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 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
168 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
169 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
170 -fallow-single-precision -fcond-mismatch @gol
171 -fsigned-bitfields -fsigned-char @gol
172 @c APPLE LOCAL -Wnewline-eof 2001-08-23 --sts **
173 -Wnewline-eof (Apple compatible) @gol
174 -funsigned-bitfields -funsigned-char}
176 @item C++ Language Options
177 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
178 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
179 -fconserve-space -ffriend-injection @gol
180 -fno-elide-constructors @gol
181 -fno-enforce-eh-specs @gol
182 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
183 -fno-implicit-templates @gol
184 -fno-implicit-inline-templates @gol
185 -fno-implement-inlines -fms-extensions @gol
186 -fno-nonansi-builtins -fno-operator-names @gol
187 -fno-optional-diags -fpermissive @gol
188 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
189 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
190 -fno-default-inline -fvisibility-inlines-hidden @gol
191 -Wabi -Wctor-dtor-privacy @gol
192 -Wnon-virtual-dtor -Wreorder @gol
193 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel @gol
194 -Wno-non-template-friend -Wold-style-cast @gol
195 -Woverloaded-virtual -Wno-pmf-conversions @gol
198 @item Language Independent Options
199 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
200 @gccoptlist{-fmessage-length=@var{n} @gol
201 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
202 -fdiagnostics-show-option}
204 @item Warning Options
205 @xref{Warning Options,,Options to Request or Suppress Warnings}.
206 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
207 -w -Wextra -Wall -Waddress -Waggregate-return -Wno-attributes @gol
208 -Wc++-compat -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
209 -Wconversion -Wno-deprecated-declarations @gol
210 -Wdisabled-optimization -Wno-div-by-zero -Wno-endif-labels @gol
211 -Werror -Werror=* -Werror-implicit-function-declaration @gol
212 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
213 -Wno-format-extra-args -Wformat-nonliteral @gol
214 -Wformat-security -Wformat-y2k @gol
215 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
216 -Wimport -Wno-import -Winit-self -Winline @gol
217 -Wno-int-to-pointer-cast @gol
218 -Wno-invalid-offsetof -Winvalid-pch @gol
219 -Wlarger-than-@var{len} -Wframe-larger-than-@var{len} @gol
220 -Wunsafe-loop-optimizations -Wlong-long @gol
221 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
222 -Wmissing-format-attribute -Wmissing-include-dirs @gol
223 -Wmissing-noreturn @gol
224 -Wno-multichar -Wnonnull -Wno-overflow @gol
225 -Woverlength-strings -Wpacked -Wpadded @gol
226 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
227 -Wredundant-decls @gol
228 -Wreturn-type -Wsequence-point -Wshadow @gol
229 -Wsign-compare -Wstack-protector @gol
230 -Wstrict-aliasing -Wstrict-aliasing=2 @gol
231 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
232 -Wswitch -Wswitch-default -Wswitch-enum @gol
233 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
234 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
235 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
236 -Wunused-value -Wunused-variable @gol
237 -Wvariadic-macros -Wvla @gol
238 -Wvolatile-register-var -Wwrite-strings}
240 @item C-only Warning Options
241 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
242 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
243 -Wstrict-prototypes -Wtraditional @gol
244 -Wdeclaration-after-statement -Wpointer-sign}
246 @item Debugging Options
247 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
248 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
249 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
250 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
251 -fdump-ipa-all -fdump-ipa-cgraph @gol
253 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
254 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
255 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
256 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
258 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
259 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
260 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
261 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
262 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
263 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
264 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
266 -fdump-tree-nrv -fdump-tree-vect @gol
267 -fdump-tree-sink @gol
268 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
269 -fdump-tree-salias @gol
270 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
271 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
272 -ftree-vectorizer-verbose=@var{n} @gol
273 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
274 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
275 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
276 -fmem-report -fprofile-arcs @gol
277 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
278 -ftest-coverage -ftime-report -fvar-tracking @gol
279 -g -g@var{level} -gcoff -gdwarf-2 @gol
280 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
281 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
282 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
283 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
284 -print-multi-directory -print-multi-lib @gol
285 -print-prog-name=@var{program} -print-search-dirs -Q @gol
288 @item Optimization Options
289 @xref{Optimize Options,,Options that Control Optimization}.
290 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
291 -falign-labels=@var{n} -falign-loops=@var{n} @gol
292 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
293 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
294 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
295 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
296 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
297 -fdelayed-branch -fdelete-null-pointer-checks -fearly-inlining @gol
298 -fexpensive-optimizations -ffast-math -ffloat-store @gol
299 -fforce-addr -ffunction-sections @gol
300 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
301 -fcrossjumping -fif-conversion -fif-conversion2 @gol
302 -finline-functions -finline-functions-called-once @gol
303 -finline-limit=@var{n} -fkeep-inline-functions @gol
304 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
305 -fmodulo-sched -fno-branch-count-reg @gol
306 -fno-default-inline -fno-defer-pop -fmove-loop-invariants @gol
307 -fno-function-cse -fno-guess-branch-probability @gol
308 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
309 -funsafe-math-optimizations -funsafe-loop-optimizations -ffinite-math-only @gol
310 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
311 -fomit-frame-pointer -foptimize-register-move @gol
312 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
313 -fprofile-generate -fprofile-use @gol
314 -fregmove -frename-registers @gol
315 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
316 -frerun-cse-after-loop @gol
317 -frounding-math -frtl-abstract-sequences @gol
318 -fschedule-insns -fschedule-insns2 @gol
319 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
320 -fsched-spec-load-dangerous @gol
321 -fsched-stalled-insns=@var{n} -fsched-stalled-insns-dep=@var{n} @gol
322 -fsched2-use-superblocks @gol
323 -fsched2-use-traces -fsee -freschedule-modulo-scheduled-loops @gol
324 -fsection-anchors -fsignaling-nans -fsingle-precision-constant @gol
325 -fstack-protector -fstack-protector-all @gol
326 -fstrict-aliasing -fstrict-overflow -ftracer -fthread-jumps @gol
327 -funroll-all-loops -funroll-loops -fpeel-loops @gol
328 -fsplit-ivs-in-unroller -funswitch-loops @gol
329 -fvariable-expansion-in-unroller @gol
330 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
331 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
332 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
333 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
334 -ftree-vect-loop-version -ftree-salias -fipa-pta -fweb @gol
335 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
336 --param @var{name}=@var{value}
337 -O -O0 -O1 -O2 -O3 -Os}
339 @item Preprocessor Options
340 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
341 @gccoptlist{-A@var{question}=@var{answer} @gol
342 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
343 -C -dD -dI -dM -dN @gol
344 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
345 -idirafter @var{dir} @gol
346 -include @var{file} -imacros @var{file} @gol
347 -iprefix @var{file} -iwithprefix @var{dir} @gol
348 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
349 -imultilib @var{dir} -isysroot @var{dir} @gol
350 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
351 -P -fworking-directory -remap @gol
352 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
353 -Xpreprocessor @var{option}}
355 @item Assembler Option
356 @xref{Assembler Options,,Passing Options to the Assembler}.
357 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
360 @xref{Link Options,,Options for Linking}.
361 @gccoptlist{@var{object-file-name} -l@var{library} @gol
362 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
363 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
364 -Wl,@var{option} -Xlinker @var{option} @gol
367 @item Directory Options
368 @xref{Directory Options,,Options for Directory Search}.
369 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
370 -specs=@var{file} -I- --sysroot=@var{dir}}
373 @c I wrote this xref this way to avoid overfull hbox. -- rms
374 @xref{Target Options}.
375 @gccoptlist{-V @var{version} -b @var{machine}}
377 @item Machine Dependent Options
378 @xref{Submodel Options,,Hardware Models and Configurations}.
379 @c This list is ordered alphanumerically by subsection name.
380 @c Try and put the significant identifier (CPU or system) first,
381 @c so users have a clue at guessing where the ones they want will be.
384 @gccoptlist{-EB -EL @gol
385 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
386 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
389 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
390 -mabi=@var{name} @gol
391 -mapcs-stack-check -mno-apcs-stack-check @gol
392 -mapcs-float -mno-apcs-float @gol
393 -mapcs-reentrant -mno-apcs-reentrant @gol
394 -msched-prolog -mno-sched-prolog @gol
395 -mlittle-endian -mbig-endian -mwords-little-endian @gol
396 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
397 -mthumb-interwork -mno-thumb-interwork @gol
398 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
399 -mstructure-size-boundary=@var{n} @gol
400 -mabort-on-noreturn @gol
401 -mlong-calls -mno-long-calls @gol
402 -msingle-pic-base -mno-single-pic-base @gol
403 -mpic-register=@var{reg} @gol
404 -mnop-fun-dllimport @gol
405 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
406 -mpoke-function-name @gol
408 -mtpcs-frame -mtpcs-leaf-frame @gol
409 -mcaller-super-interworking -mcallee-super-interworking @gol
413 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
414 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
416 @emph{Blackfin Options}
417 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
418 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
419 -mlow-64k -mno-low64k -mid-shared-library @gol
420 -mno-id-shared-library -mshared-library-id=@var{n} @gol
421 -mlong-calls -mno-long-calls}
424 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
425 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
426 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
427 -mstack-align -mdata-align -mconst-align @gol
428 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
429 -melf -maout -melinux -mlinux -sim -sim2 @gol
430 -mmul-bug-workaround -mno-mul-bug-workaround}
433 @gccoptlist{-mmac -mpush-args}
435 @emph{Darwin Options}
436 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
437 -arch_only -bind_at_load -bundle -bundle_loader @gol
438 -client_name -compatibility_version -current_version @gol
440 -dependency-file -dylib_file -dylinker_install_name @gol
441 -dynamic -dynamiclib -exported_symbols_list @gol
442 -filelist -flat_namespace -force_cpusubtype_ALL @gol
443 -force_flat_namespace -headerpad_max_install_names @gol
444 -image_base -init -install_name -keep_private_externs @gol
445 -multi_module -multiply_defined -multiply_defined_unused @gol
446 -noall_load -no_dead_strip_inits_and_terms @gol
447 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
448 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
449 -private_bundle -read_only_relocs -sectalign @gol
450 -sectobjectsymbols -whyload -seg1addr @gol
451 -sectcreate -sectobjectsymbols -sectorder @gol
452 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
453 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
454 -segprot -segs_read_only_addr -segs_read_write_addr @gol
455 -single_module -static -sub_library -sub_umbrella @gol
456 -twolevel_namespace -umbrella -undefined @gol
457 -unexported_symbols_list -weak_reference_mismatches @gol
458 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
459 -mkernel -mone-byte-bool}
461 @emph{DEC Alpha Options}
462 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
463 -mieee -mieee-with-inexact -mieee-conformant @gol
464 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
465 -mtrap-precision=@var{mode} -mbuild-constants @gol
466 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
467 -mbwx -mmax -mfix -mcix @gol
468 -mfloat-vax -mfloat-ieee @gol
469 -mexplicit-relocs -msmall-data -mlarge-data @gol
470 -msmall-text -mlarge-text @gol
471 -mmemory-latency=@var{time}}
473 @emph{DEC Alpha/VMS Options}
474 @gccoptlist{-mvms-return-codes}
477 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
478 -mhard-float -msoft-float @gol
479 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
480 -mdouble -mno-double @gol
481 -mmedia -mno-media -mmuladd -mno-muladd @gol
482 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
483 -mlinked-fp -mlong-calls -malign-labels @gol
484 -mlibrary-pic -macc-4 -macc-8 @gol
485 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
486 -moptimize-membar -mno-optimize-membar @gol
487 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
488 -mvliw-branch -mno-vliw-branch @gol
489 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
490 -mno-nested-cond-exec -mtomcat-stats @gol
494 @emph{GNU/Linux Options}
495 @gccoptlist{-muclibc}
497 @emph{H8/300 Options}
498 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
501 @gccoptlist{-march=@var{architecture-type} @gol
502 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
503 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
504 -mfixed-range=@var{register-range} @gol
505 -mjump-in-delay -mlinker-opt -mlong-calls @gol
506 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
507 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
508 -mno-jump-in-delay -mno-long-load-store @gol
509 -mno-portable-runtime -mno-soft-float @gol
510 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
511 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
512 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
513 -munix=@var{unix-std} -nolibdld -static -threads}
515 @emph{i386 and x86-64 Options}
516 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
517 -mfpmath=@var{unit} @gol
518 -masm=@var{dialect} -mno-fancy-math-387 @gol
519 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
520 -mno-wide-multiply -mrtd -malign-double @gol
521 -mpreferred-stack-boundary=@var{num} @gol
522 -mmmx -msse -msse2 -msse3 -mssse3 -msse4a -m3dnow -mpopcnt -mabm -maes @gol
523 -mthreads -mno-align-stringops -minline-all-stringops @gol
524 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
525 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
527 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
528 -mcmodel=@var{code-model} @gol
529 -m32 -m64 -mlarge-data-threshold=@var{num}}
532 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
533 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
534 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
535 -minline-float-divide-max-throughput @gol
536 -minline-int-divide-min-latency @gol
537 -minline-int-divide-max-throughput @gol
538 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
539 -mno-dwarf2-asm -mearly-stop-bits @gol
540 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
541 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
542 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
543 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
544 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
545 -mno-sched-prefer-non-data-spec-insns @gol
546 -mno-sched-prefer-non-control-spec-insns @gol
547 -mno-sched-count-spec-in-critical-path}
549 @emph{M32R/D Options}
550 @gccoptlist{-m32r2 -m32rx -m32r @gol
552 -malign-loops -mno-align-loops @gol
553 -missue-rate=@var{number} @gol
554 -mbranch-cost=@var{number} @gol
555 -mmodel=@var{code-size-model-type} @gol
556 -msdata=@var{sdata-type} @gol
557 -mno-flush-func -mflush-func=@var{name} @gol
558 -mno-flush-trap -mflush-trap=@var{number} @gol
562 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
564 @emph{M680x0 Options}
565 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
566 -m68060 -mcpu32 -m5200 -mcfv4e -m68881 -mbitfield @gol
567 -mc68000 -mc68020 @gol
568 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
569 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
570 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
572 @emph{M68hc1x Options}
573 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
574 -mauto-incdec -minmax -mlong-calls -mshort @gol
575 -msoft-reg-count=@var{count}}
578 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
579 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
580 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
581 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
582 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
585 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
586 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
587 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
588 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
589 -mfp32 -mfp64 -mhard-float -msoft-float @gol
590 -msingle-float -mdouble-float -mdsp -mpaired-single -mips3d @gol
591 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
592 -G@var{num} -membedded-data -mno-embedded-data @gol
593 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
594 -msplit-addresses -mno-split-addresses @gol
595 -mexplicit-relocs -mno-explicit-relocs @gol
596 -mcheck-zero-division -mno-check-zero-division @gol
597 -mdivide-traps -mdivide-breaks @gol
598 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
599 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
600 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
601 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
602 -mfix-sb1 -mno-fix-sb1 @gol
603 -mflush-func=@var{func} -mno-flush-func @gol
604 -mbranch-likely -mno-branch-likely @gol
605 -mfp-exceptions -mno-fp-exceptions @gol
606 -mvr4130-align -mno-vr4130-align}
609 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
610 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
611 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
612 -mno-base-addresses -msingle-exit -mno-single-exit}
614 @emph{MN10300 Options}
615 @gccoptlist{-mmult-bug -mno-mult-bug @gol
616 -mam33 -mno-am33 @gol
617 -mam33-2 -mno-am33-2 @gol
618 -mreturn-pointer-on-d0 @gol
622 @gccoptlist{-mno-crt0 -mbacc -msim @gol
623 -march=@var{cpu-type} }
625 @emph{PDP-11 Options}
626 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
627 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
628 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
629 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
630 -mbranch-expensive -mbranch-cheap @gol
631 -msplit -mno-split -munix-asm -mdec-asm}
633 @emph{PowerPC Options}
634 See RS/6000 and PowerPC Options.
636 @emph{RS/6000 and PowerPC Options}
637 @gccoptlist{-mcpu=@var{cpu-type} @gol
638 -mtune=@var{cpu-type} @gol
639 -mpower -mno-power -mpower2 -mno-power2 @gol
640 -mpowerpc -mpowerpc64 -mno-powerpc @gol
641 -maltivec -mno-altivec @gol
642 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
643 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
644 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
645 -mnew-mnemonics -mold-mnemonics @gol
646 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
647 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
648 -malign-power -malign-natural @gol
649 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
650 -mstring -mno-string -mupdate -mno-update @gol
651 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
652 -mstrict-align -mno-strict-align -mrelocatable @gol
653 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
654 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
655 -mdynamic-no-pic -maltivec -mswdiv @gol
656 -mprioritize-restricted-insns=@var{priority} @gol
657 -msched-costly-dep=@var{dependence_type} @gol
658 -minsert-sched-nops=@var{scheme} @gol
659 -mcall-sysv -mcall-netbsd @gol
660 -maix-struct-return -msvr4-struct-return @gol
661 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
662 -misel -mno-isel @gol
663 -misel=yes -misel=no @gol
665 -mspe=yes -mspe=no @gol
666 -mvrsave -mno-vrsave @gol
667 -mmulhw -mno-mulhw @gol
668 -mdlmzb -mno-dlmzb @gol
669 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
670 -mprototype -mno-prototype @gol
671 -msim -mmvme -mads -myellowknife -memb -msdata @gol
672 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
674 @emph{S/390 and zSeries Options}
675 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
676 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
677 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
678 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
679 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
680 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
681 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
684 @gccoptlist{-meb -mel @gol
688 -mscore5 -mscore5u -mscore7 -mscore7d}
691 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
692 -m4-nofpu -m4-single-only -m4-single -m4 @gol
693 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
694 -m5-64media -m5-64media-nofpu @gol
695 -m5-32media -m5-32media-nofpu @gol
696 -m5-compact -m5-compact-nofpu @gol
697 -mb -ml -mdalign -mrelax @gol
698 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
699 -mieee -misize -mpadstruct -mspace @gol
700 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
701 -mdivsi3_libfunc=@var{name} @gol
702 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
706 @gccoptlist{-mcpu=@var{cpu-type} @gol
707 -mtune=@var{cpu-type} @gol
708 -mcmodel=@var{code-model} @gol
709 -m32 -m64 -mapp-regs -mno-app-regs @gol
710 -mfaster-structs -mno-faster-structs @gol
711 -mfpu -mno-fpu -mhard-float -msoft-float @gol
712 -mhard-quad-float -msoft-quad-float @gol
713 -mimpure-text -mno-impure-text -mlittle-endian @gol
714 -mstack-bias -mno-stack-bias @gol
715 -munaligned-doubles -mno-unaligned-doubles @gol
716 -mv8plus -mno-v8plus -mvis -mno-vis
717 -threads -pthreads -pthread}
719 @emph{System V Options}
720 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
722 @emph{TMS320C3x/C4x Options}
723 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
724 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
725 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
726 -mparallel-insns -mparallel-mpy -mpreserve-float}
729 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
730 -mprolog-function -mno-prolog-function -mspace @gol
731 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
732 -mapp-regs -mno-app-regs @gol
733 -mdisable-callt -mno-disable-callt @gol
739 @gccoptlist{-mg -mgnu -munix}
741 @emph{x86-64 Options}
742 See i386 and x86-64 Options.
744 @emph{Xstormy16 Options}
747 @emph{Xtensa Options}
748 @gccoptlist{-mconst16 -mno-const16 @gol
749 -mfused-madd -mno-fused-madd @gol
750 -mtext-section-literals -mno-text-section-literals @gol
751 -mtarget-align -mno-target-align @gol
752 -mlongcalls -mno-longcalls}
754 @emph{zSeries Options}
755 See S/390 and zSeries Options.
757 @item Code Generation Options
758 @xref{Code Gen Options,,Options for Code Generation Conventions}.
759 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
760 -ffixed-@var{reg} -fexceptions @gol
761 -fnon-call-exceptions -funwind-tables @gol
762 -fasynchronous-unwind-tables @gol
763 -finhibit-size-directive -finstrument-functions @gol
764 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
765 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
766 -fno-common -fno-ident @gol
767 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
768 -fno-jump-tables @gol
769 -freg-struct-return -fshort-enums @gol
770 -fshort-double -fshort-wchar @gol
771 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
772 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
773 -fargument-alias -fargument-noalias @gol
774 -fargument-noalias-global -fargument-noalias-anything
775 -fleading-underscore -ftls-model=@var{model} @gol
776 -ftrapv -fwrapv -fbounds-check @gol
781 * Overall Options:: Controlling the kind of output:
782 an executable, object files, assembler files,
783 or preprocessed source.
784 * C Dialect Options:: Controlling the variant of C language compiled.
785 * C++ Dialect Options:: Variations on C++.
786 * Language Independent Options:: Controlling how diagnostics should be
788 * Warning Options:: How picky should the compiler be?
789 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
790 * Optimize Options:: How much optimization?
791 * Preprocessor Options:: Controlling header files and macro definitions.
792 Also, getting dependency information for Make.
793 * Assembler Options:: Passing options to the assembler.
794 * Link Options:: Specifying libraries and so on.
795 * Directory Options:: Where to find header files and libraries.
796 Where to find the compiler executable files.
797 * Spec Files:: How to pass switches to sub-processes.
798 * Target Options:: Running a cross-compiler, or an old version of GCC.
801 @node Overall Options
802 @section Options Controlling the Kind of Output
804 Compilation can involve up to four stages: preprocessing, compilation
805 proper, assembly and linking, always in that order. GCC is capable of
806 preprocessing and compiling several files either into several
807 assembler input files, or into one assembler input file; then each
808 assembler input file produces an object file, and linking combines all
809 the object files (those newly compiled, and those specified as input)
810 into an executable file.
812 @cindex file name suffix
813 For any given input file, the file name suffix determines what kind of
818 C source code which must be preprocessed.
821 C source code which should not be preprocessed.
824 C++ source code which should not be preprocessed.
827 C, or C++ header file to be turned into a precompiled header.
831 @itemx @var{file}.cxx
832 @itemx @var{file}.cpp
833 @itemx @var{file}.CPP
834 @itemx @var{file}.c++
836 C++ source code which must be preprocessed. Note that in @samp{.cxx},
837 the last two letters must both be literally @samp{x}. Likewise,
838 @samp{.C} refers to a literal capital C@.
842 C++ header file to be turned into a precompiled header.
845 @itemx @var{file}.for
846 @itemx @var{file}.FOR
847 Fixed form Fortran source code which should not be preprocessed.
850 @itemx @var{file}.fpp
851 @itemx @var{file}.FPP
852 Fixed form Fortran source code which must be preprocessed (with the traditional
856 @itemx @var{file}.f95
857 Free form Fortran source code which should not be preprocessed.
860 @itemx @var{file}.F95
861 Free form Fortran source code which must be preprocessed (with the
862 traditional preprocessor).
864 @c FIXME: Descriptions of Java file types.
871 Ada source code file which contains a library unit declaration (a
872 declaration of a package, subprogram, or generic, or a generic
873 instantiation), or a library unit renaming declaration (a package,
874 generic, or subprogram renaming declaration). Such files are also
877 @itemx @var{file}.adb
878 Ada source code file containing a library unit body (a subprogram or
879 package body). Such files are also called @dfn{bodies}.
881 @c GCC also knows about some suffixes for languages not yet included:
892 Assembler code which must be preprocessed.
895 An object file to be fed straight into linking.
896 Any file name with no recognized suffix is treated this way.
900 You can specify the input language explicitly with the @option{-x} option:
903 @item -x @var{language}
904 Specify explicitly the @var{language} for the following input files
905 (rather than letting the compiler choose a default based on the file
906 name suffix). This option applies to all following input files until
907 the next @option{-x} option. Possible values for @var{language} are:
909 c c-header c-cpp-output
910 c++ c++-header c++-cpp-output
911 assembler assembler-with-cpp
919 Turn off any specification of a language, so that subsequent files are
920 handled according to their file name suffixes (as they are if @option{-x}
921 has not been used at all).
923 @item -pass-exit-codes
924 @opindex pass-exit-codes
925 Normally the @command{gcc} program will exit with the code of 1 if any
926 phase of the compiler returns a non-success return code. If you specify
927 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
928 numerically highest error produced by any phase that returned an error
929 indication. The C, C++, and Fortran frontends return 4, if an internal
930 compiler error is encountered.
933 If you only want some of the stages of compilation, you can use
934 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
935 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
936 @command{gcc} is to stop. Note that some combinations (for example,
937 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
942 Compile or assemble the source files, but do not link. The linking
943 stage simply is not done. The ultimate output is in the form of an
944 object file for each source file.
946 By default, the object file name for a source file is made by replacing
947 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
949 Unrecognized input files, not requiring compilation or assembly, are
954 Stop after the stage of compilation proper; do not assemble. The output
955 is in the form of an assembler code file for each non-assembler input
958 By default, the assembler file name for a source file is made by
959 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
961 Input files that don't require compilation are ignored.
965 Stop after the preprocessing stage; do not run the compiler proper. The
966 output is in the form of preprocessed source code, which is sent to the
969 Input files which don't require preprocessing are ignored.
971 @cindex output file option
974 Place output in file @var{file}. This applies regardless to whatever
975 sort of output is being produced, whether it be an executable file,
976 an object file, an assembler file or preprocessed C code.
978 If @option{-o} is not specified, the default is to put an executable
979 file in @file{a.out}, the object file for
980 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
981 assembler file in @file{@var{source}.s}, a precompiled header file in
982 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
987 Print (on standard error output) the commands executed to run the stages
988 of compilation. Also print the version number of the compiler driver
989 program and of the preprocessor and the compiler proper.
993 Like @option{-v} except the commands are not executed and all command
994 arguments are quoted. This is useful for shell scripts to capture the
995 driver-generated command lines.
999 Use pipes rather than temporary files for communication between the
1000 various stages of compilation. This fails to work on some systems where
1001 the assembler is unable to read from a pipe; but the GNU assembler has
1006 If you are compiling multiple source files, this option tells the driver
1007 to pass all the source files to the compiler at once (for those
1008 languages for which the compiler can handle this). This will allow
1009 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1010 language for which this is supported is C@. If you pass source files for
1011 multiple languages to the driver, using this option, the driver will invoke
1012 the compiler(s) that support IMA once each, passing each compiler all the
1013 source files appropriate for it. For those languages that do not support
1014 IMA this option will be ignored, and the compiler will be invoked once for
1015 each source file in that language. If you use this option in conjunction
1016 with @option{-save-temps}, the compiler will generate multiple
1018 (one for each source file), but only one (combined) @file{.o} or
1023 Print (on the standard output) a description of the command line options
1024 understood by @command{gcc}. If the @option{-v} option is also specified
1025 then @option{--help} will also be passed on to the various processes
1026 invoked by @command{gcc}, so that they can display the command line options
1027 they accept. If the @option{-Wextra} option is also specified then command
1028 line options which have no documentation associated with them will also
1032 @opindex target-help
1033 Print (on the standard output) a description of target specific command
1034 line options for each tool.
1038 Display the version number and copyrights of the invoked GCC@.
1040 @include @value{srcdir}/../libiberty/at-file.texi
1044 @section Compiling C++ Programs
1046 @cindex suffixes for C++ source
1047 @cindex C++ source file suffixes
1048 C++ source files conventionally use one of the suffixes @samp{.C},
1049 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1050 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1051 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1052 files with these names and compiles them as C++ programs even if you
1053 call the compiler the same way as for compiling C programs (usually
1054 with the name @command{gcc}).
1058 However, the use of @command{gcc} does not add the C++ library.
1059 @command{g++} is a program that calls GCC and treats @samp{.c},
1060 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1061 files unless @option{-x} is used, and automatically specifies linking
1062 against the C++ library. This program is also useful when
1063 precompiling a C header file with a @samp{.h} extension for use in C++
1064 compilations. On many systems, @command{g++} is also installed with
1065 the name @command{c++}.
1067 @cindex invoking @command{g++}
1068 When you compile C++ programs, you may specify many of the same
1069 command-line options that you use for compiling programs in any
1070 language; or command-line options meaningful for C and related
1071 languages; or options that are meaningful only for C++ programs.
1072 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1073 explanations of options for languages related to C@.
1074 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1075 explanations of options that are meaningful only for C++ programs.
1077 @node C Dialect Options
1078 @section Options Controlling C Dialect
1079 @cindex dialect options
1080 @cindex language dialect options
1081 @cindex options, dialect
1083 The following options control the dialect of C (or languages derived
1084 from C, such as C++) that the compiler accepts:
1087 @cindex ANSI support
1091 In C mode, support all ISO C90 programs. In C++ mode,
1092 remove GNU extensions that conflict with ISO C++.
1094 This turns off certain features of GCC that are incompatible with ISO
1095 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1096 such as the @code{asm} and @code{typeof} keywords, and
1097 predefined macros such as @code{unix} and @code{vax} that identify the
1098 type of system you are using. It also enables the undesirable and
1099 rarely used ISO trigraph feature. For the C compiler,
1100 it disables recognition of C++ style @samp{//} comments as well as
1101 the @code{inline} keyword.
1103 The alternate keywords @code{__asm__}, @code{__extension__},
1104 @code{__inline__} and @code{__typeof__} continue to work despite
1105 @option{-ansi}. You would not want to use them in an ISO C program, of
1106 course, but it is useful to put them in header files that might be included
1107 in compilations done with @option{-ansi}. Alternate predefined macros
1108 such as @code{__unix__} and @code{__vax__} are also available, with or
1109 without @option{-ansi}.
1111 The @option{-ansi} option does not cause non-ISO programs to be
1112 rejected gratuitously. For that, @option{-pedantic} is required in
1113 addition to @option{-ansi}. @xref{Warning Options}.
1115 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1116 option is used. Some header files may notice this macro and refrain
1117 from declaring certain functions or defining certain macros that the
1118 ISO standard doesn't call for; this is to avoid interfering with any
1119 programs that might use these names for other things.
1121 Functions which would normally be built in but do not have semantics
1122 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1123 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1124 built-in functions provided by GCC}, for details of the functions
1129 Determine the language standard. This option is currently only
1130 supported when compiling C or C++. A value for this option must be
1131 provided; possible values are
1136 ISO C90 (same as @option{-ansi}).
1138 @item iso9899:199409
1139 ISO C90 as modified in amendment 1.
1145 ISO C99. Note that this standard is not yet fully supported; see
1146 @w{@uref{http://gcc.gnu.org/gcc-4.2/c99status.html}} for more information. The
1147 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1150 Default, ISO C90 plus GNU extensions (including some C99 features).
1154 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1155 this will become the default. The name @samp{gnu9x} is deprecated.
1158 The 1998 ISO C++ standard plus amendments.
1161 The same as @option{-std=c++98} plus GNU extensions. This is the
1162 default for C++ code.
1165 Even when this option is not specified, you can still use some of the
1166 features of newer standards in so far as they do not conflict with
1167 previous C standards. For example, you may use @code{__restrict__} even
1168 when @option{-std=c99} is not specified.
1170 The @option{-std} options specifying some version of ISO C have the same
1171 effects as @option{-ansi}, except that features that were not in ISO C90
1172 but are in the specified version (for example, @samp{//} comments and
1173 the @code{inline} keyword in ISO C99) are not disabled.
1175 @xref{Standards,,Language Standards Supported by GCC}, for details of
1176 these standard versions.
1178 @item -fgnu89-inline
1179 @opindex fgnu89-inline
1180 The option @option{-fgnu89-inline} tells GCC to use the traditional
1181 GNU semantics for @code{inline} functions when in C99 mode.
1182 @xref{Inline,,An Inline Function is As Fast As a Macro}. Using this
1183 option is roughly equivalent to adding the @code{gnu_inline} function
1184 attribute to all inline functions (@pxref{Function Attributes}).
1186 This option is accepted by GCC versions 4.1.3 and up. In GCC versions
1187 prior to 4.3, C99 inline semantics are not supported, and thus this
1188 option is effectively assumed to be present regardless of whether or not
1189 it is specified; the only effect of specifying it explicitly is to
1190 disable warnings about using inline functions in C99 mode. Likewise,
1191 the option @option{-fno-gnu89-inline} is not supported in versions of
1192 GCC before 4.3. It will be supported only in C99 or gnu99 mode, not in
1195 The preprocesor macros @code{__GNUC_GNU_INLINE__} and
1196 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1197 in effect for @code{inline} functions. @xref{Common Predefined
1198 Macros,,,cpp,The C Preprocessor}.
1200 @item -aux-info @var{filename}
1202 Output to the given filename prototyped declarations for all functions
1203 declared and/or defined in a translation unit, including those in header
1204 files. This option is silently ignored in any language other than C@.
1206 Besides declarations, the file indicates, in comments, the origin of
1207 each declaration (source file and line), whether the declaration was
1208 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1209 @samp{O} for old, respectively, in the first character after the line
1210 number and the colon), and whether it came from a declaration or a
1211 definition (@samp{C} or @samp{F}, respectively, in the following
1212 character). In the case of function definitions, a K&R-style list of
1213 arguments followed by their declarations is also provided, inside
1214 comments, after the declaration.
1218 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1219 keyword, so that code can use these words as identifiers. You can use
1220 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1221 instead. @option{-ansi} implies @option{-fno-asm}.
1223 In C++, this switch only affects the @code{typeof} keyword, since
1224 @code{asm} and @code{inline} are standard keywords. You may want to
1225 use the @option{-fno-gnu-keywords} flag instead, which has the same
1226 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1227 switch only affects the @code{asm} and @code{typeof} keywords, since
1228 @code{inline} is a standard keyword in ISO C99.
1231 @itemx -fno-builtin-@var{function}
1232 @opindex fno-builtin
1233 @cindex built-in functions
1234 Don't recognize built-in functions that do not begin with
1235 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1236 functions provided by GCC}, for details of the functions affected,
1237 including those which are not built-in functions when @option{-ansi} or
1238 @option{-std} options for strict ISO C conformance are used because they
1239 do not have an ISO standard meaning.
1241 GCC normally generates special code to handle certain built-in functions
1242 more efficiently; for instance, calls to @code{alloca} may become single
1243 instructions that adjust the stack directly, and calls to @code{memcpy}
1244 may become inline copy loops. The resulting code is often both smaller
1245 and faster, but since the function calls no longer appear as such, you
1246 cannot set a breakpoint on those calls, nor can you change the behavior
1247 of the functions by linking with a different library. In addition,
1248 when a function is recognized as a built-in function, GCC may use
1249 information about that function to warn about problems with calls to
1250 that function, or to generate more efficient code, even if the
1251 resulting code still contains calls to that function. For example,
1252 warnings are given with @option{-Wformat} for bad calls to
1253 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1254 known not to modify global memory.
1256 With the @option{-fno-builtin-@var{function}} option
1257 only the built-in function @var{function} is
1258 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1259 function is named this is not built-in in this version of GCC, this
1260 option is ignored. There is no corresponding
1261 @option{-fbuiltin-@var{function}} option; if you wish to enable
1262 built-in functions selectively when using @option{-fno-builtin} or
1263 @option{-ffreestanding}, you may define macros such as:
1266 #define abs(n) __builtin_abs ((n))
1267 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1272 @cindex hosted environment
1274 Assert that compilation takes place in a hosted environment. This implies
1275 @option{-fbuiltin}. A hosted environment is one in which the
1276 entire standard library is available, and in which @code{main} has a return
1277 type of @code{int}. Examples are nearly everything except a kernel.
1278 This is equivalent to @option{-fno-freestanding}.
1280 @item -ffreestanding
1281 @opindex ffreestanding
1282 @cindex hosted environment
1284 Assert that compilation takes place in a freestanding environment. This
1285 implies @option{-fno-builtin}. A freestanding environment
1286 is one in which the standard library may not exist, and program startup may
1287 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1288 This is equivalent to @option{-fno-hosted}.
1290 @xref{Standards,,Language Standards Supported by GCC}, for details of
1291 freestanding and hosted environments.
1295 @cindex openmp parallel
1296 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1297 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1298 compiler generates parallel code according to the OpenMP Application
1299 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}.
1301 @item -fms-extensions
1302 @opindex fms-extensions
1303 Accept some non-standard constructs used in Microsoft header files.
1305 Some cases of unnamed fields in structures and unions are only
1306 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1307 fields within structs/unions}, for details.
1311 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1312 options for strict ISO C conformance) implies @option{-trigraphs}.
1314 @item -no-integrated-cpp
1315 @opindex no-integrated-cpp
1316 Performs a compilation in two passes: preprocessing and compiling. This
1317 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1318 @option{-B} option. The user supplied compilation step can then add in
1319 an additional preprocessing step after normal preprocessing but before
1320 compiling. The default is to use the integrated cpp (internal cpp)
1322 The semantics of this option will change if "cc1", "cc1plus", and
1323 "cc1obj" are merged.
1325 @cindex traditional C language
1326 @cindex C language, traditional
1328 @itemx -traditional-cpp
1329 @opindex traditional-cpp
1330 @opindex traditional
1331 Formerly, these options caused GCC to attempt to emulate a pre-standard
1332 C compiler. They are now only supported with the @option{-E} switch.
1333 The preprocessor continues to support a pre-standard mode. See the GNU
1334 CPP manual for details.
1336 @item -fcond-mismatch
1337 @opindex fcond-mismatch
1338 Allow conditional expressions with mismatched types in the second and
1339 third arguments. The value of such an expression is void. This option
1340 is not supported for C++.
1342 @item -funsigned-char
1343 @opindex funsigned-char
1344 Let the type @code{char} be unsigned, like @code{unsigned char}.
1346 Each kind of machine has a default for what @code{char} should
1347 be. It is either like @code{unsigned char} by default or like
1348 @code{signed char} by default.
1350 Ideally, a portable program should always use @code{signed char} or
1351 @code{unsigned char} when it depends on the signedness of an object.
1352 But many programs have been written to use plain @code{char} and
1353 expect it to be signed, or expect it to be unsigned, depending on the
1354 machines they were written for. This option, and its inverse, let you
1355 make such a program work with the opposite default.
1357 The type @code{char} is always a distinct type from each of
1358 @code{signed char} or @code{unsigned char}, even though its behavior
1359 is always just like one of those two.
1362 @opindex fsigned-char
1363 Let the type @code{char} be signed, like @code{signed char}.
1365 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1366 the negative form of @option{-funsigned-char}. Likewise, the option
1367 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1369 @item -fsigned-bitfields
1370 @itemx -funsigned-bitfields
1371 @itemx -fno-signed-bitfields
1372 @itemx -fno-unsigned-bitfields
1373 @opindex fsigned-bitfields
1374 @opindex funsigned-bitfields
1375 @opindex fno-signed-bitfields
1376 @opindex fno-unsigned-bitfields
1377 These options control whether a bit-field is signed or unsigned, when the
1378 declaration does not use either @code{signed} or @code{unsigned}. By
1379 default, such a bit-field is signed, because this is consistent: the
1380 basic integer types such as @code{int} are signed types.
1383 @node C++ Dialect Options
1384 @section Options Controlling C++ Dialect
1386 @cindex compiler options, C++
1387 @cindex C++ options, command line
1388 @cindex options, C++
1389 This section describes the command-line options that are only meaningful
1390 for C++ programs; but you can also use most of the GNU compiler options
1391 regardless of what language your program is in. For example, you
1392 might compile a file @code{firstClass.C} like this:
1395 g++ -g -frepo -O -c firstClass.C
1399 In this example, only @option{-frepo} is an option meant
1400 only for C++ programs; you can use the other options with any
1401 language supported by GCC@.
1403 Here is a list of options that are @emph{only} for compiling C++ programs:
1407 @item -fabi-version=@var{n}
1408 @opindex fabi-version
1409 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1410 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1411 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1412 the version that conforms most closely to the C++ ABI specification.
1413 Therefore, the ABI obtained using version 0 will change as ABI bugs
1416 The default is version 2.
1418 @item -fno-access-control
1419 @opindex fno-access-control
1420 Turn off all access checking. This switch is mainly useful for working
1421 around bugs in the access control code.
1425 Check that the pointer returned by @code{operator new} is non-null
1426 before attempting to modify the storage allocated. This check is
1427 normally unnecessary because the C++ standard specifies that
1428 @code{operator new} will only return @code{0} if it is declared
1429 @samp{throw()}, in which case the compiler will always check the
1430 return value even without this option. In all other cases, when
1431 @code{operator new} has a non-empty exception specification, memory
1432 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1433 @samp{new (nothrow)}.
1435 @item -fconserve-space
1436 @opindex fconserve-space
1437 Put uninitialized or runtime-initialized global variables into the
1438 common segment, as C does. This saves space in the executable at the
1439 cost of not diagnosing duplicate definitions. If you compile with this
1440 flag and your program mysteriously crashes after @code{main()} has
1441 completed, you may have an object that is being destroyed twice because
1442 two definitions were merged.
1444 This option is no longer useful on most targets, now that support has
1445 been added for putting variables into BSS without making them common.
1447 @item -ffriend-injection
1448 @opindex ffriend-injection
1449 Inject friend functions into the enclosing namespace, so that they are
1450 visible outside the scope of the class in which they are declared.
1451 Friend functions were documented to work this way in the old Annotated
1452 C++ Reference Manual, and versions of G++ before 4.1 always worked
1453 that way. However, in ISO C++ a friend function which is not declared
1454 in an enclosing scope can only be found using argument dependent
1455 lookup. This option causes friends to be injected as they were in
1458 This option is for compatibility, and may be removed in a future
1461 @item -fno-elide-constructors
1462 @opindex fno-elide-constructors
1463 The C++ standard allows an implementation to omit creating a temporary
1464 which is only used to initialize another object of the same type.
1465 Specifying this option disables that optimization, and forces G++ to
1466 call the copy constructor in all cases.
1468 @item -fno-enforce-eh-specs
1469 @opindex fno-enforce-eh-specs
1470 Don't generate code to check for violation of exception specifications
1471 at runtime. This option violates the C++ standard, but may be useful
1472 for reducing code size in production builds, much like defining
1473 @samp{NDEBUG}. This does not give user code permission to throw
1474 exceptions in violation of the exception specifications; the compiler
1475 will still optimize based on the specifications, so throwing an
1476 unexpected exception will result in undefined behavior.
1479 @itemx -fno-for-scope
1481 @opindex fno-for-scope
1482 If @option{-ffor-scope} is specified, the scope of variables declared in
1483 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1484 as specified by the C++ standard.
1485 If @option{-fno-for-scope} is specified, the scope of variables declared in
1486 a @i{for-init-statement} extends to the end of the enclosing scope,
1487 as was the case in old versions of G++, and other (traditional)
1488 implementations of C++.
1490 The default if neither flag is given to follow the standard,
1491 but to allow and give a warning for old-style code that would
1492 otherwise be invalid, or have different behavior.
1494 @item -fno-gnu-keywords
1495 @opindex fno-gnu-keywords
1496 Do not recognize @code{typeof} as a keyword, so that code can use this
1497 word as an identifier. You can use the keyword @code{__typeof__} instead.
1498 @option{-ansi} implies @option{-fno-gnu-keywords}.
1500 @item -fno-implicit-templates
1501 @opindex fno-implicit-templates
1502 Never emit code for non-inline templates which are instantiated
1503 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1504 @xref{Template Instantiation}, for more information.
1506 @item -fno-implicit-inline-templates
1507 @opindex fno-implicit-inline-templates
1508 Don't emit code for implicit instantiations of inline templates, either.
1509 The default is to handle inlines differently so that compiles with and
1510 without optimization will need the same set of explicit instantiations.
1512 @item -fno-implement-inlines
1513 @opindex fno-implement-inlines
1514 To save space, do not emit out-of-line copies of inline functions
1515 controlled by @samp{#pragma implementation}. This will cause linker
1516 errors if these functions are not inlined everywhere they are called.
1518 @item -fms-extensions
1519 @opindex fms-extensions
1520 Disable pedantic warnings about constructs used in MFC, such as implicit
1521 int and getting a pointer to member function via non-standard syntax.
1523 @item -fno-nonansi-builtins
1524 @opindex fno-nonansi-builtins
1525 Disable built-in declarations of functions that are not mandated by
1526 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1527 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1529 @item -fno-operator-names
1530 @opindex fno-operator-names
1531 Do not treat the operator name keywords @code{and}, @code{bitand},
1532 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1533 synonyms as keywords.
1535 @item -fno-optional-diags
1536 @opindex fno-optional-diags
1537 Disable diagnostics that the standard says a compiler does not need to
1538 issue. Currently, the only such diagnostic issued by G++ is the one for
1539 a name having multiple meanings within a class.
1542 @opindex fpermissive
1543 Downgrade some diagnostics about nonconformant code from errors to
1544 warnings. Thus, using @option{-fpermissive} will allow some
1545 nonconforming code to compile.
1549 Enable automatic template instantiation at link time. This option also
1550 implies @option{-fno-implicit-templates}. @xref{Template
1551 Instantiation}, for more information.
1555 Disable generation of information about every class with virtual
1556 functions for use by the C++ runtime type identification features
1557 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1558 of the language, you can save some space by using this flag. Note that
1559 exception handling uses the same information, but it will generate it as
1560 needed. The @samp{dynamic_cast} operator can still be used for casts that
1561 do not require runtime type information, i.e. casts to @code{void *} or to
1562 unambiguous base classes.
1566 Emit statistics about front-end processing at the end of the compilation.
1567 This information is generally only useful to the G++ development team.
1569 @item -ftemplate-depth-@var{n}
1570 @opindex ftemplate-depth
1571 Set the maximum instantiation depth for template classes to @var{n}.
1572 A limit on the template instantiation depth is needed to detect
1573 endless recursions during template class instantiation. ANSI/ISO C++
1574 conforming programs must not rely on a maximum depth greater than 17.
