors, but will only work if your C library supports
1621 @code{__cxa_atexit}.
1623 @item -fno-use-cxa-get-exception-ptr
1624 @opindex fno-use-cxa-get-exception-ptr
1625 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1626 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1627 if the runtime routine is not available.
1629 @item -fvisibility-inlines-hidden
1630 @opindex fvisibility-inlines-hidden
1631 This switch declares that the user does not attempt to compare
1632 pointers to inline methods where the addresses of the two functions
1633 were taken in different shared objects.
1635 The effect of this is that GCC may, effectively, mark inline methods with
1636 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1637 appear in the export table of a DSO and do not require a PLT indirection
1638 when used within the DSO@. Enabling this option can have a dramatic effect
1639 on load and link times of a DSO as it massively reduces the size of the
1640 dynamic export table when the library makes heavy use of templates.
1642 The behaviour of this switch is not quite the same as marking the
1643 methods as hidden directly, because it does not affect static variables
1644 local to the function or cause the compiler to deduce that
1645 the function is defined in only one shared object.
1647 You may mark a method as having a visibility explicitly to negate the
1648 effect of the switch for that method. For example, if you do want to
1649 compare pointers to a particular inline method, you might mark it as
1650 having default visibility. Marking the enclosing class with explicit
1651 visibility will have no effect.
1653 Explicitly instantiated inline methods are unaffected by this option
1654 as their linkage might otherwise cross a shared library boundary.
1655 @xref{Template Instantiation}.
1657 @item -fvisibility-ms-compat
1658 @opindex fvisibility-ms-compat
1659 This flag attempts to use visibility settings to make GCC's C++
1660 linkage model compatible with that of Microsoft Visual Studio.
1662 The flag makes these changes to GCC's linkage model:
1666 It sets the default visibility to @code{hidden}, like
1667 @option{-fvisibility=hidden}.
1670 Types, but not their members, are not hidden by default.
1673 The One Definition Rule is relaxed for types without explicit
1674 visibility specifications which are defined in more than one different
1675 shared object: those declarations are permitted if they would have
1676 been permitted when this option was not used.
1679 In new code it is better to use @option{-fvisibility=hidden} and
1680 export those classes which are intended to be externally visible.
1681 Unfortunately it is possible for code to rely, perhaps accidentally,
1682 on the Visual Studio behaviour.
1684 Among the consequences of these changes are that static data members
1685 of the same type with the same name but defined in different shared
1686 objects will be different, so changing one will not change the other;
1687 and that pointers to function members defined in different shared
1688 objects may not compare equal. When this flag is given, it is a
1689 violation of the ODR to define types with the same name differently.
1693 Do not use weak symbol support, even if it is provided by the linker.
1694 By default, G++ will use weak symbols if they are available. This
1695 option exists only for testing, and should not be used by end-users;
1696 it will result in inferior code and has no benefits. This option may
1697 be removed in a future release of G++.
1701 Do not search for header files in the standard directories specific to
1702 C++, but do still search the other standard directories. (This option
1703 is used when building the C++ library.)
1706 In addition, these optimization, warning, and code generation options
1707 have meanings only for C++ programs:
1710 @item -fno-default-inline
1711 @opindex fno-default-inline
1712 Do not assume @samp{inline} for functions defined inside a class scope.
1713 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1714 functions will have linkage like inline functions; they just won't be
1717 @item -Wabi @r{(C++ only)}
1719 Warn when G++ generates code that is probably not compatible with the
1720 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1721 all such cases, there are probably some cases that are not warned about,
1722 even though G++ is generating incompatible code. There may also be
1723 cases where warnings are emitted even though the code that is generated
1726 You should rewrite your code to avoid these warnings if you are
1727 concerned about the fact that code generated by G++ may not be binary
1728 compatible with code generated by other compilers.
1730 The known incompatibilities at this point include:
1735 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1736 pack data into the same byte as a base class. For example:
1739 struct A @{ virtual void f(); int f1 : 1; @};
1740 struct B : public A @{ int f2 : 1; @};
1744 In this case, G++ will place @code{B::f2} into the same byte
1745 as@code{A::f1}; other compilers will not. You can avoid this problem
1746 by explicitly padding @code{A} so that its size is a multiple of the
1747 byte size on your platform; that will cause G++ and other compilers to
1748 layout @code{B} identically.
1751 Incorrect handling of tail-padding for virtual bases. G++ does not use
1752 tail padding when laying out virtual bases. For example:
1755 struct A @{ virtual void f(); char c1; @};
1756 struct B @{ B(); char c2; @};
1757 struct C : public A, public virtual B @{@};
1761 In this case, G++ will not place @code{B} into the tail-padding for
1762 @code{A}; other compilers will. You can avoid this problem by
1763 explicitly padding @code{A} so that its size is a multiple of its
1764 alignment (ignoring virtual base classes); that will cause G++ and other
1765 compilers to layout @code{C} identically.
1768 Incorrect handling of bit-fields with declared widths greater than that
1769 of their underlying types, when the bit-fields appear in a union. For
1773 union U @{ int i : 4096; @};
1777 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1778 union too small by the number of bits in an @code{int}.
1781 Empty classes can be placed at incorrect offsets. For example:
1791 struct C : public B, public A @{@};
1795 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1796 it should be placed at offset zero. G++ mistakenly believes that the
1797 @code{A} data member of @code{B} is already at offset zero.
1800 Names of template functions whose types involve @code{typename} or
1801 template template parameters can be mangled incorrectly.
1804 template <typename Q>
1805 void f(typename Q::X) @{@}
1807 template <template <typename> class Q>
1808 void f(typename Q<int>::X) @{@}
1812 Instantiations of these templates may be mangled incorrectly.
1816 @item -Wctor-dtor-privacy @r{(C++ only)}
1817 @opindex Wctor-dtor-privacy
1818 Warn when a class seems unusable because all the constructors or
1819 destructors in that class are private, and it has neither friends nor
1820 public static member functions.
1822 @item -Wnon-virtual-dtor @r{(C++ only)}
1823 @opindex Wnon-virtual-dtor
1824 Warn when a class appears to be polymorphic, thereby requiring a virtual
1825 destructor, yet it declares a non-virtual one. This warning is also
1826 enabled if -Weffc++ is specified.
1828 @item -Wreorder @r{(C++ only)}
1830 @cindex reordering, warning
1831 @cindex warning for reordering of member initializers
1832 Warn when the order of member initializers given in the code does not
1833 match the order in which they must be executed. For instance:
1839 A(): j (0), i (1) @{ @}
1843 The compiler will rearrange the member initializers for @samp{i}
1844 and @samp{j} to match the declaration order of the members, emitting
1845 a warning to that effect. This warning is enabled by @option{-Wall}.
1848 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1851 @item -Weffc++ @r{(C++ only)}
1853 Warn about violations of the following style guidelines from Scott Meyers'
1854 @cite{Effective C++} book:
1858 Item 11: Define a copy constructor and an assignment operator for classes
1859 with dynamically allocated memory.
1862 Item 12: Prefer initialization to assignment in constructors.
1865 Item 14: Make destructors virtual in base classes.
1868 Item 15: Have @code{operator=} return a reference to @code{*this}.
1871 Item 23: Don't try to return a reference when you must return an object.
1875 Also warn about violations of the following style guidelines from
1876 Scott Meyers' @cite{More Effective C++} book:
1880 Item 6: Distinguish between prefix and postfix forms of increment and
1881 decrement operators.
1884 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1888 When selecting this option, be aware that the standard library
1889 headers do not obey all of these guidelines; use @samp{grep -v}
1890 to filter out those warnings.
1892 @item -Wno-deprecated @r{(C++ only)}
1893 @opindex Wno-deprecated
1894 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1896 @item -Wstrict-null-sentinel @r{(C++ only)}
1897 @opindex Wstrict-null-sentinel
1898 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1899 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1900 to @code{__null}. Although it is a null pointer constant not a null pointer,
1901 it is guaranteed to of the same size as a pointer. But this use is
1902 not portable across different compilers.
1904 @item -Wno-non-template-friend @r{(C++ only)}
1905 @opindex Wno-non-template-friend
1906 Disable warnings when non-templatized friend functions are declared
1907 within a template. Since the advent of explicit template specification
1908 support in G++, if the name of the friend is an unqualified-id (i.e.,
1909 @samp{friend foo(int)}), the C++ language specification demands that the
1910 friend declare or define an ordinary, nontemplate function. (Section
1911 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1912 could be interpreted as a particular specialization of a templatized
1913 function. Because this non-conforming behavior is no longer the default
1914 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1915 check existing code for potential trouble spots and is on by default.
1916 This new compiler behavior can be turned off with
1917 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1918 but disables the helpful warning.
1920 @item -Wold-style-cast @r{(C++ only)}
1921 @opindex Wold-style-cast
1922 Warn if an old-style (C-style) cast to a non-void type is used within
1923 a C++ program. The new-style casts (@samp{dynamic_cast},
1924 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
1925 less vulnerable to unintended effects and much easier to search for.
1927 @item -Woverloaded-virtual @r{(C++ only)}
1928 @opindex Woverloaded-virtual
1929 @cindex overloaded virtual fn, warning
1930 @cindex warning for overloaded virtual fn
1931 Warn when a function declaration hides virtual functions from a
1932 base class. For example, in:
1939 struct B: public A @{
1944 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1952 will fail to compile.
1954 @item -Wno-pmf-conversions @r{(C++ only)}
1955 @opindex Wno-pmf-conversions
1956 Disable the diagnostic for converting a bound pointer to member function
1959 @item -Wsign-promo @r{(C++ only)}
1960 @opindex Wsign-promo
1961 Warn when overload resolution chooses a promotion from unsigned or
1962 enumerated type to a signed type, over a conversion to an unsigned type of
1963 the same size. Previous versions of G++ would try to preserve
1964 unsignedness, but the standard mandates the current behavior.
1969 A& operator = (int);
1979 In this example, G++ will synthesize a default @samp{A& operator =
1980 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1983 @node Language Independent Options
1984 @section Options to Control Diagnostic Messages Formatting
1985 @cindex options to control diagnostics formatting
1986 @cindex diagnostic messages
1987 @cindex message formatting
1989 Traditionally, diagnostic messages have been formatted irrespective of
1990 the output device's aspect (e.g.@: its width, @dots{}). The options described
1991 below can be used to control the diagnostic messages formatting
1992 algorithm, e.g.@: how many characters per line, how often source location
1993 information should be reported. Right now, only the C++ front end can
1994 honor these options. However it is expected, in the near future, that
1995 the remaining front ends would be able to digest them correctly.
1998 @item -fmessage-length=@var{n}
1999 @opindex fmessage-length
2000 Try to format error messages so that they fit on lines of about @var{n}
2001 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2002 the front ends supported by GCC@. If @var{n} is zero, then no
2003 line-wrapping will be done; each error message will appear on a single
2006 @opindex fdiagnostics-show-location
2007 @item -fdiagnostics-show-location=once
2008 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2009 reporter to emit @emph{once} source location information; that is, in
2010 case the message is too long to fit on a single physical line and has to
2011 be wrapped, the source location won't be emitted (as prefix) again,
2012 over and over, in subsequent continuation lines. This is the default
2015 @item -fdiagnostics-show-location=every-line
2016 Only meaningful in line-wrapping mode. Instructs the diagnostic
2017 messages reporter to emit the same source location information (as
2018 prefix) for physical lines that result from the process of breaking
2019 a message which is too long to fit on a single line.
2021 @item -fdiagnostics-show-option
2022 @opindex fdiagnostics-show-option
2023 This option instructs the diagnostic machinery to add text to each
2024 diagnostic emitted, which indicates which command line option directly
2025 controls that diagnostic, when such an option is known to the
2026 diagnostic machinery.
2030 @node Warning Options
2031 @section Options to Request or Suppress Warnings
2032 @cindex options to control warnings
2033 @cindex warning messages
2034 @cindex messages, warning
2035 @cindex suppressing warnings
2037 Warnings are diagnostic messages that report constructions which
2038 are not inherently erroneous but which are risky or suggest there
2039 may have been an error.
2041 You can request many specific warnings with options beginning @samp{-W},
2042 for example @option{-Wimplicit} to request warnings on implicit
2043 declarations. Each of these specific warning options also has a
2044 negative form beginning @samp{-Wno-} to turn off warnings;
2045 for example, @option{-Wno-implicit}. This manual lists only one of the
2046 two forms, whichever is not the default.
2048 The following options control the amount and kinds of warnings produced
2049 by GCC; for further, language-specific options also refer to
2050 @ref{C++ Dialect Options}.
2053 @cindex syntax checking
2055 @opindex fsyntax-only
2056 Check the code for syntax errors, but don't do anything beyond that.
2060 Issue all the warnings demanded by strict ISO C and ISO C++;
2061 reject all programs that use forbidden extensions, and some other
2062 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2063 version of the ISO C standard specified by any @option{-std} option used.
2065 Valid ISO C and ISO C++ programs should compile properly with or without
2066 this option (though a rare few will require @option{-ansi} or a
2067 @option{-std} option specifying the required version of ISO C)@. However,
2068 without this option, certain GNU extensions and traditional C and C++
2069 features are supported as well. With this option, they are rejected.
2071 @option{-pedantic} does not cause warning messages for use of the
2072 alternate keywords whose names begin and end with @samp{__}. Pedantic
2073 warnings are also disabled in the expression that follows
2074 @code{__extension__}. However, only system header files should use
2075 these escape routes; application programs should avoid them.
2076 @xref{Alternate Keywords}.
2078 Some users try to use @option{-pedantic} to check programs for strict ISO
2079 C conformance. They soon find that it does not do quite what they want:
2080 it finds some non-ISO practices, but not all---only those for which
2081 ISO C @emph{requires} a diagnostic, and some others for which
2082 diagnostics have been added.
2084 A feature to report any failure to conform to ISO C might be useful in
2085 some instances, but would require considerable additional work and would
2086 be quite different from @option{-pedantic}. We don't have plans to
2087 support such a feature in the near future.
2089 Where the standard specified with @option{-std} represents a GNU
2090 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2091 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2092 extended dialect is based. Warnings from @option{-pedantic} are given
2093 where they are required by the base standard. (It would not make sense
2094 for such warnings to be given only for features not in the specified GNU
2095 C dialect, since by definition the GNU dialects of C include all
2096 features the compiler supports with the given option, and there would be
2097 nothing to warn about.)
2099 @item -pedantic-errors
2100 @opindex pedantic-errors
2101 Like @option{-pedantic}, except that errors are produced rather than
2106 Inhibit all warning messages.
2110 Inhibit warning messages about the use of @samp{#import}.
2112 @c APPLE LOCAL begin -Wnewline-eof 2001-08-23 --sts **
2114 @opindex Wnewline-eof
2115 Warn about files missing a newline at the end of the file. (Apple compatible)
2116 @c APPLE LOCAL end -Wnewline-eof 2001-08-23 --sts **
2118 @item -Wchar-subscripts
2119 @opindex Wchar-subscripts
2120 Warn if an array subscript has type @code{char}. This is a common cause
2121 of error, as programmers often forget that this type is signed on some
2123 This warning is enabled by @option{-Wall}.
2127 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2128 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2129 This warning is enabled by @option{-Wall}.
2131 @item -Wfatal-errors
2132 @opindex Wfatal-errors
2133 This option causes the compiler to abort compilation on the first error
2134 occurred rather than trying to keep going and printing further error
2139 @opindex ffreestanding
2140 @opindex fno-builtin
2141 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2142 the arguments supplied have types appropriate to the format string
2143 specified, and that the conversions specified in the format string make
2144 sense. This includes standard functions, and others specified by format
2145 attributes (@pxref{Function Attributes}), in the @code{printf},
2146 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2147 not in the C standard) families (or other target-specific families).
2148 Which functions are checked without format attributes having been
2149 specified depends on the standard version selected, and such checks of
2150 functions without the attribute specified are disabled by
2151 @option{-ffreestanding} or @option{-fno-builtin}.
2153 The formats are checked against the format features supported by GNU
2154 libc version 2.2. These include all ISO C90 and C99 features, as well
2155 as features from the Single Unix Specification and some BSD and GNU
2156 extensions. Other library implementations may not support all these
2157 features; GCC does not support warning about features that go beyond a
2158 particular library's limitations. However, if @option{-pedantic} is used
2159 with @option{-Wformat}, warnings will be given about format features not
2160 in the selected standard version (but not for @code{strfmon} formats,
2161 since those are not in any version of the C standard). @xref{C Dialect
2162 Options,,Options Controlling C Dialect}.
2164 Since @option{-Wformat} also checks for null format arguments for
2165 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2167 @option{-Wformat} is included in @option{-Wall}. For more control over some
2168 aspects of format checking, the options @option{-Wformat-y2k},
2169 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2170 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2171 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2174 @opindex Wformat-y2k
2175 If @option{-Wformat} is specified, also warn about @code{strftime}
2176 formats which may yield only a two-digit year.
2178 @item -Wno-format-extra-args
2179 @opindex Wno-format-extra-args
2180 If @option{-Wformat} is specified, do not warn about excess arguments to a
2181 @code{printf} or @code{scanf} format function. The C standard specifies
2182 that such arguments are ignored.
2184 Where the unused arguments lie between used arguments that are
2185 specified with @samp{$} operand number specifications, normally
2186 warnings are still given, since the implementation could not know what
2187 type to pass to @code{va_arg} to skip the unused arguments. However,
2188 in the case of @code{scanf} formats, this option will suppress the
2189 warning if the unused arguments are all pointers, since the Single
2190 Unix Specification says that such unused arguments are allowed.
2192 @item -Wno-format-zero-length
2193 @opindex Wno-format-zero-length
2194 If @option{-Wformat} is specified, do not warn about zero-length formats.
2195 The C standard specifies that zero-length formats are allowed.
2197 @item -Wformat-nonliteral
2198 @opindex Wformat-nonliteral
2199 If @option{-Wformat} is specified, also warn if the format string is not a
2200 string literal and so cannot be checked, unless the format function
2201 takes its format arguments as a @code{va_list}.
2203 @item -Wformat-security
2204 @opindex Wformat-security
2205 If @option{-Wformat} is specified, also warn about uses of format
2206 functions that represent possible security problems. At present, this
2207 warns about calls to @code{printf} and @code{scanf} functions where the
2208 format string is not a string literal and there are no format arguments,
2209 as in @code{printf (foo);}. This may be a security hole if the format
2210 string came from untrusted input and contains @samp{%n}. (This is
2211 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2212 in future warnings may be added to @option{-Wformat-security} that are not
2213 included in @option{-Wformat-nonliteral}.)
2217 Enable @option{-Wformat} plus format checks not included in
2218 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2219 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2223 Warn about passing a null pointer for arguments marked as
2224 requiring a non-null value by the @code{nonnull} function attribute.
2226 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2227 can be disabled with the @option{-Wno-nonnull} option.
2229 @item -Winit-self @r{(C and C++ only)}
2231 Warn about uninitialized variables which are initialized with themselves.
2232 Note this option can only be used with the @option{-Wuninitialized} option,
2233 which in turn only works with @option{-O1} and above.
2235 For example, GCC will warn about @code{i} being uninitialized in the
2236 following snippet only when @option{-Winit-self} has been specified:
2247 @item -Wimplicit-int
2248 @opindex Wimplicit-int
2249 Warn when a declaration does not specify a type.
2250 This warning is enabled by @option{-Wall}.
2252 @item -Wimplicit-function-declaration
2253 @itemx -Werror-implicit-function-declaration
2254 @opindex Wimplicit-function-declaration
2255 @opindex Werror-implicit-function-declaration
2256 Give a warning (or error) whenever a function is used before being
2257 declared. The form @option{-Wno-error-implicit-function-declaration}
2259 This warning is enabled by @option{-Wall} (as a warning, not an error).
2263 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2264 This warning is enabled by @option{-Wall}.
2268 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2269 function with external linkage, returning int, taking either zero
2270 arguments, two, or three arguments of appropriate types.
2271 This warning is enabled by @option{-Wall}.
2273 @item -Wmissing-braces
2274 @opindex Wmissing-braces
2275 Warn if an aggregate or union initializer is not fully bracketed. In
2276 the following example, the initializer for @samp{a} is not fully
2277 bracketed, but that for @samp{b} is fully bracketed.
2280 int a[2][2] = @{ 0, 1, 2, 3 @};
2281 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2284 This warning is enabled by @option{-Wall}.
2286 @item -Wmissing-include-dirs @r{(C and C++ only)}
2287 @opindex Wmissing-include-dirs
2288 Warn if a user-supplied include directory does not exist.
2291 @opindex Wparentheses
2292 Warn if parentheses are omitted in certain contexts, such
2293 as when there is an assignment in a context where a truth value
2294 is expected, or when operators are nested whose precedence people
2295 often get confused about.
2297 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2298 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2299 interpretation from that of ordinary mathematical notation.
2301 Also warn about constructions where there may be confusion to which
2302 @code{if} statement an @code{else} branch belongs. Here is an example of
2317 In C/C++, every @code{else} branch belongs to the innermost possible
2318 @code{if} statement, which in this example is @code{if (b)}. This is
2319 often not what the programmer expected, as illustrated in the above
2320 example by indentation the programmer chose. When there is the
2321 potential for this confusion, GCC will issue a warning when this flag
2322 is specified. To eliminate the warning, add explicit braces around
2323 the innermost @code{if} statement so there is no way the @code{else}
2324 could belong to the enclosing @code{if}. The resulting code would
2341 This warning is enabled by @option{-Wall}.
2343 @item -Wsequence-point
2344 @opindex Wsequence-point
2345 Warn about code that may have undefined semantics because of violations
2346 of sequence point rules in the C and C++ standards.
2348 The C and C++ standards defines the order in which expressions in a C/C++
2349 program are evaluated in terms of @dfn{sequence points}, which represent
2350 a partial ordering between the execution of parts of the program: those
2351 executed before the sequence point, and those executed after it. These
2352 occur after the evaluation of a full expression (one which is not part
2353 of a larger expression), after the evaluation of the first operand of a
2354 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2355 function is called (but after the evaluation of its arguments and the
2356 expression denoting the called function), and in certain other places.
2357 Other than as expressed by the sequence point rules, the order of
2358 evaluation of subexpressions of an expression is not specified. All
2359 these rules describe only a partial order rather than a total order,
2360 since, for example, if two functions are called within one expression
2361 with no sequence point between them, the order in which the functions
2362 are called is not specified. However, the standards committee have
2363 ruled that function calls do not overlap.
2365 It is not specified when between sequence points modifications to the
2366 values of objects take effect. Programs whose behavior depends on this
2367 have undefined behavior; the C and C++ standards specify that ``Between
2368 the previous and next sequence point an object shall have its stored
2369 value modified at most once by the evaluation of an expression.
2370 Furthermore, the prior value shall be read only to determine the value
2371 to be stored.''. If a program breaks these rules, the results on any
2372 particular implementation are entirely unpredictable.
2374 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2375 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2376 diagnosed by this option, and it may give an occasional false positive
2377 result, but in general it has been found fairly effective at detecting
2378 this sort of problem in programs.
2380 The standard is worded confusingly, therefore there is some debate
2381 over the precise meaning of the sequence point rules in subtle cases.
2382 Links to discussions of the problem, including proposed formal
2383 definitions, may be found on the GCC readings page, at
2384 @w{@uref{http://gcc.gnu.org/readings.html}}.
2386 This warning is enabled by @option{-Wall} for C and C++.
2389 @opindex Wreturn-type
2390 Warn whenever a function is defined with a return-type that defaults to
2391 @code{int}. Also warn about any @code{return} statement with no
2392 return-value in a function whose return-type is not @code{void}.
2394 For C, also warn if the return type of a function has a type qualifier
2395 such as @code{const}. Such a type qualifier has no effect, since the
2396 value returned by a function is not an lvalue. ISO C prohibits
2397 qualified @code{void} return types on function definitions, so such
2398 return types always receive a warning even without this option.
2400 For C++, a function without return type always produces a diagnostic
2401 message, even when @option{-Wno-return-type} is specified. The only
2402 exceptions are @samp{main} and functions defined in system headers.
2404 This warning is enabled by @option{-Wall}.
2408 Warn whenever a @code{switch} statement has an index of enumerated type
2409 and lacks a @code{case} for one or more of the named codes of that
2410 enumeration. (The presence of a @code{default} label prevents this
2411 warning.) @code{case} labels outside the enumeration range also
2412 provoke warnings when this option is used.
2413 This warning is enabled by @option{-Wall}.
2415 @item -Wswitch-default
2416 @opindex Wswitch-switch
2417 Warn whenever a @code{switch} statement does not have a @code{default}
2421 @opindex Wswitch-enum
2422 Warn whenever a @code{switch} statement has an index of enumerated type
2423 and lacks a @code{case} for one or more of the named codes of that
2424 enumeration. @code{case} labels outside the enumeration range also
2425 provoke warnings when this option is used.
2429 Warn if any trigraphs are encountered that might change the meaning of
2430 the program (trigraphs within comments are not warned about).
2431 This warning is enabled by @option{-Wall}.
2433 @item -Wunused-function
2434 @opindex Wunused-function
2435 Warn whenever a static function is declared but not defined or a
2436 non-inline static function is unused.
2437 This warning is enabled by @option{-Wall}.
2439 @item -Wunused-label
2440 @opindex Wunused-label
2441 Warn whenever a label is declared but not used.
2442 This warning is enabled by @option{-Wall}.
2444 To suppress this warning use the @samp{unused} attribute
2445 (@pxref{Variable Attributes}).
2447 @item -Wunused-parameter
2448 @opindex Wunused-parameter
2449 Warn whenever a function parameter is unused aside from its declaration.
2451 To suppress this warning use the @samp{unused} attribute
2452 (@pxref{Variable Attributes}).
2454 @item -Wunused-variable
2455 @opindex Wunused-variable
2456 Warn whenever a local variable or non-constant static variable is unused
2457 aside from its declaration.
2458 This warning is enabled by @option{-Wall}.
2460 To suppress this warning use the @samp{unused} attribute
2461 (@pxref{Variable Attributes}).
2463 @item -Wunused-value
2464 @opindex Wunused-value
2465 Warn whenever a statement computes a result that is explicitly not used.
2466 This warning is enabled by @option{-Wall}.
2468 To suppress this warning cast the expression to @samp{void}.
2472 All the above @option{-Wunused} options combined.
2474 In order to get a warning about an unused function parameter, you must
2475 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2476 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2478 @item -Wuninitialized
2479 @opindex Wuninitialized
2480 Warn if an automatic variable is used without first being initialized or
2481 if a variable may be clobbered by a @code{setjmp} call.
2483 These warnings are possible only in optimizing compilation,
2484 because they require data flow information that is computed only
2485 when optimizing. If you do not specify @option{-O}, you will not get
2486 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
2487 requiring @option{-O}.
2489 If you want to warn about code which uses the uninitialized value of the
2490 variable in its own initializer, use the @option{-Winit-self} option.
2492 These warnings occur for individual uninitialized or clobbered
2493 elements of structure, union or array variables as well as for
2494 variables which are uninitialized or clobbered as a whole. They do
2495 not occur for variables or elements declared @code{volatile}. Because
2496 these warnings depend on optimization, the exact variables or elements
2497 for which there are warnings will depend on the precise optimization
2498 options and version of GCC used.
2500 Note that there may be no warning about a variable that is used only
2501 to compute a value that itself is never used, because such
2502 computations may be deleted by data flow analysis before the warnings
2505 These warnings are made optional because GCC is not smart
2506 enough to see all the reasons why the code might be correct
2507 despite appearing to have an error. Here is one example of how
2528 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2529 always initialized, but GCC doesn't know this. Here is
2530 another common case:
2535 if (change_y) save_y = y, y = new_y;
2537 if (change_y) y = save_y;
2542 This has no bug because @code{save_y} is used only if it is set.
2544 @cindex @code{longjmp} warnings
2545 This option also warns when a non-volatile automatic variable might be
2546 changed by a call to @code{longjmp}. These warnings as well are possible
2547 only in optimizing compilation.
2549 The compiler sees only the calls to @code{setjmp}. It cannot know
2550 where @code{longjmp} will be called; in fact, a signal handler could
2551 call it at any point in the code. As a result, you may get a warning
2552 even when there is in fact no problem because @code{longjmp} cannot
2553 in fact be called at the place which would cause a problem.
2555 Some spurious warnings can be avoided if you declare all the functions
2556 you use that never return as @code{noreturn}. @xref{Function
2559 This warning is enabled by @option{-Wall}.
2561 @item -Wunknown-pragmas
2562 @opindex Wunknown-pragmas
2563 @cindex warning for unknown pragmas
2564 @cindex unknown pragmas, warning
2565 @cindex pragmas, warning of unknown
2566 Warn when a #pragma directive is encountered which is not understood by
2567 GCC@. If this command line option is used, warnings will even be issued
2568 for unknown pragmas in system header files. This is not the case if
2569 the warnings were only enabled by the @option{-Wall} command line option.
2572 @opindex Wno-pragmas
2574 Do not warn about misuses of pragmas, such as incorrect parameters,
2575 invalid syntax, or conflicts between pragmas. See also
2576 @samp{-Wunknown-pragmas}.
2578 @item -Wstrict-aliasing
2579 @opindex Wstrict-aliasing
2580 This option is only active when @option{-fstrict-aliasing} is active.
2581 It warns about code which might break the strict aliasing rules that the
2582 compiler is using for optimization. The warning does not catch all
2583 cases, but does attempt to catch the more common pitfalls. It is
2584 included in @option{-Wall}.
2585 It is equivalent to -Wstrict-aliasing=3
2587 @item -Wstrict-aliasing=n
2588 @opindex Wstrict-aliasing=n
2589 This option is only active when @option{-fstrict-aliasing} is active.
2590 It warns about code which might break the strict aliasing rules that the
2591 compiler is using for optimization.
2592 Higher levels correspond to higher accuracy (fewer false positives).
2593 Higher levels also correspond to more effort, similar to the way -O works.
2594 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
2597 Level 1: Most aggressive, quick, least accurate.
2598 Possibly useful when higher levels
2599 do not warn but -fstrict-aliasing still breaks the code, as it has very few
2600 false negatives. However, it has many false positives.
2601 Warns for all pointer conversions between possibly incompatible types,
2602 even if never dereferenced. Runs in the frontend only.
2604 Level 2: Aggressive, quick, not too precise.
2605 May still have many false positives (not as many as level 1 though),
2606 and few false negatives (but possibly more than level 1).
2607 Unlike level 1, it only warns when an address is taken. Warns about
2608 incomplete types. Runs in the frontend only.
2610 Level 3 (default for @option{-Wstrict-aliasing}):
2611 Should have very few false positives and few false
2612 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
2613 Takes care of the common punn+dereference pattern in the frontend:
2614 @code{*(int*)&some_float}.
2615 If optimization is enabled, it also runs in the backend, where it deals
2616 with multiple statement cases using flow-sensitive points-to information.
2617 Only warns when the converted pointer is dereferenced.
2618 Does not warn about incomplete types.
2620 @item -Wstrict-overflow
2621 @item -Wstrict-overflow=@var{n}
2622 @opindex Wstrict-overflow
2623 This option is only active when @option{-fstrict-overflow} is active.
2624 It warns about cases where the compiler optimizes based on the
2625 assumption that signed overflow does not occur. Note that it does not
2626 warn about all cases where the code might overflow: it only warns
2627 about cases where the compiler implements some optimization. Thus
2628 this warning depends on the optimization level.
2630 An optimization which assumes that signed overflow does not occur is
2631 perfectly safe if the values of the variables involved are such that
2632 overflow never does, in fact, occur. Therefore this warning can
2633 easily give a false positive: a warning about code which is not
2634 actually a problem. To help focus on important issues, several
2635 warning levels are defined. No warnings are issued for the use of
2636 undefined signed overflow when estimating how many iterations a loop
2637 will require, in particular when determining whether a loop will be
2641 @item -Wstrict-overflow=1
2642 Warn about cases which are both questionable and easy to avoid. For
2643 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
2644 compiler will simplify this to @code{1}. This level of
2645 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
2646 are not, and must be explicitly requested.
2648 @item -Wstrict-overflow=2
2649 Also warn about other cases where a comparison is simplified to a
2650 constant. For example: @code{abs (x) >= 0}. This can only be
2651 simplified when @option{-fstrict-overflow} is in effect, because
2652 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
2653 zero. @option{-Wstrict-overflow} (with no level) is the same as
2654 @option{-Wstrict-overflow=2}.
2656 @item -Wstrict-overflow=3
2657 Also warn about other cases where a comparison is simplified. For
2658 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
2660 @item -Wstrict-overflow=4
2661 Also warn about other simplifications not covered by the above cases.
2662 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
2664 @item -Wstrict-overflow=5
2665 Also warn about cases where the compiler reduces the magnitude of a
2666 constant involved in a comparison. For example: @code{x + 2 > y} will
2667 be simplified to @code{x + 1 >= y}. This is reported only at the
2668 highest warning level because this simplification applies to many
2669 comparisons, so this warning level will give a very large number of
2675 All of the above @samp{-W} options combined. This enables all the
2676 warnings about constructions that some users consider questionable, and
2677 that are easy to avoid (or modify to prevent the warning), even in
2678 conjunction with macros. This also enables some language-specific
2679 warnings described in @ref{C++ Dialect Options}.
2680 @c APPLE LOCAL begin -Wmost
2683 This is equivalent to -Wall -Wno-parentheses. (Apple compatible)
2685 @c APPLE LOCAL end -Wmost
2687 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2688 Some of them warn about constructions that users generally do not
2689 consider questionable, but which occasionally you might wish to check
2690 for; others warn about constructions that are necessary or hard to avoid
2691 in some cases, and there is no simple way to modify the code to suppress
2698 (This option used to be called @option{-W}. The older name is still
2699 supported, but the newer name is more descriptive.) Print extra warning
2700 messages for these events:
2704 A function can return either with or without a value. (Falling
2705 off the end of the function body is considered returning without
2706 a value.) For example, this function would evoke such a
2720 An expression-statement or the left-hand side of a comma expression
2721 contains no side effects.
2722 To suppress the warning, cast the unused expression to void.
2723 For example, an expression such as @samp{x[i,j]} will cause a warning,
2724 but @samp{x[(void)i,j]} will not.
2727 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2730 Storage-class specifiers like @code{static} are not the first things in
2731 a declaration. According to the C Standard, this usage is obsolescent.
2734 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2738 A comparison between signed and unsigned values could produce an
2739 incorrect result when the signed value is converted to unsigned.
2740 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2743 An aggregate has an initializer which does not initialize all members.
2744 This warning can be independently controlled by
2745 @option{-Wmissing-field-initializers}.
2748 An initialized field without side effects is overridden when using
2749 designated initializers (@pxref{Designated Inits, , Designated
2750 Initializers}). This warning can be independently controlled by
2751 @option{-Woverride-init}.
2754 A function parameter is declared without a type specifier in K&R-style
2762 An empty body occurs in an @samp{if} or @samp{else} statement.
2765 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2766 @samp{>}, or @samp{>=}.
2769 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2771 @item @r{(C++ only)}
2772 An enumerator and a non-enumerator both appear in a conditional expression.
2774 @item @r{(C++ only)}
2775 A non-static reference or non-static @samp{const} member appears in a
2776 class without constructors.
2778 @item @r{(C++ only)}
2779 Ambiguous virtual bases.
2781 @item @r{(C++ only)}
2782 Subscripting an array which has been declared @samp{register}.
2784 @item @r{(C++ only)}
2785 Taking the address of a variable which has been declared @samp{register}.
2787 @item @r{(C++ only)}
2788 A base class is not initialized in a derived class' copy constructor.
2791 @item -Wno-div-by-zero
2792 @opindex Wno-div-by-zero
2793 @opindex Wdiv-by-zero
2794 Do not warn about compile-time integer division by zero. Floating point
2795 division by zero is not warned about, as it can be a legitimate way of
2796 obtaining infinities and NaNs.
2798 @item -Wsystem-headers
2799 @opindex Wsystem-headers
2800 @cindex warnings from system headers
2801 @cindex system headers, warnings from
2802 Print warning messages for constructs found in system header files.
2803 Warnings from system headers are normally suppressed, on the assumption
2804 that they usually do not indicate real problems and would only make the
2805 compiler output harder to read. Using this command line option tells
2806 GCC to emit warnings from system headers as if they occurred in user
2807 code. However, note that using @option{-Wall} in conjunction with this
2808 option will @emph{not} warn about unknown pragmas in system
2809 headers---for that, @option{-Wunknown-pragmas} must also be used.
2812 @opindex Wfloat-equal
2813 Warn if floating point values are used in equality comparisons.
2815 The idea behind this is that sometimes it is convenient (for the
2816 programmer) to consider floating-point values as approximations to
2817 infinitely precise real numbers. If you are doing this, then you need
2818 to compute (by analyzing the code, or in some other way) the maximum or
2819 likely maximum error that the computation introduces, and allow for it
2820 when performing comparisons (and when producing output, but that's a
2821 different problem). In particular, instead of testing for equality, you
2822 would check to see whether the two values have ranges that overlap; and
2823 this is done with the relational operators, so equality comparisons are
2826 @item -Wtraditional @r{(C only)}
2827 @opindex Wtraditional
2828 Warn about certain constructs that behave differently in traditional and
2829 ISO C@. Also warn about ISO C constructs that have no traditional C
2830 equivalent, and/or problematic constructs which should be avoided.
2834 Macro parameters that appear within string literals in the macro body.
2835 In traditional C macro replacement takes place within string literals,
2836 but does not in ISO C@.
2839 In traditional C, some preprocessor directives did not exist.
2840 Traditional preprocessors would only consider a line to be a directive
2841 if the @samp{#} appeared in column 1 on the line. Therefore
2842 @option{-Wtraditional} warns about directives that traditional C
2843 understands but would ignore because the @samp{#} does not appear as the
2844 first character on the line. It also suggests you hide directives like
2845 @samp{#pragma} not understood by traditional C by indenting them. Some
2846 traditional implementations would not recognize @samp{#elif}, so it
2847 suggests avoiding it altogether.
2850 A function-like macro that appears without arguments.
2853 The unary plus operator.
2856 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2857 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2858 constants.) Note, these suffixes appear in macros defined in the system
2859 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2860 Use of these macros in user code might normally lead to spurious
2861 warnings, however GCC's integrated preprocessor has enough context to
2862 avoid warning in these cases.
2865 A function declared external in one block and then used after the end of
2869 A @code{switch} statement has an operand of type @code{long}.
2872 A non-@code{static} function declaration follows a @code{static} one.
2873 This construct is not accepted by some traditional C compilers.
2876 The ISO type of an integer constant has a different width or
2877 signedness from its traditional type. This warning is only issued if
2878 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2879 typically represent bit patterns, are not warned about.
2882 Usage of ISO string concatenation is detected.
2885 Initialization of automatic aggregates.
2888 Identifier conflicts with labels. Traditional C lacks a separate
2889 namespace for labels.
2892 Initialization of unions. If the initializer is zero, the warning is
2893 omitted. This is done under the assumption that the zero initializer in
2894 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2895 initializer warnings and relies on default initialization to zero in the
2899 Conversions by prototypes between fixed/floating point values and vice
2900 versa. The absence of these prototypes when compiling with traditional
2901 C would cause serious problems. This is a subset of the possible
2902 conversion warnings, for the full set use @option{-Wconversion}.
2905 Use of ISO C style function definitions. This warning intentionally is
2906 @emph{not} issued for prototype declarations or variadic functions
2907 because these ISO C features will appear in your code when using
2908 libiberty's traditional C compatibility macros, @code{PARAMS} and
2909 @code{VPARAMS}. This warning is also bypassed for nested functions
2910 because that feature is already a GCC extension and thus not relevant to
2911 traditional C compatibility.
2914 @item -Wdeclaration-after-statement @r{(C only)}
2915 @opindex Wdeclaration-after-statement
2916 Warn when a declaration is found after a statement in a block. This
2917 construct, known from C++, was introduced with ISO C99 and is by default
2918 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2919 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2923 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2925 @item -Wno-endif-labels
2926 @opindex Wno-endif-labels
2927 @opindex Wendif-labels
2928 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2932 Warn whenever a local variable shadows another local variable, parameter or
2933 global variable or whenever a built-in function is shadowed.
2935 @item -Wlarger-than-@var{len}
2936 @opindex Wlarger-than
2937 Warn whenever an object of larger than @var{len} bytes is defined.
2939 @item -Wframe-larger-than-@var{len}
2940 @opindex Wframe-larger-than
2941 Warn whenever the frame size of a function is larger than @var{len} bytes.
2943 @item -Wunsafe-loop-optimizations
2944 @opindex Wunsafe-loop-optimizations
2945 Warn if the loop cannot be optimized because the compiler could not
2946 assume anything on the bounds of the loop indices. With
2947 @option{-funsafe-loop-optimizations} warn if the compiler made
2950 @item -Wpointer-arith
2951 @opindex Wpointer-arith
2952 Warn about anything that depends on the ``size of'' a function type or
2953 of @code{void}. GNU C assigns these types a size of 1, for
2954 convenience in calculations with @code{void *} pointers and pointers
2957 @item -Wbad-function-cast @r{(C only)}
2958 @opindex Wbad-function-cast
2959 Warn whenever a function call is cast to a non-matching type.
2960 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2963 Warn about ISO C constructs that are outside of the common subset of
2964 ISO C and ISO C++, e.g.@: request for implicit conversion from
2965 @code{void *} to a pointer to non-@code{void} type.
2969 Warn whenever a pointer is cast so as to remove a type qualifier from
2970 the target type. For example, warn if a @code{const char *} is cast
2971 to an ordinary @code{char *}.
2974 @opindex Wcast-align
2975 Warn whenever a pointer is cast such that the required alignment of the
2976 target is increased. For example, warn if a @code{char *} is cast to
2977 an @code{int *} on machines where integers can only be accessed at
2978 two- or four-byte boundaries.
2980 @item -Wwrite-strings
2981 @opindex Wwrite-strings
2982 When compiling C, give string constants the type @code{const
2983 char[@var{length}]} so that
2984 copying the address of one into a non-@code{const} @code{char *}
2985 pointer will get a warning; when compiling C++, warn about the
2986 deprecated conversion from string literals to @code{char *}. This
2987 warning, by default, is enabled for C++ programs.
2988 These warnings will help you find at
2989 compile time code that can try to write into a string constant, but
2990 only if you have been very careful about using @code{const} in
2991 declarations and prototypes. Otherwise, it will just be a nuisance;
2992 this is why we did not make @option{-Wall} request these warnings.
2995 @opindex Wconversion
2996 Warn if a prototype causes a type conversion that is different from what
2997 would happen to the same argument in the absence of a prototype. This
2998 includes conversions of fixed point to floating and vice versa, and
2999 conversions changing the width or signedness of a fixed point argument
3000 except when the same as the default promotion.
3002 Also, warn if a negative integer constant expression is implicitly
3003 converted to an unsigned type. For example, warn about the assignment
3004 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
3005 casts like @code{(unsigned) -1}.
3007 @item -Wsign-compare
3008 @opindex Wsign-compare
3009 @cindex warning for comparison of signed and unsigned values
3010 @cindex comparison of signed and unsigned values, warning
3011 @cindex signed and unsigned values, comparison warning
3012 Warn when a comparison between signed and unsigned values could produce
3013 an incorrect result when the signed value is converted to unsigned.
3014 This warning is also enabled by @option{-Wextra}; to get the other warnings
3015 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3019 @opindex Wno-address
3020 Warn about suspicious uses of memory addresses. These include using
3021 the address of a function in a conditional expression, such as
3022 @code{void func(void); if (func)}, and comparisons against the memory
3023 address of a string literal, such as @code{if (x == "abc")}. Such
3024 uses typically indicate a programmer error: the address of a function
3025 always evaluates to true, so their use in a conditional usually
3026 indicate that the programmer forgot the parentheses in a function
3027 call; and comparisons against string literals result in unspecified
3028 behavior and are not portable in C, so they usually indicate that the
3029 programmer intended to use @code{strcmp}. This warning is enabled by
3032 @item -Waggregate-return
3033 @opindex Waggregate-return
3034 Warn if any functions that return structures or unions are defined or
3035 called. (In languages where you can return an array, this also elicits
3038 @item -Wno-attributes
3039 @opindex Wno-attributes
3040 @opindex Wattributes
3041 Do not warn if an unexpected @code{__attribute__} is used, such as
3042 unrecognized attributes, function attributes applied to variables,
3043 etc. This will not stop errors for incorrect use of supported
3046 @item -Wstrict-prototypes @r{(C only)}
3047 @opindex Wstrict-prototypes
3048 Warn if a function is declared or defined without specifying the
3049 argument types. (An old-style function definition is permitted without
3050 a warning if preceded by a declaration which specifies the argument
3053 @item -Wold-style-definition @r{(C only)}
3054 @opindex Wold-style-definition
3055 Warn if an old-style function definition is used. A warning is given
3056 even if there is a previous prototype.
3058 @c APPLE LOCAL warn missing prototype 6261539
3059 @item -Wmissing-prototypes
3060 @opindex Wmissing-prototypes
3061 Warn if a global function is defined without a previous prototype
3062 declaration. This warning is issued even if the definition itself
3063 provides a prototype. The aim is to detect global functions that fail
3064 to be declared in header files.
3066 @item -Wmissing-declarations @r{(C only)}
3067 @opindex Wmissing-declarations
3068 Warn if a global function is defined without a previous declaration.
3069 Do so even if the definition itself provides a prototype.
3070 Use this option to detect global functions that are not declared in
3073 @item -Wmissing-field-initializers
3074 @opindex Wmissing-field-initializers
3077 Warn if a structure's initializer has some fields missing. For
3078 example, the following code would cause such a warning, because
3079 @code{x.h} is implicitly zero:
3082 struct s @{ int f, g, h; @};
3083 struct s x = @{ 3, 4 @};
3086 This option does not warn about designated initializers, so the following
3087 modification would not trigger a warning:
3090 struct s @{ int f, g, h; @};
3091 struct s x = @{ .f = 3, .g = 4 @};
3094 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3095 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3097 @item -Wmissing-noreturn
3098 @opindex Wmissing-noreturn
3099 Warn about functions which might be candidates for attribute @code{noreturn}.
3100 Note these are only possible candidates, not absolute ones. Care should
3101 be taken to manually verify functions actually do not ever return before
3102 adding the @code{noreturn} attribute, otherwise subtle code generation
3103 bugs could be introduced. You will not get a warning for @code{main} in
3104 hosted C environments.
3106 @item -Wmissing-format-attribute
3107 @opindex Wmissing-format-attribute
3109 Warn about function pointers which might be candidates for @code{format}
3110 attributes. Note these are only possible candidates, not absolute ones.
3111 GCC will guess that function pointers with @code{format} attributes that
3112 are used in assignment, initialization, parameter passing or return
3113 statements should have a corresponding @code{format} attribute in the
3114 resulting type. I.e.@: the left-hand side of the assignment or
3115 initialization, the type of the parameter variable, or the return type
3116 of the containing function respectively should also have a @code{format}
3117 attribute to avoid the warning.
3119 GCC will also warn about function definitions which might be
3120 candidates for @code{format} attributes. Again, these are only
3121 possible candidates. GCC will guess that @code{format} attributes
3122 might be appropriate for any function that calls a function like
3123 @code{vprintf} or @code{vscanf}, but this might not always be the
3124 case, and some functions for which @code{format} attributes are
3125 appropriate may not be detected.
3127 @item -Wno-multichar
3128 @opindex Wno-multichar
3130 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3131 Usually they indicate a typo in the user's code, as they have
3132 implementation-defined values, and should not be used in portable code.
3134 @item -Wnormalized=<none|id|nfc|nfkc>
3135 @opindex Wnormalized
3138 @cindex character set, input normalization
3139 In ISO C and ISO C++, two identifiers are different if they are
3140 different sequences of characters. However, sometimes when characters
3141 outside the basic ASCII character set are used, you can have two
3142 different character sequences that look the same. To avoid confusion,
3143 the ISO 10646 standard sets out some @dfn{normalization rules} which
3144 when applied ensure that two sequences that look the same are turned into
3145 the same sequence. GCC can warn you if you are using identifiers which
3146 have not been normalized; this option controls that warning.
3148 There are four levels of warning that GCC supports. The default is
3149 @option{-Wnormalized=nfc}, which warns about any identifier which is
3150 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3151 recommended form for most uses.
3153 Unfortunately, there are some characters which ISO C and ISO C++ allow
3154 in identifiers that when turned into NFC aren't allowable as
3155 identifiers. That is, there's no way to use these symbols in portable
3156 ISO C or C++ and have all your identifiers in NFC.
3157 @option{-Wnormalized=id} suppresses the warning for these characters.
3158 It is hoped that future versions of the standards involved will correct
3159 this, which is why this option is not the default.
3161 You can switch the warning off for all characters by writing
3162 @option{-Wnormalized=none}. You would only want to do this if you
3163 were using some other normalization scheme (like ``D''), because
3164 otherwise you can easily create bugs that are literally impossible to see.
3166 Some characters in ISO 10646 have distinct meanings but look identical
3167 in some fonts or display methodologies, especially once formatting has
3168 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3169 LETTER N'', will display just like a regular @code{n} which has been
3170 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3171 normalization scheme to convert all these into a standard form as
3172 well, and GCC will warn if your code is not in NFKC if you use
3173 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3174 about every identifier that contains the letter O because it might be
3175 confused with the digit 0, and so is not the default, but may be
3176 useful as a local coding convention if the programming environment is
3177 unable to be fixed to display these characters distinctly.
3179 @item -Wno-deprecated-declarations
3180 @opindex Wno-deprecated-declarations
3181 Do not warn about uses of functions (@pxref{Function Attributes}),
3182 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3183 Attributes}) marked as deprecated by using the @code{deprecated}
3187 @opindex Wno-overflow
3188 Do not warn about compile-time overflow in constant expressions.
3190 @item -Woverride-init
3191 @opindex Woverride-init
3194 Warn if an initialized field without side effects is overridden when
3195 using designated initializers (@pxref{Designated Inits, , Designated
3198 This warning is included in @option{-Wextra}. To get other
3199 @option{-Wextra} warnings without this one, use @samp{-Wextra
3200 -Wno-override-init}.
3204 Warn if a structure is given the packed attribute, but the packed
3205 attribute has no effect on the layout or size of the structure.
3206 Such structures may be mis-aligned for little benefit. For
3207 instance, in this code, the variable @code{f.x} in @code{struct bar}
3208 will be misaligned even though @code{struct bar} does not itself
3209 have the packed attribute:
3216 @} __attribute__((packed));
3226 Warn if padding is included in a structure, either to align an element
3227 of the structure or to align the whole structure. Sometimes when this
3228 happens it is possible to rearrange the fields of the structure to
3229 reduce the padding and so make the structure smaller.
3231 @item -Wredundant-decls
3232 @opindex Wredundant-decls
3233 Warn if anything is declared more than once in the same scope, even in
3234 cases where multiple declaration is valid and changes nothing.
3236 @item -Wnested-externs @r{(C only)}
3237 @opindex Wnested-externs
3238 Warn if an @code{extern} declaration is encountered within a function.
3240 @item -Wunreachable-code
3241 @opindex Wunreachable-code
3242 Warn if the compiler detects that code will never be executed.
3244 This option is intended to warn when the compiler detects that at
3245 least a whole line of source code will never be executed, because
3246 some condition is never satisfied or because it is after a
3247 procedure that never returns.
3249 It is possible for this option to produce a warning even though there
3250 are circumstances under which part of the affected line can be executed,
3251 so care should be taken when removing apparently-unreachable code.
3253 For instance, when a function is inlined, a warning may mean that the
3254 line is unreachable in only one inlined copy of the function.
3256 This option is not made part of @option{-Wall} because in a debugging
3257 version of a program there is often substantial code which checks
3258 correct functioning of the program and is, hopefully, unreachable
3259 because the program does work. Another common use of unreachable
3260 code is to provide behavior which is selectable at compile-time.
3264 Warn if a function can not be inlined and it was declared as inline.
3265 Even with this option, the compiler will not warn about failures to
3266 inline functions declared in system headers.
3268 The compiler uses a variety of heuristics to determine whether or not
3269 to inline a function. For example, the compiler takes into account
3270 the size of the function being inlined and the amount of inlining
3271 that has already been done in the current function. Therefore,
3272 seemingly insignificant changes in the source program can cause the
3273 warnings produced by @option{-Winline} to appear or disappear.
3275 @item -Wno-invalid-offsetof @r{(C++ only)}
3276 @opindex Wno-invalid-offsetof
3277 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3278 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3279 to a non-POD type is undefined. In existing C++ implementations,
3280 however, @samp{offsetof} typically gives meaningful results even when
3281 applied to certain kinds of non-POD types. (Such as a simple
3282 @samp{struct} that fails to be a POD type only by virtue of having a
3283 constructor.) This flag is for users who are aware that they are
3284 writing nonportable code and who have deliberately chosen to ignore the
3287 The restrictions on @samp{offsetof} may be relaxed in a future version
3288 of the C++ standard.
3290 @item -Wno-int-to-pointer-cast @r{(C only)}
3291 @opindex Wno-int-to-pointer-cast
3292 Suppress warnings from casts to pointer type of an integer of a
3295 @item -Wno-pointer-to-int-cast @r{(C only)}
3296 @opindex Wno-pointer-to-int-cast
3297 Suppress warnings from casts from a pointer to an integer type of a
3301 @opindex Winvalid-pch
3302 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3303 the search path but can't be used.
3307 @opindex Wno-long-long
3308 Warn if @samp{long long} type is used. This is default. To inhibit
3309 the warning messages, use @option{-Wno-long-long}. Flags
3310 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3311 only when @option{-pedantic} flag is used.
3313 @item -Wvariadic-macros
3314 @opindex Wvariadic-macros
3315 @opindex Wno-variadic-macros
3316 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3317 alternate syntax when in pedantic ISO C99 mode. This is default.
3318 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3323 Warn if variable length array is used in the code.
3324 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
3325 the variable length array.
3327 @item -Wvolatile-register-var
3328 @opindex Wvolatile-register-var
3329 @opindex Wno-volatile-register-var
3330 Warn if a register variable is declared volatile. The volatile
3331 modifier does not inhibit all optimizations that may eliminate reads
3332 and/or writes to register variables.
3334 @item -Wdisabled-optimization
3335 @opindex Wdisabled-optimization
3336 Warn if a requested optimization pass is disabled. This warning does
3337 not generally indicate that there is anything wrong with your code; it
3338 merely indicates that GCC's optimizers were unable to handle the code
3339 effectively. Often, the problem is that your code is too big or too
3340 complex; GCC will refuse to optimize programs when the optimization
3341 itself is likely to take inordinate amounts of time.
3343 @item -Wpointer-sign
3344 @opindex Wpointer-sign
3345 @opindex Wno-pointer-sign
3346 Warn for pointer argument passing or assignment with different signedness.
3347 This option is only supported for C. It is implied by @option{-Wall}
3348 and by @option{-pedantic}, which can be disabled with
3349 @option{-Wno-pointer-sign}.
3353 Make all warnings into errors.
3357 Make the specified warning into an errors. The specifier for a
3358 warning is appended, for example @option{-Werror=switch} turns the
3359 warnings controlled by @option{-Wswitch} into errors. This switch
3360 takes a negative form, to be used to negate @option{-Werror} for
3361 specific warnings, for example @option{-Wno-error=switch} makes
3362 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3363 is in effect. You can use the @option{-fdiagnostics-show-option}
3364 option to have each controllable warning amended with the option which
3365 controls it, to determine what to use with this option.
3367 Note that specifying @option{-Werror=}@var{foo} automatically implies
3368 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3371 @item -Wstack-protector
3372 @opindex Wstack-protector
3373 This option is only active when @option{-fstack-protector} is active. It
3374 warns about functions that will not be protected against stack smashing.
3376 @item -Woverlength-strings
3377 @opindex Woverlength-strings
3378 Warn about string constants which are longer than the ``minimum
3379 maximum'' length specified in the C standard. Modern compilers
3380 generally allow string constants which are much longer than the
3381 standard's minimum limit, but very portable programs should avoid
3382 using longer strings.
3384 The limit applies @emph{after} string constant concatenation, and does
3385 not count the trailing NUL@. In C89, the limit was 509 characters; in
3386 C99, it was raised to 4095. C++98 does not specify a normative
3387 minimum maximum, so we do not diagnose overlength strings in C++@.
3389 This option is implied by @option{-pedantic}, and can be disabled with
3390 @option{-Wno-overlength-strings}.
3393 @node Debugging Options
3394 @section Options for Debugging Your Program or GCC
3395 @cindex options, debugging
3396 @cindex debugging information options
3398 GCC has various special options that are used for debugging
3399 either your program or GCC:
3404 Produce debugging information in the operating system's native format
3405 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3408 On most systems that use stabs format, @option{-g} enables use of extra
3409 debugging information that only GDB can use; this extra information
3410 makes debugging work better in GDB but will probably make other debuggers
3412 refuse to read the program. If you want to control for certain whether
3413 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3414 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3416 GCC allows you to use @option{-g} with
3417 @option{-O}. The shortcuts taken by optimized code may occasionally
3418 produce surprising results: some variables you declared may not exist
3419 at all; flow of control may briefly move where you did not expect it;
3420 some statements may not be executed because they compute constant
3421 results or their values were already at hand; some statements may
3422 execute in different places because they were moved out of loops.
3424 Nevertheless it proves possible to debug optimized output. This makes
3425 it reasonable to use the optimizer for programs that might have bugs.
3427 The following options are useful when GCC is generated with the
3428 capability for more than one debugging format.
3432 Produce debugging information for use by GDB@. This means to use the
3433 most expressive format available (DWARF 2, stabs, or the native format
3434 if neither of those are supported), including GDB extensions if at all
3439 Produce debugging information in stabs format (if that is supported),
3440 without GDB extensions. This is the format used by DBX on most BSD
3441 systems. On MIPS, Alpha and System V Release 4 systems this option
3442 produces stabs debugging output which is not understood by DBX or SDB@.
3443 On System V Release 4 systems this option requires the GNU assembler.
3445 @item -feliminate-unused-debug-symbols
3446 @opindex feliminate-unused-debug-symbols
3447 Produce debugging information in stabs format (if that is supported),
3448 for only symbols that are actually used.
3450 @item -femit-class-debug-always
3451 Instead of emitting debugging information for a C++ class in only one
3452 object file, emit it in all object files using the class. This option
3453 should be used only with debuggers that are unable to handle the way GCC
3454 normally emits debugging information for classes because using this
3455 option will increase the size of debugging information by as much as a
3460 Produce debugging information in stabs format (if that is supported),
3461 using GNU extensions understood only by the GNU debugger (GDB)@. The
3462 use of these extensions is likely to make other debuggers crash or
3463 refuse to read the program.
3467 Produce debugging information in COFF format (if that is supported).
3468 This is the format used by SDB on most System V systems prior to
3473 Produce debugging information in XCOFF format (if that is supported).
3474 This is the format used by the DBX debugger on IBM RS/6000 systems.
3478 Produce debugging information in XCOFF format (if that is supported),
3479 using GNU extensions understood only by the GNU debugger (GDB)@. The
3480 use of these extensions is likely to make other debuggers crash or
3481 refuse to read the program, and may cause assemblers other than the GNU
3482 assembler (GAS) to fail with an error.
3486 Produce debugging information in DWARF version 2 format (if that is
3487 supported). This is the format used by DBX on IRIX 6. With this
3488 option, GCC uses features of DWARF version 3 when they are useful;
3489 version 3 is upward compatible with version 2, but may still cause
3490 problems for older debuggers.
3494 Produce debugging information in VMS debug format (if that is
3495 supported). This is the format used by DEBUG on VMS systems.
3498 @itemx -ggdb@var{level}
3499 @itemx -gstabs@var{level}
3500 @itemx -gcoff@var{level}
3501 @itemx -gxcoff@var{level}
3502 @itemx -gvms@var{level}
3503 Request debugging information and also use @var{level} to specify how
3504 much information. The default level is 2.
3506 Level 1 produces minimal information, enough for making backtraces in
3507 parts of the program that you don't plan to debug. This includes
3508 descriptions of functions and external variables, but no information
3509 about local variables and no line numbers.
3511 Level 3 includes extra information, such as all the macro definitions
3512 present in the program. Some debuggers support macro expansion when
3513 you use @option{-g3}.
3515 @option{-gdwarf-2} does not accept a concatenated debug level, because
3516 GCC used to support an option @option{-gdwarf} that meant to generate
3517 debug information in version 1 of the DWARF format (which is very
3518 different from version 2), and it would have been too confusing. That
3519 debug format is long obsolete, but the option cannot be changed now.
3520 Instead use an additional @option{-g@var{level}} option to change the
3521 debug level for DWARF2.
3523 @item -feliminate-dwarf2-dups
3524 @opindex feliminate-dwarf2-dups
3525 Compress DWARF2 debugging information by eliminating duplicated
3526 information about each symbol. This option only makes sense when
3527 generating DWARF2 debugging information with @option{-gdwarf-2}.
3529 @item -femit-struct-debug-baseonly
3530 Emit debug information for struct-like types
3531 only when the base name of the compilation source file
3532 matches the base name of file in which the struct was defined.
3534 This option substantially reduces the size of debugging information,
3535 but at significant potential loss in type information to the debugger.
3536 See @option{-femit-struct-debug-reduced} for a less aggressive option.
3537 See @option{-femit-struct-debug-detailed} for more detailed control.
3539 This option works only with DWARF 2.
3541 @item -femit-struct-debug-reduced
3542 Emit debug information for struct-like types
3543 only when the base name of the compilation source file
3544 matches the base name of file in which the type was defined,
3545 unless the struct is a template or defined in a system header.
3547 This option significantly reduces the size of debugging information,
3548 with some potential loss in type information to the debugger.
3549 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
3550 See @option{-femit-struct-debug-detailed} for more detailed control.
3552 This option works only with DWARF 2.
3554 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
3555 Specify the struct-like types
3556 for which the compiler will generate debug information.
3557 The intent is to reduce duplicate struct debug information
3558 between different object files within the same program.
3560 This option is a detailed version of
3561 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
3562 which will serve for most needs.
3564 A specification has the syntax
3565 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
3567 The optional first word limits the specification to
3568 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
3569 A struct type is used directly when it is the type of a variable, member.
3570 Indirect uses arise through pointers to structs.
3571 That is, when use of an incomplete struct would be legal, the use is indirect.
3573 @samp{struct one direct; struct two * indirect;}.
3575 The optional second word limits the specification to
3576 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
3577 Generic structs are a bit complicated to explain.
3578 For C++, these are non-explicit specializations of template classes,
3579 or non-template classes within the above.
3580 Other programming languages have generics,
3581 but @samp{-femit-struct-debug-detailed} does not yet implement them.
3583 The third word specifies the source files for those
3584 structs for which the compiler will emit debug information.
3585 The values @samp{none} and @samp{any} have the normal meaning.
3586 The value @samp{base} means that
3587 the base of name of the file in which the type declaration appears
3588 must match the base of the name of the main compilation file.
3589 In practice, this means that
3590 types declared in @file{foo.c} and @file{foo.h} will have debug information,
3591 but types declared in other header will not.
3592 The value @samp{sys} means those types satisfying @samp{base}
3593 or declared in system or compiler headers.
3595 You may need to experiment to determine the best settings for your application.
3597 The default is @samp{-femit-struct-debug-detailed=all}.
3599 This option works only with DWARF 2.
3601 @cindex @command{prof}
3604 Generate extra code to write profile information suitable for the
3605 analysis program @command{prof}. You must use this option when compiling
3606 the source files you want data about, and you must also use it when
3609 @cindex @command{gprof}
3612 Generate extra code to write profile information suitable for the
3613 analysis program @command{gprof}. You must use this option when compiling
3614 the source files you want data about, and you must also use it when
3619 Makes the compiler print out each function name as it is compiled, and
3620 print some statistics about each pass when it finishes.
3623 @opindex ftime-report
3624 Makes the compiler print some statistics about the time consumed by each
3625 pass when it finishes.
3628 @opindex fmem-report
3629 Makes the compiler print some statistics about permanent memory
3630 allocation when it finishes.
3632 @item -fprofile-arcs
3633 @opindex fprofile-arcs
3634 Add code so that program flow @dfn{arcs} are instrumented. During
3635 execution the program records how many times each branch and call is
3636 executed and how many times it is taken or returns. When the compiled
3637 program exits it saves this data to a file called
3638 @file{@var{auxname}.gcda} for each source file. The data may be used for
3639 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3640 test coverage analysis (@option{-ftest-coverage}). Each object file's
3641 @var{auxname} is generated from the name of the output file, if
3642 explicitly specified and it is not the final executable, otherwise it is
3643 the basename of the source file. In both cases any suffix is removed
3644 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3645 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3646 @xref{Cross-profiling}.
3648 @cindex @command{gcov}
3652 This option is used to compile and link code instrumented for coverage
3653 analysis. The option is a synonym for @option{-fprofile-arcs}
3654 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3655 linking). See the documentation for those options for more details.
3660 Compile the source files with @option{-fprofile-arcs} plus optimization
3661 and code generation options. For test coverage analysis, use the
3662 additional @option{-ftest-coverage} option. You do not need to profile
3663 every source file in a program.
3666 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3667 (the latter implies the former).
3670 Run the program on a representative workload to generate the arc profile
3671 information. This may be repeated any number of times. You can run
3672 concurrent instances of your program, and provided that the file system
3673 supports locking, the data files will be correctly updated. Also
3674 @code{fork} calls are detected and correctly handled (double counting
3678 For profile-directed optimizations, compile the source files again with
3679 the same optimization and code generation options plus
3680 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3681 Control Optimization}).
3684 For test coverage analysis, use @command{gcov} to produce human readable
3685 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3686 @command{gcov} documentation for further information.
3690 With @option{-fprofile-arcs}, for each function of your program GCC
3691 creates a program flow graph, then finds a spanning tree for the graph.
3692 Only arcs that are not on the spanning tree have to be instrumented: the
3693 compiler adds code to count the number of times that these arcs are
3694 executed. When an arc is the only exit or only entrance to a block, the
3695 instrumentation code can be added to the block; otherwise, a new basic
3696 block must be created to hold the instrumentation code.
3699 @item -ftest-coverage
3700 @opindex ftest-coverage
3701 Produce a notes file that the @command{gcov} code-coverage utility
3702 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3703 show program coverage. Each source file's note file is called
3704 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3705 above for a description of @var{auxname} and instructions on how to
3706 generate test coverage data. Coverage data will match the source files
3707 more closely, if you do not optimize.
3709 @item -d@var{letters}
3710 @item -fdump-rtl-@var{pass}
3712 Says to make debugging dumps during compilation at times specified by
3713 @var{letters}. This is used for debugging the RTL-based passes of the
3714 compiler. The file names for most of the dumps are made by appending a
3715 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3716 from the name of the output file, if explicitly specified and it is not
3717 an executable, otherwise it is the basename of the source file. These
3718 switches may have different effects when @option{-E} is used for
3721 Most debug dumps can be enabled either passing a letter to the @option{-d}
3722 option, or with a long @option{-fdump-rtl} switch; here are the possible
3723 letters for use in @var{letters} and @var{pass}, and their meanings:
3728 Annotate the assembler output with miscellaneous debugging information.
3731 @itemx -fdump-rtl-bbro
3733 @opindex fdump-rtl-bbro
3734 Dump after block reordering, to @file{@var{file}.148r.bbro}.
3737 @itemx -fdump-rtl-combine
3739 @opindex fdump-rtl-combine
3740 Dump after instruction combination, to the file @file{@var{file}.129r.combine}.
3743 @itemx -fdump-rtl-ce1
3744 @itemx -fdump-rtl-ce2
3746 @opindex fdump-rtl-ce1
3747 @opindex fdump-rtl-ce2
3748 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3749 first if conversion, to the file @file{@var{file}.117r.ce1}. @option{-dC}
3750 and @option{-fdump-rtl-ce2} enable dumping after the second if
3751 conversion, to the file @file{@var{file}.130r.ce2}.
3754 @itemx -fdump-rtl-btl
3755 @itemx -fdump-rtl-dbr
3757 @opindex fdump-rtl-btl
3758 @opindex fdump-rtl-dbr
3759 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3760 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3761 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3762 scheduling, to @file{@var{file}.36.dbr}.
3766 Dump all macro definitions, at the end of preprocessing, in addition to
3770 @itemx -fdump-rtl-ce3
3772 @opindex fdump-rtl-ce3
3773 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
3776 @itemx -fdump-rtl-cfg
3777 @itemx -fdump-rtl-life
3779 @opindex fdump-rtl-cfg
3780 @opindex fdump-rtl-life
3781 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3782 and data flow analysis, to @file{@var{file}.116r.cfg}. @option{-df}
3783 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3784 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
3787 @itemx -fdump-rtl-greg
3789 @opindex fdump-rtl-greg
3790 Dump after global register allocation, to @file{@var{file}.139r.greg}.
3793 @itemx -fdump-rtl-gcse
3794 @itemx -fdump-rtl-bypass
3796 @opindex fdump-rtl-gcse
3797 @opindex fdump-rtl-bypass
3798 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3799 @file{@var{file}.114r.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3800 enable dumping after jump bypassing and control flow optimizations, to
3801 @file{@var{file}.115r.bypass}.
3804 @itemx -fdump-rtl-eh
3806 @opindex fdump-rtl-eh
3807 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3810 @itemx -fdump-rtl-sibling
3812 @opindex fdump-rtl-sibling
3813 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
3816 @itemx -fdump-rtl-jump
3818 @opindex fdump-rtl-jump
3819 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
3822 @itemx -fdump-rtl-stack
3824 @opindex fdump-rtl-stack
3825 Dump after conversion from registers to stack, to @file{@var{file}.152r.stack}.
3828 @itemx -fdump-rtl-lreg
3830 @opindex fdump-rtl-lreg
3831 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
3834 @itemx -fdump-rtl-loop2
3836 @opindex fdump-rtl-loop2
3837 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
3838 loop optimization pass, to @file{@var{file}.119r.loop2},
3839 @file{@var{file}.120r.loop2_init},
3840 @file{@var{file}.121r.loop2_invariant}, and
3841 @file{@var{file}.125r.loop2_done}.
3844 @itemx -fdump-rtl-sms
3846 @opindex fdump-rtl-sms
3847 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
3850 @itemx -fdump-rtl-mach
3852 @opindex fdump-rtl-mach
3853 Dump after performing the machine dependent reorganization pass, to
3854 @file{@var{file}.155r.mach} if that pass exists.
3857 @itemx -fdump-rtl-rnreg
3859 @opindex fdump-rtl-rnreg
3860 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
3863 @itemx -fdump-rtl-regmove
3865 @opindex fdump-rtl-regmove
3866 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
3869 @itemx -fdump-rtl-postreload
3871 @opindex fdump-rtl-postreload
3872 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3875 @itemx -fdump-rtl-expand
3877 @opindex fdump-rtl-expand
3878 Dump after RTL generation, to @file{@var{file}.104r.expand}.
3881 @itemx -fdump-rtl-sched2
3883 @opindex fdump-rtl-sched2
3884 Dump after the second scheduling pass, to @file{@var{file}.150r.sched2}.
3887 @itemx -fdump-rtl-cse
3889 @opindex fdump-rtl-cse
3890 Dump after CSE (including the jump optimization that sometimes follows
3891 CSE), to @file{@var{file}.113r.cse}.
3894 @itemx -fdump-rtl-sched
3896 @opindex fdump-rtl-sched
3897 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3900 @itemx -fdump-rtl-cse2
3902 @opindex fdump-rtl-cse2
3903 Dump after the second CSE pass (including the jump optimization that
3904 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
3907 @itemx -fdump-rtl-tracer
3909 @opindex fdump-rtl-tracer
3910 Dump after running tracer, to @file{@var{file}.118r.tracer}.
3913 @itemx -fdump-rtl-vpt
3914 @itemx -fdump-rtl-vartrack
3916 @opindex fdump-rtl-vpt
3917 @opindex fdump-rtl-vartrack
3918 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3919 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3920 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3921 to @file{@var{file}.154r.vartrack}.
3924 @itemx -fdump-rtl-flow2
3926 @opindex fdump-rtl-flow2
3927 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
3930 @itemx -fdump-rtl-peephole2
3932 @opindex fdump-rtl-peephole2
3933 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
3936 @itemx -fdump-rtl-web
3938 @opindex fdump-rtl-web
3939 Dump after live range splitting, to @file{@var{file}.126r.web}.
3942 @itemx -fdump-rtl-all
3944 @opindex fdump-rtl-all
3945 Produce all the dumps listed above.
3949 Produce a core dump whenever an error occurs.
3953 Print statistics on memory usage, at the end of the run, to
3958 Annotate the assembler output with a comment indicating which
3959 pattern and alternative was used. The length of each instruction is
3964 Dump the RTL in the assembler output as a comment before each instruction.
3965 Also turns on @option{-dp} annotation.
3969 For each of the other indicated dump files (either with @option{-d} or
3970 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3971 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3975 Just generate RTL for a function instead of compiling it. Usually used
3976 with @samp{r} (@option{-fdump-rtl-expand}).
3980 Dump debugging information during parsing, to standard error.
3984 @opindex fdump-noaddr
3985 When doing debugging dumps (see @option{-d} option above), suppress
3986 address output. This makes it more feasible to use diff on debugging
3987 dumps for compiler invocations with different compiler binaries and/or
3988 different text / bss / data / heap / stack / dso start locations.
3990 @item -fdump-unnumbered
3991 @opindex fdump-unnumbered
3992 When doing debugging dumps (see @option{-d} option above), suppress instruction
3993 numbers, line number note and address output. This makes it more feasible to
3994 use diff on debugging dumps for compiler invocations with different
3995 options, in particular with and without @option{-g}.
3997 @item -fdump-translation-unit @r{(C++ only)}
3998 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
3999 @opindex fdump-translation-unit
4000 Dump a representation of the tree structure for the entire translation
4001 unit to a file. The file name is made by appending @file{.tu} to the
4002 source file name. If the @samp{-@var{options}} form is used, @var{options}
4003 controls the details of the dump as described for the
4004 @option{-fdump-tree} options.
4006 @item -fdump-class-hierarchy @r{(C++ only)}
4007 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4008 @opindex fdump-class-hierarchy
4009 Dump a representation of each class's hierarchy and virtual function
4010 table layout to a file. The file name is made by appending @file{.class}
4011 to the source file name. If the @samp{-@var{options}} form is used,
4012 @var{options} controls the details of the dump as described for the
4013 @option{-fdump-tree} options.
4015 @item -fdump-ipa-@var{switch}
4017 Control the dumping at various stages of inter-procedural analysis
4018 language tree to a file. The file name is generated by appending a switch
4019 specific suffix to the source file name. The following dumps are possible:
4023 Enables all inter-procedural analysis dumps; currently the only produced
4024 dump is the @samp{cgraph} dump.
4027 Dumps information about call-graph optimization, unused function removal,
4028 and inlining decisions.
4031 @item -fdump-tree-@var{switch}
4032 @itemx -fdump-tree-@var{switch}-@var{options}
4034 Control the dumping at various stages of processing the intermediate
4035 language tree to a file. The file name is generated by appending a switch
4036 specific suffix to the source file name. If the @samp{-@var{options}}
4037 form is used, @var{options} is a list of @samp{-} separated options that
4038 control the details of the dump. Not all options are applicable to all
4039 dumps, those which are not meaningful will be ignored. The following
4040 options are available
4044 Print the address of each node. Usually this is not meaningful as it
4045 changes according to the environment and source file. Its primary use
4046 is for tying up a dump file with a debug environment.
4048 Inhibit dumping of members of a scope or body of a function merely
4049 because that scope has been reached. Only dump such items when they
4050 are directly reachable by some other path. When dumping pretty-printed
4051 trees, this option inhibits dumping the bodies of control structures.
4053 Print a raw representation of the tree. By default, trees are
4054 pretty-printed into a C-like representation.
4056 Enable more detailed dumps (not honored by every dump option).
4058 Enable dumping various statistics about the pass (not honored by every dump
4061 Enable showing basic block boundaries (disabled in raw dumps).
4063 Enable showing virtual operands for every statement.
4065 Enable showing line numbers for statements.
4067 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4069 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4072 The following tree dumps are possible:
4076 Dump before any tree based optimization, to @file{@var{file}.original}.
4079 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4082 Dump after function inlining, to @file{@var{file}.inlined}.
4085 @opindex fdump-tree-gimple
4086 Dump each function before and after the gimplification pass to a file. The
4087 file name is made by appending @file{.gimple} to the source file name.
4090 @opindex fdump-tree-cfg
4091 Dump the control flow graph of each function to a file. The file name is
4092 made by appending @file{.cfg} to the source file name.
4095 @opindex fdump-tree-vcg
4096 Dump the control flow graph of each function to a file in VCG format. The
4097 file name is made by appending @file{.vcg} to the source file name. Note
4098 that if the file contains more than one function, the generated file cannot
4099 be used directly by VCG@. You will need to cut and paste each function's
4100 graph into its own separate file first.
4103 @opindex fdump-tree-ch
4104 Dump each function after copying loop headers. The file name is made by
4105 appending @file{.ch} to the source file name.
4108 @opindex fdump-tree-ssa
4109 Dump SSA related information to a file. The file name is made by appending
4110 @file{.ssa} to the source file name.
4113 @opindex fdump-tree-salias
4114 Dump structure aliasing variable information to a file. This file name
4115 is made by appending @file{.salias} to the source file name.
4118 @opindex fdump-tree-alias
4119 Dump aliasing information for each function. The file name is made by
4120 appending @file{.alias} to the source file name.
4123 @opindex fdump-tree-ccp
4124 Dump each function after CCP@. The file name is made by appending
4125 @file{.ccp} to the source file name.
4128 @opindex fdump-tree-storeccp
4129 Dump each function after STORE-CCP. The file name is made by appending
4130 @file{.storeccp} to the source file name.
4133 @opindex fdump-tree-pre
4134 Dump trees after partial redundancy elimination. The file name is made
4135 by appending @file{.pre} to the source file name.
4138 @opindex fdump-tree-fre
4139 Dump trees after full redundancy elimination. The file name is made
4140 by appending @file{.fre} to the source file name.
4143 @opindex fdump-tree-copyprop
4144 Dump trees after copy propagation. The file name is made
4145 by appending @file{.copyprop} to the source file name.
4147 @item store_copyprop
4148 @opindex fdump-tree-store_copyprop
4149 Dump trees after store copy-propagation. The file name is made
4150 by appending @file{.store_copyprop} to the source file name.
4153 @opindex fdump-tree-dce
4154 Dump each function after dead code elimination. The file name is made by
4155 appending @file{.dce} to the source file name.
4158 @opindex fdump-tree-mudflap
4159 Dump each function after adding mudflap instrumentation. The file name is
4160 made by appending @file{.mudflap} to the source file name.
4163 @opindex fdump-tree-sra
4164 Dump each function after performing scalar replacement of aggregates. The
4165 file name is made by appending @file{.sra} to the source file name.
4168 @opindex fdump-tree-sink
4169 Dump each function after performing code sinking. The file name is made
4170 by appending @file{.sink} to the source file name.
4173 @opindex fdump-tree-dom
4174 Dump each function after applying dominator tree optimizations. The file
4175 name is made by appending @file{.dom} to the source file name.
4178 @opindex fdump-tree-dse
4179 Dump each function after applying dead store elimination. The file
4180 name is made by appending @file{.dse} to the source file name.
4183 @opindex fdump-tree-phiopt
4184 Dump each function after optimizing PHI nodes into straightline code. The file
4185 name is made by appending @file{.phiopt} to the source file name.
4188 @opindex fdump-tree-forwprop
4189 Dump each function after forward propagating single use variables. The file
4190 name is made by appending @file{.forwprop} to the source file name.
4193 @opindex fdump-tree-copyrename
4194 Dump each function after applying the copy rename optimization. The file
4195 name is made by appending @file{.copyrename} to the source file name.
4198 @opindex fdump-tree-nrv
4199 Dump each function after applying the named return value optimization on
4200 generic trees. The file name is made by appending @file{.nrv} to the source
4204 @opindex fdump-tree-vect
4205 Dump each function after applying vectorization of loops. The file name is
4206 made by appending @file{.vect} to the source file name.
4209 @opindex fdump-tree-vrp
4210 Dump each function after Value Range Propagation (VRP). The file name
4211 is made by appending @file{.vrp} to the source file name.
4214 @opindex fdump-tree-all
4215 Enable all the available tree dumps with the flags provided in this option.
4218 @item -ftree-vectorizer-verbose=@var{n}
4219 @opindex ftree-vectorizer-verbose
4220 This option controls the amount of debugging output the vectorizer prints.
4221 This information is written to standard error, unless
4222 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4223 in which case it is output to the usual dump listing file, @file{.vect}.
4224 For @var{n}=0 no diagnostic information is reported.
4225 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4226 and the total number of loops that got vectorized.
4227 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4228 the first analysis phase (vect_analyze_loop_form) - i.e. countable,
4229 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4230 level that @option{-fdump-tree-vect-stats} uses.
4231 Higher verbosity levels mean either more information dumped for each
4232 reported loop, or same amount of information reported for more loops:
4233 If @var{n}=3, alignment related information is added to the reports.
4234 If @var{n}=4, data-references related information (e.g. memory dependences,
4235 memory access-patterns) is added to the reports.
4236 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4237 that did not pass the first analysis phase (i.e. may not be countable, or
4238 may have complicated control-flow).
4239 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4240 For @var{n}=7, all the information the vectorizer generates during its
4241 analysis and transformation is reported. This is the same verbosity level
4242 that @option{-fdump-tree-vect-details} uses.
4244 @item -frandom-seed=@var{string}
4245 @opindex frandom-string
4246 This option provides a seed that GCC uses when it would otherwise use
4247 random numbers. It is used to generate certain symbol names
4248 that have to be different in every compiled file. It is also used to
4249 place unique stamps in coverage data files and the object files that
4250 produce them. You can use the @option{-frandom-seed} option to produce
4251 reproducibly identical object files.
4253 The @var{string} should be different for every file you compile.
4255 @item -fsched-verbose=@var{n}
4256 @opindex fsched-verbose
4257 On targets that use instruction scheduling, this option controls the
4258 amount of debugging output the scheduler prints. This information is
4259 written to standard error, unless @option{-dS} or @option{-dR} is
4260 specified, in which case it is output to the usual dump
4261 listing file, @file{.sched} or @file{.sched2} respectively. However
4262 for @var{n} greater than nine, the output is always printed to standard
4265 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4266 same information as @option{-dRS}. For @var{n} greater than one, it
4267 also output basic block probabilities, detailed ready list information
4268 and unit/insn info. For @var{n} greater than two, it includes RTL
4269 at abort point, control-flow and regions info. And for @var{n} over
4270 four, @option{-fsched-verbose} also includes dependence info.
4274 Store the usual ``temporary'' intermediate files permanently; place them
4275 in the current directory and name them based on the source file. Thus,
4276 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4277 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4278 preprocessed @file{foo.i} output file even though the compiler now
4279 normally uses an integrated preprocessor.
4281 When used in combination with the @option{-x} command line option,
4282 @option{-save-temps} is sensible enough to avoid over writing an
4283 input source file with the same extension as an intermediate file.
4284 The corresponding intermediate file may be obtained by renaming the
4285 source file before using @option{-save-temps}.
4289 Report the CPU time taken by each subprocess in the compilation
4290 sequence. For C source files, this is the compiler proper and assembler
4291 (plus the linker if linking is done). The output looks like this:
4298 The first number on each line is the ``user time'', that is time spent
4299 executing the program itself. The second number is ``system time'',
4300 time spent executing operating system routines on behalf of the program.
4301 Both numbers are in seconds.
4303 @item -fvar-tracking
4304 @opindex fvar-tracking
4305 Run variable tracking pass. It computes where variables are stored at each
4306 position in code. Better debugging information is then generated
4307 (if the debugging information format supports this information).
4309 It is enabled by default when compiling with optimization (@option{-Os},
4310 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4311 the debug info format supports it.
4313 @item -print-file-name=@var{library}
4314 @opindex print-file-name
4315 Print the full absolute name of the library file @var{library} that
4316 would be used when linking---and don't do anything else. With this
4317 option, GCC does not compile or link anything; it just prints the
4320 @item -print-multi-directory
4321 @opindex print-multi-directory
4322 Print the directory name corresponding to the multilib selected by any
4323 other switches present in the command line. This directory is supposed
4324 to exist in @env{GCC_EXEC_PREFIX}.
4326 @item -print-multi-lib
4327 @opindex print-multi-lib
4328 Print the mapping from multilib directory names to compiler switches
4329 that enable them. The directory name is separated from the switches by
4330 @samp{;}, and each switch starts with an @samp{@@} instead of the
4331 @samp{-}, without spaces between multiple switches. This is supposed to
4332 ease shell-processing.
4334 @item -print-prog-name=@var{program}
4335 @opindex print-prog-name
4336 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4338 @item -print-libgcc-file-name
4339 @opindex print-libgcc-file-name
4340 Same as @option{-print-file-name=libgcc.a}.
4342 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4343 but you do want to link with @file{libgcc.a}. You can do
4346 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4349 @item -print-search-dirs
4350 @opindex print-search-dirs
4351 Print the name of the configured installation directory and a list of
4352 program and library directories @command{gcc} will search---and don't do anything else.
4354 This is useful when @command{gcc} prints the error message
4355 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4356 To resolve this you either need to put @file{cpp0} and the other compiler
4357 components where @command{gcc} expects to find them, or you can set the environment
4358 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4359 Don't forget the trailing @samp{/}.
4360 @xref{Environment Variables}.
4363 @opindex dumpmachine
4364 Print the compiler's target machine (for example,
4365 @samp{i686-pc-linux-gnu})---and don't do anything else.
4368 @opindex dumpversion
4369 Print the compiler version (for example, @samp{3.0})---and don't do
4374 Print the compiler's built-in specs---and don't do anything else. (This
4375 is used when GCC itself is being built.) @xref{Spec Files}.
4377 @item -feliminate-unused-debug-types
4378 @opindex feliminate-unused-debug-types
4379 Normally, when producing DWARF2 output, GCC will emit debugging
4380 information for all types declared in a compilation
4381 unit, regardless of whether or not they are actually used
4382 in that compilation unit. Sometimes this is useful, such as
4383 if, in the debugger, you want to cast a value to a type that is
4384 not actually used in your program (but is declared). More often,
4385 however, this results in a significant amount of wasted space.
4386 With this option, GCC will avoid producing debug symbol output
4387 for types that are nowhere used in the source file being compiled.
4390 @node Optimize Options
4391 @section Options That Control Optimization
4392 @cindex optimize options
4393 @cindex options, optimization
4395 These options control various sorts of optimizations.
4397 Without any optimization option, the compiler's goal is to reduce the
4398 cost of compilation and to make debugging produce the expected
4399 results. Statements are independent: if you stop the program with a
4400 breakpoint between statements, you can then assign a new value to any
4401 variable or change the program counter to any other statement in the
4402 function and get exactly the results you would expect from the source
4405 Turning on optimization flags makes the compiler attempt to improve
4406 the performance and/or code size at the expense of compilation time
4407 and possibly the ability to debug the program.
4409 The compiler performs optimization based on the knowledge it has of
4410 the program. Optimization levels @option{-O} and above, in
4411 particular, enable @emph{unit-at-a-time} mode, which allows the
4412 compiler to consider information gained from later functions in
4413 the file when compiling a function. Compiling multiple files at
4414 once to a single output file in @emph{unit-at-a-time} mode allows
4415 the compiler to use information gained from all of the files when
4416 compiling each of them.
4418 Not all optimizations are controlled directly by a flag. Only
4419 optimizations that have a flag are listed.
4426 Optimize. Optimizing compilation takes somewhat more time, and a lot
4427 more memory for a large function.
4429 With @option{-O}, the compiler tries to reduce code size and execution
4430 time, without performing any optimizations that take a great deal of
4433 @option{-O} turns on the following optimization flags:
4434 @gccoptlist{-fdefer-pop @gol
4435 -fdelayed-branch @gol
4436 -fguess-branch-probability @gol
4437 -fcprop-registers @gol
4438 -fif-conversion @gol
4439 -fif-conversion2 @gol
4442 -ftree-dominator-opts @gol
4447 -ftree-copyrename @gol
4450 -funit-at-a-time @gol
4453 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4454 where doing so does not interfere with debugging.
4458 Optimize even more. GCC performs nearly all supported optimizations
4459 that do not involve a space-speed tradeoff. The compiler does not
4460 perform loop unrolling or function inlining when you specify @option{-O2}.
4461 As compared to @option{-O}, this option increases both compilation time
4462 and the performance of the generated code.
4464 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4465 also turns on the following optimization flags:
4466 @gccoptlist{-fthread-jumps @gol
4468 -foptimize-sibling-calls @gol
4469 -fcse-follow-jumps -fcse-skip-blocks @gol
4470 -fgcse -fgcse-lm @gol
4471 -fexpensive-optimizations @gol
4472 -frerun-cse-after-loop @gol
4475 -fschedule-insns -fschedule-insns2 @gol
4476 -fsched-interblock -fsched-spec @gol
4478 -fstrict-aliasing -fstrict-overflow @gol
4479 -fdelete-null-pointer-checks @gol
4480 -freorder-blocks -freorder-functions @gol
4481 -falign-functions -falign-jumps @gol
4482 -falign-loops -falign-labels @gol
4486 Please note the warning under @option{-fgcse} about
4487 invoking @option{-O2} on programs that use computed gotos.
4489 @option{-O2} doesn't turn on @option{-ftree-vrp} for the Ada compiler.
4490 This option must be explicitly specified on the command line to be
4491 enabled for the Ada compiler.
4495 Optimize yet more. @option{-O3} turns on all optimizations specified by
4496 @option{-O2} and also turns on the @option{-finline-functions},
4497 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4501 Do not optimize. This is the default.
4505 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4506 do not typically increase code size. It also performs further
4507 optimizations designed to reduce code size.
4509 @option{-Os} disables the following optimization flags:
4510 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4511 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
4512 -fprefetch-loop-arrays -ftree-vect-loop-version}
4514 If you use multiple @option{-O} options, with or without level numbers,
4515 the last such option is the one that is effective.
4518 Options of the form @option{-f@var{flag}} specify machine-independent
4519 flags. Most flags have both positive and negative forms; the negative
4520 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4521 below, only one of the forms is listed---the one you typically will
4522 use. You can figure out the other form by either removing @samp{no-}
4525 The following options control specific optimizations. They are either
4526 activated by @option{-O} options or are related to ones that are. You
4527 can use the following flags in the rare cases when ``fine-tuning'' of
4528 optimizations to be performed is desired.
4531 @item -fno-default-inline
4532 @opindex fno-default-inline
4533 Do not make member functions inline by default merely because they are
4534 defined inside the class scope (C++ only). Otherwise, when you specify
4535 @w{@option{-O}}, member functions defined inside class scope are compiled
4536 inline by default; i.e., you don't need to add @samp{inline} in front of
4537 the member function name.
4539 @item -fno-defer-pop
4540 @opindex fno-defer-pop
4541 Always pop the arguments to each function call as soon as that function
4542 returns. For machines which must pop arguments after a function call,
4543 the compiler normally lets arguments accumulate on the stack for several
4544 function calls and pops them all at once.
4546 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4550 Force memory operands to be copied into registers before doing
4551 arithmetic on them. This produces better code by making all memory
4552 references potential common subexpressions. When they are not common
4553 subexpressions, instruction combination should eliminate the separate
4554 register-load. This option is now a nop and will be removed in 4.3.
4557 @opindex fforce-addr
4558 Force memory address constants to be copied into registers before
4559 doing arithmetic on them.
4561 @item -fomit-frame-pointer
4562 @opindex fomit-frame-pointer
4563 Don't keep the frame pointer in a register for functions that
4564 don't need one. This avoids the instructions to save, set up and
4565 restore frame pointers; it also makes an extra register available
4566 in many functions. @strong{It also makes debugging impossible on
4569 On some machines, such as the VAX, this flag has no effect, because
4570 the standard calling sequence automatically handles the frame pointer
4571 and nothing is saved by pretending it doesn't exist. The
4572 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4573 whether a target machine supports this flag. @xref{Registers,,Register
4574 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4576 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4578 @item -foptimize-sibling-calls
4579 @opindex foptimize-sibling-calls
4580 Optimize sibling and tail recursive calls.
4582 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4586 Don't pay attention to the @code{inline} keyword. Normally this option
4587 is used to keep the compiler from expanding any functions inline.
4588 Note that if you are not optimizing, no functions can be expanded inline.
4590 @item -finline-functions
4591 @opindex finline-functions
4592 Integrate all simple functions into their callers. The compiler
4593 heuristically decides which functions are simple enough to be worth
4594 integrating in this way.
4596 If all calls to a given function are integrated, and the function is
4597 declared @code{static}, then the function is normally not output as
4598 assembler code in its own right.
4600 Enabled at level @option{-O3}.
4602 @item -finline-functions-called-once
4603 @opindex finline-functions-called-once
4604 Consider all @code{static} functions called once for inlining into their
4605 caller even if they are not marked @code{inline}. If a call to a given
4606 function is integrated, then the function is not output as assembler code
4609 Enabled if @option{-funit-at-a-time} is enabled.
4611 @item -fearly-inlining
4612 @opindex fearly-inlining
4613 Inline functions marked by @code{always_inline} and functions whose body seems
4614 smaller than the function call overhead early before doing
4615 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
4616 makes profiling significantly cheaper and usually inlining faster on programs
4617 having large chains of nested wrapper functions.
4621 @item -finline-limit=@var{n}
4622 @opindex finline-limit
4623 By default, GCC limits the size of functions that can be inlined. This flag
4624 allows the control of this limit for functions that are explicitly marked as
4625 inline (i.e., marked with the inline keyword or defined within the class
4626 definition in c++). @var{n} is the size of functions that can be inlined in
4627 number of pseudo instructions (not counting parameter handling). The default
4628 value of @var{n} is 600.
4629 Increasing this value can result in more inlined code at
4630 the cost of compilation time and memory consumption. Decreasing usually makes
4631 the compilation faster and less code will be inlined (which presumably
4632 means slower programs). This option is particularly useful for programs that
4633 use inlining heavily such as those based on recursive templates with C++.
4635 Inlining is actually controlled by a number of parameters, which may be
4636 specified individually by using @option{--param @var{name}=@var{value}}.
4637 The @option{-finline-limit=@var{n}} option sets some of these parameters
4641 @item max-inline-insns-single
4642 is set to @var{n}/2.
4643 @item max-inline-insns-auto
4644 is set to @var{n}/2.
4645 @item min-inline-insns
4646 is set to 130 or @var{n}/4, whichever is smaller.
4647 @item max-inline-insns-rtl
4651 See below for a documentation of the individual
4652 parameters controlling inlining.
4654 @emph{Note:} pseudo instruction represents, in this particular context, an
4655 abstract measurement of function's size. In no way does it represent a count
4656 of assembly instructions and as such its exact meaning might change from one
4657 release to an another.
4659 @item -fkeep-inline-functions
4660 @opindex fkeep-inline-functions
4661 In C, emit @code{static} functions that are declared @code{inline}
4662 into the object file, even if the function has been inlined into all
4663 of its callers. This switch does not affect functions using the
4664 @code{extern inline} extension in GNU C@. In C++, emit any and all
4665 inline functions into the object file.
4667 @item -fkeep-static-consts
4668 @opindex fkeep-static-consts
4669 Emit variables declared @code{static const} when optimization isn't turned
4670 on, even if the variables aren't referenced.
4672 GCC enables this option by default. If you want to force the compiler to
4673 check if the variable was referenced, regardless of whether or not
4674 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4676 @item -fmerge-constants
4677 Attempt to merge identical constants (string constants and floating point
4678 constants) across compilation units.
4680 This option is the default for optimized compilation if the assembler and
4681 linker support it. Use @option{-fno-merge-constants} to inhibit this
4684 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4686 @item -fmerge-all-constants
4687 Attempt to merge identical constants and identical variables.
4689 This option implies @option{-fmerge-constants}. In addition to
4690 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4691 arrays or initialized constant variables with integral or floating point
4692 types. Languages like C or C++ require each non-automatic variable to
4693 have distinct location, so using this option will result in non-conforming
4696 @item -fmodulo-sched
4697 @opindex fmodulo-sched
4698 Perform swing modulo scheduling immediately before the first scheduling
4699 pass. This pass looks at innermost loops and reorders their
4700 instructions by overlapping different iterations.