1576 @item -fno-threadsafe-statics
1577 @opindex fno-threadsafe-statics
1578 Do not emit the extra code to use the routines specified in the C++
1579 ABI for thread-safe initialization of local statics. You can use this
1580 option to reduce code size slightly in code that doesn't need to be
1583 @item -fuse-cxa-atexit
1584 @opindex fuse-cxa-atexit
1585 Register destructors for objects with static storage duration with the
1586 @code{__cxa_atexit} function rather than the @code{atexit} function.
1587 This option is required for fully standards-compliant handling of static
1588 destructors, but will only work if your C library supports
1589 @code{__cxa_atexit}.
1591 @item -fno-use-cxa-get-exception-ptr
1592 @opindex fno-use-cxa-get-exception-ptr
1593 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1594 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1595 if the runtime routine is not available.
1597 @item -fvisibility-inlines-hidden
1598 @opindex fvisibility-inlines-hidden
1599 This switch declares that the user does not attempt to compare
1600 pointers to inline methods where the addresses of the two functions
1601 were taken in different shared objects.
1603 The effect of this is that GCC may, effectively, mark inline methods with
1604 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1605 appear in the export table of a DSO and do not require a PLT indirection
1606 when used within the DSO@. Enabling this option can have a dramatic effect
1607 on load and link times of a DSO as it massively reduces the size of the
1608 dynamic export table when the library makes heavy use of templates.
1610 The behaviour of this switch is not quite the same as marking the
1611 methods as hidden directly, because it does not affect static variables
1612 local to the function or cause the compiler to deduce that
1613 the function is defined in only one shared object.
1615 You may mark a method as having a visibility explicitly to negate the
1616 effect of the switch for that method. For example, if you do want to
1617 compare pointers to a particular inline method, you might mark it as
1618 having default visibility. Marking the enclosing class with explicit
1619 visibility will have no effect.
1621 Explicitly instantiated inline methods are unaffected by this option
1622 as their linkage might otherwise cross a shared library boundary.
1623 @xref{Template Instantiation}.
1627 Do not use weak symbol support, even if it is provided by the linker.
1628 By default, G++ will use weak symbols if they are available. This
1629 option exists only for testing, and should not be used by end-users;
1630 it will result in inferior code and has no benefits. This option may
1631 be removed in a future release of G++.
1635 Do not search for header files in the standard directories specific to
1636 C++, but do still search the other standard directories. (This option
1637 is used when building the C++ library.)
1640 In addition, these optimization, warning, and code generation options
1641 have meanings only for C++ programs:
1644 @item -fno-default-inline
1645 @opindex fno-default-inline
1646 Do not assume @samp{inline} for functions defined inside a class scope.
1647 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1648 functions will have linkage like inline functions; they just won't be
1651 @item -Wabi @r{(C++ only)}
1653 Warn when G++ generates code that is probably not compatible with the
1654 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1655 all such cases, there are probably some cases that are not warned about,
1656 even though G++ is generating incompatible code. There may also be
1657 cases where warnings are emitted even though the code that is generated
1660 You should rewrite your code to avoid these warnings if you are
1661 concerned about the fact that code generated by G++ may not be binary
1662 compatible with code generated by other compilers.
1664 The known incompatibilities at this point include:
1669 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1670 pack data into the same byte as a base class. For example:
1673 struct A @{ virtual void f(); int f1 : 1; @};
1674 struct B : public A @{ int f2 : 1; @};
1678 In this case, G++ will place @code{B::f2} into the same byte
1679 as@code{A::f1}; other compilers will not. You can avoid this problem
1680 by explicitly padding @code{A} so that its size is a multiple of the
1681 byte size on your platform; that will cause G++ and other compilers to
1682 layout @code{B} identically.
1685 Incorrect handling of tail-padding for virtual bases. G++ does not use
1686 tail padding when laying out virtual bases. For example:
1689 struct A @{ virtual void f(); char c1; @};
1690 struct B @{ B(); char c2; @};
1691 struct C : public A, public virtual B @{@};
1695 In this case, G++ will not place @code{B} into the tail-padding for
1696 @code{A}; other compilers will. You can avoid this problem by
1697 explicitly padding @code{A} so that its size is a multiple of its
1698 alignment (ignoring virtual base classes); that will cause G++ and other
1699 compilers to layout @code{C} identically.
1702 Incorrect handling of bit-fields with declared widths greater than that
1703 of their underlying types, when the bit-fields appear in a union. For
1707 union U @{ int i : 4096; @};
1711 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1712 union too small by the number of bits in an @code{int}.
1715 Empty classes can be placed at incorrect offsets. For example:
1725 struct C : public B, public A @{@};
1729 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1730 it should be placed at offset zero. G++ mistakenly believes that the
1731 @code{A} data member of @code{B} is already at offset zero.
1734 Names of template functions whose types involve @code{typename} or
1735 template template parameters can be mangled incorrectly.
1738 template <typename Q>
1739 void f(typename Q::X) @{@}
1741 template <template <typename> class Q>
1742 void f(typename Q<int>::X) @{@}
1746 Instantiations of these templates may be mangled incorrectly.
1750 @item -Wctor-dtor-privacy @r{(C++ only)}
1751 @opindex Wctor-dtor-privacy
1752 Warn when a class seems unusable because all the constructors or
1753 destructors in that class are private, and it has neither friends nor
1754 public static member functions.
1756 @item -Wnon-virtual-dtor @r{(C++ only)}
1757 @opindex Wnon-virtual-dtor
1758 Warn when a class appears to be polymorphic, thereby requiring a virtual
1759 destructor, yet it declares a non-virtual one. This warning is also
1760 enabled if -Weffc++ is specified.
1762 @item -Wreorder @r{(C++ only)}
1764 @cindex reordering, warning
1765 @cindex warning for reordering of member initializers
1766 Warn when the order of member initializers given in the code does not
1767 match the order in which they must be executed. For instance:
1773 A(): j (0), i (1) @{ @}
1777 The compiler will rearrange the member initializers for @samp{i}
1778 and @samp{j} to match the declaration order of the members, emitting
1779 a warning to that effect. This warning is enabled by @option{-Wall}.
1782 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1785 @item -Weffc++ @r{(C++ only)}
1787 Warn about violations of the following style guidelines from Scott Meyers'
1788 @cite{Effective C++} book:
1792 Item 11: Define a copy constructor and an assignment operator for classes
1793 with dynamically allocated memory.
1796 Item 12: Prefer initialization to assignment in constructors.
1799 Item 14: Make destructors virtual in base classes.
1802 Item 15: Have @code{operator=} return a reference to @code{*this}.
1805 Item 23: Don't try to return a reference when you must return an object.
1809 Also warn about violations of the following style guidelines from
1810 Scott Meyers' @cite{More Effective C++} book:
1814 Item 6: Distinguish between prefix and postfix forms of increment and
1815 decrement operators.
1818 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1822 When selecting this option, be aware that the standard library
1823 headers do not obey all of these guidelines; use @samp{grep -v}
1824 to filter out those warnings.
1826 @item -Wno-deprecated @r{(C++ only)}
1827 @opindex Wno-deprecated
1828 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1830 @item -Wstrict-null-sentinel @r{(C++ only)}
1831 @opindex Wstrict-null-sentinel
1832 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1833 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1834 to @code{__null}. Although it is a null pointer constant not a null pointer,
1835 it is guaranteed to of the same size as a pointer. But this use is
1836 not portable across different compilers.
1838 @item -Wno-non-template-friend @r{(C++ only)}
1839 @opindex Wno-non-template-friend
1840 Disable warnings when non-templatized friend functions are declared
1841 within a template. Since the advent of explicit template specification
1842 support in G++, if the name of the friend is an unqualified-id (i.e.,
1843 @samp{friend foo(int)}), the C++ language specification demands that the
1844 friend declare or define an ordinary, nontemplate function. (Section
1845 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1846 could be interpreted as a particular specialization of a templatized
1847 function. Because this non-conforming behavior is no longer the default
1848 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1849 check existing code for potential trouble spots and is on by default.
1850 This new compiler behavior can be turned off with
1851 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1852 but disables the helpful warning.
1854 @item -Wold-style-cast @r{(C++ only)}
1855 @opindex Wold-style-cast
1856 Warn if an old-style (C-style) cast to a non-void type is used within
1857 a C++ program. The new-style casts (@samp{dynamic_cast},
1858 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
1859 less vulnerable to unintended effects and much easier to search for.
1861 @item -Woverloaded-virtual @r{(C++ only)}
1862 @opindex Woverloaded-virtual
1863 @cindex overloaded virtual fn, warning
1864 @cindex warning for overloaded virtual fn
1865 Warn when a function declaration hides virtual functions from a
1866 base class. For example, in:
1873 struct B: public A @{
1878 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1886 will fail to compile.
1888 @item -Wno-pmf-conversions @r{(C++ only)}
1889 @opindex Wno-pmf-conversions
1890 Disable the diagnostic for converting a bound pointer to member function
1893 @item -Wsign-promo @r{(C++ only)}
1894 @opindex Wsign-promo
1895 Warn when overload resolution chooses a promotion from unsigned or
1896 enumerated type to a signed type, over a conversion to an unsigned type of
1897 the same size. Previous versions of G++ would try to preserve
1898 unsignedness, but the standard mandates the current behavior.
1903 A& operator = (int);
1913 In this example, G++ will synthesize a default @samp{A& operator =
1914 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1917 @node Language Independent Options
1918 @section Options to Control Diagnostic Messages Formatting
1919 @cindex options to control diagnostics formatting
1920 @cindex diagnostic messages
1921 @cindex message formatting
1923 Traditionally, diagnostic messages have been formatted irrespective of
1924 the output device's aspect (e.g.@: its width, @dots{}). The options described
1925 below can be used to control the diagnostic messages formatting
1926 algorithm, e.g.@: how many characters per line, how often source location
1927 information should be reported. Right now, only the C++ front end can
1928 honor these options. However it is expected, in the near future, that
1929 the remaining front ends would be able to digest them correctly.
1932 @item -fmessage-length=@var{n}
1933 @opindex fmessage-length
1934 Try to format error messages so that they fit on lines of about @var{n}
1935 characters. The default is 72 characters for @command{g++} and 0 for the rest of
1936 the front ends supported by GCC@. If @var{n} is zero, then no
1937 line-wrapping will be done; each error message will appear on a single
1940 @opindex fdiagnostics-show-location
1941 @item -fdiagnostics-show-location=once
1942 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
1943 reporter to emit @emph{once} source location information; that is, in
1944 case the message is too long to fit on a single physical line and has to
1945 be wrapped, the source location won't be emitted (as prefix) again,
1946 over and over, in subsequent continuation lines. This is the default
1949 @item -fdiagnostics-show-location=every-line
1950 Only meaningful in line-wrapping mode. Instructs the diagnostic
1951 messages reporter to emit the same source location information (as
1952 prefix) for physical lines that result from the process of breaking
1953 a message which is too long to fit on a single line.
1955 @item -fdiagnostics-show-option
1956 @opindex fdiagnostics-show-option
1957 This option instructs the diagnostic machinery to add text to each
1958 diagnostic emitted, which indicates which command line option directly
1959 controls that diagnostic, when such an option is known to the
1960 diagnostic machinery.
1964 @node Warning Options
1965 @section Options to Request or Suppress Warnings
1966 @cindex options to control warnings
1967 @cindex warning messages
1968 @cindex messages, warning
1969 @cindex suppressing warnings
1971 Warnings are diagnostic messages that report constructions which
1972 are not inherently erroneous but which are risky or suggest there
1973 may have been an error.
1975 You can request many specific warnings with options beginning @samp{-W},
1976 for example @option{-Wimplicit} to request warnings on implicit
1977 declarations. Each of these specific warning options also has a
1978 negative form beginning @samp{-Wno-} to turn off warnings;
1979 for example, @option{-Wno-implicit}. This manual lists only one of the
1980 two forms, whichever is not the default.
1982 The following options control the amount and kinds of warnings produced
1983 by GCC; for further, language-specific options also refer to
1984 @ref{C++ Dialect Options}.
1987 @cindex syntax checking
1989 @opindex fsyntax-only
1990 Check the code for syntax errors, but don't do anything beyond that.
1994 Issue all the warnings demanded by strict ISO C and ISO C++;
1995 reject all programs that use forbidden extensions, and some other
1996 programs that do not follow ISO C and ISO C++. For ISO C, follows the
1997 version of the ISO C standard specified by any @option{-std} option used.
1999 Valid ISO C and ISO C++ programs should compile properly with or without
2000 this option (though a rare few will require @option{-ansi} or a
2001 @option{-std} option specifying the required version of ISO C)@. However,
2002 without this option, certain GNU extensions and traditional C and C++
2003 features are supported as well. With this option, they are rejected.
2005 @option{-pedantic} does not cause warning messages for use of the
2006 alternate keywords whose names begin and end with @samp{__}. Pedantic
2007 warnings are also disabled in the expression that follows
2008 @code{__extension__}. However, only system header files should use
2009 these escape routes; application programs should avoid them.
2010 @xref{Alternate Keywords}.
2012 Some users try to use @option{-pedantic} to check programs for strict ISO
2013 C conformance. They soon find that it does not do quite what they want:
2014 it finds some non-ISO practices, but not all---only those for which
2015 ISO C @emph{requires} a diagnostic, and some others for which
2016 diagnostics have been added.
2018 A feature to report any failure to conform to ISO C might be useful in
2019 some instances, but would require considerable additional work and would
2020 be quite different from @option{-pedantic}. We don't have plans to
2021 support such a feature in the near future.
2023 Where the standard specified with @option{-std} represents a GNU
2024 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2025 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2026 extended dialect is based. Warnings from @option{-pedantic} are given
2027 where they are required by the base standard. (It would not make sense
2028 for such warnings to be given only for features not in the specified GNU
2029 C dialect, since by definition the GNU dialects of C include all
2030 features the compiler supports with the given option, and there would be
2031 nothing to warn about.)
2033 @item -pedantic-errors
2034 @opindex pedantic-errors
2035 Like @option{-pedantic}, except that errors are produced rather than
2040 Inhibit all warning messages.
2044 Inhibit warning messages about the use of @samp{#import}.
2046 @c APPLE LOCAL begin -Wnewline-eof 2001-08-23 --sts **
2048 @opindex Wnewline-eof
2049 Warn about files missing a newline at the end of the file. (Apple compatible)
2050 @c APPLE LOCAL end -Wnewline-eof 2001-08-23 --sts **
2052 @item -Wchar-subscripts
2053 @opindex Wchar-subscripts
2054 Warn if an array subscript has type @code{char}. This is a common cause
2055 of error, as programmers often forget that this type is signed on some
2057 This warning is enabled by @option{-Wall}.
2061 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2062 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2063 This warning is enabled by @option{-Wall}.
2065 @item -Wfatal-errors
2066 @opindex Wfatal-errors
2067 This option causes the compiler to abort compilation on the first error
2068 occurred rather than trying to keep going and printing further error
2073 @opindex ffreestanding
2074 @opindex fno-builtin
2075 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2076 the arguments supplied have types appropriate to the format string
2077 specified, and that the conversions specified in the format string make
2078 sense. This includes standard functions, and others specified by format
2079 attributes (@pxref{Function Attributes}), in the @code{printf},
2080 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2081 not in the C standard) families (or other target-specific families).
2082 Which functions are checked without format attributes having been
2083 specified depends on the standard version selected, and such checks of
2084 functions without the attribute specified are disabled by
2085 @option{-ffreestanding} or @option{-fno-builtin}.
2087 The formats are checked against the format features supported by GNU
2088 libc version 2.2. These include all ISO C90 and C99 features, as well
2089 as features from the Single Unix Specification and some BSD and GNU
2090 extensions. Other library implementations may not support all these
2091 features; GCC does not support warning about features that go beyond a
2092 particular library's limitations. However, if @option{-pedantic} is used
2093 with @option{-Wformat}, warnings will be given about format features not
2094 in the selected standard version (but not for @code{strfmon} formats,
2095 since those are not in any version of the C standard). @xref{C Dialect
2096 Options,,Options Controlling C Dialect}.
2098 Since @option{-Wformat} also checks for null format arguments for
2099 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2101 @option{-Wformat} is included in @option{-Wall}. For more control over some
2102 aspects of format checking, the options @option{-Wformat-y2k},
2103 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2104 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2105 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2108 @opindex Wformat-y2k
2109 If @option{-Wformat} is specified, also warn about @code{strftime}
2110 formats which may yield only a two-digit year.
2112 @item -Wno-format-extra-args
2113 @opindex Wno-format-extra-args
2114 If @option{-Wformat} is specified, do not warn about excess arguments to a
2115 @code{printf} or @code{scanf} format function. The C standard specifies
2116 that such arguments are ignored.
2118 Where the unused arguments lie between used arguments that are
2119 specified with @samp{$} operand number specifications, normally
2120 warnings are still given, since the implementation could not know what
2121 type to pass to @code{va_arg} to skip the unused arguments. However,
2122 in the case of @code{scanf} formats, this option will suppress the
2123 warning if the unused arguments are all pointers, since the Single
2124 Unix Specification says that such unused arguments are allowed.
2126 @item -Wno-format-zero-length
2127 @opindex Wno-format-zero-length
2128 If @option{-Wformat} is specified, do not warn about zero-length formats.
2129 The C standard specifies that zero-length formats are allowed.
2131 @item -Wformat-nonliteral
2132 @opindex Wformat-nonliteral
2133 If @option{-Wformat} is specified, also warn if the format string is not a
2134 string literal and so cannot be checked, unless the format function
2135 takes its format arguments as a @code{va_list}.
2137 @item -Wformat-security
2138 @opindex Wformat-security
2139 If @option{-Wformat} is specified, also warn about uses of format
2140 functions that represent possible security problems. At present, this
2141 warns about calls to @code{printf} and @code{scanf} functions where the
2142 format string is not a string literal and there are no format arguments,
2143 as in @code{printf (foo);}. This may be a security hole if the format
2144 string came from untrusted input and contains @samp{%n}. (This is
2145 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2146 in future warnings may be added to @option{-Wformat-security} that are not
2147 included in @option{-Wformat-nonliteral}.)
2151 Enable @option{-Wformat} plus format checks not included in
2152 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2153 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2157 Warn about passing a null pointer for arguments marked as
2158 requiring a non-null value by the @code{nonnull} function attribute.
2160 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2161 can be disabled with the @option{-Wno-nonnull} option.
2163 @item -Winit-self @r{(C and C++ only)}
2165 Warn about uninitialized variables which are initialized with themselves.
2166 Note this option can only be used with the @option{-Wuninitialized} option,
2167 which in turn only works with @option{-O1} and above.
2169 For example, GCC will warn about @code{i} being uninitialized in the
2170 following snippet only when @option{-Winit-self} has been specified:
2181 @item -Wimplicit-int
2182 @opindex Wimplicit-int
2183 Warn when a declaration does not specify a type.
2184 This warning is enabled by @option{-Wall}.
2186 @item -Wimplicit-function-declaration
2187 @itemx -Werror-implicit-function-declaration
2188 @opindex Wimplicit-function-declaration
2189 @opindex Werror-implicit-function-declaration
2190 Give a warning (or error) whenever a function is used before being
2191 declared. The form @option{-Wno-error-implicit-function-declaration}
2193 This warning is enabled by @option{-Wall} (as a warning, not an error).
2197 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2198 This warning is enabled by @option{-Wall}.
2202 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2203 function with external linkage, returning int, taking either zero
2204 arguments, two, or three arguments of appropriate types.
2205 This warning is enabled by @option{-Wall}.
2207 @item -Wmissing-braces
2208 @opindex Wmissing-braces
2209 Warn if an aggregate or union initializer is not fully bracketed. In
2210 the following example, the initializer for @samp{a} is not fully
2211 bracketed, but that for @samp{b} is fully bracketed.
2214 int a[2][2] = @{ 0, 1, 2, 3 @};
2215 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2218 This warning is enabled by @option{-Wall}.
2220 @item -Wmissing-include-dirs @r{(C and C++ only)}
2221 @opindex Wmissing-include-dirs
2222 Warn if a user-supplied include directory does not exist.
2225 @opindex Wparentheses
2226 Warn if parentheses are omitted in certain contexts, such
2227 as when there is an assignment in a context where a truth value
2228 is expected, or when operators are nested whose precedence people
2229 often get confused about.
2231 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2232 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2233 interpretation from that of ordinary mathematical notation.
2235 Also warn about constructions where there may be confusion to which
2236 @code{if} statement an @code{else} branch belongs. Here is an example of
2251 In C/C++, every @code{else} branch belongs to the innermost possible
2252 @code{if} statement, which in this example is @code{if (b)}. This is
2253 often not what the programmer expected, as illustrated in the above
2254 example by indentation the programmer chose. When there is the
2255 potential for this confusion, GCC will issue a warning when this flag
2256 is specified. To eliminate the warning, add explicit braces around
2257 the innermost @code{if} statement so there is no way the @code{else}
2258 could belong to the enclosing @code{if}. The resulting code would
2275 This warning is enabled by @option{-Wall}.
2277 @item -Wsequence-point
2278 @opindex Wsequence-point
2279 Warn about code that may have undefined semantics because of violations
2280 of sequence point rules in the C and C++ standards.
2282 The C and C++ standards defines the order in which expressions in a C/C++
2283 program are evaluated in terms of @dfn{sequence points}, which represent
2284 a partial ordering between the execution of parts of the program: those
2285 executed before the sequence point, and those executed after it. These
2286 occur after the evaluation of a full expression (one which is not part
2287 of a larger expression), after the evaluation of the first operand of a
2288 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2289 function is called (but after the evaluation of its arguments and the
2290 expression denoting the called function), and in certain other places.
2291 Other than as expressed by the sequence point rules, the order of
2292 evaluation of subexpressions of an expression is not specified. All
2293 these rules describe only a partial order rather than a total order,
2294 since, for example, if two functions are called within one expression
2295 with no sequence point between them, the order in which the functions
2296 are called is not specified. However, the standards committee have
2297 ruled that function calls do not overlap.
2299 It is not specified when between sequence points modifications to the
2300 values of objects take effect. Programs whose behavior depends on this
2301 have undefined behavior; the C and C++ standards specify that ``Between
2302 the previous and next sequence point an object shall have its stored
2303 value modified at most once by the evaluation of an expression.
2304 Furthermore, the prior value shall be read only to determine the value
2305 to be stored.''. If a program breaks these rules, the results on any
2306 particular implementation are entirely unpredictable.
2308 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2309 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2310 diagnosed by this option, and it may give an occasional false positive
2311 result, but in general it has been found fairly effective at detecting
2312 this sort of problem in programs.
2314 The standard is worded confusingly, therefore there is some debate
2315 over the precise meaning of the sequence point rules in subtle cases.
2316 Links to discussions of the problem, including proposed formal
2317 definitions, may be found on the GCC readings page, at
2318 @w{@uref{http://gcc.gnu.org/readings.html}}.
2320 This warning is enabled by @option{-Wall} for C and C++.
2323 @opindex Wreturn-type
2324 Warn whenever a function is defined with a return-type that defaults to
2325 @code{int}. Also warn about any @code{return} statement with no
2326 return-value in a function whose return-type is not @code{void}.
2328 For C, also warn if the return type of a function has a type qualifier
2329 such as @code{const}. Such a type qualifier has no effect, since the
2330 value returned by a function is not an lvalue. ISO C prohibits
2331 qualified @code{void} return types on function definitions, so such
2332 return types always receive a warning even without this option.
2334 For C++, a function without return type always produces a diagnostic
2335 message, even when @option{-Wno-return-type} is specified. The only
2336 exceptions are @samp{main} and functions defined in system headers.
2338 This warning is enabled by @option{-Wall}.
2342 Warn whenever a @code{switch} statement has an index of enumerated type
2343 and lacks a @code{case} for one or more of the named codes of that
2344 enumeration. (The presence of a @code{default} label prevents this
2345 warning.) @code{case} labels outside the enumeration range also
2346 provoke warnings when this option is used.
2347 This warning is enabled by @option{-Wall}.
2349 @item -Wswitch-default
2350 @opindex Wswitch-switch
2351 Warn whenever a @code{switch} statement does not have a @code{default}
2355 @opindex Wswitch-enum
2356 Warn whenever a @code{switch} statement has an index of enumerated type
2357 and lacks a @code{case} for one or more of the named codes of that
2358 enumeration. @code{case} labels outside the enumeration range also
2359 provoke warnings when this option is used.
2363 Warn if any trigraphs are encountered that might change the meaning of
2364 the program (trigraphs within comments are not warned about).
2365 This warning is enabled by @option{-Wall}.
2367 @item -Wunused-function
2368 @opindex Wunused-function
2369 Warn whenever a static function is declared but not defined or a
2370 non-inline static function is unused.
2371 This warning is enabled by @option{-Wall}.
2373 @item -Wunused-label
2374 @opindex Wunused-label
2375 Warn whenever a label is declared but not used.
2376 This warning is enabled by @option{-Wall}.
2378 To suppress this warning use the @samp{unused} attribute
2379 (@pxref{Variable Attributes}).
2381 @item -Wunused-parameter
2382 @opindex Wunused-parameter
2383 Warn whenever a function parameter is unused aside from its declaration.
2385 To suppress this warning use the @samp{unused} attribute
2386 (@pxref{Variable Attributes}).
2388 @item -Wunused-variable
2389 @opindex Wunused-variable
2390 Warn whenever a local variable or non-constant static variable is unused
2391 aside from its declaration.
2392 This warning is enabled by @option{-Wall}.
2394 To suppress this warning use the @samp{unused} attribute
2395 (@pxref{Variable Attributes}).
2397 @item -Wunused-value
2398 @opindex Wunused-value
2399 Warn whenever a statement computes a result that is explicitly not used.
2400 This warning is enabled by @option{-Wall}.
2402 To suppress this warning cast the expression to @samp{void}.
2406 All the above @option{-Wunused} options combined.
2408 In order to get a warning about an unused function parameter, you must
2409 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2410 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2412 @item -Wuninitialized
2413 @opindex Wuninitialized
2414 Warn if an automatic variable is used without first being initialized or
2415 if a variable may be clobbered by a @code{setjmp} call.
2417 These warnings are possible only in optimizing compilation,
2418 because they require data flow information that is computed only
2419 when optimizing. If you do not specify @option{-O}, you will not get
2420 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
2421 requiring @option{-O}.
2423 If you want to warn about code which uses the uninitialized value of the
2424 variable in its own initializer, use the @option{-Winit-self} option.
2426 These warnings occur for individual uninitialized or clobbered
2427 elements of structure, union or array variables as well as for
2428 variables which are uninitialized or clobbered as a whole. They do
2429 not occur for variables or elements declared @code{volatile}. Because
2430 these warnings depend on optimization, the exact variables or elements
2431 for which there are warnings will depend on the precise optimization
2432 options and version of GCC used.
2434 Note that there may be no warning about a variable that is used only
2435 to compute a value that itself is never used, because such
2436 computations may be deleted by data flow analysis before the warnings
2439 These warnings are made optional because GCC is not smart
2440 enough to see all the reasons why the code might be correct
2441 despite appearing to have an error. Here is one example of how
2462 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2463 always initialized, but GCC doesn't know this. Here is
2464 another common case:
2469 if (change_y) save_y = y, y = new_y;
2471 if (change_y) y = save_y;
2476 This has no bug because @code{save_y} is used only if it is set.
2478 @cindex @code{longjmp} warnings
2479 This option also warns when a non-volatile automatic variable might be
2480 changed by a call to @code{longjmp}. These warnings as well are possible
2481 only in optimizing compilation.
2483 The compiler sees only the calls to @code{setjmp}. It cannot know
2484 where @code{longjmp} will be called; in fact, a signal handler could
2485 call it at any point in the code. As a result, you may get a warning
2486 even when there is in fact no problem because @code{longjmp} cannot
2487 in fact be called at the place which would cause a problem.
2489 Some spurious warnings can be avoided if you declare all the functions
2490 you use that never return as @code{noreturn}. @xref{Function
2493 This warning is enabled by @option{-Wall}.
2495 @item -Wunknown-pragmas
2496 @opindex Wunknown-pragmas
2497 @cindex warning for unknown pragmas
2498 @cindex unknown pragmas, warning
2499 @cindex pragmas, warning of unknown
2500 Warn when a #pragma directive is encountered which is not understood by
2501 GCC@. If this command line option is used, warnings will even be issued
2502 for unknown pragmas in system header files. This is not the case if
2503 the warnings were only enabled by the @option{-Wall} command line option.
2506 @opindex Wno-pragmas
2508 Do not warn about misuses of pragmas, such as incorrect parameters,
2509 invalid syntax, or conflicts between pragmas. See also
2510 @samp{-Wunknown-pragmas}.
2512 @item -Wstrict-aliasing
2513 @opindex Wstrict-aliasing
2514 This option is only active when @option{-fstrict-aliasing} is active.
2515 It warns about code which might break the strict aliasing rules that the
2516 compiler is using for optimization. The warning does not catch all
2517 cases, but does attempt to catch the more common pitfalls. It is
2518 included in @option{-Wall}.
2519 It is equivalent to -Wstrict-aliasing=3
2521 @item -Wstrict-aliasing=n
2522 @opindex Wstrict-aliasing=n
2523 This option is only active when @option{-fstrict-aliasing} is active.
2524 It warns about code which might break the strict aliasing rules that the
2525 compiler is using for optimization.
2526 Higher levels correspond to higher accuracy (fewer false positives).
2527 Higher levels also correspond to more effort, similar to the way -O works.
2528 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
2531 Level 1: Most aggressive, quick, least accurate.
2532 Possibly useful when higher levels
2533 do not warn but -fstrict-aliasing still breaks the code, as it has very few
2534 false negatives. However, it has many false positives.
2535 Warns for all pointer conversions between possibly incompatible types,
2536 even if never dereferenced. Runs in the frontend only.
2538 Level 2: Aggressive, quick, not too precise.
2539 May still have many false positives (not as many as level 1 though),
2540 and few false negatives (but possibly more than level 1).
2541 Unlike level 1, it only warns when an address is taken. Warns about
2542 incomplete types. Runs in the frontend only.
2544 Level 3 (default for @option{-Wstrict-aliasing}):
2545 Should have very few false positives and few false
2546 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
2547 Takes care of the common punn+dereference pattern in the frontend:
2548 @code{*(int*)&some_float}.
2549 If optimization is enabled, it also runs in the backend, where it deals
2550 with multiple statement cases using flow-sensitive points-to information.
2551 Only warns when the converted pointer is dereferenced.
2552 Does not warn about incomplete types.
2554 @item -Wstrict-overflow
2555 @item -Wstrict-overflow=@var{n}
2556 @opindex Wstrict-overflow
2557 This option is only active when @option{-fstrict-overflow} is active.
2558 It warns about cases where the compiler optimizes based on the
2559 assumption that signed overflow does not occur. Note that it does not
2560 warn about all cases where the code might overflow: it only warns
2561 about cases where the compiler implements some optimization. Thus
2562 this warning depends on the optimization level.
2564 An optimization which assumes that signed overflow does not occur is
2565 perfectly safe if the values of the variables involved are such that
2566 overflow never does, in fact, occur. Therefore this warning can
2567 easily give a false positive: a warning about code which is not
2568 actually a problem. To help focus on important issues, several
2569 warning levels are defined. No warnings are issued for the use of
2570 undefined signed overflow when estimating how many iterations a loop
2571 will require, in particular when determining whether a loop will be
2575 @item -Wstrict-overflow=1
2576 Warn about cases which are both questionable and easy to avoid. For
2577 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
2578 compiler will simplify this to @code{1}. This level of
2579 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
2580 are not, and must be explicitly requested.
2582 @item -Wstrict-overflow=2
2583 Also warn about other cases where a comparison is simplified to a
2584 constant. For example: @code{abs (x) >= 0}. This can only be
2585 simplified when @option{-fstrict-overflow} is in effect, because
2586 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
2587 zero. @option{-Wstrict-overflow} (with no level) is the same as
2588 @option{-Wstrict-overflow=2}.
2590 @item -Wstrict-overflow=3
2591 Also warn about other cases where a comparison is simplified. For
2592 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
2594 @item -Wstrict-overflow=4
2595 Also warn about other simplifications not covered by the above cases.
2596 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
2598 @item -Wstrict-overflow=5
2599 Also warn about cases where the compiler reduces the magnitude of a
2600 constant involved in a comparison. For example: @code{x + 2 > y} will
2601 be simplified to @code{x + 1 >= y}. This is reported only at the
2602 highest warning level because this simplification applies to many
2603 comparisons, so this warning level will give a very large number of
2609 All of the above @samp{-W} options combined. This enables all the
2610 warnings about constructions that some users consider questionable, and
2611 that are easy to avoid (or modify to prevent the warning), even in
2612 conjunction with macros. This also enables some language-specific
2613 warnings described in @ref{C++ Dialect Options}.
2616 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2617 Some of them warn about constructions that users generally do not
2618 consider questionable, but which occasionally you might wish to check
2619 for; others warn about constructions that are necessary or hard to avoid
2620 in some cases, and there is no simple way to modify the code to suppress
2627 (This option used to be called @option{-W}. The older name is still
2628 supported, but the newer name is more descriptive.) Print extra warning
2629 messages for these events:
2633 A function can return either with or without a value. (Falling
2634 off the end of the function body is considered returning without
2635 a value.) For example, this function would evoke such a
2649 An expression-statement or the left-hand side of a comma expression
2650 contains no side effects.
2651 To suppress the warning, cast the unused expression to void.
2652 For example, an expression such as @samp{x[i,j]} will cause a warning,
2653 but @samp{x[(void)i,j]} will not.
2656 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2659 Storage-class specifiers like @code{static} are not the first things in
2660 a declaration. According to the C Standard, this usage is obsolescent.
2663 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2667 A comparison between signed and unsigned values could produce an
2668 incorrect result when the signed value is converted to unsigned.
2669 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2672 An aggregate has an initializer which does not initialize all members.
2673 This warning can be independently controlled by
2674 @option{-Wmissing-field-initializers}.
2677 An initialized field without side effects is overridden when using
2678 designated initializers (@pxref{Designated Inits, , Designated
2679 Initializers}). This warning can be independently controlled by
2680 @option{-Woverride-init}.
2683 A function parameter is declared without a type specifier in K&R-style
2691 An empty body occurs in an @samp{if} or @samp{else} statement.
2694 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2695 @samp{>}, or @samp{>=}.
2698 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2700 @item @r{(C++ only)}
2701 An enumerator and a non-enumerator both appear in a conditional expression.
2703 @item @r{(C++ only)}
2704 A non-static reference or non-static @samp{const} member appears in a
2705 class without constructors.
2707 @item @r{(C++ only)}
2708 Ambiguous virtual bases.
2710 @item @r{(C++ only)}
2711 Subscripting an array which has been declared @samp{register}.
2713 @item @r{(C++ only)}
2714 Taking the address of a variable which has been declared @samp{register}.
2716 @item @r{(C++ only)}
2717 A base class is not initialized in a derived class' copy constructor.
2720 @item -Wno-div-by-zero
2721 @opindex Wno-div-by-zero
2722 @opindex Wdiv-by-zero
2723 Do not warn about compile-time integer division by zero. Floating point
2724 division by zero is not warned about, as it can be a legitimate way of
2725 obtaining infinities and NaNs.
2727 @item -Wsystem-headers
2728 @opindex Wsystem-headers
2729 @cindex warnings from system headers
2730 @cindex system headers, warnings from
2731 Print warning messages for constructs found in system header files.
2732 Warnings from system headers are normally suppressed, on the assumption
2733 that they usually do not indicate real problems and would only make the
2734 compiler output harder to read. Using this command line option tells
2735 GCC to emit warnings from system headers as if they occurred in user
2736 code. However, note that using @option{-Wall} in conjunction with this
2737 option will @emph{not} warn about unknown pragmas in system
2738 headers---for that, @option{-Wunknown-pragmas} must also be used.
2741 @opindex Wfloat-equal
2742 Warn if floating point values are used in equality comparisons.
2744 The idea behind this is that sometimes it is convenient (for the
2745 programmer) to consider floating-point values as approximations to
2746 infinitely precise real numbers. If you are doing this, then you need
2747 to compute (by analyzing the code, or in some other way) the maximum or
2748 likely maximum error that the computation introduces, and allow for it
2749 when performing comparisons (and when producing output, but that's a
2750 different problem). In particular, instead of testing for equality, you
2751 would check to see whether the two values have ranges that overlap; and
2752 this is done with the relational operators, so equality comparisons are
2755 @item -Wtraditional @r{(C only)}
2756 @opindex Wtraditional
2757 Warn about certain constructs that behave differently in traditional and
2758 ISO C@. Also warn about ISO C constructs that have no traditional C
2759 equivalent, and/or problematic constructs which should be avoided.
2763 Macro parameters that appear within string literals in the macro body.
2764 In traditional C macro replacement takes place within string literals,
2765 but does not in ISO C@.
2768 In traditional C, some preprocessor directives did not exist.
2769 Traditional preprocessors would only consider a line to be a directive
2770 if the @samp{#} appeared in column 1 on the line. Therefore
2771 @option{-Wtraditional} warns about directives that traditional C
2772 understands but would ignore because the @samp{#} does not appear as the
2773 first character on the line. It also suggests you hide directives like
2774 @samp{#pragma} not understood by traditional C by indenting them. Some
2775 traditional implementations would not recognize @samp{#elif}, so it
2776 suggests avoiding it altogether.
2779 A function-like macro that appears without arguments.
2782 The unary plus operator.
2785 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2786 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2787 constants.) Note, these suffixes appear in macros defined in the system
2788 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2789 Use of these macros in user code might normally lead to spurious
2790 warnings, however GCC's integrated preprocessor has enough context to
2791 avoid warning in these cases.
2794 A function declared external in one block and then used after the end of
2798 A @code{switch} statement has an operand of type @code{long}.
2801 A non-@code{static} function declaration follows a @code{static} one.
2802 This construct is not accepted by some traditional C compilers.
2805 The ISO type of an integer constant has a different width or
2806 signedness from its traditional type. This warning is only issued if
2807 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2808 typically represent bit patterns, are not warned about.
2811 Usage of ISO string concatenation is detected.
2814 Initialization of automatic aggregates.
2817 Identifier conflicts with labels. Traditional C lacks a separate
2818 namespace for labels.
2821 Initialization of unions. If the initializer is zero, the warning is
2822 omitted. This is done under the assumption that the zero initializer in
2823 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2824 initializer warnings and relies on default initialization to zero in the
2828 Conversions by prototypes between fixed/floating point values and vice
2829 versa. The absence of these prototypes when compiling with traditional
2830 C would cause serious problems. This is a subset of the possible
2831 conversion warnings, for the full set use @option{-Wconversion}.
2834 Use of ISO C style function definitions. This warning intentionally is
2835 @emph{not} issued for prototype declarations or variadic functions
2836 because these ISO C features will appear in your code when using
2837 libiberty's traditional C compatibility macros, @code{PARAMS} and
2838 @code{VPARAMS}. This warning is also bypassed for nested functions
2839 because that feature is already a GCC extension and thus not relevant to
2840 traditional C compatibility.
2843 @item -Wdeclaration-after-statement @r{(C only)}
2844 @opindex Wdeclaration-after-statement
2845 Warn when a declaration is found after a statement in a block. This
2846 construct, known from C++, was introduced with ISO C99 and is by default
2847 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2848 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2852 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2854 @item -Wno-endif-labels
2855 @opindex Wno-endif-labels
2856 @opindex Wendif-labels
2857 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2861 Warn whenever a local variable shadows another local variable, parameter or
2862 global variable or whenever a built-in function is shadowed.
2864 @item -Wlarger-than-@var{len}
2865 @opindex Wlarger-than
2866 Warn whenever an object of larger than @var{len} bytes is defined.
2868 @item -Wframe-larger-than-@var{len}
2869 @opindex Wframe-larger-than
2870 Warn whenever the frame size of a function is larger than @var{len} bytes.
2872 @item -Wunsafe-loop-optimizations
2873 @opindex Wunsafe-loop-optimizations
2874 Warn if the loop cannot be optimized because the compiler could not
2875 assume anything on the bounds of the loop indices. With
2876 @option{-funsafe-loop-optimizations} warn if the compiler made
2879 @item -Wpointer-arith
2880 @opindex Wpointer-arith
2881 Warn about anything that depends on the ``size of'' a function type or
2882 of @code{void}. GNU C assigns these types a size of 1, for
2883 convenience in calculations with @code{void *} pointers and pointers
2886 @item -Wbad-function-cast @r{(C only)}
2887 @opindex Wbad-function-cast
2888 Warn whenever a function call is cast to a non-matching type.
2889 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2892 Warn about ISO C constructs that are outside of the common subset of
2893 ISO C and ISO C++, e.g.@: request for implicit conversion from
2894 @code{void *} to a pointer to non-@code{void} type.
2898 Warn whenever a pointer is cast so as to remove a type qualifier from
2899 the target type. For example, warn if a @code{const char *} is cast
2900 to an ordinary @code{char *}.
2903 @opindex Wcast-align
2904 Warn whenever a pointer is cast such that the required alignment of the
2905 target is increased. For example, warn if a @code{char *} is cast to
2906 an @code{int *} on machines where integers can only be accessed at
2907 two- or four-byte boundaries.
2909 @item -Wwrite-strings
2910 @opindex Wwrite-strings
2911 When compiling C, give string constants the type @code{const
2912 char[@var{length}]} so that
2913 copying the address of one into a non-@code{const} @code{char *}
2914 pointer will get a warning; when compiling C++, warn about the
2915 deprecated conversion from string literals to @code{char *}. This
2916 warning, by default, is enabled for C++ programs.
2917 These warnings will help you find at
2918 compile time code that can try to write into a string constant, but
2919 only if you have been very careful about using @code{const} in
2920 declarations and prototypes. Otherwise, it will just be a nuisance;
2921 this is why we did not make @option{-Wall} request these warnings.
2924 @opindex Wconversion
2925 Warn if a prototype causes a type conversion that is different from what
2926 would happen to the same argument in the absence of a prototype. This
2927 includes conversions of fixed point to floating and vice versa, and
2928 conversions changing the width or signedness of a fixed point argument
2929 except when the same as the default promotion.
2931 Also, warn if a negative integer constant expression is implicitly
2932 converted to an unsigned type. For example, warn about the assignment
2933 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2934 casts like @code{(unsigned) -1}.
2936 @item -Wsign-compare
2937 @opindex Wsign-compare
2938 @cindex warning for comparison of signed and unsigned values
2939 @cindex comparison of signed and unsigned values, warning
2940 @cindex signed and unsigned values, comparison warning
2941 Warn when a comparison between signed and unsigned values could produce
2942 an incorrect result when the signed value is converted to unsigned.
2943 This warning is also enabled by @option{-Wextra}; to get the other warnings
2944 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2948 @opindex Wno-address
2949 Warn about suspicious uses of memory addresses. These include using
2950 the address of a function in a conditional expression, such as
2951 @code{void func(void); if (func)}, and comparisons against the memory
2952 address of a string literal, such as @code{if (x == "abc")}. Such
2953 uses typically indicate a programmer error: the address of a function
2954 always evaluates to true, so their use in a conditional usually
2955 indicate that the programmer forgot the parentheses in a function
2956 call; and comparisons against string literals result in unspecified
2957 behavior and are not portable in C, so they usually indicate that the
2958 programmer intended to use @code{strcmp}. This warning is enabled by
2961 @item -Waggregate-return
2962 @opindex Waggregate-return
2963 Warn if any functions that return structures or unions are defined or
2964 called. (In languages where you can return an array, this also elicits
2967 @item -Wno-attributes
2968 @opindex Wno-attributes
2969 @opindex Wattributes
2970 Do not warn if an unexpected @code{__attribute__} is used, such as
2971 unrecognized attributes, function attributes applied to variables,
2972 etc. This will not stop errors for incorrect use of supported
2975 @item -Wstrict-prototypes @r{(C only)}
2976 @opindex Wstrict-prototypes
2977 Warn if a function is declared or defined without specifying the
2978 argument types. (An old-style function definition is permitted without
2979 a warning if preceded by a declaration which specifies the argument
2982 @item -Wold-style-definition @r{(C only)}
2983 @opindex Wold-style-definition
2984 Warn if an old-style function definition is used. A warning is given
2985 even if there is a previous prototype.
2987 @item -Wmissing-prototypes @r{(C only)}
2988 @opindex Wmissing-prototypes
2989 Warn if a global function is defined without a previous prototype
2990 declaration. This warning is issued even if the definition itself
2991 provides a prototype. The aim is to detect global functions that fail
2992 to be declared in header files.
2994 @item -Wmissing-declarations @r{(C only)}
2995 @opindex Wmissing-declarations
2996 Warn if a global function is defined without a previous declaration.
2997 Do so even if the definition itself provides a prototype.
2998 Use this option to detect global functions that are not declared in
3001 @item -Wmissing-field-initializers
3002 @opindex Wmissing-field-initializers
3005 Warn if a structure's initializer has some fields missing. For
3006 example, the following code would cause such a warning, because
3007 @code{x.h} is implicitly zero:
3010 struct s @{ int f, g, h; @};
3011 struct s x = @{ 3, 4 @};
3014 This option does not warn about designated initializers, so the following
3015 modification would not trigger a warning:
3018 struct s @{ int f, g, h; @};
3019 struct s x = @{ .f = 3, .g = 4 @};
3022 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3023 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3025 @item -Wmissing-noreturn
3026 @opindex Wmissing-noreturn
3027 Warn about functions which might be candidates for attribute @code{noreturn}.
3028 Note these are only possible candidates, not absolute ones. Care should
3029 be taken to manually verify functions actually do not ever return before
3030 adding the @code{noreturn} attribute, otherwise subtle code generation
3031 bugs could be introduced. You will not get a warning for @code{main} in
3032 hosted C environments.