4702 @item -fno-branch-count-reg
4703 @opindex fno-branch-count-reg
4704 Do not use ``decrement and branch'' instructions on a count register,
4705 but instead generate a sequence of instructions that decrement a
4706 register, compare it against zero, then branch based upon the result.
4707 This option is only meaningful on architectures that support such
4708 instructions, which include x86, PowerPC, IA-64 and S/390.
4710 The default is @option{-fbranch-count-reg}.
4712 @item -fno-function-cse
4713 @opindex fno-function-cse
4714 Do not put function addresses in registers; make each instruction that
4715 calls a constant function contain the function's address explicitly.
4717 This option results in less efficient code, but some strange hacks
4718 that alter the assembler output may be confused by the optimizations
4719 performed when this option is not used.
4721 The default is @option{-ffunction-cse}
4723 @item -fno-zero-initialized-in-bss
4724 @opindex fno-zero-initialized-in-bss
4725 If the target supports a BSS section, GCC by default puts variables that
4726 are initialized to zero into BSS@. This can save space in the resulting
4729 This option turns off this behavior because some programs explicitly
4730 rely on variables going to the data section. E.g., so that the
4731 resulting executable can find the beginning of that section and/or make
4732 assumptions based on that.
4734 The default is @option{-fzero-initialized-in-bss}.
4736 @item -fbounds-check
4737 @opindex fbounds-check
4738 For front-ends that support it, generate additional code to check that
4739 indices used to access arrays are within the declared range. This is
4740 currently only supported by the Java and Fortran front-ends, where
4741 this option defaults to true and false respectively.
4743 @item -fmudflap -fmudflapth -fmudflapir
4747 @cindex bounds checking
4749 For front-ends that support it (C and C++), instrument all risky
4750 pointer/array dereferencing operations, some standard library
4751 string/heap functions, and some other associated constructs with
4752 range/validity tests. Modules so instrumented should be immune to
4753 buffer overflows, invalid heap use, and some other classes of C/C++
4754 programming errors. The instrumentation relies on a separate runtime
4755 library (@file{libmudflap}), which will be linked into a program if
4756 @option{-fmudflap} is given at link time. Run-time behavior of the
4757 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4758 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4761 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4762 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4763 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4764 instrumentation should ignore pointer reads. This produces less
4765 instrumentation (and therefore faster execution) and still provides
4766 some protection against outright memory corrupting writes, but allows
4767 erroneously read data to propagate within a program.
4769 @item -fthread-jumps
4770 @opindex fthread-jumps
4771 Perform optimizations where we check to see if a jump branches to a
4772 location where another comparison subsumed by the first is found. If
4773 so, the first branch is redirected to either the destination of the
4774 second branch or a point immediately following it, depending on whether
4775 the condition is known to be true or false.
4777 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4779 @item -fcse-follow-jumps
4780 @opindex fcse-follow-jumps
4781 In common subexpression elimination, scan through jump instructions
4782 when the target of the jump is not reached by any other path. For
4783 example, when CSE encounters an @code{if} statement with an
4784 @code{else} clause, CSE will follow the jump when the condition
4787 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4789 @item -fcse-skip-blocks
4790 @opindex fcse-skip-blocks
4791 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4792 follow jumps which conditionally skip over blocks. When CSE
4793 encounters a simple @code{if} statement with no else clause,
4794 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4795 body of the @code{if}.
4797 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4799 @item -frerun-cse-after-loop
4800 @opindex frerun-cse-after-loop
4801 Re-run common subexpression elimination after loop optimizations has been
4804 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4808 Perform a global common subexpression elimination pass.
4809 This pass also performs global constant and copy propagation.
4811 @emph{Note:} When compiling a program using computed gotos, a GCC
4812 extension, you may get better runtime performance if you disable
4813 the global common subexpression elimination pass by adding
4814 @option{-fno-gcse} to the command line.
4816 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4820 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4821 attempt to move loads which are only killed by stores into themselves. This
4822 allows a loop containing a load/store sequence to be changed to a load outside
4823 the loop, and a copy/store within the loop.
4825 Enabled by default when gcse is enabled.
4829 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4830 global common subexpression elimination. This pass will attempt to move
4831 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4832 loops containing a load/store sequence can be changed to a load before
4833 the loop and a store after the loop.
4835 Not enabled at any optimization level.
4839 When @option{-fgcse-las} is enabled, the global common subexpression
4840 elimination pass eliminates redundant loads that come after stores to the
4841 same memory location (both partial and full redundancies).
4843 Not enabled at any optimization level.
4845 @item -fgcse-after-reload
4846 @opindex fgcse-after-reload
4847 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4848 pass is performed after reload. The purpose of this pass is to cleanup
4851 @item -funsafe-loop-optimizations
4852 @opindex funsafe-loop-optimizations
4853 If given, the loop optimizer will assume that loop indices do not
4854 overflow, and that the loops with nontrivial exit condition are not
4855 infinite. This enables a wider range of loop optimizations even if
4856 the loop optimizer itself cannot prove that these assumptions are valid.
4857 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
4858 if it finds this kind of loop.
4860 @item -fcrossjumping
4861 @opindex crossjumping
4862 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4863 resulting code may or may not perform better than without cross-jumping.
4865 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4867 @item -fif-conversion
4868 @opindex if-conversion
4869 Attempt to transform conditional jumps into branch-less equivalents. This
4870 include use of conditional moves, min, max, set flags and abs instructions, and
4871 some tricks doable by standard arithmetics. The use of conditional execution
4872 on chips where it is available is controlled by @code{if-conversion2}.
4874 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4876 @item -fif-conversion2
4877 @opindex if-conversion2
4878 Use conditional execution (where available) to transform conditional jumps into
4879 branch-less equivalents.
4881 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4883 @item -fdelete-null-pointer-checks
4884 @opindex fdelete-null-pointer-checks
4885 Use global dataflow analysis to identify and eliminate useless checks
4886 for null pointers. The compiler assumes that dereferencing a null
4887 pointer would have halted the program. If a pointer is checked after
4888 it has already been dereferenced, it cannot be null.
4890 In some environments, this assumption is not true, and programs can
4891 safely dereference null pointers. Use
4892 @option{-fno-delete-null-pointer-checks} to disable this optimization
4893 for programs which depend on that behavior.
4895 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4897 @item -fexpensive-optimizations
4898 @opindex fexpensive-optimizations
4899 Perform a number of minor optimizations that are relatively expensive.
4901 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4903 @item -foptimize-register-move
4905 @opindex foptimize-register-move
4907 Attempt to reassign register numbers in move instructions and as
4908 operands of other simple instructions in order to maximize the amount of
4909 register tying. This is especially helpful on machines with two-operand
4912 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4915 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4917 @item -fdelayed-branch
4918 @opindex fdelayed-branch
4919 If supported for the target machine, attempt to reorder instructions
4920 to exploit instruction slots available after delayed branch
4923 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4925 @item -fschedule-insns
4926 @opindex fschedule-insns
4927 If supported for the target machine, attempt to reorder instructions to
4928 eliminate execution stalls due to required data being unavailable. This
4929 helps machines that have slow floating point or memory load instructions
4930 by allowing other instructions to be issued until the result of the load
4931 or floating point instruction is required.
4933 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4935 @item -fschedule-insns2
4936 @opindex fschedule-insns2
4937 Similar to @option{-fschedule-insns}, but requests an additional pass of
4938 instruction scheduling after register allocation has been done. This is
4939 especially useful on machines with a relatively small number of
4940 registers and where memory load instructions take more than one cycle.
4942 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4944 @item -fno-sched-interblock
4945 @opindex fno-sched-interblock
4946 Don't schedule instructions across basic blocks. This is normally
4947 enabled by default when scheduling before register allocation, i.e.@:
4948 with @option{-fschedule-insns} or at @option{-O2} or higher.
4950 @item -fno-sched-spec
4951 @opindex fno-sched-spec
4952 Don't allow speculative motion of non-load instructions. This is normally
4953 enabled by default when scheduling before register allocation, i.e.@:
4954 with @option{-fschedule-insns} or at @option{-O2} or higher.
4956 @item -fsched-spec-load
4957 @opindex fsched-spec-load
4958 Allow speculative motion of some load instructions. This only makes
4959 sense when scheduling before register allocation, i.e.@: with
4960 @option{-fschedule-insns} or at @option{-O2} or higher.
4962 @item -fsched-spec-load-dangerous
4963 @opindex fsched-spec-load-dangerous
4964 Allow speculative motion of more load instructions. This only makes
4965 sense when scheduling before register allocation, i.e.@: with
4966 @option{-fschedule-insns} or at @option{-O2} or higher.
4968 @item -fsched-stalled-insns=@var{n}
4969 @opindex fsched-stalled-insns
4970 Define how many insns (if any) can be moved prematurely from the queue
4971 of stalled insns into the ready list, during the second scheduling pass.
4973 @item -fsched-stalled-insns-dep=@var{n}
4974 @opindex fsched-stalled-insns-dep
4975 Define how many insn groups (cycles) will be examined for a dependency
4976 on a stalled insn that is candidate for premature removal from the queue
4977 of stalled insns. Has an effect only during the second scheduling pass,
4978 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4980 @item -fsched2-use-superblocks
4981 @opindex fsched2-use-superblocks
4982 When scheduling after register allocation, do use superblock scheduling
4983 algorithm. Superblock scheduling allows motion across basic block boundaries
4984 resulting on faster schedules. This option is experimental, as not all machine
4985 descriptions used by GCC model the CPU closely enough to avoid unreliable
4986 results from the algorithm.
4988 This only makes sense when scheduling after register allocation, i.e.@: with
4989 @option{-fschedule-insns2} or at @option{-O2} or higher.
4991 @item -fsched2-use-traces
4992 @opindex fsched2-use-traces
4993 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4994 allocation and additionally perform code duplication in order to increase the
4995 size of superblocks using tracer pass. See @option{-ftracer} for details on
4998 This mode should produce faster but significantly longer programs. Also
4999 without @option{-fbranch-probabilities} the traces constructed may not
5000 match the reality and hurt the performance. This only makes
5001 sense when scheduling after register allocation, i.e.@: with
5002 @option{-fschedule-insns2} or at @option{-O2} or higher.
5006 Eliminates redundant extension instructions and move the non redundant
5007 ones to optimal placement using LCM.
5009 @item -freschedule-modulo-scheduled-loops
5010 @opindex fscheduling-in-modulo-scheduled-loops
5011 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
5012 we may want to prevent the later scheduling passes from changing its schedule, we use this
5013 option to control that.
5015 @item -fcaller-saves
5016 @opindex fcaller-saves
5017 Enable values to be allocated in registers that will be clobbered by
5018 function calls, by emitting extra instructions to save and restore the
5019 registers around such calls. Such allocation is done only when it
5020 seems to result in better code than would otherwise be produced.
5022 This option is always enabled by default on certain machines, usually
5023 those which have no call-preserved registers to use instead.
5025 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5028 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
5029 enabled by default at @option{-O2} and @option{-O3}.
5032 Perform Full Redundancy Elimination (FRE) on trees. The difference
5033 between FRE and PRE is that FRE only considers expressions
5034 that are computed on all paths leading to the redundant computation.
5035 This analysis faster than PRE, though it exposes fewer redundancies.
5036 This flag is enabled by default at @option{-O} and higher.
5038 @item -ftree-copy-prop
5039 Perform copy propagation on trees. This pass eliminates unnecessary
5040 copy operations. This flag is enabled by default at @option{-O} and
5043 @item -ftree-store-copy-prop
5044 Perform copy propagation of memory loads and stores. This pass
5045 eliminates unnecessary copy operations in memory references
5046 (structures, global variables, arrays, etc). This flag is enabled by
5047 default at @option{-O2} and higher.
5050 Perform structural alias analysis on trees. This flag
5051 is enabled by default at @option{-O} and higher.
5054 Perform interprocedural pointer analysis.
5057 Perform forward store motion on trees. This flag is
5058 enabled by default at @option{-O} and higher.
5061 Perform sparse conditional constant propagation (CCP) on trees. This
5062 pass only operates on local scalar variables and is enabled by default
5063 at @option{-O} and higher.
5065 @item -ftree-store-ccp
5066 Perform sparse conditional constant propagation (CCP) on trees. This
5067 pass operates on both local scalar variables and memory stores and
5068 loads (global variables, structures, arrays, etc). This flag is
5069 enabled by default at @option{-O2} and higher.
5072 Perform dead code elimination (DCE) on trees. This flag is enabled by
5073 default at @option{-O} and higher.
5075 @item -ftree-dominator-opts
5076 Perform a variety of simple scalar cleanups (constant/copy
5077 propagation, redundancy elimination, range propagation and expression
5078 simplification) based on a dominator tree traversal. This also
5079 performs jump threading (to reduce jumps to jumps). This flag is
5080 enabled by default at @option{-O} and higher.
5083 Perform loop header copying on trees. This is beneficial since it increases
5084 effectiveness of code motion optimizations. It also saves one jump. This flag
5085 is enabled by default at @option{-O} and higher. It is not enabled
5086 for @option{-Os}, since it usually increases code size.
5088 @item -ftree-loop-optimize
5089 Perform loop optimizations on trees. This flag is enabled by default
5090 at @option{-O} and higher.
5092 @item -ftree-loop-linear
5093 Perform linear loop transformations on tree. This flag can improve cache
5094 performance and allow further loop optimizations to take place.
5096 @item -ftree-loop-im
5097 Perform loop invariant motion on trees. This pass moves only invariants that
5098 would be hard to handle at RTL level (function calls, operations that expand to
5099 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5100 operands of conditions that are invariant out of the loop, so that we can use
5101 just trivial invariantness analysis in loop unswitching. The pass also includes
5104 @item -ftree-loop-ivcanon
5105 Create a canonical counter for number of iterations in the loop for that
5106 determining number of iterations requires complicated analysis. Later
5107 optimizations then may determine the number easily. Useful especially
5108 in connection with unrolling.
5111 Perform induction variable optimizations (strength reduction, induction
5112 variable merging and induction variable elimination) on trees.
5115 Perform scalar replacement of aggregates. This pass replaces structure
5116 references with scalars to prevent committing structures to memory too
5117 early. This flag is enabled by default at @option{-O} and higher.
5119 @item -ftree-copyrename
5120 Perform copy renaming on trees. This pass attempts to rename compiler
5121 temporaries to other variables at copy locations, usually resulting in
5122 variable names which more closely resemble the original variables. This flag
5123 is enabled by default at @option{-O} and higher.
5126 Perform temporary expression replacement during the SSA->normal phase. Single
5127 use/single def temporaries are replaced at their use location with their
5128 defining expression. This results in non-GIMPLE code, but gives the expanders
5129 much more complex trees to work on resulting in better RTL generation. This is
5130 enabled by default at @option{-O} and higher.
5133 Perform live range splitting during the SSA->normal phase. Distinct live
5134 ranges of a variable are split into unique variables, allowing for better
5135 optimization later. This is enabled by default at @option{-O} and higher.
5137 @item -ftree-vectorize
5138 Perform loop vectorization on trees.
5140 @item -ftree-vect-loop-version
5141 @opindex ftree-vect-loop-version
5142 Perform loop versioning when doing loop vectorization on trees. When a loop
5143 appears to be vectorizable except that data alignment or data dependence cannot
5144 be determined at compile time then vectorized and non-vectorized versions of
5145 the loop are generated along with runtime checks for alignment or dependence
5146 to control which version is executed. This option is enabled by default
5147 except at level @option{-Os} where it is disabled.
5150 Perform Value Range Propagation on trees. This is similar to the
5151 constant propagation pass, but instead of values, ranges of values are
5152 propagated. This allows the optimizers to remove unnecessary range
5153 checks like array bound checks and null pointer checks. This is
5154 enabled by default at @option{-O2} and higher. Null pointer check
5155 elimination is only done if @option{-fdelete-null-pointer-checks} is
5160 Perform tail duplication to enlarge superblock size. This transformation
5161 simplifies the control flow of the function allowing other optimizations to do
5164 @item -funroll-loops
5165 @opindex funroll-loops
5166 Unroll loops whose number of iterations can be determined at compile
5167 time or upon entry to the loop. @option{-funroll-loops} implies
5168 @option{-frerun-cse-after-loop}. This option makes code larger,
5169 and may or may not make it run faster.
5171 @item -funroll-all-loops
5172 @opindex funroll-all-loops
5173 Unroll all loops, even if their number of iterations is uncertain when
5174 the loop is entered. This usually makes programs run more slowly.
5175 @option{-funroll-all-loops} implies the same options as
5176 @option{-funroll-loops},
5178 @item -fsplit-ivs-in-unroller
5179 @opindex fsplit-ivs-in-unroller
5180 Enables expressing of values of induction variables in later iterations
5181 of the unrolled loop using the value in the first iteration. This breaks
5182 long dependency chains, thus improving efficiency of the scheduling passes.
5184 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5185 same effect. However in cases the loop body is more complicated than
5186 a single basic block, this is not reliable. It also does not work at all
5187 on some of the architectures due to restrictions in the CSE pass.
5189 This optimization is enabled by default.
5191 @item -fvariable-expansion-in-unroller
5192 @opindex fvariable-expansion-in-unroller
5193 With this option, the compiler will create multiple copies of some
5194 local variables when unrolling a loop which can result in superior code.
5196 @item -fprefetch-loop-arrays
5197 @opindex fprefetch-loop-arrays
5198 If supported by the target machine, generate instructions to prefetch
5199 memory to improve the performance of loops that access large arrays.
5201 This option may generate better or worse code; results are highly
5202 dependent on the structure of loops within the source code.
5204 Disabled at level @option{-Os}.
5207 @itemx -fno-peephole2
5208 @opindex fno-peephole
5209 @opindex fno-peephole2
5210 Disable any machine-specific peephole optimizations. The difference
5211 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5212 are implemented in the compiler; some targets use one, some use the
5213 other, a few use both.
5215 @option{-fpeephole} is enabled by default.
5216 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5218 @item -fno-guess-branch-probability
5219 @opindex fno-guess-branch-probability
5220 Do not guess branch probabilities using heuristics.
5222 GCC will use heuristics to guess branch probabilities if they are
5223 not provided by profiling feedback (@option{-fprofile-arcs}). These
5224 heuristics are based on the control flow graph. If some branch probabilities
5225 are specified by @samp{__builtin_expect}, then the heuristics will be
5226 used to guess branch probabilities for the rest of the control flow graph,
5227 taking the @samp{__builtin_expect} info into account. The interactions
5228 between the heuristics and @samp{__builtin_expect} can be complex, and in
5229 some cases, it may be useful to disable the heuristics so that the effects
5230 of @samp{__builtin_expect} are easier to understand.
5232 The default is @option{-fguess-branch-probability} at levels
5233 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5235 @item -freorder-blocks
5236 @opindex freorder-blocks
5237 Reorder basic blocks in the compiled function in order to reduce number of
5238 taken branches and improve code locality.
5240 Enabled at levels @option{-O2}, @option{-O3}.
5242 @item -freorder-blocks-and-partition
5243 @opindex freorder-blocks-and-partition
5244 In addition to reordering basic blocks in the compiled function, in order
5245 to reduce number of taken branches, partitions hot and cold basic blocks
5246 into separate sections of the assembly and .o files, to improve
5247 paging and cache locality performance.
5249 This optimization is automatically turned off in the presence of
5250 exception handling, for linkonce sections, for functions with a user-defined
5251 section attribute and on any architecture that does not support named
5254 @item -freorder-functions
5255 @opindex freorder-functions
5256 Reorder functions in the object file in order to
5257 improve code locality. This is implemented by using special
5258 subsections @code{.text.hot} for most frequently executed functions and
5259 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5260 the linker so object file format must support named sections and linker must
5261 place them in a reasonable way.
5263 Also profile feedback must be available in to make this option effective. See
5264 @option{-fprofile-arcs} for details.
5266 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5268 @item -fstrict-aliasing
5269 @opindex fstrict-aliasing
5270 Allows the compiler to assume the strictest aliasing rules applicable to
5271 the language being compiled. For C (and C++), this activates
5272 optimizations based on the type of expressions. In particular, an
5273 object of one type is assumed never to reside at the same address as an
5274 object of a different type, unless the types are almost the same. For
5275 example, an @code{unsigned int} can alias an @code{int}, but not a
5276 @code{void*} or a @code{double}. A character type may alias any other
5279 Pay special attention to code like this:
5292 The practice of reading from a different union member than the one most
5293 recently written to (called ``type-punning'') is common. Even with
5294 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5295 is accessed through the union type. So, the code above will work as
5296 expected. However, this code might not:
5307 Every language that wishes to perform language-specific alias analysis
5308 should define a function that computes, given an @code{tree}
5309 node, an alias set for the node. Nodes in different alias sets are not
5310 allowed to alias. For an example, see the C front-end function
5311 @code{c_get_alias_set}.
5313 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5315 @item -fstrict-overflow
5316 @opindex fstrict-overflow
5317 Allow the compiler to assume strict signed overflow rules, depending
5318 on the language being compiled. For C (and C++) this means that
5319 overflow when doing arithmetic with signed numbers is undefined, which
5320 means that the compiler may assume that it will not happen. This
5321 permits various optimizations. For example, the compiler will assume
5322 that an expression like @code{i + 10 > i} will always be true for
5323 signed @code{i}. This assumption is only valid if signed overflow is
5324 undefined, as the expression is false if @code{i + 10} overflows when
5325 using twos complement arithmetic. When this option is in effect any
5326 attempt to determine whether an operation on signed numbers will
5327 overflow must be written carefully to not actually involve overflow.
5329 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
5330 that signed overflow is fully defined: it wraps. When
5331 @option{-fwrapv} is used, there is no difference between
5332 @option{-fstrict-overflow} and @option{-fno-strict-overflow}. With
5333 @option{-fwrapv} certain types of overflow are permitted. For
5334 example, if the compiler gets an overflow when doing arithmetic on
5335 constants, the overflowed value can still be used with
5336 @option{-fwrapv}, but not otherwise.
5338 The @option{-fstrict-overflow} option is enabled at levels
5339 @option{-O2}, @option{-O3}, @option{-Os}.
5341 @item -falign-functions
5342 @itemx -falign-functions=@var{n}
5343 @opindex falign-functions
5344 Align the start of functions to the next power-of-two greater than
5345 @var{n}, skipping up to @var{n} bytes. For instance,
5346 @option{-falign-functions=32} aligns functions to the next 32-byte
5347 boundary, but @option{-falign-functions=24} would align to the next
5348 32-byte boundary only if this can be done by skipping 23 bytes or less.
5350 @option{-fno-align-functions} and @option{-falign-functions=1} are
5351 equivalent and mean that functions will not be aligned.
5353 Some assemblers only support this flag when @var{n} is a power of two;
5354 in that case, it is rounded up.
5356 If @var{n} is not specified or is zero, use a machine-dependent default.
5358 Enabled at levels @option{-O2}, @option{-O3}.
5360 @item -falign-labels
5361 @itemx -falign-labels=@var{n}
5362 @opindex falign-labels
5363 Align all branch targets to a power-of-two boundary, skipping up to
5364 @var{n} bytes like @option{-falign-functions}. This option can easily
5365 make code slower, because it must insert dummy operations for when the
5366 branch target is reached in the usual flow of the code.
5368 @option{-fno-align-labels} and @option{-falign-labels=1} are
5369 equivalent and mean that labels will not be aligned.
5371 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5372 are greater than this value, then their values are used instead.
5374 If @var{n} is not specified or is zero, use a machine-dependent default
5375 which is very likely to be @samp{1}, meaning no alignment.
5377 Enabled at levels @option{-O2}, @option{-O3}.
5380 @itemx -falign-loops=@var{n}
5381 @opindex falign-loops
5382 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5383 like @option{-falign-functions}. The hope is that the loop will be
5384 executed many times, which will make up for any execution of the dummy
5387 @option{-fno-align-loops} and @option{-falign-loops=1} are
5388 equivalent and mean that loops will not be aligned.
5390 If @var{n} is not specified or is zero, use a machine-dependent default.
5392 Enabled at levels @option{-O2}, @option{-O3}.
5395 @itemx -falign-jumps=@var{n}
5396 @opindex falign-jumps
5397 Align branch targets to a power-of-two boundary, for branch targets
5398 where the targets can only be reached by jumping, skipping up to @var{n}
5399 bytes like @option{-falign-functions}. In this case, no dummy operations
5402 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5403 equivalent and mean that loops will not be aligned.
5405 If @var{n} is not specified or is zero, use a machine-dependent default.
5407 Enabled at levels @option{-O2}, @option{-O3}.
5409 @item -funit-at-a-time
5410 @opindex funit-at-a-time
5411 Parse the whole compilation unit before starting to produce code.
5412 This allows some extra optimizations to take place but consumes
5413 more memory (in general). There are some compatibility issues
5414 with @emph{unit-at-a-time} mode:
5417 enabling @emph{unit-at-a-time} mode may change the order
5418 in which functions, variables, and top-level @code{asm} statements
5419 are emitted, and will likely break code relying on some particular
5420 ordering. The majority of such top-level @code{asm} statements,
5421 though, can be replaced by @code{section} attributes. The
5422 @option{fno-toplevel-reorder} option may be used to keep the ordering
5423 used in the input file, at the cost of some optimizations.
5426 @emph{unit-at-a-time} mode removes unreferenced static variables
5427 and functions. This may result in undefined references
5428 when an @code{asm} statement refers directly to variables or functions
5429 that are otherwise unused. In that case either the variable/function
5430 shall be listed as an operand of the @code{asm} statement operand or,
5431 in the case of top-level @code{asm} statements the attribute @code{used}
5432 shall be used on the declaration.
5435 Static functions now can use non-standard passing conventions that
5436 may break @code{asm} statements calling functions directly. Again,
5437 attribute @code{used} will prevent this behavior.
5440 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5441 but this scheme may not be supported by future releases of GCC@.
5443 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5445 @item -fno-toplevel-reorder
5446 Do not reorder top-level functions, variables, and @code{asm}
5447 statements. Output them in the same order that they appear in the
5448 input file. When this option is used, unreferenced static variables
5449 will not be removed. This option is intended to support existing code
5450 which relies on a particular ordering. For new code, it is better to
5455 Constructs webs as commonly used for register allocation purposes and assign
5456 each web individual pseudo register. This allows the register allocation pass
5457 to operate on pseudos directly, but also strengthens several other optimization
5458 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5459 however, make debugging impossible, since variables will no longer stay in a
5462 Enabled by default with @option{-funroll-loops}.
5464 @item -fwhole-program
5465 @opindex fwhole-program
5466 Assume that the current compilation unit represents whole program being
5467 compiled. All public functions and variables with the exception of @code{main}
5468 and those merged by attribute @code{externally_visible} become static functions
5469 and in a affect gets more aggressively optimized by interprocedural optimizers.
5470 While this option is equivalent to proper use of @code{static} keyword for
5471 programs consisting of single file, in combination with option
5472 @option{--combine} this flag can be used to compile most of smaller scale C
5473 programs since the functions and variables become local for the whole combined
5474 compilation unit, not for the single source file itself.
5477 @item -fno-cprop-registers
5478 @opindex fno-cprop-registers
5479 After register allocation and post-register allocation instruction splitting,
5480 we perform a copy-propagation pass to try to reduce scheduling dependencies
5481 and occasionally eliminate the copy.
5483 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5485 @item -fprofile-generate
5486 @opindex fprofile-generate
5488 Enable options usually used for instrumenting application to produce
5489 profile useful for later recompilation with profile feedback based
5490 optimization. You must use @option{-fprofile-generate} both when
5491 compiling and when linking your program.
5493 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5496 @opindex fprofile-use
5497 Enable profile feedback directed optimizations, and optimizations
5498 generally profitable only with profile feedback available.
5500 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
5501 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
5505 The following options control compiler behavior regarding floating
5506 point arithmetic. These options trade off between speed and
5507 correctness. All must be specifically enabled.
5511 @opindex ffloat-store
5512 Do not store floating point variables in registers, and inhibit other
5513 options that might change whether a floating point value is taken from a
5516 @cindex floating point precision
5517 This option prevents undesirable excess precision on machines such as
5518 the 68000 where the floating registers (of the 68881) keep more
5519 precision than a @code{double} is supposed to have. Similarly for the
5520 x86 architecture. For most programs, the excess precision does only
5521 good, but a few programs rely on the precise definition of IEEE floating
5522 point. Use @option{-ffloat-store} for such programs, after modifying
5523 them to store all pertinent intermediate computations into variables.
5527 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5528 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5529 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5530 and @option{fcx-limited-range}.
5532 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5534 This option should never be turned on by any @option{-O} option since
5535 it can result in incorrect output for programs which depend on
5536 an exact implementation of IEEE or ISO rules/specifications for
5539 @item -fno-math-errno
5540 @opindex fno-math-errno
5541 Do not set ERRNO after calling math functions that are executed
5542 with a single instruction, e.g., sqrt. A program that relies on
5543 IEEE exceptions for math error handling may want to use this flag
5544 for speed while maintaining IEEE arithmetic compatibility.
5546 This option should never be turned on by any @option{-O} option since
5547 it can result in incorrect output for programs which depend on
5548 an exact implementation of IEEE or ISO rules/specifications for
5551 The default is @option{-fmath-errno}.
5553 On Darwin and FreeBSD systems, the math library never sets @code{errno}.
5555 no reason for the compiler to consider the possibility that it might,
5556 and @option{-fno-math-errno} is the default.
5558 @item -funsafe-math-optimizations
5559 @opindex funsafe-math-optimizations
5560 Allow optimizations for floating-point arithmetic that (a) assume
5561 that arguments and results are valid and (b) may violate IEEE or
5562 ANSI standards. When used at link-time, it may include libraries
5563 or startup files that change the default FPU control word or other
5564 similar optimizations.
5566 This option should never be turned on by any @option{-O} option since
5567 it can result in incorrect output for programs which depend on
5568 an exact implementation of IEEE or ISO rules/specifications for
5571 The default is @option{-fno-unsafe-math-optimizations}.
5573 @item -ffinite-math-only
5574 @opindex ffinite-math-only
5575 Allow optimizations for floating-point arithmetic that assume
5576 that arguments and results are not NaNs or +-Infs.
5578 This option should never be turned on by any @option{-O} option since
5579 it can result in incorrect output for programs which depend on
5580 an exact implementation of IEEE or ISO rules/specifications.
5582 The default is @option{-fno-finite-math-only}.
5584 @item -fno-trapping-math
5585 @opindex fno-trapping-math
5586 Compile code assuming that floating-point operations cannot generate
5587 user-visible traps. These traps include division by zero, overflow,
5588 underflow, inexact result and invalid operation. This option implies
5589 @option{-fno-signaling-nans}. Setting this option may allow faster
5590 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5592 This option should never be turned on by any @option{-O} option since
5593 it can result in incorrect output for programs which depend on
5594 an exact implementation of IEEE or ISO rules/specifications for
5597 The default is @option{-ftrapping-math}.
5599 @item -frounding-math
5600 @opindex frounding-math
5601 Disable transformations and optimizations that assume default floating
5602 point rounding behavior. This is round-to-zero for all floating point
5603 to integer conversions, and round-to-nearest for all other arithmetic
5604 truncations. This option should be specified for programs that change
5605 the FP rounding mode dynamically, or that may be executed with a
5606 non-default rounding mode. This option disables constant folding of
5607 floating point expressions at compile-time (which may be affected by
5608 rounding mode) and arithmetic transformations that are unsafe in the
5609 presence of sign-dependent rounding modes.
5611 The default is @option{-fno-rounding-math}.
5613 This option is experimental and does not currently guarantee to
5614 disable all GCC optimizations that are affected by rounding mode.
5615 Future versions of GCC may provide finer control of this setting
5616 using C99's @code{FENV_ACCESS} pragma. This command line option
5617 will be used to specify the default state for @code{FENV_ACCESS}.
5619 @item -frtl-abstract-sequences
5620 @opindex frtl-abstract-sequences
5621 It is a size optimization method. This option is to find identical
5622 sequences of code, which can be turned into pseudo-procedures and
5623 then replace all occurrences with calls to the newly created
5624 subroutine. It is kind of an opposite of @option{-finline-functions}.
5625 This optimization runs at RTL level.
5627 @item -fsignaling-nans
5628 @opindex fsignaling-nans
5629 Compile code assuming that IEEE signaling NaNs may generate user-visible
5630 traps during floating-point operations. Setting this option disables
5631 optimizations that may change the number of exceptions visible with
5632 signaling NaNs. This option implies @option{-ftrapping-math}.
5634 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5637 The default is @option{-fno-signaling-nans}.
5639 This option is experimental and does not currently guarantee to
5640 disable all GCC optimizations that affect signaling NaN behavior.
5642 @item -fsingle-precision-constant
5643 @opindex fsingle-precision-constant
5644 Treat floating point constant as single precision constant instead of
5645 implicitly converting it to double precision constant.
5647 @item -fcx-limited-range
5648 @itemx -fno-cx-limited-range
5649 @opindex fcx-limited-range
5650 @opindex fno-cx-limited-range
5651 When enabled, this option states that a range reduction step is not
5652 needed when performing complex division. The default is
5653 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5655 This option controls the default setting of the ISO C99
5656 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5661 The following options control optimizations that may improve
5662 performance, but are not enabled by any @option{-O} options. This
5663 section includes experimental options that may produce broken code.
5666 @item -fbranch-probabilities
5667 @opindex fbranch-probabilities
5668 After running a program compiled with @option{-fprofile-arcs}
5669 (@pxref{Debugging Options,, Options for Debugging Your Program or
5670 @command{gcc}}), you can compile it a second time using
5671 @option{-fbranch-probabilities}, to improve optimizations based on
5672 the number of times each branch was taken. When the program
5673 compiled with @option{-fprofile-arcs} exits it saves arc execution
5674 counts to a file called @file{@var{sourcename}.gcda} for each source
5675 file The information in this data file is very dependent on the
5676 structure of the generated code, so you must use the same source code
5677 and the same optimization options for both compilations.
5679 With @option{-fbranch-probabilities}, GCC puts a
5680 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5681 These can be used to improve optimization. Currently, they are only
5682 used in one place: in @file{reorg.c}, instead of guessing which path a
5683 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5684 exactly determine which path is taken more often.
5686 @item -fprofile-values
5687 @opindex fprofile-values
5688 If combined with @option{-fprofile-arcs}, it adds code so that some
5689 data about values of expressions in the program is gathered.
5691 With @option{-fbranch-probabilities}, it reads back the data gathered
5692 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5693 notes to instructions for their later usage in optimizations.
5695 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5699 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5700 a code to gather information about values of expressions.
5702 With @option{-fbranch-probabilities}, it reads back the data gathered
5703 and actually performs the optimizations based on them.
5704 Currently the optimizations include specialization of division operation
5705 using the knowledge about the value of the denominator.
5707 @item -frename-registers
5708 @opindex frename-registers
5709 Attempt to avoid false dependencies in scheduled code by making use
5710 of registers left over after register allocation. This optimization
5711 will most benefit processors with lots of registers. Depending on the
5712 debug information format adopted by the target, however, it can
5713 make debugging impossible, since variables will no longer stay in
5714 a ``home register''.
5716 Enabled by default with @option{-funroll-loops}.
5720 Perform tail duplication to enlarge superblock size. This transformation
5721 simplifies the control flow of the function allowing other optimizations to do
5724 Enabled with @option{-fprofile-use}.
5726 @item -funroll-loops
5727 @opindex funroll-loops
5728 Unroll loops whose number of iterations can be determined at compile time or
5729 upon entry to the loop. @option{-funroll-loops} implies
5730 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
5731 It also turns on complete loop peeling (i.e.@: complete removal of loops with
5732 small constant number of iterations). This option makes code larger, and may
5733 or may not make it run faster.
5735 Enabled with @option{-fprofile-use}.
5737 @item -funroll-all-loops
5738 @opindex funroll-all-loops
5739 Unroll all loops, even if their number of iterations is uncertain when
5740 the loop is entered. This usually makes programs run more slowly.
5741 @option{-funroll-all-loops} implies the same options as
5742 @option{-funroll-loops}.
5745 @opindex fpeel-loops
5746 Peels the loops for that there is enough information that they do not
5747 roll much (from profile feedback). It also turns on complete loop peeling
5748 (i.e.@: complete removal of loops with small constant number of iterations).
5750 Enabled with @option{-fprofile-use}.
5752 @item -fmove-loop-invariants
5753 @opindex fmove-loop-invariants
5754 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
5755 at level @option{-O1}
5757 @item -funswitch-loops
5758 @opindex funswitch-loops
5759 Move branches with loop invariant conditions out of the loop, with duplicates
5760 of the loop on both branches (modified according to result of the condition).
5762 @item -ffunction-sections
5763 @itemx -fdata-sections
5764 @opindex ffunction-sections
5765 @opindex fdata-sections
5766 Place each function or data item into its own section in the output
5767 file if the target supports arbitrary sections. The name of the
5768 function or the name of the data item determines the section's name
5771 Use these options on systems where the linker can perform optimizations
5772 to improve locality of reference in the instruction space. Most systems
5773 using the ELF object format and SPARC processors running Solaris 2 have
5774 linkers with such optimizations. AIX may have these optimizations in
5777 Only use these options when there are significant benefits from doing
5778 so. When you specify these options, the assembler and linker will
5779 create larger object and executable files and will also be slower.
5780 You will not be able to use @code{gprof} on all systems if you
5781 specify this option and you may have problems with debugging if
5782 you specify both this option and @option{-g}.
5784 @item -fbranch-target-load-optimize
5785 @opindex fbranch-target-load-optimize
5786 Perform branch target register load optimization before prologue / epilogue
5788 The use of target registers can typically be exposed only during reload,
5789 thus hoisting loads out of loops and doing inter-block scheduling needs
5790 a separate optimization pass.
5792 @item -fbranch-target-load-optimize2
5793 @opindex fbranch-target-load-optimize2
5794 Perform branch target register load optimization after prologue / epilogue
5797 @item -fbtr-bb-exclusive
5798 @opindex fbtr-bb-exclusive
5799 When performing branch target register load optimization, don't reuse
5800 branch target registers in within any basic block.
5802 @item -fstack-protector
5803 Emit extra code to check for buffer overflows, such as stack smashing
5804 attacks. This is done by adding a guard variable to functions with
5805 vulnerable objects. This includes functions that call alloca, and
5806 functions with buffers larger than 8 bytes. The guards are initialized
5807 when a function is entered and then checked when the function exits.
5808 If a guard check fails, an error message is printed and the program exits.
5810 @item -fstack-protector-all
5811 Like @option{-fstack-protector} except that all functions are protected.
5813 @item -fstack-protector-strong
5814 Like @option{-fstack-protector} but includes additional functions to
5815 be protected --- those that have local array definitions, or have
5816 references to local frame addresses.
5818 @item -fsection-anchors
5819 @opindex fsection-anchors
5820 Try to reduce the number of symbolic address calculations by using
5821 shared ``anchor'' symbols to address nearby objects. This transformation
5822 can help to reduce the number of GOT entries and GOT accesses on some
5825 For example, the implementation of the following function @code{foo}:
5829 int foo (void) @{ return a + b + c; @}
5832 would usually calculate the addresses of all three variables, but if you
5833 compile it with @option{-fsection-anchors}, it will access the variables
5834 from a common anchor point instead. The effect is similar to the
5835 following pseudocode (which isn't valid C):
5840 register int *xr = &x;
5841 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5845 Not all targets support this option.
5847 @item --param @var{name}=@var{value}
5849 In some places, GCC uses various constants to control the amount of
5850 optimization that is done. For example, GCC will not inline functions
5851 that contain more that a certain number of instructions. You can
5852 control some of these constants on the command-line using the
5853 @option{--param} option.
5855 The names of specific parameters, and the meaning of the values, are
5856 tied to the internals of the compiler, and are subject to change
5857 without notice in future releases.
5859 In each case, the @var{value} is an integer. The allowable choices for
5860 @var{name} are given in the following table:
5863 @item salias-max-implicit-fields
5864 The maximum number of fields in a variable without direct
5865 structure accesses for which structure aliasing will consider trying
5866 to track each field. The default is 5
5868 @item salias-max-array-elements
5869 The maximum number of elements an array can have and its elements
5870 still be tracked individually by structure aliasing. The default is 4
5872 @item sra-max-structure-size
5873 The maximum structure size, in bytes, at which the scalar replacement
5874 of aggregates (SRA) optimization will perform block copies. The
5875 default value, 0, implies that GCC will select the most appropriate
5878 @item sra-field-structure-ratio
5879 The threshold ratio (as a percentage) between instantiated fields and
5880 the complete structure size. We say that if the ratio of the number
5881 of bytes in instantiated fields to the number of bytes in the complete
5882 structure exceeds this parameter, then block copies are not used. The
5885 @item max-crossjump-edges
5886 The maximum number of incoming edges to consider for crossjumping.
5887 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5888 the number of edges incoming to each block. Increasing values mean
5889 more aggressive optimization, making the compile time increase with
5890 probably small improvement in executable size.
5892 @item min-crossjump-insns
5893 The minimum number of instructions which must be matched at the end
5894 of two blocks before crossjumping will be performed on them. This
5895 value is ignored in the case where all instructions in the block being
5896 crossjumped from are matched. The default value is 5.
5898 @item max-grow-copy-bb-insns
5899 The maximum code size expansion factor when copying basic blocks
5900 instead of jumping. The expansion is relative to a jump instruction.
5901 The default value is 8.
5903 @item max-goto-duplication-insns
5904 The maximum number of instructions to duplicate to a block that jumps
5905 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5906 passes, GCC factors computed gotos early in the compilation process,
5907 and unfactors them as late as possible. Only computed jumps at the
5908 end of a basic blocks with no more than max-goto-duplication-insns are
5909 unfactored. The default value is 8.
5911 @item max-delay-slot-insn-search
5912 The maximum number of instructions to consider when looking for an
5913 instruction to fill a delay slot. If more than this arbitrary number of
5914 instructions is searched, the time savings from filling the delay slot
5915 will be minimal so stop searching. Increasing values mean more
5916 aggressive optimization, making the compile time increase with probably
5917 small improvement in executable run time.
5919 @item max-delay-slot-live-search
5920 When trying to fill delay slots, the maximum number of instructions to
5921 consider when searching for a block with valid live register
5922 information. Increasing this arbitrarily chosen value means more
5923 aggressive optimization, increasing the compile time. This parameter
5924 should be removed when the delay slot code is rewritten to maintain the
5927 @item max-gcse-memory
5928 The approximate maximum amount of memory that will be allocated in
5929 order to perform the global common subexpression elimination
5930 optimization. If more memory than specified is required, the
5931 optimization will not be done.
5933 @item max-gcse-passes
5934 The maximum number of passes of GCSE to run. The default is 1.
5936 @item max-pending-list-length
5937 The maximum number of pending dependencies scheduling will allow
5938 before flushing the current state and starting over. Large functions
5939 with few branches or calls can create excessively large lists which
5940 needlessly consume memory and resources.
5942 @item max-inline-insns-single
5943 Several parameters control the tree inliner used in gcc.
5944 This number sets the maximum number of instructions (counted in GCC's
5945 internal representation) in a single function that the tree inliner
5946 will consider for inlining. This only affects functions declared
5947 inline and methods implemented in a class declaration (C++).
5948 The default value is 450.
5950 @item max-inline-insns-auto
5951 When you use @option{-finline-functions} (included in @option{-O3}),
5952 a lot of functions that would otherwise not be considered for inlining
5953 by the compiler will be investigated. To those functions, a different
5954 (more restrictive) limit compared to functions declared inline can
5956 The default value is 90.
5958 @item large-function-insns
5959 The limit specifying really large functions. For functions larger than this
5960 limit after inlining inlining is constrained by
5961 @option{--param large-function-growth}. This parameter is useful primarily
5962 to avoid extreme compilation time caused by non-linear algorithms used by the
5964 This parameter is ignored when @option{-funit-at-a-time} is not used.
5965 The default value is 2700.
5967 @item large-function-growth
5968 Specifies maximal growth of large function caused by inlining in percents.
5969 This parameter is ignored when @option{-funit-at-a-time} is not used.
5970 The default value is 100 which limits large function growth to 2.0 times
5973 @item large-unit-insns
5974 The limit specifying large translation unit. Growth caused by inlining of
5975 units larger than this limit is limited by @option{--param inline-unit-growth}.
5976 For small units this might be too tight (consider unit consisting of function A
5977 that is inline and B that just calls A three time. If B is small relative to
5978 A, the growth of unit is 300\% and yet such inlining is very sane. For very
5979 large units consisting of small inlininable functions however the overall unit
5980 growth limit is needed to avoid exponential explosion of code size. Thus for
5981 smaller units, the size is increased to @option{--param large-unit-insns}
5982 before applying @option{--param inline-unit-growth}. The default is 10000
5984 @item inline-unit-growth
5985 Specifies maximal overall growth of the compilation unit caused by inlining.
5986 This parameter is ignored when @option{-funit-at-a-time} is not used.
5987 The default value is 50 which limits unit growth to 1.5 times the original
5990 @item max-inline-insns-recursive
5991 @itemx max-inline-insns-recursive-auto
5992 Specifies maximum number of instructions out-of-line copy of self recursive inline
5993 function can grow into by performing recursive inlining.
5995 For functions declared inline @option{--param max-inline-insns-recursive} is
5996 taken into account. For function not declared inline, recursive inlining
5997 happens only when @option{-finline-functions} (included in @option{-O3}) is
5998 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5999 default value is 450.
6001 @item max-inline-recursive-depth
6002 @itemx max-inline-recursive-depth-auto
6003 Specifies maximum recursion depth used by the recursive inlining.
6005 For functions declared inline @option{--param max-inline-recursive-depth} is
6006 taken into account. For function not declared inline, recursive inlining
6007 happens only when @option{-finline-functions} (included in @option{-O3}) is
6008 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
6009 default value is 450.
6011 @item min-inline-recursive-probability
6012 Recursive inlining is profitable only for function having deep recursion
6013 in average and can hurt for function having little recursion depth by
6014 increasing the prologue size or complexity of function body to other
6017 When profile feedback is available (see @option{-fprofile-generate}) the actual
6018 recursion depth can be guessed from probability that function will recurse via
6019 given call expression. This parameter limits inlining only to call expression
6020 whose probability exceeds given threshold (in percents). The default value is
6023 @item inline-call-cost
6024 Specify cost of call instruction relative to simple arithmetics operations
6025 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
6026 functions and at the same time increases size of leaf function that is believed to
6027 reduce function size by being inlined. In effect it increases amount of
6028 inlining for code having large abstraction penalty (many functions that just
6029 pass the arguments to other functions) and decrease inlining for code with low
6030 abstraction penalty. The default value is 16.
6032 @item max-unrolled-insns
6033 The maximum number of instructions that a loop should have if that loop
6034 is unrolled, and if the loop is unrolled, it determines how many times
6035 the loop code is unrolled.
6037 @item max-average-unrolled-insns
6038 The maximum number of instructions biased by probabilities of their execution
6039 that a loop should have if that loop is unrolled, and if the loop is unrolled,
6040 it determines how many times the loop code is unrolled.
6042 @item max-unroll-times
6043 The maximum number of unrollings of a single loop.
6045 @item max-peeled-insns
6046 The maximum number of instructions that a loop should have if that loop
6047 is peeled, and if the loop is peeled, it determines how many times
6048 the loop code is peeled.
6050 @item max-peel-times
6051 The maximum number of peelings of a single loop.
6053 @item max-completely-peeled-insns
6054 The maximum number of insns of a completely peeled loop.
6056 @item max-completely-peel-times
6057 The maximum number of iterations of a loop to be suitable for complete peeling.
6059 @item max-unswitch-insns
6060 The maximum number of insns of an unswitched loop.
6062 @item max-unswitch-level
6063 The maximum number of branches unswitched in a single loop.
6066 The minimum cost of an expensive expression in the loop invariant motion.
6068 @item iv-consider-all-candidates-bound
6069 Bound on number of candidates for induction variables below that
6070 all candidates are considered for each use in induction variable
6071 optimizations. Only the most relevant candidates are considered
6072 if there are more candidates, to avoid quadratic time complexity.
6074 @item iv-max-considered-uses
6075 The induction variable optimizations give up on loops that contain more
6076 induction variable uses.