3034 @item -Wmissing-format-attribute
3035 @opindex Wmissing-format-attribute
3037 Warn about function pointers which might be candidates for @code{format}
3038 attributes. Note these are only possible candidates, not absolute ones.
3039 GCC will guess that function pointers with @code{format} attributes that
3040 are used in assignment, initialization, parameter passing or return
3041 statements should have a corresponding @code{format} attribute in the
3042 resulting type. I.e.@: the left-hand side of the assignment or
3043 initialization, the type of the parameter variable, or the return type
3044 of the containing function respectively should also have a @code{format}
3045 attribute to avoid the warning.
3047 GCC will also warn about function definitions which might be
3048 candidates for @code{format} attributes. Again, these are only
3049 possible candidates. GCC will guess that @code{format} attributes
3050 might be appropriate for any function that calls a function like
3051 @code{vprintf} or @code{vscanf}, but this might not always be the
3052 case, and some functions for which @code{format} attributes are
3053 appropriate may not be detected.
3055 @item -Wno-multichar
3056 @opindex Wno-multichar
3058 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3059 Usually they indicate a typo in the user's code, as they have
3060 implementation-defined values, and should not be used in portable code.
3062 @item -Wnormalized=<none|id|nfc|nfkc>
3063 @opindex Wnormalized
3066 @cindex character set, input normalization
3067 In ISO C and ISO C++, two identifiers are different if they are
3068 different sequences of characters. However, sometimes when characters
3069 outside the basic ASCII character set are used, you can have two
3070 different character sequences that look the same. To avoid confusion,
3071 the ISO 10646 standard sets out some @dfn{normalization rules} which
3072 when applied ensure that two sequences that look the same are turned into
3073 the same sequence. GCC can warn you if you are using identifiers which
3074 have not been normalized; this option controls that warning.
3076 There are four levels of warning that GCC supports. The default is
3077 @option{-Wnormalized=nfc}, which warns about any identifier which is
3078 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3079 recommended form for most uses.
3081 Unfortunately, there are some characters which ISO C and ISO C++ allow
3082 in identifiers that when turned into NFC aren't allowable as
3083 identifiers. That is, there's no way to use these symbols in portable
3084 ISO C or C++ and have all your identifiers in NFC.
3085 @option{-Wnormalized=id} suppresses the warning for these characters.
3086 It is hoped that future versions of the standards involved will correct
3087 this, which is why this option is not the default.
3089 You can switch the warning off for all characters by writing
3090 @option{-Wnormalized=none}. You would only want to do this if you
3091 were using some other normalization scheme (like ``D''), because
3092 otherwise you can easily create bugs that are literally impossible to see.
3094 Some characters in ISO 10646 have distinct meanings but look identical
3095 in some fonts or display methodologies, especially once formatting has
3096 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3097 LETTER N'', will display just like a regular @code{n} which has been
3098 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3099 normalization scheme to convert all these into a standard form as
3100 well, and GCC will warn if your code is not in NFKC if you use
3101 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3102 about every identifier that contains the letter O because it might be
3103 confused with the digit 0, and so is not the default, but may be
3104 useful as a local coding convention if the programming environment is
3105 unable to be fixed to display these characters distinctly.
3107 @item -Wno-deprecated-declarations
3108 @opindex Wno-deprecated-declarations
3109 Do not warn about uses of functions (@pxref{Function Attributes}),
3110 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3111 Attributes}) marked as deprecated by using the @code{deprecated}
3115 @opindex Wno-overflow
3116 Do not warn about compile-time overflow in constant expressions.
3118 @item -Woverride-init
3119 @opindex Woverride-init
3122 Warn if an initialized field without side effects is overridden when
3123 using designated initializers (@pxref{Designated Inits, , Designated
3126 This warning is included in @option{-Wextra}. To get other
3127 @option{-Wextra} warnings without this one, use @samp{-Wextra
3128 -Wno-override-init}.
3132 Warn if a structure is given the packed attribute, but the packed
3133 attribute has no effect on the layout or size of the structure.
3134 Such structures may be mis-aligned for little benefit. For
3135 instance, in this code, the variable @code{f.x} in @code{struct bar}
3136 will be misaligned even though @code{struct bar} does not itself
3137 have the packed attribute:
3144 @} __attribute__((packed));
3154 Warn if padding is included in a structure, either to align an element
3155 of the structure or to align the whole structure. Sometimes when this
3156 happens it is possible to rearrange the fields of the structure to
3157 reduce the padding and so make the structure smaller.
3159 @item -Wredundant-decls
3160 @opindex Wredundant-decls
3161 Warn if anything is declared more than once in the same scope, even in
3162 cases where multiple declaration is valid and changes nothing.
3164 @item -Wnested-externs @r{(C only)}
3165 @opindex Wnested-externs
3166 Warn if an @code{extern} declaration is encountered within a function.
3168 @item -Wunreachable-code
3169 @opindex Wunreachable-code
3170 Warn if the compiler detects that code will never be executed.
3172 This option is intended to warn when the compiler detects that at
3173 least a whole line of source code will never be executed, because
3174 some condition is never satisfied or because it is after a
3175 procedure that never returns.
3177 It is possible for this option to produce a warning even though there
3178 are circumstances under which part of the affected line can be executed,
3179 so care should be taken when removing apparently-unreachable code.
3181 For instance, when a function is inlined, a warning may mean that the
3182 line is unreachable in only one inlined copy of the function.
3184 This option is not made part of @option{-Wall} because in a debugging
3185 version of a program there is often substantial code which checks
3186 correct functioning of the program and is, hopefully, unreachable
3187 because the program does work. Another common use of unreachable
3188 code is to provide behavior which is selectable at compile-time.
3192 Warn if a function can not be inlined and it was declared as inline.
3193 Even with this option, the compiler will not warn about failures to
3194 inline functions declared in system headers.
3196 The compiler uses a variety of heuristics to determine whether or not
3197 to inline a function. For example, the compiler takes into account
3198 the size of the function being inlined and the amount of inlining
3199 that has already been done in the current function. Therefore,
3200 seemingly insignificant changes in the source program can cause the
3201 warnings produced by @option{-Winline} to appear or disappear.
3203 @item -Wno-invalid-offsetof @r{(C++ only)}
3204 @opindex Wno-invalid-offsetof
3205 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3206 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3207 to a non-POD type is undefined. In existing C++ implementations,
3208 however, @samp{offsetof} typically gives meaningful results even when
3209 applied to certain kinds of non-POD types. (Such as a simple
3210 @samp{struct} that fails to be a POD type only by virtue of having a
3211 constructor.) This flag is for users who are aware that they are
3212 writing nonportable code and who have deliberately chosen to ignore the
3215 The restrictions on @samp{offsetof} may be relaxed in a future version
3216 of the C++ standard.
3218 @item -Wno-int-to-pointer-cast @r{(C only)}
3219 @opindex Wno-int-to-pointer-cast
3220 Suppress warnings from casts to pointer type of an integer of a
3223 @item -Wno-pointer-to-int-cast @r{(C only)}
3224 @opindex Wno-pointer-to-int-cast
3225 Suppress warnings from casts from a pointer to an integer type of a
3229 @opindex Winvalid-pch
3230 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3231 the search path but can't be used.
3235 @opindex Wno-long-long
3236 Warn if @samp{long long} type is used. This is default. To inhibit
3237 the warning messages, use @option{-Wno-long-long}. Flags
3238 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3239 only when @option{-pedantic} flag is used.
3241 @item -Wvariadic-macros
3242 @opindex Wvariadic-macros
3243 @opindex Wno-variadic-macros
3244 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3245 alternate syntax when in pedantic ISO C99 mode. This is default.
3246 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3251 Warn if variable length array is used in the code.
3252 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
3253 the variable length array.
3255 @item -Wvolatile-register-var
3256 @opindex Wvolatile-register-var
3257 @opindex Wno-volatile-register-var
3258 Warn if a register variable is declared volatile. The volatile
3259 modifier does not inhibit all optimizations that may eliminate reads
3260 and/or writes to register variables.
3262 @item -Wdisabled-optimization
3263 @opindex Wdisabled-optimization
3264 Warn if a requested optimization pass is disabled. This warning does
3265 not generally indicate that there is anything wrong with your code; it
3266 merely indicates that GCC's optimizers were unable to handle the code
3267 effectively. Often, the problem is that your code is too big or too
3268 complex; GCC will refuse to optimize programs when the optimization
3269 itself is likely to take inordinate amounts of time.
3271 @item -Wpointer-sign
3272 @opindex Wpointer-sign
3273 @opindex Wno-pointer-sign
3274 Warn for pointer argument passing or assignment with different signedness.
3275 This option is only supported for C. It is implied by @option{-Wall}
3276 and by @option{-pedantic}, which can be disabled with
3277 @option{-Wno-pointer-sign}.
3281 Make all warnings into errors.
3285 Make the specified warning into an errors. The specifier for a
3286 warning is appended, for example @option{-Werror=switch} turns the
3287 warnings controlled by @option{-Wswitch} into errors. This switch
3288 takes a negative form, to be used to negate @option{-Werror} for
3289 specific warnings, for example @option{-Wno-error=switch} makes
3290 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3291 is in effect. You can use the @option{-fdiagnostics-show-option}
3292 option to have each controllable warning amended with the option which
3293 controls it, to determine what to use with this option.
3295 Note that specifying @option{-Werror=}@var{foo} automatically implies
3296 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3299 @item -Wstack-protector
3300 @opindex Wstack-protector
3301 This option is only active when @option{-fstack-protector} is active. It
3302 warns about functions that will not be protected against stack smashing.
3304 @item -Woverlength-strings
3305 @opindex Woverlength-strings
3306 Warn about string constants which are longer than the ``minimum
3307 maximum'' length specified in the C standard. Modern compilers
3308 generally allow string constants which are much longer than the
3309 standard's minimum limit, but very portable programs should avoid
3310 using longer strings.
3312 The limit applies @emph{after} string constant concatenation, and does
3313 not count the trailing NUL@. In C89, the limit was 509 characters; in
3314 C99, it was raised to 4095. C++98 does not specify a normative
3315 minimum maximum, so we do not diagnose overlength strings in C++@.
3317 This option is implied by @option{-pedantic}, and can be disabled with
3318 @option{-Wno-overlength-strings}.
3321 @node Debugging Options
3322 @section Options for Debugging Your Program or GCC
3323 @cindex options, debugging
3324 @cindex debugging information options
3326 GCC has various special options that are used for debugging
3327 either your program or GCC:
3332 Produce debugging information in the operating system's native format
3333 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3336 On most systems that use stabs format, @option{-g} enables use of extra
3337 debugging information that only GDB can use; this extra information
3338 makes debugging work better in GDB but will probably make other debuggers
3340 refuse to read the program. If you want to control for certain whether
3341 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3342 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3344 GCC allows you to use @option{-g} with
3345 @option{-O}. The shortcuts taken by optimized code may occasionally
3346 produce surprising results: some variables you declared may not exist
3347 at all; flow of control may briefly move where you did not expect it;
3348 some statements may not be executed because they compute constant
3349 results or their values were already at hand; some statements may
3350 execute in different places because they were moved out of loops.
3352 Nevertheless it proves possible to debug optimized output. This makes
3353 it reasonable to use the optimizer for programs that might have bugs.
3355 The following options are useful when GCC is generated with the
3356 capability for more than one debugging format.
3360 Produce debugging information for use by GDB@. This means to use the
3361 most expressive format available (DWARF 2, stabs, or the native format
3362 if neither of those are supported), including GDB extensions if at all
3367 Produce debugging information in stabs format (if that is supported),
3368 without GDB extensions. This is the format used by DBX on most BSD
3369 systems. On MIPS, Alpha and System V Release 4 systems this option
3370 produces stabs debugging output which is not understood by DBX or SDB@.
3371 On System V Release 4 systems this option requires the GNU assembler.
3373 @item -feliminate-unused-debug-symbols
3374 @opindex feliminate-unused-debug-symbols
3375 Produce debugging information in stabs format (if that is supported),
3376 for only symbols that are actually used.
3378 @item -femit-class-debug-always
3379 Instead of emitting debugging information for a C++ class in only one
3380 object file, emit it in all object files using the class. This option
3381 should be used only with debuggers that are unable to handle the way GCC
3382 normally emits debugging information for classes because using this
3383 option will increase the size of debugging information by as much as a
3388 Produce debugging information in stabs format (if that is supported),
3389 using GNU extensions understood only by the GNU debugger (GDB)@. The
3390 use of these extensions is likely to make other debuggers crash or
3391 refuse to read the program.
3395 Produce debugging information in COFF format (if that is supported).
3396 This is the format used by SDB on most System V systems prior to
3401 Produce debugging information in XCOFF format (if that is supported).
3402 This is the format used by the DBX debugger on IBM RS/6000 systems.
3406 Produce debugging information in XCOFF format (if that is supported),
3407 using GNU extensions understood only by the GNU debugger (GDB)@. The
3408 use of these extensions is likely to make other debuggers crash or
3409 refuse to read the program, and may cause assemblers other than the GNU
3410 assembler (GAS) to fail with an error.
3414 Produce debugging information in DWARF version 2 format (if that is
3415 supported). This is the format used by DBX on IRIX 6. With this
3416 option, GCC uses features of DWARF version 3 when they are useful;
3417 version 3 is upward compatible with version 2, but may still cause
3418 problems for older debuggers.
3422 Produce debugging information in VMS debug format (if that is
3423 supported). This is the format used by DEBUG on VMS systems.
3426 @itemx -ggdb@var{level}
3427 @itemx -gstabs@var{level}
3428 @itemx -gcoff@var{level}
3429 @itemx -gxcoff@var{level}
3430 @itemx -gvms@var{level}
3431 Request debugging information and also use @var{level} to specify how
3432 much information. The default level is 2.
3434 Level 1 produces minimal information, enough for making backtraces in
3435 parts of the program that you don't plan to debug. This includes
3436 descriptions of functions and external variables, but no information
3437 about local variables and no line numbers.
3439 Level 3 includes extra information, such as all the macro definitions
3440 present in the program. Some debuggers support macro expansion when
3441 you use @option{-g3}.
3443 @option{-gdwarf-2} does not accept a concatenated debug level, because
3444 GCC used to support an option @option{-gdwarf} that meant to generate
3445 debug information in version 1 of the DWARF format (which is very
3446 different from version 2), and it would have been too confusing. That
3447 debug format is long obsolete, but the option cannot be changed now.
3448 Instead use an additional @option{-g@var{level}} option to change the
3449 debug level for DWARF2.
3451 @item -feliminate-dwarf2-dups
3452 @opindex feliminate-dwarf2-dups
3453 Compress DWARF2 debugging information by eliminating duplicated
3454 information about each symbol. This option only makes sense when
3455 generating DWARF2 debugging information with @option{-gdwarf-2}.
3457 @item -femit-struct-debug-baseonly
3458 Emit debug information for struct-like types
3459 only when the base name of the compilation source file
3460 matches the base name of file in which the struct was defined.
3462 This option substantially reduces the size of debugging information,
3463 but at significant potential loss in type information to the debugger.
3464 See @option{-femit-struct-debug-reduced} for a less aggressive option.
3465 See @option{-femit-struct-debug-detailed} for more detailed control.
3467 This option works only with DWARF 2.
3469 @item -femit-struct-debug-reduced
3470 Emit debug information for struct-like types
3471 only when the base name of the compilation source file
3472 matches the base name of file in which the type was defined,
3473 unless the struct is a template or defined in a system header.
3475 This option significantly reduces the size of debugging information,
3476 with some potential loss in type information to the debugger.
3477 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
3478 See @option{-femit-struct-debug-detailed} for more detailed control.
3480 This option works only with DWARF 2.
3482 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
3483 Specify the struct-like types
3484 for which the compiler will generate debug information.
3485 The intent is to reduce duplicate struct debug information
3486 between different object files within the same program.
3488 This option is a detailed version of
3489 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
3490 which will serve for most needs.
3492 A specification has the syntax
3493 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
3495 The optional first word limits the specification to
3496 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
3497 A struct type is used directly when it is the type of a variable, member.
3498 Indirect uses arise through pointers to structs.
3499 That is, when use of an incomplete struct would be legal, the use is indirect.
3501 @samp{struct one direct; struct two * indirect;}.
3503 The optional second word limits the specification to
3504 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
3505 Generic structs are a bit complicated to explain.
3506 For C++, these are non-explicit specializations of template classes,
3507 or non-template classes within the above.
3508 Other programming languages have generics,
3509 but @samp{-femit-struct-debug-detailed} does not yet implement them.
3511 The third word specifies the source files for those
3512 structs for which the compiler will emit debug information.
3513 The values @samp{none} and @samp{any} have the normal meaning.
3514 The value @samp{base} means that
3515 the base of name of the file in which the type declaration appears
3516 must match the base of the name of the main compilation file.
3517 In practice, this means that
3518 types declared in @file{foo.c} and @file{foo.h} will have debug information,
3519 but types declared in other header will not.
3520 The value @samp{sys} means those types satisfying @samp{base}
3521 or declared in system or compiler headers.
3523 You may need to experiment to determine the best settings for your application.
3525 The default is @samp{-femit-struct-debug-detailed=all}.
3527 This option works only with DWARF 2.
3529 @cindex @command{prof}
3532 Generate extra code to write profile information suitable for the
3533 analysis program @command{prof}. You must use this option when compiling
3534 the source files you want data about, and you must also use it when
3537 @cindex @command{gprof}
3540 Generate extra code to write profile information suitable for the
3541 analysis program @command{gprof}. You must use this option when compiling
3542 the source files you want data about, and you must also use it when
3547 Makes the compiler print out each function name as it is compiled, and
3548 print some statistics about each pass when it finishes.
3551 @opindex ftime-report
3552 Makes the compiler print some statistics about the time consumed by each
3553 pass when it finishes.
3556 @opindex fmem-report
3557 Makes the compiler print some statistics about permanent memory
3558 allocation when it finishes.
3560 @item -fprofile-arcs
3561 @opindex fprofile-arcs
3562 Add code so that program flow @dfn{arcs} are instrumented. During
3563 execution the program records how many times each branch and call is
3564 executed and how many times it is taken or returns. When the compiled
3565 program exits it saves this data to a file called
3566 @file{@var{auxname}.gcda} for each source file. The data may be used for
3567 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3568 test coverage analysis (@option{-ftest-coverage}). Each object file's
3569 @var{auxname} is generated from the name of the output file, if
3570 explicitly specified and it is not the final executable, otherwise it is
3571 the basename of the source file. In both cases any suffix is removed
3572 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3573 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3574 @xref{Cross-profiling}.
3576 @cindex @command{gcov}
3580 This option is used to compile and link code instrumented for coverage
3581 analysis. The option is a synonym for @option{-fprofile-arcs}
3582 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3583 linking). See the documentation for those options for more details.
3588 Compile the source files with @option{-fprofile-arcs} plus optimization
3589 and code generation options. For test coverage analysis, use the
3590 additional @option{-ftest-coverage} option. You do not need to profile
3591 every source file in a program.
3594 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3595 (the latter implies the former).
3598 Run the program on a representative workload to generate the arc profile
3599 information. This may be repeated any number of times. You can run
3600 concurrent instances of your program, and provided that the file system
3601 supports locking, the data files will be correctly updated. Also
3602 @code{fork} calls are detected and correctly handled (double counting
3606 For profile-directed optimizations, compile the source files again with
3607 the same optimization and code generation options plus
3608 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3609 Control Optimization}).
3612 For test coverage analysis, use @command{gcov} to produce human readable
3613 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3614 @command{gcov} documentation for further information.
3618 With @option{-fprofile-arcs}, for each function of your program GCC
3619 creates a program flow graph, then finds a spanning tree for the graph.
3620 Only arcs that are not on the spanning tree have to be instrumented: the
3621 compiler adds code to count the number of times that these arcs are
3622 executed. When an arc is the only exit or only entrance to a block, the
3623 instrumentation code can be added to the block; otherwise, a new basic
3624 block must be created to hold the instrumentation code.
3627 @item -ftest-coverage
3628 @opindex ftest-coverage
3629 Produce a notes file that the @command{gcov} code-coverage utility
3630 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3631 show program coverage. Each source file's note file is called
3632 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3633 above for a description of @var{auxname} and instructions on how to
3634 generate test coverage data. Coverage data will match the source files
3635 more closely, if you do not optimize.
3637 @item -d@var{letters}
3638 @item -fdump-rtl-@var{pass}
3640 Says to make debugging dumps during compilation at times specified by
3641 @var{letters}. This is used for debugging the RTL-based passes of the
3642 compiler. The file names for most of the dumps are made by appending a
3643 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3644 from the name of the output file, if explicitly specified and it is not
3645 an executable, otherwise it is the basename of the source file.
3647 Most debug dumps can be enabled either passing a letter to the @option{-d}
3648 option, or with a long @option{-fdump-rtl} switch; here are the possible
3649 letters for use in @var{letters} and @var{pass}, and their meanings:
3654 Annotate the assembler output with miscellaneous debugging information.
3657 @itemx -fdump-rtl-bbro
3659 @opindex fdump-rtl-bbro
3660 Dump after block reordering, to @file{@var{file}.148r.bbro}.
3663 @itemx -fdump-rtl-combine
3665 @opindex fdump-rtl-combine
3666 Dump after instruction combination, to the file @file{@var{file}.129r.combine}.
3669 @itemx -fdump-rtl-ce1
3670 @itemx -fdump-rtl-ce2
3672 @opindex fdump-rtl-ce1
3673 @opindex fdump-rtl-ce2
3674 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3675 first if conversion, to the file @file{@var{file}.117r.ce1}. @option{-dC}
3676 and @option{-fdump-rtl-ce2} enable dumping after the second if
3677 conversion, to the file @file{@var{file}.130r.ce2}.
3680 @itemx -fdump-rtl-btl
3681 @itemx -fdump-rtl-dbr
3683 @opindex fdump-rtl-btl
3684 @opindex fdump-rtl-dbr
3685 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3686 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3687 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3688 scheduling, to @file{@var{file}.36.dbr}.
3692 Dump all macro definitions, at the end of preprocessing, in addition to
3696 @itemx -fdump-rtl-ce3
3698 @opindex fdump-rtl-ce3
3699 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
3702 @itemx -fdump-rtl-cfg
3703 @itemx -fdump-rtl-life
3705 @opindex fdump-rtl-cfg
3706 @opindex fdump-rtl-life
3707 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3708 and data flow analysis, to @file{@var{file}.116r.cfg}. @option{-df}
3709 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3710 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
3713 @itemx -fdump-rtl-greg
3715 @opindex fdump-rtl-greg
3716 Dump after global register allocation, to @file{@var{file}.139r.greg}.
3719 @itemx -fdump-rtl-gcse
3720 @itemx -fdump-rtl-bypass
3722 @opindex fdump-rtl-gcse
3723 @opindex fdump-rtl-bypass
3724 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3725 @file{@var{file}.114r.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3726 enable dumping after jump bypassing and control flow optimizations, to
3727 @file{@var{file}.115r.bypass}.
3730 @itemx -fdump-rtl-eh
3732 @opindex fdump-rtl-eh
3733 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3736 @itemx -fdump-rtl-sibling
3738 @opindex fdump-rtl-sibling
3739 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
3742 @itemx -fdump-rtl-jump
3744 @opindex fdump-rtl-jump
3745 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
3748 @itemx -fdump-rtl-stack
3750 @opindex fdump-rtl-stack
3751 Dump after conversion from registers to stack, to @file{@var{file}.152r.stack}.
3754 @itemx -fdump-rtl-lreg
3756 @opindex fdump-rtl-lreg
3757 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
3760 @itemx -fdump-rtl-loop2
3762 @opindex fdump-rtl-loop2
3763 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
3764 loop optimization pass, to @file{@var{file}.119r.loop2},
3765 @file{@var{file}.120r.loop2_init},
3766 @file{@var{file}.121r.loop2_invariant}, and
3767 @file{@var{file}.125r.loop2_done}.
3770 @itemx -fdump-rtl-sms
3772 @opindex fdump-rtl-sms
3773 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
3776 @itemx -fdump-rtl-mach
3778 @opindex fdump-rtl-mach
3779 Dump after performing the machine dependent reorganization pass, to
3780 @file{@var{file}.155r.mach}.
3783 @itemx -fdump-rtl-rnreg
3785 @opindex fdump-rtl-rnreg
3786 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
3789 @itemx -fdump-rtl-regmove
3791 @opindex fdump-rtl-regmove
3792 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
3795 @itemx -fdump-rtl-postreload
3797 @opindex fdump-rtl-postreload
3798 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3801 @itemx -fdump-rtl-expand
3803 @opindex fdump-rtl-expand
3804 Dump after RTL generation, to @file{@var{file}.104r.expand}.
3807 @itemx -fdump-rtl-sched2
3809 @opindex fdump-rtl-sched2
3810 Dump after the second scheduling pass, to @file{@var{file}.150r.sched2}.
3813 @itemx -fdump-rtl-cse
3815 @opindex fdump-rtl-cse
3816 Dump after CSE (including the jump optimization that sometimes follows
3817 CSE), to @file{@var{file}.113r.cse}.
3820 @itemx -fdump-rtl-sched
3822 @opindex fdump-rtl-sched
3823 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3826 @itemx -fdump-rtl-cse2
3828 @opindex fdump-rtl-cse2
3829 Dump after the second CSE pass (including the jump optimization that
3830 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
3833 @itemx -fdump-rtl-tracer
3835 @opindex fdump-rtl-tracer
3836 Dump after running tracer, to @file{@var{file}.118r.tracer}.
3839 @itemx -fdump-rtl-vpt
3840 @itemx -fdump-rtl-vartrack
3842 @opindex fdump-rtl-vpt
3843 @opindex fdump-rtl-vartrack
3844 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3845 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3846 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3847 to @file{@var{file}.154r.vartrack}.
3850 @itemx -fdump-rtl-flow2
3852 @opindex fdump-rtl-flow2
3853 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
3856 @itemx -fdump-rtl-peephole2
3858 @opindex fdump-rtl-peephole2
3859 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
3862 @itemx -fdump-rtl-web
3864 @opindex fdump-rtl-web
3865 Dump after live range splitting, to @file{@var{file}.126r.web}.
3868 @itemx -fdump-rtl-all
3870 @opindex fdump-rtl-all
3871 Produce all the dumps listed above.
3875 Produce a core dump whenever an error occurs.
3879 Print statistics on memory usage, at the end of the run, to
3884 Annotate the assembler output with a comment indicating which
3885 pattern and alternative was used. The length of each instruction is
3890 Dump the RTL in the assembler output as a comment before each instruction.
3891 Also turns on @option{-dp} annotation.
3895 For each of the other indicated dump files (either with @option{-d} or
3896 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3897 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3901 Just generate RTL for a function instead of compiling it. Usually used
3902 with @samp{r} (@option{-fdump-rtl-expand}).
3906 Dump debugging information during parsing, to standard error.
3910 @opindex fdump-noaddr
3911 When doing debugging dumps (see @option{-d} option above), suppress
3912 address output. This makes it more feasible to use diff on debugging
3913 dumps for compiler invocations with different compiler binaries and/or
3914 different text / bss / data / heap / stack / dso start locations.
3916 @item -fdump-unnumbered
3917 @opindex fdump-unnumbered
3918 When doing debugging dumps (see @option{-d} option above), suppress instruction
3919 numbers, line number note and address output. This makes it more feasible to
3920 use diff on debugging dumps for compiler invocations with different
3921 options, in particular with and without @option{-g}.
3923 @item -fdump-translation-unit @r{(C++ only)}
3924 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
3925 @opindex fdump-translation-unit
3926 Dump a representation of the tree structure for the entire translation
3927 unit to a file. The file name is made by appending @file{.tu} to the
3928 source file name. If the @samp{-@var{options}} form is used, @var{options}
3929 controls the details of the dump as described for the
3930 @option{-fdump-tree} options.
3932 @item -fdump-class-hierarchy @r{(C++ only)}
3933 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3934 @opindex fdump-class-hierarchy
3935 Dump a representation of each class's hierarchy and virtual function
3936 table layout to a file. The file name is made by appending @file{.class}
3937 to the source file name. If the @samp{-@var{options}} form is used,
3938 @var{options} controls the details of the dump as described for the
3939 @option{-fdump-tree} options.
3941 @item -fdump-ipa-@var{switch}
3943 Control the dumping at various stages of inter-procedural analysis
3944 language tree to a file. The file name is generated by appending a switch
3945 specific suffix to the source file name. The following dumps are possible:
3949 Enables all inter-procedural analysis dumps; currently the only produced
3950 dump is the @samp{cgraph} dump.
3953 Dumps information about call-graph optimization, unused function removal,
3954 and inlining decisions.
3957 @item -fdump-tree-@var{switch}
3958 @itemx -fdump-tree-@var{switch}-@var{options}
3960 Control the dumping at various stages of processing the intermediate
3961 language tree to a file. The file name is generated by appending a switch
3962 specific suffix to the source file name. If the @samp{-@var{options}}
3963 form is used, @var{options} is a list of @samp{-} separated options that
3964 control the details of the dump. Not all options are applicable to all
3965 dumps, those which are not meaningful will be ignored. The following
3966 options are available
3970 Print the address of each node. Usually this is not meaningful as it
3971 changes according to the environment and source file. Its primary use
3972 is for tying up a dump file with a debug environment.
3974 Inhibit dumping of members of a scope or body of a function merely
3975 because that scope has been reached. Only dump such items when they
3976 are directly reachable by some other path. When dumping pretty-printed
3977 trees, this option inhibits dumping the bodies of control structures.
3979 Print a raw representation of the tree. By default, trees are
3980 pretty-printed into a C-like representation.
3982 Enable more detailed dumps (not honored by every dump option).
3984 Enable dumping various statistics about the pass (not honored by every dump
3987 Enable showing basic block boundaries (disabled in raw dumps).
3989 Enable showing virtual operands for every statement.
3991 Enable showing line numbers for statements.
3993 Enable showing the unique ID (@code{DECL_UID}) for each variable.
3995 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
3998 The following tree dumps are possible:
4002 Dump before any tree based optimization, to @file{@var{file}.original}.
4005 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4008 Dump after function inlining, to @file{@var{file}.inlined}.
4011 @opindex fdump-tree-gimple
4012 Dump each function before and after the gimplification pass to a file. The
4013 file name is made by appending @file{.gimple} to the source file name.
4016 @opindex fdump-tree-cfg
4017 Dump the control flow graph of each function to a file. The file name is
4018 made by appending @file{.cfg} to the source file name.
4021 @opindex fdump-tree-vcg
4022 Dump the control flow graph of each function to a file in VCG format. The
4023 file name is made by appending @file{.vcg} to the source file name. Note
4024 that if the file contains more than one function, the generated file cannot
4025 be used directly by VCG@. You will need to cut and paste each function's
4026 graph into its own separate file first.
4029 @opindex fdump-tree-ch
4030 Dump each function after copying loop headers. The file name is made by
4031 appending @file{.ch} to the source file name.
4034 @opindex fdump-tree-ssa
4035 Dump SSA related information to a file. The file name is made by appending
4036 @file{.ssa} to the source file name.
4039 @opindex fdump-tree-salias
4040 Dump structure aliasing variable information to a file. This file name
4041 is made by appending @file{.salias} to the source file name.
4044 @opindex fdump-tree-alias
4045 Dump aliasing information for each function. The file name is made by
4046 appending @file{.alias} to the source file name.
4049 @opindex fdump-tree-ccp
4050 Dump each function after CCP@. The file name is made by appending
4051 @file{.ccp} to the source file name.
4054 @opindex fdump-tree-storeccp
4055 Dump each function after STORE-CCP. The file name is made by appending
4056 @file{.storeccp} to the source file name.
4059 @opindex fdump-tree-pre
4060 Dump trees after partial redundancy elimination. The file name is made
4061 by appending @file{.pre} to the source file name.
4064 @opindex fdump-tree-fre
4065 Dump trees after full redundancy elimination. The file name is made
4066 by appending @file{.fre} to the source file name.
4069 @opindex fdump-tree-copyprop
4070 Dump trees after copy propagation. The file name is made
4071 by appending @file{.copyprop} to the source file name.
4073 @item store_copyprop
4074 @opindex fdump-tree-store_copyprop
4075 Dump trees after store copy-propagation. The file name is made
4076 by appending @file{.store_copyprop} to the source file name.
4079 @opindex fdump-tree-dce
4080 Dump each function after dead code elimination. The file name is made by
4081 appending @file{.dce} to the source file name.
4084 @opindex fdump-tree-mudflap
4085 Dump each function after adding mudflap instrumentation. The file name is
4086 made by appending @file{.mudflap} to the source file name.
4089 @opindex fdump-tree-sra
4090 Dump each function after performing scalar replacement of aggregates. The
4091 file name is made by appending @file{.sra} to the source file name.
4094 @opindex fdump-tree-sink
4095 Dump each function after performing code sinking. The file name is made
4096 by appending @file{.sink} to the source file name.
4099 @opindex fdump-tree-dom
4100 Dump each function after applying dominator tree optimizations. The file
4101 name is made by appending @file{.dom} to the source file name.
4104 @opindex fdump-tree-dse
4105 Dump each function after applying dead store elimination. The file
4106 name is made by appending @file{.dse} to the source file name.
4109 @opindex fdump-tree-phiopt
4110 Dump each function after optimizing PHI nodes into straightline code. The file
4111 name is made by appending @file{.phiopt} to the source file name.
4114 @opindex fdump-tree-forwprop
4115 Dump each function after forward propagating single use variables. The file
4116 name is made by appending @file{.forwprop} to the source file name.
4119 @opindex fdump-tree-copyrename
4120 Dump each function after applying the copy rename optimization. The file
4121 name is made by appending @file{.copyrename} to the source file name.
4124 @opindex fdump-tree-nrv
4125 Dump each function after applying the named return value optimization on
4126 generic trees. The file name is made by appending @file{.nrv} to the source
4130 @opindex fdump-tree-vect
4131 Dump each function after applying vectorization of loops. The file name is
4132 made by appending @file{.vect} to the source file name.
4135 @opindex fdump-tree-vrp
4136 Dump each function after Value Range Propagation (VRP). The file name
4137 is made by appending @file{.vrp} to the source file name.
4140 @opindex fdump-tree-all
4141 Enable all the available tree dumps with the flags provided in this option.
4144 @item -ftree-vectorizer-verbose=@var{n}
4145 @opindex ftree-vectorizer-verbose
4146 This option controls the amount of debugging output the vectorizer prints.
4147 This information is written to standard error, unless
4148 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4149 in which case it is output to the usual dump listing file, @file{.vect}.
4150 For @var{n}=0 no diagnostic information is reported.
4151 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4152 and the total number of loops that got vectorized.
4153 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4154 the first analysis phase (vect_analyze_loop_form) - i.e. countable,
4155 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4156 level that @option{-fdump-tree-vect-stats} uses.
4157 Higher verbosity levels mean either more information dumped for each
4158 reported loop, or same amount of information reported for more loops:
4159 If @var{n}=3, alignment related information is added to the reports.
4160 If @var{n}=4, data-references related information (e.g. memory dependences,
4161 memory access-patterns) is added to the reports.
4162 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4163 that did not pass the first analysis phase (i.e. may not be countable, or
4164 may have complicated control-flow).
4165 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4166 For @var{n}=7, all the information the vectorizer generates during its
4167 analysis and transformation is reported. This is the same verbosity level
4168 that @option{-fdump-tree-vect-details} uses.
4170 @item -frandom-seed=@var{string}
4171 @opindex frandom-string
4172 This option provides a seed that GCC uses when it would otherwise use
4173 random numbers. It is used to generate certain symbol names
4174 that have to be different in every compiled file. It is also used to
4175 place unique stamps in coverage data files and the object files that
4176 produce them. You can use the @option{-frandom-seed} option to produce
4177 reproducibly identical object files.
4179 The @var{string} should be different for every file you compile.
4181 @item -fsched-verbose=@var{n}
4182 @opindex fsched-verbose
4183 On targets that use instruction scheduling, this option controls the
4184 amount of debugging output the scheduler prints. This information is
4185 written to standard error, unless @option{-dS} or @option{-dR} is
4186 specified, in which case it is output to the usual dump
4187 listing file, @file{.sched} or @file{.sched2} respectively. However
4188 for @var{n} greater than nine, the output is always printed to standard
4191 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4192 same information as @option{-dRS}. For @var{n} greater than one, it
4193 also output basic block probabilities, detailed ready list information
4194 and unit/insn info. For @var{n} greater than two, it includes RTL
4195 at abort point, control-flow and regions info. And for @var{n} over
4196 four, @option{-fsched-verbose} also includes dependence info.
4200 Store the usual ``temporary'' intermediate files permanently; place them
4201 in the current directory and name them based on the source file. Thus,
4202 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4203 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4204 preprocessed @file{foo.i} output file even though the compiler now
4205 normally uses an integrated preprocessor.
4207 When used in combination with the @option{-x} command line option,
4208 @option{-save-temps} is sensible enough to avoid over writing an
4209 input source file with the same extension as an intermediate file.
4210 The corresponding intermediate file may be obtained by renaming the
4211 source file before using @option{-save-temps}.
4215 Report the CPU time taken by each subprocess in the compilation
4216 sequence. For C source files, this is the compiler proper and assembler
4217 (plus the linker if linking is done). The output looks like this:
4224 The first number on each line is the ``user time'', that is time spent
4225 executing the program itself. The second number is ``system time'',
4226 time spent executing operating system routines on behalf of the program.
4227 Both numbers are in seconds.
4229 @item -fvar-tracking
4230 @opindex fvar-tracking
4231 Run variable tracking pass. It computes where variables are stored at each
4232 position in code. Better debugging information is then generated
4233 (if the debugging information format supports this information).
4235 It is enabled by default when compiling with optimization (@option{-Os},
4236 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4237 the debug info format supports it.
4239 @item -print-file-name=@var{library}
4240 @opindex print-file-name
4241 Print the full absolute name of the library file @var{library} that
4242 would be used when linking---and don't do anything else. With this
4243 option, GCC does not compile or link anything; it just prints the
4246 @item -print-multi-directory
4247 @opindex print-multi-directory
4248 Print the directory name corresponding to the multilib selected by any
4249 other switches present in the command line. This directory is supposed
4250 to exist in @env{GCC_EXEC_PREFIX}.
4252 @item -print-multi-lib
4253 @opindex print-multi-lib
4254 Print the mapping from multilib directory names to compiler switches
4255 that enable them. The directory name is separated from the switches by
4256 @samp{;}, and each switch starts with an @samp{@@} instead of the
4257 @samp{-}, without spaces between multiple switches. This is supposed to
4258 ease shell-processing.
4260 @item -print-prog-name=@var{program}
4261 @opindex print-prog-name
4262 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4264 @item -print-libgcc-file-name
4265 @opindex print-libgcc-file-name
4266 Same as @option{-print-file-name=libgcc.a}.
4268 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4269 but you do want to link with @file{libgcc.a}. You can do
4272 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4275 @item -print-search-dirs
4276 @opindex print-search-dirs
4277 Print the name of the configured installation directory and a list of
4278 program and library directories @command{gcc} will search---and don't do anything else.
4280 This is useful when @command{gcc} prints the error message
4281 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4282 To resolve this you either need to put @file{cpp0} and the other compiler
4283 components where @command{gcc} expects to find them, or you can set the environment
4284 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4285 Don't forget the trailing @samp{/}.
4286 @xref{Environment Variables}.
4289 @opindex dumpmachine
4290 Print the compiler's target machine (for example,
4291 @samp{i686-pc-linux-gnu})---and don't do anything else.
4294 @opindex dumpversion
4295 Print the compiler version (for example, @samp{3.0})---and don't do
4300 Print the compiler's built-in specs---and don't do anything else. (This
4301 is used when GCC itself is being built.) @xref{Spec Files}.
4303 @item -feliminate-unused-debug-types
4304 @opindex feliminate-unused-debug-types
4305 Normally, when producing DWARF2 output, GCC will emit debugging
4306 information for all types declared in a compilation
4307 unit, regardless of whether or not they are actually used
4308 in that compilation unit. Sometimes this is useful, such as
4309 if, in the debugger, you want to cast a value to a type that is
4310 not actually used in your program (but is declared). More often,
4311 however, this results in a significant amount of wasted space.
4312 With this option, GCC will avoid producing debug symbol output
4313 for types that are nowhere used in the source file being compiled.
4316 @node Optimize Options
4317 @section Options That Control Optimization
4318 @cindex optimize options
4319 @cindex options, optimization
4321 These options control various sorts of optimizations.
4323 Without any optimization option, the compiler's goal is to reduce the
4324 cost of compilation and to make debugging produce the expected
4325 results. Statements are independent: if you stop the program with a
4326 breakpoint between statements, you can then assign a new value to any
4327 variable or change the program counter to any other statement in the
4328 function and get exactly the results you would expect from the source
4331 Turning on optimization flags makes the compiler attempt to improve
4332 the performance and/or code size at the expense of compilation time
4333 and possibly the ability to debug the program.
4335 The compiler performs optimization based on the knowledge it has of
4336 the program. Optimization levels @option{-O} and above, in
4337 particular, enable @emph{unit-at-a-time} mode, which allows the
4338 compiler to consider information gained from later functions in
4339 the file when compiling a function. Compiling multiple files at
4340 once to a single output file in @emph{unit-at-a-time} mode allows
4341 the compiler to use information gained from all of the files when
4342 compiling each of them.
4344 Not all optimizations are controlled directly by a flag. Only
4345 optimizations that have a flag are listed.
4352 Optimize. Optimizing compilation takes somewhat more time, and a lot
4353 more memory for a large function.
4355 With @option{-O}, the compiler tries to reduce code size and execution
4356 time, without performing any optimizations that take a great deal of
4359 @option{-O} turns on the following optimization flags:
4360 @gccoptlist{-fdefer-pop @gol
4361 -fdelayed-branch @gol
4362 -fguess-branch-probability @gol
4363 -fcprop-registers @gol
4364 -fif-conversion @gol
4365 -fif-conversion2 @gol
4368 -ftree-dominator-opts @gol
4373 -ftree-copyrename @gol
4376 -funit-at-a-time @gol
4379 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4380 where doing so does not interfere with debugging.
4384 Optimize even more. GCC performs nearly all supported optimizations
4385 that do not involve a space-speed tradeoff. The compiler does not
4386 perform loop unrolling or function inlining when you specify @option{-O2}.
4387 As compared to @option{-O}, this option increases both compilation time
4388 and the performance of the generated code.
4390 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4391 also turns on the following optimization flags:
4392 @gccoptlist{-fthread-jumps @gol
4394 -foptimize-sibling-calls @gol
4395 -fcse-follow-jumps -fcse-skip-blocks @gol
4396 -fgcse -fgcse-lm @gol
4397 -fexpensive-optimizations @gol
4398 -frerun-cse-after-loop @gol
4401 -fschedule-insns -fschedule-insns2 @gol
4402 -fsched-interblock -fsched-spec @gol
4404 -fstrict-aliasing -fstrict-overflow @gol
4405 -fdelete-null-pointer-checks @gol
4406 -freorder-blocks -freorder-functions @gol
4407 -falign-functions -falign-jumps @gol
4408 -falign-loops -falign-labels @gol
4412 Please note the warning under @option{-fgcse} about
4413 invoking @option{-O2} on programs that use computed gotos.
4415 @option{-O2} doesn't turn on @option{-ftree-vrp} for the Ada compiler.
4416 This option must be explicitly specified on the command line to be
4417 enabled for the Ada compiler.
4421 Optimize yet more. @option{-O3} turns on all optimizations specified by
4422 @option{-O2} and also turns on the @option{-finline-functions},
4423 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4427 Do not optimize. This is the default.
4431 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4432 do not typically increase code size. It also performs further
4433 optimizations designed to reduce code size.
4435 @option{-Os} disables the following optimization flags:
4436 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4437 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
4438 -fprefetch-loop-arrays -ftree-vect-loop-version}
4440 If you use multiple @option{-O} options, with or without level numbers,
4441 the last such option is the one that is effective.
4444 Options of the form @option{-f@var{flag}} specify machine-independent
4445 flags. Most flags have both positive and negative forms; the negative
4446 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4447 below, only one of the forms is listed---the one you typically will
4448 use. You can figure out the other form by either removing @samp{no-}
4451 The following options control specific optimizations. They are either
4452 activated by @option{-O} options or are related to ones that are. You
4453 can use the following flags in the rare cases when ``fine-tuning'' of
4454 optimizations to be performed is desired.
4457 @item -fno-default-inline
4458 @opindex fno-default-inline
4459 Do not make member functions inline by default merely because they are
4460 defined inside the class scope (C++ only). Otherwise, when you specify
4461 @w{@option{-O}}, member functions defined inside class scope are compiled
4462 inline by default; i.e., you don't need to add @samp{inline} in front of
4463 the member function name.
4465 @item -fno-defer-pop
4466 @opindex fno-defer-pop
4467 Always pop the arguments to each function call as soon as that function
4468 returns. For machines which must pop arguments after a function call,
4469 the compiler normally lets arguments accumulate on the stack for several
4470 function calls and pops them all at once.
4472 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4476 Force memory operands to be copied into registers before doing
4477 arithmetic on them. This produces better code by making all memory
4478 references potential common subexpressions. When they are not common
4479 subexpressions, instruction combination should eliminate the separate
4480 register-load. This option is now a nop and will be removed in 4.3.
4483 @opindex fforce-addr
4484 Force memory address constants to be copied into registers before
4485 doing arithmetic on them.