6078 @item iv-always-prune-cand-set-bound
6079 If number of candidates in the set is smaller than this value,
6080 we always try to remove unnecessary ivs from the set during its
6081 optimization when a new iv is added to the set.
6083 @item scev-max-expr-size
6084 Bound on size of expressions used in the scalar evolutions analyzer.
6085 Large expressions slow the analyzer.
6087 @item vect-max-version-checks
6088 The maximum number of runtime checks that can be performed when doing
6089 loop versioning in the vectorizer. See option ftree-vect-loop-version
6090 for more information.
6092 @item max-iterations-to-track
6094 The maximum number of iterations of a loop the brute force algorithm
6095 for analysis of # of iterations of the loop tries to evaluate.
6097 @item hot-bb-count-fraction
6098 Select fraction of the maximal count of repetitions of basic block in program
6099 given basic block needs to have to be considered hot.
6101 @item hot-bb-frequency-fraction
6102 Select fraction of the maximal frequency of executions of basic block in
6103 function given basic block needs to have to be considered hot
6105 @item max-predicted-iterations
6106 The maximum number of loop iterations we predict statically. This is useful
6107 in cases where function contain single loop with known bound and other loop
6108 with unknown. We predict the known number of iterations correctly, while
6109 the unknown number of iterations average to roughly 10. This means that the
6110 loop without bounds would appear artificially cold relative to the other one.
6112 @item tracer-dynamic-coverage
6113 @itemx tracer-dynamic-coverage-feedback
6115 This value is used to limit superblock formation once the given percentage of
6116 executed instructions is covered. This limits unnecessary code size
6119 The @option{tracer-dynamic-coverage-feedback} is used only when profile
6120 feedback is available. The real profiles (as opposed to statically estimated
6121 ones) are much less balanced allowing the threshold to be larger value.
6123 @item tracer-max-code-growth
6124 Stop tail duplication once code growth has reached given percentage. This is
6125 rather hokey argument, as most of the duplicates will be eliminated later in
6126 cross jumping, so it may be set to much higher values than is the desired code
6129 @item tracer-min-branch-ratio
6131 Stop reverse growth when the reverse probability of best edge is less than this
6132 threshold (in percent).
6134 @item tracer-min-branch-ratio
6135 @itemx tracer-min-branch-ratio-feedback
6137 Stop forward growth if the best edge do have probability lower than this
6140 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
6141 compilation for profile feedback and one for compilation without. The value
6142 for compilation with profile feedback needs to be more conservative (higher) in
6143 order to make tracer effective.
6145 @item max-cse-path-length
6147 Maximum number of basic blocks on path that cse considers. The default is 10.
6150 The maximum instructions CSE process before flushing. The default is 1000.
6152 @item global-var-threshold
6154 Counts the number of function calls (@var{n}) and the number of
6155 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
6156 single artificial variable will be created to represent all the
6157 call-clobbered variables at function call sites. This artificial
6158 variable will then be made to alias every call-clobbered variable.
6159 (done as @code{int * size_t} on the host machine; beware overflow).
6161 @item max-aliased-vops
6163 Maximum number of virtual operands allowed to represent aliases
6164 before triggering the alias grouping heuristic. Alias grouping
6165 reduces compile times and memory consumption needed for aliasing at
6166 the expense of precision loss in alias information.
6168 @item ggc-min-expand
6170 GCC uses a garbage collector to manage its own memory allocation. This
6171 parameter specifies the minimum percentage by which the garbage
6172 collector's heap should be allowed to expand between collections.
6173 Tuning this may improve compilation speed; it has no effect on code
6176 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6177 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6178 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6179 GCC is not able to calculate RAM on a particular platform, the lower
6180 bound of 30% is used. Setting this parameter and
6181 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6182 every opportunity. This is extremely slow, but can be useful for
6185 @item ggc-min-heapsize
6187 Minimum size of the garbage collector's heap before it begins bothering
6188 to collect garbage. The first collection occurs after the heap expands
6189 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6190 tuning this may improve compilation speed, and has no effect on code
6193 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6194 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6195 with a lower bound of 4096 (four megabytes) and an upper bound of
6196 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6197 particular platform, the lower bound is used. Setting this parameter
6198 very large effectively disables garbage collection. Setting this
6199 parameter and @option{ggc-min-expand} to zero causes a full collection
6200 to occur at every opportunity.
6202 @item max-reload-search-insns
6203 The maximum number of instruction reload should look backward for equivalent
6204 register. Increasing values mean more aggressive optimization, making the
6205 compile time increase with probably slightly better performance. The default
6208 @item max-cselib-memory-locations
6209 The maximum number of memory locations cselib should take into account.
6210 Increasing values mean more aggressive optimization, making the compile time
6211 increase with probably slightly better performance. The default value is 500.
6213 @item max-flow-memory-locations
6214 Similar as @option{max-cselib-memory-locations} but for dataflow liveness.
6215 The default value is 100.
6217 @item reorder-blocks-duplicate
6218 @itemx reorder-blocks-duplicate-feedback
6220 Used by basic block reordering pass to decide whether to use unconditional
6221 branch or duplicate the code on its destination. Code is duplicated when its
6222 estimated size is smaller than this value multiplied by the estimated size of
6223 unconditional jump in the hot spots of the program.
6225 The @option{reorder-block-duplicate-feedback} is used only when profile
6226 feedback is available and may be set to higher values than
6227 @option{reorder-block-duplicate} since information about the hot spots is more
6230 @item max-sched-ready-insns
6231 The maximum number of instructions ready to be issued the scheduler should
6232 consider at any given time during the first scheduling pass. Increasing
6233 values mean more thorough searches, making the compilation time increase
6234 with probably little benefit. The default value is 100.
6236 @item max-sched-region-blocks
6237 The maximum number of blocks in a region to be considered for
6238 interblock scheduling. The default value is 10.
6240 @item max-sched-region-insns
6241 The maximum number of insns in a region to be considered for
6242 interblock scheduling. The default value is 100.
6245 The minimum probability (in percents) of reaching a source block
6246 for interblock speculative scheduling. The default value is 40.
6248 @item max-sched-extend-regions-iters
6249 The maximum number of iterations through CFG to extend regions.
6250 0 - disable region extension,
6251 N - do at most N iterations.
6252 The default value is 0.
6254 @item max-sched-insn-conflict-delay
6255 The maximum conflict delay for an insn to be considered for speculative motion.
6256 The default value is 3.
6258 @item sched-spec-prob-cutoff
6259 The minimal probability of speculation success (in percents), so that
6260 speculative insn will be scheduled.
6261 The default value is 40.
6263 @item max-last-value-rtl
6265 The maximum size measured as number of RTLs that can be recorded in an expression
6266 in combiner for a pseudo register as last known value of that register. The default
6269 @item integer-share-limit
6270 Small integer constants can use a shared data structure, reducing the
6271 compiler's memory usage and increasing its speed. This sets the maximum
6272 value of a shared integer constant's. The default value is 256.
6274 @item min-virtual-mappings
6275 Specifies the minimum number of virtual mappings in the incremental
6276 SSA updater that should be registered to trigger the virtual mappings
6277 heuristic defined by virtual-mappings-ratio. The default value is
6280 @item virtual-mappings-ratio
6281 If the number of virtual mappings is virtual-mappings-ratio bigger
6282 than the number of virtual symbols to be updated, then the incremental
6283 SSA updater switches to a full update for those symbols. The default
6286 @item ssp-buffer-size
6287 The minimum size of buffers (i.e. arrays) that will receive stack smashing
6288 protection when @option{-fstack-protection} is used.
6290 @item max-jump-thread-duplication-stmts
6291 Maximum number of statements allowed in a block that needs to be
6292 duplicated when threading jumps.
6294 @item max-fields-for-field-sensitive
6295 Maximum number of fields in a structure we will treat in
6296 a field sensitive manner during pointer analysis.
6301 @node Preprocessor Options
6302 @section Options Controlling the Preprocessor
6303 @cindex preprocessor options
6304 @cindex options, preprocessor
6306 These options control the C preprocessor, which is run on each C source
6307 file before actual compilation.
6309 If you use the @option{-E} option, nothing is done except preprocessing.
6310 Some of these options make sense only together with @option{-E} because
6311 they cause the preprocessor output to be unsuitable for actual
6316 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6317 and pass @var{option} directly through to the preprocessor. If
6318 @var{option} contains commas, it is split into multiple options at the
6319 commas. However, many options are modified, translated or interpreted
6320 by the compiler driver before being passed to the preprocessor, and
6321 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6322 interface is undocumented and subject to change, so whenever possible
6323 you should avoid using @option{-Wp} and let the driver handle the
6326 @item -Xpreprocessor @var{option}
6327 @opindex preprocessor
6328 Pass @var{option} as an option to the preprocessor. You can use this to
6329 supply system-specific preprocessor options which GCC does not know how to
6332 If you want to pass an option that takes an argument, you must use
6333 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6336 @include cppopts.texi
6338 @node Assembler Options
6339 @section Passing Options to the Assembler
6341 @c prevent bad page break with this line
6342 You can pass options to the assembler.
6345 @item -Wa,@var{option}
6347 Pass @var{option} as an option to the assembler. If @var{option}
6348 contains commas, it is split into multiple options at the commas.
6350 @item -Xassembler @var{option}
6352 Pass @var{option} as an option to the assembler. You can use this to
6353 supply system-specific assembler options which GCC does not know how to
6356 If you want to pass an option that takes an argument, you must use
6357 @option{-Xassembler} twice, once for the option and once for the argument.
6362 @section Options for Linking
6363 @cindex link options
6364 @cindex options, linking
6366 These options come into play when the compiler links object files into
6367 an executable output file. They are meaningless if the compiler is
6368 not doing a link step.
6372 @item @var{object-file-name}
6373 A file name that does not end in a special recognized suffix is
6374 considered to name an object file or library. (Object files are
6375 distinguished from libraries by the linker according to the file
6376 contents.) If linking is done, these object files are used as input
6385 If any of these options is used, then the linker is not run, and
6386 object file names should not be used as arguments. @xref{Overall
6390 @item -l@var{library}
6391 @itemx -l @var{library}
6393 Search the library named @var{library} when linking. (The second
6394 alternative with the library as a separate argument is only for
6395 POSIX compliance and is not recommended.)
6397 It makes a difference where in the command you write this option; the
6398 linker searches and processes libraries and object files in the order they
6399 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6400 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6401 to functions in @samp{z}, those functions may not be loaded.
6403 The linker searches a standard list of directories for the library,
6404 which is actually a file named @file{lib@var{library}.a}. The linker
6405 then uses this file as if it had been specified precisely by name.
6407 The directories searched include several standard system directories
6408 plus any that you specify with @option{-L}.
6410 Normally the files found this way are library files---archive files
6411 whose members are object files. The linker handles an archive file by
6412 scanning through it for members which define symbols that have so far
6413 been referenced but not defined. But if the file that is found is an
6414 ordinary object file, it is linked in the usual fashion. The only
6415 difference between using an @option{-l} option and specifying a file name
6416 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6417 and searches several directories.
6420 @opindex nostartfiles
6421 Do not use the standard system startup files when linking.
6422 The standard system libraries are used normally, unless @option{-nostdlib}
6423 or @option{-nodefaultlibs} is used.
6425 @item -nodefaultlibs
6426 @opindex nodefaultlibs
6427 Do not use the standard system libraries when linking.
6428 Only the libraries you specify will be passed to the linker.
6429 The standard startup files are used normally, unless @option{-nostartfiles}
6430 is used. The compiler may generate calls to @code{memcmp},
6431 @code{memset}, @code{memcpy} and @code{memmove}.
6432 These entries are usually resolved by entries in
6433 libc. These entry points should be supplied through some other
6434 mechanism when this option is specified.
6438 Do not use the standard system startup files or libraries when linking.
6439 No startup files and only the libraries you specify will be passed to
6440 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6441 @code{memcpy} and @code{memmove}.
6442 These entries are usually resolved by entries in
6443 libc. These entry points should be supplied through some other
6444 mechanism when this option is specified.
6446 @cindex @option{-lgcc}, use with @option{-nostdlib}
6447 @cindex @option{-nostdlib} and unresolved references
6448 @cindex unresolved references and @option{-nostdlib}
6449 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6450 @cindex @option{-nodefaultlibs} and unresolved references
6451 @cindex unresolved references and @option{-nodefaultlibs}
6452 One of the standard libraries bypassed by @option{-nostdlib} and
6453 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6454 that GCC uses to overcome shortcomings of particular machines, or special
6455 needs for some languages.
6456 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6457 Collection (GCC) Internals},
6458 for more discussion of @file{libgcc.a}.)
6459 In most cases, you need @file{libgcc.a} even when you want to avoid
6460 other standard libraries. In other words, when you specify @option{-nostdlib}
6461 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6462 This ensures that you have no unresolved references to internal GCC
6463 library subroutines. (For example, @samp{__main}, used to ensure C++
6464 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6465 GNU Compiler Collection (GCC) Internals}.)
6469 Produce a position independent executable on targets which support it.
6470 For predictable results, you must also specify the same set of options
6471 that were used to generate code (@option{-fpie}, @option{-fPIE},
6472 or model suboptions) when you specify this option.
6476 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
6477 that support it. This instructs the linker to add all symbols, not
6478 only used ones, to the dynamic symbol table. This option is needed
6479 for some uses of @code{dlopen} or to allow obtaining backtraces
6480 from within a program.
6484 Remove all symbol table and relocation information from the executable.
6488 On systems that support dynamic linking, this prevents linking with the shared
6489 libraries. On other systems, this option has no effect.
6493 Produce a shared object which can then be linked with other objects to
6494 form an executable. Not all systems support this option. For predictable
6495 results, you must also specify the same set of options that were used to
6496 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6497 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6498 needs to build supplementary stub code for constructors to work. On
6499 multi-libbed systems, @samp{gcc -shared} must select the correct support
6500 libraries to link against. Failing to supply the correct flags may lead
6501 to subtle defects. Supplying them in cases where they are not necessary
6504 @item -shared-libgcc
6505 @itemx -static-libgcc
6506 @opindex shared-libgcc
6507 @opindex static-libgcc
6508 On systems that provide @file{libgcc} as a shared library, these options
6509 force the use of either the shared or static version respectively.
6510 If no shared version of @file{libgcc} was built when the compiler was
6511 configured, these options have no effect.
6513 There are several situations in which an application should use the
6514 shared @file{libgcc} instead of the static version. The most common
6515 of these is when the application wishes to throw and catch exceptions
6516 across different shared libraries. In that case, each of the libraries
6517 as well as the application itself should use the shared @file{libgcc}.
6519 Therefore, the G++ and GCJ drivers automatically add
6520 @option{-shared-libgcc} whenever you build a shared library or a main
6521 executable, because C++ and Java programs typically use exceptions, so
6522 this is the right thing to do.
6524 If, instead, you use the GCC driver to create shared libraries, you may
6525 find that they will not always be linked with the shared @file{libgcc}.
6526 If GCC finds, at its configuration time, that you have a non-GNU linker
6527 or a GNU linker that does not support option @option{--eh-frame-hdr},
6528 it will link the shared version of @file{libgcc} into shared libraries
6529 by default. Otherwise, it will take advantage of the linker and optimize
6530 away the linking with the shared version of @file{libgcc}, linking with
6531 the static version of libgcc by default. This allows exceptions to
6532 propagate through such shared libraries, without incurring relocation
6533 costs at library load time.
6535 However, if a library or main executable is supposed to throw or catch
6536 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6537 for the languages used in the program, or using the option
6538 @option{-shared-libgcc}, such that it is linked with the shared
6543 Bind references to global symbols when building a shared object. Warn
6544 about any unresolved references (unless overridden by the link editor
6545 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6548 @item -Xlinker @var{option}
6550 Pass @var{option} as an option to the linker. You can use this to
6551 supply system-specific linker options which GCC does not know how to
6554 If you want to pass an option that takes an argument, you must use
6555 @option{-Xlinker} twice, once for the option and once for the argument.
6556 For example, to pass @option{-assert definitions}, you must write
6557 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6558 @option{-Xlinker "-assert definitions"}, because this passes the entire
6559 string as a single argument, which is not what the linker expects.
6561 @item -Wl,@var{option}
6563 Pass @var{option} as an option to the linker. If @var{option} contains
6564 commas, it is split into multiple options at the commas.
6566 @item -u @var{symbol}
6568 Pretend the symbol @var{symbol} is undefined, to force linking of
6569 library modules to define it. You can use @option{-u} multiple times with
6570 different symbols to force loading of additional library modules.
6573 @node Directory Options
6574 @section Options for Directory Search
6575 @cindex directory options
6576 @cindex options, directory search
6579 These options specify directories to search for header files, for
6580 libraries and for parts of the compiler:
6585 Add the directory @var{dir} to the head of the list of directories to be
6586 searched for header files. This can be used to override a system header
6587 file, substituting your own version, since these directories are
6588 searched before the system header file directories. However, you should
6589 not use this option to add directories that contain vendor-supplied
6590 system header files (use @option{-isystem} for that). If you use more than
6591 one @option{-I} option, the directories are scanned in left-to-right
6592 order; the standard system directories come after.
6594 If a standard system include directory, or a directory specified with
6595 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6596 option will be ignored. The directory will still be searched but as a
6597 system directory at its normal position in the system include chain.
6598 This is to ensure that GCC's procedure to fix buggy system headers and
6599 the ordering for the include_next directive are not inadvertently changed.
6600 If you really need to change the search order for system directories,
6601 use the @option{-nostdinc} and/or @option{-isystem} options.
6603 @item -iquote@var{dir}
6605 Add the directory @var{dir} to the head of the list of directories to
6606 be searched for header files only for the case of @samp{#include
6607 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6608 otherwise just like @option{-I}.
6612 Add directory @var{dir} to the list of directories to be searched
6615 @item -B@var{prefix}
6617 This option specifies where to find the executables, libraries,
6618 include files, and data files of the compiler itself.
6620 The compiler driver program runs one or more of the subprograms
6621 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6622 @var{prefix} as a prefix for each program it tries to run, both with and
6623 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6625 For each subprogram to be run, the compiler driver first tries the
6626 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6627 was not specified, the driver tries two standard prefixes, which are
6628 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6629 those results in a file name that is found, the unmodified program
6630 name is searched for using the directories specified in your
6631 @env{PATH} environment variable.
6633 The compiler will check to see if the path provided by the @option{-B}
6634 refers to a directory, and if necessary it will add a directory
6635 separator character at the end of the path.
6637 @option{-B} prefixes that effectively specify directory names also apply
6638 to libraries in the linker, because the compiler translates these
6639 options into @option{-L} options for the linker. They also apply to
6640 includes files in the preprocessor, because the compiler translates these
6641 options into @option{-isystem} options for the preprocessor. In this case,
6642 the compiler appends @samp{include} to the prefix.
6644 The run-time support file @file{libgcc.a} can also be searched for using
6645 the @option{-B} prefix, if needed. If it is not found there, the two
6646 standard prefixes above are tried, and that is all. The file is left
6647 out of the link if it is not found by those means.
6649 Another way to specify a prefix much like the @option{-B} prefix is to use
6650 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6653 As a special kludge, if the path provided by @option{-B} is
6654 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6655 9, then it will be replaced by @file{[dir/]include}. This is to help
6656 with boot-strapping the compiler.
6658 @item -specs=@var{file}
6660 Process @var{file} after the compiler reads in the standard @file{specs}
6661 file, in order to override the defaults that the @file{gcc} driver
6662 program uses when determining what switches to pass to @file{cc1},
6663 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6664 @option{-specs=@var{file}} can be specified on the command line, and they
6665 are processed in order, from left to right.
6667 @item --sysroot=@var{dir}
6669 Use @var{dir} as the logical root directory for headers and libraries.
6670 For example, if the compiler would normally search for headers in
6671 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
6672 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
6674 If you use both this option and the @option{-isysroot} option, then
6675 the @option{--sysroot} option will apply to libraries, but the
6676 @option{-isysroot} option will apply to header files.
6678 The GNU linker (beginning with version 2.16) has the necessary support
6679 for this option. If your linker does not support this option, the
6680 header file aspect of @option{--sysroot} will still work, but the
6681 library aspect will not.
6685 This option has been deprecated. Please use @option{-iquote} instead for
6686 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6687 Any directories you specify with @option{-I} options before the @option{-I-}
6688 option are searched only for the case of @samp{#include "@var{file}"};
6689 they are not searched for @samp{#include <@var{file}>}.
6691 If additional directories are specified with @option{-I} options after
6692 the @option{-I-}, these directories are searched for all @samp{#include}
6693 directives. (Ordinarily @emph{all} @option{-I} directories are used
6696 In addition, the @option{-I-} option inhibits the use of the current
6697 directory (where the current input file came from) as the first search
6698 directory for @samp{#include "@var{file}"}. There is no way to
6699 override this effect of @option{-I-}. With @option{-I.} you can specify
6700 searching the directory which was current when the compiler was
6701 invoked. That is not exactly the same as what the preprocessor does
6702 by default, but it is often satisfactory.
6704 @option{-I-} does not inhibit the use of the standard system directories
6705 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6712 @section Specifying subprocesses and the switches to pass to them
6715 @command{gcc} is a driver program. It performs its job by invoking a
6716 sequence of other programs to do the work of compiling, assembling and
6717 linking. GCC interprets its command-line parameters and uses these to
6718 deduce which programs it should invoke, and which command-line options
6719 it ought to place on their command lines. This behavior is controlled
6720 by @dfn{spec strings}. In most cases there is one spec string for each
6721 program that GCC can invoke, but a few programs have multiple spec
6722 strings to control their behavior. The spec strings built into GCC can
6723 be overridden by using the @option{-specs=} command-line switch to specify
6726 @dfn{Spec files} are plaintext files that are used to construct spec
6727 strings. They consist of a sequence of directives separated by blank
6728 lines. The type of directive is determined by the first non-whitespace
6729 character on the line and it can be one of the following:
6732 @item %@var{command}
6733 Issues a @var{command} to the spec file processor. The commands that can
6737 @item %include <@var{file}>
6739 Search for @var{file} and insert its text at the current point in the
6742 @item %include_noerr <@var{file}>
6743 @cindex %include_noerr
6744 Just like @samp{%include}, but do not generate an error message if the include
6745 file cannot be found.
6747 @item %rename @var{old_name} @var{new_name}
6749 Rename the spec string @var{old_name} to @var{new_name}.
6753 @item *[@var{spec_name}]:
6754 This tells the compiler to create, override or delete the named spec
6755 string. All lines after this directive up to the next directive or
6756 blank line are considered to be the text for the spec string. If this
6757 results in an empty string then the spec will be deleted. (Or, if the
6758 spec did not exist, then nothing will happened.) Otherwise, if the spec
6759 does not currently exist a new spec will be created. If the spec does
6760 exist then its contents will be overridden by the text of this
6761 directive, unless the first character of that text is the @samp{+}
6762 character, in which case the text will be appended to the spec.
6764 @item [@var{suffix}]:
6765 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6766 and up to the next directive or blank line are considered to make up the
6767 spec string for the indicated suffix. When the compiler encounters an
6768 input file with the named suffix, it will processes the spec string in
6769 order to work out how to compile that file. For example:
6776 This says that any input file whose name ends in @samp{.ZZ} should be
6777 passed to the program @samp{z-compile}, which should be invoked with the
6778 command-line switch @option{-input} and with the result of performing the
6779 @samp{%i} substitution. (See below.)
6781 As an alternative to providing a spec string, the text that follows a
6782 suffix directive can be one of the following:
6785 @item @@@var{language}
6786 This says that the suffix is an alias for a known @var{language}. This is
6787 similar to using the @option{-x} command-line switch to GCC to specify a
6788 language explicitly. For example:
6795 Says that .ZZ files are, in fact, C++ source files.
6798 This causes an error messages saying:
6801 @var{name} compiler not installed on this system.
6805 GCC already has an extensive list of suffixes built into it.
6806 This directive will add an entry to the end of the list of suffixes, but
6807 since the list is searched from the end backwards, it is effectively
6808 possible to override earlier entries using this technique.
6812 GCC has the following spec strings built into it. Spec files can
6813 override these strings or create their own. Note that individual
6814 targets can also add their own spec strings to this list.
6817 asm Options to pass to the assembler
6818 asm_final Options to pass to the assembler post-processor
6819 cpp Options to pass to the C preprocessor
6820 cc1 Options to pass to the C compiler
6821 cc1plus Options to pass to the C++ compiler
6822 endfile Object files to include at the end of the link
6823 link Options to pass to the linker
6824 lib Libraries to include on the command line to the linker
6825 libgcc Decides which GCC support library to pass to the linker
6826 linker Sets the name of the linker
6827 predefines Defines to be passed to the C preprocessor
6828 signed_char Defines to pass to CPP to say whether @code{char} is signed
6830 startfile Object files to include at the start of the link
6833 Here is a small example of a spec file:
6839 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6842 This example renames the spec called @samp{lib} to @samp{old_lib} and
6843 then overrides the previous definition of @samp{lib} with a new one.
6844 The new definition adds in some extra command-line options before
6845 including the text of the old definition.
6847 @dfn{Spec strings} are a list of command-line options to be passed to their
6848 corresponding program. In addition, the spec strings can contain
6849 @samp{%}-prefixed sequences to substitute variable text or to
6850 conditionally insert text into the command line. Using these constructs
6851 it is possible to generate quite complex command lines.
6853 Here is a table of all defined @samp{%}-sequences for spec
6854 strings. Note that spaces are not generated automatically around the
6855 results of expanding these sequences. Therefore you can concatenate them
6856 together or combine them with constant text in a single argument.
6860 Substitute one @samp{%} into the program name or argument.
6863 Substitute the name of the input file being processed.
6866 Substitute the basename of the input file being processed.
6867 This is the substring up to (and not including) the last period
6868 and not including the directory.
6871 This is the same as @samp{%b}, but include the file suffix (text after
6875 Marks the argument containing or following the @samp{%d} as a
6876 temporary file name, so that that file will be deleted if GCC exits
6877 successfully. Unlike @samp{%g}, this contributes no text to the
6880 @item %g@var{suffix}
6881 Substitute a file name that has suffix @var{suffix} and is chosen
6882 once per compilation, and mark the argument in the same way as
6883 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6884 name is now chosen in a way that is hard to predict even when previously
6885 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6886 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6887 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6888 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6889 was simply substituted with a file name chosen once per compilation,
6890 without regard to any appended suffix (which was therefore treated
6891 just like ordinary text), making such attacks more likely to succeed.
6893 @item %u@var{suffix}
6894 Like @samp{%g}, but generates a new temporary file name even if
6895 @samp{%u@var{suffix}} was already seen.
6897 @item %U@var{suffix}
6898 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6899 new one if there is no such last file name. In the absence of any
6900 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6901 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6902 would involve the generation of two distinct file names, one
6903 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6904 simply substituted with a file name chosen for the previous @samp{%u},
6905 without regard to any appended suffix.
6907 @item %j@var{suffix}
6908 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6909 writable, and if save-temps is off; otherwise, substitute the name
6910 of a temporary file, just like @samp{%u}. This temporary file is not
6911 meant for communication between processes, but rather as a junk
6914 @item %|@var{suffix}
6915 @itemx %m@var{suffix}
6916 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6917 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6918 all. These are the two most common ways to instruct a program that it
6919 should read from standard input or write to standard output. If you
6920 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6921 construct: see for example @file{f/lang-specs.h}.
6923 @item %.@var{SUFFIX}
6924 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6925 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6926 terminated by the next space or %.
6929 Marks the argument containing or following the @samp{%w} as the
6930 designated output file of this compilation. This puts the argument
6931 into the sequence of arguments that @samp{%o} will substitute later.
6934 Substitutes the names of all the output files, with spaces
6935 automatically placed around them. You should write spaces
6936 around the @samp{%o} as well or the results are undefined.
6937 @samp{%o} is for use in the specs for running the linker.
6938 Input files whose names have no recognized suffix are not compiled
6939 at all, but they are included among the output files, so they will
6943 Substitutes the suffix for object files. Note that this is
6944 handled specially when it immediately follows @samp{%g, %u, or %U},
6945 because of the need for those to form complete file names. The
6946 handling is such that @samp{%O} is treated exactly as if it had already
6947 been substituted, except that @samp{%g, %u, and %U} do not currently
6948 support additional @var{suffix} characters following @samp{%O} as they would
6949 following, for example, @samp{.o}.
6952 Substitutes the standard macro predefinitions for the
6953 current target machine. Use this when running @code{cpp}.
6956 Like @samp{%p}, but puts @samp{__} before and after the name of each
6957 predefined macro, except for macros that start with @samp{__} or with
6958 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6962 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6963 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
6964 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6965 and @option{-imultilib} as necessary.
6968 Current argument is the name of a library or startup file of some sort.
6969 Search for that file in a standard list of directories and substitute
6970 the full name found.
6973 Print @var{str} as an error message. @var{str} is terminated by a newline.
6974 Use this when inconsistent options are detected.
6977 Substitute the contents of spec string @var{name} at this point.
6980 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6982 @item %x@{@var{option}@}
6983 Accumulate an option for @samp{%X}.
6986 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6990 Output the accumulated assembler options specified by @option{-Wa}.
6993 Output the accumulated preprocessor options specified by @option{-Wp}.
6996 Process the @code{asm} spec. This is used to compute the
6997 switches to be passed to the assembler.
7000 Process the @code{asm_final} spec. This is a spec string for
7001 passing switches to an assembler post-processor, if such a program is
7005 Process the @code{link} spec. This is the spec for computing the
7006 command line passed to the linker. Typically it will make use of the
7007 @samp{%L %G %S %D and %E} sequences.
7010 Dump out a @option{-L} option for each directory that GCC believes might
7011 contain startup files. If the target supports multilibs then the
7012 current multilib directory will be prepended to each of these paths.
7015 Process the @code{lib} spec. This is a spec string for deciding which
7016 libraries should be included on the command line to the linker.
7019 Process the @code{libgcc} spec. This is a spec string for deciding
7020 which GCC support library should be included on the command line to the linker.
7023 Process the @code{startfile} spec. This is a spec for deciding which
7024 object files should be the first ones passed to the linker. Typically
7025 this might be a file named @file{crt0.o}.
7028 Process the @code{endfile} spec. This is a spec string that specifies
7029 the last object files that will be passed to the linker.
7032 Process the @code{cpp} spec. This is used to construct the arguments
7033 to be passed to the C preprocessor.
7036 Process the @code{cc1} spec. This is used to construct the options to be
7037 passed to the actual C compiler (@samp{cc1}).
7040 Process the @code{cc1plus} spec. This is used to construct the options to be
7041 passed to the actual C++ compiler (@samp{cc1plus}).
7044 Substitute the variable part of a matched option. See below.
7045 Note that each comma in the substituted string is replaced by
7049 Remove all occurrences of @code{-S} from the command line. Note---this
7050 command is position dependent. @samp{%} commands in the spec string
7051 before this one will see @code{-S}, @samp{%} commands in the spec string
7052 after this one will not.
7054 @item %:@var{function}(@var{args})
7055 Call the named function @var{function}, passing it @var{args}.
7056 @var{args} is first processed as a nested spec string, then split
7057 into an argument vector in the usual fashion. The function returns
7058 a string which is processed as if it had appeared literally as part
7059 of the current spec.
7061 The following built-in spec functions are provided:
7064 @item @code{if-exists}
7065 The @code{if-exists} spec function takes one argument, an absolute
7066 pathname to a file. If the file exists, @code{if-exists} returns the
7067 pathname. Here is a small example of its usage:
7071 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
7074 @item @code{if-exists-else}
7075 The @code{if-exists-else} spec function is similar to the @code{if-exists}
7076 spec function, except that it takes two arguments. The first argument is
7077 an absolute pathname to a file. If the file exists, @code{if-exists-else}
7078 returns the pathname. If it does not exist, it returns the second argument.
7079 This way, @code{if-exists-else} can be used to select one file or another,
7080 based on the existence of the first. Here is a small example of its usage:
7084 crt0%O%s %:if-exists(crti%O%s) \
7085 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
7088 @item @code{replace-outfile}
7089 The @code{replace-outfile} spec function takes two arguments. It looks for the
7090 first argument in the outfiles array and replaces it with the second argument. Here
7091 is a small example of its usage:
7094 %@{static|static-libgcc|static-libstdc++:%:replace-outfile(-lstdc++ \
7102 Substitutes the @code{-S} switch, if that switch was given to GCC@.
7103 If that switch was not specified, this substitutes nothing. Note that
7104 the leading dash is omitted when specifying this option, and it is
7105 automatically inserted if the substitution is performed. Thus the spec
7106 string @samp{%@{foo@}} would match the command-line option @option{-foo}
7107 and would output the command line option @option{-foo}.
7109 @item %W@{@code{S}@}
7110 Like %@{@code{S}@} but mark last argument supplied within as a file to be
7113 @item %@{@code{S}*@}
7114 Substitutes all the switches specified to GCC whose names start
7115 with @code{-S}, but which also take an argument. This is used for
7116 switches like @option{-o}, @option{-D}, @option{-I}, etc.
7117 GCC considers @option{-o foo} as being
7118 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
7119 text, including the space. Thus two arguments would be generated.
7121 @item %@{@code{S}*&@code{T}*@}
7122 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
7123 (the order of @code{S} and @code{T} in the spec is not significant).
7124 There can be any number of ampersand-separated variables; for each the
7125 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
7127 @item %@{@code{S}:@code{X}@}
7128 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
7130 @item %@{!@code{S}:@code{X}@}
7131 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
7133 @item %@{@code{S}*:@code{X}@}
7134 Substitutes @code{X} if one or more switches whose names start with
7135 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
7136 once, no matter how many such switches appeared. However, if @code{%*}
7137 appears somewhere in @code{X}, then @code{X} will be substituted once
7138 for each matching switch, with the @code{%*} replaced by the part of
7139 that switch that matched the @code{*}.
7141 @item %@{.@code{S}:@code{X}@}
7142 Substitutes @code{X}, if processing a file with suffix @code{S}.
7144 @item %@{!.@code{S}:@code{X}@}
7145 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
7147 @item %@{@code{S}|@code{P}:@code{X}@}
7148 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
7149 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
7150 although they have a stronger binding than the @samp{|}. If @code{%*}
7151 appears in @code{X}, all of the alternatives must be starred, and only
7152 the first matching alternative is substituted.
7154 For example, a spec string like this:
7157 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
7160 will output the following command-line options from the following input
7161 command-line options:
7166 -d fred.c -foo -baz -boggle
7167 -d jim.d -bar -baz -boggle
7170 @item %@{S:X; T:Y; :D@}
7172 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
7173 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
7174 be as many clauses as you need. This may be combined with @code{.},
7175 @code{!}, @code{|}, and @code{*} as needed.
7180 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
7181 construct may contain other nested @samp{%} constructs or spaces, or
7182 even newlines. They are processed as usual, as described above.
7183 Trailing white space in @code{X} is ignored. White space may also
7184 appear anywhere on the left side of the colon in these constructs,
7185 except between @code{.} or @code{*} and the corresponding word.
7187 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
7188 handled specifically in these constructs. If another value of
7189 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
7190 @option{-W} switch is found later in the command line, the earlier
7191 switch value is ignored, except with @{@code{S}*@} where @code{S} is
7192 just one letter, which passes all matching options.
7194 The character @samp{|} at the beginning of the predicate text is used to
7195 indicate that a command should be piped to the following command, but
7196 only if @option{-pipe} is specified.
7198 It is built into GCC which switches take arguments and which do not.
7199 (You might think it would be useful to generalize this to allow each
7200 compiler's spec to say which switches take arguments. But this cannot
7201 be done in a consistent fashion. GCC cannot even decide which input
7202 files have been specified without knowing which switches take arguments,
7203 and it must know which input files to compile in order to tell which
7206 GCC also knows implicitly that arguments starting in @option{-l} are to be
7207 treated as compiler output files, and passed to the linker in their
7208 proper position among the other output files.
7210 @c man begin OPTIONS
7212 @node Target Options
7213 @section Specifying Target Machine and Compiler Version
7214 @cindex target options
7215 @cindex cross compiling
7216 @cindex specifying machine version
7217 @cindex specifying compiler version and target machine
7218 @cindex compiler version, specifying
7219 @cindex target machine, specifying
7221 The usual way to run GCC is to run the executable called @file{gcc}, or
7222 @file{<machine>-gcc} when cross-compiling, or
7223 @file{<machine>-gcc-<version>} to run a version other than the one that
7224 was installed last. Sometimes this is inconvenient, so GCC provides
7225 options that will switch to another cross-compiler or version.
7228 @item -b @var{machine}
7230 The argument @var{machine} specifies the target machine for compilation.
7232 The value to use for @var{machine} is the same as was specified as the
7233 machine type when configuring GCC as a cross-compiler. For
7234 example, if a cross-compiler was configured with @samp{configure
7235 arm-elf}, meaning to compile for an arm processor with elf binaries,
7236 then you would specify @option{-b arm-elf} to run that cross compiler.
7237 Because there are other options beginning with @option{-b}, the
7238 configuration must contain a hyphen.
7240 @item -V @var{version}
7242 The argument @var{version} specifies which version of GCC to run.
7243 This is useful when multiple versions are installed. For example,
7244 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
7247 The @option{-V} and @option{-b} options work by running the
7248 @file{<machine>-gcc-<version>} executable, so there's no real reason to
7249 use them if you can just run that directly.
7251 @node Submodel Options
7252 @section Hardware Models and Configurations
7253 @cindex submodel options
7254 @cindex specifying hardware config
7255 @cindex hardware models and configurations, specifying
7256 @cindex machine dependent options
7258 Earlier we discussed the standard option @option{-b} which chooses among
7259 different installed compilers for completely different target
7260 machines, such as VAX vs.@: 68000 vs.@: 80386.
7262 In addition, each of these target machine types can have its own
7263 special options, starting with @samp{-m}, to choose among various
7264 hardware models or configurations---for example, 68010 vs 68020,
7265 floating coprocessor or none. A single installed version of the
7266 compiler can compile for any model or configuration, according to the
7269 Some configurations of the compiler also support additional special
7270 options, usually for compatibility with other compilers on the same
7273 @c This list is ordered alphanumerically by subsection name.
7274 @c It should be the same order and spelling as these options are listed
7275 @c in Machine Dependent Options
7281 * Blackfin Options::
7285 * DEC Alpha Options::
7286 * DEC Alpha/VMS Options::
7288 * GNU/Linux Options::
7291 * i386 and x86-64 Options::
7304 * RS/6000 and PowerPC Options::
7305 * S/390 and zSeries Options::
7309 * System V Options::
7310 * TMS320C3x/C4x Options::
7314 * Xstormy16 Options::
7320 @subsection ARC Options
7323 These options are defined for ARC implementations:
7328 Compile code for little endian mode. This is the default.
7332 Compile code for big endian mode.
7335 @opindex mmangle-cpu
7336 Prepend the name of the cpu to all public symbol names.
7337 In multiple-processor systems, there are many ARC variants with different
7338 instruction and register set characteristics. This flag prevents code
7339 compiled for one cpu to be linked with code compiled for another.
7340 No facility exists for handling variants that are ``almost identical''.
7341 This is an all or nothing option.
7343 @item -mcpu=@var{cpu}
7345 Compile code for ARC variant @var{cpu}.
7346 Which variants are supported depend on the configuration.
7347 All variants support @option{-mcpu=base}, this is the default.
7349 @item -mtext=@var{text-section}
7350 @itemx -mdata=@var{data-section}
7351 @itemx -mrodata=@var{readonly-data-section}
7355 Put functions, data, and readonly data in @var{text-section},
7356 @var{data-section}, and @var{readonly-data-section} respectively
7357 by default. This can be overridden with the @code{section} attribute.
7358 @xref{Variable Attributes}.
7363 @subsection ARM Options
7366 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7370 @item -mabi=@var{name}
7372 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7373 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
7376 @opindex mapcs-frame
7377 Generate a stack frame that is compliant with the ARM Procedure Call
7378 Standard for all functions, even if this is not strictly necessary for
7379 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7380 with this option will cause the stack frames not to be generated for
7381 leaf functions. The default is @option{-mno-apcs-frame}.
7385 This is a synonym for @option{-mapcs-frame}.
7388 @c not currently implemented
7389 @item -mapcs-stack-check
7390 @opindex mapcs-stack-check
7391 Generate code to check the amount of stack space available upon entry to
7392 every function (that actually uses some stack space). If there is
7393 insufficient space available then either the function
7394 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7395 called, depending upon the amount of stack space required. The run time
7396 system is required to provide these functions. The default is
7397 @option{-mno-apcs-stack-check}, since this produces smaller code.
7399 @c not currently implemented
7401 @opindex mapcs-float
7402 Pass floating point arguments using the float point registers. This is
7403 one of the variants of the APCS@. This option is recommended if the
7404 target hardware has a floating point unit or if a lot of floating point
7405 arithmetic is going to be performed by the code. The default is
7406 @option{-mno-apcs-float}, since integer only code is slightly increased in
7407 size if @option{-mapcs-float} is used.
7409 @c not currently implemented
7410 @item -mapcs-reentrant
7411 @opindex mapcs-reentrant
7412 Generate reentrant, position independent code. The default is
7413 @option{-mno-apcs-reentrant}.
7416 @item -mthumb-interwork
7417 @opindex mthumb-interwork
7418 Generate code which supports calling between the ARM and Thumb
7419 instruction sets. Without this option the two instruction sets cannot
7420 be reliably used inside one program. The default is
7421 @option{-mno-thumb-interwork}, since slightly larger code is generated
7422 when @option{-mthumb-interwork} is specified.
7424 @item -mno-sched-prolog
7425 @opindex mno-sched-prolog
7426 Prevent the reordering of instructions in the function prolog, or the
7427 merging of those instruction with the instructions in the function's
7428 body. This means that all functions will start with a recognizable set
7429 of instructions (or in fact one of a choice from a small set of
7430 different function prologues), and this information can be used to
7431 locate the start if functions inside an executable piece of code. The
7432 default is @option{-msched-prolog}.
7435 @opindex mhard-float
7436 Generate output containing floating point instructions. This is the
7440 @opindex msoft-float
7441 Generate output containing library calls for floating point.
7442 @strong{Warning:} the requisite libraries are not available for all ARM
7443 targets. Normally the facilities of the machine's usual C compiler are
7444 used, but this cannot be done directly in cross-compilation. You must make
7445 your own arrangements to provide suitable library functions for
7448 @option{-msoft-float} changes the calling convention in the output file;
7449 therefore, it is only useful if you compile @emph{all} of a program with
7450 this option. In particular, you need to compile @file{libgcc.a}, the
7451 library that comes with GCC, with @option{-msoft-float} in order for
7454 @item -mfloat-abi=@var{name}
7456 Specifies which ABI to use for floating point values. Permissible values
7457 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7459 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7460 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7461 of floating point instructions, but still uses the soft-float calling
7464 @item -mlittle-endian
7465 @opindex mlittle-endian
7466 Generate code for a processor running in little-endian mode. This is
7467 the default for all standard configurations.
7470 @opindex mbig-endian
7471 Generate code for a processor running in big-endian mode; the default is
7472 to compile code for a little-endian processor.
7474 @item -mwords-little-endian
7475 @opindex mwords-little-endian
7476 This option only applies when generating code for big-endian processors.
7477 Generate code for a little-endian word order but a big-endian byte
7478 order. That is, a byte order of the form @samp{32107654}. Note: this
7479 option should only be used if you require compatibility with code for
7480 big-endian ARM processors generated by versions of the compiler prior to
7483 @item -mcpu=@var{name}
7485 This specifies the name of the target ARM processor. GCC uses this name
7486 to determine what kind of instructions it can emit when generating
7487 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7488 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7489 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7490 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7491 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7492 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7493 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7494 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7495 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7496 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7497 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7498 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7499 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7500 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
7503 @itemx -mtune=@var{name}
7505 This option is very similar to the @option{-mcpu=} option, except that
7506 instead of specifying the actual target processor type, and hence
7507 restricting which instructions can be used, it specifies that GCC should
7508 tune the performance of the code as if the target were of the type
7509 specified in this option, but still choosing the instructions that it
7510 will generate based on the cpu specified by a @option{-mcpu=} option.
7511 For some ARM implementations better performance can be obtained by using
7514 @item -march=@var{name}
7516 This specifies the name of the target ARM architecture. GCC uses this
7517 name to determine what kind of instructions it can emit when generating
7518 assembly code. This option can be used in conjunction with or instead
7519 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7520 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7521 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7522 @samp{iwmmxt}, @samp{ep9312}.
7524 @item -mfpu=@var{name}
7525 @itemx -mfpe=@var{number}
7526 @itemx -mfp=@var{number}
7530 This specifies what floating point hardware (or hardware emulation) is
7531 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7532 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7533 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7534 with older versions of GCC@.
7536 If @option{-msoft-float} is specified this specifies the format of
7537 floating point values.
7539 @item -mstructure-size-boundary=@var{n}
7540 @opindex mstructure-size-boundary
7541 The size of all structures and unions will be rounded up to a multiple
7542 of the number of bits set by this option. Permissible values are 8, 32
7543 and 64. The default value varies for different toolchains. For the COFF
7544 targeted toolchain the default value is 8. A value of 64 is only allowed
7545 if the underlying ABI supports it.
7547 Specifying the larger number can produce faster, more efficient code, but
7548 can also increase the size of the program. Different values are potentially
7549 incompatible. Code compiled with one value cannot necessarily expect to
7550 work with code or libraries compiled with another value, if they exchange
7551 information using structures or unions.
7553 @item -mabort-on-noreturn
7554 @opindex mabort-on-noreturn
7555 Generate a call to the function @code{abort} at the end of a
7556 @code{noreturn} function. It will be executed if the function tries to
7560 @itemx -mno-long-calls
7561 @opindex mlong-calls
7562 @opindex mno-long-calls
7563 Tells the compiler to perform function calls by first loading the
7564 address of the function into a register and then performing a subroutine
7565 call on this register. This switch is needed if the target function
7566 will lie outside of the 64 megabyte addressing range of the offset based
7567 version of subroutine call instruction.
7569 Even if this switch is enabled, not all function calls will be turned
7570 into long calls. The heuristic is that static functions, functions
7571 which have the @samp{short-call} attribute, functions that are inside
7572 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7573 definitions have already been compiled within the current compilation
7574 unit, will not be turned into long calls. The exception to this rule is
7575 that weak function definitions, functions with the @samp{long-call}
7576 attribute or the @samp{section} attribute, and functions that are within
7577 the scope of a @samp{#pragma long_calls} directive, will always be
7578 turned into long calls.
7580 This feature is not enabled by default. Specifying
7581 @option{-mno-long-calls} will restore the default behavior, as will
7582 placing the function calls within the scope of a @samp{#pragma
7583 long_calls_off} directive. Note these switches have no effect on how
7584 the compiler generates code to handle function calls via function
7587 @item -mnop-fun-dllimport
7588 @opindex mnop-fun-dllimport
7589 Disable support for the @code{dllimport} attribute.
7591 @item -msingle-pic-base
7592 @opindex msingle-pic-base
7593 Treat the register used for PIC addressing as read-only, rather than
7594 loading it in the prologue for each function. The run-time system is
7595 responsible for initializing this register with an appropriate value
7596 before execution begins.
7598 @item -mpic-register=@var{reg}
7599 @opindex mpic-register
7600 Specify the register to be used for PIC addressing. The default is R10
7601 unless stack-checking is enabled, when R9 is used.
7603 @item -mcirrus-fix-invalid-insns
7604 @opindex mcirrus-fix-invalid-insns
7605 @opindex mno-cirrus-fix-invalid-insns
7606 Insert NOPs into the instruction stream to in order to work around
7607 problems with invalid Maverick instruction combinations. This option
7608 is only valid if the @option{-mcpu=ep9312} option has been used to
7609 enable generation of instructions for the Cirrus Maverick floating
7610 point co-processor. This option is not enabled by default, since the
7611 problem is only present in older Maverick implementations. The default
7612 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7615 @item -mpoke-function-name
7616 @opindex mpoke-function-name
7617 Write the name of each function into the text section, directly
7618 preceding the function prologue. The generated code is similar to this:
7622 .ascii "arm_poke_function_name", 0
7625 .word 0xff000000 + (t1 - t0)
7626 arm_poke_function_name
7628 stmfd sp!, @{fp, ip, lr, pc@}
7632 When performing a stack backtrace, code can inspect the value of
7633 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7634 location @code{pc - 12} and the top 8 bits are set, then we know that
7635 there is a function name embedded immediately preceding this location
7636 and has length @code{((pc[-3]) & 0xff000000)}.
7640 Generate code for the 16-bit Thumb instruction set. The default is to
7641 use the 32-bit ARM instruction set.
7644 @opindex mtpcs-frame
7645 Generate a stack frame that is compliant with the Thumb Procedure Call
7646 Standard for all non-leaf functions. (A leaf function is one that does
7647 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7649 @item -mtpcs-leaf-frame
7650 @opindex mtpcs-leaf-frame
7651 Generate a stack frame that is compliant with the Thumb Procedure Call
7652 Standard for all leaf functions. (A leaf function is one that does
7653 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7655 @item -mcallee-super-interworking
7656 @opindex mcallee-super-interworking
7657 Gives all externally visible functions in the file being compiled an ARM
7658 instruction set header which switches to Thumb mode before executing the
7659 rest of the function. This allows these functions to be called from
7660 non-interworking code.
7662 @item -mcaller-super-interworking
7663 @opindex mcaller-super-interworking
7664 Allows calls via function pointers (including virtual functions) to
7665 execute correctly regardless of whether the target code has been
7666 compiled for interworking or not. There is a small overhead in the cost
7667 of executing a function pointer if this option is enabled.
7669 @item -mtp=@var{name}
7671 Specify the access model for the thread local storage pointer. The valid
7672 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
7673 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
7674 (supported in the arm6k architecture), and @option{auto}, which uses the
7675 best available method for the selected processor. The default setting is
7681 @subsection AVR Options
7684 These options are defined for AVR implementations:
7687 @item -mmcu=@var{mcu}
7689 Specify ATMEL AVR instruction set or MCU type.
7691 Instruction set avr1 is for the minimal AVR core, not supported by the C
7692 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7693 attiny11, attiny12, attiny15, attiny28).
7695 Instruction set avr2 (default) is for the classic AVR core with up to
7696 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7697 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7698 at90c8534, at90s8535).
7700 Instruction set avr3 is for the classic AVR core with up to 128K program
7701 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7703 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7704 memory space (MCU types: atmega8, atmega83, atmega85).
7706 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7707 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7708 atmega64, atmega128, at43usb355, at94k).
7712 Output instruction sizes to the asm file.
7714 @item -minit-stack=@var{N}
7715 @opindex minit-stack
7716 Specify the initial stack address, which may be a symbol or numeric value,
7717 @samp{__stack} is the default.
7719 @item -mno-interrupts
7720 @opindex mno-interrupts
7721 Generated code is not compatible with hardware interrupts.
7722 Code size will be smaller.
7724 @item -mcall-prologues
7725 @opindex mcall-prologues
7726 Functions prologues/epilogues expanded as call to appropriate
7727 subroutines. Code size will be smaller.
7729 @item -mno-tablejump
7730 @opindex mno-tablejump
7731 Do not generate tablejump insns which sometimes increase code size.
7734 @opindex mtiny-stack
7735 Change only the low 8 bits of the stack pointer.
7739 Assume int to be 8 bit integer. This affects the sizes of all types: A
7740 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7741 and long long will be 4 bytes. Please note that this option does not
7742 comply to the C standards, but it will provide you with smaller code
7746 @node Blackfin Options
7747 @subsection Blackfin Options
7748 @cindex Blackfin Options
7751 @item -momit-leaf-frame-pointer
7752 @opindex momit-leaf-frame-pointer
7753 Don't keep the frame pointer in a register for leaf functions. This
7754 avoids the instructions to save, set up and restore frame pointers and
7755 makes an extra register available in leaf functions. The option
7756 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7757 which might make debugging harder.
7759 @item -mspecld-anomaly
7760 @opindex mspecld-anomaly
7761 When enabled, the compiler will ensure that the generated code does not
7762 contain speculative loads after jump instructions. This option is enabled
7765 @item -mno-specld-anomaly
7766 @opindex mno-specld-anomaly
7767 Don't generate extra code to prevent speculative loads from occurring.
7769 @item -mcsync-anomaly
7770 @opindex mcsync-anomaly
7771 When enabled, the compiler will ensure that the generated code does not
7772 contain CSYNC or SSYNC instructions too soon after conditional branches.
7773 This option is enabled by default.
7775 @item -mno-csync-anomaly
7776 @opindex mno-csync-anomaly
7777 Don't generate extra code to prevent CSYNC or SSYNC instructions from
7778 occurring too soon after a conditional branch.
7782 When enabled, the compiler is free to take advantage of the knowledge that
7783 the entire program fits into the low 64k of memory.
7786 @opindex mno-low-64k
7787 Assume that the program is arbitrarily large. This is the default.
7789 @item -mid-shared-library
7790 @opindex mid-shared-library
7791 Generate code that supports shared libraries via the library ID method.
7792 This allows for execute in place and shared libraries in an environment
7793 without virtual memory management. This option implies @option{-fPIC}.
7795 @item -mno-id-shared-library
7796 @opindex mno-id-shared-library
7797 Generate code that doesn't assume ID based shared libraries are being used.
7798 This is the default.
7800 @item -mshared-library-id=n
7801 @opindex mshared-library-id
7802 Specified the identification number of the ID based shared library being
7803 compiled. Specifying a value of 0 will generate more compact code, specifying
7804 other values will force the allocation of that number to the current
7805 library but is no more space or time efficient than omitting this option.
7808 @itemx -mno-long-calls
7809 @opindex mlong-calls
7810 @opindex mno-long-calls
7811 Tells the compiler to perform function calls by first loading the
7812 address of the function into a register and then performing a subroutine
7813 call on this register. This switch is needed if the target function
7814 will lie outside of the 24 bit addressing range of the offset based
7815 version of subroutine call instruction.
7817 This feature is not enabled by default. Specifying
7818 @option{-mno-long-calls} will restore the default behavior. Note these
7819 switches have no effect on how the compiler generates code to handle
7820 function calls via function pointers.
7824 @subsection CRIS Options
7825 @cindex CRIS Options
7827 These options are defined specifically for the CRIS ports.
7830 @item -march=@var{architecture-type}
7831 @itemx -mcpu=@var{architecture-type}
7834 Generate code for the specified architecture. The choices for
7835 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7836 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7837 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7840 @item -mtune=@var{architecture-type}
7842 Tune to @var{architecture-type} everything applicable about the generated
7843 code, except for the ABI and the set of available instructions. The
7844 choices for @var{architecture-type} are the same as for
7845 @option{-march=@var{architecture-type}}.
7847 @item -mmax-stack-frame=@var{n}
7848 @opindex mmax-stack-frame
7849 Warn when the stack frame of a function exceeds @var{n} bytes.
7851 @item -melinux-stacksize=@var{n}
7852 @opindex melinux-stacksize
7853 Only available with the @samp{cris-axis-aout} target. Arranges for
7854 indications in the program to the kernel loader that the stack of the
7855 program should be set to @var{n} bytes.
7861 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7862 @option{-march=v3} and @option{-march=v8} respectively.
7864 @item -mmul-bug-workaround
7865 @itemx -mno-mul-bug-workaround
7866 @opindex mmul-bug-workaround
7867 @opindex mno-mul-bug-workaround
7868 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7869 models where it applies. This option is active by default.
7873 Enable CRIS-specific verbose debug-related information in the assembly
7874 code. This option also has the effect to turn off the @samp{#NO_APP}
7875 formatted-code indicator to the assembler at the beginning of the
7880 Do not use condition-code results from previous instruction; always emit
7881 compare and test instructions before use of condition codes.
7883 @item -mno-side-effects
7884 @opindex mno-side-effects
7885 Do not emit instructions with side-effects in addressing modes other than
7889 @itemx -mno-stack-align
7891 @itemx -mno-data-align
7892 @itemx -mconst-align
7893 @itemx -mno-const-align
7894 @opindex mstack-align
7895 @opindex mno-stack-align
7896 @opindex mdata-align
7897 @opindex mno-data-align
7898 @opindex mconst-align
7899 @opindex mno-const-align
7900 These options (no-options) arranges (eliminate arrangements) for the
7901 stack-frame, individual data and constants to be aligned for the maximum
7902 single data access size for the chosen CPU model. The default is to
7903 arrange for 32-bit alignment. ABI details such as structure layout are
7904 not affected by these options.
7912 Similar to the stack- data- and const-align options above, these options
7913 arrange for stack-frame, writable data and constants to all be 32-bit,
7914 16-bit or 8-bit aligned. The default is 32-bit alignment.
7916 @item -mno-prologue-epilogue
7917 @itemx -mprologue-epilogue
7918 @opindex mno-prologue-epilogue
7919 @opindex mprologue-epilogue
7920 With @option{-mno-prologue-epilogue}, the normal function prologue and
7921 epilogue that sets up the stack-frame are omitted and no return
7922 instructions or return sequences are generated in the code. Use this
7923 option only together with visual inspection of the compiled code: no
7924 warnings or errors are generated when call-saved registers must be saved,
7925 or storage for local variable needs to be allocated.
7931 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7932 instruction sequences that load addresses for functions from the PLT part
7933 of the GOT rather than (traditional on other architectures) calls to the
7934 PLT@. The default is @option{-mgotplt}.
7938 Legacy no-op option only recognized with the cris-axis-aout target.
7942 Legacy no-op option only recognized with the cris-axis-elf and
7943 cris-axis-linux-gnu targets.
7947 Only recognized with the cris-axis-aout target, where it selects a
7948 GNU/linux-like multilib, include files and instruction set for
7953 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7957 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7958 to link with input-output functions from a simulator library. Code,
7959 initialized data and zero-initialized data are allocated consecutively.
7963 Like @option{-sim}, but pass linker options to locate initialized data at
7964 0x40000000 and zero-initialized data at 0x80000000.
7968 @subsection CRX Options
7971 These options are defined specifically for the CRX ports.
7977 Enable the use of multiply-accumulate instructions. Disabled by default.
7981 Push instructions will be used to pass outgoing arguments when functions
7982 are called. Enabled by default.
7985 @node Darwin Options
7986 @subsection Darwin Options
7987 @cindex Darwin options
7989 These options are defined for all architectures running the Darwin operating
7992 FSF GCC on Darwin does not create ``fat'' object files; it will create
7993 an object file for the single architecture that it was built to
7994 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7995 @option{-arch} options are used; it does so by running the compiler or
7996 linker multiple times and joining the results together with
7999 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
8000 @samp{i686}) is determined by the flags that specify the ISA
8001 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
8002 @option{-force_cpusubtype_ALL} option can be used to override this.
8004 The Darwin tools vary in their behavior when presented with an ISA
8005 mismatch. The assembler, @file{as}, will only permit instructions to
8006 be used that are valid for the subtype of the file it is generating,
8007 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
8008 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
8009 and print an error if asked to create a shared library with a less
8010 restrictive subtype than its input files (for instance, trying to put
8011 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
8012 for executables, @file{ld}, will quietly give the executable the most
8013 restrictive subtype of any of its input files.
8018 Add the framework directory @var{dir} to the head of the list of
8019 directories to be searched for header files. These directories are
8020 interleaved with those specified by @option{-I} options and are
8021 scanned in a left-to-right order.
8023 A framework directory is a directory with frameworks in it. A
8024 framework is a directory with a @samp{"Headers"} and/or
8025 @samp{"PrivateHeaders"} directory contained directly in it that ends
8026 in @samp{".framework"}. The name of a framework is the name of this
8027 directory excluding the @samp{".framework"}. Headers associated with
8028 the framework are found in one of those two directories, with
8029 @samp{"Headers"} being searched first. A subframework is a framework
8030 directory that is in a framework's @samp{"Frameworks"} directory.
8031 Includes of subframework headers can only appear in a header of a
8032 framework that contains the subframework, or in a sibling subframework
8033 header. Two subframeworks are siblings if they occur in the same
8034 framework. A subframework should not have the same name as a
8035 framework, a warning will be issued if this is violated. Currently a
8036 subframework cannot have subframeworks, in the future, the mechanism
8037 may be extended to support this. The standard frameworks can be found
8038 in @samp{"/System/Library/Frameworks"} and
8039 @samp{"/Library/Frameworks"}. An example include looks like
8040 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
8041 the name of the framework and header.h is found in the
8042 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
8046 Emit debugging information for symbols that are used. For STABS
8047 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
8048 This is by default ON@.
8052 Emit debugging information for all symbols and types.
8054 @item -mmacosx-version-min=@var{version}
8055 The earliest version of MacOS X that this executable will run on
8056 is @var{version}. Typical values of @var{version} include @code{10.1},
8057 @code{10.2}, and @code{10.3.9}.
8059 The default for this option is to make choices that seem to be most
8064 Enable kernel development mode. The @option{-mkernel} option sets
8065 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
8066 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
8067 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
8068 applicable. This mode also sets @option{-mno-altivec},
8069 @option{-msoft-float}, @option{-fno-builtin} and
8070 @option{-mlong-branch} for PowerPC targets.
8072 @item -mone-byte-bool
8073 @opindex mone-byte-bool
8074 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
8075 By default @samp{sizeof(bool)} is @samp{4} when compiling for
8076 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
8077 option has no effect on x86.
8079 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
8080 to generate code that is not binary compatible with code generated
8081 without that switch. Using this switch may require recompiling all
8082 other modules in a program, including system libraries. Use this
8083 switch to conform to a non-default data model.
8085 @item -mfix-and-continue
8086 @itemx -ffix-and-continue
8087 @itemx -findirect-data
8088 @opindex mfix-and-continue
8089 @opindex ffix-and-continue
8090 @opindex findirect-data
8091 Generate code suitable for fast turn around development. Needed to
8092 enable gdb to dynamically load @code{.o} files into already running
8093 programs. @option{-findirect-data} and @option{-ffix-and-continue}
8094 are provided for backwards compatibility.
8098 Loads all members of static archive libraries.
8099 See man ld(1) for more information.
8101 @item -arch_errors_fatal
8102 @opindex arch_errors_fatal
8103 Cause the errors having to do with files that have the wrong architecture
8107 @opindex bind_at_load
8108 Causes the output file to be marked such that the dynamic linker will
8109 bind all undefined references when the file is loaded or launched.
8113 Produce a Mach-o bundle format file.
8114 See man ld(1) for more information.
8116 @item -bundle_loader @var{executable}
8117 @opindex bundle_loader
8118 This option specifies the @var{executable} that will be loading the build
8119 output file being linked. See man ld(1) for more information.
8123 When passed this option, GCC will produce a dynamic library instead of
8124 an executable when linking, using the Darwin @file{libtool} command.
8126 @item -force_cpusubtype_ALL
8127 @opindex force_cpusubtype_ALL
8128 This causes GCC's output file to have the @var{ALL} subtype, instead of
8129 one controlled by the @option{-mcpu} or @option{-march} option.
8131 @item -allowable_client @var{client_name}
8133 @itemx -compatibility_version
8134 @itemx -current_version
8136 @itemx -dependency-file
8138 @itemx -dylinker_install_name
8140 @itemx -exported_symbols_list
8142 @itemx -flat_namespace
8143 @itemx -force_flat_namespace
8144 @itemx -headerpad_max_install_names
8147 @itemx -install_name
8148 @itemx -keep_private_externs
8149 @itemx -multi_module
8150 @itemx -multiply_defined
8151 @itemx -multiply_defined_unused
8153 @itemx -no_dead_strip_inits_and_terms
8154 @itemx -nofixprebinding
8157 @itemx -noseglinkedit
8158 @itemx -pagezero_size
8160 @itemx -prebind_all_twolevel_modules
8161 @itemx -private_bundle
8162 @itemx -read_only_relocs
8164 @itemx -sectobjectsymbols
8168 @itemx -sectobjectsymbols
8171 @itemx -segs_read_only_addr
8172 @itemx -segs_read_write_addr
8173 @itemx -seg_addr_table
8174 @itemx -seg_addr_table_filename
8177 @itemx -segs_read_only_addr
8178 @itemx -segs_read_write_addr
8179 @itemx -single_module
8182 @itemx -sub_umbrella
8183 @itemx -twolevel_namespace
8186 @itemx -unexported_symbols_list
8187 @itemx -weak_reference_mismatches
8190 @opindex allowable_client
8191 @opindex client_name
8192 @opindex compatibility_version
8193 @opindex current_version
8195 @opindex dependency-file
8197 @opindex dylinker_install_name
8199 @opindex exported_symbols_list
8201 @opindex flat_namespace
8202 @opindex force_flat_namespace
8203 @opindex headerpad_max_install_names
8206 @opindex install_name
8207 @opindex keep_private_externs
8208 @opindex multi_module
8209 @opindex multiply_defined
8210 @opindex multiply_defined_unused
8212 @opindex no_dead_strip_inits_and_terms
8213 @opindex nofixprebinding
8214 @opindex nomultidefs
8216 @opindex noseglinkedit
8217 @opindex pagezero_size
8219 @opindex prebind_all_twolevel_modules
8220 @opindex private_bundle
8221 @opindex read_only_relocs
8223 @opindex sectobjectsymbols
8227 @opindex sectobjectsymbols
8230 @opindex segs_read_only_addr
8231 @opindex segs_read_write_addr
8232 @opindex seg_addr_table
8233 @opindex seg_addr_table_filename
8234 @opindex seglinkedit
8236 @opindex segs_read_only_addr
8237 @opindex segs_read_write_addr
8238 @opindex single_module
8240 @opindex sub_library
8241 @opindex sub_umbrella
8242 @opindex twolevel_namespace
8245 @opindex unexported_symbols_list
8246 @opindex weak_reference_mismatches
8247 @opindex whatsloaded
8249 These options are passed to the Darwin linker. The Darwin linker man page
8250 describes them in detail.
8253 @node DEC Alpha Options
8254 @subsection DEC Alpha Options
8256 These @samp{-m} options are defined for the DEC Alpha implementations:
8259 @item -mno-soft-float
8261 @opindex mno-soft-float
8262 @opindex msoft-float
8263 Use (do not use) the hardware floating-point instructions for
8264 floating-point operations. When @option{-msoft-float} is specified,
8265 functions in @file{libgcc.a} will be used to perform floating-point
8266 operations. Unless they are replaced by routines that emulate the
8267 floating-point operations, or compiled in such a way as to call such
8268 emulations routines, these routines will issue floating-point
8269 operations. If you are compiling for an Alpha without floating-point
8270 operations, you must ensure that the library is built so as not to call
8273 Note that Alpha implementations without floating-point operations are
8274 required to have floating-point registers.
8279 @opindex mno-fp-regs
8280 Generate code that uses (does not use) the floating-point register set.
8281 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8282 register set is not used, floating point operands are passed in integer
8283 registers as if they were integers and floating-point results are passed
8284 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8285 so any function with a floating-point argument or return value called by code
8286 compiled with @option{-mno-fp-regs} must also be compiled with that
8289 A typical use of this option is building a kernel that does not use,
8290 and hence need not save and restore, any floating-point registers.
8294 The Alpha architecture implements floating-point hardware optimized for
8295 maximum performance. It is mostly compliant with the IEEE floating
8296 point standard. However, for full compliance, software assistance is
8297 required. This option generates code fully IEEE compliant code
8298 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8299 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8300 defined during compilation. The resulting code is less efficient but is
8301 able to correctly support denormalized numbers and exceptional IEEE
8302 values such as not-a-number and plus/minus infinity. Other Alpha
8303 compilers call this option @option{-ieee_with_no_inexact}.
8305 @item -mieee-with-inexact
8306 @opindex mieee-with-inexact
8307 This is like @option{-mieee} except the generated code also maintains
8308 the IEEE @var{inexact-flag}. Turning on this option causes the
8309 generated code to implement fully-compliant IEEE math. In addition to
8310 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8311 macro. On some Alpha implementations the resulting code may execute
8312 significantly slower than the code generated by default. Since there is
8313 very little code that depends on the @var{inexact-flag}, you should
8314 normally not specify this option. Other Alpha compilers call this
8315 option @option{-ieee_with_inexact}.
8317 @item -mfp-trap-mode=@var{trap-mode}
8318 @opindex mfp-trap-mode
8319 This option controls what floating-point related traps are enabled.
8320 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8321 The trap mode can be set to one of four values:
8325 This is the default (normal) setting. The only traps that are enabled
8326 are the ones that cannot be disabled in software (e.g., division by zero
8330 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8334 Like @samp{u}, but the instructions are marked to be safe for software
8335 completion (see Alpha architecture manual for details).
8338 Like @samp{su}, but inexact traps are enabled as well.
8341 @item -mfp-rounding-mode=@var{rounding-mode}
8342 @opindex mfp-rounding-mode
8343 Selects the IEEE rounding mode. Other Alpha compilers call this option
8344 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8349 Normal IEEE rounding mode. Floating point numbers are rounded towards
8350 the nearest machine number or towards the even machine number in case
8354 Round towards minus infinity.
8357 Chopped rounding mode. Floating point numbers are rounded towards zero.
8360 Dynamic rounding mode. A field in the floating point control register
8361 (@var{fpcr}, see Alpha architecture reference manual) controls the
8362 rounding mode in effect. The C library initializes this register for
8363 rounding towards plus infinity. Thus, unless your program modifies the
8364 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8367 @item -mtrap-precision=@var{trap-precision}
8368 @opindex mtrap-precision
8369 In the Alpha architecture, floating point traps are imprecise. This
8370 means without software assistance it is impossible to recover from a
8371 floating trap and program execution normally needs to be terminated.
8372 GCC can generate code that can assist operating system trap handlers
8373 in determining the exact location that caused a floating point trap.
8374 Depending on the requirements of an application, different levels of
8375 precisions can be selected:
8379 Program precision. This option is the default and means a trap handler
8380 can only identify which program caused a floating point exception.
8383 Function precision. The trap handler can determine the function that
8384 caused a floating point exception.
8387 Instruction precision. The trap handler can determine the exact
8388 instruction that caused a floating point exception.
8391 Other Alpha compilers provide the equivalent options called
8392 @option{-scope_safe} and @option{-resumption_safe}.
8394 @item -mieee-conformant
8395 @opindex mieee-conformant
8396 This option marks the generated code as IEEE conformant. You must not
8397 use this option unless you also specify @option{-mtrap-precision=i} and either
8398 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8399 is to emit the line @samp{.eflag 48} in the function prologue of the
8400 generated assembly file. Under DEC Unix, this has the effect that
8401 IEEE-conformant math library routines will be linked in.
8403 @item -mbuild-constants
8404 @opindex mbuild-constants
8405 Normally GCC examines a 32- or 64-bit integer constant to
8406 see if it can construct it from smaller constants in two or three
8407 instructions. If it cannot, it will output the constant as a literal and
8408 generate code to load it from the data segment at runtime.
8410 Use this option to require GCC to construct @emph{all} integer constants
8411 using code, even if it takes more instructions (the maximum is six).
8413 You would typically use this option to build a shared library dynamic
8414 loader. Itself a shared library, it must relocate itself in memory
8415 before it can find the variables and constants in its own data segment.
8421 Select whether to generate code to be assembled by the vendor-supplied
8422 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8440 Indicate whether GCC should generate code to use the optional BWX,
8441 CIX, FIX and MAX instruction sets. The default is to use the instruction
8442 sets supported by the CPU type specified via @option{-mcpu=} option or that
8443 of the CPU on which GCC was built if none was specified.
8448 @opindex mfloat-ieee
8449 Generate code that uses (does not use) VAX F and G floating point
8450 arithmetic instead of IEEE single and double precision.
8452 @item -mexplicit-relocs
8453 @itemx -mno-explicit-relocs
8454 @opindex mexplicit-relocs
8455 @opindex mno-explicit-relocs
8456 Older Alpha assemblers provided no way to generate symbol relocations
8457 except via assembler macros. Use of these macros does not allow
8458 optimal instruction scheduling. GNU binutils as of version 2.12
8459 supports a new syntax that allows the compiler to explicitly mark
8460 which relocations should apply to which instructions. This option
8461 is mostly useful for debugging, as GCC detects the capabilities of
8462 the assembler when it is built and sets the default accordingly.
8466 @opindex msmall-data
8467 @opindex mlarge-data
8468 When @option{-mexplicit-relocs} is in effect, static data is
8469 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8470 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8471 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8472 16-bit relocations off of the @code{$gp} register. This limits the
8473 size of the small data area to 64KB, but allows the variables to be
8474 directly accessed via a single instruction.
8476 The default is @option{-mlarge-data}. With this option the data area
8477 is limited to just below 2GB@. Programs that require more than 2GB of
8478 data must use @code{malloc} or @code{mmap} to allocate the data in the
8479 heap instead of in the program's data segment.
8481 When generating code for shared libraries, @option{-fpic} implies
8482 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8486 @opindex msmall-text
8487 @opindex mlarge-text
8488 When @option{-msmall-text} is used, the compiler assumes that the
8489 code of the entire program (or shared library) fits in 4MB, and is
8490 thus reachable with a branch instruction. When @option{-msmall-data}
8491 is used, the compiler can assume that all local symbols share the
8492 same @code{$gp} value, and thus reduce the number of instructions
8493 required for a function call from 4 to 1.
8495 The default is @option{-mlarge-text}.
8497 @item -mcpu=@var{cpu_type}
8499 Set the instruction set and instruction scheduling parameters for
8500 machine type @var{cpu_type}. You can specify either the @samp{EV}
8501 style name or the corresponding chip number. GCC supports scheduling
8502 parameters for the EV4, EV5 and EV6 family of processors and will
8503 choose the default values for the instruction set from the processor
8504 you specify. If you do not specify a processor type, GCC will default
8505 to the processor on which the compiler was built.
8507 Supported values for @var{cpu_type} are
8513 Schedules as an EV4 and has no instruction set extensions.
8517 Schedules as an EV5 and has no instruction set extensions.
8521 Schedules as an EV5 and supports the BWX extension.
8526 Schedules as an EV5 and supports the BWX and MAX extensions.
8530 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8534 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8537 @item -mtune=@var{cpu_type}
8539 Set only the instruction scheduling parameters for machine type
8540 @var{cpu_type}. The instruction set is not changed.
8542 @item -mmemory-latency=@var{time}
8543 @opindex mmemory-latency
8544 Sets the latency the scheduler should assume for typical memory
8545 references as seen by the application. This number is highly
8546 dependent on the memory access patterns used by the application
8547 and the size of the external cache on the machine.
8549 Valid options for @var{time} are
8553 A decimal number representing clock cycles.
8559 The compiler contains estimates of the number of clock cycles for
8560 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8561 (also called Dcache, Scache, and Bcache), as well as to main memory.
8562 Note that L3 is only valid for EV5.
8567 @node DEC Alpha/VMS Options
8568 @subsection DEC Alpha/VMS Options
8570 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8573 @item -mvms-return-codes
8574 @opindex mvms-return-codes
8575 Return VMS condition codes from main. The default is to return POSIX
8576 style condition (e.g.@ error) codes.
8580 @subsection FRV Options
8587 Only use the first 32 general purpose registers.
8592 Use all 64 general purpose registers.
8597 Use only the first 32 floating point registers.
8602 Use all 64 floating point registers
8605 @opindex mhard-float
8607 Use hardware instructions for floating point operations.
8610 @opindex msoft-float
8612 Use library routines for floating point operations.
8617 Dynamically allocate condition code registers.
8622 Do not try to dynamically allocate condition code registers, only
8623 use @code{icc0} and @code{fcc0}.
8628 Change ABI to use double word insns.
8633 Do not use double word instructions.
8638 Use floating point double instructions.
8643 Do not use floating point double instructions.
8648 Use media instructions.
8653 Do not use media instructions.
8658 Use multiply and add/subtract instructions.
8663 Do not use multiply and add/subtract instructions.
8668 Select the FDPIC ABI, that uses function descriptors to represent
8669 pointers to functions. Without any PIC/PIE-related options, it
8670 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8671 assumes GOT entries and small data are within a 12-bit range from the
8672 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8673 are computed with 32 bits.
8676 @opindex minline-plt
8678 Enable inlining of PLT entries in function calls to functions that are
8679 not known to bind locally. It has no effect without @option{-mfdpic}.
8680 It's enabled by default if optimizing for speed and compiling for
8681 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8682 optimization option such as @option{-O3} or above is present in the
8688 Assume a large TLS segment when generating thread-local code.
8693 Do not assume a large TLS segment when generating thread-local code.
8698 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8699 that is known to be in read-only sections. It's enabled by default,
8700 except for @option{-fpic} or @option{-fpie}: even though it may help
8701 make the global offset table smaller, it trades 1 instruction for 4.
8702 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8703 one of which may be shared by multiple symbols, and it avoids the need
8704 for a GOT entry for the referenced symbol, so it's more likely to be a
8705 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8707 @item -multilib-library-pic
8708 @opindex multilib-library-pic
8710 Link with the (library, not FD) pic libraries. It's implied by
8711 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8712 @option{-fpic} without @option{-mfdpic}. You should never have to use
8718 Follow the EABI requirement of always creating a frame pointer whenever
8719 a stack frame is allocated. This option is enabled by default and can
8720 be disabled with @option{-mno-linked-fp}.
8723 @opindex mlong-calls
8725 Use indirect addressing to call functions outside the current
8726 compilation unit. This allows the functions to be placed anywhere
8727 within the 32-bit address space.
8729 @item -malign-labels
8730 @opindex malign-labels
8732 Try to align labels to an 8-byte boundary by inserting nops into the
8733 previous packet. This option only has an effect when VLIW packing
8734 is enabled. It doesn't create new packets; it merely adds nops to
8738 @opindex mlibrary-pic
8740 Generate position-independent EABI code.
8745 Use only the first four media accumulator registers.
8750 Use all eight media accumulator registers.
8755 Pack VLIW instructions.
8760 Do not pack VLIW instructions.
8765 Do not mark ABI switches in e_flags.
8770 Enable the use of conditional-move instructions (default).
8772 This switch is mainly for debugging the compiler and will likely be removed
8773 in a future version.
8775 @item -mno-cond-move
8776 @opindex mno-cond-move
8778 Disable the use of conditional-move instructions.
8780 This switch is mainly for debugging the compiler and will likely be removed
8781 in a future version.
8786 Enable the use of conditional set instructions (default).
8788 This switch is mainly for debugging the compiler and will likely be removed
8789 in a future version.
8794 Disable the use of conditional set instructions.
8796 This switch is mainly for debugging the compiler and will likely be removed
8797 in a future version.
8802 Enable the use of conditional execution (default).
8804 This switch is mainly for debugging the compiler and will likely be removed
8805 in a future version.
8807 @item -mno-cond-exec
8808 @opindex mno-cond-exec
8810 Disable the use of conditional execution.
8812 This switch is mainly for debugging the compiler and will likely be removed
8813 in a future version.
8816 @opindex mvliw-branch
8818 Run a pass to pack branches into VLIW instructions (default).
8820 This switch is mainly for debugging the compiler and will likely be removed
8821 in a future version.
8823 @item -mno-vliw-branch
8824 @opindex mno-vliw-branch
8826 Do not run a pass to pack branches into VLIW instructions.
8828 This switch is mainly for debugging the compiler and will likely be removed
8829 in a future version.
8831 @item -mmulti-cond-exec
8832 @opindex mmulti-cond-exec
8834 Enable optimization of @code{&&} and @code{||} in conditional execution
8837 This switch is mainly for debugging the compiler and will likely be removed
8838 in a future version.
8840 @item -mno-multi-cond-exec
8841 @opindex mno-multi-cond-exec
8843 Disable optimization of @code{&&} and @code{||} in conditional execution.
8845 This switch is mainly for debugging the compiler and will likely be removed
8846 in a future version.
8848 @item -mnested-cond-exec
8849 @opindex mnested-cond-exec
8851 Enable nested conditional execution optimizations (default).
8853 This switch is mainly for debugging the compiler and will likely be removed
8854 in a future version.
8856 @item -mno-nested-cond-exec
8857 @opindex mno-nested-cond-exec
8859 Disable nested conditional execution optimizations.
8861 This switch is mainly for debugging the compiler and will likely be removed
8862 in a future version.
8864 @item -moptimize-membar
8865 @opindex moptimize-membar
8867 This switch removes redundant @code{membar} instructions from the
8868 compiler generated code. It is enabled by default.
8870 @item -mno-optimize-membar
8871 @opindex mno-optimize-membar
8873 This switch disables the automatic removal of redundant @code{membar}
8874 instructions from the generated code.
8876 @item -mtomcat-stats
8877 @opindex mtomcat-stats
8879 Cause gas to print out tomcat statistics.
8881 @item -mcpu=@var{cpu}
8884 Select the processor type for which to generate code. Possible values are
8885 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8886 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8890 @node GNU/Linux Options
8891 @subsection GNU/Linux Options
8893 These @samp{-m} options are defined for GNU/Linux targets:
8898 Use the GNU C library instead of uClibc. This is the default except
8899 on @samp{*-*-linux-*uclibc*} targets.
8903 Use uClibc instead of the GNU C library. This is the default on
8904 @samp{*-*-linux-*uclibc*} targets.
8907 @node H8/300 Options
8908 @subsection H8/300 Options
8910 These @samp{-m} options are defined for the H8/300 implementations:
8915 Shorten some address references at link time, when possible; uses the
8916 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8917 ld, Using ld}, for a fuller description.
8921 Generate code for the H8/300H@.
8925 Generate code for the H8S@.
8929 Generate code for the H8S and H8/300H in the normal mode. This switch
8930 must be used either with @option{-mh} or @option{-ms}.
8934 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8938 Make @code{int} data 32 bits by default.
8942 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8943 The default for the H8/300H and H8S is to align longs and floats on 4
8945 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8946 This option has no effect on the H8/300.
8950 @subsection HPPA Options
8951 @cindex HPPA Options
8953 These @samp{-m} options are defined for the HPPA family of computers:
8956 @item -march=@var{architecture-type}
8958 Generate code for the specified architecture. The choices for
8959 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8960 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8961 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8962 architecture option for your machine. Code compiled for lower numbered
8963 architectures will run on higher numbered architectures, but not the
8967 @itemx -mpa-risc-1-1
8968 @itemx -mpa-risc-2-0
8969 @opindex mpa-risc-1-0
8970 @opindex mpa-risc-1-1
8971 @opindex mpa-risc-2-0
8972 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8975 @opindex mbig-switch
8976 Generate code suitable for big switch tables. Use this option only if
8977 the assembler/linker complain about out of range branches within a switch
8980 @item -mjump-in-delay
8981 @opindex mjump-in-delay
8982 Fill delay slots of function calls with unconditional jump instructions
8983 by modifying the return pointer for the function call to be the target
8984 of the conditional jump.
8986 @item -mdisable-fpregs
8987 @opindex mdisable-fpregs
8988 Prevent floating point registers from being used in any manner. This is
8989 necessary for compiling kernels which perform lazy context switching of
8990 floating point registers. If you use this option and attempt to perform
8991 floating point operations, the compiler will abort.
8993 @item -mdisable-indexing
8994 @opindex mdisable-indexing
8995 Prevent the compiler from using indexing address modes. This avoids some
8996 rather obscure problems when compiling MIG generated code under MACH@.
8998 @item -mno-space-regs
8999 @opindex mno-space-regs
9000 Generate code that assumes the target has no space registers. This allows
9001 GCC to generate faster indirect calls and use unscaled index address modes.
9003 Such code is suitable for level 0 PA systems and kernels.
9005 @item -mfast-indirect-calls
9006 @opindex mfast-indirect-calls
9007 Generate code that assumes calls never cross space boundaries. This
9008 allows GCC to emit code which performs faster indirect calls.
9010 This option will not work in the presence of shared libraries or nested
9013 @item -mfixed-range=@var{register-range}
9014 @opindex mfixed-range
9015 Generate code treating the given register range as fixed registers.
9016 A fixed register is one that the register allocator can not use. This is
9017 useful when compiling kernel code. A register range is specified as
9018 two registers separated by a dash. Multiple register ranges can be
9019 specified separated by a comma.
9021 @item -mlong-load-store
9022 @opindex mlong-load-store
9023 Generate 3-instruction load and store sequences as sometimes required by
9024 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
9027 @item -mportable-runtime
9028 @opindex mportable-runtime
9029 Use the portable calling conventions proposed by HP for ELF systems.
9033 Enable the use of assembler directives only GAS understands.
9035 @item -mschedule=@var{cpu-type}
9037 Schedule code according to the constraints for the machine type
9038 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
9039 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
9040 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
9041 proper scheduling option for your machine. The default scheduling is
9045 @opindex mlinker-opt
9046 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
9047 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
9048 linkers in which they give bogus error messages when linking some programs.
9051 @opindex msoft-float
9052 Generate output containing library calls for floating point.
9053 @strong{Warning:} the requisite libraries are not available for all HPPA
9054 targets. Normally the facilities of the machine's usual C compiler are
9055 used, but this cannot be done directly in cross-compilation. You must make
9056 your own arrangements to provide suitable library functions for
9057 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
9058 does provide software floating point support.
9060 @option{-msoft-float} changes the calling convention in the output file;
9061 therefore, it is only useful if you compile @emph{all} of a program with
9062 this option. In particular, you need to compile @file{libgcc.a}, the
9063 library that comes with GCC, with @option{-msoft-float} in order for
9068 Generate the predefine, @code{_SIO}, for server IO@. The default is
9069 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
9070 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
9071 options are available under HP-UX and HI-UX@.
9075 Use GNU ld specific options. This passes @option{-shared} to ld when
9076 building a shared library. It is the default when GCC is configured,
9077 explicitly or implicitly, with the GNU linker. This option does not
9078 have any affect on which ld is called, it only changes what parameters
9079 are passed to that ld. The ld that is called is determined by the
9080 @option{--with-ld} configure option, GCC's program search path, and
9081 finally by the user's @env{PATH}. The linker used by GCC can be printed
9082 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
9083 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9087 Use HP ld specific options. This passes @option{-b} to ld when building
9088 a shared library and passes @option{+Accept TypeMismatch} to ld on all
9089 links. It is the default when GCC is configured, explicitly or
9090 implicitly, with the HP linker. This option does not have any affect on
9091 which ld is called, it only changes what parameters are passed to that
9092 ld. The ld that is called is determined by the @option{--with-ld}
9093 configure option, GCC's program search path, and finally by the user's
9094 @env{PATH}. The linker used by GCC can be printed using @samp{which
9095 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
9096 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9099 @opindex mno-long-calls
9100 Generate code that uses long call sequences. This ensures that a call
9101 is always able to reach linker generated stubs. The default is to generate
9102 long calls only when the distance from the call site to the beginning
9103 of the function or translation unit, as the case may be, exceeds a
9104 predefined limit set by the branch type being used. The limits for
9105 normal calls are 7,600,000 and 240,000 bytes, respectively for the
9106 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
9109 Distances are measured from the beginning of functions when using the
9110 @option{-ffunction-sections} option, or when using the @option{-mgas}
9111 and @option{-mno-portable-runtime} options together under HP-UX with
9114 It is normally not desirable to use this option as it will degrade
9115 performance. However, it may be useful in large applications,
9116 particularly when partial linking is used to build the application.
9118 The types of long calls used depends on the capabilities of the
9119 assembler and linker, and the type of code being generated. The
9120 impact on systems that support long absolute calls, and long pic
9121 symbol-difference or pc-relative calls should be relatively small.
9122 However, an indirect call is used on 32-bit ELF systems in pic code
9123 and it is quite long.
9125 @item -munix=@var{unix-std}
9127 Generate compiler predefines and select a startfile for the specified
9128 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
9129 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
9130 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
9131 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
9132 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
9135 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
9136 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
9137 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
9138 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
9139 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
9140 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
9142 It is @emph{important} to note that this option changes the interfaces
9143 for various library routines. It also affects the operational behavior
9144 of the C library. Thus, @emph{extreme} care is needed in using this
9147 Library code that is intended to operate with more than one UNIX
9148 standard must test, set and restore the variable @var{__xpg4_extended_mask}
9149 as appropriate. Most GNU software doesn't provide this capability.
9153 Suppress the generation of link options to search libdld.sl when the
9154 @option{-static} option is specified on HP-UX 10 and later.
9158 The HP-UX implementation of setlocale in libc has a dependency on
9159 libdld.sl. There isn't an archive version of libdld.sl. Thus,
9160 when the @option{-static} option is specified, special link options
9161 are needed to resolve this dependency.
9163 On HP-UX 10 and later, the GCC driver adds the necessary options to
9164 link with libdld.sl when the @option{-static} option is specified.
9165 This causes the resulting binary to be dynamic. On the 64-bit port,
9166 the linkers generate dynamic binaries by default in any case. The
9167 @option{-nolibdld} option can be used to prevent the GCC driver from
9168 adding these link options.
9172 Add support for multithreading with the @dfn{dce thread} library
9173 under HP-UX@. This option sets flags for both the preprocessor and
9177 @node i386 and x86-64 Options
9178 @subsection Intel 386 and AMD x86-64 Options
9179 @cindex i386 Options
9180 @cindex x86-64 Options
9181 @cindex Intel 386 Options
9182 @cindex AMD x86-64 Options
9184 These @samp{-m} options are defined for the i386 and x86-64 family of
9188 @item -mtune=@var{cpu-type}
9190 Tune to @var{cpu-type} everything applicable about the generated code, except
9191 for the ABI and the set of available instructions. The choices for
9195 Produce code optimized for the most common IA32/AMD64/EM64T processors.
9196 If you know the CPU on which your code will run, then you should use
9197 the corresponding @option{-mtune} option instead of
9198 @option{-mtune=generic}. But, if you do not know exactly what CPU users
9199 of your application will have, then you should use this option.
9201 As new processors are deployed in the marketplace, the behavior of this
9202 option will change. Therefore, if you upgrade to a newer version of
9203 GCC, the code generated option will change to reflect the processors
9204 that were most common when that version of GCC was released.
9206 There is no @option{-march=generic} option because @option{-march}
9207 indicates the instruction set the compiler can use, and there is no
9208 generic instruction set applicable to all processors. In contrast,
9209 @option{-mtune} indicates the processor (or, in this case, collection of
9210 processors) for which the code is optimized.
9212 This selects the CPU to tune for at compilation time by determining
9213 the processor type of the compiling machine. Using @option{-mtune=native}
9214 will produce code optimized for the local machine under the constraints
9215 of the selected instruction set. Using @option{-march=native} will
9216 enable all instruction subsets supported by the local machine (hence
9217 the result might not run on different machines).
9219 Original Intel's i386 CPU@.
9221 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
9223 Intel Pentium CPU with no MMX support.
9225 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
9227 Intel PentiumPro CPU@.
9229 Same as @code{generic}, but when used as @code{march} option, PentiumPro
9230 instruction set will be used, so the code will run on all i686 family chips.
9232 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
9233 @item pentium3, pentium3m
9234 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
9237 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
9238 support. Used by Centrino notebooks.
9239 @item pentium4, pentium4m
9240 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
9242 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
9245 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
9246 SSE2 and SSE3 instruction set support.
9248 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
9249 instruction set support.
9251 AMD K6 CPU with MMX instruction set support.
9253 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
9254 @item athlon, athlon-tbird
9255 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
9257 @item athlon-4, athlon-xp, athlon-mp
9258 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
9259 instruction set support.
9260 @item k8, opteron, athlon64, athlon-fx
9261 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
9262 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
9263 @item k8-sse3, opteron-sse3, athlon64-sse3
9264 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
9265 @item amdfam10, barcelona
9266 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
9267 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
9268 instruction set extensions.)
9270 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
9273 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
9274 instruction set support.
9276 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
9277 implemented for this chip.)
9279 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
9280 implemented for this chip.)
9282 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
9285 While picking a specific @var{cpu-type} will schedule things appropriately
9286 for that particular chip, the compiler will not generate any code that
9287 does not run on the i386 without the @option{-march=@var{cpu-type}} option
9290 @item -march=@var{cpu-type}
9292 Generate instructions for the machine type @var{cpu-type}. The choices
9293 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
9294 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
9296 @item -mcpu=@var{cpu-type}
9298 A deprecated synonym for @option{-mtune}.
9307 @opindex mpentiumpro
9308 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
9309 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
9310 These synonyms are deprecated.
9312 @item -mfpmath=@var{unit}
9314 Generate floating point arithmetics for selected unit @var{unit}. The choices
9319 Use the standard 387 floating point coprocessor present majority of chips and
9320 emulated otherwise. Code compiled with this option will run almost everywhere.
9321 The temporary results are computed in 80bit precision instead of precision
9322 specified by the type resulting in slightly different results compared to most
9323 of other chips. See @option{-ffloat-store} for more detailed description.
9325 This is the default choice for i386 compiler.
9328 Use scalar floating point instructions present in the SSE instruction set.
9329 This instruction set is supported by Pentium3 and newer chips, in the AMD line
9330 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
9331 instruction set supports only single precision arithmetics, thus the double and
9332 extended precision arithmetics is still done using 387. Later version, present
9333 only in Pentium4 and the future AMD x86-64 chips supports double precision
9336 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
9337 or @option{-msse2} switches to enable SSE extensions and make this option
9338 effective. For the x86-64 compiler, these extensions are enabled by default.
9340 The resulting code should be considerably faster in the majority of cases and avoid
9341 the numerical instability problems of 387 code, but may break some existing
9342 code that expects temporaries to be 80bit.
9344 This is the default choice for the x86-64 compiler.
9347 Attempt to utilize both instruction sets at once. This effectively double the
9348 amount of available registers and on chips with separate execution units for
9349 387 and SSE the execution resources too. Use this option with care, as it is
9350 still experimental, because the GCC register allocator does not model separate
9351 functional units well resulting in instable performance.
9354 @item -masm=@var{dialect}
9355 @opindex masm=@var{dialect}
9356 Output asm instructions using selected @var{dialect}. Supported
9357 choices are @samp{intel} or @samp{att} (the default one). Darwin does
9358 not support @samp{intel}.
9363 @opindex mno-ieee-fp
9364 Control whether or not the compiler uses IEEE floating point
9365 comparisons. These handle correctly the case where the result of a
9366 comparison is unordered.
9369 @opindex msoft-float
9370 Generate output containing library calls for floating point.
9371 @strong{Warning:} the requisite libraries are not part of GCC@.
9372 Normally the facilities of the machine's usual C compiler are used, but
9373 this can't be done directly in cross-compilation. You must make your
9374 own arrangements to provide suitable library functions for
9377 On machines where a function returns floating point results in the 80387
9378 register stack, some floating point opcodes may be emitted even if
9379 @option{-msoft-float} is used.
9381 @item -mno-fp-ret-in-387
9382 @opindex mno-fp-ret-in-387
9383 Do not use the FPU registers for return values of functions.
9385 The usual calling convention has functions return values of types
9386 @code{float} and @code{double} in an FPU register, even if there
9387 is no FPU@. The idea is that the operating system should emulate
9390 The option @option{-mno-fp-ret-in-387} causes such values to be returned
9391 in ordinary CPU registers instead.
9393 @item -mno-fancy-math-387
9394 @opindex mno-fancy-math-387
9395 Some 387 emulators do not support the @code{sin}, @code{cos} and
9396 @code{sqrt} instructions for the 387. Specify this option to avoid
9397 generating those instructions. This option is the default on
9398 OpenBSD and NetBSD@. This option is overridden when @option{-march}
9399 indicates that the target cpu will always have an FPU and so the
9400 instruction will not need emulation. As of revision 2.6.1, these
9401 instructions are not generated unless you also use the
9402 @option{-funsafe-math-optimizations} switch.
9404 @item -malign-double
9405 @itemx -mno-align-double
9406 @opindex malign-double
9407 @opindex mno-align-double
9408 Control whether GCC aligns @code{double}, @code{long double}, and
9409 @code{long long} variables on a two word boundary or a one word
9410 boundary. Aligning @code{double} variables on a two word boundary will
9411 produce code that runs somewhat faster on a @samp{Pentium} at the
9412 expense of more memory.
9414 On x86-64, @option{-malign-double} is enabled by default.
9416 @strong{Warning:} if you use the @option{-malign-double} switch,
9417 structures containing the above types will be aligned differently than
9418 the published application binary interface specifications for the 386
9419 and will not be binary compatible with structures in code compiled
9420 without that switch.
9422 @item -m96bit-long-double
9423 @itemx -m128bit-long-double
9424 @opindex m96bit-long-double
9425 @opindex m128bit-long-double
9426 These switches control the size of @code{long double} type. The i386
9427 application binary interface specifies the size to be 96 bits,
9428 so @option{-m96bit-long-double} is the default in 32 bit mode.