4487 @item -fomit-frame-pointer
4488 @opindex fomit-frame-pointer
4489 Don't keep the frame pointer in a register for functions that
4490 don't need one. This avoids the instructions to save, set up and
4491 restore frame pointers; it also makes an extra register available
4492 in many functions. @strong{It also makes debugging impossible on
4495 On some machines, such as the VAX, this flag has no effect, because
4496 the standard calling sequence automatically handles the frame pointer
4497 and nothing is saved by pretending it doesn't exist. The
4498 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4499 whether a target machine supports this flag. @xref{Registers,,Register
4500 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4502 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4504 @item -foptimize-sibling-calls
4505 @opindex foptimize-sibling-calls
4506 Optimize sibling and tail recursive calls.
4508 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4512 Don't pay attention to the @code{inline} keyword. Normally this option
4513 is used to keep the compiler from expanding any functions inline.
4514 Note that if you are not optimizing, no functions can be expanded inline.
4516 @item -finline-functions
4517 @opindex finline-functions
4518 Integrate all simple functions into their callers. The compiler
4519 heuristically decides which functions are simple enough to be worth
4520 integrating in this way.
4522 If all calls to a given function are integrated, and the function is
4523 declared @code{static}, then the function is normally not output as
4524 assembler code in its own right.
4526 Enabled at level @option{-O3}.
4528 @item -finline-functions-called-once
4529 @opindex finline-functions-called-once
4530 Consider all @code{static} functions called once for inlining into their
4531 caller even if they are not marked @code{inline}. If a call to a given
4532 function is integrated, then the function is not output as assembler code
4535 Enabled if @option{-funit-at-a-time} is enabled.
4537 @item -fearly-inlining
4538 @opindex fearly-inlining
4539 Inline functions marked by @code{always_inline} and functions whose body seems
4540 smaller than the function call overhead early before doing
4541 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
4542 makes profiling significantly cheaper and usually inlining faster on programs
4543 having large chains of nested wrapper functions.
4547 @item -finline-limit=@var{n}
4548 @opindex finline-limit
4549 By default, GCC limits the size of functions that can be inlined. This flag
4550 allows the control of this limit for functions that are explicitly marked as
4551 inline (i.e., marked with the inline keyword or defined within the class
4552 definition in c++). @var{n} is the size of functions that can be inlined in
4553 number of pseudo instructions (not counting parameter handling). The default
4554 value of @var{n} is 600.
4555 Increasing this value can result in more inlined code at
4556 the cost of compilation time and memory consumption. Decreasing usually makes
4557 the compilation faster and less code will be inlined (which presumably
4558 means slower programs). This option is particularly useful for programs that
4559 use inlining heavily such as those based on recursive templates with C++.
4561 Inlining is actually controlled by a number of parameters, which may be
4562 specified individually by using @option{--param @var{name}=@var{value}}.
4563 The @option{-finline-limit=@var{n}} option sets some of these parameters
4567 @item max-inline-insns-single
4568 is set to @var{n}/2.
4569 @item max-inline-insns-auto
4570 is set to @var{n}/2.
4571 @item min-inline-insns
4572 is set to 130 or @var{n}/4, whichever is smaller.
4573 @item max-inline-insns-rtl
4577 See below for a documentation of the individual
4578 parameters controlling inlining.
4580 @emph{Note:} pseudo instruction represents, in this particular context, an
4581 abstract measurement of function's size. In no way does it represent a count
4582 of assembly instructions and as such its exact meaning might change from one
4583 release to an another.
4585 @item -fkeep-inline-functions
4586 @opindex fkeep-inline-functions
4587 In C, emit @code{static} functions that are declared @code{inline}
4588 into the object file, even if the function has been inlined into all
4589 of its callers. This switch does not affect functions using the
4590 @code{extern inline} extension in GNU C@. In C++, emit any and all
4591 inline functions into the object file.
4593 @item -fkeep-static-consts
4594 @opindex fkeep-static-consts
4595 Emit variables declared @code{static const} when optimization isn't turned
4596 on, even if the variables aren't referenced.
4598 GCC enables this option by default. If you want to force the compiler to
4599 check if the variable was referenced, regardless of whether or not
4600 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4602 @item -fmerge-constants
4603 Attempt to merge identical constants (string constants and floating point
4604 constants) across compilation units.
4606 This option is the default for optimized compilation if the assembler and
4607 linker support it. Use @option{-fno-merge-constants} to inhibit this
4610 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4612 @item -fmerge-all-constants
4613 Attempt to merge identical constants and identical variables.
4615 This option implies @option{-fmerge-constants}. In addition to
4616 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4617 arrays or initialized constant variables with integral or floating point
4618 types. Languages like C or C++ require each non-automatic variable to
4619 have distinct location, so using this option will result in non-conforming
4622 @item -fmodulo-sched
4623 @opindex fmodulo-sched
4624 Perform swing modulo scheduling immediately before the first scheduling
4625 pass. This pass looks at innermost loops and reorders their
4626 instructions by overlapping different iterations.
4628 @item -fno-branch-count-reg
4629 @opindex fno-branch-count-reg
4630 Do not use ``decrement and branch'' instructions on a count register,
4631 but instead generate a sequence of instructions that decrement a
4632 register, compare it against zero, then branch based upon the result.
4633 This option is only meaningful on architectures that support such
4634 instructions, which include x86, PowerPC, IA-64 and S/390.
4636 The default is @option{-fbranch-count-reg}.
4638 @item -fno-function-cse
4639 @opindex fno-function-cse
4640 Do not put function addresses in registers; make each instruction that
4641 calls a constant function contain the function's address explicitly.
4643 This option results in less efficient code, but some strange hacks
4644 that alter the assembler output may be confused by the optimizations
4645 performed when this option is not used.
4647 The default is @option{-ffunction-cse}
4649 @item -fno-zero-initialized-in-bss
4650 @opindex fno-zero-initialized-in-bss
4651 If the target supports a BSS section, GCC by default puts variables that
4652 are initialized to zero into BSS@. This can save space in the resulting
4655 This option turns off this behavior because some programs explicitly
4656 rely on variables going to the data section. E.g., so that the
4657 resulting executable can find the beginning of that section and/or make
4658 assumptions based on that.
4660 The default is @option{-fzero-initialized-in-bss}.
4662 @item -fbounds-check
4663 @opindex fbounds-check
4664 For front-ends that support it, generate additional code to check that
4665 indices used to access arrays are within the declared range. This is
4666 currently only supported by the Java and Fortran front-ends, where
4667 this option defaults to true and false respectively.
4669 @item -fmudflap -fmudflapth -fmudflapir
4673 @cindex bounds checking
4675 For front-ends that support it (C and C++), instrument all risky
4676 pointer/array dereferencing operations, some standard library
4677 string/heap functions, and some other associated constructs with
4678 range/validity tests. Modules so instrumented should be immune to
4679 buffer overflows, invalid heap use, and some other classes of C/C++
4680 programming errors. The instrumentation relies on a separate runtime
4681 library (@file{libmudflap}), which will be linked into a program if
4682 @option{-fmudflap} is given at link time. Run-time behavior of the
4683 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4684 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4687 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4688 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4689 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4690 instrumentation should ignore pointer reads. This produces less
4691 instrumentation (and therefore faster execution) and still provides
4692 some protection against outright memory corrupting writes, but allows
4693 erroneously read data to propagate within a program.
4695 @item -fthread-jumps
4696 @opindex fthread-jumps
4697 Perform optimizations where we check to see if a jump branches to a
4698 location where another comparison subsumed by the first is found. If
4699 so, the first branch is redirected to either the destination of the
4700 second branch or a point immediately following it, depending on whether
4701 the condition is known to be true or false.
4703 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4705 @item -fcse-follow-jumps
4706 @opindex fcse-follow-jumps
4707 In common subexpression elimination, scan through jump instructions
4708 when the target of the jump is not reached by any other path. For
4709 example, when CSE encounters an @code{if} statement with an
4710 @code{else} clause, CSE will follow the jump when the condition
4713 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4715 @item -fcse-skip-blocks
4716 @opindex fcse-skip-blocks
4717 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4718 follow jumps which conditionally skip over blocks. When CSE
4719 encounters a simple @code{if} statement with no else clause,
4720 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4721 body of the @code{if}.
4723 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4725 @item -frerun-cse-after-loop
4726 @opindex frerun-cse-after-loop
4727 Re-run common subexpression elimination after loop optimizations has been
4730 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4734 Perform a global common subexpression elimination pass.
4735 This pass also performs global constant and copy propagation.
4737 @emph{Note:} When compiling a program using computed gotos, a GCC
4738 extension, you may get better runtime performance if you disable
4739 the global common subexpression elimination pass by adding
4740 @option{-fno-gcse} to the command line.
4742 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4746 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4747 attempt to move loads which are only killed by stores into themselves. This
4748 allows a loop containing a load/store sequence to be changed to a load outside
4749 the loop, and a copy/store within the loop.
4751 Enabled by default when gcse is enabled.
4755 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4756 global common subexpression elimination. This pass will attempt to move
4757 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4758 loops containing a load/store sequence can be changed to a load before
4759 the loop and a store after the loop.
4761 Not enabled at any optimization level.
4765 When @option{-fgcse-las} is enabled, the global common subexpression
4766 elimination pass eliminates redundant loads that come after stores to the
4767 same memory location (both partial and full redundancies).
4769 Not enabled at any optimization level.
4771 @item -fgcse-after-reload
4772 @opindex fgcse-after-reload
4773 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4774 pass is performed after reload. The purpose of this pass is to cleanup
4777 @item -funsafe-loop-optimizations
4778 @opindex funsafe-loop-optimizations
4779 If given, the loop optimizer will assume that loop indices do not
4780 overflow, and that the loops with nontrivial exit condition are not
4781 infinite. This enables a wider range of loop optimizations even if
4782 the loop optimizer itself cannot prove that these assumptions are valid.
4783 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
4784 if it finds this kind of loop.
4786 @item -fcrossjumping
4787 @opindex crossjumping
4788 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4789 resulting code may or may not perform better than without cross-jumping.
4791 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4793 @item -fif-conversion
4794 @opindex if-conversion
4795 Attempt to transform conditional jumps into branch-less equivalents. This
4796 include use of conditional moves, min, max, set flags and abs instructions, and
4797 some tricks doable by standard arithmetics. The use of conditional execution
4798 on chips where it is available is controlled by @code{if-conversion2}.
4800 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4802 @item -fif-conversion2
4803 @opindex if-conversion2
4804 Use conditional execution (where available) to transform conditional jumps into
4805 branch-less equivalents.
4807 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4809 @item -fdelete-null-pointer-checks
4810 @opindex fdelete-null-pointer-checks
4811 Use global dataflow analysis to identify and eliminate useless checks
4812 for null pointers. The compiler assumes that dereferencing a null
4813 pointer would have halted the program. If a pointer is checked after
4814 it has already been dereferenced, it cannot be null.
4816 In some environments, this assumption is not true, and programs can
4817 safely dereference null pointers. Use
4818 @option{-fno-delete-null-pointer-checks} to disable this optimization
4819 for programs which depend on that behavior.
4821 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4823 @item -fexpensive-optimizations
4824 @opindex fexpensive-optimizations
4825 Perform a number of minor optimizations that are relatively expensive.
4827 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4829 @item -foptimize-register-move
4831 @opindex foptimize-register-move
4833 Attempt to reassign register numbers in move instructions and as
4834 operands of other simple instructions in order to maximize the amount of
4835 register tying. This is especially helpful on machines with two-operand
4838 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4841 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4843 @item -fdelayed-branch
4844 @opindex fdelayed-branch
4845 If supported for the target machine, attempt to reorder instructions
4846 to exploit instruction slots available after delayed branch
4849 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4851 @item -fschedule-insns
4852 @opindex fschedule-insns
4853 If supported for the target machine, attempt to reorder instructions to
4854 eliminate execution stalls due to required data being unavailable. This
4855 helps machines that have slow floating point or memory load instructions
4856 by allowing other instructions to be issued until the result of the load
4857 or floating point instruction is required.
4859 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4861 @item -fschedule-insns2
4862 @opindex fschedule-insns2
4863 Similar to @option{-fschedule-insns}, but requests an additional pass of
4864 instruction scheduling after register allocation has been done. This is
4865 especially useful on machines with a relatively small number of
4866 registers and where memory load instructions take more than one cycle.
4868 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4870 @item -fno-sched-interblock
4871 @opindex fno-sched-interblock
4872 Don't schedule instructions across basic blocks. This is normally
4873 enabled by default when scheduling before register allocation, i.e.@:
4874 with @option{-fschedule-insns} or at @option{-O2} or higher.
4876 @item -fno-sched-spec
4877 @opindex fno-sched-spec
4878 Don't allow speculative motion of non-load instructions. This is normally
4879 enabled by default when scheduling before register allocation, i.e.@:
4880 with @option{-fschedule-insns} or at @option{-O2} or higher.
4882 @item -fsched-spec-load
4883 @opindex fsched-spec-load
4884 Allow speculative motion of some load instructions. This only makes
4885 sense when scheduling before register allocation, i.e.@: with
4886 @option{-fschedule-insns} or at @option{-O2} or higher.
4888 @item -fsched-spec-load-dangerous
4889 @opindex fsched-spec-load-dangerous
4890 Allow speculative motion of more load instructions. This only makes
4891 sense when scheduling before register allocation, i.e.@: with
4892 @option{-fschedule-insns} or at @option{-O2} or higher.
4894 @item -fsched-stalled-insns=@var{n}
4895 @opindex fsched-stalled-insns
4896 Define how many insns (if any) can be moved prematurely from the queue
4897 of stalled insns into the ready list, during the second scheduling pass.
4899 @item -fsched-stalled-insns-dep=@var{n}
4900 @opindex fsched-stalled-insns-dep
4901 Define how many insn groups (cycles) will be examined for a dependency
4902 on a stalled insn that is candidate for premature removal from the queue
4903 of stalled insns. Has an effect only during the second scheduling pass,
4904 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4906 @item -fsched2-use-superblocks
4907 @opindex fsched2-use-superblocks
4908 When scheduling after register allocation, do use superblock scheduling
4909 algorithm. Superblock scheduling allows motion across basic block boundaries
4910 resulting on faster schedules. This option is experimental, as not all machine
4911 descriptions used by GCC model the CPU closely enough to avoid unreliable
4912 results from the algorithm.
4914 This only makes sense when scheduling after register allocation, i.e.@: with
4915 @option{-fschedule-insns2} or at @option{-O2} or higher.
4917 @item -fsched2-use-traces
4918 @opindex fsched2-use-traces
4919 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4920 allocation and additionally perform code duplication in order to increase the
4921 size of superblocks using tracer pass. See @option{-ftracer} for details on
4924 This mode should produce faster but significantly longer programs. Also
4925 without @option{-fbranch-probabilities} the traces constructed may not
4926 match the reality and hurt the performance. This only makes
4927 sense when scheduling after register allocation, i.e.@: with
4928 @option{-fschedule-insns2} or at @option{-O2} or higher.
4932 Eliminates redundant extension instructions and move the non redundant
4933 ones to optimal placement using LCM.
4935 @item -freschedule-modulo-scheduled-loops
4936 @opindex fscheduling-in-modulo-scheduled-loops
4937 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4938 we may want to prevent the later scheduling passes from changing its schedule, we use this
4939 option to control that.
4941 @item -fcaller-saves
4942 @opindex fcaller-saves
4943 Enable values to be allocated in registers that will be clobbered by
4944 function calls, by emitting extra instructions to save and restore the
4945 registers around such calls. Such allocation is done only when it
4946 seems to result in better code than would otherwise be produced.
4948 This option is always enabled by default on certain machines, usually
4949 those which have no call-preserved registers to use instead.
4951 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4954 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4955 enabled by default at @option{-O2} and @option{-O3}.
4958 Perform Full Redundancy Elimination (FRE) on trees. The difference
4959 between FRE and PRE is that FRE only considers expressions
4960 that are computed on all paths leading to the redundant computation.
4961 This analysis faster than PRE, though it exposes fewer redundancies.
4962 This flag is enabled by default at @option{-O} and higher.
4964 @item -ftree-copy-prop
4965 Perform copy propagation on trees. This pass eliminates unnecessary
4966 copy operations. This flag is enabled by default at @option{-O} and
4969 @item -ftree-store-copy-prop
4970 Perform copy propagation of memory loads and stores. This pass
4971 eliminates unnecessary copy operations in memory references
4972 (structures, global variables, arrays, etc). This flag is enabled by
4973 default at @option{-O2} and higher.
4976 Perform structural alias analysis on trees. This flag
4977 is enabled by default at @option{-O} and higher.
4980 Perform interprocedural pointer analysis.
4983 Perform forward store motion on trees. This flag is
4984 enabled by default at @option{-O} and higher.
4987 Perform sparse conditional constant propagation (CCP) on trees. This
4988 pass only operates on local scalar variables and is enabled by default
4989 at @option{-O} and higher.
4991 @item -ftree-store-ccp
4992 Perform sparse conditional constant propagation (CCP) on trees. This
4993 pass operates on both local scalar variables and memory stores and
4994 loads (global variables, structures, arrays, etc). This flag is
4995 enabled by default at @option{-O2} and higher.
4998 Perform dead code elimination (DCE) on trees. This flag is enabled by
4999 default at @option{-O} and higher.
5001 @item -ftree-dominator-opts
5002 Perform a variety of simple scalar cleanups (constant/copy
5003 propagation, redundancy elimination, range propagation and expression
5004 simplification) based on a dominator tree traversal. This also
5005 performs jump threading (to reduce jumps to jumps). This flag is
5006 enabled by default at @option{-O} and higher.
5009 Perform loop header copying on trees. This is beneficial since it increases
5010 effectiveness of code motion optimizations. It also saves one jump. This flag
5011 is enabled by default at @option{-O} and higher. It is not enabled
5012 for @option{-Os}, since it usually increases code size.
5014 @item -ftree-loop-optimize
5015 Perform loop optimizations on trees. This flag is enabled by default
5016 at @option{-O} and higher.
5018 @item -ftree-loop-linear
5019 Perform linear loop transformations on tree. This flag can improve cache
5020 performance and allow further loop optimizations to take place.
5022 @item -ftree-loop-im
5023 Perform loop invariant motion on trees. This pass moves only invariants that
5024 would be hard to handle at RTL level (function calls, operations that expand to
5025 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5026 operands of conditions that are invariant out of the loop, so that we can use
5027 just trivial invariantness analysis in loop unswitching. The pass also includes
5030 @item -ftree-loop-ivcanon
5031 Create a canonical counter for number of iterations in the loop for that
5032 determining number of iterations requires complicated analysis. Later
5033 optimizations then may determine the number easily. Useful especially
5034 in connection with unrolling.
5037 Perform induction variable optimizations (strength reduction, induction
5038 variable merging and induction variable elimination) on trees.
5041 Perform scalar replacement of aggregates. This pass replaces structure
5042 references with scalars to prevent committing structures to memory too
5043 early. This flag is enabled by default at @option{-O} and higher.
5045 @item -ftree-copyrename
5046 Perform copy renaming on trees. This pass attempts to rename compiler
5047 temporaries to other variables at copy locations, usually resulting in
5048 variable names which more closely resemble the original variables. This flag
5049 is enabled by default at @option{-O} and higher.
5052 Perform temporary expression replacement during the SSA->normal phase. Single
5053 use/single def temporaries are replaced at their use location with their
5054 defining expression. This results in non-GIMPLE code, but gives the expanders
5055 much more complex trees to work on resulting in better RTL generation. This is
5056 enabled by default at @option{-O} and higher.
5059 Perform live range splitting during the SSA->normal phase. Distinct live
5060 ranges of a variable are split into unique variables, allowing for better
5061 optimization later. This is enabled by default at @option{-O} and higher.
5063 @item -ftree-vectorize
5064 Perform loop vectorization on trees.
5066 @item -ftree-vect-loop-version
5067 @opindex ftree-vect-loop-version
5068 Perform loop versioning when doing loop vectorization on trees. When a loop
5069 appears to be vectorizable except that data alignment or data dependence cannot
5070 be determined at compile time then vectorized and non-vectorized versions of
5071 the loop are generated along with runtime checks for alignment or dependence
5072 to control which version is executed. This option is enabled by default
5073 except at level @option{-Os} where it is disabled.
5076 Perform Value Range Propagation on trees. This is similar to the
5077 constant propagation pass, but instead of values, ranges of values are
5078 propagated. This allows the optimizers to remove unnecessary range
5079 checks like array bound checks and null pointer checks. This is
5080 enabled by default at @option{-O2} and higher. Null pointer check
5081 elimination is only done if @option{-fdelete-null-pointer-checks} is
5086 Perform tail duplication to enlarge superblock size. This transformation
5087 simplifies the control flow of the function allowing other optimizations to do
5090 @item -funroll-loops
5091 @opindex funroll-loops
5092 Unroll loops whose number of iterations can be determined at compile
5093 time or upon entry to the loop. @option{-funroll-loops} implies
5094 @option{-frerun-cse-after-loop}. This option makes code larger,
5095 and may or may not make it run faster.
5097 @item -funroll-all-loops
5098 @opindex funroll-all-loops
5099 Unroll all loops, even if their number of iterations is uncertain when
5100 the loop is entered. This usually makes programs run more slowly.
5101 @option{-funroll-all-loops} implies the same options as
5102 @option{-funroll-loops},
5104 @item -fsplit-ivs-in-unroller
5105 @opindex fsplit-ivs-in-unroller
5106 Enables expressing of values of induction variables in later iterations
5107 of the unrolled loop using the value in the first iteration. This breaks
5108 long dependency chains, thus improving efficiency of the scheduling passes.
5110 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5111 same effect. However in cases the loop body is more complicated than
5112 a single basic block, this is not reliable. It also does not work at all
5113 on some of the architectures due to restrictions in the CSE pass.
5115 This optimization is enabled by default.
5117 @item -fvariable-expansion-in-unroller
5118 @opindex fvariable-expansion-in-unroller
5119 With this option, the compiler will create multiple copies of some
5120 local variables when unrolling a loop which can result in superior code.
5122 @item -fprefetch-loop-arrays
5123 @opindex fprefetch-loop-arrays
5124 If supported by the target machine, generate instructions to prefetch
5125 memory to improve the performance of loops that access large arrays.
5127 This option may generate better or worse code; results are highly
5128 dependent on the structure of loops within the source code.
5130 Disabled at level @option{-Os}.
5133 @itemx -fno-peephole2
5134 @opindex fno-peephole
5135 @opindex fno-peephole2
5136 Disable any machine-specific peephole optimizations. The difference
5137 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5138 are implemented in the compiler; some targets use one, some use the
5139 other, a few use both.
5141 @option{-fpeephole} is enabled by default.
5142 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5144 @item -fno-guess-branch-probability
5145 @opindex fno-guess-branch-probability
5146 Do not guess branch probabilities using heuristics.
5148 GCC will use heuristics to guess branch probabilities if they are
5149 not provided by profiling feedback (@option{-fprofile-arcs}). These
5150 heuristics are based on the control flow graph. If some branch probabilities
5151 are specified by @samp{__builtin_expect}, then the heuristics will be
5152 used to guess branch probabilities for the rest of the control flow graph,
5153 taking the @samp{__builtin_expect} info into account. The interactions
5154 between the heuristics and @samp{__builtin_expect} can be complex, and in
5155 some cases, it may be useful to disable the heuristics so that the effects
5156 of @samp{__builtin_expect} are easier to understand.
5158 The default is @option{-fguess-branch-probability} at levels
5159 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5161 @item -freorder-blocks
5162 @opindex freorder-blocks
5163 Reorder basic blocks in the compiled function in order to reduce number of
5164 taken branches and improve code locality.
5166 Enabled at levels @option{-O2}, @option{-O3}.
5168 @item -freorder-blocks-and-partition
5169 @opindex freorder-blocks-and-partition
5170 In addition to reordering basic blocks in the compiled function, in order
5171 to reduce number of taken branches, partitions hot and cold basic blocks
5172 into separate sections of the assembly and .o files, to improve
5173 paging and cache locality performance.
5175 This optimization is automatically turned off in the presence of
5176 exception handling, for linkonce sections, for functions with a user-defined
5177 section attribute and on any architecture that does not support named
5180 @item -freorder-functions
5181 @opindex freorder-functions
5182 Reorder functions in the object file in order to
5183 improve code locality. This is implemented by using special
5184 subsections @code{.text.hot} for most frequently executed functions and
5185 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5186 the linker so object file format must support named sections and linker must
5187 place them in a reasonable way.
5189 Also profile feedback must be available in to make this option effective. See
5190 @option{-fprofile-arcs} for details.
5192 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5194 @item -fstrict-aliasing
5195 @opindex fstrict-aliasing
5196 Allows the compiler to assume the strictest aliasing rules applicable to
5197 the language being compiled. For C (and C++), this activates
5198 optimizations based on the type of expressions. In particular, an
5199 object of one type is assumed never to reside at the same address as an
5200 object of a different type, unless the types are almost the same. For
5201 example, an @code{unsigned int} can alias an @code{int}, but not a
5202 @code{void*} or a @code{double}. A character type may alias any other
5205 Pay special attention to code like this:
5218 The practice of reading from a different union member than the one most
5219 recently written to (called ``type-punning'') is common. Even with
5220 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5221 is accessed through the union type. So, the code above will work as
5222 expected. However, this code might not:
5233 Every language that wishes to perform language-specific alias analysis
5234 should define a function that computes, given an @code{tree}
5235 node, an alias set for the node. Nodes in different alias sets are not
5236 allowed to alias. For an example, see the C front-end function
5237 @code{c_get_alias_set}.
5239 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5241 @item -fstrict-overflow
5242 @opindex fstrict-overflow
5243 Allow the compiler to assume strict signed overflow rules, depending
5244 on the language being compiled. For C (and C++) this means that
5245 overflow when doing arithmetic with signed numbers is undefined, which
5246 means that the compiler may assume that it will not happen. This
5247 permits various optimizations. For example, the compiler will assume
5248 that an expression like @code{i + 10 > i} will always be true for
5249 signed @code{i}. This assumption is only valid if signed overflow is
5250 undefined, as the expression is false if @code{i + 10} overflows when
5251 using twos complement arithmetic. When this option is in effect any
5252 attempt to determine whether an operation on signed numbers will
5253 overflow must be written carefully to not actually involve overflow.
5255 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
5256 that signed overflow is fully defined: it wraps. When
5257 @option{-fwrapv} is used, there is no difference between
5258 @option{-fstrict-overflow} and @option{-fno-strict-overflow}. With
5259 @option{-fwrapv} certain types of overflow are permitted. For
5260 example, if the compiler gets an overflow when doing arithmetic on
5261 constants, the overflowed value can still be used with
5262 @option{-fwrapv}, but not otherwise.
5264 The @option{-fstrict-overflow} option is enabled at levels
5265 @option{-O2}, @option{-O3}, @option{-Os}.
5267 @item -falign-functions
5268 @itemx -falign-functions=@var{n}
5269 @opindex falign-functions
5270 Align the start of functions to the next power-of-two greater than
5271 @var{n}, skipping up to @var{n} bytes. For instance,
5272 @option{-falign-functions=32} aligns functions to the next 32-byte
5273 boundary, but @option{-falign-functions=24} would align to the next
5274 32-byte boundary only if this can be done by skipping 23 bytes or less.
5276 @option{-fno-align-functions} and @option{-falign-functions=1} are
5277 equivalent and mean that functions will not be aligned.
5279 Some assemblers only support this flag when @var{n} is a power of two;
5280 in that case, it is rounded up.
5282 If @var{n} is not specified or is zero, use a machine-dependent default.
5284 Enabled at levels @option{-O2}, @option{-O3}.
5286 @item -falign-labels
5287 @itemx -falign-labels=@var{n}
5288 @opindex falign-labels
5289 Align all branch targets to a power-of-two boundary, skipping up to
5290 @var{n} bytes like @option{-falign-functions}. This option can easily
5291 make code slower, because it must insert dummy operations for when the
5292 branch target is reached in the usual flow of the code.
5294 @option{-fno-align-labels} and @option{-falign-labels=1} are
5295 equivalent and mean that labels will not be aligned.
5297 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5298 are greater than this value, then their values are used instead.
5300 If @var{n} is not specified or is zero, use a machine-dependent default
5301 which is very likely to be @samp{1}, meaning no alignment.
5303 Enabled at levels @option{-O2}, @option{-O3}.
5306 @itemx -falign-loops=@var{n}
5307 @opindex falign-loops
5308 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5309 like @option{-falign-functions}. The hope is that the loop will be
5310 executed many times, which will make up for any execution of the dummy
5313 @option{-fno-align-loops} and @option{-falign-loops=1} are
5314 equivalent and mean that loops will not be aligned.
5316 If @var{n} is not specified or is zero, use a machine-dependent default.
5318 Enabled at levels @option{-O2}, @option{-O3}.
5321 @itemx -falign-jumps=@var{n}
5322 @opindex falign-jumps
5323 Align branch targets to a power-of-two boundary, for branch targets
5324 where the targets can only be reached by jumping, skipping up to @var{n}
5325 bytes like @option{-falign-functions}. In this case, no dummy operations
5328 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5329 equivalent and mean that loops will not be aligned.
5331 If @var{n} is not specified or is zero, use a machine-dependent default.
5333 Enabled at levels @option{-O2}, @option{-O3}.
5335 @item -funit-at-a-time
5336 @opindex funit-at-a-time
5337 Parse the whole compilation unit before starting to produce code.
5338 This allows some extra optimizations to take place but consumes
5339 more memory (in general). There are some compatibility issues
5340 with @emph{unit-at-a-time} mode:
5343 enabling @emph{unit-at-a-time} mode may change the order
5344 in which functions, variables, and top-level @code{asm} statements
5345 are emitted, and will likely break code relying on some particular
5346 ordering. The majority of such top-level @code{asm} statements,
5347 though, can be replaced by @code{section} attributes. The
5348 @option{fno-toplevel-reorder} option may be used to keep the ordering
5349 used in the input file, at the cost of some optimizations.
5352 @emph{unit-at-a-time} mode removes unreferenced static variables
5353 and functions. This may result in undefined references
5354 when an @code{asm} statement refers directly to variables or functions
5355 that are otherwise unused. In that case either the variable/function
5356 shall be listed as an operand of the @code{asm} statement operand or,
5357 in the case of top-level @code{asm} statements the attribute @code{used}
5358 shall be used on the declaration.
5361 Static functions now can use non-standard passing conventions that
5362 may break @code{asm} statements calling functions directly. Again,
5363 attribute @code{used} will prevent this behavior.
5366 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5367 but this scheme may not be supported by future releases of GCC@.
5369 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5371 @item -fno-toplevel-reorder
5372 Do not reorder top-level functions, variables, and @code{asm}
5373 statements. Output them in the same order that they appear in the
5374 input file. When this option is used, unreferenced static variables
5375 will not be removed. This option is intended to support existing code
5376 which relies on a particular ordering. For new code, it is better to
5381 Constructs webs as commonly used for register allocation purposes and assign
5382 each web individual pseudo register. This allows the register allocation pass
5383 to operate on pseudos directly, but also strengthens several other optimization
5384 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5385 however, make debugging impossible, since variables will no longer stay in a
5388 Enabled by default with @option{-funroll-loops}.
5390 @item -fwhole-program
5391 @opindex fwhole-program
5392 Assume that the current compilation unit represents whole program being
5393 compiled. All public functions and variables with the exception of @code{main}
5394 and those merged by attribute @code{externally_visible} become static functions
5395 and in a affect gets more aggressively optimized by interprocedural optimizers.
5396 While this option is equivalent to proper use of @code{static} keyword for
5397 programs consisting of single file, in combination with option
5398 @option{--combine} this flag can be used to compile most of smaller scale C
5399 programs since the functions and variables become local for the whole combined
5400 compilation unit, not for the single source file itself.
5403 @item -fno-cprop-registers
5404 @opindex fno-cprop-registers
5405 After register allocation and post-register allocation instruction splitting,
5406 we perform a copy-propagation pass to try to reduce scheduling dependencies
5407 and occasionally eliminate the copy.
5409 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5411 @item -fprofile-generate
5412 @opindex fprofile-generate
5414 Enable options usually used for instrumenting application to produce
5415 profile useful for later recompilation with profile feedback based
5416 optimization. You must use @option{-fprofile-generate} both when
5417 compiling and when linking your program.
5419 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5422 @opindex fprofile-use
5423 Enable profile feedback directed optimizations, and optimizations
5424 generally profitable only with profile feedback available.
5426 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
5427 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
5431 The following options control compiler behavior regarding floating
5432 point arithmetic. These options trade off between speed and
5433 correctness. All must be specifically enabled.
5437 @opindex ffloat-store
5438 Do not store floating point variables in registers, and inhibit other
5439 options that might change whether a floating point value is taken from a
5442 @cindex floating point precision
5443 This option prevents undesirable excess precision on machines such as
5444 the 68000 where the floating registers (of the 68881) keep more
5445 precision than a @code{double} is supposed to have. Similarly for the
5446 x86 architecture. For most programs, the excess precision does only
5447 good, but a few programs rely on the precise definition of IEEE floating
5448 point. Use @option{-ffloat-store} for such programs, after modifying
5449 them to store all pertinent intermediate computations into variables.
5453 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5454 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5455 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5456 and @option{fcx-limited-range}.
5458 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5460 This option should never be turned on by any @option{-O} option since
5461 it can result in incorrect output for programs which depend on
5462 an exact implementation of IEEE or ISO rules/specifications for
5465 @item -fno-math-errno
5466 @opindex fno-math-errno
5467 Do not set ERRNO after calling math functions that are executed
5468 with a single instruction, e.g., sqrt. A program that relies on
5469 IEEE exceptions for math error handling may want to use this flag
5470 for speed while maintaining IEEE arithmetic compatibility.
5472 This option should never be turned on by any @option{-O} option since
5473 it can result in incorrect output for programs which depend on
5474 an exact implementation of IEEE or ISO rules/specifications for
5477 The default is @option{-fmath-errno}.
5479 On Darwin and FreeBSD systems, the math library never sets @code{errno}.
5481 no reason for the compiler to consider the possibility that it might,
5482 and @option{-fno-math-errno} is the default.
5484 @item -funsafe-math-optimizations
5485 @opindex funsafe-math-optimizations
5486 Allow optimizations for floating-point arithmetic that (a) assume
5487 that arguments and results are valid and (b) may violate IEEE or
5488 ANSI standards. When used at link-time, it may include libraries
5489 or startup files that change the default FPU control word or other
5490 similar optimizations.
5492 This option should never be turned on by any @option{-O} option since
5493 it can result in incorrect output for programs which depend on
5494 an exact implementation of IEEE or ISO rules/specifications for
5497 The default is @option{-fno-unsafe-math-optimizations}.
5499 @item -ffinite-math-only
5500 @opindex ffinite-math-only
5501 Allow optimizations for floating-point arithmetic that assume
5502 that arguments and results are not NaNs or +-Infs.
5504 This option should never be turned on by any @option{-O} option since
5505 it can result in incorrect output for programs which depend on
5506 an exact implementation of IEEE or ISO rules/specifications.
5508 The default is @option{-fno-finite-math-only}.
5510 @item -fno-trapping-math
5511 @opindex fno-trapping-math
5512 Compile code assuming that floating-point operations cannot generate
5513 user-visible traps. These traps include division by zero, overflow,
5514 underflow, inexact result and invalid operation. This option implies
5515 @option{-fno-signaling-nans}. Setting this option may allow faster
5516 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5518 This option should never be turned on by any @option{-O} option since
5519 it can result in incorrect output for programs which depend on
5520 an exact implementation of IEEE or ISO rules/specifications for
5523 The default is @option{-ftrapping-math}.
5525 @item -frounding-math
5526 @opindex frounding-math
5527 Disable transformations and optimizations that assume default floating
5528 point rounding behavior. This is round-to-zero for all floating point
5529 to integer conversions, and round-to-nearest for all other arithmetic
5530 truncations. This option should be specified for programs that change
5531 the FP rounding mode dynamically, or that may be executed with a
5532 non-default rounding mode. This option disables constant folding of
5533 floating point expressions at compile-time (which may be affected by
5534 rounding mode) and arithmetic transformations that are unsafe in the
5535 presence of sign-dependent rounding modes.
5537 The default is @option{-fno-rounding-math}.
5539 This option is experimental and does not currently guarantee to
5540 disable all GCC optimizations that are affected by rounding mode.
5541 Future versions of GCC may provide finer control of this setting
5542 using C99's @code{FENV_ACCESS} pragma. This command line option
5543 will be used to specify the default state for @code{FENV_ACCESS}.
5545 @item -frtl-abstract-sequences
5546 @opindex frtl-abstract-sequences
5547 It is a size optimization method. This option is to find identical
5548 sequences of code, which can be turned into pseudo-procedures and
5549 then replace all occurrences with calls to the newly created
5550 subroutine. It is kind of an opposite of @option{-finline-functions}.
5551 This optimization runs at RTL level.
5553 @item -fsignaling-nans
5554 @opindex fsignaling-nans
5555 Compile code assuming that IEEE signaling NaNs may generate user-visible
5556 traps during floating-point operations. Setting this option disables
5557 optimizations that may change the number of exceptions visible with
5558 signaling NaNs. This option implies @option{-ftrapping-math}.
5560 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5563 The default is @option{-fno-signaling-nans}.
5565 This option is experimental and does not currently guarantee to
5566 disable all GCC optimizations that affect signaling NaN behavior.
5568 @item -fsingle-precision-constant
5569 @opindex fsingle-precision-constant
5570 Treat floating point constant as single precision constant instead of
5571 implicitly converting it to double precision constant.
5573 @item -fcx-limited-range
5574 @itemx -fno-cx-limited-range
5575 @opindex fcx-limited-range
5576 @opindex fno-cx-limited-range
5577 When enabled, this option states that a range reduction step is not
5578 needed when performing complex division. The default is
5579 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5581 This option controls the default setting of the ISO C99
5582 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5587 The following options control optimizations that may improve
5588 performance, but are not enabled by any @option{-O} options. This
5589 section includes experimental options that may produce broken code.
5592 @item -fbranch-probabilities
5593 @opindex fbranch-probabilities
5594 After running a program compiled with @option{-fprofile-arcs}
5595 (@pxref{Debugging Options,, Options for Debugging Your Program or
5596 @command{gcc}}), you can compile it a second time using
5597 @option{-fbranch-probabilities}, to improve optimizations based on
5598 the number of times each branch was taken. When the program
5599 compiled with @option{-fprofile-arcs} exits it saves arc execution
5600 counts to a file called @file{@var{sourcename}.gcda} for each source
5601 file The information in this data file is very dependent on the
5602 structure of the generated code, so you must use the same source code
5603 and the same optimization options for both compilations.
5605 With @option{-fbranch-probabilities}, GCC puts a
5606 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5607 These can be used to improve optimization. Currently, they are only
5608 used in one place: in @file{reorg.c}, instead of guessing which path a
5609 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5610 exactly determine which path is taken more often.
5612 @item -fprofile-values
5613 @opindex fprofile-values
5614 If combined with @option{-fprofile-arcs}, it adds code so that some
5615 data about values of expressions in the program is gathered.
5617 With @option{-fbranch-probabilities}, it reads back the data gathered
5618 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5619 notes to instructions for their later usage in optimizations.
5621 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5625 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5626 a code to gather information about values of expressions.
5628 With @option{-fbranch-probabilities}, it reads back the data gathered
5629 and actually performs the optimizations based on them.
5630 Currently the optimizations include specialization of division operation
5631 using the knowledge about the value of the denominator.
5633 @item -frename-registers
5634 @opindex frename-registers
5635 Attempt to avoid false dependencies in scheduled code by making use
5636 of registers left over after register allocation. This optimization
5637 will most benefit processors with lots of registers. Depending on the
5638 debug information format adopted by the target, however, it can
5639 make debugging impossible, since variables will no longer stay in
5640 a ``home register''.
5642 Enabled by default with @option{-funroll-loops}.
5646 Perform tail duplication to enlarge superblock size. This transformation
5647 simplifies the control flow of the function allowing other optimizations to do
5650 Enabled with @option{-fprofile-use}.
5652 @item -funroll-loops
5653 @opindex funroll-loops
5654 Unroll loops whose number of iterations can be determined at compile time or
5655 upon entry to the loop. @option{-funroll-loops} implies
5656 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
5657 It also turns on complete loop peeling (i.e.@: complete removal of loops with
5658 small constant number of iterations). This option makes code larger, and may
5659 or may not make it run faster.
5661 Enabled with @option{-fprofile-use}.
5663 @item -funroll-all-loops
5664 @opindex funroll-all-loops
5665 Unroll all loops, even if their number of iterations is uncertain when
5666 the loop is entered. This usually makes programs run more slowly.
5667 @option{-funroll-all-loops} implies the same options as
5668 @option{-funroll-loops}.
5671 @opindex fpeel-loops
5672 Peels the loops for that there is enough information that they do not
5673 roll much (from profile feedback). It also turns on complete loop peeling
5674 (i.e.@: complete removal of loops with small constant number of iterations).
5676 Enabled with @option{-fprofile-use}.
5678 @item -fmove-loop-invariants
5679 @opindex fmove-loop-invariants
5680 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
5681 at level @option{-O1}
5683 @item -funswitch-loops
5684 @opindex funswitch-loops
5685 Move branches with loop invariant conditions out of the loop, with duplicates
5686 of the loop on both branches (modified according to result of the condition).
5688 @item -ffunction-sections
5689 @itemx -fdata-sections
5690 @opindex ffunction-sections
5691 @opindex fdata-sections
5692 Place each function or data item into its own section in the output
5693 file if the target supports arbitrary sections. The name of the
5694 function or the name of the data item determines the section's name
5697 Use these options on systems where the linker can perform optimizations
5698 to improve locality of reference in the instruction space. Most systems
5699 using the ELF object format and SPARC processors running Solaris 2 have
5700 linkers with such optimizations. AIX may have these optimizations in
5703 Only use these options when there are significant benefits from doing
5704 so. When you specify these options, the assembler and linker will
5705 create larger object and executable files and will also be slower.
5706 You will not be able to use @code{gprof} on all systems if you
5707 specify this option and you may have problems with debugging if
5708 you specify both this option and @option{-g}.
5710 @item -fbranch-target-load-optimize
5711 @opindex fbranch-target-load-optimize
5712 Perform branch target register load optimization before prologue / epilogue
5714 The use of target registers can typically be exposed only during reload,
5715 thus hoisting loads out of loops and doing inter-block scheduling needs
5716 a separate optimization pass.
5718 @item -fbranch-target-load-optimize2
5719 @opindex fbranch-target-load-optimize2
5720 Perform branch target register load optimization after prologue / epilogue
5723 @item -fbtr-bb-exclusive
5724 @opindex fbtr-bb-exclusive
5725 When performing branch target register load optimization, don't reuse
5726 branch target registers in within any basic block.
5728 @item -fstack-protector
5729 Emit extra code to check for buffer overflows, such as stack smashing
5730 attacks. This is done by adding a guard variable to functions with
5731 vulnerable objects. This includes functions that call alloca, and
5732 functions with buffers larger than 8 bytes. The guards are initialized
5733 when a function is entered and then checked when the function exits.
5734 If a guard check fails, an error message is printed and the program exits.
5736 @item -fstack-protector-all
5737 Like @option{-fstack-protector} except that all functions are protected.
5739 @item -fsection-anchors
5740 @opindex fsection-anchors
5741 Try to reduce the number of symbolic address calculations by using
5742 shared ``anchor'' symbols to address nearby objects. This transformation
5743 can help to reduce the number of GOT entries and GOT accesses on some
5746 For example, the implementation of the following function @code{foo}:
5750 int foo (void) @{ return a + b + c; @}
5753 would usually calculate the addresses of all three variables, but if you
5754 compile it with @option{-fsection-anchors}, it will access the variables
5755 from a common anchor point instead. The effect is similar to the
5756 following pseudocode (which isn't valid C):
5761 register int *xr = &x;
5762 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5766 Not all targets support this option.
5768 @item --param @var{name}=@var{value}
5770 In some places, GCC uses various constants to control the amount of
5771 optimization that is done. For example, GCC will not inline functions
5772 that contain more that a certain number of instructions. You can
5773 control some of these constants on the command-line using the
5774 @option{--param} option.
5776 The names of specific parameters, and the meaning of the values, are
5777 tied to the internals of the compiler, and are subject to change
5778 without notice in future releases.
5780 In each case, the @var{value} is an integer. The allowable choices for
5781 @var{name} are given in the following table:
5784 @item salias-max-implicit-fields
5785 The maximum number of fields in a variable without direct
5786 structure accesses for which structure aliasing will consider trying
5787 to track each field. The default is 5
5789 @item salias-max-array-elements
5790 The maximum number of elements an array can have and its elements
5791 still be tracked individually by structure aliasing. The default is 4
5793 @item sra-max-structure-size
5794 The maximum structure size, in bytes, at which the scalar replacement
5795 of aggregates (SRA) optimization will perform block copies. The
5796 default value, 0, implies that GCC will select the most appropriate
5799 @item sra-field-structure-ratio
5800 The threshold ratio (as a percentage) between instantiated fields and
5801 the complete structure size. We say that if the ratio of the number
5802 of bytes in instantiated fields to the number of bytes in the complete
5803 structure exceeds this parameter, then block copies are not used. The
5806 @item max-crossjump-edges
5807 The maximum number of incoming edges to consider for crossjumping.
5808 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5809 the number of edges incoming to each block. Increasing values mean
5810 more aggressive optimization, making the compile time increase with
5811 probably small improvement in executable size.
5813 @item min-crossjump-insns
5814 The minimum number of instructions which must be matched at the end
5815 of two blocks before crossjumping will be performed on them. This
5816 value is ignored in the case where all instructions in the block being
5817 crossjumped from are matched. The default value is 5.
5819 @item max-grow-copy-bb-insns
5820 The maximum code size expansion factor when copying basic blocks
5821 instead of jumping. The expansion is relative to a jump instruction.
5822 The default value is 8.
5824 @item max-goto-duplication-insns
5825 The maximum number of instructions to duplicate to a block that jumps
5826 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5827 passes, GCC factors computed gotos early in the compilation process,
5828 and unfactors them as late as possible. Only computed jumps at the
5829 end of a basic blocks with no more than max-goto-duplication-insns are
5830 unfactored. The default value is 8.
5832 @item max-delay-slot-insn-search
5833 The maximum number of instructions to consider when looking for an
5834 instruction to fill a delay slot. If more than this arbitrary number of
5835 instructions is searched, the time savings from filling the delay slot
5836 will be minimal so stop searching. Increasing values mean more
5837 aggressive optimization, making the compile time increase with probably
5838 small improvement in executable run time.
5840 @item max-delay-slot-live-search
5841 When trying to fill delay slots, the maximum number of instructions to
5842 consider when searching for a block with valid live register
5843 information. Increasing this arbitrarily chosen value means more
5844 aggressive optimization, increasing the compile time. This parameter
5845 should be removed when the delay slot code is rewritten to maintain the
5848 @item max-gcse-memory
5849 The approximate maximum amount of memory that will be allocated in
5850 order to perform the global common subexpression elimination
5851 optimization. If more memory than specified is required, the
5852 optimization will not be done.
5854 @item max-gcse-passes
5855 The maximum number of passes of GCSE to run. The default is 1.
5857 @item max-pending-list-length
5858 The maximum number of pending dependencies scheduling will allow
5859 before flushing the current state and starting over. Large functions
5860 with few branches or calls can create excessively large lists which
5861 needlessly consume memory and resources.
5863 @item max-inline-insns-single
5864 Several parameters control the tree inliner used in gcc.
5865 This number sets the maximum number of instructions (counted in GCC's
5866 internal representation) in a single function that the tree inliner
5867 will consider for inlining. This only affects functions declared
5868 inline and methods implemented in a class declaration (C++).
5869 The default value is 450.
5871 @item max-inline-insns-auto
5872 When you use @option{-finline-functions} (included in @option{-O3}),
5873 a lot of functions that would otherwise not be considered for inlining
5874 by the compiler will be investigated. To those functions, a different
5875 (more restrictive) limit compared to functions declared inline can
5877 The default value is 90.
5879 @item large-function-insns
5880 The limit specifying really large functions. For functions larger than this
5881 limit after inlining inlining is constrained by
5882 @option{--param large-function-growth}. This parameter is useful primarily
5883 to avoid extreme compilation time caused by non-linear algorithms used by the
5885 This parameter is ignored when @option{-funit-at-a-time} is not used.
5886 The default value is 2700.
5888 @item large-function-growth
5889 Specifies maximal growth of large function caused by inlining in percents.
5890 This parameter is ignored when @option{-funit-at-a-time} is not used.
5891 The default value is 100 which limits large function growth to 2.0 times
5894 @item large-unit-insns
5895 The limit specifying large translation unit. Growth caused by inlining of
5896 units larger than this limit is limited by @option{--param inline-unit-growth}.
5897 For small units this might be too tight (consider unit consisting of function A
5898 that is inline and B that just calls A three time. If B is small relative to
5899 A, the growth of unit is 300\% and yet such inlining is very sane. For very
5900 large units consisting of small inlininable functions however the overall unit
5901 growth limit is needed to avoid exponential explosion of code size. Thus for
5902 smaller units, the size is increased to @option{--param large-unit-insns}
5903 before applying @option{--param inline-unit-growth}. The default is 10000
5905 @item inline-unit-growth
5906 Specifies maximal overall growth of the compilation unit caused by inlining.
5907 This parameter is ignored when @option{-funit-at-a-time} is not used.
5908 The default value is 50 which limits unit growth to 1.5 times the original
5911 @item max-inline-insns-recursive
5912 @itemx max-inline-insns-recursive-auto
5913 Specifies maximum number of instructions out-of-line copy of self recursive inline
5914 function can grow into by performing recursive inlining.
5916 For functions declared inline @option{--param max-inline-insns-recursive} is
5917 taken into account. For function not declared inline, recursive inlining
5918 happens only when @option{-finline-functions} (included in @option{-O3}) is
5919 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5920 default value is 450.
5922 @item max-inline-recursive-depth
5923 @itemx max-inline-recursive-depth-auto
5924 Specifies maximum recursion depth used by the recursive inlining.
5926 For functions declared inline @option{--param max-inline-recursive-depth} is
5927 taken into account. For function not declared inline, recursive inlining
5928 happens only when @option{-finline-functions} (included in @option{-O3}) is
5929 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5930 default value is 450.
5932 @item min-inline-recursive-probability
5933 Recursive inlining is profitable only for function having deep recursion
5934 in average and can hurt for function having little recursion depth by
5935 increasing the prologue size or complexity of function body to other
5938 When profile feedback is available (see @option{-fprofile-generate}) the actual
5939 recursion depth can be guessed from probability that function will recurse via
5940 given call expression. This parameter limits inlining only to call expression
5941 whose probability exceeds given threshold (in percents). The default value is
5944 @item inline-call-cost
5945 Specify cost of call instruction relative to simple arithmetics operations
5946 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
5947 functions and at the same time increases size of leaf function that is believed to
5948 reduce function size by being inlined. In effect it increases amount of
5949 inlining for code having large abstraction penalty (many functions that just
5950 pass the arguments to other functions) and decrease inlining for code with low
5951 abstraction penalty. The default value is 16.
5953 @item max-unrolled-insns
5954 The maximum number of instructions that a loop should have if that loop
5955 is unrolled, and if the loop is unrolled, it determines how many times
5956 the loop code is unrolled.
5958 @item max-average-unrolled-insns
5959 The maximum number of instructions biased by probabilities of their execution
5960 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5961 it determines how many times the loop code is unrolled.
5963 @item max-unroll-times
5964 The maximum number of unrollings of a single loop.
5966 @item max-peeled-insns
5967 The maximum number of instructions that a loop should have if that loop
5968 is peeled, and if the loop is peeled, it determines how many times
5969 the loop code is peeled.
5971 @item max-peel-times
5972 The maximum number of peelings of a single loop.
5974 @item max-completely-peeled-insns
5975 The maximum number of insns of a completely peeled loop.
5977 @item max-completely-peel-times
5978 The maximum number of iterations of a loop to be suitable for complete peeling.
5980 @item max-unswitch-insns
5981 The maximum number of insns of an unswitched loop.
5983 @item max-unswitch-level
5984 The maximum number of branches unswitched in a single loop.
5987 The minimum cost of an expensive expression in the loop invariant motion.
5989 @item iv-consider-all-candidates-bound
5990 Bound on number of candidates for induction variables below that
5991 all candidates are considered for each use in induction variable
5992 optimizations. Only the most relevant candidates are considered
5993 if there are more candidates, to avoid quadratic time complexity.
5995 @item iv-max-considered-uses
5996 The induction variable optimizations give up on loops that contain more
5997 induction variable uses.
5999 @item iv-always-prune-cand-set-bound
6000 If number of candidates in the set is smaller than this value,
6001 we always try to remove unnecessary ivs from the set during its
6002 optimization when a new iv is added to the set.
6004 @item scev-max-expr-size
6005 Bound on size of expressions used in the scalar evolutions analyzer.
6006 Large expressions slow the analyzer.
6008 @item vect-max-version-checks
6009 The maximum number of runtime checks that can be performed when doing
6010 loop versioning in the vectorizer. See option ftree-vect-loop-version
6011 for more information.
6013 @item max-iterations-to-track
6015 The maximum number of iterations of a loop the brute force algorithm
6016 for analysis of # of iterations of the loop tries to evaluate.
6018 @item hot-bb-count-fraction
6019 Select fraction of the maximal count of repetitions of basic block in program
6020 given basic block needs to have to be considered hot.
6022 @item hot-bb-frequency-fraction
6023 Select fraction of the maximal frequency of executions of basic block in
6024 function given basic block needs to have to be considered hot
6026 @item max-predicted-iterations
6027 The maximum number of loop iterations we predict statically. This is useful
6028 in cases where function contain single loop with known bound and other loop
6029 with unknown. We predict the known number of iterations correctly, while
6030 the unknown number of iterations average to roughly 10. This means that the
6031 loop without bounds would appear artificially cold relative to the other one.
6033 @item tracer-dynamic-coverage
6034 @itemx tracer-dynamic-coverage-feedback
6036 This value is used to limit superblock formation once the given percentage of
6037 executed instructions is covered. This limits unnecessary code size
6040 The @option{tracer-dynamic-coverage-feedback} is used only when profile
6041 feedback is available. The real profiles (as opposed to statically estimated
6042 ones) are much less balanced allowing the threshold to be larger value.
6044 @item tracer-max-code-growth
6045 Stop tail duplication once code growth has reached given percentage. This is
6046 rather hokey argument, as most of the duplicates will be eliminated later in
6047 cross jumping, so it may be set to much higher values than is the desired code
6050 @item tracer-min-branch-ratio
6052 Stop reverse growth when the reverse probability of best edge is less than this
6053 threshold (in percent).
6055 @item tracer-min-branch-ratio
6056 @itemx tracer-min-branch-ratio-feedback
6058 Stop forward growth if the best edge do have probability lower than this
6061 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
6062 compilation for profile feedback and one for compilation without. The value
6063 for compilation with profile feedback needs to be more conservative (higher) in
6064 order to make tracer effective.
6066 @item max-cse-path-length
6068 Maximum number of basic blocks on path that cse considers. The default is 10.
6071 The maximum instructions CSE process before flushing. The default is 1000.
6073 @item global-var-threshold
6075 Counts the number of function calls (@var{n}) and the number of
6076 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
6077 single artificial variable will be created to represent all the
6078 call-clobbered variables at function call sites. This artificial
6079 variable will then be made to alias every call-clobbered variable.
6080 (done as @code{int * size_t} on the host machine; beware overflow).
6082 @item max-aliased-vops
6084 Maximum number of virtual operands allowed to represent aliases
6085 before triggering the alias grouping heuristic. Alias grouping
6086 reduces compile times and memory consumption needed for aliasing at
6087 the expense of precision loss in alias information.
6089 @item ggc-min-expand
6091 GCC uses a garbage collector to manage its own memory allocation. This
6092 parameter specifies the minimum percentage by which the garbage
6093 collector's heap should be allowed to expand between collections.
6094 Tuning this may improve compilation speed; it has no effect on code
6097 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6098 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6099 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6100 GCC is not able to calculate RAM on a particular platform, the lower
6101 bound of 30% is used. Setting this parameter and
6102 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6103 every opportunity. This is extremely slow, but can be useful for
6106 @item ggc-min-heapsize
6108 Minimum size of the garbage collector's heap before it begins bothering
6109 to collect garbage. The first collection occurs after the heap expands
6110 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6111 tuning this may improve compilation speed, and has no effect on code
6114 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6115 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6116 with a lower bound of 4096 (four megabytes) and an upper bound of
6117 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6118 particular platform, the lower bound is used. Setting this parameter
6119 very large effectively disables garbage collection. Setting this
6120 parameter and @option{ggc-min-expand} to zero causes a full collection
6121 to occur at every opportunity.
6123 @item max-reload-search-insns
6124 The maximum number of instruction reload should look backward for equivalent
6125 register. Increasing values mean more aggressive optimization, making the
6126 compile time increase with probably slightly better performance. The default
6129 @item max-cselib-memory-locations
6130 The maximum number of memory locations cselib should take into account.
6131 Increasing values mean more aggressive optimization, making the compile time
6132 increase with probably slightly better performance. The default value is 500.
6134 @item max-flow-memory-locations
6135 Similar as @option{max-cselib-memory-locations} but for dataflow liveness.
6136 The default value is 100.
6138 @item reorder-blocks-duplicate
6139 @itemx reorder-blocks-duplicate-feedback
6141 Used by basic block reordering pass to decide whether to use unconditional
6142 branch or duplicate the code on its destination. Code is duplicated when its
6143 estimated size is smaller than this value multiplied by the estimated size of
6144 unconditional jump in the hot spots of the program.
6146 The @option{reorder-block-duplicate-feedback} is used only when profile
6147 feedback is available and may be set to higher values than
6148 @option{reorder-block-duplicate} since information about the hot spots is more
6151 @item max-sched-ready-insns
6152 The maximum number of instructions ready to be issued the scheduler should
6153 consider at any given time during the first scheduling pass. Increasing
6154 values mean more thorough searches, making the compilation time increase
6155 with probably little benefit. The default value is 100.
6157 @item max-sched-region-blocks
6158 The maximum number of blocks in a region to be considered for
6159 interblock scheduling. The default value is 10.
6161 @item max-sched-region-insns
6162 The maximum number of insns in a region to be considered for
6163 interblock scheduling. The default value is 100.
6166 The minimum probability (in percents) of reaching a source block
6167 for interblock speculative scheduling. The default value is 40.
6169 @item max-sched-extend-regions-iters
6170 The maximum number of iterations through CFG to extend regions.
6171 0 - disable region extension,
6172 N - do at most N iterations.
6173 The default value is 0.
6175 @item max-sched-insn-conflict-delay
6176 The maximum conflict delay for an insn to be considered for speculative motion.
6177 The default value is 3.
6179 @item sched-spec-prob-cutoff
6180 The minimal probability of speculation success (in percents), so that
6181 speculative insn will be scheduled.
6182 The default value is 40.
6184 @item max-last-value-rtl
6186 The maximum size measured as number of RTLs that can be recorded in an expression
6187 in combiner for a pseudo register as last known value of that register. The default
6190 @item integer-share-limit
6191 Small integer constants can use a shared data structure, reducing the
6192 compiler's memory usage and increasing its speed. This sets the maximum
6193 value of a shared integer constant's. The default value is 256.
6195 @item min-virtual-mappings
6196 Specifies the minimum number of virtual mappings in the incremental
6197 SSA updater that should be registered to trigger the virtual mappings
6198 heuristic defined by virtual-mappings-ratio. The default value is
6201 @item virtual-mappings-ratio
6202 If the number of virtual mappings is virtual-mappings-ratio bigger
6203 than the number of virtual symbols to be updated, then the incremental
6204 SSA updater switches to a full update for those symbols. The default
6207 @item ssp-buffer-size
6208 The minimum size of buffers (i.e. arrays) that will receive stack smashing
6209 protection when @option{-fstack-protection} is used.
6211 @item max-jump-thread-duplication-stmts
6212 Maximum number of statements allowed in a block that needs to be
6213 duplicated when threading jumps.
6215 @item max-fields-for-field-sensitive
6216 Maximum number of fields in a structure we will treat in
6217 a field sensitive manner during pointer analysis.
6222 @node Preprocessor Options
6223 @section Options Controlling the Preprocessor
6224 @cindex preprocessor options
6225 @cindex options, preprocessor
6227 These options control the C preprocessor, which is run on each C source
6228 file before actual compilation.
6230 If you use the @option{-E} option, nothing is done except preprocessing.
6231 Some of these options make sense only together with @option{-E} because
6232 they cause the preprocessor output to be unsuitable for actual
6237 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6238 and pass @var{option} directly through to the preprocessor. If
6239 @var{option} contains commas, it is split into multiple options at the
6240 commas. However, many options are modified, translated or interpreted
6241 by the compiler driver before being passed to the preprocessor, and
6242 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6243 interface is undocumented and subject to change, so whenever possible
6244 you should avoid using @option{-Wp} and let the driver handle the
6247 @item -Xpreprocessor @var{option}
6248 @opindex preprocessor
6249 Pass @var{option} as an option to the preprocessor. You can use this to
6250 supply system-specific preprocessor options which GCC does not know how to
6253 If you want to pass an option that takes an argument, you must use
6254 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6257 @include cppopts.texi
6259 @node Assembler Options
6260 @section Passing Options to the Assembler
6262 @c prevent bad page break with this line
6263 You can pass options to the assembler.
6266 @item -Wa,@var{option}
6268 Pass @var{option} as an option to the assembler. If @var{option}
6269 contains commas, it is split into multiple options at the commas.
6271 @item -Xassembler @var{option}
6273 Pass @var{option} as an option to the assembler. You can use this to
6274 supply system-specific assembler options which GCC does not know how to
6277 If you want to pass an option that takes an argument, you must use
6278 @option{-Xassembler} twice, once for the option and once for the argument.
6283 @section Options for Linking
6284 @cindex link options
6285 @cindex options, linking
6287 These options come into play when the compiler links object files into
6288 an executable output file. They are meaningless if the compiler is
6289 not doing a link step.
6293 @item @var{object-file-name}
6294 A file name that does not end in a special recognized suffix is
6295 considered to name an object file or library. (Object files are
6296 distinguished from libraries by the linker according to the file
6297 contents.) If linking is done, these object files are used as input
6306 If any of these options is used, then the linker is not run, and
6307 object file names should not be used as arguments. @xref{Overall
6311 @item -l@var{library}
6312 @itemx -l @var{library}
6314 Search the library named @var{library} when linking. (The second
6315 alternative with the library as a separate argument is only for
6316 POSIX compliance and is not recommended.)
6318 It makes a difference where in the command you write this option; the
6319 linker searches and processes libraries and object files in the order they
6320 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6321 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6322 to functions in @samp{z}, those functions may not be loaded.
6324 The linker searches a standard list of directories for the library,
6325 which is actually a file named @file{lib@var{library}.a}. The linker
6326 then uses this file as if it had been specified precisely by name.
6328 The directories searched include several standard system directories
6329 plus any that you specify with @option{-L}.
6331 Normally the files found this way are library files---archive files
6332 whose members are object files. The linker handles an archive file by
6333 scanning through it for members which define symbols that have so far
6334 been referenced but not defined. But if the file that is found is an
6335 ordinary object file, it is linked in the usual fashion. The only
6336 difference between using an @option{-l} option and specifying a file name
6337 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6338 and searches several directories.
6341 @opindex nostartfiles
6342 Do not use the standard system startup files when linking.
6343 The standard system libraries are used normally, unless @option{-nostdlib}
6344 or @option{-nodefaultlibs} is used.
6346 @item -nodefaultlibs
6347 @opindex nodefaultlibs
6348 Do not use the standard system libraries when linking.
6349 Only the libraries you specify will be passed to the linker.
6350 The standard startup files are used normally, unless @option{-nostartfiles}
6351 is used. The compiler may generate calls to @code{memcmp},
6352 @code{memset}, @code{memcpy} and @code{memmove}.
6353 These entries are usually resolved by entries in
6354 libc. These entry points should be supplied through some other
6355 mechanism when this option is specified.
6359 Do not use the standard system startup files or libraries when linking.
6360 No startup files and only the libraries you specify will be passed to
6361 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6362 @code{memcpy} and @code{memmove}.
6363 These entries are usually resolved by entries in
6364 libc. These entry points should be supplied through some other
6365 mechanism when this option is specified.
6367 @cindex @option{-lgcc}, use with @option{-nostdlib}
6368 @cindex @option{-nostdlib} and unresolved references
6369 @cindex unresolved references and @option{-nostdlib}
6370 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6371 @cindex @option{-nodefaultlibs} and unresolved references
6372 @cindex unresolved references and @option{-nodefaultlibs}
6373 One of the standard libraries bypassed by @option{-nostdlib} and
6374 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6375 that GCC uses to overcome shortcomings of particular machines, or special
6376 needs for some languages.
6377 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6378 Collection (GCC) Internals},
6379 for more discussion of @file{libgcc.a}.)
6380 In most cases, you need @file{libgcc.a} even when you want to avoid
6381 other standard libraries. In other words, when you specify @option{-nostdlib}
6382 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6383 This ensures that you have no unresolved references to internal GCC
6384 library subroutines. (For example, @samp{__main}, used to ensure C++
6385 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6386 GNU Compiler Collection (GCC) Internals}.)
6390 Produce a position independent executable on targets which support it.
6391 For predictable results, you must also specify the same set of options
6392 that were used to generate code (@option{-fpie}, @option{-fPIE},
6393 or model suboptions) when you specify this option.
6397 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
6398 that support it. This instructs the linker to add all symbols, not
6399 only used ones, to the dynamic symbol table. This option is needed
6400 for some uses of @code{dlopen} or to allow obtaining backtraces
6401 from within a program.
6405 Remove all symbol table and relocation information from the executable.
6409 On systems that support dynamic linking, this prevents linking with the shared
6410 libraries. On other systems, this option has no effect.
6414 Produce a shared object which can then be linked with other objects to
6415 form an executable. Not all systems support this option. For predictable
6416 results, you must also specify the same set of options that were used to
6417 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6418 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6419 needs to build supplementary stub code for constructors to work. On
6420 multi-libbed systems, @samp{gcc -shared} must select the correct support
6421 libraries to link against. Failing to supply the correct flags may lead
6422 to subtle defects. Supplying them in cases where they are not necessary
6425 @item -shared-libgcc
6426 @itemx -static-libgcc
6427 @opindex shared-libgcc
6428 @opindex static-libgcc
6429 On systems that provide @file{libgcc} as a shared library, these options
6430 force the use of either the shared or static version respectively.
6431 If no shared version of @file{libgcc} was built when the compiler was
6432 configured, these options have no effect.
6434 There are several situations in which an application should use the
6435 shared @file{libgcc} instead of the static version. The most common
6436 of these is when the application wishes to throw and catch exceptions
6437 across different shared libraries. In that case, each of the libraries
6438 as well as the application itself should use the shared @file{libgcc}.
6440 Therefore, the G++ and GCJ drivers automatically add
6441 @option{-shared-libgcc} whenever you build a shared library or a main
6442 executable, because C++ and Java programs typically use exceptions, so
6443 this is the right thing to do.
6445 If, instead, you use the GCC driver to create shared libraries, you may
6446 find that they will not always be linked with the shared @file{libgcc}.
6447 If GCC finds, at its configuration time, that you have a non-GNU linker
6448 or a GNU linker that does not support option @option{--eh-frame-hdr},
6449 it will link the shared version of @file{libgcc} into shared libraries
6450 by default. Otherwise, it will take advantage of the linker and optimize
6451 away the linking with the shared version of @file{libgcc}, linking with
6452 the static version of libgcc by default. This allows exceptions to
6453 propagate through such shared libraries, without incurring relocation
6454 costs at library load time.
6456 However, if a library or main executable is supposed to throw or catch
6457 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6458 for the languages used in the program, or using the option
6459 @option{-shared-libgcc}, such that it is linked with the shared
6464 Bind references to global symbols when building a shared object. Warn
6465 about any unresolved references (unless overridden by the link editor
6466 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6469 @item -Xlinker @var{option}
6471 Pass @var{option} as an option to the linker. You can use this to
6472 supply system-specific linker options which GCC does not know how to
6475 If you want to pass an option that takes an argument, you must use
6476 @option{-Xlinker} twice, once for the option and once for the argument.
6477 For example, to pass @option{-assert definitions}, you must write
6478 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6479 @option{-Xlinker "-assert definitions"}, because this passes the entire
6480 string as a single argument, which is not what the linker expects.
6482 @item -Wl,@var{option}
6484 Pass @var{option} as an option to the linker. If @var{option} contains
6485 commas, it is split into multiple options at the commas.
6487 @item -u @var{symbol}
6489 Pretend the symbol @var{symbol} is undefined, to force linking of
6490 library modules to define it. You can use @option{-u} multiple times with
6491 different symbols to force loading of additional library modules.
6494 @node Directory Options
6495 @section Options for Directory Search
6496 @cindex directory options
6497 @cindex options, directory search
6500 These options specify directories to search for header files, for
6501 libraries and for parts of the compiler:
6506 Add the directory @var{dir} to the head of the list of directories to be
6507 searched for header files. This can be used to override a system header
6508 file, substituting your own version, since these directories are
6509 searched before the system header file directories. However, you should
6510 not use this option to add directories that contain vendor-supplied
6511 system header files (use @option{-isystem} for that). If you use more than
6512 one @option{-I} option, the directories are scanned in left-to-right
6513 order; the standard system directories come after.
6515 If a standard system include directory, or a directory specified with
6516 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6517 option will be ignored. The directory will still be searched but as a
6518 system directory at its normal position in the system include chain.
6519 This is to ensure that GCC's procedure to fix buggy system headers and
6520 the ordering for the include_next directive are not inadvertently changed.
6521 If you really need to change the search order for system directories,
6522 use the @option{-nostdinc} and/or @option{-isystem} options.
6524 @item -iquote@var{dir}
6526 Add the directory @var{dir} to the head of the list of directories to
6527 be searched for header files only for the case of @samp{#include
6528 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6529 otherwise just like @option{-I}.
6533 Add directory @var{dir} to the list of directories to be searched
6536 @item -B@var{prefix}
6538 This option specifies where to find the executables, libraries,
6539 include files, and data files of the compiler itself.
6541 The compiler driver program runs one or more of the subprograms
6542 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6543 @var{prefix} as a prefix for each program it tries to run, both with and
6544 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6546 For each subprogram to be run, the compiler driver first tries the
6547 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6548 was not specified, the driver tries two standard prefixes, which are
6549 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6550 those results in a file name that is found, the unmodified program
6551 name is searched for using the directories specified in your
6552 @env{PATH} environment variable.
6554 The compiler will check to see if the path provided by the @option{-B}
6555 refers to a directory, and if necessary it will add a directory
6556 separator character at the end of the path.
6558 @option{-B} prefixes that effectively specify directory names also apply
6559 to libraries in the linker, because the compiler translates these
6560 options into @option{-L} options for the linker. They also apply to
6561 includes files in the preprocessor, because the compiler translates these
6562 options into @option{-isystem} options for the preprocessor. In this case,
6563 the compiler appends @samp{include} to the prefix.
6565 The run-time support file @file{libgcc.a} can also be searched for using
6566 the @option{-B} prefix, if needed. If it is not found there, the two
6567 standard prefixes above are tried, and that is all. The file is left
6568 out of the link if it is not found by those means.
6570 Another way to specify a prefix much like the @option{-B} prefix is to use
6571 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6574 As a special kludge, if the path provided by @option{-B} is
6575 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6576 9, then it will be replaced by @file{[dir/]include}. This is to help
6577 with boot-strapping the compiler.
6579 @item -specs=@var{file}
6581 Process @var{file} after the compiler reads in the standard @file{specs}
6582 file, in order to override the defaults that the @file{gcc} driver
6583 program uses when determining what switches to pass to @file{cc1},
6584 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6585 @option{-specs=@var{file}} can be specified on the command line, and they
6586 are processed in order, from left to right.
6588 @item --sysroot=@var{dir}
6590 Use @var{dir} as the logical root directory for headers and libraries.
6591 For example, if the compiler would normally search for headers in
6592 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
6593 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
6595 If you use both this option and the @option{-isysroot} option, then
6596 the @option{--sysroot} option will apply to libraries, but the
6597 @option{-isysroot} option will apply to header files.
6599 The GNU linker (beginning with version 2.16) has the necessary support
6600 for this option. If your linker does not support this option, the
6601 header file aspect of @option{--sysroot} will still work, but the
6602 library aspect will not.
6606 This option has been deprecated. Please use @option{-iquote} instead for
6607 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6608 Any directories you specify with @option{-I} options before the @option{-I-}
6609 option are searched only for the case of @samp{#include "@var{file}"};
6610 they are not searched for @samp{#include <@var{file}>}.
6612 If additional directories are specified with @option{-I} options after
6613 the @option{-I-}, these directories are searched for all @samp{#include}
6614 directives. (Ordinarily @emph{all} @option{-I} directories are used
6617 In addition, the @option{-I-} option inhibits the use of the current
6618 directory (where the current input file came from) as the first search
6619 directory for @samp{#include "@var{file}"}. There is no way to
6620 override this effect of @option{-I-}. With @option{-I.} you can specify
6621 searching the directory which was current when the compiler was
6622 invoked. That is not exactly the same as what the preprocessor does
6623 by default, but it is often satisfactory.
6625 @option{-I-} does not inhibit the use of the standard system directories
6626 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6633 @section Specifying subprocesses and the switches to pass to them
6636 @command{gcc} is a driver program. It performs its job by invoking a
6637 sequence of other programs to do the work of compiling, assembling and
6638 linking. GCC interprets its command-line parameters and uses these to
6639 deduce which programs it should invoke, and which command-line options
6640 it ought to place on their command lines. This behavior is controlled
6641 by @dfn{spec strings}. In most cases there is one spec string for each
6642 program that GCC can invoke, but a few programs have multiple spec
6643 strings to control their behavior. The spec strings built into GCC can
6644 be overridden by using the @option{-specs=} command-line switch to specify
6647 @dfn{Spec files} are plaintext files that are used to construct spec
6648 strings. They consist of a sequence of directives separated by blank
6649 lines. The type of directive is determined by the first non-whitespace
6650 character on the line and it can be one of the following:
6653 @item %@var{command}
6654 Issues a @var{command} to the spec file processor. The commands that can
6658 @item %include <@var{file}>
6660 Search for @var{file} and insert its text at the current point in the
6663 @item %include_noerr <@var{file}>
6664 @cindex %include_noerr
6665 Just like @samp{%include}, but do not generate an error message if the include
6666 file cannot be found.
6668 @item %rename @var{old_name} @var{new_name}
6670 Rename the spec string @var{old_name} to @var{new_name}.
6674 @item *[@var{spec_name}]:
6675 This tells the compiler to create, override or delete the named spec
6676 string. All lines after this directive up to the next directive or
6677 blank line are considered to be the text for the spec string. If this
6678 results in an empty string then the spec will be deleted. (Or, if the
6679 spec did not exist, then nothing will happened.) Otherwise, if the spec
6680 does not currently exist a new spec will be created. If the spec does
6681 exist then its contents will be overridden by the text of this
6682 directive, unless the first character of that text is the @samp{+}
6683 character, in which case the text will be appended to the spec.
6685 @item [@var{suffix}]:
6686 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6687 and up to the next directive or blank line are considered to make up the
6688 spec string for the indicated suffix. When the compiler encounters an
6689 input file with the named suffix, it will processes the spec string in
6690 order to work out how to compile that file. For example:
6697 This says that any input file whose name ends in @samp{.ZZ} should be
6698 passed to the program @samp{z-compile}, which should be invoked with the
6699 command-line switch @option{-input} and with the result of performing the
6700 @samp{%i} substitution. (See below.)
6702 As an alternative to providing a spec string, the text that follows a
6703 suffix directive can be one of the following:
6706 @item @@@var{language}
6707 This says that the suffix is an alias for a known @var{language}. This is
6708 similar to using the @option{-x} command-line switch to GCC to specify a
6709 language explicitly. For example:
6716 Says that .ZZ files are, in fact, C++ source files.
6719 This causes an error messages saying:
6722 @var{name} compiler not installed on this system.
6726 GCC already has an extensive list of suffixes built into it.
6727 This directive will add an entry to the end of the list of suffixes, but
6728 since the list is searched from the end backwards, it is effectively
6729 possible to override earlier entries using this technique.
6733 GCC has the following spec strings built into it. Spec files can
6734 override these strings or create their own. Note that individual
6735 targets can also add their own spec strings to this list.
6738 asm Options to pass to the assembler
6739 asm_final Options to pass to the assembler post-processor
6740 cpp Options to pass to the C preprocessor
6741 cc1 Options to pass to the C compiler
6742 cc1plus Options to pass to the C++ compiler
6743 endfile Object files to include at the end of the link
6744 link Options to pass to the linker
6745 lib Libraries to include on the command line to the linker
6746 libgcc Decides which GCC support library to pass to the linker
6747 linker Sets the name of the linker
6748 predefines Defines to be passed to the C preprocessor
6749 signed_char Defines to pass to CPP to say whether @code{char} is signed
6751 startfile Object files to include at the start of the link
6754 Here is a small example of a spec file:
6760 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6763 This example renames the spec called @samp{lib} to @samp{old_lib} and
6764 then overrides the previous definition of @samp{lib} with a new one.
6765 The new definition adds in some extra command-line options before
6766 including the text of the old definition.
6768 @dfn{Spec strings} are a list of command-line options to be passed to their
6769 corresponding program. In addition, the spec strings can contain
6770 @samp{%}-prefixed sequences to substitute variable text or to
6771 conditionally insert text into the command line. Using these constructs
6772 it is possible to generate quite complex command lines.
6774 Here is a table of all defined @samp{%}-sequences for spec
6775 strings. Note that spaces are not generated automatically around the
6776 results of expanding these sequences. Therefore you can concatenate them
6777 together or combine them with constant text in a single argument.
6781 Substitute one @samp{%} into the program name or argument.
6784 Substitute the name of the input file being processed.
6787 Substitute the basename of the input file being processed.
6788 This is the substring up to (and not including) the last period
6789 and not including the directory.
6792 This is the same as @samp{%b}, but include the file suffix (text after
6796 Marks the argument containing or following the @samp{%d} as a
6797 temporary file name, so that that file will be deleted if GCC exits
6798 successfully. Unlike @samp{%g}, this contributes no text to the
6801 @item %g@var{suffix}
6802 Substitute a file name that has suffix @var{suffix} and is chosen
6803 once per compilation, and mark the argument in the same way as
6804 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6805 name is now chosen in a way that is hard to predict even when previously
6806 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6807 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6808 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6809 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6810 was simply substituted with a file name chosen once per compilation,
6811 without regard to any appended suffix (which was therefore treated
6812 just like ordinary text), making such attacks more likely to succeed.
6814 @item %u@var{suffix}
6815 Like @samp{%g}, but generates a new temporary file name even if
6816 @samp{%u@var{suffix}} was already seen.
6818 @item %U@var{suffix}
6819 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6820 new one if there is no such last file name. In the absence of any
6821 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6822 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6823 would involve the generation of two distinct file names, one
6824 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6825 simply substituted with a file name chosen for the previous @samp{%u},
6826 without regard to any appended suffix.
6828 @item %j@var{suffix}
6829 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6830 writable, and if save-temps is off; otherwise, substitute the name
6831 of a temporary file, just like @samp{%u}. This temporary file is not
6832 meant for communication between processes, but rather as a junk
6835 @item %|@var{suffix}
6836 @itemx %m@var{suffix}
6837 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6838 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6839 all. These are the two most common ways to instruct a program that it
6840 should read from standard input or write to standard output. If you
6841 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6842 construct: see for example @file{f/lang-specs.h}.
6844 @item %.@var{SUFFIX}
6845 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6846 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6847 terminated by the next space or %.
6850 Marks the argument containing or following the @samp{%w} as the
6851 designated output file of this compilation. This puts the argument
6852 into the sequence of arguments that @samp{%o} will substitute later.
6855 Substitutes the names of all the output files, with spaces
6856 automatically placed around them. You should write spaces
6857 around the @samp{%o} as well or the results are undefined.
6858 @samp{%o} is for use in the specs for running the linker.
6859 Input files whose names have no recognized suffix are not compiled
6860 at all, but they are included among the output files, so they will
6864 Substitutes the suffix for object files. Note that this is
6865 handled specially when it immediately follows @samp{%g, %u, or %U},
6866 because of the need for those to form complete file names. The
6867 handling is such that @samp{%O} is treated exactly as if it had already
6868 been substituted, except that @samp{%g, %u, and %U} do not currently
6869 support additional @var{suffix} characters following @samp{%O} as they would
6870 following, for example, @samp{.o}.
6873 Substitutes the standard macro predefinitions for the
6874 current target machine. Use this when running @code{cpp}.
6877 Like @samp{%p}, but puts @samp{__} before and after the name of each
6878 predefined macro, except for macros that start with @samp{__} or with
6879 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6883 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6884 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
6885 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6886 and @option{-imultilib} as necessary.
6889 Current argument is the name of a library or startup file of some sort.
6890 Search for that file in a standard list of directories and substitute
6891 the full name found.
6894 Print @var{str} as an error message. @var{str} is terminated by a newline.
6895 Use this when inconsistent options are detected.
6898 Substitute the contents of spec string @var{name} at this point.
6901 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6903 @item %x@{@var{option}@}
6904 Accumulate an option for @samp{%X}.
6907 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6911 Output the accumulated assembler options specified by @option{-Wa}.
6914 Output the accumulated preprocessor options specified by @option{-Wp}.
6917 Process the @code{asm} spec. This is used to compute the
6918 switches to be passed to the assembler.
6921 Process the @code{asm_final} spec. This is a spec string for
6922 passing switches to an assembler post-processor, if such a program is
6926 Process the @code{link} spec. This is the spec for computing the
6927 command line passed to the linker. Typically it will make use of the
6928 @samp{%L %G %S %D and %E} sequences.
6931 Dump out a @option{-L} option for each directory that GCC believes might
6932 contain startup files. If the target supports multilibs then the
6933 current multilib directory will be prepended to each of these paths.
6936 Process the @code{lib} spec. This is a spec string for deciding which
6937 libraries should be included on the command line to the linker.
6940 Process the @code{libgcc} spec. This is a spec string for deciding
6941 which GCC support library should be included on the command line to the linker.
6944 Process the @code{startfile} spec. This is a spec for deciding which
6945 object files should be the first ones passed to the linker. Typically
6946 this might be a file named @file{crt0.o}.
6949 Process the @code{endfile} spec. This is a spec string that specifies
6950 the last object files that will be passed to the linker.
6953 Process the @code{cpp} spec. This is used to construct the arguments
6954 to be passed to the C preprocessor.
6957 Process the @code{cc1} spec. This is used to construct the options to be
6958 passed to the actual C compiler (@samp{cc1}).
6961 Process the @code{cc1plus} spec. This is used to construct the options to be
6962 passed to the actual C++ compiler (@samp{cc1plus}).
6965 Substitute the variable part of a matched option. See below.
6966 Note that each comma in the substituted string is replaced by
6970 Remove all occurrences of @code{-S} from the command line. Note---this
6971 command is position dependent. @samp{%} commands in the spec string
6972 before this one will see @code{-S}, @samp{%} commands in the spec string
6973 after this one will not.
6975 @item %:@var{function}(@var{args})
6976 Call the named function @var{function}, passing it @var{args}.
6977 @var{args} is first processed as a nested spec string, then split
6978 into an argument vector in the usual fashion. The function returns
6979 a string which is processed as if it had appeared literally as part
6980 of the current spec.
6982 The following built-in spec functions are provided:
6985 @item @code{if-exists}
6986 The @code{if-exists} spec function takes one argument, an absolute
6987 pathname to a file. If the file exists, @code{if-exists} returns the
6988 pathname. Here is a small example of its usage:
6992 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6995 @item @code{if-exists-else}
6996 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6997 spec function, except that it takes two arguments. The first argument is
6998 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6999 returns the pathname. If it does not exist, it returns the second argument.
7000 This way, @code{if-exists-else} can be used to select one file or another,
7001 based on the existence of the first. Here is a small example of its usage:
7005 crt0%O%s %:if-exists(crti%O%s) \
7006 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
7009 @item @code{replace-outfile}
7010 The @code{replace-outfile} spec function takes two arguments. It looks for the
7011 first argument in the outfiles array and replaces it with the second argument. Here
7012 is a small example of its usage:
7015 %@{static|static-libgcc|static-libstdc++:%:replace-outfile(-lstdc++ \
7023 Substitutes the @code{-S} switch, if that switch was given to GCC@.
7024 If that switch was not specified, this substitutes nothing. Note that
7025 the leading dash is omitted when specifying this option, and it is
7026 automatically inserted if the substitution is performed. Thus the spec
7027 string @samp{%@{foo@}} would match the command-line option @option{-foo}
7028 and would output the command line option @option{-foo}.
7030 @item %W@{@code{S}@}
7031 Like %@{@code{S}@} but mark last argument supplied within as a file to be
7034 @item %@{@code{S}*@}
7035 Substitutes all the switches specified to GCC whose names start
7036 with @code{-S}, but which also take an argument. This is used for
7037 switches like @option{-o}, @option{-D}, @option{-I}, etc.
7038 GCC considers @option{-o foo} as being
7039 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
7040 text, including the space. Thus two arguments would be generated.
7042 @item %@{@code{S}*&@code{T}*@}
7043 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
7044 (the order of @code{S} and @code{T} in the spec is not significant).
7045 There can be any number of ampersand-separated variables; for each the
7046 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
7048 @item %@{@code{S}:@code{X}@}
7049 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
7051 @item %@{!@code{S}:@code{X}@}
7052 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
7054 @item %@{@code{S}*:@code{X}@}
7055 Substitutes @code{X} if one or more switches whose names start with
7056 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
7057 once, no matter how many such switches appeared. However, if @code{%*}
7058 appears somewhere in @code{X}, then @code{X} will be substituted once
7059 for each matching switch, with the @code{%*} replaced by the part of
7060 that switch that matched the @code{*}.
7062 @item %@{.@code{S}:@code{X}@}
7063 Substitutes @code{X}, if processing a file with suffix @code{S}.
7065 @item %@{!.@code{S}:@code{X}@}
7066 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
7068 @item %@{@code{S}|@code{P}:@code{X}@}
7069 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
7070 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
7071 although they have a stronger binding than the @samp{|}. If @code{%*}
7072 appears in @code{X}, all of the alternatives must be starred, and only
7073 the first matching alternative is substituted.
7075 For example, a spec string like this:
7078 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
7081 will output the following command-line options from the following input
7082 command-line options:
7087 -d fred.c -foo -baz -boggle
7088 -d jim.d -bar -baz -boggle
7091 @item %@{S:X; T:Y; :D@}
7093 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
7094 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
7095 be as many clauses as you need. This may be combined with @code{.},
7096 @code{!}, @code{|}, and @code{*} as needed.
7101 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
7102 construct may contain other nested @samp{%} constructs or spaces, or
7103 even newlines. They are processed as usual, as described above.
7104 Trailing white space in @code{X} is ignored. White space may also
7105 appear anywhere on the left side of the colon in these constructs,
7106 except between @code{.} or @code{*} and the corresponding word.
7108 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
7109 handled specifically in these constructs. If another value of
7110 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
7111 @option{-W} switch is found later in the command line, the earlier
7112 switch value is ignored, except with @{@code{S}*@} where @code{S} is
7113 just one letter, which passes all matching options.
7115 The character @samp{|} at the beginning of the predicate text is used to
7116 indicate that a command should be piped to the following command, but
7117 only if @option{-pipe} is specified.
7119 It is built into GCC which switches take arguments and which do not.
7120 (You might think it would be useful to generalize this to allow each
7121 compiler's spec to say which switches take arguments. But this cannot
7122 be done in a consistent fashion. GCC cannot even decide which input
7123 files have been specified without knowing which switches take arguments,
7124 and it must know which input files to compile in order to tell which
7127 GCC also knows implicitly that arguments starting in @option{-l} are to be
7128 treated as compiler output files, and passed to the linker in their
7129 proper position among the other output files.
7131 @c man begin OPTIONS
7133 @node Target Options
7134 @section Specifying Target Machine and Compiler Version
7135 @cindex target options
7136 @cindex cross compiling
7137 @cindex specifying machine version
7138 @cindex specifying compiler version and target machine
7139 @cindex compiler version, specifying
7140 @cindex target machine, specifying
7142 The usual way to run GCC is to run the executable called @file{gcc}, or
7143 @file{<machine>-gcc} when cross-compiling, or
7144 @file{<machine>-gcc-<version>} to run a version other than the one that
7145 was installed last. Sometimes this is inconvenient, so GCC provides
7146 options that will switch to another cross-compiler or version.
7149 @item -b @var{machine}
7151 The argument @var{machine} specifies the target machine for compilation.
7153 The value to use for @var{machine} is the same as was specified as the
7154 machine type when configuring GCC as a cross-compiler. For
7155 example, if a cross-compiler was configured with @samp{configure
7156 arm-elf}, meaning to compile for an arm processor with elf binaries,
7157 then you would specify @option{-b arm-elf} to run that cross compiler.
7158 Because there are other options beginning with @option{-b}, the
7159 configuration must contain a hyphen.
7161 @item -V @var{version}
7163 The argument @var{version} specifies which version of GCC to run.
7164 This is useful when multiple versions are installed. For example,
7165 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
7168 The @option{-V} and @option{-b} options work by running the
7169 @file{<machine>-gcc-<version>} executable, so there's no real reason to
7170 use them if you can just run that directly.
7172 @node Submodel Options
7173 @section Hardware Models and Configurations
7174 @cindex submodel options
7175 @cindex specifying hardware config
7176 @cindex hardware models and configurations, specifying
7177 @cindex machine dependent options
7179 Earlier we discussed the standard option @option{-b} which chooses among
7180 different installed compilers for completely different target
7181 machines, such as VAX vs.@: 68000 vs.@: 80386.
7183 In addition, each of these target machine types can have its own
7184 special options, starting with @samp{-m}, to choose among various
7185 hardware models or configurations---for example, 68010 vs 68020,
7186 floating coprocessor or none. A single installed version of the
7187 compiler can compile for any model or configuration, according to the
7190 Some configurations of the compiler also support additional special
7191 options, usually for compatibility with other compilers on the same
7194 @c This list is ordered alphanumerically by subsection name.
7195 @c It should be the same order and spelling as these options are listed
7196 @c in Machine Dependent Options
7202 * Blackfin Options::
7206 * DEC Alpha Options::
7207 * DEC Alpha/VMS Options::
7209 * GNU/Linux Options::
7212 * i386 and x86-64 Options::
7225 * RS/6000 and PowerPC Options::
7226 * S/390 and zSeries Options::
7230 * System V Options::
7231 * TMS320C3x/C4x Options::
7235 * Xstormy16 Options::
7241 @subsection ARC Options
7244 These options are defined for ARC implementations:
7249 Compile code for little endian mode. This is the default.
7253 Compile code for big endian mode.
7256 @opindex mmangle-cpu
7257 Prepend the name of the cpu to all public symbol names.
7258 In multiple-processor systems, there are many ARC variants with different
7259 instruction and register set characteristics. This flag prevents code
7260 compiled for one cpu to be linked with code compiled for another.