9430 Modern architectures (Pentium and newer) would prefer @code{long double}
9431 to be aligned to an 8 or 16 byte boundary. In arrays or structures
9432 conforming to the ABI, this would not be possible. So specifying a
9433 @option{-m128bit-long-double} will align @code{long double}
9434 to a 16 byte boundary by padding the @code{long double} with an additional
9437 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
9438 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
9440 Notice that neither of these options enable any extra precision over the x87
9441 standard of 80 bits for a @code{long double}.
9443 @strong{Warning:} if you override the default value for your target ABI, the
9444 structures and arrays containing @code{long double} variables will change
9445 their size as well as function calling convention for function taking
9446 @code{long double} will be modified. Hence they will not be binary
9447 compatible with arrays or structures in code compiled without that switch.
9449 @item -mmlarge-data-threshold=@var{number}
9450 @opindex mlarge-data-threshold=@var{number}
9451 When @option{-mcmodel=medium} is specified, the data greater than
9452 @var{threshold} are placed in large data section. This value must be the
9453 same across all object linked into the binary and defaults to 65535.
9456 @itemx -mno-svr3-shlib
9457 @opindex msvr3-shlib
9458 @opindex mno-svr3-shlib
9459 Control whether GCC places uninitialized local variables into the
9460 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
9461 into @code{bss}. These options are meaningful only on System V Release 3.
9465 Use a different function-calling convention, in which functions that
9466 take a fixed number of arguments return with the @code{ret} @var{num}
9467 instruction, which pops their arguments while returning. This saves one
9468 instruction in the caller since there is no need to pop the arguments
9471 You can specify that an individual function is called with this calling
9472 sequence with the function attribute @samp{stdcall}. You can also
9473 override the @option{-mrtd} option by using the function attribute
9474 @samp{cdecl}. @xref{Function Attributes}.
9476 @strong{Warning:} this calling convention is incompatible with the one
9477 normally used on Unix, so you cannot use it if you need to call
9478 libraries compiled with the Unix compiler.
9480 Also, you must provide function prototypes for all functions that
9481 take variable numbers of arguments (including @code{printf});
9482 otherwise incorrect code will be generated for calls to those
9485 In addition, seriously incorrect code will result if you call a
9486 function with too many arguments. (Normally, extra arguments are
9487 harmlessly ignored.)
9489 @item -mregparm=@var{num}
9491 Control how many registers are used to pass integer arguments. By
9492 default, no registers are used to pass arguments, and at most 3
9493 registers can be used. You can control this behavior for a specific
9494 function by using the function attribute @samp{regparm}.
9495 @xref{Function Attributes}.
9497 @strong{Warning:} if you use this switch, and
9498 @var{num} is nonzero, then you must build all modules with the same
9499 value, including any libraries. This includes the system libraries and
9503 @opindex msseregparm
9504 Use SSE register passing conventions for float and double arguments
9505 and return values. You can control this behavior for a specific
9506 function by using the function attribute @samp{sseregparm}.
9507 @xref{Function Attributes}.
9509 @strong{Warning:} if you use this switch then you must build all
9510 modules with the same value, including any libraries. This includes
9511 the system libraries and startup modules.
9513 @item -mstackrealign
9514 @opindex mstackrealign
9515 Realign the stack at entry. On the Intel x86, the
9516 @option{-mstackrealign} option will generate an alternate prologue and
9517 epilogue that realigns the runtime stack. This supports mixing legacy
9518 codes that keep a 4-byte aligned stack with modern codes that keep a
9519 16-byte stack for SSE compatibility. The alternate prologue and
9520 epilogue are slower and bigger than the regular ones, and the
9521 alternate prologue requires an extra scratch register; this lowers the
9522 number of registers available if used in conjunction with the
9523 @code{regparm} attribute. The @option{-mstackrealign} option is
9524 incompatible with the nested function prologue; this is considered a
9525 hard error. See also the attribute @code{force_align_arg_pointer},
9526 applicable to individual functions.
9528 @item -mpreferred-stack-boundary=@var{num}
9529 @opindex mpreferred-stack-boundary
9530 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
9531 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
9532 the default is 4 (16 bytes or 128 bits).
9534 On Pentium and PentiumPro, @code{double} and @code{long double} values
9535 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
9536 suffer significant run time performance penalties. On Pentium III, the
9537 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
9538 properly if it is not 16 byte aligned.
9540 To ensure proper alignment of this values on the stack, the stack boundary
9541 must be as aligned as that required by any value stored on the stack.
9542 Further, every function must be generated such that it keeps the stack
9543 aligned. Thus calling a function compiled with a higher preferred
9544 stack boundary from a function compiled with a lower preferred stack
9545 boundary will most likely misalign the stack. It is recommended that
9546 libraries that use callbacks always use the default setting.
9548 This extra alignment does consume extra stack space, and generally
9549 increases code size. Code that is sensitive to stack space usage, such
9550 as embedded systems and operating system kernels, may want to reduce the
9551 preferred alignment to @option{-mpreferred-stack-boundary=2}.
9579 These switches enable or disable the use of instructions in the MMX,
9580 SSE, SSE2, SSE3, SSSE3, SSE4A, ABM, AES or 3DNow! extended
9581 instruction sets. These extensions are also available as built-in
9582 functions: see @ref{X86 Built-in Functions}, for details of the functions
9583 enabled and disabled by these switches.
9585 To have SSE/SSE2 instructions generated automatically from floating-point
9586 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
9588 These options will enable GCC to use these extended instructions in
9589 generated code, even without @option{-mfpmath=sse}. Applications which
9590 perform runtime CPU detection must compile separate files for each
9591 supported architecture, using the appropriate flags. In particular,
9592 the file containing the CPU detection code should be compiled without
9596 @itemx -mno-push-args
9598 @opindex mno-push-args
9599 Use PUSH operations to store outgoing parameters. This method is shorter
9600 and usually equally fast as method using SUB/MOV operations and is enabled
9601 by default. In some cases disabling it may improve performance because of
9602 improved scheduling and reduced dependencies.
9604 @item -maccumulate-outgoing-args
9605 @opindex maccumulate-outgoing-args
9606 If enabled, the maximum amount of space required for outgoing arguments will be
9607 computed in the function prologue. This is faster on most modern CPUs
9608 because of reduced dependencies, improved scheduling and reduced stack usage
9609 when preferred stack boundary is not equal to 2. The drawback is a notable
9610 increase in code size. This switch implies @option{-mno-push-args}.
9614 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
9615 on thread-safe exception handling must compile and link all code with the
9616 @option{-mthreads} option. When compiling, @option{-mthreads} defines
9617 @option{-D_MT}; when linking, it links in a special thread helper library
9618 @option{-lmingwthrd} which cleans up per thread exception handling data.
9620 @item -mno-align-stringops
9621 @opindex mno-align-stringops
9622 Do not align destination of inlined string operations. This switch reduces
9623 code size and improves performance in case the destination is already aligned,
9624 but GCC doesn't know about it.
9626 @item -minline-all-stringops
9627 @opindex minline-all-stringops
9628 By default GCC inlines string operations only when destination is known to be
9629 aligned at least to 4 byte boundary. This enables more inlining, increase code
9630 size, but may improve performance of code that depends on fast memcpy, strlen
9631 and memset for short lengths.
9633 @item -momit-leaf-frame-pointer
9634 @opindex momit-leaf-frame-pointer
9635 Don't keep the frame pointer in a register for leaf functions. This
9636 avoids the instructions to save, set up and restore frame pointers and
9637 makes an extra register available in leaf functions. The option
9638 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9639 which might make debugging harder.
9641 @item -mtls-direct-seg-refs
9642 @itemx -mno-tls-direct-seg-refs
9643 @opindex mtls-direct-seg-refs
9644 Controls whether TLS variables may be accessed with offsets from the
9645 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
9646 or whether the thread base pointer must be added. Whether or not this
9647 is legal depends on the operating system, and whether it maps the
9648 segment to cover the entire TLS area.
9650 For systems that use GNU libc, the default is on.
9653 These @samp{-m} switches are supported in addition to the above
9654 on AMD x86-64 processors in 64-bit environments.
9661 Generate code for a 32-bit or 64-bit environment.
9662 The 32-bit environment sets int, long and pointer to 32 bits and
9663 generates code that runs on any i386 system.
9664 The 64-bit environment sets int to 32 bits and long and pointer
9665 to 64 bits and generates code for AMD's x86-64 architecture. For
9666 darwin only the -m64 option turns off the @option{-fno-pic} and
9667 @option{-mdynamic-no-pic} options.
9670 @opindex no-red-zone
9671 Do not use a so called red zone for x86-64 code. The red zone is mandated
9672 by the x86-64 ABI, it is a 128-byte area beyond the location of the
9673 stack pointer that will not be modified by signal or interrupt handlers
9674 and therefore can be used for temporary data without adjusting the stack
9675 pointer. The flag @option{-mno-red-zone} disables this red zone.
9677 @item -mcmodel=small
9678 @opindex mcmodel=small
9679 Generate code for the small code model: the program and its symbols must
9680 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
9681 Programs can be statically or dynamically linked. This is the default
9684 @item -mcmodel=kernel
9685 @opindex mcmodel=kernel
9686 Generate code for the kernel code model. The kernel runs in the
9687 negative 2 GB of the address space.
9688 This model has to be used for Linux kernel code.
9690 @item -mcmodel=medium
9691 @opindex mcmodel=medium
9692 Generate code for the medium model: The program is linked in the lower 2
9693 GB of the address space but symbols can be located anywhere in the
9694 address space. Programs can be statically or dynamically linked, but
9695 building of shared libraries are not supported with the medium model.
9697 @item -mcmodel=large
9698 @opindex mcmodel=large
9699 Generate code for the large model: This model makes no assumptions
9700 about addresses and sizes of sections. Currently GCC does not implement
9705 @subsection IA-64 Options
9706 @cindex IA-64 Options
9708 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9712 @opindex mbig-endian
9713 Generate code for a big endian target. This is the default for HP-UX@.
9715 @item -mlittle-endian
9716 @opindex mlittle-endian
9717 Generate code for a little endian target. This is the default for AIX5
9724 Generate (or don't) code for the GNU assembler. This is the default.
9725 @c Also, this is the default if the configure option @option{--with-gnu-as}
9732 Generate (or don't) code for the GNU linker. This is the default.
9733 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9738 Generate code that does not use a global pointer register. The result
9739 is not position independent code, and violates the IA-64 ABI@.
9741 @item -mvolatile-asm-stop
9742 @itemx -mno-volatile-asm-stop
9743 @opindex mvolatile-asm-stop
9744 @opindex mno-volatile-asm-stop
9745 Generate (or don't) a stop bit immediately before and after volatile asm
9748 @item -mregister-names
9749 @itemx -mno-register-names
9750 @opindex mregister-names
9751 @opindex mno-register-names
9752 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9753 the stacked registers. This may make assembler output more readable.
9759 Disable (or enable) optimizations that use the small data section. This may
9760 be useful for working around optimizer bugs.
9763 @opindex mconstant-gp
9764 Generate code that uses a single constant global pointer value. This is
9765 useful when compiling kernel code.
9769 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9770 This is useful when compiling firmware code.
9772 @item -minline-float-divide-min-latency
9773 @opindex minline-float-divide-min-latency
9774 Generate code for inline divides of floating point values
9775 using the minimum latency algorithm.
9777 @item -minline-float-divide-max-throughput
9778 @opindex minline-float-divide-max-throughput
9779 Generate code for inline divides of floating point values
9780 using the maximum throughput algorithm.
9782 @item -minline-int-divide-min-latency
9783 @opindex minline-int-divide-min-latency
9784 Generate code for inline divides of integer values
9785 using the minimum latency algorithm.
9787 @item -minline-int-divide-max-throughput
9788 @opindex minline-int-divide-max-throughput
9789 Generate code for inline divides of integer values
9790 using the maximum throughput algorithm.
9792 @item -minline-sqrt-min-latency
9793 @opindex minline-sqrt-min-latency
9794 Generate code for inline square roots
9795 using the minimum latency algorithm.
9797 @item -minline-sqrt-max-throughput
9798 @opindex minline-sqrt-max-throughput
9799 Generate code for inline square roots
9800 using the maximum throughput algorithm.
9802 @item -mno-dwarf2-asm
9804 @opindex mno-dwarf2-asm
9805 @opindex mdwarf2-asm
9806 Don't (or do) generate assembler code for the DWARF2 line number debugging
9807 info. This may be useful when not using the GNU assembler.
9809 @item -mearly-stop-bits
9810 @itemx -mno-early-stop-bits
9811 @opindex mearly-stop-bits
9812 @opindex mno-early-stop-bits
9813 Allow stop bits to be placed earlier than immediately preceding the
9814 instruction that triggered the stop bit. This can improve instruction
9815 scheduling, but does not always do so.
9817 @item -mfixed-range=@var{register-range}
9818 @opindex mfixed-range
9819 Generate code treating the given register range as fixed registers.
9820 A fixed register is one that the register allocator can not use. This is
9821 useful when compiling kernel code. A register range is specified as
9822 two registers separated by a dash. Multiple register ranges can be
9823 specified separated by a comma.
9825 @item -mtls-size=@var{tls-size}
9827 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9830 @item -mtune=@var{cpu-type}
9832 Tune the instruction scheduling for a particular CPU, Valid values are
9833 itanium, itanium1, merced, itanium2, and mckinley.
9839 Add support for multithreading using the POSIX threads library. This
9840 option sets flags for both the preprocessor and linker. It does
9841 not affect the thread safety of object code produced by the compiler or
9842 that of libraries supplied with it. These are HP-UX specific flags.
9848 Generate code for a 32-bit or 64-bit environment.
9849 The 32-bit environment sets int, long and pointer to 32 bits.
9850 The 64-bit environment sets int to 32 bits and long and pointer
9851 to 64 bits. These are HP-UX specific flags.
9853 @item -mno-sched-br-data-spec
9854 @itemx -msched-br-data-spec
9855 @opindex mno-sched-br-data-spec
9856 @opindex msched-br-data-spec
9857 (Dis/En)able data speculative scheduling before 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 'disable'.
9862 @item -msched-ar-data-spec
9863 @itemx -mno-sched-ar-data-spec
9864 @opindex msched-ar-data-spec
9865 @opindex mno-sched-ar-data-spec
9866 (En/Dis)able data speculative scheduling after reload.
9867 This will result in generation of the ld.a instructions and
9868 the corresponding check instructions (ld.c / chk.a).
9869 The default is 'enable'.
9871 @item -mno-sched-control-spec
9872 @itemx -msched-control-spec
9873 @opindex mno-sched-control-spec
9874 @opindex msched-control-spec
9875 (Dis/En)able control speculative scheduling. This feature is
9876 available only during region scheduling (i.e. before reload).
9877 This will result in generation of the ld.s instructions and
9878 the corresponding check instructions chk.s .
9879 The default is 'disable'.
9881 @item -msched-br-in-data-spec
9882 @itemx -mno-sched-br-in-data-spec
9883 @opindex msched-br-in-data-spec
9884 @opindex mno-sched-br-in-data-spec
9885 (En/Dis)able speculative scheduling of the instructions that
9886 are dependent on the data speculative loads before reload.
9887 This is effective only with @option{-msched-br-data-spec} enabled.
9888 The default is 'enable'.
9890 @item -msched-ar-in-data-spec
9891 @itemx -mno-sched-ar-in-data-spec
9892 @opindex msched-ar-in-data-spec
9893 @opindex mno-sched-ar-in-data-spec
9894 (En/Dis)able speculative scheduling of the instructions that
9895 are dependent on the data speculative loads after reload.
9896 This is effective only with @option{-msched-ar-data-spec} enabled.
9897 The default is 'enable'.
9899 @item -msched-in-control-spec
9900 @itemx -mno-sched-in-control-spec
9901 @opindex msched-in-control-spec
9902 @opindex mno-sched-in-control-spec
9903 (En/Dis)able speculative scheduling of the instructions that
9904 are dependent on the control speculative loads.
9905 This is effective only with @option{-msched-control-spec} enabled.
9906 The default is 'enable'.
9909 @itemx -mno-sched-ldc
9911 @opindex mno-sched-ldc
9912 (En/Dis)able use of simple data speculation checks ld.c .
9913 If disabled, only chk.a instructions will be emitted to check
9914 data speculative loads.
9915 The default is 'enable'.
9917 @item -mno-sched-control-ldc
9918 @itemx -msched-control-ldc
9919 @opindex mno-sched-control-ldc
9920 @opindex msched-control-ldc
9921 (Dis/En)able use of ld.c instructions to check control speculative loads.
9922 If enabled, in case of control speculative load with no speculatively
9923 scheduled dependent instructions this load will be emitted as ld.sa and
9924 ld.c will be used to check it.
9925 The default is 'disable'.
9927 @item -mno-sched-spec-verbose
9928 @itemx -msched-spec-verbose
9929 @opindex mno-sched-spec-verbose
9930 @opindex msched-spec-verbose
9931 (Dis/En)able printing of the information about speculative motions.
9933 @item -mno-sched-prefer-non-data-spec-insns
9934 @itemx -msched-prefer-non-data-spec-insns
9935 @opindex mno-sched-prefer-non-data-spec-insns
9936 @opindex msched-prefer-non-data-spec-insns
9937 If enabled, data 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 data speculation much more conservative.
9940 The default is 'disable'.
9942 @item -mno-sched-prefer-non-control-spec-insns
9943 @itemx -msched-prefer-non-control-spec-insns
9944 @opindex mno-sched-prefer-non-control-spec-insns
9945 @opindex msched-prefer-non-control-spec-insns
9946 If enabled, control speculative instructions will be chosen for schedule
9947 only if there are no other choices at the moment. This will make
9948 the use of the control speculation much more conservative.
9949 The default is 'disable'.
9951 @item -mno-sched-count-spec-in-critical-path
9952 @itemx -msched-count-spec-in-critical-path
9953 @opindex mno-sched-count-spec-in-critical-path
9954 @opindex msched-count-spec-in-critical-path
9955 If enabled, speculative dependencies will be considered during
9956 computation of the instructions priorities. This will make the use of the
9957 speculation a bit more conservative.
9958 The default is 'disable'.
9963 @subsection M32C Options
9964 @cindex M32C options
9967 @item -mcpu=@var{name}
9969 Select the CPU for which code is generated. @var{name} may be one of
9970 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
9971 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
9976 Specifies that the program will be run on the simulator. This causes
9977 an alternate runtime library to be linked in which supports, for
9978 example, file I/O. You must not use this option when generating
9979 programs that will run on real hardware; you must provide your own
9980 runtime library for whatever I/O functions are needed.
9982 @item -memregs=@var{number}
9984 Specifies the number of memory-based pseudo-registers GCC will use
9985 during code generation. These pseudo-registers will be used like real
9986 registers, so there is a tradeoff between GCC's ability to fit the
9987 code into available registers, and the performance penalty of using
9988 memory instead of registers. Note that all modules in a program must
9989 be compiled with the same value for this option. Because of that, you
9990 must not use this option with the default runtime libraries gcc
9995 @node M32R/D Options
9996 @subsection M32R/D Options
9997 @cindex M32R/D options
9999 These @option{-m} options are defined for Renesas M32R/D architectures:
10004 Generate code for the M32R/2@.
10008 Generate code for the M32R/X@.
10012 Generate code for the M32R@. This is the default.
10014 @item -mmodel=small
10015 @opindex mmodel=small
10016 Assume all objects live in the lower 16MB of memory (so that their addresses
10017 can be loaded with the @code{ld24} instruction), and assume all subroutines
10018 are reachable with the @code{bl} instruction.
10019 This is the default.
10021 The addressability of a particular object can be set with the
10022 @code{model} attribute.
10024 @item -mmodel=medium
10025 @opindex mmodel=medium
10026 Assume objects may be anywhere in the 32-bit address space (the compiler
10027 will generate @code{seth/add3} instructions to load their addresses), and
10028 assume all subroutines are reachable with the @code{bl} instruction.
10030 @item -mmodel=large
10031 @opindex mmodel=large
10032 Assume objects may be anywhere in the 32-bit address space (the compiler
10033 will generate @code{seth/add3} instructions to load their addresses), and
10034 assume subroutines may not be reachable with the @code{bl} instruction
10035 (the compiler will generate the much slower @code{seth/add3/jl}
10036 instruction sequence).
10039 @opindex msdata=none
10040 Disable use of the small data area. Variables will be put into
10041 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
10042 @code{section} attribute has been specified).
10043 This is the default.
10045 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
10046 Objects may be explicitly put in the small data area with the
10047 @code{section} attribute using one of these sections.
10049 @item -msdata=sdata
10050 @opindex msdata=sdata
10051 Put small global and static data in the small data area, but do not
10052 generate special code to reference them.
10055 @opindex msdata=use
10056 Put small global and static data in the small data area, and generate
10057 special instructions to reference them.
10061 @cindex smaller data references
10062 Put global and static objects less than or equal to @var{num} bytes
10063 into the small data or bss sections instead of the normal data or bss
10064 sections. The default value of @var{num} is 8.
10065 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
10066 for this option to have any effect.
10068 All modules should be compiled with the same @option{-G @var{num}} value.
10069 Compiling with different values of @var{num} may or may not work; if it
10070 doesn't the linker will give an error message---incorrect code will not be
10075 Makes the M32R specific code in the compiler display some statistics
10076 that might help in debugging programs.
10078 @item -malign-loops
10079 @opindex malign-loops
10080 Align all loops to a 32-byte boundary.
10082 @item -mno-align-loops
10083 @opindex mno-align-loops
10084 Do not enforce a 32-byte alignment for loops. This is the default.
10086 @item -missue-rate=@var{number}
10087 @opindex missue-rate=@var{number}
10088 Issue @var{number} instructions per cycle. @var{number} can only be 1
10091 @item -mbranch-cost=@var{number}
10092 @opindex mbranch-cost=@var{number}
10093 @var{number} can only be 1 or 2. If it is 1 then branches will be
10094 preferred over conditional code, if it is 2, then the opposite will
10097 @item -mflush-trap=@var{number}
10098 @opindex mflush-trap=@var{number}
10099 Specifies the trap number to use to flush the cache. The default is
10100 12. Valid numbers are between 0 and 15 inclusive.
10102 @item -mno-flush-trap
10103 @opindex mno-flush-trap
10104 Specifies that the cache cannot be flushed by using a trap.
10106 @item -mflush-func=@var{name}
10107 @opindex mflush-func=@var{name}
10108 Specifies the name of the operating system function to call to flush
10109 the cache. The default is @emph{_flush_cache}, but a function call
10110 will only be used if a trap is not available.
10112 @item -mno-flush-func
10113 @opindex mno-flush-func
10114 Indicates that there is no OS function for flushing the cache.
10118 @node M680x0 Options
10119 @subsection M680x0 Options
10120 @cindex M680x0 options
10122 These are the @samp{-m} options defined for the 68000 series. The default
10123 values for these options depends on which style of 68000 was selected when
10124 the compiler was configured; the defaults for the most common choices are
10132 Generate output for a 68000. This is the default
10133 when the compiler is configured for 68000-based systems.
10135 Use this option for microcontrollers with a 68000 or EC000 core,
10136 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
10142 Generate output for a 68020. This is the default
10143 when the compiler is configured for 68020-based systems.
10147 Generate output containing 68881 instructions for floating point.
10148 This is the default for most 68020 systems unless @option{--nfp} was
10149 specified when the compiler was configured.
10153 Generate output for a 68030. This is the default when the compiler is
10154 configured for 68030-based systems.
10158 Generate output for a 68040. This is the default when the compiler is
10159 configured for 68040-based systems.
10161 This option inhibits the use of 68881/68882 instructions that have to be
10162 emulated by software on the 68040. Use this option if your 68040 does not
10163 have code to emulate those instructions.
10167 Generate output for a 68060. This is the default when the compiler is
10168 configured for 68060-based systems.
10170 This option inhibits the use of 68020 and 68881/68882 instructions that
10171 have to be emulated by software on the 68060. Use this option if your 68060
10172 does not have code to emulate those instructions.
10176 Generate output for a CPU32. This is the default
10177 when the compiler is configured for CPU32-based systems.
10179 Use this option for microcontrollers with a
10180 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
10181 68336, 68340, 68341, 68349 and 68360.
10185 Generate output for a 520X ``coldfire'' family cpu. This is the default
10186 when the compiler is configured for 520X-based systems.
10188 Use this option for microcontroller with a 5200 core, including
10189 the MCF5202, MCF5203, MCF5204 and MCF5202.
10193 Generate output for a ColdFire V4e family cpu (e.g.@: 547x/548x).
10194 This includes use of hardware floating point instructions.
10198 Generate output for a 68040, without using any of the new instructions.
10199 This results in code which can run relatively efficiently on either a
10200 68020/68881 or a 68030 or a 68040. The generated code does use the
10201 68881 instructions that are emulated on the 68040.
10205 Generate output for a 68060, without using any of the new instructions.
10206 This results in code which can run relatively efficiently on either a
10207 68020/68881 or a 68030 or a 68040. The generated code does use the
10208 68881 instructions that are emulated on the 68060.
10211 @opindex msoft-float
10212 Generate output containing library calls for floating point.
10213 @strong{Warning:} the requisite libraries are not available for all m68k
10214 targets. Normally the facilities of the machine's usual C compiler are
10215 used, but this can't be done directly in cross-compilation. You must
10216 make your own arrangements to provide suitable library functions for
10217 cross-compilation. The embedded targets @samp{m68k-*-aout} and
10218 @samp{m68k-*-coff} do provide software floating point support.
10222 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10223 Additionally, parameters passed on the stack are also aligned to a
10224 16-bit boundary even on targets whose API mandates promotion to 32-bit.
10227 @opindex mnobitfield
10228 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
10229 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
10233 Do use the bit-field instructions. The @option{-m68020} option implies
10234 @option{-mbitfield}. This is the default if you use a configuration
10235 designed for a 68020.
10239 Use a different function-calling convention, in which functions
10240 that take a fixed number of arguments return with the @code{rtd}
10241 instruction, which pops their arguments while returning. This
10242 saves one instruction in the caller since there is no need to pop
10243 the arguments there.
10245 This calling convention is incompatible with the one normally
10246 used on Unix, so you cannot use it if you need to call libraries
10247 compiled with the Unix compiler.
10249 Also, you must provide function prototypes for all functions that
10250 take variable numbers of arguments (including @code{printf});
10251 otherwise incorrect code will be generated for calls to those
10254 In addition, seriously incorrect code will result if you call a
10255 function with too many arguments. (Normally, extra arguments are
10256 harmlessly ignored.)
10258 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
10259 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
10262 @itemx -mno-align-int
10263 @opindex malign-int
10264 @opindex mno-align-int
10265 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
10266 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
10267 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
10268 Aligning variables on 32-bit boundaries produces code that runs somewhat
10269 faster on processors with 32-bit busses at the expense of more memory.
10271 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
10272 align structures containing the above types differently than
10273 most published application binary interface specifications for the m68k.
10277 Use the pc-relative addressing mode of the 68000 directly, instead of
10278 using a global offset table. At present, this option implies @option{-fpic},
10279 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
10280 not presently supported with @option{-mpcrel}, though this could be supported for
10281 68020 and higher processors.
10283 @item -mno-strict-align
10284 @itemx -mstrict-align
10285 @opindex mno-strict-align
10286 @opindex mstrict-align
10287 Do not (do) assume that unaligned memory references will be handled by
10291 Generate code that allows the data segment to be located in a different
10292 area of memory from the text segment. This allows for execute in place in
10293 an environment without virtual memory management. This option implies
10296 @item -mno-sep-data
10297 Generate code that assumes that the data segment follows the text segment.
10298 This is the default.
10300 @item -mid-shared-library
10301 Generate code that supports shared libraries via the library ID method.
10302 This allows for execute in place and shared libraries in an environment
10303 without virtual memory management. This option implies @option{-fPIC}.
10305 @item -mno-id-shared-library
10306 Generate code that doesn't assume ID based shared libraries are being used.
10307 This is the default.
10309 @item -mshared-library-id=n
10310 Specified the identification number of the ID based shared library being
10311 compiled. Specifying a value of 0 will generate more compact code, specifying
10312 other values will force the allocation of that number to the current
10313 library but is no more space or time efficient than omitting this option.
10317 @node M68hc1x Options
10318 @subsection M68hc1x Options
10319 @cindex M68hc1x options
10321 These are the @samp{-m} options defined for the 68hc11 and 68hc12
10322 microcontrollers. The default values for these options depends on
10323 which style of microcontroller was selected when the compiler was configured;
10324 the defaults for the most common choices are given below.
10331 Generate output for a 68HC11. This is the default
10332 when the compiler is configured for 68HC11-based systems.
10338 Generate output for a 68HC12. This is the default
10339 when the compiler is configured for 68HC12-based systems.
10345 Generate output for a 68HCS12.
10347 @item -mauto-incdec
10348 @opindex mauto-incdec
10349 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
10356 Enable the use of 68HC12 min and max instructions.
10359 @itemx -mno-long-calls
10360 @opindex mlong-calls
10361 @opindex mno-long-calls
10362 Treat all calls as being far away (near). If calls are assumed to be
10363 far away, the compiler will use the @code{call} instruction to
10364 call a function and the @code{rtc} instruction for returning.
10368 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10370 @item -msoft-reg-count=@var{count}
10371 @opindex msoft-reg-count
10372 Specify the number of pseudo-soft registers which are used for the
10373 code generation. The maximum number is 32. Using more pseudo-soft
10374 register may or may not result in better code depending on the program.
10375 The default is 4 for 68HC11 and 2 for 68HC12.
10379 @node MCore Options
10380 @subsection MCore Options
10381 @cindex MCore options
10383 These are the @samp{-m} options defined for the Motorola M*Core
10389 @itemx -mno-hardlit
10391 @opindex mno-hardlit
10392 Inline constants into the code stream if it can be done in two
10393 instructions or less.
10399 Use the divide instruction. (Enabled by default).
10401 @item -mrelax-immediate
10402 @itemx -mno-relax-immediate
10403 @opindex mrelax-immediate
10404 @opindex mno-relax-immediate
10405 Allow arbitrary sized immediates in bit operations.
10407 @item -mwide-bitfields
10408 @itemx -mno-wide-bitfields
10409 @opindex mwide-bitfields
10410 @opindex mno-wide-bitfields
10411 Always treat bit-fields as int-sized.
10413 @item -m4byte-functions
10414 @itemx -mno-4byte-functions
10415 @opindex m4byte-functions
10416 @opindex mno-4byte-functions
10417 Force all functions to be aligned to a four byte boundary.
10419 @item -mcallgraph-data
10420 @itemx -mno-callgraph-data
10421 @opindex mcallgraph-data
10422 @opindex mno-callgraph-data
10423 Emit callgraph information.
10426 @itemx -mno-slow-bytes
10427 @opindex mslow-bytes
10428 @opindex mno-slow-bytes
10429 Prefer word access when reading byte quantities.
10431 @item -mlittle-endian
10432 @itemx -mbig-endian
10433 @opindex mlittle-endian
10434 @opindex mbig-endian
10435 Generate code for a little endian target.
10441 Generate code for the 210 processor.
10445 @subsection MIPS Options
10446 @cindex MIPS options
10452 Generate big-endian code.
10456 Generate little-endian code. This is the default for @samp{mips*el-*-*}
10459 @item -march=@var{arch}
10461 Generate code that will run on @var{arch}, which can be the name of a
10462 generic MIPS ISA, or the name of a particular processor.
10464 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
10465 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
10466 The processor names are:
10467 @samp{4kc}, @samp{4km}, @samp{4kp},
10468 @samp{5kc}, @samp{5kf},
10470 @samp{24k}, @samp{24kc}, @samp{24kf}, @samp{24kx},
10473 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
10474 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
10475 @samp{rm7000}, @samp{rm9000},
10478 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
10479 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
10480 The special value @samp{from-abi} selects the
10481 most compatible architecture for the selected ABI (that is,
10482 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
10484 In processor names, a final @samp{000} can be abbreviated as @samp{k}
10485 (for example, @samp{-march=r2k}). Prefixes are optional, and
10486 @samp{vr} may be written @samp{r}.
10488 GCC defines two macros based on the value of this option. The first
10489 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
10490 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
10491 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
10492 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
10493 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
10495 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
10496 above. In other words, it will have the full prefix and will not
10497 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
10498 the macro names the resolved architecture (either @samp{"mips1"} or
10499 @samp{"mips3"}). It names the default architecture when no
10500 @option{-march} option is given.
10502 @item -mtune=@var{arch}
10504 Optimize for @var{arch}. Among other things, this option controls
10505 the way instructions are scheduled, and the perceived cost of arithmetic
10506 operations. The list of @var{arch} values is the same as for
10509 When this option is not used, GCC will optimize for the processor
10510 specified by @option{-march}. By using @option{-march} and
10511 @option{-mtune} together, it is possible to generate code that will
10512 run on a family of processors, but optimize the code for one
10513 particular member of that family.
10515 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
10516 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
10517 @samp{-march} ones described above.
10521 Equivalent to @samp{-march=mips1}.
10525 Equivalent to @samp{-march=mips2}.
10529 Equivalent to @samp{-march=mips3}.
10533 Equivalent to @samp{-march=mips4}.
10537 Equivalent to @samp{-march=mips32}.
10541 Equivalent to @samp{-march=mips32r2}.
10545 Equivalent to @samp{-march=mips64}.
10550 @opindex mno-mips16
10551 Generate (do not generate) MIPS16 code. If GCC is targetting a
10552 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
10564 Generate code for the given ABI@.
10566 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
10567 generates 64-bit code when you select a 64-bit architecture, but you
10568 can use @option{-mgp32} to get 32-bit code instead.
10570 For information about the O64 ABI, see
10571 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
10574 @itemx -mno-abicalls
10576 @opindex mno-abicalls
10577 Generate (do not generate) code that is suitable for SVR4-style
10578 dynamic objects. @option{-mabicalls} is the default for SVR4-based
10583 Generate (do not generate) code that is fully position-independent,
10584 and that can therefore be linked into shared libraries. This option
10585 only affects @option{-mabicalls}.
10587 All @option{-mabicalls} code has traditionally been position-independent,
10588 regardless of options like @option{-fPIC} and @option{-fpic}. However,
10589 as an extension, the GNU toolchain allows executables to use absolute
10590 accesses for locally-binding symbols. It can also use shorter GP
10591 initialization sequences and generate direct calls to locally-defined
10592 functions. This mode is selected by @option{-mno-shared}.
10594 @option{-mno-shared} depends on binutils 2.16 or higher and generates
10595 objects that can only be linked by the GNU linker. However, the option
10596 does not affect the ABI of the final executable; it only affects the ABI
10597 of relocatable objects. Using @option{-mno-shared} will generally make
10598 executables both smaller and quicker.
10600 @option{-mshared} is the default.
10606 Lift (do not lift) the usual restrictions on the size of the global
10609 GCC normally uses a single instruction to load values from the GOT@.
10610 While this is relatively efficient, it will only work if the GOT
10611 is smaller than about 64k. Anything larger will cause the linker
10612 to report an error such as:
10614 @cindex relocation truncated to fit (MIPS)
10616 relocation truncated to fit: R_MIPS_GOT16 foobar
10619 If this happens, you should recompile your code with @option{-mxgot}.
10620 It should then work with very large GOTs, although it will also be
10621 less efficient, since it will take three instructions to fetch the
10622 value of a global symbol.
10624 Note that some linkers can create multiple GOTs. If you have such a
10625 linker, you should only need to use @option{-mxgot} when a single object
10626 file accesses more than 64k's worth of GOT entries. Very few do.
10628 These options have no effect unless GCC is generating position
10633 Assume that general-purpose registers are 32 bits wide.
10637 Assume that general-purpose registers are 64 bits wide.
10641 Assume that floating-point registers are 32 bits wide.
10645 Assume that floating-point registers are 64 bits wide.
10648 @opindex mhard-float
10649 Use floating-point coprocessor instructions.
10652 @opindex msoft-float
10653 Do not use floating-point coprocessor instructions. Implement
10654 floating-point calculations using library calls instead.
10656 @item -msingle-float
10657 @opindex msingle-float
10658 Assume that the floating-point coprocessor only supports single-precision
10661 @itemx -mdouble-float
10662 @opindex mdouble-float
10663 Assume that the floating-point coprocessor supports double-precision
10664 operations. This is the default.
10670 Use (do not use) the MIPS DSP ASE. @xref{MIPS DSP Built-in Functions}.
10672 @itemx -mpaired-single
10673 @itemx -mno-paired-single
10674 @opindex mpaired-single
10675 @opindex mno-paired-single
10676 Use (do not use) paired-single floating-point instructions.
10677 @xref{MIPS Paired-Single Support}. This option can only be used
10678 when generating 64-bit code and requires hardware floating-point
10679 support to be enabled.
10684 @opindex mno-mips3d
10685 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
10686 The option @option{-mips3d} implies @option{-mpaired-single}.
10690 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
10691 an explanation of the default and the way that the pointer size is
10696 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
10698 The default size of @code{int}s, @code{long}s and pointers depends on
10699 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
10700 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
10701 32-bit @code{long}s. Pointers are the same size as @code{long}s,
10702 or the same size as integer registers, whichever is smaller.
10708 Assume (do not assume) that all symbols have 32-bit values, regardless
10709 of the selected ABI@. This option is useful in combination with
10710 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
10711 to generate shorter and faster references to symbolic addresses.
10715 @cindex smaller data references (MIPS)
10716 @cindex gp-relative references (MIPS)
10717 Put global and static items less than or equal to @var{num} bytes into
10718 the small data or bss section instead of the normal data or bss section.
10719 This allows the data to be accessed using a single instruction.
10721 All modules should be compiled with the same @option{-G @var{num}}
10724 @item -membedded-data
10725 @itemx -mno-embedded-data
10726 @opindex membedded-data
10727 @opindex mno-embedded-data
10728 Allocate variables to the read-only data section first if possible, then
10729 next in the small data section if possible, otherwise in data. This gives
10730 slightly slower code than the default, but reduces the amount of RAM required
10731 when executing, and thus may be preferred for some embedded systems.
10733 @item -muninit-const-in-rodata
10734 @itemx -mno-uninit-const-in-rodata
10735 @opindex muninit-const-in-rodata
10736 @opindex mno-uninit-const-in-rodata
10737 Put uninitialized @code{const} variables in the read-only data section.
10738 This option is only meaningful in conjunction with @option{-membedded-data}.
10740 @item -msplit-addresses
10741 @itemx -mno-split-addresses
10742 @opindex msplit-addresses
10743 @opindex mno-split-addresses
10744 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
10745 relocation operators. This option has been superseded by
10746 @option{-mexplicit-relocs} but is retained for backwards compatibility.
10748 @item -mexplicit-relocs
10749 @itemx -mno-explicit-relocs
10750 @opindex mexplicit-relocs
10751 @opindex mno-explicit-relocs
10752 Use (do not use) assembler relocation operators when dealing with symbolic
10753 addresses. The alternative, selected by @option{-mno-explicit-relocs},
10754 is to use assembler macros instead.
10756 @option{-mexplicit-relocs} is the default if GCC was configured
10757 to use an assembler that supports relocation operators.
10759 @item -mcheck-zero-division
10760 @itemx -mno-check-zero-division
10761 @opindex mcheck-zero-division
10762 @opindex mno-check-zero-division
10763 Trap (do not trap) on integer division by zero. The default is
10764 @option{-mcheck-zero-division}.
10766 @item -mdivide-traps
10767 @itemx -mdivide-breaks
10768 @opindex mdivide-traps
10769 @opindex mdivide-breaks
10770 MIPS systems check for division by zero by generating either a
10771 conditional trap or a break instruction. Using traps results in
10772 smaller code, but is only supported on MIPS II and later. Also, some
10773 versions of the Linux kernel have a bug that prevents trap from
10774 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
10775 allow conditional traps on architectures that support them and
10776 @option{-mdivide-breaks} to force the use of breaks.
10778 The default is usually @option{-mdivide-traps}, but this can be
10779 overridden at configure time using @option{--with-divide=breaks}.
10780 Divide-by-zero checks can be completely disabled using
10781 @option{-mno-check-zero-division}.
10786 @opindex mno-memcpy
10787 Force (do not force) the use of @code{memcpy()} for non-trivial block
10788 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
10789 most constant-sized copies.
10792 @itemx -mno-long-calls
10793 @opindex mlong-calls
10794 @opindex mno-long-calls
10795 Disable (do not disable) use of the @code{jal} instruction. Calling
10796 functions using @code{jal} is more efficient but requires the caller
10797 and callee to be in the same 256 megabyte segment.
10799 This option has no effect on abicalls code. The default is
10800 @option{-mno-long-calls}.
10806 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
10807 instructions, as provided by the R4650 ISA@.
10810 @itemx -mno-fused-madd
10811 @opindex mfused-madd
10812 @opindex mno-fused-madd
10813 Enable (disable) use of the floating point multiply-accumulate
10814 instructions, when they are available. The default is
10815 @option{-mfused-madd}.
10817 When multiply-accumulate instructions are used, the intermediate
10818 product is calculated to infinite precision and is not subject to
10819 the FCSR Flush to Zero bit. This may be undesirable in some
10824 Tell the MIPS assembler to not run its preprocessor over user
10825 assembler files (with a @samp{.s} suffix) when assembling them.
10828 @itemx -mno-fix-r4000
10829 @opindex mfix-r4000
10830 @opindex mno-fix-r4000
10831 Work around certain R4000 CPU errata:
10834 A double-word or a variable shift may give an incorrect result if executed
10835 immediately after starting an integer division.
10837 A double-word or a variable shift may give an incorrect result if executed
10838 while an integer multiplication is in progress.
10840 An integer division may give an incorrect result if started in a delay slot
10841 of a taken branch or a jump.
10845 @itemx -mno-fix-r4400
10846 @opindex mfix-r4400
10847 @opindex mno-fix-r4400
10848 Work around certain R4400 CPU errata:
10851 A double-word or a variable shift may give an incorrect result if executed
10852 immediately after starting an integer division.
10856 @itemx -mno-fix-vr4120
10857 @opindex mfix-vr4120
10858 Work around certain VR4120 errata:
10861 @code{dmultu} does not always produce the correct result.
10863 @code{div} and @code{ddiv} do not always produce the correct result if one
10864 of the operands is negative.
10866 The workarounds for the division errata rely on special functions in
10867 @file{libgcc.a}. At present, these functions are only provided by
10868 the @code{mips64vr*-elf} configurations.
10870 Other VR4120 errata require a nop to be inserted between certain pairs of
10871 instructions. These errata are handled by the assembler, not by GCC itself.
10874 @opindex mfix-vr4130
10875 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
10876 workarounds are implemented by the assembler rather than by GCC,
10877 although GCC will avoid using @code{mflo} and @code{mfhi} if the
10878 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
10879 instructions are available instead.
10882 @itemx -mno-fix-sb1
10884 Work around certain SB-1 CPU core errata.
10885 (This flag currently works around the SB-1 revision 2
10886 ``F1'' and ``F2'' floating point errata.)
10888 @item -mflush-func=@var{func}
10889 @itemx -mno-flush-func
10890 @opindex mflush-func
10891 Specifies the function to call to flush the I and D caches, or to not
10892 call any such function. If called, the function must take the same
10893 arguments as the common @code{_flush_func()}, that is, the address of the
10894 memory range for which the cache is being flushed, the size of the
10895 memory range, and the number 3 (to flush both caches). The default
10896 depends on the target GCC was configured for, but commonly is either
10897 @samp{_flush_func} or @samp{__cpu_flush}.
10899 @item -mbranch-likely
10900 @itemx -mno-branch-likely
10901 @opindex mbranch-likely
10902 @opindex mno-branch-likely
10903 Enable or disable use of Branch Likely instructions, regardless of the
10904 default for the selected architecture. By default, Branch Likely
10905 instructions may be generated if they are supported by the selected
10906 architecture. An exception is for the MIPS32 and MIPS64 architectures
10907 and processors which implement those architectures; for those, Branch
10908 Likely instructions will not be generated by default because the MIPS32
10909 and MIPS64 architectures specifically deprecate their use.
10911 @item -mfp-exceptions
10912 @itemx -mno-fp-exceptions
10913 @opindex mfp-exceptions
10914 Specifies whether FP exceptions are enabled. This affects how we schedule
10915 FP instructions for some processors. The default is that FP exceptions are
10918 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
10919 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
10922 @item -mvr4130-align
10923 @itemx -mno-vr4130-align
10924 @opindex mvr4130-align
10925 The VR4130 pipeline is two-way superscalar, but can only issue two
10926 instructions together if the first one is 8-byte aligned. When this
10927 option is enabled, GCC will align pairs of instructions that it
10928 thinks should execute in parallel.
10930 This option only has an effect when optimizing for the VR4130.
10931 It normally makes code faster, but at the expense of making it bigger.
10932 It is enabled by default at optimization level @option{-O3}.
10936 @subsection MMIX Options
10937 @cindex MMIX Options
10939 These options are defined for the MMIX:
10943 @itemx -mno-libfuncs
10945 @opindex mno-libfuncs
10946 Specify that intrinsic library functions are being compiled, passing all
10947 values in registers, no matter the size.
10950 @itemx -mno-epsilon
10952 @opindex mno-epsilon
10953 Generate floating-point comparison instructions that compare with respect
10954 to the @code{rE} epsilon register.
10956 @item -mabi=mmixware
10958 @opindex mabi-mmixware
10960 Generate code that passes function parameters and return values that (in
10961 the called function) are seen as registers @code{$0} and up, as opposed to
10962 the GNU ABI which uses global registers @code{$231} and up.
10964 @item -mzero-extend
10965 @itemx -mno-zero-extend
10966 @opindex mzero-extend
10967 @opindex mno-zero-extend
10968 When reading data from memory in sizes shorter than 64 bits, use (do not
10969 use) zero-extending load instructions by default, rather than
10970 sign-extending ones.
10973 @itemx -mno-knuthdiv
10975 @opindex mno-knuthdiv
10976 Make the result of a division yielding a remainder have the same sign as
10977 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10978 remainder follows the sign of the dividend. Both methods are
10979 arithmetically valid, the latter being almost exclusively used.
10981 @item -mtoplevel-symbols
10982 @itemx -mno-toplevel-symbols
10983 @opindex mtoplevel-symbols
10984 @opindex mno-toplevel-symbols
10985 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10986 code can be used with the @code{PREFIX} assembly directive.
10990 Generate an executable in the ELF format, rather than the default
10991 @samp{mmo} format used by the @command{mmix} simulator.
10993 @item -mbranch-predict
10994 @itemx -mno-branch-predict
10995 @opindex mbranch-predict
10996 @opindex mno-branch-predict
10997 Use (do not use) the probable-branch instructions, when static branch
10998 prediction indicates a probable branch.
11000 @item -mbase-addresses
11001 @itemx -mno-base-addresses
11002 @opindex mbase-addresses
11003 @opindex mno-base-addresses
11004 Generate (do not generate) code that uses @emph{base addresses}. Using a
11005 base address automatically generates a request (handled by the assembler
11006 and the linker) for a constant to be set up in a global register. The
11007 register is used for one or more base address requests within the range 0
11008 to 255 from the value held in the register. The generally leads to short
11009 and fast code, but the number of different data items that can be
11010 addressed is limited. This means that a program that uses lots of static
11011 data may require @option{-mno-base-addresses}.
11013 @item -msingle-exit
11014 @itemx -mno-single-exit
11015 @opindex msingle-exit
11016 @opindex mno-single-exit
11017 Force (do not force) generated code to have a single exit point in each
11021 @node MN10300 Options
11022 @subsection MN10300 Options
11023 @cindex MN10300 options
11025 These @option{-m} options are defined for Matsushita MN10300 architectures:
11030 Generate code to avoid bugs in the multiply instructions for the MN10300
11031 processors. This is the default.
11033 @item -mno-mult-bug
11034 @opindex mno-mult-bug
11035 Do not generate code to avoid bugs in the multiply instructions for the
11036 MN10300 processors.
11040 Generate code which uses features specific to the AM33 processor.
11044 Do not generate code which uses features specific to the AM33 processor. This
11047 @item -mreturn-pointer-on-d0
11048 @opindex mreturn-pointer-on-d0
11049 When generating a function which returns a pointer, return the pointer
11050 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
11051 only in a0, and attempts to call such functions without a prototype
11052 would result in errors. Note that this option is on by default; use
11053 @option{-mno-return-pointer-on-d0} to disable it.