7261 No facility exists for handling variants that are ``almost identical''.
7262 This is an all or nothing option.
7264 @item -mcpu=@var{cpu}
7266 Compile code for ARC variant @var{cpu}.
7267 Which variants are supported depend on the configuration.
7268 All variants support @option{-mcpu=base}, this is the default.
7270 @item -mtext=@var{text-section}
7271 @itemx -mdata=@var{data-section}
7272 @itemx -mrodata=@var{readonly-data-section}
7276 Put functions, data, and readonly data in @var{text-section},
7277 @var{data-section}, and @var{readonly-data-section} respectively
7278 by default. This can be overridden with the @code{section} attribute.
7279 @xref{Variable Attributes}.
7284 @subsection ARM Options
7287 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7291 @item -mabi=@var{name}
7293 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7294 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
7297 @opindex mapcs-frame
7298 Generate a stack frame that is compliant with the ARM Procedure Call
7299 Standard for all functions, even if this is not strictly necessary for
7300 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7301 with this option will cause the stack frames not to be generated for
7302 leaf functions. The default is @option{-mno-apcs-frame}.
7306 This is a synonym for @option{-mapcs-frame}.
7309 @c not currently implemented
7310 @item -mapcs-stack-check
7311 @opindex mapcs-stack-check
7312 Generate code to check the amount of stack space available upon entry to
7313 every function (that actually uses some stack space). If there is
7314 insufficient space available then either the function
7315 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7316 called, depending upon the amount of stack space required. The run time
7317 system is required to provide these functions. The default is
7318 @option{-mno-apcs-stack-check}, since this produces smaller code.
7320 @c not currently implemented
7322 @opindex mapcs-float
7323 Pass floating point arguments using the float point registers. This is
7324 one of the variants of the APCS@. This option is recommended if the
7325 target hardware has a floating point unit or if a lot of floating point
7326 arithmetic is going to be performed by the code. The default is
7327 @option{-mno-apcs-float}, since integer only code is slightly increased in
7328 size if @option{-mapcs-float} is used.
7330 @c not currently implemented
7331 @item -mapcs-reentrant
7332 @opindex mapcs-reentrant
7333 Generate reentrant, position independent code. The default is
7334 @option{-mno-apcs-reentrant}.
7337 @item -mthumb-interwork
7338 @opindex mthumb-interwork
7339 Generate code which supports calling between the ARM and Thumb
7340 instruction sets. Without this option the two instruction sets cannot
7341 be reliably used inside one program. The default is
7342 @option{-mno-thumb-interwork}, since slightly larger code is generated
7343 when @option{-mthumb-interwork} is specified.
7345 @item -mno-sched-prolog
7346 @opindex mno-sched-prolog
7347 Prevent the reordering of instructions in the function prolog, or the
7348 merging of those instruction with the instructions in the function's
7349 body. This means that all functions will start with a recognizable set
7350 of instructions (or in fact one of a choice from a small set of
7351 different function prologues), and this information can be used to
7352 locate the start if functions inside an executable piece of code. The
7353 default is @option{-msched-prolog}.
7356 @opindex mhard-float
7357 Generate output containing floating point instructions. This is the
7361 @opindex msoft-float
7362 Generate output containing library calls for floating point.
7363 @strong{Warning:} the requisite libraries are not available for all ARM
7364 targets. Normally the facilities of the machine's usual C compiler are
7365 used, but this cannot be done directly in cross-compilation. You must make
7366 your own arrangements to provide suitable library functions for
7369 @option{-msoft-float} changes the calling convention in the output file;
7370 therefore, it is only useful if you compile @emph{all} of a program with
7371 this option. In particular, you need to compile @file{libgcc.a}, the
7372 library that comes with GCC, with @option{-msoft-float} in order for
7375 @item -mfloat-abi=@var{name}
7377 Specifies which ABI to use for floating point values. Permissible values
7378 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7380 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7381 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7382 of floating point instructions, but still uses the soft-float calling
7385 @item -mlittle-endian
7386 @opindex mlittle-endian
7387 Generate code for a processor running in little-endian mode. This is
7388 the default for all standard configurations.
7391 @opindex mbig-endian
7392 Generate code for a processor running in big-endian mode; the default is
7393 to compile code for a little-endian processor.
7395 @item -mwords-little-endian
7396 @opindex mwords-little-endian
7397 This option only applies when generating code for big-endian processors.
7398 Generate code for a little-endian word order but a big-endian byte
7399 order. That is, a byte order of the form @samp{32107654}. Note: this
7400 option should only be used if you require compatibility with code for
7401 big-endian ARM processors generated by versions of the compiler prior to
7404 @item -mcpu=@var{name}
7406 This specifies the name of the target ARM processor. GCC uses this name
7407 to determine what kind of instructions it can emit when generating
7408 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7409 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7410 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7411 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7412 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7413 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7414 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7415 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7416 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7417 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7418 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7419 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7420 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7421 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
7424 @itemx -mtune=@var{name}
7426 This option is very similar to the @option{-mcpu=} option, except that
7427 instead of specifying the actual target processor type, and hence
7428 restricting which instructions can be used, it specifies that GCC should
7429 tune the performance of the code as if the target were of the type
7430 specified in this option, but still choosing the instructions that it
7431 will generate based on the cpu specified by a @option{-mcpu=} option.
7432 For some ARM implementations better performance can be obtained by using
7435 @item -march=@var{name}
7437 This specifies the name of the target ARM architecture. GCC uses this
7438 name to determine what kind of instructions it can emit when generating
7439 assembly code. This option can be used in conjunction with or instead
7440 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7441 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7442 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7443 @samp{iwmmxt}, @samp{ep9312}.
7445 @item -mfpu=@var{name}
7446 @itemx -mfpe=@var{number}
7447 @itemx -mfp=@var{number}
7451 This specifies what floating point hardware (or hardware emulation) is
7452 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7453 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7454 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7455 with older versions of GCC@.
7457 If @option{-msoft-float} is specified this specifies the format of
7458 floating point values.
7460 @item -mstructure-size-boundary=@var{n}
7461 @opindex mstructure-size-boundary
7462 The size of all structures and unions will be rounded up to a multiple
7463 of the number of bits set by this option. Permissible values are 8, 32
7464 and 64. The default value varies for different toolchains. For the COFF
7465 targeted toolchain the default value is 8. A value of 64 is only allowed
7466 if the underlying ABI supports it.
7468 Specifying the larger number can produce faster, more efficient code, but
7469 can also increase the size of the program. Different values are potentially
7470 incompatible. Code compiled with one value cannot necessarily expect to
7471 work with code or libraries compiled with another value, if they exchange
7472 information using structures or unions.
7474 @item -mabort-on-noreturn
7475 @opindex mabort-on-noreturn
7476 Generate a call to the function @code{abort} at the end of a
7477 @code{noreturn} function. It will be executed if the function tries to
7481 @itemx -mno-long-calls
7482 @opindex mlong-calls
7483 @opindex mno-long-calls
7484 Tells the compiler to perform function calls by first loading the
7485 address of the function into a register and then performing a subroutine
7486 call on this register. This switch is needed if the target function
7487 will lie outside of the 64 megabyte addressing range of the offset based
7488 version of subroutine call instruction.
7490 Even if this switch is enabled, not all function calls will be turned
7491 into long calls. The heuristic is that static functions, functions
7492 which have the @samp{short-call} attribute, functions that are inside
7493 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7494 definitions have already been compiled within the current compilation
7495 unit, will not be turned into long calls. The exception to this rule is
7496 that weak function definitions, functions with the @samp{long-call}
7497 attribute or the @samp{section} attribute, and functions that are within
7498 the scope of a @samp{#pragma long_calls} directive, will always be
7499 turned into long calls.
7501 This feature is not enabled by default. Specifying
7502 @option{-mno-long-calls} will restore the default behavior, as will
7503 placing the function calls within the scope of a @samp{#pragma
7504 long_calls_off} directive. Note these switches have no effect on how
7505 the compiler generates code to handle function calls via function
7508 @item -mnop-fun-dllimport
7509 @opindex mnop-fun-dllimport
7510 Disable support for the @code{dllimport} attribute.
7512 @item -msingle-pic-base
7513 @opindex msingle-pic-base
7514 Treat the register used for PIC addressing as read-only, rather than
7515 loading it in the prologue for each function. The run-time system is
7516 responsible for initializing this register with an appropriate value
7517 before execution begins.
7519 @item -mpic-register=@var{reg}
7520 @opindex mpic-register
7521 Specify the register to be used for PIC addressing. The default is R10
7522 unless stack-checking is enabled, when R9 is used.
7524 @item -mcirrus-fix-invalid-insns
7525 @opindex mcirrus-fix-invalid-insns
7526 @opindex mno-cirrus-fix-invalid-insns
7527 Insert NOPs into the instruction stream to in order to work around
7528 problems with invalid Maverick instruction combinations. This option
7529 is only valid if the @option{-mcpu=ep9312} option has been used to
7530 enable generation of instructions for the Cirrus Maverick floating
7531 point co-processor. This option is not enabled by default, since the
7532 problem is only present in older Maverick implementations. The default
7533 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7536 @item -mpoke-function-name
7537 @opindex mpoke-function-name
7538 Write the name of each function into the text section, directly
7539 preceding the function prologue. The generated code is similar to this:
7543 .ascii "arm_poke_function_name", 0
7546 .word 0xff000000 + (t1 - t0)
7547 arm_poke_function_name
7549 stmfd sp!, @{fp, ip, lr, pc@}
7553 When performing a stack backtrace, code can inspect the value of
7554 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7555 location @code{pc - 12} and the top 8 bits are set, then we know that
7556 there is a function name embedded immediately preceding this location
7557 and has length @code{((pc[-3]) & 0xff000000)}.
7561 Generate code for the 16-bit Thumb instruction set. The default is to
7562 use the 32-bit ARM instruction set.
7565 @opindex mtpcs-frame
7566 Generate a stack frame that is compliant with the Thumb Procedure Call
7567 Standard for all non-leaf functions. (A leaf function is one that does
7568 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7570 @item -mtpcs-leaf-frame
7571 @opindex mtpcs-leaf-frame
7572 Generate a stack frame that is compliant with the Thumb Procedure Call
7573 Standard for all leaf functions. (A leaf function is one that does
7574 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7576 @item -mcallee-super-interworking
7577 @opindex mcallee-super-interworking
7578 Gives all externally visible functions in the file being compiled an ARM
7579 instruction set header which switches to Thumb mode before executing the
7580 rest of the function. This allows these functions to be called from
7581 non-interworking code.
7583 @item -mcaller-super-interworking
7584 @opindex mcaller-super-interworking
7585 Allows calls via function pointers (including virtual functions) to
7586 execute correctly regardless of whether the target code has been
7587 compiled for interworking or not. There is a small overhead in the cost
7588 of executing a function pointer if this option is enabled.
7590 @item -mtp=@var{name}
7592 Specify the access model for the thread local storage pointer. The valid
7593 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
7594 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
7595 (supported in the arm6k architecture), and @option{auto}, which uses the
7596 best available method for the selected processor. The default setting is
7602 @subsection AVR Options
7605 These options are defined for AVR implementations:
7608 @item -mmcu=@var{mcu}
7610 Specify ATMEL AVR instruction set or MCU type.
7612 Instruction set avr1 is for the minimal AVR core, not supported by the C
7613 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7614 attiny11, attiny12, attiny15, attiny28).
7616 Instruction set avr2 (default) is for the classic AVR core with up to
7617 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7618 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7619 at90c8534, at90s8535).
7621 Instruction set avr3 is for the classic AVR core with up to 128K program
7622 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7624 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7625 memory space (MCU types: atmega8, atmega83, atmega85).
7627 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7628 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7629 atmega64, atmega128, at43usb355, at94k).
7633 Output instruction sizes to the asm file.
7635 @item -minit-stack=@var{N}
7636 @opindex minit-stack
7637 Specify the initial stack address, which may be a symbol or numeric value,
7638 @samp{__stack} is the default.
7640 @item -mno-interrupts
7641 @opindex mno-interrupts
7642 Generated code is not compatible with hardware interrupts.
7643 Code size will be smaller.
7645 @item -mcall-prologues
7646 @opindex mcall-prologues
7647 Functions prologues/epilogues expanded as call to appropriate
7648 subroutines. Code size will be smaller.
7650 @item -mno-tablejump
7651 @opindex mno-tablejump
7652 Do not generate tablejump insns which sometimes increase code size.
7655 @opindex mtiny-stack
7656 Change only the low 8 bits of the stack pointer.
7660 Assume int to be 8 bit integer. This affects the sizes of all types: A
7661 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7662 and long long will be 4 bytes. Please note that this option does not
7663 comply to the C standards, but it will provide you with smaller code
7667 @node Blackfin Options
7668 @subsection Blackfin Options
7669 @cindex Blackfin Options
7672 @item -momit-leaf-frame-pointer
7673 @opindex momit-leaf-frame-pointer
7674 Don't keep the frame pointer in a register for leaf functions. This
7675 avoids the instructions to save, set up and restore frame pointers and
7676 makes an extra register available in leaf functions. The option
7677 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7678 which might make debugging harder.
7680 @item -mspecld-anomaly
7681 @opindex mspecld-anomaly
7682 When enabled, the compiler will ensure that the generated code does not
7683 contain speculative loads after jump instructions. This option is enabled
7686 @item -mno-specld-anomaly
7687 @opindex mno-specld-anomaly
7688 Don't generate extra code to prevent speculative loads from occurring.
7690 @item -mcsync-anomaly
7691 @opindex mcsync-anomaly
7692 When enabled, the compiler will ensure that the generated code does not
7693 contain CSYNC or SSYNC instructions too soon after conditional branches.
7694 This option is enabled by default.
7696 @item -mno-csync-anomaly
7697 @opindex mno-csync-anomaly
7698 Don't generate extra code to prevent CSYNC or SSYNC instructions from
7699 occurring too soon after a conditional branch.
7703 When enabled, the compiler is free to take advantage of the knowledge that
7704 the entire program fits into the low 64k of memory.
7707 @opindex mno-low-64k
7708 Assume that the program is arbitrarily large. This is the default.
7710 @item -mid-shared-library
7711 @opindex mid-shared-library
7712 Generate code that supports shared libraries via the library ID method.
7713 This allows for execute in place and shared libraries in an environment
7714 without virtual memory management. This option implies @option{-fPIC}.
7716 @item -mno-id-shared-library
7717 @opindex mno-id-shared-library
7718 Generate code that doesn't assume ID based shared libraries are being used.
7719 This is the default.
7721 @item -mshared-library-id=n
7722 @opindex mshared-library-id
7723 Specified the identification number of the ID based shared library being
7724 compiled. Specifying a value of 0 will generate more compact code, specifying
7725 other values will force the allocation of that number to the current
7726 library but is no more space or time efficient than omitting this option.
7729 @itemx -mno-long-calls
7730 @opindex mlong-calls
7731 @opindex mno-long-calls
7732 Tells the compiler to perform function calls by first loading the
7733 address of the function into a register and then performing a subroutine
7734 call on this register. This switch is needed if the target function
7735 will lie outside of the 24 bit addressing range of the offset based
7736 version of subroutine call instruction.
7738 This feature is not enabled by default. Specifying
7739 @option{-mno-long-calls} will restore the default behavior. Note these
7740 switches have no effect on how the compiler generates code to handle
7741 function calls via function pointers.
7745 @subsection CRIS Options
7746 @cindex CRIS Options
7748 These options are defined specifically for the CRIS ports.
7751 @item -march=@var{architecture-type}
7752 @itemx -mcpu=@var{architecture-type}
7755 Generate code for the specified architecture. The choices for
7756 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7757 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7758 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7761 @item -mtune=@var{architecture-type}
7763 Tune to @var{architecture-type} everything applicable about the generated
7764 code, except for the ABI and the set of available instructions. The
7765 choices for @var{architecture-type} are the same as for
7766 @option{-march=@var{architecture-type}}.
7768 @item -mmax-stack-frame=@var{n}
7769 @opindex mmax-stack-frame
7770 Warn when the stack frame of a function exceeds @var{n} bytes.
7772 @item -melinux-stacksize=@var{n}
7773 @opindex melinux-stacksize
7774 Only available with the @samp{cris-axis-aout} target. Arranges for
7775 indications in the program to the kernel loader that the stack of the
7776 program should be set to @var{n} bytes.
7782 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7783 @option{-march=v3} and @option{-march=v8} respectively.
7785 @item -mmul-bug-workaround
7786 @itemx -mno-mul-bug-workaround
7787 @opindex mmul-bug-workaround
7788 @opindex mno-mul-bug-workaround
7789 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7790 models where it applies. This option is active by default.
7794 Enable CRIS-specific verbose debug-related information in the assembly
7795 code. This option also has the effect to turn off the @samp{#NO_APP}
7796 formatted-code indicator to the assembler at the beginning of the
7801 Do not use condition-code results from previous instruction; always emit
7802 compare and test instructions before use of condition codes.
7804 @item -mno-side-effects
7805 @opindex mno-side-effects
7806 Do not emit instructions with side-effects in addressing modes other than
7810 @itemx -mno-stack-align
7812 @itemx -mno-data-align
7813 @itemx -mconst-align
7814 @itemx -mno-const-align
7815 @opindex mstack-align
7816 @opindex mno-stack-align
7817 @opindex mdata-align
7818 @opindex mno-data-align
7819 @opindex mconst-align
7820 @opindex mno-const-align
7821 These options (no-options) arranges (eliminate arrangements) for the
7822 stack-frame, individual data and constants to be aligned for the maximum
7823 single data access size for the chosen CPU model. The default is to
7824 arrange for 32-bit alignment. ABI details such as structure layout are
7825 not affected by these options.
7833 Similar to the stack- data- and const-align options above, these options
7834 arrange for stack-frame, writable data and constants to all be 32-bit,
7835 16-bit or 8-bit aligned. The default is 32-bit alignment.
7837 @item -mno-prologue-epilogue
7838 @itemx -mprologue-epilogue
7839 @opindex mno-prologue-epilogue
7840 @opindex mprologue-epilogue
7841 With @option{-mno-prologue-epilogue}, the normal function prologue and
7842 epilogue that sets up the stack-frame are omitted and no return
7843 instructions or return sequences are generated in the code. Use this
7844 option only together with visual inspection of the compiled code: no
7845 warnings or errors are generated when call-saved registers must be saved,
7846 or storage for local variable needs to be allocated.
7852 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7853 instruction sequences that load addresses for functions from the PLT part
7854 of the GOT rather than (traditional on other architectures) calls to the
7855 PLT@. The default is @option{-mgotplt}.
7859 Legacy no-op option only recognized with the cris-axis-aout target.
7863 Legacy no-op option only recognized with the cris-axis-elf and
7864 cris-axis-linux-gnu targets.
7868 Only recognized with the cris-axis-aout target, where it selects a
7869 GNU/linux-like multilib, include files and instruction set for
7874 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7878 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7879 to link with input-output functions from a simulator library. Code,
7880 initialized data and zero-initialized data are allocated consecutively.
7884 Like @option{-sim}, but pass linker options to locate initialized data at
7885 0x40000000 and zero-initialized data at 0x80000000.
7889 @subsection CRX Options
7892 These options are defined specifically for the CRX ports.
7898 Enable the use of multiply-accumulate instructions. Disabled by default.
7902 Push instructions will be used to pass outgoing arguments when functions
7903 are called. Enabled by default.
7906 @node Darwin Options
7907 @subsection Darwin Options
7908 @cindex Darwin options
7910 These options are defined for all architectures running the Darwin operating
7913 FSF GCC on Darwin does not create ``fat'' object files; it will create
7914 an object file for the single architecture that it was built to
7915 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7916 @option{-arch} options are used; it does so by running the compiler or
7917 linker multiple times and joining the results together with
7920 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7921 @samp{i686}) is determined by the flags that specify the ISA
7922 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7923 @option{-force_cpusubtype_ALL} option can be used to override this.
7925 The Darwin tools vary in their behavior when presented with an ISA
7926 mismatch. The assembler, @file{as}, will only permit instructions to
7927 be used that are valid for the subtype of the file it is generating,
7928 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7929 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
7930 and print an error if asked to create a shared library with a less
7931 restrictive subtype than its input files (for instance, trying to put
7932 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
7933 for executables, @file{ld}, will quietly give the executable the most
7934 restrictive subtype of any of its input files.
7939 Add the framework directory @var{dir} to the head of the list of
7940 directories to be searched for header files. These directories are
7941 interleaved with those specified by @option{-I} options and are
7942 scanned in a left-to-right order.
7944 A framework directory is a directory with frameworks in it. A
7945 framework is a directory with a @samp{"Headers"} and/or
7946 @samp{"PrivateHeaders"} directory contained directly in it that ends
7947 in @samp{".framework"}. The name of a framework is the name of this
7948 directory excluding the @samp{".framework"}. Headers associated with
7949 the framework are found in one of those two directories, with
7950 @samp{"Headers"} being searched first. A subframework is a framework
7951 directory that is in a framework's @samp{"Frameworks"} directory.
7952 Includes of subframework headers can only appear in a header of a
7953 framework that contains the subframework, or in a sibling subframework
7954 header. Two subframeworks are siblings if they occur in the same
7955 framework. A subframework should not have the same name as a
7956 framework, a warning will be issued if this is violated. Currently a
7957 subframework cannot have subframeworks, in the future, the mechanism
7958 may be extended to support this. The standard frameworks can be found
7959 in @samp{"/System/Library/Frameworks"} and
7960 @samp{"/Library/Frameworks"}. An example include looks like
7961 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
7962 the name of the framework and header.h is found in the
7963 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
7967 Emit debugging information for symbols that are used. For STABS
7968 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
7969 This is by default ON@.
7973 Emit debugging information for all symbols and types.
7975 @item -mmacosx-version-min=@var{version}
7976 The earliest version of MacOS X that this executable will run on
7977 is @var{version}. Typical values of @var{version} include @code{10.1},
7978 @code{10.2}, and @code{10.3.9}.
7980 The default for this option is to make choices that seem to be most
7985 Enable kernel development mode. The @option{-mkernel} option sets
7986 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
7987 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
7988 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
7989 applicable. This mode also sets @option{-mno-altivec},
7990 @option{-msoft-float}, @option{-fno-builtin} and
7991 @option{-mlong-branch} for PowerPC targets.
7993 @item -mone-byte-bool
7994 @opindex mone-byte-bool
7995 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7996 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7997 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7998 option has no effect on x86.
8000 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
8001 to generate code that is not binary compatible with code generated
8002 without that switch. Using this switch may require recompiling all
8003 other modules in a program, including system libraries. Use this
8004 switch to conform to a non-default data model.
8006 @item -mfix-and-continue
8007 @itemx -ffix-and-continue
8008 @itemx -findirect-data
8009 @opindex mfix-and-continue
8010 @opindex ffix-and-continue
8011 @opindex findirect-data
8012 Generate code suitable for fast turn around development. Needed to
8013 enable gdb to dynamically load @code{.o} files into already running
8014 programs. @option{-findirect-data} and @option{-ffix-and-continue}
8015 are provided for backwards compatibility.
8019 Loads all members of static archive libraries.
8020 See man ld(1) for more information.
8022 @item -arch_errors_fatal
8023 @opindex arch_errors_fatal
8024 Cause the errors having to do with files that have the wrong architecture
8028 @opindex bind_at_load
8029 Causes the output file to be marked such that the dynamic linker will
8030 bind all undefined references when the file is loaded or launched.
8034 Produce a Mach-o bundle format file.
8035 See man ld(1) for more information.
8037 @item -bundle_loader @var{executable}
8038 @opindex bundle_loader
8039 This option specifies the @var{executable} that will be loading the build
8040 output file being linked. See man ld(1) for more information.
8044 When passed this option, GCC will produce a dynamic library instead of
8045 an executable when linking, using the Darwin @file{libtool} command.
8047 @item -force_cpusubtype_ALL
8048 @opindex force_cpusubtype_ALL
8049 This causes GCC's output file to have the @var{ALL} subtype, instead of
8050 one controlled by the @option{-mcpu} or @option{-march} option.
8052 @item -allowable_client @var{client_name}
8054 @itemx -compatibility_version
8055 @itemx -current_version
8057 @itemx -dependency-file
8059 @itemx -dylinker_install_name
8061 @itemx -exported_symbols_list
8063 @itemx -flat_namespace
8064 @itemx -force_flat_namespace
8065 @itemx -headerpad_max_install_names
8068 @itemx -install_name
8069 @itemx -keep_private_externs
8070 @itemx -multi_module
8071 @itemx -multiply_defined
8072 @itemx -multiply_defined_unused
8074 @itemx -no_dead_strip_inits_and_terms
8075 @itemx -nofixprebinding
8078 @itemx -noseglinkedit
8079 @itemx -pagezero_size
8081 @itemx -prebind_all_twolevel_modules
8082 @itemx -private_bundle
8083 @itemx -read_only_relocs
8085 @itemx -sectobjectsymbols
8089 @itemx -sectobjectsymbols
8092 @itemx -segs_read_only_addr
8093 @itemx -segs_read_write_addr
8094 @itemx -seg_addr_table
8095 @itemx -seg_addr_table_filename
8098 @itemx -segs_read_only_addr
8099 @itemx -segs_read_write_addr
8100 @itemx -single_module
8103 @itemx -sub_umbrella
8104 @itemx -twolevel_namespace
8107 @itemx -unexported_symbols_list
8108 @itemx -weak_reference_mismatches
8111 @opindex allowable_client
8112 @opindex client_name
8113 @opindex compatibility_version
8114 @opindex current_version
8116 @opindex dependency-file
8118 @opindex dylinker_install_name
8120 @opindex exported_symbols_list
8122 @opindex flat_namespace
8123 @opindex force_flat_namespace
8124 @opindex headerpad_max_install_names
8127 @opindex install_name
8128 @opindex keep_private_externs
8129 @opindex multi_module
8130 @opindex multiply_defined
8131 @opindex multiply_defined_unused
8133 @opindex no_dead_strip_inits_and_terms
8134 @opindex nofixprebinding
8135 @opindex nomultidefs
8137 @opindex noseglinkedit
8138 @opindex pagezero_size
8140 @opindex prebind_all_twolevel_modules
8141 @opindex private_bundle
8142 @opindex read_only_relocs
8144 @opindex sectobjectsymbols
8148 @opindex sectobjectsymbols
8151 @opindex segs_read_only_addr
8152 @opindex segs_read_write_addr
8153 @opindex seg_addr_table
8154 @opindex seg_addr_table_filename
8155 @opindex seglinkedit
8157 @opindex segs_read_only_addr
8158 @opindex segs_read_write_addr
8159 @opindex single_module
8161 @opindex sub_library
8162 @opindex sub_umbrella
8163 @opindex twolevel_namespace
8166 @opindex unexported_symbols_list
8167 @opindex weak_reference_mismatches
8168 @opindex whatsloaded
8170 These options are passed to the Darwin linker. The Darwin linker man page
8171 describes them in detail.
8174 @node DEC Alpha Options
8175 @subsection DEC Alpha Options
8177 These @samp{-m} options are defined for the DEC Alpha implementations:
8180 @item -mno-soft-float
8182 @opindex mno-soft-float
8183 @opindex msoft-float
8184 Use (do not use) the hardware floating-point instructions for
8185 floating-point operations. When @option{-msoft-float} is specified,
8186 functions in @file{libgcc.a} will be used to perform floating-point
8187 operations. Unless they are replaced by routines that emulate the
8188 floating-point operations, or compiled in such a way as to call such
8189 emulations routines, these routines will issue floating-point
8190 operations. If you are compiling for an Alpha without floating-point
8191 operations, you must ensure that the library is built so as not to call
8194 Note that Alpha implementations without floating-point operations are
8195 required to have floating-point registers.
8200 @opindex mno-fp-regs
8201 Generate code that uses (does not use) the floating-point register set.
8202 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8203 register set is not used, floating point operands are passed in integer
8204 registers as if they were integers and floating-point results are passed
8205 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8206 so any function with a floating-point argument or return value called by code
8207 compiled with @option{-mno-fp-regs} must also be compiled with that
8210 A typical use of this option is building a kernel that does not use,
8211 and hence need not save and restore, any floating-point registers.
8215 The Alpha architecture implements floating-point hardware optimized for
8216 maximum performance. It is mostly compliant with the IEEE floating
8217 point standard. However, for full compliance, software assistance is
8218 required. This option generates code fully IEEE compliant code
8219 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8220 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8221 defined during compilation. The resulting code is less efficient but is
8222 able to correctly support denormalized numbers and exceptional IEEE
8223 values such as not-a-number and plus/minus infinity. Other Alpha
8224 compilers call this option @option{-ieee_with_no_inexact}.
8226 @item -mieee-with-inexact
8227 @opindex mieee-with-inexact
8228 This is like @option{-mieee} except the generated code also maintains
8229 the IEEE @var{inexact-flag}. Turning on this option causes the
8230 generated code to implement fully-compliant IEEE math. In addition to
8231 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8232 macro. On some Alpha implementations the resulting code may execute
8233 significantly slower than the code generated by default. Since there is
8234 very little code that depends on the @var{inexact-flag}, you should
8235 normally not specify this option. Other Alpha compilers call this
8236 option @option{-ieee_with_inexact}.
8238 @item -mfp-trap-mode=@var{trap-mode}
8239 @opindex mfp-trap-mode
8240 This option controls what floating-point related traps are enabled.
8241 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8242 The trap mode can be set to one of four values:
8246 This is the default (normal) setting. The only traps that are enabled
8247 are the ones that cannot be disabled in software (e.g., division by zero
8251 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8255 Like @samp{u}, but the instructions are marked to be safe for software
8256 completion (see Alpha architecture manual for details).
8259 Like @samp{su}, but inexact traps are enabled as well.
8262 @item -mfp-rounding-mode=@var{rounding-mode}
8263 @opindex mfp-rounding-mode
8264 Selects the IEEE rounding mode. Other Alpha compilers call this option
8265 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8270 Normal IEEE rounding mode. Floating point numbers are rounded towards
8271 the nearest machine number or towards the even machine number in case
8275 Round towards minus infinity.
8278 Chopped rounding mode. Floating point numbers are rounded towards zero.
8281 Dynamic rounding mode. A field in the floating point control register
8282 (@var{fpcr}, see Alpha architecture reference manual) controls the
8283 rounding mode in effect. The C library initializes this register for
8284 rounding towards plus infinity. Thus, unless your program modifies the
8285 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8288 @item -mtrap-precision=@var{trap-precision}
8289 @opindex mtrap-precision
8290 In the Alpha architecture, floating point traps are imprecise. This
8291 means without software assistance it is impossible to recover from a
8292 floating trap and program execution normally needs to be terminated.
8293 GCC can generate code that can assist operating system trap handlers
8294 in determining the exact location that caused a floating point trap.
8295 Depending on the requirements of an application, different levels of
8296 precisions can be selected:
8300 Program precision. This option is the default and means a trap handler
8301 can only identify which program caused a floating point exception.
8304 Function precision. The trap handler can determine the function that
8305 caused a floating point exception.
8308 Instruction precision. The trap handler can determine the exact
8309 instruction that caused a floating point exception.
8312 Other Alpha compilers provide the equivalent options called
8313 @option{-scope_safe} and @option{-resumption_safe}.
8315 @item -mieee-conformant
8316 @opindex mieee-conformant
8317 This option marks the generated code as IEEE conformant. You must not
8318 use this option unless you also specify @option{-mtrap-precision=i} and either
8319 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8320 is to emit the line @samp{.eflag 48} in the function prologue of the
8321 generated assembly file. Under DEC Unix, this has the effect that
8322 IEEE-conformant math library routines will be linked in.
8324 @item -mbuild-constants
8325 @opindex mbuild-constants
8326 Normally GCC examines a 32- or 64-bit integer constant to
8327 see if it can construct it from smaller constants in two or three
8328 instructions. If it cannot, it will output the constant as a literal and
8329 generate code to load it from the data segment at runtime.
8331 Use this option to require GCC to construct @emph{all} integer constants
8332 using code, even if it takes more instructions (the maximum is six).
8334 You would typically use this option to build a shared library dynamic
8335 loader. Itself a shared library, it must relocate itself in memory
8336 before it can find the variables and constants in its own data segment.
8342 Select whether to generate code to be assembled by the vendor-supplied
8343 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8361 Indicate whether GCC should generate code to use the optional BWX,
8362 CIX, FIX and MAX instruction sets. The default is to use the instruction
8363 sets supported by the CPU type specified via @option{-mcpu=} option or that
8364 of the CPU on which GCC was built if none was specified.
8369 @opindex mfloat-ieee
8370 Generate code that uses (does not use) VAX F and G floating point
8371 arithmetic instead of IEEE single and double precision.
8373 @item -mexplicit-relocs
8374 @itemx -mno-explicit-relocs
8375 @opindex mexplicit-relocs
8376 @opindex mno-explicit-relocs
8377 Older Alpha assemblers provided no way to generate symbol relocations
8378 except via assembler macros. Use of these macros does not allow
8379 optimal instruction scheduling. GNU binutils as of version 2.12
8380 supports a new syntax that allows the compiler to explicitly mark
8381 which relocations should apply to which instructions. This option
8382 is mostly useful for debugging, as GCC detects the capabilities of
8383 the assembler when it is built and sets the default accordingly.
8387 @opindex msmall-data
8388 @opindex mlarge-data
8389 When @option{-mexplicit-relocs} is in effect, static data is
8390 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8391 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8392 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8393 16-bit relocations off of the @code{$gp} register. This limits the
8394 size of the small data area to 64KB, but allows the variables to be
8395 directly accessed via a single instruction.
8397 The default is @option{-mlarge-data}. With this option the data area
8398 is limited to just below 2GB@. Programs that require more than 2GB of
8399 data must use @code{malloc} or @code{mmap} to allocate the data in the
8400 heap instead of in the program's data segment.
8402 When generating code for shared libraries, @option{-fpic} implies
8403 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8407 @opindex msmall-text
8408 @opindex mlarge-text
8409 When @option{-msmall-text} is used, the compiler assumes that the
8410 code of the entire program (or shared library) fits in 4MB, and is
8411 thus reachable with a branch instruction. When @option{-msmall-data}
8412 is used, the compiler can assume that all local symbols share the
8413 same @code{$gp} value, and thus reduce the number of instructions
8414 required for a function call from 4 to 1.
8416 The default is @option{-mlarge-text}.
8418 @item -mcpu=@var{cpu_type}
8420 Set the instruction set and instruction scheduling parameters for
8421 machine type @var{cpu_type}. You can specify either the @samp{EV}
8422 style name or the corresponding chip number. GCC supports scheduling
8423 parameters for the EV4, EV5 and EV6 family of processors and will
8424 choose the default values for the instruction set from the processor
8425 you specify. If you do not specify a processor type, GCC will default
8426 to the processor on which the compiler was built.
8428 Supported values for @var{cpu_type} are
8434 Schedules as an EV4 and has no instruction set extensions.
8438 Schedules as an EV5 and has no instruction set extensions.
8442 Schedules as an EV5 and supports the BWX extension.
8447 Schedules as an EV5 and supports the BWX and MAX extensions.
8451 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8455 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8458 @item -mtune=@var{cpu_type}
8460 Set only the instruction scheduling parameters for machine type
8461 @var{cpu_type}. The instruction set is not changed.
8463 @item -mmemory-latency=@var{time}
8464 @opindex mmemory-latency
8465 Sets the latency the scheduler should assume for typical memory
8466 references as seen by the application. This number is highly
8467 dependent on the memory access patterns used by the application
8468 and the size of the external cache on the machine.
8470 Valid options for @var{time} are
8474 A decimal number representing clock cycles.
8480 The compiler contains estimates of the number of clock cycles for
8481 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8482 (also called Dcache, Scache, and Bcache), as well as to main memory.
8483 Note that L3 is only valid for EV5.
8488 @node DEC Alpha/VMS Options
8489 @subsection DEC Alpha/VMS Options
8491 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8494 @item -mvms-return-codes
8495 @opindex mvms-return-codes
8496 Return VMS condition codes from main. The default is to return POSIX
8497 style condition (e.g.@ error) codes.
8501 @subsection FRV Options
8508 Only use the first 32 general purpose registers.
8513 Use all 64 general purpose registers.
8518 Use only the first 32 floating point registers.
8523 Use all 64 floating point registers
8526 @opindex mhard-float
8528 Use hardware instructions for floating point operations.
8531 @opindex msoft-float
8533 Use library routines for floating point operations.
8538 Dynamically allocate condition code registers.
8543 Do not try to dynamically allocate condition code registers, only
8544 use @code{icc0} and @code{fcc0}.
8549 Change ABI to use double word insns.
8554 Do not use double word instructions.
8559 Use floating point double instructions.
8564 Do not use floating point double instructions.
8569 Use media instructions.
8574 Do not use media instructions.
8579 Use multiply and add/subtract instructions.
8584 Do not use multiply and add/subtract instructions.
8589 Select the FDPIC ABI, that uses function descriptors to represent
8590 pointers to functions. Without any PIC/PIE-related options, it
8591 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8592 assumes GOT entries and small data are within a 12-bit range from the
8593 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8594 are computed with 32 bits.
8597 @opindex minline-plt
8599 Enable inlining of PLT entries in function calls to functions that are
8600 not known to bind locally. It has no effect without @option{-mfdpic}.
8601 It's enabled by default if optimizing for speed and compiling for
8602 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8603 optimization option such as @option{-O3} or above is present in the
8609 Assume a large TLS segment when generating thread-local code.
8614 Do not assume a large TLS segment when generating thread-local code.
8619 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8620 that is known to be in read-only sections. It's enabled by default,
8621 except for @option{-fpic} or @option{-fpie}: even though it may help
8622 make the global offset table smaller, it trades 1 instruction for 4.
8623 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8624 one of which may be shared by multiple symbols, and it avoids the need
8625 for a GOT entry for the referenced symbol, so it's more likely to be a
8626 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8628 @item -multilib-library-pic
8629 @opindex multilib-library-pic
8631 Link with the (library, not FD) pic libraries. It's implied by
8632 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8633 @option{-fpic} without @option{-mfdpic}. You should never have to use
8639 Follow the EABI requirement of always creating a frame pointer whenever
8640 a stack frame is allocated. This option is enabled by default and can
8641 be disabled with @option{-mno-linked-fp}.
8644 @opindex mlong-calls
8646 Use indirect addressing to call functions outside the current
8647 compilation unit. This allows the functions to be placed anywhere
8648 within the 32-bit address space.
8650 @item -malign-labels
8651 @opindex malign-labels
8653 Try to align labels to an 8-byte boundary by inserting nops into the
8654 previous packet. This option only has an effect when VLIW packing
8655 is enabled. It doesn't create new packets; it merely adds nops to
8659 @opindex mlibrary-pic
8661 Generate position-independent EABI code.
8666 Use only the first four media accumulator registers.
8671 Use all eight media accumulator registers.
8676 Pack VLIW instructions.
8681 Do not pack VLIW instructions.
8686 Do not mark ABI switches in e_flags.
8691 Enable the use of conditional-move instructions (default).
8693 This switch is mainly for debugging the compiler and will likely be removed
8694 in a future version.
8696 @item -mno-cond-move
8697 @opindex mno-cond-move
8699 Disable the use of conditional-move instructions.
8701 This switch is mainly for debugging the compiler and will likely be removed
8702 in a future version.
8707 Enable the use of conditional set instructions (default).
8709 This switch is mainly for debugging the compiler and will likely be removed
8710 in a future version.
8715 Disable the use of conditional set instructions.
8717 This switch is mainly for debugging the compiler and will likely be removed
8718 in a future version.
8723 Enable the use of conditional execution (default).
8725 This switch is mainly for debugging the compiler and will likely be removed
8726 in a future version.
8728 @item -mno-cond-exec
8729 @opindex mno-cond-exec
8731 Disable the use of conditional execution.
8733 This switch is mainly for debugging the compiler and will likely be removed
8734 in a future version.
8737 @opindex mvliw-branch
8739 Run a pass to pack branches into VLIW instructions (default).
8741 This switch is mainly for debugging the compiler and will likely be removed
8742 in a future version.
8744 @item -mno-vliw-branch
8745 @opindex mno-vliw-branch
8747 Do not run a pass to pack branches into VLIW instructions.
8749 This switch is mainly for debugging the compiler and will likely be removed
8750 in a future version.
8752 @item -mmulti-cond-exec
8753 @opindex mmulti-cond-exec
8755 Enable optimization of @code{&&} and @code{||} in conditional execution
8758 This switch is mainly for debugging the compiler and will likely be removed
8759 in a future version.
8761 @item -mno-multi-cond-exec
8762 @opindex mno-multi-cond-exec
8764 Disable optimization of @code{&&} and @code{||} in conditional execution.
8766 This switch is mainly for debugging the compiler and will likely be removed
8767 in a future version.
8769 @item -mnested-cond-exec
8770 @opindex mnested-cond-exec
8772 Enable nested conditional execution optimizations (default).
8774 This switch is mainly for debugging the compiler and will likely be removed
8775 in a future version.
8777 @item -mno-nested-cond-exec
8778 @opindex mno-nested-cond-exec
8780 Disable nested conditional execution optimizations.
8782 This switch is mainly for debugging the compiler and will likely be removed
8783 in a future version.
8785 @item -moptimize-membar
8786 @opindex moptimize-membar
8788 This switch removes redundant @code{membar} instructions from the
8789 compiler generated code. It is enabled by default.
8791 @item -mno-optimize-membar
8792 @opindex mno-optimize-membar
8794 This switch disables the automatic removal of redundant @code{membar}
8795 instructions from the generated code.
8797 @item -mtomcat-stats
8798 @opindex mtomcat-stats
8800 Cause gas to print out tomcat statistics.
8802 @item -mcpu=@var{cpu}
8805 Select the processor type for which to generate code. Possible values are
8806 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8807 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8811 @node GNU/Linux Options
8812 @subsection GNU/Linux Options
8814 These @samp{-m} options are defined for GNU/Linux targets:
8819 Use the GNU C library instead of uClibc. This is the default except
8820 on @samp{*-*-linux-*uclibc*} targets.
8824 Use uClibc instead of the GNU C library. This is the default on
8825 @samp{*-*-linux-*uclibc*} targets.
8828 @node H8/300 Options
8829 @subsection H8/300 Options
8831 These @samp{-m} options are defined for the H8/300 implementations:
8836 Shorten some address references at link time, when possible; uses the
8837 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8838 ld, Using ld}, for a fuller description.
8842 Generate code for the H8/300H@.
8846 Generate code for the H8S@.
8850 Generate code for the H8S and H8/300H in the normal mode. This switch
8851 must be used either with @option{-mh} or @option{-ms}.
8855 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8859 Make @code{int} data 32 bits by default.
8863 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8864 The default for the H8/300H and H8S is to align longs and floats on 4
8866 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8867 This option has no effect on the H8/300.
8871 @subsection HPPA Options
8872 @cindex HPPA Options
8874 These @samp{-m} options are defined for the HPPA family of computers:
8877 @item -march=@var{architecture-type}
8879 Generate code for the specified architecture. The choices for
8880 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8881 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8882 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8883 architecture option for your machine. Code compiled for lower numbered
8884 architectures will run on higher numbered architectures, but not the
8888 @itemx -mpa-risc-1-1
8889 @itemx -mpa-risc-2-0
8890 @opindex mpa-risc-1-0
8891 @opindex mpa-risc-1-1
8892 @opindex mpa-risc-2-0
8893 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8896 @opindex mbig-switch
8897 Generate code suitable for big switch tables. Use this option only if
8898 the assembler/linker complain about out of range branches within a switch
8901 @item -mjump-in-delay
8902 @opindex mjump-in-delay
8903 Fill delay slots of function calls with unconditional jump instructions
8904 by modifying the return pointer for the function call to be the target
8905 of the conditional jump.
8907 @item -mdisable-fpregs
8908 @opindex mdisable-fpregs
8909 Prevent floating point registers from being used in any manner. This is
8910 necessary for compiling kernels which perform lazy context switching of
8911 floating point registers. If you use this option and attempt to perform
8912 floating point operations, the compiler will abort.
8914 @item -mdisable-indexing
8915 @opindex mdisable-indexing
8916 Prevent the compiler from using indexing address modes. This avoids some
8917 rather obscure problems when compiling MIG generated code under MACH@.
8919 @item -mno-space-regs
8920 @opindex mno-space-regs
8921 Generate code that assumes the target has no space registers. This allows
8922 GCC to generate faster indirect calls and use unscaled index address modes.
8924 Such code is suitable for level 0 PA systems and kernels.
8926 @item -mfast-indirect-calls
8927 @opindex mfast-indirect-calls
8928 Generate code that assumes calls never cross space boundaries. This
8929 allows GCC to emit code which performs faster indirect calls.
8931 This option will not work in the presence of shared libraries or nested
8934 @item -mfixed-range=@var{register-range}
8935 @opindex mfixed-range
8936 Generate code treating the given register range as fixed registers.
8937 A fixed register is one that the register allocator can not use. This is
8938 useful when compiling kernel code. A register range is specified as
8939 two registers separated by a dash. Multiple register ranges can be
8940 specified separated by a comma.
8942 @item -mlong-load-store
8943 @opindex mlong-load-store
8944 Generate 3-instruction load and store sequences as sometimes required by
8945 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8948 @item -mportable-runtime
8949 @opindex mportable-runtime
8950 Use the portable calling conventions proposed by HP for ELF systems.
8954 Enable the use of assembler directives only GAS understands.
8956 @item -mschedule=@var{cpu-type}
8958 Schedule code according to the constraints for the machine type
8959 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8960 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8961 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8962 proper scheduling option for your machine. The default scheduling is
8966 @opindex mlinker-opt
8967 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8968 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8969 linkers in which they give bogus error messages when linking some programs.
8972 @opindex msoft-float
8973 Generate output containing library calls for floating point.
8974 @strong{Warning:} the requisite libraries are not available for all HPPA
8975 targets. Normally the facilities of the machine's usual C compiler are
8976 used, but this cannot be done directly in cross-compilation. You must make
8977 your own arrangements to provide suitable library functions for
8978 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8979 does provide software floating point support.
8981 @option{-msoft-float} changes the calling convention in the output file;
8982 therefore, it is only useful if you compile @emph{all} of a program with
8983 this option. In particular, you need to compile @file{libgcc.a}, the
8984 library that comes with GCC, with @option{-msoft-float} in order for
8989 Generate the predefine, @code{_SIO}, for server IO@. The default is
8990 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8991 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
8992 options are available under HP-UX and HI-UX@.
8996 Use GNU ld specific options. This passes @option{-shared} to ld when
8997 building a shared library. It is the default when GCC is configured,
8998 explicitly or implicitly, with the GNU linker. This option does not
8999 have any affect on which ld is called, it only changes what parameters
9000 are passed to that ld. The ld that is called is determined by the
9001 @option{--with-ld} configure option, GCC's program search path, and
9002 finally by the user's @env{PATH}. The linker used by GCC can be printed
9003 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
9004 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9008 Use HP ld specific options. This passes @option{-b} to ld when building
9009 a shared library and passes @option{+Accept TypeMismatch} to ld on all
9010 links. It is the default when GCC is configured, explicitly or
9011 implicitly, with the HP linker. This option does not have any affect on
9012 which ld is called, it only changes what parameters are passed to that
9013 ld. The ld that is called is determined by the @option{--with-ld}
9014 configure option, GCC's program search path, and finally by the user's
9015 @env{PATH}. The linker used by GCC can be printed using @samp{which
9016 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
9017 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9020 @opindex mno-long-calls
9021 Generate code that uses long call sequences. This ensures that a call
9022 is always able to reach linker generated stubs. The default is to generate
9023 long calls only when the distance from the call site to the beginning
9024 of the function or translation unit, as the case may be, exceeds a
9025 predefined limit set by the branch type being used. The limits for
9026 normal calls are 7,600,000 and 240,000 bytes, respectively for the
9027 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
9030 Distances are measured from the beginning of functions when using the
9031 @option{-ffunction-sections} option, or when using the @option{-mgas}
9032 and @option{-mno-portable-runtime} options together under HP-UX with
9035 It is normally not desirable to use this option as it will degrade
9036 performance. However, it may be useful in large applications,
9037 particularly when partial linking is used to build the application.
9039 The types of long calls used depends on the capabilities of the
9040 assembler and linker, and the type of code being generated. The
9041 impact on systems that support long absolute calls, and long pic
9042 symbol-difference or pc-relative calls should be relatively small.
9043 However, an indirect call is used on 32-bit ELF systems in pic code
9044 and it is quite long.
9046 @item -munix=@var{unix-std}
9048 Generate compiler predefines and select a startfile for the specified
9049 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
9050 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
9051 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
9052 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
9053 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
9056 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
9057 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
9058 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
9059 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
9060 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
9061 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
9063 It is @emph{important} to note that this option changes the interfaces
9064 for various library routines. It also affects the operational behavior
9065 of the C library. Thus, @emph{extreme} care is needed in using this
9068 Library code that is intended to operate with more than one UNIX
9069 standard must test, set and restore the variable @var{__xpg4_extended_mask}
9070 as appropriate. Most GNU software doesn't provide this capability.
9074 Suppress the generation of link options to search libdld.sl when the
9075 @option{-static} option is specified on HP-UX 10 and later.
9079 The HP-UX implementation of setlocale in libc has a dependency on
9080 libdld.sl. There isn't an archive version of libdld.sl. Thus,
9081 when the @option{-static} option is specified, special link options
9082 are needed to resolve this dependency.
9084 On HP-UX 10 and later, the GCC driver adds the necessary options to
9085 link with libdld.sl when the @option{-static} option is specified.
9086 This causes the resulting binary to be dynamic. On the 64-bit port,
9087 the linkers generate dynamic binaries by default in any case. The
9088 @option{-nolibdld} option can be used to prevent the GCC driver from
9089 adding these link options.
9093 Add support for multithreading with the @dfn{dce thread} library
9094 under HP-UX@. This option sets flags for both the preprocessor and
9098 @node i386 and x86-64 Options
9099 @subsection Intel 386 and AMD x86-64 Options
9100 @cindex i386 Options
9101 @cindex x86-64 Options
9102 @cindex Intel 386 Options
9103 @cindex AMD x86-64 Options
9105 These @samp{-m} options are defined for the i386 and x86-64 family of
9109 @item -mtune=@var{cpu-type}
9111 Tune to @var{cpu-type} everything applicable about the generated code, except
9112 for the ABI and the set of available instructions. The choices for
9116 Produce code optimized for the most common IA32/AMD64/EM64T processors.
9117 If you know the CPU on which your code will run, then you should use
9118 the corresponding @option{-mtune} option instead of
9119 @option{-mtune=generic}. But, if you do not know exactly what CPU users
9120 of your application will have, then you should use this option.
9122 As new processors are deployed in the marketplace, the behavior of this
9123 option will change. Therefore, if you upgrade to a newer version of
9124 GCC, the code generated option will change to reflect the processors
9125 that were most common when that version of GCC was released.
9127 There is no @option{-march=generic} option because @option{-march}
9128 indicates the instruction set the compiler can use, and there is no
9129 generic instruction set applicable to all processors. In contrast,
9130 @option{-mtune} indicates the processor (or, in this case, collection of
9131 processors) for which the code is optimized.
9133 This selects the CPU to tune for at compilation time by determining
9134 the processor type of the compiling machine. Using @option{-mtune=native}
9135 will produce code optimized for the local machine under the constraints
9136 of the selected instruction set. Using @option{-march=native} will
9137 enable all instruction subsets supported by the local machine (hence
9138 the result might not run on different machines).
9140 Original Intel's i386 CPU@.
9142 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
9144 Intel Pentium CPU with no MMX support.
9146 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
9148 Intel PentiumPro CPU@.
9150 Same as @code{generic}, but when used as @code{march} option, PentiumPro
9151 instruction set will be used, so the code will run on all i686 family chips.
9153 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
9154 @item pentium3, pentium3m
9155 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
9158 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
9159 support. Used by Centrino notebooks.
9160 @item pentium4, pentium4m
9161 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
9163 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
9166 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
9167 SSE2 and SSE3 instruction set support.
9169 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
9170 instruction set support.
9172 AMD K6 CPU with MMX instruction set support.
9174 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
9175 @item athlon, athlon-tbird
9176 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
9178 @item athlon-4, athlon-xp, athlon-mp
9179 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
9180 instruction set support.
9181 @item k8, opteron, athlon64, athlon-fx
9182 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
9183 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
9184 @item k8-sse3, opteron-sse3, athlon64-sse3
9185 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
9186 @item amdfam10, barcelona
9187 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
9188 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
9189 instruction set extensions.)
9191 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
9194 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
9195 instruction set support.
9197 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
9198 implemented for this chip.)
9200 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
9201 implemented for this chip.)
9203 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
9206 While picking a specific @var{cpu-type} will schedule things appropriately
9207 for that particular chip, the compiler will not generate any code that
9208 does not run on the i386 without the @option{-march=@var{cpu-type}} option
9211 @item -march=@var{cpu-type}
9213 Generate instructions for the machine type @var{cpu-type}. The choices
9214 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
9215 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
9217 @item -mcpu=@var{cpu-type}
9219 A deprecated synonym for @option{-mtune}.
9228 @opindex mpentiumpro
9229 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
9230 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
9231 These synonyms are deprecated.
9233 @item -mfpmath=@var{unit}
9235 Generate floating point arithmetics for selected unit @var{unit}. The choices
9240 Use the standard 387 floating point coprocessor present majority of chips and
9241 emulated otherwise. Code compiled with this option will run almost everywhere.
9242 The temporary results are computed in 80bit precision instead of precision
9243 specified by the type resulting in slightly different results compared to most
9244 of other chips. See @option{-ffloat-store} for more detailed description.
9246 This is the default choice for i386 compiler.
9249 Use scalar floating point instructions present in the SSE instruction set.
9250 This instruction set is supported by Pentium3 and newer chips, in the AMD line
9251 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
9252 instruction set supports only single precision arithmetics, thus the double and
9253 extended precision arithmetics is still done using 387. Later version, present
9254 only in Pentium4 and the future AMD x86-64 chips supports double precision
9257 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
9258 or @option{-msse2} switches to enable SSE extensions and make this option
9259 effective. For the x86-64 compiler, these extensions are enabled by default.
9261 The resulting code should be considerably faster in the majority of cases and avoid
9262 the numerical instability problems of 387 code, but may break some existing
9263 code that expects temporaries to be 80bit.
9265 This is the default choice for the x86-64 compiler.
9268 Attempt to utilize both instruction sets at once. This effectively double the
9269 amount of available registers and on chips with separate execution units for
9270 387 and SSE the execution resources too. Use this option with care, as it is
9271 still experimental, because the GCC register allocator does not model separate
9272 functional units well resulting in instable performance.
9275 @item -masm=@var{dialect}
9276 @opindex masm=@var{dialect}
9277 Output asm instructions using selected @var{dialect}. Supported
9278 choices are @samp{intel} or @samp{att} (the default one). Darwin does
9279 not support @samp{intel}.
9284 @opindex mno-ieee-fp
9285 Control whether or not the compiler uses IEEE floating point
9286 comparisons. These handle correctly the case where the result of a
9287 comparison is unordered.
9290 @opindex msoft-float
9291 Generate output containing library calls for floating point.
9292 @strong{Warning:} the requisite libraries are not part of GCC@.
9293 Normally the facilities of the machine's usual C compiler are used, but
9294 this can't be done directly in cross-compilation. You must make your
9295 own arrangements to provide suitable library functions for
9298 On machines where a function returns floating point results in the 80387
9299 register stack, some floating point opcodes may be emitted even if
9300 @option{-msoft-float} is used.
9302 @item -mno-fp-ret-in-387
9303 @opindex mno-fp-ret-in-387
9304 Do not use the FPU registers for return values of functions.
9306 The usual calling convention has functions return values of types
9307 @code{float} and @code{double} in an FPU register, even if there
9308 is no FPU@. The idea is that the operating system should emulate
9311 The option @option{-mno-fp-ret-in-387} causes such values to be returned
9312 in ordinary CPU registers instead.
9314 @item -mno-fancy-math-387
9315 @opindex mno-fancy-math-387
9316 Some 387 emulators do not support the @code{sin}, @code{cos} and
9317 @code{sqrt} instructions for the 387. Specify this option to avoid
9318 generating those instructions. This option is the default on
9319 OpenBSD and NetBSD@. This option is overridden when @option{-march}
9320 indicates that the target cpu will always have an FPU and so the
9321 instruction will not need emulation. As of revision 2.6.1, these
9322 instructions are not generated unless you also use the
9323 @option{-funsafe-math-optimizations} switch.
9325 @item -malign-double
9326 @itemx -mno-align-double
9327 @opindex malign-double
9328 @opindex mno-align-double
9329 Control whether GCC aligns @code{double}, @code{long double}, and
9330 @code{long long} variables on a two word boundary or a one word
9331 boundary. Aligning @code{double} variables on a two word boundary will
9332 produce code that runs somewhat faster on a @samp{Pentium} at the
9333 expense of more memory.
9335 On x86-64, @option{-malign-double} is enabled by default.
9337 @strong{Warning:} if you use the @option{-malign-double} switch,
9338 structures containing the above types will be aligned differently than
9339 the published application binary interface specifications for the 386
9340 and will not be binary compatible with structures in code compiled
9341 without that switch.
9343 @item -m96bit-long-double
9344 @itemx -m128bit-long-double
9345 @opindex m96bit-long-double
9346 @opindex m128bit-long-double
9347 These switches control the size of @code{long double} type. The i386
9348 application binary interface specifies the size to be 96 bits,
9349 so @option{-m96bit-long-double} is the default in 32 bit mode.
9351 Modern architectures (Pentium and newer) would prefer @code{long double}
9352 to be aligned to an 8 or 16 byte boundary. In arrays or structures
9353 conforming to the ABI, this would not be possible. So specifying a
9354 @option{-m128bit-long-double} will align @code{long double}
9355 to a 16 byte boundary by padding the @code{long double} with an additional
9358 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
9359 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
9361 Notice that neither of these options enable any extra precision over the x87
9362 standard of 80 bits for a @code{long double}.
9364 @strong{Warning:} if you override the default value for your target ABI, the
9365 structures and arrays containing @code{long double} variables will change
9366 their size as well as function calling convention for function taking
9367 @code{long double} will be modified. Hence they will not be binary
9368 compatible with arrays or structures in code compiled without that switch.
9370 @item -mmlarge-data-threshold=@var{number}
9371 @opindex mlarge-data-threshold=@var{number}
9372 When @option{-mcmodel=medium} is specified, the data greater than
9373 @var{threshold} are placed in large data section. This value must be the
9374 same across all object linked into the binary and defaults to 65535.
9377 @itemx -mno-svr3-shlib
9378 @opindex msvr3-shlib
9379 @opindex mno-svr3-shlib
9380 Control whether GCC places uninitialized local variables into the
9381 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
9382 into @code{bss}. These options are meaningful only on System V Release 3.
9386 Use a different function-calling convention, in which functions that
9387 take a fixed number of arguments return with the @code{ret} @var{num}
9388 instruction, which pops their arguments while returning. This saves one
9389 instruction in the caller since there is no need to pop the arguments
9392 You can specify that an individual function is called with this calling
9393 sequence with the function attribute @samp{stdcall}. You can also
9394 override the @option{-mrtd} option by using the function attribute
9395 @samp{cdecl}. @xref{Function Attributes}.
9397 @strong{Warning:} this calling convention is incompatible with the one
9398 normally used on Unix, so you cannot use it if you need to call
9399 libraries compiled with the Unix compiler.
9401 Also, you must provide function prototypes for all functions that
9402 take variable numbers of arguments (including @code{printf});
9403 otherwise incorrect code will be generated for calls to those
9406 In addition, seriously incorrect code will result if you call a
9407 function with too many arguments. (Normally, extra arguments are
9408 harmlessly ignored.)
9410 @item -mregparm=@var{num}
9412 Control how many registers are used to pass integer arguments. By
9413 default, no registers are used to pass arguments, and at most 3
9414 registers can be used. You can control this behavior for a specific
9415 function by using the function attribute @samp{regparm}.
9416 @xref{Function Attributes}.
9418 @strong{Warning:} if you use this switch, and
9419 @var{num} is nonzero, then you must build all modules with the same
9420 value, including any libraries. This includes the system libraries and
9424 @opindex msseregparm
9425 Use SSE register passing conventions for float and double arguments
9426 and return values. You can control this behavior for a specific
9427 function by using the function attribute @samp{sseregparm}.
9428 @xref{Function Attributes}.
9430 @strong{Warning:} if you use this switch then you must build all
9431 modules with the same value, including any libraries. This includes
9432 the system libraries and startup modules.
9434 @item -mstackrealign
9435 @opindex mstackrealign
9436 Realign the stack at entry. On the Intel x86, the
9437 @option{-mstackrealign} option will generate an alternate prologue and
9438 epilogue that realigns the runtime stack. This supports mixing legacy
9439 codes that keep a 4-byte aligned stack with modern codes that keep a
9440 16-byte stack for SSE compatibility. The alternate prologue and
9441 epilogue are slower and bigger than the regular ones, and the
9442 alternate prologue requires an extra scratch register; this lowers the
9443 number of registers available if used in conjunction with the
9444 @code{regparm} attribute. The @option{-mstackrealign} option is
9445 incompatible with the nested function prologue; this is considered a
9446 hard error. See also the attribute @code{force_align_arg_pointer},
9447 applicable to individual functions.
9449 @item -mpreferred-stack-boundary=@var{num}
9450 @opindex mpreferred-stack-boundary
9451 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
9452 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
9453 the default is 4 (16 bytes or 128 bits).
9455 On Pentium and PentiumPro, @code{double} and @code{long double} values
9456 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
9457 suffer significant run time performance penalties. On Pentium III, the
9458 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
9459 properly if it is not 16 byte aligned.
9461 To ensure proper alignment of this values on the stack, the stack boundary
9462 must be as aligned as that required by any value stored on the stack.
9463 Further, every function must be generated such that it keeps the stack
9464 aligned. Thus calling a function compiled with a higher preferred
9465 stack boundary from a function compiled with a lower preferred stack
9466 boundary will most likely misalign the stack. It is recommended that
9467 libraries that use callbacks always use the default setting.
9469 This extra alignment does consume extra stack space, and generally
9470 increases code size. Code that is sensitive to stack space usage, such
9471 as embedded systems and operating system kernels, may want to reduce the
9472 preferred alignment to @option{-mpreferred-stack-boundary=2}.
9500 These switches enable or disable the use of instructions in the MMX,
9501 SSE, SSE2, SSE3, SSSE3, SSE4A, ABM, AES or 3DNow! extended
9502 instruction sets. These extensions are also available as built-in
9503 functions: see @ref{X86 Built-in Functions}, for details of the functions
9504 enabled and disabled by these switches.
9506 To have SSE/SSE2 instructions generated automatically from floating-point
9507 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
9509 These options will enable GCC to use these extended instructions in
9510 generated code, even without @option{-mfpmath=sse}. Applications which
9511 perform runtime CPU detection must compile separate files for each
9512 supported architecture, using the appropriate flags. In particular,
9513 the file containing the CPU detection code should be compiled without
9517 @itemx -mno-push-args
9519 @opindex mno-push-args
9520 Use PUSH operations to store outgoing parameters. This method is shorter
9521 and usually equally fast as method using SUB/MOV operations and is enabled
9522 by default. In some cases disabling it may improve performance because of
9523 improved scheduling and reduced dependencies.
9525 @item -maccumulate-outgoing-args
9526 @opindex maccumulate-outgoing-args
9527 If enabled, the maximum amount of space required for outgoing arguments will be
9528 computed in the function prologue. This is faster on most modern CPUs
9529 because of reduced dependencies, improved scheduling and reduced stack usage
9530 when preferred stack boundary is not equal to 2. The drawback is a notable
9531 increase in code size. This switch implies @option{-mno-push-args}.
9535 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
9536 on thread-safe exception handling must compile and link all code with the
9537 @option{-mthreads} option. When compiling, @option{-mthreads} defines
9538 @option{-D_MT}; when linking, it links in a special thread helper library
9539 @option{-lmingwthrd} which cleans up per thread exception handling data.
9541 @item -mno-align-stringops
9542 @opindex mno-align-stringops
9543 Do not align destination of inlined string operations. This switch reduces
9544 code size and improves performance in case the destination is already aligned,
9545 but GCC doesn't know about it.
9547 @item -minline-all-stringops
9548 @opindex minline-all-stringops
9549 By default GCC inlines string operations only when destination is known to be
9550 aligned at least to 4 byte boundary. This enables more inlining, increase code
9551 size, but may improve performance of code that depends on fast memcpy, strlen
9552 and memset for short lengths.
9554 @item -momit-leaf-frame-pointer
9555 @opindex momit-leaf-frame-pointer
9556 Don't keep the frame pointer in a register for leaf functions. This
9557 avoids the instructions to save, set up and restore frame pointers and
9558 makes an extra register available in leaf functions. The option
9559 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9560 which might make debugging harder.
9562 @item -mtls-direct-seg-refs
9563 @itemx -mno-tls-direct-seg-refs
9564 @opindex mtls-direct-seg-refs
9565 Controls whether TLS variables may be accessed with offsets from the
9566 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
9567 or whether the thread base pointer must be added. Whether or not this
9568 is legal depends on the operating system, and whether it maps the
9569 segment to cover the entire TLS area.
9571 For systems that use GNU libc, the default is on.
9574 These @samp{-m} switches are supported in addition to the above
9575 on AMD x86-64 processors in 64-bit environments.
9582 Generate code for a 32-bit or 64-bit environment.
9583 The 32-bit environment sets int, long and pointer to 32 bits and
9584 generates code that runs on any i386 system.
9585 The 64-bit environment sets int to 32 bits and long and pointer
9586 to 64 bits and generates code for AMD's x86-64 architecture. For
9587 darwin only the -m64 option turns off the @option{-fno-pic} and
9588 @option{-mdynamic-no-pic} options.
9591 @opindex no-red-zone
9592 Do not use a so called red zone for x86-64 code. The red zone is mandated
9593 by the x86-64 ABI, it is a 128-byte area beyond the location of the
9594 stack pointer that will not be modified by signal or interrupt handlers
9595 and therefore can be used for temporary data without adjusting the stack
9596 pointer. The flag @option{-mno-red-zone} disables this red zone.
9598 @item -mcmodel=small
9599 @opindex mcmodel=small
9600 Generate code for the small code model: the program and its symbols must
9601 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
9602 Programs can be statically or dynamically linked. This is the default
9605 @item -mcmodel=kernel
9606 @opindex mcmodel=kernel
9607 Generate code for the kernel code model. The kernel runs in the
9608 negative 2 GB of the address space.
9609 This model has to be used for Linux kernel code.
9611 @item -mcmodel=medium
9612 @opindex mcmodel=medium
9613 Generate code for the medium model: The program is linked in the lower 2
9614 GB of the address space but symbols can be located anywhere in the
9615 address space. Programs can be statically or dynamically linked, but
9616 building of shared libraries are not supported with the medium model.
9618 @item -mcmodel=large
9619 @opindex mcmodel=large
9620 Generate code for the large model: This model makes no assumptions
9621 about addresses and sizes of sections. Currently GCC does not implement
9626 @subsection IA-64 Options
9627 @cindex IA-64 Options
9629 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9633 @opindex mbig-endian
9634 Generate code for a big endian target. This is the default for HP-UX@.
9636 @item -mlittle-endian
9637 @opindex mlittle-endian
9638 Generate code for a little endian target. This is the default for AIX5
9645 Generate (or don't) code for the GNU assembler. This is the default.
9646 @c Also, this is the default if the configure option @option{--with-gnu-as}
9653 Generate (or don't) code for the GNU linker. This is the default.
9654 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9659 Generate code that does not use a global pointer register. The result
9660 is not position independent code, and violates the IA-64 ABI@.
9662 @item -mvolatile-asm-stop
9663 @itemx -mno-volatile-asm-stop
9664 @opindex mvolatile-asm-stop
9665 @opindex mno-volatile-asm-stop
9666 Generate (or don't) a stop bit immediately before and after volatile asm
9669 @item -mregister-names
9670 @itemx -mno-register-names
9671 @opindex mregister-names
9672 @opindex mno-register-names
9673 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9674 the stacked registers. This may make assembler output more readable.
9680 Disable (or enable) optimizations that use the small data section. This may
9681 be useful for working around optimizer bugs.
9684 @opindex mconstant-gp
9685 Generate code that uses a single constant global pointer value. This is
9686 useful when compiling kernel code.
9690 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9691 This is useful when compiling firmware code.
9693 @item -minline-float-divide-min-latency
9694 @opindex minline-float-divide-min-latency
9695 Generate code for inline divides of floating point values
9696 using the minimum latency algorithm.
9698 @item -minline-float-divide-max-throughput
9699 @opindex minline-float-divide-max-throughput
9700 Generate code for inline divides of floating point values
9701 using the maximum throughput algorithm.
9703 @item -minline-int-divide-min-latency
9704 @opindex minline-int-divide-min-latency
9705 Generate code for inline divides of integer values
9706 using the minimum latency algorithm.
9708 @item -minline-int-divide-max-throughput
9709 @opindex minline-int-divide-max-throughput
9710 Generate code for inline divides of integer values
9711 using the maximum throughput algorithm.
9713 @item -minline-sqrt-min-latency
9714 @opindex minline-sqrt-min-latency
9715 Generate code for inline square roots
9716 using the minimum latency algorithm.
9718 @item -minline-sqrt-max-throughput
9719 @opindex minline-sqrt-max-throughput
9720 Generate code for inline square roots
9721 using the maximum throughput algorithm.
9723 @item -mno-dwarf2-asm
9725 @opindex mno-dwarf2-asm
9726 @opindex mdwarf2-asm
9727 Don't (or do) generate assembler code for the DWARF2 line number debugging
9728 info. This may be useful when not using the GNU assembler.
9730 @item -mearly-stop-bits
9731 @itemx -mno-early-stop-bits
9732 @opindex mearly-stop-bits
9733 @opindex mno-early-stop-bits
9734 Allow stop bits to be placed earlier than immediately preceding the
9735 instruction that triggered the stop bit. This can improve instruction
9736 scheduling, but does not always do so.
9738 @item -mfixed-range=@var{register-range}
9739 @opindex mfixed-range
9740 Generate code treating the given register range as fixed registers.
9741 A fixed register is one that the register allocator can not use. This is
9742 useful when compiling kernel code. A register range is specified as
9743 two registers separated by a dash. Multiple register ranges can be
9744 specified separated by a comma.
9746 @item -mtls-size=@var{tls-size}
9748 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9751 @item -mtune=@var{cpu-type}
9753 Tune the instruction scheduling for a particular CPU, Valid values are
9754 itanium, itanium1, merced, itanium2, and mckinley.
9760 Add support for multithreading using the POSIX threads library. This
9761 option sets flags for both the preprocessor and linker. It does
9762 not affect the thread safety of object code produced by the compiler or
9763 that of libraries supplied with it. These are HP-UX specific flags.
9769 Generate code for a 32-bit or 64-bit environment.
9770 The 32-bit environment sets int, long and pointer to 32 bits.
9771 The 64-bit environment sets int to 32 bits and long and pointer
9772 to 64 bits. These are HP-UX specific flags.
9774 @item -mno-sched-br-data-spec
9775 @itemx -msched-br-data-spec
9776 @opindex mno-sched-br-data-spec
9777 @opindex msched-br-data-spec
9778 (Dis/En)able data speculative scheduling before reload.
9779 This will result in generation of the ld.a instructions and
9780 the corresponding check instructions (ld.c / chk.a).
9781 The default is 'disable'.
9783 @item -msched-ar-data-spec
9784 @itemx -mno-sched-ar-data-spec
9785 @opindex msched-ar-data-spec
9786 @opindex mno-sched-ar-data-spec
9787 (En/Dis)able data speculative scheduling after reload.
9788 This will result in generation of the ld.a instructions and
9789 the corresponding check instructions (ld.c / chk.a).
9790 The default is 'enable'.
9792 @item -mno-sched-control-spec
9793 @itemx -msched-control-spec
9794 @opindex mno-sched-control-spec
9795 @opindex msched-control-spec
9796 (Dis/En)able control speculative scheduling. This feature is
9797 available only during region scheduling (i.e. before reload).
9798 This will result in generation of the ld.s instructions and
9799 the corresponding check instructions chk.s .
9800 The default is 'disable'.
9802 @item -msched-br-in-data-spec
9803 @itemx -mno-sched-br-in-data-spec
9804 @opindex msched-br-in-data-spec
9805 @opindex mno-sched-br-in-data-spec
9806 (En/Dis)able speculative scheduling of the instructions that
9807 are dependent on the data speculative loads before reload.
9808 This is effective only with @option{-msched-br-data-spec} enabled.
9809 The default is 'enable'.
9811 @item -msched-ar-in-data-spec
9812 @itemx -mno-sched-ar-in-data-spec
9813 @opindex msched-ar-in-data-spec
9814 @opindex mno-sched-ar-in-data-spec
9815 (En/Dis)able speculative scheduling of the instructions that
9816 are dependent on the data speculative loads after reload.
9817 This is effective only with @option{-msched-ar-data-spec} enabled.
9818 The default is 'enable'.
9820 @item -msched-in-control-spec
9821 @itemx -mno-sched-in-control-spec
9822 @opindex msched-in-control-spec
9823 @opindex mno-sched-in-control-spec
9824 (En/Dis)able speculative scheduling of the instructions that
9825 are dependent on the control speculative loads.
9826 This is effective only with @option{-msched-control-spec} enabled.
9827 The default is 'enable'.
9830 @itemx -mno-sched-ldc
9832 @opindex mno-sched-ldc
9833 (En/Dis)able use of simple data speculation checks ld.c .
9834 If disabled, only chk.a instructions will be emitted to check
9835 data speculative loads.
9836 The default is 'enable'.
9838 @item -mno-sched-control-ldc
9839 @itemx -msched-control-ldc
9840 @opindex mno-sched-control-ldc
9841 @opindex msched-control-ldc
9842 (Dis/En)able use of ld.c instructions to check control speculative loads.
9843 If enabled, in case of control speculative load with no speculatively
9844 scheduled dependent instructions this load will be emitted as ld.sa and
9845 ld.c will be used to check it.
9846 The default is 'disable'.
9848 @item -mno-sched-spec-verbose
9849 @itemx -msched-spec-verbose
9850 @opindex mno-sched-spec-verbose
9851 @opindex msched-spec-verbose
9852 (Dis/En)able printing of the information about speculative motions.
9854 @item -mno-sched-prefer-non-data-spec-insns
9855 @itemx -msched-prefer-non-data-spec-insns
9856 @opindex mno-sched-prefer-non-data-spec-insns
9857 @opindex msched-prefer-non-data-spec-insns
9858 If enabled, data speculative instructions will be chosen for schedule
9859 only if there are no other choices at the moment. This will make
9860 the use of the data speculation much more conservative.
9861 The default is 'disable'.
9863 @item -mno-sched-prefer-non-control-spec-insns
9864 @itemx -msched-prefer-non-control-spec-insns
9865 @opindex mno-sched-prefer-non-control-spec-insns
9866 @opindex msched-prefer-non-control-spec-insns
9867 If enabled, control speculative instructions will be chosen for schedule
9868 only if there are no other choices at the moment. This will make
9869 the use of the control speculation much more conservative.
9870 The default is 'disable'.
9872 @item -mno-sched-count-spec-in-critical-path
9873 @itemx -msched-count-spec-in-critical-path
9874 @opindex mno-sched-count-spec-in-critical-path
9875 @opindex msched-count-spec-in-critical-path
9876 If enabled, speculative dependencies will be considered during
9877 computation of the instructions priorities. This will make the use of the
9878 speculation a bit more conservative.
9879 The default is 'disable'.
9884 @subsection M32C Options
9885 @cindex M32C options
9888 @item -mcpu=@var{name}
9890 Select the CPU for which code is generated. @var{name} may be one of
9891 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
9892 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
9897 Specifies that the program will be run on the simulator. This causes
9898 an alternate runtime library to be linked in which supports, for
9899 example, file I/O. You must not use this option when generating
9900 programs that will run on real hardware; you must provide your own
9901 runtime library for whatever I/O functions are needed.
9903 @item -memregs=@var{number}
9905 Specifies the number of memory-based pseudo-registers GCC will use
9906 during code generation. These pseudo-registers will be used like real
9907 registers, so there is a tradeoff between GCC's ability to fit the
9908 code into available registers, and the performance penalty of using
9909 memory instead of registers. Note that all modules in a program must
9910 be compiled with the same value for this option. Because of that, you
9911 must not use this option with the default runtime libraries gcc
9916 @node M32R/D Options
9917 @subsection M32R/D Options
9918 @cindex M32R/D options
9920 These @option{-m} options are defined for Renesas M32R/D architectures:
9925 Generate code for the M32R/2@.
9929 Generate code for the M32R/X@.
9933 Generate code for the M32R@. This is the default.
9936 @opindex mmodel=small
9937 Assume all objects live in the lower 16MB of memory (so that their addresses
9938 can be loaded with the @code{ld24} instruction), and assume all subroutines
9939 are reachable with the @code{bl} instruction.
9940 This is the default.
9942 The addressability of a particular object can be set with the
9943 @code{model} attribute.
9945 @item -mmodel=medium
9946 @opindex mmodel=medium
9947 Assume objects may be anywhere in the 32-bit address space (the compiler
9948 will generate @code{seth/add3} instructions to load their addresses), and
9949 assume all subroutines are reachable with the @code{bl} instruction.
9952 @opindex mmodel=large
9953 Assume objects may be anywhere in the 32-bit address space (the compiler
9954 will generate @code{seth/add3} instructions to load their addresses), and
9955 assume subroutines may not be reachable with the @code{bl} instruction
9956 (the compiler will generate the much slower @code{seth/add3/jl}
9957 instruction sequence).
9960 @opindex msdata=none
9961 Disable use of the small data area. Variables will be put into
9962 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
9963 @code{section} attribute has been specified).
9964 This is the default.
9966 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
9967 Objects may be explicitly put in the small data area with the
9968 @code{section} attribute using one of these sections.
9971 @opindex msdata=sdata
9972 Put small global and static data in the small data area, but do not
9973 generate special code to reference them.
9977 Put small global and static data in the small data area, and generate
9978 special instructions to reference them.
9982 @cindex smaller data references
9983 Put global and static objects less than or equal to @var{num} bytes
9984 into the small data or bss sections instead of the normal data or bss
9985 sections. The default value of @var{num} is 8.
9986 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
9987 for this option to have any effect.
9989 All modules should be compiled with the same @option{-G @var{num}} value.
9990 Compiling with different values of @var{num} may or may not work; if it
9991 doesn't the linker will give an error message---incorrect code will not be
9996 Makes the M32R specific code in the compiler display some statistics
9997 that might help in debugging programs.
10000 @opindex malign-loops
10001 Align all loops to a 32-byte boundary.
10003 @item -mno-align-loops
10004 @opindex mno-align-loops
10005 Do not enforce a 32-byte alignment for loops. This is the default.
10007 @item -missue-rate=@var{number}
10008 @opindex missue-rate=@var{number}
10009 Issue @var{number} instructions per cycle. @var{number} can only be 1
10012 @item -mbranch-cost=@var{number}
10013 @opindex mbranch-cost=@var{number}
10014 @var{number} can only be 1 or 2. If it is 1 then branches will be
10015 preferred over conditional code, if it is 2, then the opposite will
10018 @item -mflush-trap=@var{number}
10019 @opindex mflush-trap=@var{number}
10020 Specifies the trap number to use to flush the cache. The default is
10021 12. Valid numbers are between 0 and 15 inclusive.
10023 @item -mno-flush-trap
10024 @opindex mno-flush-trap
10025 Specifies that the cache cannot be flushed by using a trap.
10027 @item -mflush-func=@var{name}
10028 @opindex mflush-func=@var{name}
10029 Specifies the name of the operating system function to call to flush
10030 the cache. The default is @emph{_flush_cache}, but a function call
10031 will only be used if a trap is not available.
10033 @item -mno-flush-func
10034 @opindex mno-flush-func
10035 Indicates that there is no OS function for flushing the cache.
10039 @node M680x0 Options
10040 @subsection M680x0 Options
10041 @cindex M680x0 options
10043 These are the @samp{-m} options defined for the 68000 series. The default
10044 values for these options depends on which style of 68000 was selected when
10045 the compiler was configured; the defaults for the most common choices are
10053 Generate output for a 68000. This is the default
10054 when the compiler is configured for 68000-based systems.
10056 Use this option for microcontrollers with a 68000 or EC000 core,
10057 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
10063 Generate output for a 68020. This is the default
10064 when the compiler is configured for 68020-based systems.
10068 Generate output containing 68881 instructions for floating point.
10069 This is the default for most 68020 systems unless @option{--nfp} was
10070 specified when the compiler was configured.
10074 Generate output for a 68030. This is the default when the compiler is
10075 configured for 68030-based systems.
10079 Generate output for a 68040. This is the default when the compiler is
10080 configured for 68040-based systems.
10082 This option inhibits the use of 68881/68882 instructions that have to be
10083 emulated by software on the 68040. Use this option if your 68040 does not
10084 have code to emulate those instructions.
10088 Generate output for a 68060. This is the default when the compiler is
10089 configured for 68060-based systems.
10091 This option inhibits the use of 68020 and 68881/68882 instructions that
10092 have to be emulated by software on the 68060. Use this option if your 68060
10093 does not have code to emulate those instructions.
10097 Generate output for a CPU32. This is the default
10098 when the compiler is configured for CPU32-based systems.
10100 Use this option for microcontrollers with a
10101 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
10102 68336, 68340, 68341, 68349 and 68360.
10106 Generate output for a 520X ``coldfire'' family cpu. This is the default
10107 when the compiler is configured for 520X-based systems.
10109 Use this option for microcontroller with a 5200 core, including
10110 the MCF5202, MCF5203, MCF5204 and MCF5202.
10114 Generate output for a ColdFire V4e family cpu (e.g.@: 547x/548x).
10115 This includes use of hardware floating point instructions.
10119 Generate output for a 68040, without using any of the new instructions.
10120 This results in code which can run relatively efficiently on either a
10121 68020/68881 or a 68030 or a 68040. The generated code does use the
10122 68881 instructions that are emulated on the 68040.
10126 Generate output for a 68060, without using any of the new instructions.
10127 This results in code which can run relatively efficiently on either a
10128 68020/68881 or a 68030 or a 68040. The generated code does use the
10129 68881 instructions that are emulated on the 68060.
10132 @opindex msoft-float
10133 Generate output containing library calls for floating point.
10134 @strong{Warning:} the requisite libraries are not available for all m68k
10135 targets. Normally the facilities of the machine's usual C compiler are
10136 used, but this can't be done directly in cross-compilation. You must
10137 make your own arrangements to provide suitable library functions for
10138 cross-compilation. The embedded targets @samp{m68k-*-aout} and
10139 @samp{m68k-*-coff} do provide software floating point support.
10143 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10144 Additionally, parameters passed on the stack are also aligned to a
10145 16-bit boundary even on targets whose API mandates promotion to 32-bit.
10148 @opindex mnobitfield
10149 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
10150 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
10154 Do use the bit-field instructions. The @option{-m68020} option implies
10155 @option{-mbitfield}. This is the default if you use a configuration
10156 designed for a 68020.
10160 Use a different function-calling convention, in which functions
10161 that take a fixed number of arguments return with the @code{rtd}
10162 instruction, which pops their arguments while returning. This
10163 saves one instruction in the caller since there is no need to pop
10164 the arguments there.
10166 This calling convention is incompatible with the one normally
10167 used on Unix, so you cannot use it if you need to call libraries
10168 compiled with the Unix compiler.
10170 Also, you must provide function prototypes for all functions that
10171 take variable numbers of arguments (including @code{printf});
10172 otherwise incorrect code will be generated for calls to those
10175 In addition, seriously incorrect code will result if you call a
10176 function with too many arguments. (Normally, extra arguments are
10177 harmlessly ignored.)
10179 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
10180 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
10183 @itemx -mno-align-int
10184 @opindex malign-int
10185 @opindex mno-align-int
10186 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
10187 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
10188 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
10189 Aligning variables on 32-bit boundaries produces code that runs somewhat
10190 faster on processors with 32-bit busses at the expense of more memory.
10192 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
10193 align structures containing the above types differently than
10194 most published application binary interface specifications for the m68k.
10198 Use the pc-relative addressing mode of the 68000 directly, instead of
10199 using a global offset table. At present, this option implies @option{-fpic},
10200 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
10201 not presently supported with @option{-mpcrel}, though this could be supported for
10202 68020 and higher processors.
10204 @item -mno-strict-align
10205 @itemx -mstrict-align
10206 @opindex mno-strict-align
10207 @opindex mstrict-align
10208 Do not (do) assume that unaligned memory references will be handled by
10212 Generate code that allows the data segment to be located in a different
10213 area of memory from the text segment. This allows for execute in place in
10214 an environment without virtual memory management. This option implies
10217 @item -mno-sep-data
10218 Generate code that assumes that the data segment follows the text segment.
10219 This is the default.
10221 @item -mid-shared-library
10222 Generate code that supports shared libraries via the library ID method.
10223 This allows for execute in place and shared libraries in an environment
10224 without virtual memory management. This option implies @option{-fPIC}.
10226 @item -mno-id-shared-library
10227 Generate code that doesn't assume ID based shared libraries are being used.
10228 This is the default.
10230 @item -mshared-library-id=n
10231 Specified the identification number of the ID based shared library being
10232 compiled. Specifying a value of 0 will generate more compact code, specifying
10233 other values will force the allocation of that number to the current
10234 library but is no more space or time efficient than omitting this option.
10238 @node M68hc1x Options
10239 @subsection M68hc1x Options
10240 @cindex M68hc1x options
10242 These are the @samp{-m} options defined for the 68hc11 and 68hc12
10243 microcontrollers. The default values for these options depends on
10244 which style of microcontroller was selected when the compiler was configured;
10245 the defaults for the most common choices are given below.
10252 Generate output for a 68HC11. This is the default
10253 when the compiler is configured for 68HC11-based systems.
10259 Generate output for a 68HC12. This is the default
10260 when the compiler is configured for 68HC12-based systems.
10266 Generate output for a 68HCS12.
10268 @item -mauto-incdec
10269 @opindex mauto-incdec
10270 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
10277 Enable the use of 68HC12 min and max instructions.
10280 @itemx -mno-long-calls
10281 @opindex mlong-calls
10282 @opindex mno-long-calls
10283 Treat all calls as being far away (near). If calls are assumed to be
10284 far away, the compiler will use the @code{call} instruction to
10285 call a function and the @code{rtc} instruction for returning.
10289 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10291 @item -msoft-reg-count=@var{count}
10292 @opindex msoft-reg-count
10293 Specify the number of pseudo-soft registers which are used for the
10294 code generation. The maximum number is 32. Using more pseudo-soft
10295 register may or may not result in better code depending on the program.
10296 The default is 4 for 68HC11 and 2 for 68HC12.
10300 @node MCore Options
10301 @subsection MCore Options
10302 @cindex MCore options
10304 These are the @samp{-m} options defined for the Motorola M*Core
10310 @itemx -mno-hardlit
10312 @opindex mno-hardlit
10313 Inline constants into the code stream if it can be done in two
10314 instructions or less.
10320 Use the divide instruction. (Enabled by default).
10322 @item -mrelax-immediate
10323 @itemx -mno-relax-immediate
10324 @opindex mrelax-immediate
10325 @opindex mno-relax-immediate
10326 Allow arbitrary sized immediates in bit operations.
10328 @item -mwide-bitfields
10329 @itemx -mno-wide-bitfields
10330 @opindex mwide-bitfields
10331 @opindex mno-wide-bitfields
10332 Always treat bit-fields as int-sized.
10334 @item -m4byte-functions
10335 @itemx -mno-4byte-functions
10336 @opindex m4byte-functions
10337 @opindex mno-4byte-functions
10338 Force all functions to be aligned to a four byte boundary.
10340 @item -mcallgraph-data
10341 @itemx -mno-callgraph-data
10342 @opindex mcallgraph-data
10343 @opindex mno-callgraph-data
10344 Emit callgraph information.
10347 @itemx -mno-slow-bytes
10348 @opindex mslow-bytes
10349 @opindex mno-slow-bytes
10350 Prefer word access when reading byte quantities.
10352 @item -mlittle-endian
10353 @itemx -mbig-endian
10354 @opindex mlittle-endian
10355 @opindex mbig-endian
10356 Generate code for a little endian target.
10362 Generate code for the 210 processor.
10366 @subsection MIPS Options
10367 @cindex MIPS options
10373 Generate big-endian code.
10377 Generate little-endian code. This is the default for @samp{mips*el-*-*}
10380 @item -march=@var{arch}
10382 Generate code that will run on @var{arch}, which can be the name of a
10383 generic MIPS ISA, or the name of a particular processor.
10385 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
10386 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
10387 The processor names are:
10388 @samp{4kc}, @samp{4km}, @samp{4kp},
10389 @samp{5kc}, @samp{5kf},
10391 @samp{24k}, @samp{24kc}, @samp{24kf}, @samp{24kx},
10394 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
10395 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
10396 @samp{rm7000}, @samp{rm9000},
10399 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
10400 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
10401 The special value @samp{from-abi} selects the
10402 most compatible architecture for the selected ABI (that is,
10403 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
10405 In processor names, a final @samp{000} can be abbreviated as @samp{k}
10406 (for example, @samp{-march=r2k}). Prefixes are optional, and
10407 @samp{vr} may be written @samp{r}.
10409 GCC defines two macros based on the value of this option. The first
10410 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
10411 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
10412 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
10413 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
10414 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
10416 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
10417 above. In other words, it will have the full prefix and will not
10418 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
10419 the macro names the resolved architecture (either @samp{"mips1"} or
10420 @samp{"mips3"}). It names the default architecture when no
10421 @option{-march} option is given.
10423 @item -mtune=@var{arch}
10425 Optimize for @var{arch}. Among other things, this option controls
10426 the way instructions are scheduled, and the perceived cost of arithmetic
10427 operations. The list of @var{arch} values is the same as for
10430 When this option is not used, GCC will optimize for the processor
10431 specified by @option{-march}. By using @option{-march} and
10432 @option{-mtune} together, it is possible to generate code that will
10433 run on a family of processors, but optimize the code for one
10434 particular member of that family.
10436 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
10437 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
10438 @samp{-march} ones described above.