11057 Do not link in the C run-time initialization object file.
11061 Indicate to the linker that it should perform a relaxation optimization pass
11062 to shorten branches, calls and absolute memory addresses. This option only
11063 has an effect when used on the command line for the final link step.
11065 This option makes symbolic debugging impossible.
11069 @subsection MT Options
11072 These @option{-m} options are defined for Morpho MT architectures:
11076 @item -march=@var{cpu-type}
11078 Generate code that will run on @var{cpu-type}, which is the name of a system
11079 representing a certain processor type. Possible values for
11080 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
11081 @samp{ms1-16-003} and @samp{ms2}.
11083 When this option is not used, the default is @option{-march=ms1-16-002}.
11087 Use byte loads and stores when generating code.
11091 Do not use byte loads and stores when generating code.
11095 Use simulator runtime
11099 Do not link in the C run-time initialization object file
11100 @file{crti.o}. Other run-time initialization and termination files
11101 such as @file{startup.o} and @file{exit.o} are still included on the
11102 linker command line.
11106 @node PDP-11 Options
11107 @subsection PDP-11 Options
11108 @cindex PDP-11 Options
11110 These options are defined for the PDP-11:
11115 Use hardware FPP floating point. This is the default. (FIS floating
11116 point on the PDP-11/40 is not supported.)
11119 @opindex msoft-float
11120 Do not use hardware floating point.
11124 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
11128 Return floating-point results in memory. This is the default.
11132 Generate code for a PDP-11/40.
11136 Generate code for a PDP-11/45. This is the default.
11140 Generate code for a PDP-11/10.
11142 @item -mbcopy-builtin
11143 @opindex bcopy-builtin
11144 Use inline @code{movmemhi} patterns for copying memory. This is the
11149 Do not use inline @code{movmemhi} patterns for copying memory.
11155 Use 16-bit @code{int}. This is the default.
11161 Use 32-bit @code{int}.
11164 @itemx -mno-float32
11166 @opindex mno-float32
11167 Use 64-bit @code{float}. This is the default.
11170 @itemx -mno-float64
11172 @opindex mno-float64
11173 Use 32-bit @code{float}.
11177 Use @code{abshi2} pattern. This is the default.
11181 Do not use @code{abshi2} pattern.
11183 @item -mbranch-expensive
11184 @opindex mbranch-expensive
11185 Pretend that branches are expensive. This is for experimenting with
11186 code generation only.
11188 @item -mbranch-cheap
11189 @opindex mbranch-cheap
11190 Do not pretend that branches are expensive. This is the default.
11194 Generate code for a system with split I&D@.
11198 Generate code for a system without split I&D@. This is the default.
11202 Use Unix assembler syntax. This is the default when configured for
11203 @samp{pdp11-*-bsd}.
11207 Use DEC assembler syntax. This is the default when configured for any
11208 PDP-11 target other than @samp{pdp11-*-bsd}.
11211 @node PowerPC Options
11212 @subsection PowerPC Options
11213 @cindex PowerPC options
11215 These are listed under @xref{RS/6000 and PowerPC Options}.
11217 @node RS/6000 and PowerPC Options
11218 @subsection IBM RS/6000 and PowerPC Options
11219 @cindex RS/6000 and PowerPC Options
11220 @cindex IBM RS/6000 and PowerPC Options
11222 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
11229 @itemx -mno-powerpc
11230 @itemx -mpowerpc-gpopt
11231 @itemx -mno-powerpc-gpopt
11232 @itemx -mpowerpc-gfxopt
11233 @itemx -mno-powerpc-gfxopt
11235 @itemx -mno-powerpc64
11239 @itemx -mno-popcntb
11245 @opindex mno-power2
11247 @opindex mno-powerpc
11248 @opindex mpowerpc-gpopt
11249 @opindex mno-powerpc-gpopt
11250 @opindex mpowerpc-gfxopt
11251 @opindex mno-powerpc-gfxopt
11252 @opindex mpowerpc64
11253 @opindex mno-powerpc64
11257 @opindex mno-popcntb
11260 GCC supports two related instruction set architectures for the
11261 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
11262 instructions supported by the @samp{rios} chip set used in the original
11263 RS/6000 systems and the @dfn{PowerPC} instruction set is the
11264 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
11265 the IBM 4xx, 6xx, and follow-on microprocessors.
11267 Neither architecture is a subset of the other. However there is a
11268 large common subset of instructions supported by both. An MQ
11269 register is included in processors supporting the POWER architecture.
11271 You use these options to specify which instructions are available on the
11272 processor you are using. The default value of these options is
11273 determined when configuring GCC@. Specifying the
11274 @option{-mcpu=@var{cpu_type}} overrides the specification of these
11275 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
11276 rather than the options listed above.
11278 The @option{-mpower} option allows GCC to generate instructions that
11279 are found only in the POWER architecture and to use the MQ register.
11280 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
11281 to generate instructions that are present in the POWER2 architecture but
11282 not the original POWER architecture.
11284 The @option{-mpowerpc} option allows GCC to generate instructions that
11285 are found only in the 32-bit subset of the PowerPC architecture.
11286 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
11287 GCC to use the optional PowerPC architecture instructions in the
11288 General Purpose group, including floating-point square root. Specifying
11289 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
11290 use the optional PowerPC architecture instructions in the Graphics
11291 group, including floating-point select.
11293 The @option{-mmfcrf} option allows GCC to generate the move from
11294 condition register field instruction implemented on the POWER4
11295 processor and other processors that support the PowerPC V2.01
11297 The @option{-mpopcntb} option allows GCC to generate the popcount and
11298 double precision FP reciprocal estimate instruction implemented on the
11299 POWER5 processor and other processors that support the PowerPC V2.02
11301 The @option{-mfprnd} option allows GCC to generate the FP round to
11302 integer instructions implemented on the POWER5+ processor and other
11303 processors that support the PowerPC V2.03 architecture.
11305 The @option{-mpowerpc64} option allows GCC to generate the additional
11306 64-bit instructions that are found in the full PowerPC64 architecture
11307 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
11308 @option{-mno-powerpc64}.
11310 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
11311 will use only the instructions in the common subset of both
11312 architectures plus some special AIX common-mode calls, and will not use
11313 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
11314 permits GCC to use any instruction from either architecture and to
11315 allow use of the MQ register; specify this for the Motorola MPC601.
11317 @item -mnew-mnemonics
11318 @itemx -mold-mnemonics
11319 @opindex mnew-mnemonics
11320 @opindex mold-mnemonics
11321 Select which mnemonics to use in the generated assembler code. With
11322 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
11323 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
11324 assembler mnemonics defined for the POWER architecture. Instructions
11325 defined in only one architecture have only one mnemonic; GCC uses that
11326 mnemonic irrespective of which of these options is specified.
11328 GCC defaults to the mnemonics appropriate for the architecture in
11329 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
11330 value of these option. Unless you are building a cross-compiler, you
11331 should normally not specify either @option{-mnew-mnemonics} or
11332 @option{-mold-mnemonics}, but should instead accept the default.
11334 @item -mcpu=@var{cpu_type}
11336 Set architecture type, register usage, choice of mnemonics, and
11337 instruction scheduling parameters for machine type @var{cpu_type}.
11338 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
11339 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
11340 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
11341 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
11342 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
11343 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
11344 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
11345 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6},
11346 @samp{common}, @samp{powerpc}, @samp{powerpc64},
11347 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
11349 @option{-mcpu=common} selects a completely generic processor. Code
11350 generated under this option will run on any POWER or PowerPC processor.
11351 GCC will use only the instructions in the common subset of both
11352 architectures, and will not use the MQ register. GCC assumes a generic
11353 processor model for scheduling purposes.
11355 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
11356 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
11357 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
11358 types, with an appropriate, generic processor model assumed for
11359 scheduling purposes.
11361 The other options specify a specific processor. Code generated under
11362 those options will run best on that processor, and may not run at all on
11365 The @option{-mcpu} options automatically enable or disable the
11366 following options: @option{-maltivec}, @option{-mfprnd},
11367 @option{-mhard-float}, @option{-mmfcrf}, @option{-mmultiple},
11368 @option{-mnew-mnemonics}, @option{-mpopcntb}, @option{-mpower},
11369 @option{-mpower2}, @option{-mpowerpc64}, @option{-mpowerpc-gpopt},
11370 @option{-mpowerpc-gfxopt}, @option{-mstring}, @option{-mmulhw}, @option{-mdlmzb}.
11371 The particular options
11372 set for any particular CPU will vary between compiler versions,
11373 depending on what setting seems to produce optimal code for that CPU;
11374 it doesn't necessarily reflect the actual hardware's capabilities. If
11375 you wish to set an individual option to a particular value, you may
11376 specify it after the @option{-mcpu} option, like @samp{-mcpu=970
11379 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
11380 not enabled or disabled by the @option{-mcpu} option at present because
11381 AIX does not have full support for these options. You may still
11382 enable or disable them individually if you're sure it'll work in your
11385 @item -mtune=@var{cpu_type}
11387 Set the instruction scheduling parameters for machine type
11388 @var{cpu_type}, but do not set the architecture type, register usage, or
11389 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
11390 values for @var{cpu_type} are used for @option{-mtune} as for
11391 @option{-mcpu}. If both are specified, the code generated will use the
11392 architecture, registers, and mnemonics set by @option{-mcpu}, but the
11393 scheduling parameters set by @option{-mtune}.
11399 Generate code to compute division as reciprocal estimate and iterative
11400 refinement, creating opportunities for increased throughput. This
11401 feature requires: optional PowerPC Graphics instruction set for single
11402 precision and FRE instruction for double precision, assuming divides
11403 cannot generate user-visible traps, and the domain values not include
11404 Infinities, denormals or zero denominator.
11407 @itemx -mno-altivec
11409 @opindex mno-altivec
11410 Generate code that uses (does not use) AltiVec instructions, and also
11411 enable the use of built-in functions that allow more direct access to
11412 the AltiVec instruction set. You may also need to set
11413 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
11419 @opindex mno-vrsave
11420 Generate VRSAVE instructions when generating AltiVec code.
11423 @opindex msecure-plt
11424 Generate code that allows ld and ld.so to build executables and shared
11425 libraries with non-exec .plt and .got sections. This is a PowerPC
11426 32-bit SYSV ABI option.
11430 Generate code that uses a BSS .plt section that ld.so fills in, and
11431 requires .plt and .got sections that are both writable and executable.
11432 This is a PowerPC 32-bit SYSV ABI option.
11438 This switch enables or disables the generation of ISEL instructions.
11440 @item -misel=@var{yes/no}
11441 This switch has been deprecated. Use @option{-misel} and
11442 @option{-mno-isel} instead.
11448 This switch enables or disables the generation of SPE simd
11451 @item -mspe=@var{yes/no}
11452 This option has been deprecated. Use @option{-mspe} and
11453 @option{-mno-spe} instead.
11455 @item -mfloat-gprs=@var{yes/single/double/no}
11456 @itemx -mfloat-gprs
11457 @opindex mfloat-gprs
11458 This switch enables or disables the generation of floating point
11459 operations on the general purpose registers for architectures that
11462 The argument @var{yes} or @var{single} enables the use of
11463 single-precision floating point operations.
11465 The argument @var{double} enables the use of single and
11466 double-precision floating point operations.
11468 The argument @var{no} disables floating point operations on the
11469 general purpose registers.
11471 This option is currently only available on the MPC854x.
11477 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
11478 targets (including GNU/Linux). The 32-bit environment sets int, long
11479 and pointer to 32 bits and generates code that runs on any PowerPC
11480 variant. The 64-bit environment sets int to 32 bits and long and
11481 pointer to 64 bits, and generates code for PowerPC64, as for
11482 @option{-mpowerpc64}.
11485 @itemx -mno-fp-in-toc
11486 @itemx -mno-sum-in-toc
11487 @itemx -mminimal-toc
11489 @opindex mno-fp-in-toc
11490 @opindex mno-sum-in-toc
11491 @opindex mminimal-toc
11492 Modify generation of the TOC (Table Of Contents), which is created for
11493 every executable file. The @option{-mfull-toc} option is selected by
11494 default. In that case, GCC will allocate at least one TOC entry for
11495 each unique non-automatic variable reference in your program. GCC
11496 will also place floating-point constants in the TOC@. However, only
11497 16,384 entries are available in the TOC@.
11499 If you receive a linker error message that saying you have overflowed
11500 the available TOC space, you can reduce the amount of TOC space used
11501 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
11502 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
11503 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
11504 generate code to calculate the sum of an address and a constant at
11505 run-time instead of putting that sum into the TOC@. You may specify one
11506 or both of these options. Each causes GCC to produce very slightly
11507 slower and larger code at the expense of conserving TOC space.
11509 If you still run out of space in the TOC even when you specify both of
11510 these options, specify @option{-mminimal-toc} instead. This option causes
11511 GCC to make only one TOC entry for every file. When you specify this
11512 option, GCC will produce code that is slower and larger but which
11513 uses extremely little TOC space. You may wish to use this option
11514 only on files that contain less frequently executed code.
11520 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
11521 @code{long} type, and the infrastructure needed to support them.
11522 Specifying @option{-maix64} implies @option{-mpowerpc64} and
11523 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
11524 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
11527 @itemx -mno-xl-compat
11528 @opindex mxl-compat
11529 @opindex mno-xl-compat
11530 Produce code that conforms more closely to IBM XL compiler semantics
11531 when using AIX-compatible ABI. Pass floating-point arguments to
11532 prototyped functions beyond the register save area (RSA) on the stack
11533 in addition to argument FPRs. Do not assume that most significant
11534 double in 128-bit long double value is properly rounded when comparing
11535 values and converting to double. Use XL symbol names for long double
11538 The AIX calling convention was extended but not initially documented to
11539 handle an obscure K&R C case of calling a function that takes the
11540 address of its arguments with fewer arguments than declared. IBM XL
11541 compilers access floating point arguments which do not fit in the
11542 RSA from the stack when a subroutine is compiled without
11543 optimization. Because always storing floating-point arguments on the
11544 stack is inefficient and rarely needed, this option is not enabled by
11545 default and only is necessary when calling subroutines compiled by IBM
11546 XL compilers without optimization.
11550 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
11551 application written to use message passing with special startup code to
11552 enable the application to run. The system must have PE installed in the
11553 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
11554 must be overridden with the @option{-specs=} option to specify the
11555 appropriate directory location. The Parallel Environment does not
11556 support threads, so the @option{-mpe} option and the @option{-pthread}
11557 option are incompatible.
11559 @item -malign-natural
11560 @itemx -malign-power
11561 @opindex malign-natural
11562 @opindex malign-power
11563 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
11564 @option{-malign-natural} overrides the ABI-defined alignment of larger
11565 types, such as floating-point doubles, on their natural size-based boundary.
11566 The option @option{-malign-power} instructs GCC to follow the ABI-specified
11567 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
11569 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
11573 @itemx -mhard-float
11574 @opindex msoft-float
11575 @opindex mhard-float
11576 Generate code that does not use (uses) the floating-point register set.
11577 Software floating point emulation is provided if you use the
11578 @option{-msoft-float} option, and pass the option to GCC when linking.
11581 @itemx -mno-multiple
11583 @opindex mno-multiple
11584 Generate code that uses (does not use) the load multiple word
11585 instructions and the store multiple word instructions. These
11586 instructions are generated by default on POWER systems, and not
11587 generated on PowerPC systems. Do not use @option{-mmultiple} on little
11588 endian PowerPC systems, since those instructions do not work when the
11589 processor is in little endian mode. The exceptions are PPC740 and
11590 PPC750 which permit the instructions usage in little endian mode.
11595 @opindex mno-string
11596 Generate code that uses (does not use) the load string instructions
11597 and the store string word instructions to save multiple registers and
11598 do small block moves. These instructions are generated by default on
11599 POWER systems, and not generated on PowerPC systems. Do not use
11600 @option{-mstring} on little endian PowerPC systems, since those
11601 instructions do not work when the processor is in little endian mode.
11602 The exceptions are PPC740 and PPC750 which permit the instructions
11603 usage in little endian mode.
11608 @opindex mno-update
11609 Generate code that uses (does not use) the load or store instructions
11610 that update the base register to the address of the calculated memory
11611 location. These instructions are generated by default. If you use
11612 @option{-mno-update}, there is a small window between the time that the
11613 stack pointer is updated and the address of the previous frame is
11614 stored, which means code that walks the stack frame across interrupts or
11615 signals may get corrupted data.
11618 @itemx -mno-fused-madd
11619 @opindex mfused-madd
11620 @opindex mno-fused-madd
11621 Generate code that uses (does not use) the floating point multiply and
11622 accumulate instructions. These instructions are generated by default if
11623 hardware floating is used.
11629 Generate code that uses (does not use) the half-word multiply and
11630 multiply-accumulate instructions on the IBM 405 and 440 processors.
11631 These instructions are generated by default when targetting those
11638 Generate code that uses (does not use) the string-search @samp{dlmzb}
11639 instruction on the IBM 405 and 440 processors. This instruction is
11640 generated by default when targetting those processors.
11642 @item -mno-bit-align
11644 @opindex mno-bit-align
11645 @opindex mbit-align
11646 On System V.4 and embedded PowerPC systems do not (do) force structures
11647 and unions that contain bit-fields to be aligned to the base type of the
11650 For example, by default a structure containing nothing but 8
11651 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
11652 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
11653 the structure would be aligned to a 1 byte boundary and be one byte in
11656 @item -mno-strict-align
11657 @itemx -mstrict-align
11658 @opindex mno-strict-align
11659 @opindex mstrict-align
11660 On System V.4 and embedded PowerPC systems do not (do) assume that
11661 unaligned memory references will be handled by the system.
11663 @item -mrelocatable
11664 @itemx -mno-relocatable
11665 @opindex mrelocatable
11666 @opindex mno-relocatable
11667 On embedded PowerPC systems generate code that allows (does not allow)
11668 the program to be relocated to a different address at runtime. If you
11669 use @option{-mrelocatable} on any module, all objects linked together must
11670 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
11672 @item -mrelocatable-lib
11673 @itemx -mno-relocatable-lib
11674 @opindex mrelocatable-lib
11675 @opindex mno-relocatable-lib
11676 On embedded PowerPC systems generate code that allows (does not allow)
11677 the program to be relocated to a different address at runtime. Modules
11678 compiled with @option{-mrelocatable-lib} can be linked with either modules
11679 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
11680 with modules compiled with the @option{-mrelocatable} options.
11686 On System V.4 and embedded PowerPC systems do not (do) assume that
11687 register 2 contains a pointer to a global area pointing to the addresses
11688 used in the program.
11691 @itemx -mlittle-endian
11693 @opindex mlittle-endian
11694 On System V.4 and embedded PowerPC systems compile code for the
11695 processor in little endian mode. The @option{-mlittle-endian} option is
11696 the same as @option{-mlittle}.
11699 @itemx -mbig-endian
11701 @opindex mbig-endian
11702 On System V.4 and embedded PowerPC systems compile code for the
11703 processor in big endian mode. The @option{-mbig-endian} option is
11704 the same as @option{-mbig}.
11706 @item -mdynamic-no-pic
11707 @opindex mdynamic-no-pic
11708 On Darwin and Mac OS X systems, compile code so that it is not
11709 relocatable, but that its external references are relocatable. The
11710 resulting code is suitable for applications, but not shared
11713 @item -mprioritize-restricted-insns=@var{priority}
11714 @opindex mprioritize-restricted-insns
11715 This option controls the priority that is assigned to
11716 dispatch-slot restricted instructions during the second scheduling
11717 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
11718 @var{no/highest/second-highest} priority to dispatch slot restricted
11721 @item -msched-costly-dep=@var{dependence_type}
11722 @opindex msched-costly-dep
11723 This option controls which dependences are considered costly
11724 by the target during instruction scheduling. The argument
11725 @var{dependence_type} takes one of the following values:
11726 @var{no}: no dependence is costly,
11727 @var{all}: all dependences are costly,
11728 @var{true_store_to_load}: a true dependence from store to load is costly,
11729 @var{store_to_load}: any dependence from store to load is costly,
11730 @var{number}: any dependence which latency >= @var{number} is costly.
11732 @item -minsert-sched-nops=@var{scheme}
11733 @opindex minsert-sched-nops
11734 This option controls which nop insertion scheme will be used during
11735 the second scheduling pass. The argument @var{scheme} takes one of the
11737 @var{no}: Don't insert nops.
11738 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
11739 according to the scheduler's grouping.
11740 @var{regroup_exact}: Insert nops to force costly dependent insns into
11741 separate groups. Insert exactly as many nops as needed to force an insn
11742 to a new group, according to the estimated processor grouping.
11743 @var{number}: Insert nops to force costly dependent insns into
11744 separate groups. Insert @var{number} nops to force an insn to a new group.
11747 @opindex mcall-sysv
11748 On System V.4 and embedded PowerPC systems compile code using calling
11749 conventions that adheres to the March 1995 draft of the System V
11750 Application Binary Interface, PowerPC processor supplement. This is the
11751 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
11753 @item -mcall-sysv-eabi
11754 @opindex mcall-sysv-eabi
11755 Specify both @option{-mcall-sysv} and @option{-meabi} options.
11757 @item -mcall-sysv-noeabi
11758 @opindex mcall-sysv-noeabi
11759 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
11761 @item -mcall-solaris
11762 @opindex mcall-solaris
11763 On System V.4 and embedded PowerPC systems compile code for the Solaris
11767 @opindex mcall-linux
11768 On System V.4 and embedded PowerPC systems compile code for the
11769 Linux-based GNU system.
11773 On System V.4 and embedded PowerPC systems compile code for the
11774 Hurd-based GNU system.
11776 @item -mcall-netbsd
11777 @opindex mcall-netbsd
11778 On System V.4 and embedded PowerPC systems compile code for the
11779 NetBSD operating system.
11781 @item -maix-struct-return
11782 @opindex maix-struct-return
11783 Return all structures in memory (as specified by the AIX ABI)@.
11785 @item -msvr4-struct-return
11786 @opindex msvr4-struct-return
11787 Return structures smaller than 8 bytes in registers (as specified by the
11790 @item -mabi=@var{abi-type}
11792 Extend the current ABI with a particular extension, or remove such extension.
11793 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
11794 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
11798 Extend the current ABI with SPE ABI extensions. This does not change
11799 the default ABI, instead it adds the SPE ABI extensions to the current
11803 @opindex mabi=no-spe
11804 Disable Booke SPE ABI extensions for the current ABI@.
11806 @item -mabi=ibmlongdouble
11807 @opindex mabi=ibmlongdouble
11808 Change the current ABI to use IBM extended precision long double.
11809 This is a PowerPC 32-bit SYSV ABI option.
11811 @item -mabi=ieeelongdouble
11812 @opindex mabi=ieeelongdouble
11813 Change the current ABI to use IEEE extended precision long double.
11814 This is a PowerPC 32-bit Linux ABI option.
11817 @itemx -mno-prototype
11818 @opindex mprototype
11819 @opindex mno-prototype
11820 On System V.4 and embedded PowerPC systems assume that all calls to
11821 variable argument functions are properly prototyped. Otherwise, the
11822 compiler must insert an instruction before every non prototyped call to
11823 set or clear bit 6 of the condition code register (@var{CR}) to
11824 indicate whether floating point values were passed in the floating point
11825 registers in case the function takes a variable arguments. With
11826 @option{-mprototype}, only calls to prototyped variable argument functions
11827 will set or clear the bit.
11831 On embedded PowerPC systems, assume that the startup module is called
11832 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
11833 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
11838 On embedded PowerPC systems, assume that the startup module is called
11839 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
11844 On embedded PowerPC systems, assume that the startup module is called
11845 @file{crt0.o} and the standard C libraries are @file{libads.a} and
11848 @item -myellowknife
11849 @opindex myellowknife
11850 On embedded PowerPC systems, assume that the startup module is called
11851 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
11856 On System V.4 and embedded PowerPC systems, specify that you are
11857 compiling for a VxWorks system.
11861 Specify that you are compiling for the WindISS simulation environment.
11865 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
11866 header to indicate that @samp{eabi} extended relocations are used.
11872 On System V.4 and embedded PowerPC systems do (do not) adhere to the
11873 Embedded Applications Binary Interface (eabi) which is a set of
11874 modifications to the System V.4 specifications. Selecting @option{-meabi}
11875 means that the stack is aligned to an 8 byte boundary, a function
11876 @code{__eabi} is called to from @code{main} to set up the eabi
11877 environment, and the @option{-msdata} option can use both @code{r2} and
11878 @code{r13} to point to two separate small data areas. Selecting
11879 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
11880 do not call an initialization function from @code{main}, and the
11881 @option{-msdata} option will only use @code{r13} to point to a single
11882 small data area. The @option{-meabi} option is on by default if you
11883 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
11886 @opindex msdata=eabi
11887 On System V.4 and embedded PowerPC systems, put small initialized
11888 @code{const} global and static data in the @samp{.sdata2} section, which
11889 is pointed to by register @code{r2}. Put small initialized
11890 non-@code{const} global and static data in the @samp{.sdata} section,
11891 which is pointed to by register @code{r13}. Put small uninitialized
11892 global and static data in the @samp{.sbss} section, which is adjacent to
11893 the @samp{.sdata} section. The @option{-msdata=eabi} option is
11894 incompatible with the @option{-mrelocatable} option. The
11895 @option{-msdata=eabi} option also sets the @option{-memb} option.
11898 @opindex msdata=sysv
11899 On System V.4 and embedded PowerPC systems, put small global and static
11900 data in the @samp{.sdata} section, which is pointed to by register
11901 @code{r13}. Put small uninitialized global and static data in the
11902 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11903 The @option{-msdata=sysv} option is incompatible with the
11904 @option{-mrelocatable} option.
11906 @item -msdata=default
11908 @opindex msdata=default
11910 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
11911 compile code the same as @option{-msdata=eabi}, otherwise compile code the
11912 same as @option{-msdata=sysv}.
11915 @opindex msdata-data
11916 On System V.4 and embedded PowerPC systems, put small global
11917 data in the @samp{.sdata} section. Put small uninitialized global
11918 data in the @samp{.sbss} section. Do not use register @code{r13}
11919 to address small data however. This is the default behavior unless
11920 other @option{-msdata} options are used.
11924 @opindex msdata=none
11926 On embedded PowerPC systems, put all initialized global and static data
11927 in the @samp{.data} section, and all uninitialized data in the
11928 @samp{.bss} section.
11932 @cindex smaller data references (PowerPC)
11933 @cindex .sdata/.sdata2 references (PowerPC)
11934 On embedded PowerPC systems, put global and static items less than or
11935 equal to @var{num} bytes into the small data or bss sections instead of
11936 the normal data or bss section. By default, @var{num} is 8. The
11937 @option{-G @var{num}} switch is also passed to the linker.
11938 All modules should be compiled with the same @option{-G @var{num}} value.
11941 @itemx -mno-regnames
11943 @opindex mno-regnames
11944 On System V.4 and embedded PowerPC systems do (do not) emit register
11945 names in the assembly language output using symbolic forms.
11948 @itemx -mno-longcall
11950 @opindex mno-longcall
11951 By default assume that all calls are far away so that a longer more
11952 expensive calling sequence is required. This is required for calls
11953 further than 32 megabytes (33,554,432 bytes) from the current location.
11954 A short call will be generated if the compiler knows
11955 the call cannot be that far away. This setting can be overridden by
11956 the @code{shortcall} function attribute, or by @code{#pragma
11959 Some linkers are capable of detecting out-of-range calls and generating
11960 glue code on the fly. On these systems, long calls are unnecessary and
11961 generate slower code. As of this writing, the AIX linker can do this,
11962 as can the GNU linker for PowerPC/64. It is planned to add this feature
11963 to the GNU linker for 32-bit PowerPC systems as well.
11965 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11966 callee, L42'', plus a ``branch island'' (glue code). The two target
11967 addresses represent the callee and the ``branch island''. The
11968 Darwin/PPC linker will prefer the first address and generate a ``bl
11969 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11970 otherwise, the linker will generate ``bl L42'' to call the ``branch
11971 island''. The ``branch island'' is appended to the body of the
11972 calling function; it computes the full 32-bit address of the callee
11975 On Mach-O (Darwin) systems, this option directs the compiler emit to
11976 the glue for every direct call, and the Darwin linker decides whether
11977 to use or discard it.
11979 In the future, we may cause GCC to ignore all longcall specifications
11980 when the linker is known to generate glue.
11984 Adds support for multithreading with the @dfn{pthreads} library.
11985 This option sets flags for both the preprocessor and linker.
11989 @node S/390 and zSeries Options
11990 @subsection S/390 and zSeries Options
11991 @cindex S/390 and zSeries Options
11993 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11997 @itemx -msoft-float
11998 @opindex mhard-float
11999 @opindex msoft-float
12000 Use (do not use) the hardware floating-point instructions and registers
12001 for floating-point operations. When @option{-msoft-float} is specified,
12002 functions in @file{libgcc.a} will be used to perform floating-point
12003 operations. When @option{-mhard-float} is specified, the compiler
12004 generates IEEE floating-point instructions. This is the default.
12006 @item -mlong-double-64
12007 @itemx -mlong-double-128
12008 @opindex mlong-double-64
12009 @opindex mlong-double-128
12010 These switches control the size of @code{long double} type. A size
12011 of 64bit makes the @code{long double} type equivalent to the @code{double}
12012 type. This is the default.
12015 @itemx -mno-backchain
12016 @opindex mbackchain
12017 @opindex mno-backchain
12018 Store (do not store) the address of the caller's frame as backchain pointer
12019 into the callee's stack frame.
12020 A backchain may be needed to allow debugging using tools that do not understand
12021 DWARF-2 call frame information.
12022 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
12023 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
12024 the backchain is placed into the topmost word of the 96/160 byte register
12027 In general, code compiled with @option{-mbackchain} is call-compatible with
12028 code compiled with @option{-mmo-backchain}; however, use of the backchain
12029 for debugging purposes usually requires that the whole binary is built with
12030 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
12031 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12032 to build a linux kernel use @option{-msoft-float}.
12034 The default is to not maintain the backchain.
12036 @item -mpacked-stack
12037 @item -mno-packed-stack
12038 @opindex mpacked-stack
12039 @opindex mno-packed-stack
12040 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
12041 specified, the compiler uses the all fields of the 96/160 byte register save
12042 area only for their default purpose; unused fields still take up stack space.
12043 When @option{-mpacked-stack} is specified, register save slots are densely
12044 packed at the top of the register save area; unused space is reused for other
12045 purposes, allowing for more efficient use of the available stack space.
12046 However, when @option{-mbackchain} is also in effect, the topmost word of
12047 the save area is always used to store the backchain, and the return address
12048 register is always saved two words below the backchain.
12050 As long as the stack frame backchain is not used, code generated with
12051 @option{-mpacked-stack} is call-compatible with code generated with
12052 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
12053 S/390 or zSeries generated code that uses the stack frame backchain at run
12054 time, not just for debugging purposes. Such code is not call-compatible
12055 with code compiled with @option{-mpacked-stack}. Also, note that the
12056 combination of @option{-mbackchain},
12057 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12058 to build a linux kernel use @option{-msoft-float}.
12060 The default is to not use the packed stack layout.
12063 @itemx -mno-small-exec
12064 @opindex msmall-exec
12065 @opindex mno-small-exec
12066 Generate (or do not generate) code using the @code{bras} instruction
12067 to do subroutine calls.
12068 This only works reliably if the total executable size does not
12069 exceed 64k. The default is to use the @code{basr} instruction instead,
12070 which does not have this limitation.
12076 When @option{-m31} is specified, generate code compliant to the
12077 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
12078 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
12079 particular to generate 64-bit instructions. For the @samp{s390}
12080 targets, the default is @option{-m31}, while the @samp{s390x}
12081 targets default to @option{-m64}.
12087 When @option{-mzarch} is specified, generate code using the
12088 instructions available on z/Architecture.
12089 When @option{-mesa} is specified, generate code using the
12090 instructions available on ESA/390. Note that @option{-mesa} is
12091 not possible with @option{-m64}.
12092 When generating code compliant to the GNU/Linux for S/390 ABI,
12093 the default is @option{-mesa}. When generating code compliant
12094 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
12100 Generate (or do not generate) code using the @code{mvcle} instruction
12101 to perform block moves. When @option{-mno-mvcle} is specified,
12102 use a @code{mvc} loop instead. This is the default unless optimizing for
12109 Print (or do not print) additional debug information when compiling.
12110 The default is to not print debug information.
12112 @item -march=@var{cpu-type}
12114 Generate code that will run on @var{cpu-type}, which is the name of a system
12115 representing a certain processor type. Possible values for
12116 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
12117 When generating code using the instructions available on z/Architecture,
12118 the default is @option{-march=z900}. Otherwise, the default is
12119 @option{-march=g5}.
12121 @item -mtune=@var{cpu-type}
12123 Tune to @var{cpu-type} everything applicable about the generated code,
12124 except for the ABI and the set of available instructions.
12125 The list of @var{cpu-type} values is the same as for @option{-march}.
12126 The default is the value used for @option{-march}.
12129 @itemx -mno-tpf-trace
12130 @opindex mtpf-trace
12131 @opindex mno-tpf-trace
12132 Generate code that adds (does not add) in TPF OS specific branches to trace
12133 routines in the operating system. This option is off by default, even
12134 when compiling for the TPF OS@.
12137 @itemx -mno-fused-madd
12138 @opindex mfused-madd
12139 @opindex mno-fused-madd
12140 Generate code that uses (does not use) the floating point multiply and
12141 accumulate instructions. These instructions are generated by default if
12142 hardware floating point is used.
12144 @item -mwarn-framesize=@var{framesize}
12145 @opindex mwarn-framesize
12146 Emit a warning if the current function exceeds the given frame size. Because
12147 this is a compile time check it doesn't need to be a real problem when the program
12148 runs. It is intended to identify functions which most probably cause
12149 a stack overflow. It is useful to be used in an environment with limited stack
12150 size e.g.@: the linux kernel.
12152 @item -mwarn-dynamicstack
12153 @opindex mwarn-dynamicstack
12154 Emit a warning if the function calls alloca or uses dynamically
12155 sized arrays. This is generally a bad idea with a limited stack size.
12157 @item -mstack-guard=@var{stack-guard}
12158 @item -mstack-size=@var{stack-size}
12159 @opindex mstack-guard
12160 @opindex mstack-size
12161 These arguments always have to be used in conjunction. If they are present the s390
12162 back end emits additional instructions in the function prologue which trigger a trap
12163 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
12164 (remember that the stack on s390 grows downward). These options are intended to
12165 be used to help debugging stack overflow problems. The additionally emitted code
12166 causes only little overhead and hence can also be used in production like systems
12167 without greater performance degradation. The given values have to be exact
12168 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard} without
12170 In order to be efficient the extra code makes the assumption that the stack starts
12171 at an address aligned to the value given by @var{stack-size}.
12174 @node Score Options
12175 @subsection Score Options
12176 @cindex Score Options
12178 These options are defined for Score implementations:
12183 Compile code for big endian mode. This is the default.
12187 Compile code for little endian mode.
12191 Disable generate bcnz instruction.
12195 Enable generate unaligned load and store instruction.
12199 Enable the use of multiply-accumulate instructions. Disabled by default.
12203 Specify the SCORE5 as the target architecture.
12207 Specify the SCORE5U of the target architecture.
12211 Specify the SCORE7 as the target architecture. This is the default.
12215 Specify the SCORE7D as the target architecture.
12219 @subsection SH Options
12221 These @samp{-m} options are defined for the SH implementations:
12226 Generate code for the SH1.
12230 Generate code for the SH2.
12233 Generate code for the SH2e.
12237 Generate code for the SH3.
12241 Generate code for the SH3e.
12245 Generate code for the SH4 without a floating-point unit.
12247 @item -m4-single-only
12248 @opindex m4-single-only
12249 Generate code for the SH4 with a floating-point unit that only
12250 supports single-precision arithmetic.
12254 Generate code for the SH4 assuming the floating-point unit is in
12255 single-precision mode by default.
12259 Generate code for the SH4.
12263 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
12264 floating-point unit is not used.
12266 @item -m4a-single-only
12267 @opindex m4a-single-only
12268 Generate code for the SH4a, in such a way that no double-precision
12269 floating point operations are used.
12272 @opindex m4a-single
12273 Generate code for the SH4a assuming the floating-point unit is in
12274 single-precision mode by default.
12278 Generate code for the SH4a.
12282 Same as @option{-m4a-nofpu}, except that it implicitly passes
12283 @option{-dsp} to the assembler. GCC doesn't generate any DSP
12284 instructions at the moment.
12288 Compile code for the processor in big endian mode.
12292 Compile code for the processor in little endian mode.
12296 Align doubles at 64-bit boundaries. Note that this changes the calling
12297 conventions, and thus some functions from the standard C library will
12298 not work unless you recompile it first with @option{-mdalign}.
12302 Shorten some address references at link time, when possible; uses the
12303 linker option @option{-relax}.
12307 Use 32-bit offsets in @code{switch} tables. The default is to use
12312 Enable the use of the instruction @code{fmovd}.
12316 Comply with the calling conventions defined by Renesas.
12320 Comply with the calling conventions defined by Renesas.
12324 Comply with the calling conventions defined for GCC before the Renesas
12325 conventions were available. This option is the default for all
12326 targets of the SH toolchain except for @samp{sh-symbianelf}.
12329 @opindex mnomacsave
12330 Mark the @code{MAC} register as call-clobbered, even if
12331 @option{-mhitachi} is given.
12335 Increase IEEE-compliance of floating-point code.
12336 At the moment, this is equivalent to @option{-fno-finite-math-only}.
12337 When generating 16 bit SH opcodes, getting IEEE-conforming results for
12338 comparisons of NANs / infinities incurs extra overhead in every
12339 floating point comparison, therefore the default is set to
12340 @option{-ffinite-math-only}.
12344 Dump instruction size and location in the assembly code.
12347 @opindex mpadstruct
12348 This option is deprecated. It pads structures to multiple of 4 bytes,
12349 which is incompatible with the SH ABI@.
12353 Optimize for space instead of speed. Implied by @option{-Os}.
12356 @opindex mprefergot
12357 When generating position-independent code, emit function calls using
12358 the Global Offset Table instead of the Procedure Linkage Table.
12362 Generate a library function call to invalidate instruction cache
12363 entries, after fixing up a trampoline. This library function call
12364 doesn't assume it can write to the whole memory address space. This
12365 is the default when the target is @code{sh-*-linux*}.
12367 @item -multcost=@var{number}
12368 @opindex multcost=@var{number}
12369 Set the cost to assume for a multiply insn.
12371 @item -mdiv=@var{strategy}
12372 @opindex mdiv=@var{strategy}
12373 Set the division strategy to use for SHmedia code. @var{strategy} must be
12374 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
12375 inv:call2, inv:fp .
12376 "fp" performs the operation in floating point. This has a very high latency,
12377 but needs only a few instructions, so it might be a good choice if
12378 your code has enough easily exploitable ILP to allow the compiler to
12379 schedule the floating point instructions together with other instructions.
12380 Division by zero causes a floating point exception.
12381 "inv" uses integer operations to calculate the inverse of the divisor,
12382 and then multiplies the dividend with the inverse. This strategy allows
12383 cse and hoisting of the inverse calculation. Division by zero calculates
12384 an unspecified result, but does not trap.
12385 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
12386 have been found, or if the entire operation has been hoisted to the same
12387 place, the last stages of the inverse calculation are intertwined with the
12388 final multiply to reduce the overall latency, at the expense of using a few
12389 more instructions, and thus offering fewer scheduling opportunities with
12391 "call" calls a library function that usually implements the inv:minlat
12393 This gives high code density for m5-*media-nofpu compilations.
12394 "call2" uses a different entry point of the same library function, where it
12395 assumes that a pointer to a lookup table has already been set up, which
12396 exposes the pointer load to cse / code hoisting optimizations.
12397 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
12398 code generation, but if the code stays unoptimized, revert to the "call",
12399 "call2", or "fp" strategies, respectively. Note that the
12400 potentially-trapping side effect of division by zero is carried by a
12401 separate instruction, so it is possible that all the integer instructions
12402 are hoisted out, but the marker for the side effect stays where it is.
12403 A recombination to fp operations or a call is not possible in that case.
12404 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
12405 that the inverse calculation was nor separated from the multiply, they speed
12406 up division where the dividend fits into 20 bits (plus sign where applicable),
12407 by inserting a test to skip a number of operations in this case; this test
12408 slows down the case of larger dividends. inv20u assumes the case of a such
12409 a small dividend to be unlikely, and inv20l assumes it to be likely.
12411 @item -mdivsi3_libfunc=@var{name}
12412 @opindex mdivsi3_libfunc=@var{name}
12413 Set the name of the library function used for 32 bit signed division to
12414 @var{name}. This only affect the name used in the call and inv:call
12415 division strategies, and the compiler will still expect the same
12416 sets of input/output/clobbered registers as if this option was not present.
12418 @item -madjust-unroll
12419 @opindex madjust-unroll
12420 Throttle unrolling to avoid thrashing target registers.
12421 This option only has an effect if the gcc code base supports the
12422 TARGET_ADJUST_UNROLL_MAX target hook.
12424 @item -mindexed-addressing
12425 @opindex mindexed-addressing
12426 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
12427 This is only safe if the hardware and/or OS implement 32 bit wrap-around
12428 semantics for the indexed addressing mode. The architecture allows the
12429 implementation of processors with 64 bit MMU, which the OS could use to
12430 get 32 bit addressing, but since no current hardware implementation supports
12431 this or any other way to make the indexed addressing mode safe to use in
12432 the 32 bit ABI, the default is -mno-indexed-addressing.
12434 @item -mgettrcost=@var{number}
12435 @opindex mgettrcost=@var{number}
12436 Set the cost assumed for the gettr instruction to @var{number}.
12437 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
12441 Assume pt* instructions won't trap. This will generally generate better
12442 scheduled code, but is unsafe on current hardware. The current architecture
12443 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
12444 This has the unintentional effect of making it unsafe to schedule ptabs /
12445 ptrel before a branch, or hoist it out of a loop. For example,
12446 __do_global_ctors, a part of libgcc that runs constructors at program
12447 startup, calls functions in a list which is delimited by -1. With the
12448 -mpt-fixed option, the ptabs will be done before testing against -1.
12449 That means that all the constructors will be run a bit quicker, but when
12450 the loop comes to the end of the list, the program crashes because ptabs
12451 loads -1 into a target register. Since this option is unsafe for any
12452 hardware implementing the current architecture specification, the default
12453 is -mno-pt-fixed. Unless the user specifies a specific cost with
12454 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
12455 this deters register allocation using target registers for storing
12458 @item -minvalid-symbols
12459 @opindex minvalid-symbols
12460 Assume symbols might be invalid. Ordinary function symbols generated by
12461 the compiler will always be valid to load with movi/shori/ptabs or
12462 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
12463 to generate symbols that will cause ptabs / ptrel to trap.
12464 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
12465 It will then prevent cross-basic-block cse, hoisting and most scheduling
12466 of symbol loads. The default is @option{-mno-invalid-symbols}.
12469 @node SPARC Options
12470 @subsection SPARC Options
12471 @cindex SPARC options
12473 These @samp{-m} options are supported on the SPARC:
12476 @item -mno-app-regs
12478 @opindex mno-app-regs
12480 Specify @option{-mapp-regs} to generate output using the global registers
12481 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
12484 To be fully SVR4 ABI compliant at the cost of some performance loss,
12485 specify @option{-mno-app-regs}. You should compile libraries and system
12486 software with this option.
12489 @itemx -mhard-float
12491 @opindex mhard-float
12492 Generate output containing floating point instructions. This is the
12496 @itemx -msoft-float
12498 @opindex msoft-float
12499 Generate output containing library calls for floating point.
12500 @strong{Warning:} the requisite libraries are not available for all SPARC
12501 targets. Normally the facilities of the machine's usual C compiler are
12502 used, but this cannot be done directly in cross-compilation. You must make
12503 your own arrangements to provide suitable library functions for
12504 cross-compilation. The embedded targets @samp{sparc-*-aout} and
12505 @samp{sparclite-*-*} do provide software floating point support.
12507 @option{-msoft-float} changes the calling convention in the output file;
12508 therefore, it is only useful if you compile @emph{all} of a program with
12509 this option. In particular, you need to compile @file{libgcc.a}, the
12510 library that comes with GCC, with @option{-msoft-float} in order for
12513 @item -mhard-quad-float
12514 @opindex mhard-quad-float
12515 Generate output containing quad-word (long double) floating point
12518 @item -msoft-quad-float
12519 @opindex msoft-quad-float
12520 Generate output containing library calls for quad-word (long double)
12521 floating point instructions. The functions called are those specified
12522 in the SPARC ABI@. This is the default.
12524 As of this writing, there are no SPARC implementations that have hardware
12525 support for the quad-word floating point instructions. They all invoke
12526 a trap handler for one of these instructions, and then the trap handler
12527 emulates the effect of the instruction. Because of the trap handler overhead,
12528 this is much slower than calling the ABI library routines. Thus the
12529 @option{-msoft-quad-float} option is the default.
12531 @item -mno-unaligned-doubles
12532 @itemx -munaligned-doubles
12533 @opindex mno-unaligned-doubles
12534 @opindex munaligned-doubles
12535 Assume that doubles have 8 byte alignment. This is the default.
12537 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
12538 alignment only if they are contained in another type, or if they have an
12539 absolute address. Otherwise, it assumes they have 4 byte alignment.
12540 Specifying this option avoids some rare compatibility problems with code
12541 generated by other compilers. It is not the default because it results
12542 in a performance loss, especially for floating point code.
12544 @item -mno-faster-structs
12545 @itemx -mfaster-structs
12546 @opindex mno-faster-structs
12547 @opindex mfaster-structs
12548 With @option{-mfaster-structs}, the compiler assumes that structures
12549 should have 8 byte alignment. This enables the use of pairs of
12550 @code{ldd} and @code{std} instructions for copies in structure
12551 assignment, in place of twice as many @code{ld} and @code{st} pairs.
12552 However, the use of this changed alignment directly violates the SPARC
12553 ABI@. Thus, it's intended only for use on targets where the developer
12554 acknowledges that their resulting code will not be directly in line with
12555 the rules of the ABI@.
12557 @item -mimpure-text
12558 @opindex mimpure-text
12559 @option{-mimpure-text}, used in addition to @option{-shared}, tells
12560 the compiler to not pass @option{-z text} to the linker when linking a
12561 shared object. Using this option, you can link position-dependent
12562 code into a shared object.
12564 @option{-mimpure-text} suppresses the ``relocations remain against
12565 allocatable but non-writable sections'' linker error message.
12566 However, the necessary relocations will trigger copy-on-write, and the
12567 shared object is not actually shared across processes. Instead of
12568 using @option{-mimpure-text}, you should compile all source code with
12569 @option{-fpic} or @option{-fPIC}.
12571 This option is only available on SunOS and Solaris.
12573 @item -mcpu=@var{cpu_type}
12575 Set the instruction set, register set, and instruction scheduling parameters
12576 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
12577 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
12578 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
12579 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
12580 @samp{ultrasparc3}, and @samp{niagara}.
12582 Default instruction scheduling parameters are used for values that select
12583 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
12584 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
12586 Here is a list of each supported architecture and their supported
12591 v8: supersparc, hypersparc
12592 sparclite: f930, f934, sparclite86x
12594 v9: ultrasparc, ultrasparc3, niagara
12597 By default (unless configured otherwise), GCC generates code for the V7
12598 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
12599 additionally optimizes it for the Cypress CY7C602 chip, as used in the
12600 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
12601 SPARCStation 1, 2, IPX etc.
12603 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
12604 architecture. The only difference from V7 code is that the compiler emits
12605 the integer multiply and integer divide instructions which exist in SPARC-V8
12606 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
12607 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
12610 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
12611 the SPARC architecture. This adds the integer multiply, integer divide step
12612 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
12613 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
12614 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
12615 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
12616 MB86934 chip, which is the more recent SPARClite with FPU@.
12618 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
12619 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
12620 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
12621 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
12622 optimizes it for the TEMIC SPARClet chip.
12624 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
12625 architecture. This adds 64-bit integer and floating-point move instructions,
12626 3 additional floating-point condition code registers and conditional move
12627 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
12628 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
12629 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
12630 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
12631 @option{-mcpu=niagara}, the compiler additionally optimizes it for
12632 Sun UltraSPARC T1 chips.
12634 @item -mtune=@var{cpu_type}
12636 Set the instruction scheduling parameters for machine type
12637 @var{cpu_type}, but do not set the instruction set or register set that the
12638 option @option{-mcpu=@var{cpu_type}} would.
12640 The same values for @option{-mcpu=@var{cpu_type}} can be used for
12641 @option{-mtune=@var{cpu_type}}, but the only useful values are those
12642 that select a particular cpu implementation. Those are @samp{cypress},
12643 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
12644 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
12645 @samp{ultrasparc3}, and @samp{niagara}.
12650 @opindex mno-v8plus
12651 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
12652 difference from the V8 ABI is that the global and out registers are
12653 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
12654 mode for all SPARC-V9 processors.
12660 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
12661 Visual Instruction Set extensions. The default is @option{-mno-vis}.
12664 These @samp{-m} options are supported in addition to the above
12665 on SPARC-V9 processors in 64-bit environments:
12668 @item -mlittle-endian
12669 @opindex mlittle-endian
12670 Generate code for a processor running in little-endian mode. It is only
12671 available for a few configurations and most notably not on Solaris and Linux.
12677 Generate code for a 32-bit or 64-bit environment.
12678 The 32-bit environment sets int, long and pointer to 32 bits.
12679 The 64-bit environment sets int to 32 bits and long and pointer
12682 @item -mcmodel=medlow
12683 @opindex mcmodel=medlow
12684 Generate code for the Medium/Low code model: 64-bit addresses, programs
12685 must be linked in the low 32 bits of memory. Programs can be statically
12686 or dynamically linked.
12688 @item -mcmodel=medmid
12689 @opindex mcmodel=medmid
12690 Generate code for the Medium/Middle code model: 64-bit addresses, programs
12691 must be linked in the low 44 bits of memory, the text and data segments must
12692 be less than 2GB in size and the data segment must be located within 2GB of
12695 @item -mcmodel=medany
12696 @opindex mcmodel=medany
12697 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
12698 may be linked anywhere in memory, the text and data segments must be less
12699 than 2GB in size and the data segment must be located within 2GB of the
12702 @item -mcmodel=embmedany
12703 @opindex mcmodel=embmedany
12704 Generate code for the Medium/Anywhere code model for embedded systems:
12705 64-bit addresses, the text and data segments must be less than 2GB in
12706 size, both starting anywhere in memory (determined at link time). The
12707 global register %g4 points to the base of the data segment. Programs
12708 are statically linked and PIC is not supported.
12711 @itemx -mno-stack-bias
12712 @opindex mstack-bias
12713 @opindex mno-stack-bias
12714 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
12715 frame pointer if present, are offset by @minus{}2047 which must be added back
12716 when making stack frame references. This is the default in 64-bit mode.
12717 Otherwise, assume no such offset is present.
12720 These switches are supported in addition to the above on Solaris:
12725 Add support for multithreading using the Solaris threads library. This
12726 option sets flags for both the preprocessor and linker. This option does
12727 not affect the thread safety of object code produced by the compiler or
12728 that of libraries supplied with it.
12732 Add support for multithreading using the POSIX threads library. This
12733 option sets flags for both the preprocessor and linker. This option does
12734 not affect the thread safety of object code produced by the compiler or
12735 that of libraries supplied with it.
12739 This is a synonym for @option{-pthreads}.
12742 @node System V Options
12743 @subsection Options for System V
12745 These additional options are available on System V Release 4 for
12746 compatibility with other compilers on those systems:
12751 Create a shared object.
12752 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
12756 Identify the versions of each tool used by the compiler, in a
12757 @code{.ident} assembler directive in the output.
12761 Refrain from adding @code{.ident} directives to the output file (this is
12764 @item -YP,@var{dirs}
12766 Search the directories @var{dirs}, and no others, for libraries
12767 specified with @option{-l}.
12769 @item -Ym,@var{dir}
12771 Look in the directory @var{dir} to find the M4 preprocessor.
12772 The assembler uses this option.
12773 @c This is supposed to go with a -Yd for predefined M4 macro files, but
12774 @c the generic assembler that comes with Solaris takes just -Ym.
12777 @node TMS320C3x/C4x Options
12778 @subsection TMS320C3x/C4x Options
12779 @cindex TMS320C3x/C4x Options
12781 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
12785 @item -mcpu=@var{cpu_type}
12787 Set the instruction set, register set, and instruction scheduling
12788 parameters for machine type @var{cpu_type}. Supported values for
12789 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
12790 @samp{c44}. The default is @samp{c40} to generate code for the
12795 @itemx -msmall-memory
12797 @opindex mbig-memory
12799 @opindex msmall-memory
12801 Generates code for the big or small memory model. The small memory
12802 model assumed that all data fits into one 64K word page. At run-time
12803 the data page (DP) register must be set to point to the 64K page
12804 containing the .bss and .data program sections. The big memory model is
12805 the default and requires reloading of the DP register for every direct
12812 Allow (disallow) allocation of general integer operands into the block
12813 count register BK@.
12819 Enable (disable) generation of code using decrement and branch,
12820 DBcond(D), instructions. This is enabled by default for the C4x. To be
12821 on the safe side, this is disabled for the C3x, since the maximum
12822 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
12823 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
12824 that it can utilize the decrement and branch instruction, but will give
12825 up if there is more than one memory reference in the loop. Thus a loop
12826 where the loop counter is decremented can generate slightly more
12827 efficient code, in cases where the RPTB instruction cannot be utilized.
12829 @item -mdp-isr-reload
12831 @opindex mdp-isr-reload
12833 Force the DP register to be saved on entry to an interrupt service
12834 routine (ISR), reloaded to point to the data section, and restored on
12835 exit from the ISR@. This should not be required unless someone has
12836 violated the small memory model by modifying the DP register, say within
12843 For the C3x use the 24-bit MPYI instruction for integer multiplies
12844 instead of a library call to guarantee 32-bit results. Note that if one
12845 of the operands is a constant, then the multiplication will be performed
12846 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
12847 then squaring operations are performed inline instead of a library call.
12850 @itemx -mno-fast-fix
12852 @opindex mno-fast-fix
12853 The C3x/C4x FIX instruction to convert a floating point value to an
12854 integer value chooses the nearest integer less than or equal to the
12855 floating point value rather than to the nearest integer. Thus if the
12856 floating point number is negative, the result will be incorrectly
12857 truncated an additional code is necessary to detect and correct this
12858 case. This option can be used to disable generation of the additional
12859 code required to correct the result.
12865 Enable (disable) generation of repeat block sequences using the RPTB
12866 instruction for zero overhead looping. The RPTB construct is only used
12867 for innermost loops that do not call functions or jump across the loop
12868 boundaries. There is no advantage having nested RPTB loops due to the
12869 overhead required to save and restore the RC, RS, and RE registers.
12870 This is enabled by default with @option{-O2}.
12872 @item -mrpts=@var{count}
12876 Enable (disable) the use of the single instruction repeat instruction
12877 RPTS@. If a repeat block contains a single instruction, and the loop
12878 count can be guaranteed to be less than the value @var{count}, GCC will
12879 emit a RPTS instruction instead of a RPTB@. If no value is specified,
12880 then a RPTS will be emitted even if the loop count cannot be determined
12881 at compile time. Note that the repeated instruction following RPTS does
12882 not have to be reloaded from memory each iteration, thus freeing up the
12883 CPU buses for operands. However, since interrupts are blocked by this
12884 instruction, it is disabled by default.
12886 @item -mloop-unsigned
12887 @itemx -mno-loop-unsigned
12888 @opindex mloop-unsigned
12889 @opindex mno-loop-unsigned
12890 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
12891 is @math{2^{31} + 1} since these instructions test if the iteration count is
12892 negative to terminate the loop. If the iteration count is unsigned
12893 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
12894 exceeded. This switch allows an unsigned iteration count.
12898 Try to emit an assembler syntax that the TI assembler (asm30) is happy
12899 with. This also enforces compatibility with the API employed by the TI
12900 C3x C compiler. For example, long doubles are passed as structures
12901 rather than in floating point registers.
12907 Generate code that uses registers (stack) for passing arguments to functions.
12908 By default, arguments are passed in registers where possible rather
12909 than by pushing arguments on to the stack.
12911 @item -mparallel-insns
12912 @itemx -mno-parallel-insns
12913 @opindex mparallel-insns
12914 @opindex mno-parallel-insns
12915 Allow the generation of parallel instructions. This is enabled by
12916 default with @option{-O2}.
12918 @item -mparallel-mpy
12919 @itemx -mno-parallel-mpy
12920 @opindex mparallel-mpy
12921 @opindex mno-parallel-mpy
12922 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
12923 provided @option{-mparallel-insns} is also specified. These instructions have
12924 tight register constraints which can pessimize the code generation
12925 of large functions.
12930 @subsection V850 Options
12931 @cindex V850 Options
12933 These @samp{-m} options are defined for V850 implementations:
12937 @itemx -mno-long-calls
12938 @opindex mlong-calls
12939 @opindex mno-long-calls
12940 Treat all calls as being far away (near). If calls are assumed to be
12941 far away, the compiler will always load the functions address up into a
12942 register, and call indirect through the pointer.
12948 Do not optimize (do optimize) basic blocks that use the same index
12949 pointer 4 or more times to copy pointer into the @code{ep} register, and
12950 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
12951 option is on by default if you optimize.
12953 @item -mno-prolog-function
12954 @itemx -mprolog-function
12955 @opindex mno-prolog-function
12956 @opindex mprolog-function
12957 Do not use (do use) external functions to save and restore registers
12958 at the prologue and epilogue of a function. The external functions
12959 are slower, but use less code space if more than one function saves
12960 the same number of registers. The @option{-mprolog-function} option
12961 is on by default if you optimize.
12965 Try to make the code as small as possible. At present, this just turns
12966 on the @option{-mep} and @option{-mprolog-function} options.
12968 @item -mtda=@var{n}
12970 Put static or global variables whose size is @var{n} bytes or less into
12971 the tiny data area that register @code{ep} points to. The tiny data
12972 area can hold up to 256 bytes in total (128 bytes for byte references).
12974 @item -msda=@var{n}
12976 Put static or global variables whose size is @var{n} bytes or less into
12977 the small data area that register @code{gp} points to. The small data
12978 area can hold up to 64 kilobytes.
12980 @item -mzda=@var{n}
12982 Put static or global variables whose size is @var{n} bytes or less into
12983 the first 32 kilobytes of memory.
12987 Specify that the target processor is the V850.
12990 @opindex mbig-switch
12991 Generate code suitable for big switch tables. Use this option only if
12992 the assembler/linker complain about out of range branches within a switch
12997 This option will cause r2 and r5 to be used in the code generated by
12998 the compiler. This setting is the default.
13000 @item -mno-app-regs
13001 @opindex mno-app-regs
13002 This option will cause r2 and r5 to be treated as fixed registers.
13006 Specify that the target processor is the V850E1. The preprocessor
13007 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
13008 this option is used.
13012 Specify that the target processor is the V850E@. The preprocessor
13013 constant @samp{__v850e__} will be defined if this option is used.
13015 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
13016 are defined then a default target processor will be chosen and the
13017 relevant @samp{__v850*__} preprocessor constant will be defined.
13019 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
13020 defined, regardless of which processor variant is the target.
13022 @item -mdisable-callt
13023 @opindex mdisable-callt
13024 This option will suppress generation of the CALLT instruction for the
13025 v850e and v850e1 flavors of the v850 architecture. The default is
13026 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
13031 @subsection VAX Options
13032 @cindex VAX options
13034 These @samp{-m} options are defined for the VAX:
13039 Do not output certain jump instructions (@code{aobleq} and so on)
13040 that the Unix assembler for the VAX cannot handle across long
13045 Do output those jump instructions, on the assumption that you
13046 will assemble with the GNU assembler.
13050 Output code for g-format floating point numbers instead of d-format.
13053 @node x86-64 Options
13054 @subsection x86-64 Options
13055 @cindex x86-64 options
13057 These are listed under @xref{i386 and x86-64 Options}.
13059 @node Xstormy16 Options
13060 @subsection Xstormy16 Options
13061 @cindex Xstormy16 Options
13063 These options are defined for Xstormy16:
13068 Choose startup files and linker script suitable for the simulator.
13071 @node Xtensa Options
13072 @subsection Xtensa Options
13073 @cindex Xtensa Options
13075 These options are supported for Xtensa targets:
13079 @itemx -mno-const16
13081 @opindex mno-const16
13082 Enable or disable use of @code{CONST16} instructions for loading
13083 constant values. The @code{CONST16} instruction is currently not a
13084 standard option from Tensilica. When enabled, @code{CONST16}
13085 instructions are always used in place of the standard @code{L32R}
13086 instructions. The use of @code{CONST16} is enabled by default only if
13087 the @code{L32R} instruction is not available.
13090 @itemx -mno-fused-madd
13091 @opindex mfused-madd
13092 @opindex mno-fused-madd
13093 Enable or disable use of fused multiply/add and multiply/subtract
13094 instructions in the floating-point option. This has no effect if the
13095 floating-point option is not also enabled. Disabling fused multiply/add
13096 and multiply/subtract instructions forces the compiler to use separate
13097 instructions for the multiply and add/subtract operations. This may be
13098 desirable in some cases where strict IEEE 754-compliant results are
13099 required: the fused multiply add/subtract instructions do not round the
13100 intermediate result, thereby producing results with @emph{more} bits of
13101 precision than specified by the IEEE standard. Disabling fused multiply
13102 add/subtract instructions also ensures that the program output is not
13103 sensitive to the compiler's ability to combine multiply and add/subtract
13106 @item -mtext-section-literals
13107 @itemx -mno-text-section-literals
13108 @opindex mtext-section-literals
13109 @opindex mno-text-section-literals
13110 Control the treatment of literal pools. The default is
13111 @option{-mno-text-section-literals}, which places literals in a separate
13112 section in the output file. This allows the literal pool to be placed
13113 in a data RAM/ROM, and it also allows the linker to combine literal
13114 pools from separate object files to remove redundant literals and
13115 improve code size. With @option{-mtext-section-literals}, the literals
13116 are interspersed in the text section in order to keep them as close as
13117 possible to their references. This may be necessary for large assembly
13120 @item -mtarget-align
13121 @itemx -mno-target-align
13122 @opindex mtarget-align
13123 @opindex mno-target-align
13124 When this option is enabled, GCC instructs the assembler to
13125 automatically align instructions to reduce branch penalties at the
13126 expense of some code density. The assembler attempts to widen density
13127 instructions to align branch targets and the instructions following call
13128 instructions. If there are not enough preceding safe density
13129 instructions to align a target, no widening will be performed. The
13130 default is @option{-mtarget-align}. These options do not affect the
13131 treatment of auto-aligned instructions like @code{LOOP}, which the
13132 assembler will always align, either by widening density instructions or
13133 by inserting no-op instructions.
13136 @itemx -mno-longcalls
13137 @opindex mlongcalls
13138 @opindex mno-longcalls
13139 When this option is enabled, GCC instructs the assembler to translate
13140 direct calls to indirect calls unless it can determine that the target
13141 of a direct call is in the range allowed by the call instruction. This
13142 translation typically occurs for calls to functions in other source
13143 files. Specifically, the assembler translates a direct @code{CALL}
13144 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
13145 The default is @option{-mno-longcalls}. This option should be used in
13146 programs where the call target can potentially be out of range. This
13147 option is implemented in the assembler, not the compiler, so the
13148 assembly code generated by GCC will still show direct call
13149 instructions---look at the disassembled object code to see the actual
13150 instructions. Note that the assembler will use an indirect call for
13151 every cross-file call, not just those that really will be out of range.
13154 @node zSeries Options
13155 @subsection zSeries Options
13156 @cindex zSeries options
13158 These are listed under @xref{S/390 and zSeries Options}.
13160 @node Code Gen Options
13161 @section Options for Code Generation Conventions
13162 @cindex code generation conventions
13163 @cindex options, code generation
13164 @cindex run-time options
13166 These machine-independent options control the interface conventions
13167 used in code generation.
13169 Most of them have both positive and negative forms; the negative form
13170 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
13171 one of the forms is listed---the one which is not the default. You
13172 can figure out the other form by either removing @samp{no-} or adding
13176 @item -fbounds-check
13177 @opindex fbounds-check
13178 For front-ends that support it, generate additional code to check that
13179 indices used to access arrays are within the declared range. This is
13180 currently only supported by the Java and Fortran front-ends, where
13181 this option defaults to true and false respectively.
13185 This option generates traps for signed overflow on addition, subtraction,
13186 multiplication operations.
13190 This option instructs the compiler to assume that signed arithmetic
13191 overflow of addition, subtraction and multiplication wraps around
13192 using twos-complement representation. This flag enables some optimizations
13193 and disables others. This option is enabled by default for the Java
13194 front-end, as required by the Java language specification.
13197 @opindex fexceptions
13198 Enable exception handling. Generates extra code needed to propagate
13199 exceptions. For some targets, this implies GCC will generate frame
13200 unwind information for all functions, which can produce significant data
13201 size overhead, although it does not affect execution. If you do not
13202 specify this option, GCC will enable it by default for languages like
13203 C++ which normally require exception handling, and disable it for
13204 languages like C that do not normally require it. However, you may need
13205 to enable this option when compiling C code that needs to interoperate
13206 properly with exception handlers written in C++. You may also wish to
13207 disable this option if you are compiling older C++ programs that don't
13208 use exception handling.
13210 @item -fnon-call-exceptions
13211 @opindex fnon-call-exceptions
13212 Generate code that allows trapping instructions to throw exceptions.
13213 Note that this requires platform-specific runtime support that does
13214 not exist everywhere. Moreover, it only allows @emph{trapping}
13215 instructions to throw exceptions, i.e.@: memory references or floating
13216 point instructions. It does not allow exceptions to be thrown from
13217 arbitrary signal handlers such as @code{SIGALRM}.
13219 @item -funwind-tables
13220 @opindex funwind-tables
13221 Similar to @option{-fexceptions}, except that it will just generate any needed
13222 static data, but will not affect the generated code in any other way.
13223 You will normally not enable this option; instead, a language processor
13224 that needs this handling would enable it on your behalf.
13226 @item -fasynchronous-unwind-tables
13227 @opindex fasynchronous-unwind-tables
13228 Generate unwind table in dwarf2 format, if supported by target machine. The
13229 table is exact at each instruction boundary, so it can be used for stack
13230 unwinding from asynchronous events (such as debugger or garbage collector).
13232 @item -fpcc-struct-return
13233 @opindex fpcc-struct-return
13234 Return ``short'' @code{struct} and @code{union} values in memory like
13235 longer ones, rather than in registers. This convention is less
13236 efficient, but it has the advantage of allowing intercallability between
13237 GCC-compiled files and files compiled with other compilers, particularly
13238 the Portable C Compiler (pcc).
13240 The precise convention for returning structures in memory depends
13241 on the target configuration macros.
13243 Short structures and unions are those whose size and alignment match
13244 that of some integer type.
13246 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
13247 switch is not binary compatible with code compiled with the
13248 @option{-freg-struct-return} switch.
13249 Use it to conform to a non-default application binary interface.
13251 @item -freg-struct-return
13252 @opindex freg-struct-return
13253 Return @code{struct} and @code{union} values in registers when possible.
13254 This is more efficient for small structures than
13255 @option{-fpcc-struct-return}.
13257 If you specify neither @option{-fpcc-struct-return} nor
13258 @option{-freg-struct-return}, GCC defaults to whichever convention is
13259 standard for the target. If there is no standard convention, GCC
13260 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
13261 the principal compiler. In those cases, we can choose the standard, and
13262 we chose the more efficient register return alternative.
13264 @strong{Warning:} code compiled with the @option{-freg-struct-return}
13265 switch is not binary compatible with code compiled with the
13266 @option{-fpcc-struct-return} switch.
13267 Use it to conform to a non-default application binary interface.
13269 @item -fshort-enums
13270 @opindex fshort-enums
13271 Allocate to an @code{enum} type only as many bytes as it needs for the
13272 declared range of possible values. Specifically, the @code{enum} type
13273 will be equivalent to the smallest integer type which has enough room.
13275 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
13276 code that is not binary compatible with code generated without that switch.
13277 Use it to conform to a non-default application binary interface.
13279 @item -fshort-double
13280 @opindex fshort-double
13281 Use the same size for @code{double} as for @code{float}.
13283 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
13284 code that is not binary compatible with code generated without that switch.
13285 Use it to conform to a non-default application binary interface.
13287 @item -fshort-wchar
13288 @opindex fshort-wchar
13289 Override the underlying type for @samp{wchar_t} to be @samp{short
13290 unsigned int} instead of the default for the target. This option is
13291 useful for building programs to run under WINE@.
13293 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
13294 code that is not binary compatible with code generated without that switch.
13295 Use it to conform to a non-default application binary interface.
13298 @opindex fno-common
13299 In C, allocate even uninitialized global variables in the data section of the
13300 object file, rather than generating them as common blocks. This has the
13301 effect that if the same variable is declared (without @code{extern}) in
13302 two different compilations, you will get an error when you link them.
13303 The only reason this might be useful is if you wish to verify that the
13304 program will work on other systems which always work this way.
13308 Ignore the @samp{#ident} directive.
13310 @item -finhibit-size-directive
13311 @opindex finhibit-size-directive
13312 Don't output a @code{.size} assembler directive, or anything else that
13313 would cause trouble if the function is split in the middle, and the
13314 two halves are placed at locations far apart in memory. This option is
13315 used when compiling @file{crtstuff.c}; you should not need to use it
13318 @item -fverbose-asm
13319 @opindex fverbose-asm
13320 Put extra commentary information in the generated assembly code to
13321 make it more readable. This option is generally only of use to those
13322 who actually need to read the generated assembly code (perhaps while
13323 debugging the compiler itself).
13325 @option{-fno-verbose-asm}, the default, causes the
13326 extra information to be omitted and is useful when comparing two assembler
13331 @cindex global offset table
13333 Generate position-independent code (PIC) suitable for use in a shared
13334 library, if supported for the target machine. Such code accesses all
13335 constant addresses through a global offset table (GOT)@. The dynamic
13336 loader resolves the GOT entries when the program starts (the dynamic
13337 loader is not part of GCC; it is part of the operating system). If
13338 the GOT size for the linked executable exceeds a machine-specific
13339 maximum size, you get an error message from the linker indicating that
13340 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
13341 instead. (These maximums are 8k on the SPARC and 32k
13342 on the m68k and RS/6000. The 386 has no such limit.)
13344 Position-independent code requires special support, and therefore works
13345 only on certain machines. For the 386, GCC supports PIC for System V
13346 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
13347 position-independent.
13349 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13354 If supported for the target machine, emit position-independent code,
13355 suitable for dynamic linking and avoiding any limit on the size of the
13356 global offset table. This option makes a difference on the m68k,
13357 PowerPC and SPARC@.
13359 Position-independent code requires special support, and therefore works
13360 only on certain machines.
13362 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13369 These options are similar to @option{-fpic} and @option{-fPIC}, but
13370 generated position independent code can be only linked into executables.
13371 Usually these options are used when @option{-pie} GCC option will be
13372 used during linking.
13374 @item -fno-jump-tables
13375 @opindex fno-jump-tables
13376 Do not use jump tables for switch statements even where it would be
13377 more efficient than other code generation strategies. This option is
13378 of use in conjunction with @option{-fpic} or @option{-fPIC} for
13379 building code which forms part of a dynamic linker and cannot
13380 reference the address of a jump table. On some targets, jump tables
13381 do not require a GOT and this option is not needed.
13383 @item -ffixed-@var{reg}
13385 Treat the register named @var{reg} as a fixed register; generated code
13386 should never refer to it (except perhaps as a stack pointer, frame
13387 pointer or in some other fixed role).
13389 @var{reg} must be the name of a register. The register names accepted
13390 are machine-specific and are defined in the @code{REGISTER_NAMES}
13391 macro in the machine description macro file.
13393 This flag does not have a negative form, because it specifies a
13396 @item -fcall-used-@var{reg}
13397 @opindex fcall-used
13398 Treat the register named @var{reg} as an allocable register that is
13399 clobbered by function calls. It may be allocated for temporaries or
13400 variables that do not live across a call. Functions compiled this way
13401 will not save and restore the register @var{reg}.
13403 It is an error to used this flag with the frame pointer or stack pointer.
13404 Use of this flag for other registers that have fixed pervasive roles in
13405 the machine's execution model will produce disastrous results.
13407 This flag does not have a negative form, because it specifies a
13410 @item -fcall-saved-@var{reg}
13411 @opindex fcall-saved
13412 Treat the register named @var{reg} as an allocable register saved by
13413 functions. It may be allocated even for temporaries or variables that
13414 live across a call. Functions compiled this way will save and restore
13415 the register @var{reg} if they use it.
13417 It is an error to used this flag with the frame pointer or stack pointer.
13418 Use of this flag for other registers that have fixed pervasive roles in
13419 the machine's execution model will produce disastrous results.
13421 A different sort of disaster will result from the use of this flag for
13422 a register in which function values may be returned.
13424 This flag does not have a negative form, because it specifies a
13427 @item -fpack-struct[=@var{n}]
13428 @opindex fpack-struct
13429 Without a value specified, pack all structure members together without
13430 holes. When a value is specified (which must be a small power of two), pack
13431 structure members according to this value, representing the maximum
13432 alignment (that is, objects with default alignment requirements larger than
13433 this will be output potentially unaligned at the next fitting location.
13435 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
13436 code that is not binary compatible with code generated without that switch.
13437 Additionally, it makes the code suboptimal.
13438 Use it to conform to a non-default application binary interface.
13440 @item -finstrument-functions
13441 @opindex finstrument-functions
13442 Generate instrumentation calls for entry and exit to functions. Just
13443 after function entry and just before function exit, the following
13444 profiling functions will be called with the address of the current
13445 function and its call site. (On some platforms,
13446 @code{__builtin_return_address} does not work beyond the current
13447 function, so the call site information may not be available to the
13448 profiling functions otherwise.)
13451 void __cyg_profile_func_enter (void *this_fn,
13453 void __cyg_profile_func_exit (void *this_fn,
13457 The first argument is the address of the start of the current function,
13458 which may be looked up exactly in the symbol table.
13460 This instrumentation is also done for functions expanded inline in other
13461 functions. The profiling calls will indicate where, conceptually, the
13462 inline function is entered and exited. This means that addressable
13463 versions of such functions must be available. If all your uses of a
13464 function are expanded inline, this may mean an additional expansion of
13465 code size. If you use @samp{extern inline} in your C code, an
13466 addressable version of such functions must be provided. (This is
13467 normally the case anyways, but if you get lucky and the optimizer always
13468 expands the functions inline, you might have gotten away without
13469 providing static copies.)
13471 A function may be given the attribute @code{no_instrument_function}, in
13472 which case this instrumentation will not be done. This can be used, for
13473 example, for the profiling functions listed above, high-priority
13474 interrupt routines, and any functions from which the profiling functions
13475 cannot safely be called (perhaps signal handlers, if the profiling
13476 routines generate output or allocate memory).
13478 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
13479 @opindex finstrument-functions-exclude-file-list
13481 Set the list of functions that are excluded from instrumentation (see
13482 the description of @code{-finstrument-functions}). If the file that
13483 contains a function definition matches with one of @var{file}, then
13484 that function is not instrumented. The match is done on substrings:
13485 if the @var{file} parameter is a substring of the file name, it is
13486 considered to be a match.
13489 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
13490 will exclude any inline function defined in files whose pathnames
13491 contain @code{/bits/stl} or @code{include/sys}.
13493 If, for some reason, you want to include letter @code{','} in one of
13494 @var{sym}, write @code{'\,'}. For example,
13495 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
13496 (note the single quote surrounding the option).
13498 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
13499 @opindex finstrument-functions-exclude-function-list
13501 This is similar to @code{-finstrument-functions-exclude-file-list},
13502 but this option sets the list of function names to be excluded from
13503 instrumentation. The function name to be matched is its user-visible
13504 name, such as @code{vector<int> blah(const vector<int> &)}, not the
13505 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
13506 match is done on substrings: if the @var{sym} parameter is a substring
13507 of the function name, it is considered to be a match.
13509 @item -fstack-check
13510 @opindex fstack-check
13511 Generate code to verify that you do not go beyond the boundary of the
13512 stack. You should specify this flag if you are running in an
13513 environment with multiple threads, but only rarely need to specify it in
13514 a single-threaded environment since stack overflow is automatically
13515 detected on nearly all systems if there is only one stack.
13517 Note that this switch does not actually cause checking to be done; the
13518 operating system must do that. The switch causes generation of code
13519 to ensure that the operating system sees the stack being extended.
13521 @item -fstack-limit-register=@var{reg}
13522 @itemx -fstack-limit-symbol=@var{sym}
13523 @itemx -fno-stack-limit
13524 @opindex fstack-limit-register
13525 @opindex fstack-limit-symbol
13526 @opindex fno-stack-limit
13527 Generate code to ensure that the stack does not grow beyond a certain value,
13528 either the value of a register or the address of a symbol. If the stack
13529 would grow beyond the value, a signal is raised. For most targets,
13530 the signal is raised before the stack overruns the boundary, so
13531 it is possible to catch the signal without taking special precautions.
13533 For instance, if the stack starts at absolute address @samp{0x80000000}
13534 and grows downwards, you can use the flags
13535 @option{-fstack-limit-symbol=__stack_limit} and
13536 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
13537 of 128KB@. Note that this may only work with the GNU linker.
13539 @cindex aliasing of parameters
13540 @cindex parameters, aliased
13541 @item -fargument-alias
13542 @itemx -fargument-noalias
13543 @itemx -fargument-noalias-global
13544 @itemx -fargument-noalias-anything
13545 @opindex fargument-alias
13546 @opindex fargument-noalias
13547 @opindex fargument-noalias-global
13548 @opindex fargument-noalias-anything
13549 Specify the possible relationships among parameters and between
13550 parameters and global data.
13552 @option{-fargument-alias} specifies that arguments (parameters) may
13553 alias each other and may alias global storage.@*
13554 @option{-fargument-noalias} specifies that arguments do not alias
13555 each other, but may alias global storage.@*
13556 @option{-fargument-noalias-global} specifies that arguments do not
13557 alias each other and do not alias global storage.
13558 @option{-fargument-noalias-anything} specifies that arguments do not
13559 alias any other storage.
13561 Each language will automatically use whatever option is required by
13562 the language standard. You should not need to use these options yourself.
13564 @item -fleading-underscore
13565 @opindex fleading-underscore
13566 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
13567 change the way C symbols are represented in the object file. One use
13568 is to help link with legacy assembly code.
13570 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
13571 generate code that is not binary compatible with code generated without that
13572 switch. Use it to conform to a non-default application binary interface.
13573 Not all targets provide complete support for this switch.
13575 @item -ftls-model=@var{model}
13576 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
13577 The @var{model} argument should be one of @code{global-dynamic},
13578 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
13580 The default without @option{-fpic} is @code{initial-exec}; with
13581 @option{-fpic} the default is @code{global-dynamic}.
13583 @item -fvisibility=@var{default|internal|hidden|protected}
13584 @opindex fvisibility
13585 Set the default ELF image symbol visibility to the specified option---all
13586 symbols will be marked with this unless overridden within the code.
13587 Using this feature can very substantially improve linking and
13588 load times of shared object libraries, produce more optimized
13589 code, provide near-perfect API export and prevent symbol clashes.
13590 It is @strong{strongly} recommended that you use this in any shared objects
13593 Despite the nomenclature, @code{default} always means public ie;
13594 available to be linked against from outside the shared object.
13595 @code{protected} and @code{internal} are pretty useless in real-world
13596 usage so the only other commonly used option will be @code{hidden}.
13597 The default if @option{-fvisibility} isn't specified is
13598 @code{default}, i.e., make every
13599 symbol public---this causes the same behavior as previous versions of
13602 A good explanation of the benefits offered by ensuring ELF
13603 symbols have the correct visibility is given by ``How To Write
13604 Shared Libraries'' by Ulrich Drepper (which can be found at
13605 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
13606 solution made possible by this option to marking things hidden when
13607 the default is public is to make the default hidden and mark things
13608 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
13609 and @code{__attribute__ ((visibility("default")))} instead of
13610 @code{__declspec(dllexport)} you get almost identical semantics with
13611 identical syntax. This is a great boon to those working with
13612 cross-platform projects.
13614 For those adding visibility support to existing code, you may find
13615 @samp{#pragma GCC visibility} of use. This works by you enclosing
13616 the declarations you wish to set visibility for with (for example)
13617 @samp{#pragma GCC visibility push(hidden)} and
13618 @samp{#pragma GCC visibility pop}.
13619 Bear in mind that symbol visibility should be viewed @strong{as
13620 part of the API interface contract} and thus all new code should
13621 always specify visibility when it is not the default ie; declarations
13622 only for use within the local DSO should @strong{always} be marked explicitly
13623 as hidden as so to avoid PLT indirection overheads---making this
13624 abundantly clear also aids readability and self-documentation of the code.
13625 Note that due to ISO C++ specification requirements, operator new and
13626 operator delete must always be of default visibility.
13628 Be aware that headers from outside your project, in particular system
13629 headers and headers from any other library you use, may not be
13630 expecting to be compiled with visibility other than the default. You
13631 may need to explicitly say @samp{#pragma GCC visibility push(default)}
13632 before including any such headers.
13634 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
13635 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
13636 no modifications. However, this means that calls to @samp{extern}
13637 functions with no explicit visibility will use the PLT, so it is more
13638 effective to use @samp{__attribute ((visibility))} and/or
13639 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
13640 declarations should be treated as hidden.
13642 Note that @samp{-fvisibility} does affect C++ vague linkage
13643 entities. This means that, for instance, an exception class that will
13644 be thrown between DSOs must be explicitly marked with default
13645 visibility so that the @samp{type_info} nodes will be unified between
13648 An overview of these techniques, their benefits and how to use them
13649 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
13655 @node Environment Variables
13656 @section Environment Variables Affecting GCC
13657 @cindex environment variables
13659 @c man begin ENVIRONMENT
13660 This section describes several environment variables that affect how GCC
13661 operates. Some of them work by specifying directories or prefixes to use
13662 when searching for various kinds of files. Some are used to specify other
13663 aspects of the compilation environment.
13665 Note that you can also specify places to search using options such as
13666 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
13667 take precedence over places specified using environment variables, which
13668 in turn take precedence over those specified by the configuration of GCC@.
13669 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
13670 GNU Compiler Collection (GCC) Internals}.
13675 @c @itemx LC_COLLATE
13677 @c @itemx LC_MONETARY
13678 @c @itemx LC_NUMERIC
13683 @c @findex LC_COLLATE
13684 @findex LC_MESSAGES
13685 @c @findex LC_MONETARY
13686 @c @findex LC_NUMERIC
13690 These environment variables control the way that GCC uses
13691 localization information that allow GCC to work with different
13692 national conventions. GCC inspects the locale categories
13693 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
13694 so. These locale categories can be set to any value supported by your
13695 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
13696 Kingdom encoded in UTF-8.
13698 The @env{LC_CTYPE} environment variable specifies character
13699 classification. GCC uses it to determine the character boundaries in
13700 a string; this is needed for some multibyte encodings that contain quote
13701 and escape characters that would otherwise be interpreted as a string
13704 The @env{LC_MESSAGES} environment variable specifies the language to
13705 use in diagnostic messages.
13707 If the @env{LC_ALL} environment variable is set, it overrides the value
13708 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
13709 and @env{LC_MESSAGES} default to the value of the @env{LANG}
13710 environment variable. If none of these variables are set, GCC
13711 defaults to traditional C English behavior.
13715 If @env{TMPDIR} is set, it specifies the directory to use for temporary
13716 files. GCC uses temporary files to hold the output of one stage of
13717 compilation which is to be used as input to the next stage: for example,
13718 the output of the preprocessor, which is the input to the compiler
13721 @item GCC_EXEC_PREFIX
13722 @findex GCC_EXEC_PREFIX
13723 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
13724 names of the subprograms executed by the compiler. No slash is added
13725 when this prefix is combined with the name of a subprogram, but you can
13726 specify a prefix that ends with a slash if you wish.
13728 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
13729 an appropriate prefix to use based on the pathname it was invoked with.
13731 If GCC cannot find the subprogram using the specified prefix, it
13732 tries looking in the usual places for the subprogram.
13734 The default value of @env{GCC_EXEC_PREFIX} is
13735 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
13736 of @code{prefix} when you ran the @file{configure} script.
13738 Other prefixes specified with @option{-B} take precedence over this prefix.
13740 This prefix is also used for finding files such as @file{crt0.o} that are
13743 In addition, the prefix is used in an unusual way in finding the
13744 directories to search for header files. For each of the standard
13745 directories whose name normally begins with @samp{/usr/local/lib/gcc}
13746 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
13747 replacing that beginning with the specified prefix to produce an
13748 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
13749 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
13750 These alternate directories are searched first; the standard directories
13753 @item COMPILER_PATH
13754 @findex COMPILER_PATH
13755 The value of @env{COMPILER_PATH} is a colon-separated list of
13756 directories, much like @env{PATH}. GCC tries the directories thus
13757 specified when searching for subprograms, if it can't find the
13758 subprograms using @env{GCC_EXEC_PREFIX}.
13761 @findex LIBRARY_PATH
13762 The value of @env{LIBRARY_PATH} is a colon-separated list of
13763 directories, much like @env{PATH}. When configured as a native compiler,
13764 GCC tries the directories thus specified when searching for special
13765 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
13766 using GCC also uses these directories when searching for ordinary
13767 libraries for the @option{-l} option (but directories specified with
13768 @option{-L} come first).
13772 @cindex locale definition
13773 This variable is used to pass locale information to the compiler. One way in
13774 which this information is used is to determine the character set to be used
13775 when character literals, string literals and comments are parsed in C and C++.
13776 When the compiler is configured to allow multibyte characters,
13777 the following values for @env{LANG} are recognized:
13781 Recognize JIS characters.
13783 Recognize SJIS characters.
13785 Recognize EUCJP characters.
13788 If @env{LANG} is not defined, or if it has some other value, then the
13789 compiler will use mblen and mbtowc as defined by the default locale to
13790 recognize and translate multibyte characters.
13794 Some additional environments variables affect the behavior of the
13797 @include cppenv.texi
13801 @node Precompiled Headers
13802 @section Using Precompiled Headers
13803 @cindex precompiled headers
13804 @cindex speed of compilation
13806 Often large projects have many header files that are included in every
13807 source file. The time the compiler takes to process these header files
13808 over and over again can account for nearly all of the time required to
13809 build the project. To make builds faster, GCC allows users to
13810 `precompile' a header file; then, if builds can use the precompiled
13811 header file they will be much faster.
13813 To create a precompiled header file, simply compile it as you would any
13814 other file, if necessary using the @option{-x} option to make the driver
13815 treat it as a C or C++ header file. You will probably want to use a
13816 tool like @command{make} to keep the precompiled header up-to-date when
13817 the headers it contains change.
13819 A precompiled header file will be searched for when @code{#include} is
13820 seen in the compilation. As it searches for the included file
13821 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
13822 compiler looks for a precompiled header in each directory just before it
13823 looks for the include file in that directory. The name searched for is
13824 the name specified in the @code{#include} with @samp{.gch} appended. If
13825 the precompiled header file can't be used, it is ignored.
13827 For instance, if you have @code{#include "all.h"}, and you have
13828 @file{all.h.gch} in the same directory as @file{all.h}, then the
13829 precompiled header file will be used if possible, and the original
13830 header will be used otherwise.
13832 Alternatively, you might decide to put the precompiled header file in a
13833 directory and use @option{-I} to ensure that directory is searched
13834 before (or instead of) the directory containing the original header.
13835 Then, if you want to check that the precompiled header file is always
13836 used, you can put a file of the same name as the original header in this
13837 directory containing an @code{#error} command.
13839 This also works with @option{-include}. So yet another way to use
13840 precompiled headers, good for projects not designed with precompiled
13841 header files in mind, is to simply take most of the header files used by
13842 a project, include them from another header file, precompile that header
13843 file, and @option{-include} the precompiled header. If the header files
13844 have guards against multiple inclusion, they will be skipped because
13845 they've already been included (in the precompiled header).
13847 If you need to precompile the same header file for different
13848 languages, targets, or compiler options, you can instead make a
13849 @emph{directory} named like @file{all.h.gch}, and put each precompiled
13850 header in the directory, perhaps using @option{-o}. It doesn't matter
13851 what you call the files in the directory, every precompiled header in
13852 the directory will be considered. The first precompiled header
13853 encountered in the directory that is valid for this compilation will
13854 be used; they're searched in no particular order.
13856 There are many other possibilities, limited only by your imagination,
13857 good sense, and the constraints of your build system.
13859 A precompiled header file can be used only when these conditions apply:
13863 Only one precompiled header can be used in a particular compilation.
13866 A precompiled header can't be used once the first C token is seen. You
13867 can have preprocessor directives before a precompiled header; you can
13868 even include a precompiled header from inside another header, so long as
13869 there are no C tokens before the @code{#include}.
13872 The precompiled header file must be produced for the same language as
13873 the current compilation. You can't use a C precompiled header for a C++
13877 The precompiled header file must have been produced by the same compiler
13878 binary as the current compilation is using.
13881 Any macros defined before the precompiled header is included must
13882 either be defined in the same way as when the precompiled header was
13883 generated, or must not affect the precompiled header, which usually
13884 means that they don't appear in the precompiled header at all.
13886 The @option{-D} option is one way to define a macro before a
13887 precompiled header is included; using a @code{#define} can also do it.
13888 There are also some options that define macros implicitly, like
13889 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
13892 @item If debugging information is output when using the precompiled
13893 header, using @option{-g} or similar, the same kind of debugging information
13894 must have been output when building the precompiled header. However,
13895 a precompiled header built using @option{-g} can be used in a compilation
13896 when no debugging information is being output.
13898 @item The same @option{-m} options must generally be used when building
13899 and using the precompiled header. @xref{Submodel Options},
13900 for any cases where this rule is relaxed.
13902 @item Each of the following options must be the same when building and using
13903 the precompiled header:
13905 @gccoptlist{-fexceptions -funit-at-a-time}
13908 Some other command-line options starting with @option{-f},
13909 @option{-p}, or @option{-O} must be defined in the same way as when
13910 the precompiled header was generated. At present, it's not clear
13911 which options are safe to change and which are not; the safest choice
13912 is to use exactly the same options when generating and using the
13913 precompiled header. The following are known to be safe:
13915 @gccoptlist{-fmessage-length= -fpreprocessed
13916 -fsched-interblock -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
13917 -fsched-verbose=<number> -fschedule-insns -fvisibility=
13922 For all of these except the last, the compiler will automatically
13923 ignore the precompiled header if the conditions aren't met. If you
13924 find an option combination that doesn't work and doesn't cause the
13925 precompiled header to be ignored, please consider filing a bug report,
13928 If you do use differing options when generating and using the
13929 precompiled header, the actual behavior will be a mixture of the
13930 behavior for the options. For instance, if you use @option{-g} to
13931 generate the precompiled header but not when using it, you may or may
13932 not get debugging information for routines in the precompiled header.
13934 @node Running Protoize
13935 @section Running Protoize
13937 The program @code{protoize} is an optional part of GCC@. You can use
13938 it to add prototypes to a program, thus converting the program to ISO
13939 C in one respect. The companion program @code{unprotoize} does the
13940 reverse: it removes argument types from any prototypes that are found.
13942 When you run these programs, you must specify a set of source files as
13943 command line arguments. The conversion programs start out by compiling
13944 these files to see what functions they define. The information gathered
13945 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
13947 After scanning comes actual conversion. The specified files are all
13948 eligible to be converted; any files they include (whether sources or
13949 just headers) are eligible as well.
13951 But not all the eligible files are converted. By default,
13952 @code{protoize} and @code{unprotoize} convert only source and header
13953 files in the current directory. You can specify additional directories
13954 whose files should be converted with the @option{-d @var{directory}}
13955 option. You can also specify particular files to exclude with the
13956 @option{-x @var{file}} option. A file is converted if it is eligible, its
13957 directory name matches one of the specified directory names, and its
13958 name within the directory has not been excluded.
13960 Basic conversion with @code{protoize} consists of rewriting most
13961 function definitions and function declarations to specify the types of
13962 the arguments. The only ones not rewritten are those for varargs
13965 @code{protoize} optionally inserts prototype declarations at the
13966 beginning of the source file, to make them available for any calls that
13967 precede the function's definition. Or it can insert prototype
13968 declarations with block scope in the blocks where undeclared functions
13971 Basic conversion with @code{unprotoize} consists of rewriting most
13972 function declarations to remove any argument types, and rewriting
13973 function definitions to the old-style pre-ISO form.
13975 Both conversion programs print a warning for any function declaration or
13976 definition that they can't convert. You can suppress these warnings
13979 The output from @code{protoize} or @code{unprotoize} replaces the
13980 original source file. The original file is renamed to a name ending
13981 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
13982 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
13983 for DOS) file already exists, then the source file is simply discarded.
13985 @code{protoize} and @code{unprotoize} both depend on GCC itself to
13986 scan the program and collect information about the functions it uses.
13987 So neither of these programs will work until GCC is installed.
13989 Here is a table of the options you can use with @code{protoize} and
13990 @code{unprotoize}. Each option works with both programs unless
13994 @item -B @var{directory}
13995 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
13996 usual directory (normally @file{/usr/local/lib}). This file contains
13997 prototype information about standard system functions. This option
13998 applies only to @code{protoize}.
14000 @item -c @var{compilation-options}
14001 Use @var{compilation-options} as the options when running @command{gcc} to
14002 produce the @samp{.X} files. The special option @option{-aux-info} is
14003 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
14005 Note that the compilation options must be given as a single argument to
14006 @code{protoize} or @code{unprotoize}. If you want to specify several
14007 @command{gcc} options, you must quote the entire set of compilation options
14008 to make them a single word in the shell.
14010 There are certain @command{gcc} arguments that you cannot use, because they
14011 would produce the wrong kind of output. These include @option{-g},
14012 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
14013 the @var{compilation-options}, they are ignored.
14016 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
14017 systems) instead of @samp{.c}. This is convenient if you are converting
14018 a C program to C++. This option applies only to @code{protoize}.
14021 Add explicit global declarations. This means inserting explicit
14022 declarations at the beginning of each source file for each function
14023 that is called in the file and was not declared. These declarations
14024 precede the first function definition that contains a call to an
14025 undeclared function. This option applies only to @code{protoize}.
14027 @item -i @var{string}
14028 Indent old-style parameter declarations with the string @var{string}.
14029 This option applies only to @code{protoize}.
14031 @code{unprotoize} converts prototyped function definitions to old-style
14032 function definitions, where the arguments are declared between the
14033 argument list and the initial @samp{@{}. By default, @code{unprotoize}
14034 uses five spaces as the indentation. If you want to indent with just
14035 one space instead, use @option{-i " "}.
14038 Keep the @samp{.X} files. Normally, they are deleted after conversion
14042 Add explicit local declarations. @code{protoize} with @option{-l} inserts
14043 a prototype declaration for each function in each block which calls the
14044 function without any declaration. This option applies only to
14048 Make no real changes. This mode just prints information about the conversions
14049 that would have been done without @option{-n}.
14052 Make no @samp{.save} files. The original files are simply deleted.
14053 Use this option with caution.
14055 @item -p @var{program}
14056 Use the program @var{program} as the compiler. Normally, the name
14057 @file{gcc} is used.
14060 Work quietly. Most warnings are suppressed.
14063 Print the version number, just like @option{-v} for @command{gcc}.
14066 If you need special compiler options to compile one of your program's
14067 source files, then you should generate that file's @samp{.X} file
14068 specially, by running @command{gcc} on that source file with the
14069 appropriate options and the option @option{-aux-info}. Then run
14070 @code{protoize} on the entire set of files. @code{protoize} will use
14071 the existing @samp{.X} file because it is newer than the source file.
14075 gcc -Dfoo=bar file1.c -aux-info file1.X
14080 You need to include the special files along with the rest in the
14081 @code{protoize} command, even though their @samp{.X} files already
14082 exist, because otherwise they won't get converted.
14084 @xref{Protoize Caveats}, for more information on how to use
14085 @code{protoize} successfully.