10442 Equivalent to @samp{-march=mips1}.
10446 Equivalent to @samp{-march=mips2}.
10450 Equivalent to @samp{-march=mips3}.
10454 Equivalent to @samp{-march=mips4}.
10458 Equivalent to @samp{-march=mips32}.
10462 Equivalent to @samp{-march=mips32r2}.
10466 Equivalent to @samp{-march=mips64}.
10471 @opindex mno-mips16
10472 Generate (do not generate) MIPS16 code. If GCC is targetting a
10473 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
10485 Generate code for the given ABI@.
10487 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
10488 generates 64-bit code when you select a 64-bit architecture, but you
10489 can use @option{-mgp32} to get 32-bit code instead.
10491 For information about the O64 ABI, see
10492 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
10495 @itemx -mno-abicalls
10497 @opindex mno-abicalls
10498 Generate (do not generate) code that is suitable for SVR4-style
10499 dynamic objects. @option{-mabicalls} is the default for SVR4-based
10504 Generate (do not generate) code that is fully position-independent,
10505 and that can therefore be linked into shared libraries. This option
10506 only affects @option{-mabicalls}.
10508 All @option{-mabicalls} code has traditionally been position-independent,
10509 regardless of options like @option{-fPIC} and @option{-fpic}. However,
10510 as an extension, the GNU toolchain allows executables to use absolute
10511 accesses for locally-binding symbols. It can also use shorter GP
10512 initialization sequences and generate direct calls to locally-defined
10513 functions. This mode is selected by @option{-mno-shared}.
10515 @option{-mno-shared} depends on binutils 2.16 or higher and generates
10516 objects that can only be linked by the GNU linker. However, the option
10517 does not affect the ABI of the final executable; it only affects the ABI
10518 of relocatable objects. Using @option{-mno-shared} will generally make
10519 executables both smaller and quicker.
10521 @option{-mshared} is the default.
10527 Lift (do not lift) the usual restrictions on the size of the global
10530 GCC normally uses a single instruction to load values from the GOT@.
10531 While this is relatively efficient, it will only work if the GOT
10532 is smaller than about 64k. Anything larger will cause the linker
10533 to report an error such as:
10535 @cindex relocation truncated to fit (MIPS)
10537 relocation truncated to fit: R_MIPS_GOT16 foobar
10540 If this happens, you should recompile your code with @option{-mxgot}.
10541 It should then work with very large GOTs, although it will also be
10542 less efficient, since it will take three instructions to fetch the
10543 value of a global symbol.
10545 Note that some linkers can create multiple GOTs. If you have such a
10546 linker, you should only need to use @option{-mxgot} when a single object
10547 file accesses more than 64k's worth of GOT entries. Very few do.
10549 These options have no effect unless GCC is generating position
10554 Assume that general-purpose registers are 32 bits wide.
10558 Assume that general-purpose registers are 64 bits wide.
10562 Assume that floating-point registers are 32 bits wide.
10566 Assume that floating-point registers are 64 bits wide.
10569 @opindex mhard-float
10570 Use floating-point coprocessor instructions.
10573 @opindex msoft-float
10574 Do not use floating-point coprocessor instructions. Implement
10575 floating-point calculations using library calls instead.
10577 @item -msingle-float
10578 @opindex msingle-float
10579 Assume that the floating-point coprocessor only supports single-precision
10582 @itemx -mdouble-float
10583 @opindex mdouble-float
10584 Assume that the floating-point coprocessor supports double-precision
10585 operations. This is the default.
10591 Use (do not use) the MIPS DSP ASE. @xref{MIPS DSP Built-in Functions}.
10593 @itemx -mpaired-single
10594 @itemx -mno-paired-single
10595 @opindex mpaired-single
10596 @opindex mno-paired-single
10597 Use (do not use) paired-single floating-point instructions.
10598 @xref{MIPS Paired-Single Support}. This option can only be used
10599 when generating 64-bit code and requires hardware floating-point
10600 support to be enabled.
10605 @opindex mno-mips3d
10606 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
10607 The option @option{-mips3d} implies @option{-mpaired-single}.
10611 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
10612 an explanation of the default and the way that the pointer size is
10617 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
10619 The default size of @code{int}s, @code{long}s and pointers depends on
10620 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
10621 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
10622 32-bit @code{long}s. Pointers are the same size as @code{long}s,
10623 or the same size as integer registers, whichever is smaller.
10629 Assume (do not assume) that all symbols have 32-bit values, regardless
10630 of the selected ABI@. This option is useful in combination with
10631 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
10632 to generate shorter and faster references to symbolic addresses.
10636 @cindex smaller data references (MIPS)
10637 @cindex gp-relative references (MIPS)
10638 Put global and static items less than or equal to @var{num} bytes into
10639 the small data or bss section instead of the normal data or bss section.
10640 This allows the data to be accessed using a single instruction.
10642 All modules should be compiled with the same @option{-G @var{num}}
10645 @item -membedded-data
10646 @itemx -mno-embedded-data
10647 @opindex membedded-data
10648 @opindex mno-embedded-data
10649 Allocate variables to the read-only data section first if possible, then
10650 next in the small data section if possible, otherwise in data. This gives
10651 slightly slower code than the default, but reduces the amount of RAM required
10652 when executing, and thus may be preferred for some embedded systems.
10654 @item -muninit-const-in-rodata
10655 @itemx -mno-uninit-const-in-rodata
10656 @opindex muninit-const-in-rodata
10657 @opindex mno-uninit-const-in-rodata
10658 Put uninitialized @code{const} variables in the read-only data section.
10659 This option is only meaningful in conjunction with @option{-membedded-data}.
10661 @item -msplit-addresses
10662 @itemx -mno-split-addresses
10663 @opindex msplit-addresses
10664 @opindex mno-split-addresses
10665 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
10666 relocation operators. This option has been superseded by
10667 @option{-mexplicit-relocs} but is retained for backwards compatibility.
10669 @item -mexplicit-relocs
10670 @itemx -mno-explicit-relocs
10671 @opindex mexplicit-relocs
10672 @opindex mno-explicit-relocs
10673 Use (do not use) assembler relocation operators when dealing with symbolic
10674 addresses. The alternative, selected by @option{-mno-explicit-relocs},
10675 is to use assembler macros instead.
10677 @option{-mexplicit-relocs} is the default if GCC was configured
10678 to use an assembler that supports relocation operators.
10680 @item -mcheck-zero-division
10681 @itemx -mno-check-zero-division
10682 @opindex mcheck-zero-division
10683 @opindex mno-check-zero-division
10684 Trap (do not trap) on integer division by zero. The default is
10685 @option{-mcheck-zero-division}.
10687 @item -mdivide-traps
10688 @itemx -mdivide-breaks
10689 @opindex mdivide-traps
10690 @opindex mdivide-breaks
10691 MIPS systems check for division by zero by generating either a
10692 conditional trap or a break instruction. Using traps results in
10693 smaller code, but is only supported on MIPS II and later. Also, some
10694 versions of the Linux kernel have a bug that prevents trap from
10695 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
10696 allow conditional traps on architectures that support them and
10697 @option{-mdivide-breaks} to force the use of breaks.
10699 The default is usually @option{-mdivide-traps}, but this can be
10700 overridden at configure time using @option{--with-divide=breaks}.
10701 Divide-by-zero checks can be completely disabled using
10702 @option{-mno-check-zero-division}.
10707 @opindex mno-memcpy
10708 Force (do not force) the use of @code{memcpy()} for non-trivial block
10709 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
10710 most constant-sized copies.
10713 @itemx -mno-long-calls
10714 @opindex mlong-calls
10715 @opindex mno-long-calls
10716 Disable (do not disable) use of the @code{jal} instruction. Calling
10717 functions using @code{jal} is more efficient but requires the caller
10718 and callee to be in the same 256 megabyte segment.
10720 This option has no effect on abicalls code. The default is
10721 @option{-mno-long-calls}.
10727 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
10728 instructions, as provided by the R4650 ISA@.
10731 @itemx -mno-fused-madd
10732 @opindex mfused-madd
10733 @opindex mno-fused-madd
10734 Enable (disable) use of the floating point multiply-accumulate
10735 instructions, when they are available. The default is
10736 @option{-mfused-madd}.
10738 When multiply-accumulate instructions are used, the intermediate
10739 product is calculated to infinite precision and is not subject to
10740 the FCSR Flush to Zero bit. This may be undesirable in some
10745 Tell the MIPS assembler to not run its preprocessor over user
10746 assembler files (with a @samp{.s} suffix) when assembling them.
10749 @itemx -mno-fix-r4000
10750 @opindex mfix-r4000
10751 @opindex mno-fix-r4000
10752 Work around certain R4000 CPU errata:
10755 A double-word or a variable shift may give an incorrect result if executed
10756 immediately after starting an integer division.
10758 A double-word or a variable shift may give an incorrect result if executed
10759 while an integer multiplication is in progress.
10761 An integer division may give an incorrect result if started in a delay slot
10762 of a taken branch or a jump.
10766 @itemx -mno-fix-r4400
10767 @opindex mfix-r4400
10768 @opindex mno-fix-r4400
10769 Work around certain R4400 CPU errata:
10772 A double-word or a variable shift may give an incorrect result if executed
10773 immediately after starting an integer division.
10777 @itemx -mno-fix-vr4120
10778 @opindex mfix-vr4120
10779 Work around certain VR4120 errata:
10782 @code{dmultu} does not always produce the correct result.
10784 @code{div} and @code{ddiv} do not always produce the correct result if one
10785 of the operands is negative.
10787 The workarounds for the division errata rely on special functions in
10788 @file{libgcc.a}. At present, these functions are only provided by
10789 the @code{mips64vr*-elf} configurations.
10791 Other VR4120 errata require a nop to be inserted between certain pairs of
10792 instructions. These errata are handled by the assembler, not by GCC itself.
10795 @opindex mfix-vr4130
10796 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
10797 workarounds are implemented by the assembler rather than by GCC,
10798 although GCC will avoid using @code{mflo} and @code{mfhi} if the
10799 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
10800 instructions are available instead.
10803 @itemx -mno-fix-sb1
10805 Work around certain SB-1 CPU core errata.
10806 (This flag currently works around the SB-1 revision 2
10807 ``F1'' and ``F2'' floating point errata.)
10809 @item -mflush-func=@var{func}
10810 @itemx -mno-flush-func
10811 @opindex mflush-func
10812 Specifies the function to call to flush the I and D caches, or to not
10813 call any such function. If called, the function must take the same
10814 arguments as the common @code{_flush_func()}, that is, the address of the
10815 memory range for which the cache is being flushed, the size of the
10816 memory range, and the number 3 (to flush both caches). The default
10817 depends on the target GCC was configured for, but commonly is either
10818 @samp{_flush_func} or @samp{__cpu_flush}.
10820 @item -mbranch-likely
10821 @itemx -mno-branch-likely
10822 @opindex mbranch-likely
10823 @opindex mno-branch-likely
10824 Enable or disable use of Branch Likely instructions, regardless of the
10825 default for the selected architecture. By default, Branch Likely
10826 instructions may be generated if they are supported by the selected
10827 architecture. An exception is for the MIPS32 and MIPS64 architectures
10828 and processors which implement those architectures; for those, Branch
10829 Likely instructions will not be generated by default because the MIPS32
10830 and MIPS64 architectures specifically deprecate their use.
10832 @item -mfp-exceptions
10833 @itemx -mno-fp-exceptions
10834 @opindex mfp-exceptions
10835 Specifies whether FP exceptions are enabled. This affects how we schedule
10836 FP instructions for some processors. The default is that FP exceptions are
10839 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
10840 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
10843 @item -mvr4130-align
10844 @itemx -mno-vr4130-align
10845 @opindex mvr4130-align
10846 The VR4130 pipeline is two-way superscalar, but can only issue two
10847 instructions together if the first one is 8-byte aligned. When this
10848 option is enabled, GCC will align pairs of instructions that it
10849 thinks should execute in parallel.
10851 This option only has an effect when optimizing for the VR4130.
10852 It normally makes code faster, but at the expense of making it bigger.
10853 It is enabled by default at optimization level @option{-O3}.
10857 @subsection MMIX Options
10858 @cindex MMIX Options
10860 These options are defined for the MMIX:
10864 @itemx -mno-libfuncs
10866 @opindex mno-libfuncs
10867 Specify that intrinsic library functions are being compiled, passing all
10868 values in registers, no matter the size.
10871 @itemx -mno-epsilon
10873 @opindex mno-epsilon
10874 Generate floating-point comparison instructions that compare with respect
10875 to the @code{rE} epsilon register.
10877 @item -mabi=mmixware
10879 @opindex mabi-mmixware
10881 Generate code that passes function parameters and return values that (in
10882 the called function) are seen as registers @code{$0} and up, as opposed to
10883 the GNU ABI which uses global registers @code{$231} and up.
10885 @item -mzero-extend
10886 @itemx -mno-zero-extend
10887 @opindex mzero-extend
10888 @opindex mno-zero-extend
10889 When reading data from memory in sizes shorter than 64 bits, use (do not
10890 use) zero-extending load instructions by default, rather than
10891 sign-extending ones.
10894 @itemx -mno-knuthdiv
10896 @opindex mno-knuthdiv
10897 Make the result of a division yielding a remainder have the same sign as
10898 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10899 remainder follows the sign of the dividend. Both methods are
10900 arithmetically valid, the latter being almost exclusively used.
10902 @item -mtoplevel-symbols
10903 @itemx -mno-toplevel-symbols
10904 @opindex mtoplevel-symbols
10905 @opindex mno-toplevel-symbols
10906 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10907 code can be used with the @code{PREFIX} assembly directive.
10911 Generate an executable in the ELF format, rather than the default
10912 @samp{mmo} format used by the @command{mmix} simulator.
10914 @item -mbranch-predict
10915 @itemx -mno-branch-predict
10916 @opindex mbranch-predict
10917 @opindex mno-branch-predict
10918 Use (do not use) the probable-branch instructions, when static branch
10919 prediction indicates a probable branch.
10921 @item -mbase-addresses
10922 @itemx -mno-base-addresses
10923 @opindex mbase-addresses
10924 @opindex mno-base-addresses
10925 Generate (do not generate) code that uses @emph{base addresses}. Using a
10926 base address automatically generates a request (handled by the assembler
10927 and the linker) for a constant to be set up in a global register. The
10928 register is used for one or more base address requests within the range 0
10929 to 255 from the value held in the register. The generally leads to short
10930 and fast code, but the number of different data items that can be
10931 addressed is limited. This means that a program that uses lots of static
10932 data may require @option{-mno-base-addresses}.
10934 @item -msingle-exit
10935 @itemx -mno-single-exit
10936 @opindex msingle-exit
10937 @opindex mno-single-exit
10938 Force (do not force) generated code to have a single exit point in each
10942 @node MN10300 Options
10943 @subsection MN10300 Options
10944 @cindex MN10300 options
10946 These @option{-m} options are defined for Matsushita MN10300 architectures:
10951 Generate code to avoid bugs in the multiply instructions for the MN10300
10952 processors. This is the default.
10954 @item -mno-mult-bug
10955 @opindex mno-mult-bug
10956 Do not generate code to avoid bugs in the multiply instructions for the
10957 MN10300 processors.
10961 Generate code which uses features specific to the AM33 processor.
10965 Do not generate code which uses features specific to the AM33 processor. This
10968 @item -mreturn-pointer-on-d0
10969 @opindex mreturn-pointer-on-d0
10970 When generating a function which returns a pointer, return the pointer
10971 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
10972 only in a0, and attempts to call such functions without a prototype
10973 would result in errors. Note that this option is on by default; use
10974 @option{-mno-return-pointer-on-d0} to disable it.
10978 Do not link in the C run-time initialization object file.
10982 Indicate to the linker that it should perform a relaxation optimization pass
10983 to shorten branches, calls and absolute memory addresses. This option only
10984 has an effect when used on the command line for the final link step.
10986 This option makes symbolic debugging impossible.
10990 @subsection MT Options
10993 These @option{-m} options are defined for Morpho MT architectures:
10997 @item -march=@var{cpu-type}
10999 Generate code that will run on @var{cpu-type}, which is the name of a system
11000 representing a certain processor type. Possible values for
11001 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
11002 @samp{ms1-16-003} and @samp{ms2}.
11004 When this option is not used, the default is @option{-march=ms1-16-002}.
11008 Use byte loads and stores when generating code.
11012 Do not use byte loads and stores when generating code.
11016 Use simulator runtime
11020 Do not link in the C run-time initialization object file
11021 @file{crti.o}. Other run-time initialization and termination files
11022 such as @file{startup.o} and @file{exit.o} are still included on the
11023 linker command line.
11027 @node PDP-11 Options
11028 @subsection PDP-11 Options
11029 @cindex PDP-11 Options
11031 These options are defined for the PDP-11:
11036 Use hardware FPP floating point. This is the default. (FIS floating
11037 point on the PDP-11/40 is not supported.)
11040 @opindex msoft-float
11041 Do not use hardware floating point.
11045 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
11049 Return floating-point results in memory. This is the default.
11053 Generate code for a PDP-11/40.
11057 Generate code for a PDP-11/45. This is the default.
11061 Generate code for a PDP-11/10.
11063 @item -mbcopy-builtin
11064 @opindex bcopy-builtin
11065 Use inline @code{movmemhi} patterns for copying memory. This is the
11070 Do not use inline @code{movmemhi} patterns for copying memory.
11076 Use 16-bit @code{int}. This is the default.
11082 Use 32-bit @code{int}.
11085 @itemx -mno-float32
11087 @opindex mno-float32
11088 Use 64-bit @code{float}. This is the default.
11091 @itemx -mno-float64
11093 @opindex mno-float64
11094 Use 32-bit @code{float}.
11098 Use @code{abshi2} pattern. This is the default.
11102 Do not use @code{abshi2} pattern.
11104 @item -mbranch-expensive
11105 @opindex mbranch-expensive
11106 Pretend that branches are expensive. This is for experimenting with
11107 code generation only.
11109 @item -mbranch-cheap
11110 @opindex mbranch-cheap
11111 Do not pretend that branches are expensive. This is the default.
11115 Generate code for a system with split I&D@.
11119 Generate code for a system without split I&D@. This is the default.
11123 Use Unix assembler syntax. This is the default when configured for
11124 @samp{pdp11-*-bsd}.
11128 Use DEC assembler syntax. This is the default when configured for any
11129 PDP-11 target other than @samp{pdp11-*-bsd}.
11132 @node PowerPC Options
11133 @subsection PowerPC Options
11134 @cindex PowerPC options
11136 These are listed under @xref{RS/6000 and PowerPC Options}.
11138 @node RS/6000 and PowerPC Options
11139 @subsection IBM RS/6000 and PowerPC Options
11140 @cindex RS/6000 and PowerPC Options
11141 @cindex IBM RS/6000 and PowerPC Options
11143 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
11150 @itemx -mno-powerpc
11151 @itemx -mpowerpc-gpopt
11152 @itemx -mno-powerpc-gpopt
11153 @itemx -mpowerpc-gfxopt
11154 @itemx -mno-powerpc-gfxopt
11156 @itemx -mno-powerpc64
11160 @itemx -mno-popcntb
11166 @opindex mno-power2
11168 @opindex mno-powerpc
11169 @opindex mpowerpc-gpopt
11170 @opindex mno-powerpc-gpopt
11171 @opindex mpowerpc-gfxopt
11172 @opindex mno-powerpc-gfxopt
11173 @opindex mpowerpc64
11174 @opindex mno-powerpc64
11178 @opindex mno-popcntb
11181 GCC supports two related instruction set architectures for the
11182 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
11183 instructions supported by the @samp{rios} chip set used in the original
11184 RS/6000 systems and the @dfn{PowerPC} instruction set is the
11185 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
11186 the IBM 4xx, 6xx, and follow-on microprocessors.
11188 Neither architecture is a subset of the other. However there is a
11189 large common subset of instructions supported by both. An MQ
11190 register is included in processors supporting the POWER architecture.
11192 You use these options to specify which instructions are available on the
11193 processor you are using. The default value of these options is
11194 determined when configuring GCC@. Specifying the
11195 @option{-mcpu=@var{cpu_type}} overrides the specification of these
11196 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
11197 rather than the options listed above.
11199 The @option{-mpower} option allows GCC to generate instructions that
11200 are found only in the POWER architecture and to use the MQ register.
11201 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
11202 to generate instructions that are present in the POWER2 architecture but
11203 not the original POWER architecture.
11205 The @option{-mpowerpc} option allows GCC to generate instructions that
11206 are found only in the 32-bit subset of the PowerPC architecture.
11207 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
11208 GCC to use the optional PowerPC architecture instructions in the
11209 General Purpose group, including floating-point square root. Specifying
11210 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
11211 use the optional PowerPC architecture instructions in the Graphics
11212 group, including floating-point select.
11214 The @option{-mmfcrf} option allows GCC to generate the move from
11215 condition register field instruction implemented on the POWER4
11216 processor and other processors that support the PowerPC V2.01
11218 The @option{-mpopcntb} option allows GCC to generate the popcount and
11219 double precision FP reciprocal estimate instruction implemented on the
11220 POWER5 processor and other processors that support the PowerPC V2.02
11222 The @option{-mfprnd} option allows GCC to generate the FP round to
11223 integer instructions implemented on the POWER5+ processor and other
11224 processors that support the PowerPC V2.03 architecture.
11226 The @option{-mpowerpc64} option allows GCC to generate the additional
11227 64-bit instructions that are found in the full PowerPC64 architecture
11228 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
11229 @option{-mno-powerpc64}.
11231 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
11232 will use only the instructions in the common subset of both
11233 architectures plus some special AIX common-mode calls, and will not use
11234 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
11235 permits GCC to use any instruction from either architecture and to
11236 allow use of the MQ register; specify this for the Motorola MPC601.
11238 @item -mnew-mnemonics
11239 @itemx -mold-mnemonics
11240 @opindex mnew-mnemonics
11241 @opindex mold-mnemonics
11242 Select which mnemonics to use in the generated assembler code. With
11243 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
11244 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
11245 assembler mnemonics defined for the POWER architecture. Instructions
11246 defined in only one architecture have only one mnemonic; GCC uses that
11247 mnemonic irrespective of which of these options is specified.
11249 GCC defaults to the mnemonics appropriate for the architecture in
11250 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
11251 value of these option. Unless you are building a cross-compiler, you
11252 should normally not specify either @option{-mnew-mnemonics} or
11253 @option{-mold-mnemonics}, but should instead accept the default.
11255 @item -mcpu=@var{cpu_type}
11257 Set architecture type, register usage, choice of mnemonics, and
11258 instruction scheduling parameters for machine type @var{cpu_type}.
11259 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
11260 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
11261 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
11262 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
11263 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
11264 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
11265 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
11266 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6},
11267 @samp{common}, @samp{powerpc}, @samp{powerpc64},
11268 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
11270 @option{-mcpu=common} selects a completely generic processor. Code
11271 generated under this option will run on any POWER or PowerPC processor.
11272 GCC will use only the instructions in the common subset of both
11273 architectures, and will not use the MQ register. GCC assumes a generic
11274 processor model for scheduling purposes.
11276 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
11277 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
11278 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
11279 types, with an appropriate, generic processor model assumed for
11280 scheduling purposes.
11282 The other options specify a specific processor. Code generated under
11283 those options will run best on that processor, and may not run at all on
11286 The @option{-mcpu} options automatically enable or disable the
11287 following options: @option{-maltivec}, @option{-mfprnd},
11288 @option{-mhard-float}, @option{-mmfcrf}, @option{-mmultiple},
11289 @option{-mnew-mnemonics}, @option{-mpopcntb}, @option{-mpower},
11290 @option{-mpower2}, @option{-mpowerpc64}, @option{-mpowerpc-gpopt},
11291 @option{-mpowerpc-gfxopt}, @option{-mstring}, @option{-mmulhw}, @option{-mdlmzb}.
11292 The particular options
11293 set for any particular CPU will vary between compiler versions,
11294 depending on what setting seems to produce optimal code for that CPU;
11295 it doesn't necessarily reflect the actual hardware's capabilities. If
11296 you wish to set an individual option to a particular value, you may
11297 specify it after the @option{-mcpu} option, like @samp{-mcpu=970
11300 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
11301 not enabled or disabled by the @option{-mcpu} option at present because
11302 AIX does not have full support for these options. You may still
11303 enable or disable them individually if you're sure it'll work in your
11306 @item -mtune=@var{cpu_type}
11308 Set the instruction scheduling parameters for machine type
11309 @var{cpu_type}, but do not set the architecture type, register usage, or
11310 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
11311 values for @var{cpu_type} are used for @option{-mtune} as for
11312 @option{-mcpu}. If both are specified, the code generated will use the
11313 architecture, registers, and mnemonics set by @option{-mcpu}, but the
11314 scheduling parameters set by @option{-mtune}.
11320 Generate code to compute division as reciprocal estimate and iterative
11321 refinement, creating opportunities for increased throughput. This
11322 feature requires: optional PowerPC Graphics instruction set for single
11323 precision and FRE instruction for double precision, assuming divides
11324 cannot generate user-visible traps, and the domain values not include
11325 Infinities, denormals or zero denominator.
11328 @itemx -mno-altivec
11330 @opindex mno-altivec
11331 Generate code that uses (does not use) AltiVec instructions, and also
11332 enable the use of built-in functions that allow more direct access to
11333 the AltiVec instruction set. You may also need to set
11334 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
11340 @opindex mno-vrsave
11341 Generate VRSAVE instructions when generating AltiVec code.
11344 @opindex msecure-plt
11345 Generate code that allows ld and ld.so to build executables and shared
11346 libraries with non-exec .plt and .got sections. This is a PowerPC
11347 32-bit SYSV ABI option.
11351 Generate code that uses a BSS .plt section that ld.so fills in, and
11352 requires .plt and .got sections that are both writable and executable.
11353 This is a PowerPC 32-bit SYSV ABI option.
11359 This switch enables or disables the generation of ISEL instructions.
11361 @item -misel=@var{yes/no}
11362 This switch has been deprecated. Use @option{-misel} and
11363 @option{-mno-isel} instead.
11369 This switch enables or disables the generation of SPE simd
11372 @item -mspe=@var{yes/no}
11373 This option has been deprecated. Use @option{-mspe} and
11374 @option{-mno-spe} instead.
11376 @item -mfloat-gprs=@var{yes/single/double/no}
11377 @itemx -mfloat-gprs
11378 @opindex mfloat-gprs
11379 This switch enables or disables the generation of floating point
11380 operations on the general purpose registers for architectures that
11383 The argument @var{yes} or @var{single} enables the use of
11384 single-precision floating point operations.
11386 The argument @var{double} enables the use of single and
11387 double-precision floating point operations.
11389 The argument @var{no} disables floating point operations on the
11390 general purpose registers.
11392 This option is currently only available on the MPC854x.
11398 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
11399 targets (including GNU/Linux). The 32-bit environment sets int, long
11400 and pointer to 32 bits and generates code that runs on any PowerPC
11401 variant. The 64-bit environment sets int to 32 bits and long and
11402 pointer to 64 bits, and generates code for PowerPC64, as for
11403 @option{-mpowerpc64}.
11406 @itemx -mno-fp-in-toc
11407 @itemx -mno-sum-in-toc
11408 @itemx -mminimal-toc
11410 @opindex mno-fp-in-toc
11411 @opindex mno-sum-in-toc
11412 @opindex mminimal-toc
11413 Modify generation of the TOC (Table Of Contents), which is created for
11414 every executable file. The @option{-mfull-toc} option is selected by
11415 default. In that case, GCC will allocate at least one TOC entry for
11416 each unique non-automatic variable reference in your program. GCC
11417 will also place floating-point constants in the TOC@. However, only
11418 16,384 entries are available in the TOC@.
11420 If you receive a linker error message that saying you have overflowed
11421 the available TOC space, you can reduce the amount of TOC space used
11422 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
11423 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
11424 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
11425 generate code to calculate the sum of an address and a constant at
11426 run-time instead of putting that sum into the TOC@. You may specify one
11427 or both of these options. Each causes GCC to produce very slightly
11428 slower and larger code at the expense of conserving TOC space.
11430 If you still run out of space in the TOC even when you specify both of
11431 these options, specify @option{-mminimal-toc} instead. This option causes
11432 GCC to make only one TOC entry for every file. When you specify this
11433 option, GCC will produce code that is slower and larger but which
11434 uses extremely little TOC space. You may wish to use this option
11435 only on files that contain less frequently executed code.
11441 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
11442 @code{long} type, and the infrastructure needed to support them.
11443 Specifying @option{-maix64} implies @option{-mpowerpc64} and
11444 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
11445 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
11448 @itemx -mno-xl-compat
11449 @opindex mxl-compat
11450 @opindex mno-xl-compat
11451 Produce code that conforms more closely to IBM XL compiler semantics
11452 when using AIX-compatible ABI. Pass floating-point arguments to
11453 prototyped functions beyond the register save area (RSA) on the stack
11454 in addition to argument FPRs. Do not assume that most significant
11455 double in 128-bit long double value is properly rounded when comparing
11456 values and converting to double. Use XL symbol names for long double
11459 The AIX calling convention was extended but not initially documented to
11460 handle an obscure K&R C case of calling a function that takes the
11461 address of its arguments with fewer arguments than declared. IBM XL
11462 compilers access floating point arguments which do not fit in the
11463 RSA from the stack when a subroutine is compiled without
11464 optimization. Because always storing floating-point arguments on the
11465 stack is inefficient and rarely needed, this option is not enabled by
11466 default and only is necessary when calling subroutines compiled by IBM
11467 XL compilers without optimization.
11471 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
11472 application written to use message passing with special startup code to
11473 enable the application to run. The system must have PE installed in the
11474 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
11475 must be overridden with the @option{-specs=} option to specify the
11476 appropriate directory location. The Parallel Environment does not
11477 support threads, so the @option{-mpe} option and the @option{-pthread}
11478 option are incompatible.
11480 @item -malign-natural
11481 @itemx -malign-power
11482 @opindex malign-natural
11483 @opindex malign-power
11484 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
11485 @option{-malign-natural} overrides the ABI-defined alignment of larger
11486 types, such as floating-point doubles, on their natural size-based boundary.
11487 The option @option{-malign-power} instructs GCC to follow the ABI-specified
11488 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
11490 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
11494 @itemx -mhard-float
11495 @opindex msoft-float
11496 @opindex mhard-float
11497 Generate code that does not use (uses) the floating-point register set.
11498 Software floating point emulation is provided if you use the
11499 @option{-msoft-float} option, and pass the option to GCC when linking.
11502 @itemx -mno-multiple
11504 @opindex mno-multiple
11505 Generate code that uses (does not use) the load multiple word
11506 instructions and the store multiple word instructions. These
11507 instructions are generated by default on POWER systems, and not
11508 generated on PowerPC systems. Do not use @option{-mmultiple} on little
11509 endian PowerPC systems, since those instructions do not work when the
11510 processor is in little endian mode. The exceptions are PPC740 and
11511 PPC750 which permit the instructions usage in little endian mode.
11516 @opindex mno-string
11517 Generate code that uses (does not use) the load string instructions
11518 and the store string word instructions to save multiple registers and
11519 do small block moves. These instructions are generated by default on
11520 POWER systems, and not generated on PowerPC systems. Do not use
11521 @option{-mstring} on little endian PowerPC systems, since those
11522 instructions do not work when the processor is in little endian mode.
11523 The exceptions are PPC740 and PPC750 which permit the instructions
11524 usage in little endian mode.
11529 @opindex mno-update
11530 Generate code that uses (does not use) the load or store instructions
11531 that update the base register to the address of the calculated memory
11532 location. These instructions are generated by default. If you use
11533 @option{-mno-update}, there is a small window between the time that the
11534 stack pointer is updated and the address of the previous frame is
11535 stored, which means code that walks the stack frame across interrupts or
11536 signals may get corrupted data.
11539 @itemx -mno-fused-madd
11540 @opindex mfused-madd
11541 @opindex mno-fused-madd
11542 Generate code that uses (does not use) the floating point multiply and
11543 accumulate instructions. These instructions are generated by default if
11544 hardware floating is used.
11550 Generate code that uses (does not use) the half-word multiply and
11551 multiply-accumulate instructions on the IBM 405 and 440 processors.
11552 These instructions are generated by default when targetting those
11559 Generate code that uses (does not use) the string-search @samp{dlmzb}
11560 instruction on the IBM 405 and 440 processors. This instruction is
11561 generated by default when targetting those processors.
11563 @item -mno-bit-align
11565 @opindex mno-bit-align
11566 @opindex mbit-align
11567 On System V.4 and embedded PowerPC systems do not (do) force structures
11568 and unions that contain bit-fields to be aligned to the base type of the
11571 For example, by default a structure containing nothing but 8
11572 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
11573 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
11574 the structure would be aligned to a 1 byte boundary and be one byte in
11577 @item -mno-strict-align
11578 @itemx -mstrict-align
11579 @opindex mno-strict-align
11580 @opindex mstrict-align
11581 On System V.4 and embedded PowerPC systems do not (do) assume that
11582 unaligned memory references will be handled by the system.
11584 @item -mrelocatable
11585 @itemx -mno-relocatable
11586 @opindex mrelocatable
11587 @opindex mno-relocatable
11588 On embedded PowerPC systems generate code that allows (does not allow)
11589 the program to be relocated to a different address at runtime. If you
11590 use @option{-mrelocatable} on any module, all objects linked together must
11591 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
11593 @item -mrelocatable-lib
11594 @itemx -mno-relocatable-lib
11595 @opindex mrelocatable-lib
11596 @opindex mno-relocatable-lib
11597 On embedded PowerPC systems generate code that allows (does not allow)
11598 the program to be relocated to a different address at runtime. Modules
11599 compiled with @option{-mrelocatable-lib} can be linked with either modules
11600 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
11601 with modules compiled with the @option{-mrelocatable} options.
11607 On System V.4 and embedded PowerPC systems do not (do) assume that
11608 register 2 contains a pointer to a global area pointing to the addresses
11609 used in the program.
11612 @itemx -mlittle-endian
11614 @opindex mlittle-endian
11615 On System V.4 and embedded PowerPC systems compile code for the
11616 processor in little endian mode. The @option{-mlittle-endian} option is
11617 the same as @option{-mlittle}.
11620 @itemx -mbig-endian
11622 @opindex mbig-endian
11623 On System V.4 and embedded PowerPC systems compile code for the
11624 processor in big endian mode. The @option{-mbig-endian} option is
11625 the same as @option{-mbig}.
11627 @item -mdynamic-no-pic
11628 @opindex mdynamic-no-pic
11629 On Darwin and Mac OS X systems, compile code so that it is not
11630 relocatable, but that its external references are relocatable. The
11631 resulting code is suitable for applications, but not shared
11634 @item -mprioritize-restricted-insns=@var{priority}
11635 @opindex mprioritize-restricted-insns
11636 This option controls the priority that is assigned to
11637 dispatch-slot restricted instructions during the second scheduling
11638 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
11639 @var{no/highest/second-highest} priority to dispatch slot restricted
11642 @item -msched-costly-dep=@var{dependence_type}
11643 @opindex msched-costly-dep
11644 This option controls which dependences are considered costly
11645 by the target during instruction scheduling. The argument
11646 @var{dependence_type} takes one of the following values:
11647 @var{no}: no dependence is costly,
11648 @var{all}: all dependences are costly,
11649 @var{true_store_to_load}: a true dependence from store to load is costly,
11650 @var{store_to_load}: any dependence from store to load is costly,
11651 @var{number}: any dependence which latency >= @var{number} is costly.
11653 @item -minsert-sched-nops=@var{scheme}
11654 @opindex minsert-sched-nops
11655 This option controls which nop insertion scheme will be used during
11656 the second scheduling pass. The argument @var{scheme} takes one of the
11658 @var{no}: Don't insert nops.
11659 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
11660 according to the scheduler's grouping.
11661 @var{regroup_exact}: Insert nops to force costly dependent insns into
11662 separate groups. Insert exactly as many nops as needed to force an insn
11663 to a new group, according to the estimated processor grouping.
11664 @var{number}: Insert nops to force costly dependent insns into
11665 separate groups. Insert @var{number} nops to force an insn to a new group.
11668 @opindex mcall-sysv
11669 On System V.4 and embedded PowerPC systems compile code using calling
11670 conventions that adheres to the March 1995 draft of the System V
11671 Application Binary Interface, PowerPC processor supplement. This is the
11672 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
11674 @item -mcall-sysv-eabi
11675 @opindex mcall-sysv-eabi
11676 Specify both @option{-mcall-sysv} and @option{-meabi} options.
11678 @item -mcall-sysv-noeabi
11679 @opindex mcall-sysv-noeabi
11680 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
11682 @item -mcall-solaris
11683 @opindex mcall-solaris
11684 On System V.4 and embedded PowerPC systems compile code for the Solaris
11688 @opindex mcall-linux
11689 On System V.4 and embedded PowerPC systems compile code for the
11690 Linux-based GNU system.
11694 On System V.4 and embedded PowerPC systems compile code for the
11695 Hurd-based GNU system.
11697 @item -mcall-netbsd
11698 @opindex mcall-netbsd
11699 On System V.4 and embedded PowerPC systems compile code for the
11700 NetBSD operating system.
11702 @item -maix-struct-return
11703 @opindex maix-struct-return
11704 Return all structures in memory (as specified by the AIX ABI)@.
11706 @item -msvr4-struct-return
11707 @opindex msvr4-struct-return
11708 Return structures smaller than 8 bytes in registers (as specified by the
11711 @item -mabi=@var{abi-type}
11713 Extend the current ABI with a particular extension, or remove such extension.
11714 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
11715 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
11719 Extend the current ABI with SPE ABI extensions. This does not change
11720 the default ABI, instead it adds the SPE ABI extensions to the current
11724 @opindex mabi=no-spe
11725 Disable Booke SPE ABI extensions for the current ABI@.
11727 @item -mabi=ibmlongdouble
11728 @opindex mabi=ibmlongdouble
11729 Change the current ABI to use IBM extended precision long double.
11730 This is a PowerPC 32-bit SYSV ABI option.
11732 @item -mabi=ieeelongdouble
11733 @opindex mabi=ieeelongdouble
11734 Change the current ABI to use IEEE extended precision long double.
11735 This is a PowerPC 32-bit Linux ABI option.
11738 @itemx -mno-prototype
11739 @opindex mprototype
11740 @opindex mno-prototype
11741 On System V.4 and embedded PowerPC systems assume that all calls to
11742 variable argument functions are properly prototyped. Otherwise, the
11743 compiler must insert an instruction before every non prototyped call to
11744 set or clear bit 6 of the condition code register (@var{CR}) to
11745 indicate whether floating point values were passed in the floating point
11746 registers in case the function takes a variable arguments. With
11747 @option{-mprototype}, only calls to prototyped variable argument functions
11748 will set or clear the bit.
11752 On embedded PowerPC systems, assume that the startup module is called
11753 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
11754 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
11759 On embedded PowerPC systems, assume that the startup module is called
11760 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
11765 On embedded PowerPC systems, assume that the startup module is called
11766 @file{crt0.o} and the standard C libraries are @file{libads.a} and
11769 @item -myellowknife
11770 @opindex myellowknife
11771 On embedded PowerPC systems, assume that the startup module is called
11772 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
11777 On System V.4 and embedded PowerPC systems, specify that you are
11778 compiling for a VxWorks system.
11782 Specify that you are compiling for the WindISS simulation environment.
11786 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
11787 header to indicate that @samp{eabi} extended relocations are used.
11793 On System V.4 and embedded PowerPC systems do (do not) adhere to the
11794 Embedded Applications Binary Interface (eabi) which is a set of
11795 modifications to the System V.4 specifications. Selecting @option{-meabi}
11796 means that the stack is aligned to an 8 byte boundary, a function
11797 @code{__eabi} is called to from @code{main} to set up the eabi
11798 environment, and the @option{-msdata} option can use both @code{r2} and
11799 @code{r13} to point to two separate small data areas. Selecting
11800 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
11801 do not call an initialization function from @code{main}, and the
11802 @option{-msdata} option will only use @code{r13} to point to a single
11803 small data area. The @option{-meabi} option is on by default if you
11804 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
11807 @opindex msdata=eabi
11808 On System V.4 and embedded PowerPC systems, put small initialized
11809 @code{const} global and static data in the @samp{.sdata2} section, which
11810 is pointed to by register @code{r2}. Put small initialized
11811 non-@code{const} global and static data in the @samp{.sdata} section,
11812 which is pointed to by register @code{r13}. Put small uninitialized
11813 global and static data in the @samp{.sbss} section, which is adjacent to
11814 the @samp{.sdata} section. The @option{-msdata=eabi} option is
11815 incompatible with the @option{-mrelocatable} option. The
11816 @option{-msdata=eabi} option also sets the @option{-memb} option.
11819 @opindex msdata=sysv
11820 On System V.4 and embedded PowerPC systems, put small global and static
11821 data in the @samp{.sdata} section, which is pointed to by register
11822 @code{r13}. Put small uninitialized global and static data in the
11823 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11824 The @option{-msdata=sysv} option is incompatible with the
11825 @option{-mrelocatable} option.
11827 @item -msdata=default
11829 @opindex msdata=default
11831 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
11832 compile code the same as @option{-msdata=eabi}, otherwise compile code the
11833 same as @option{-msdata=sysv}.
11836 @opindex msdata-data
11837 On System V.4 and embedded PowerPC systems, put small global
11838 data in the @samp{.sdata} section. Put small uninitialized global
11839 data in the @samp{.sbss} section. Do not use register @code{r13}
11840 to address small data however. This is the default behavior unless
11841 other @option{-msdata} options are used.
11845 @opindex msdata=none
11847 On embedded PowerPC systems, put all initialized global and static data
11848 in the @samp{.data} section, and all uninitialized data in the
11849 @samp{.bss} section.
11853 @cindex smaller data references (PowerPC)
11854 @cindex .sdata/.sdata2 references (PowerPC)
11855 On embedded PowerPC systems, put global and static items less than or
11856 equal to @var{num} bytes into the small data or bss sections instead of
11857 the normal data or bss section. By default, @var{num} is 8. The
11858 @option{-G @var{num}} switch is also passed to the linker.
11859 All modules should be compiled with the same @option{-G @var{num}} value.
11862 @itemx -mno-regnames
11864 @opindex mno-regnames
11865 On System V.4 and embedded PowerPC systems do (do not) emit register
11866 names in the assembly language output using symbolic forms.
11869 @itemx -mno-longcall
11871 @opindex mno-longcall
11872 By default assume that all calls are far away so that a longer more
11873 expensive calling sequence is required. This is required for calls
11874 further than 32 megabytes (33,554,432 bytes) from the current location.
11875 A short call will be generated if the compiler knows
11876 the call cannot be that far away. This setting can be overridden by
11877 the @code{shortcall} function attribute, or by @code{#pragma
11880 Some linkers are capable of detecting out-of-range calls and generating
11881 glue code on the fly. On these systems, long calls are unnecessary and
11882 generate slower code. As of this writing, the AIX linker can do this,
11883 as can the GNU linker for PowerPC/64. It is planned to add this feature
11884 to the GNU linker for 32-bit PowerPC systems as well.
11886 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11887 callee, L42'', plus a ``branch island'' (glue code). The two target
11888 addresses represent the callee and the ``branch island''. The
11889 Darwin/PPC linker will prefer the first address and generate a ``bl
11890 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11891 otherwise, the linker will generate ``bl L42'' to call the ``branch
11892 island''. The ``branch island'' is appended to the body of the
11893 calling function; it computes the full 32-bit address of the callee
11896 On Mach-O (Darwin) systems, this option directs the compiler emit to
11897 the glue for every direct call, and the Darwin linker decides whether
11898 to use or discard it.
11900 In the future, we may cause GCC to ignore all longcall specifications
11901 when the linker is known to generate glue.
11905 Adds support for multithreading with the @dfn{pthreads} library.
11906 This option sets flags for both the preprocessor and linker.
11910 @node S/390 and zSeries Options
11911 @subsection S/390 and zSeries Options
11912 @cindex S/390 and zSeries Options
11914 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11918 @itemx -msoft-float
11919 @opindex mhard-float
11920 @opindex msoft-float
11921 Use (do not use) the hardware floating-point instructions and registers
11922 for floating-point operations. When @option{-msoft-float} is specified,
11923 functions in @file{libgcc.a} will be used to perform floating-point
11924 operations. When @option{-mhard-float} is specified, the compiler
11925 generates IEEE floating-point instructions. This is the default.
11927 @item -mlong-double-64
11928 @itemx -mlong-double-128
11929 @opindex mlong-double-64
11930 @opindex mlong-double-128
11931 These switches control the size of @code{long double} type. A size
11932 of 64bit makes the @code{long double} type equivalent to the @code{double}
11933 type. This is the default.
11936 @itemx -mno-backchain
11937 @opindex mbackchain
11938 @opindex mno-backchain
11939 Store (do not store) the address of the caller's frame as backchain pointer
11940 into the callee's stack frame.
11941 A backchain may be needed to allow debugging using tools that do not understand
11942 DWARF-2 call frame information.
11943 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
11944 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
11945 the backchain is placed into the topmost word of the 96/160 byte register
11948 In general, code compiled with @option{-mbackchain} is call-compatible with
11949 code compiled with @option{-mmo-backchain}; however, use of the backchain
11950 for debugging purposes usually requires that the whole binary is built with
11951 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
11952 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11953 to build a linux kernel use @option{-msoft-float}.
11955 The default is to not maintain the backchain.
11957 @item -mpacked-stack
11958 @item -mno-packed-stack
11959 @opindex mpacked-stack
11960 @opindex mno-packed-stack
11961 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
11962 specified, the compiler uses the all fields of the 96/160 byte register save
11963 area only for their default purpose; unused fields still take up stack space.
11964 When @option{-mpacked-stack} is specified, register save slots are densely
11965 packed at the top of the register save area; unused space is reused for other
11966 purposes, allowing for more efficient use of the available stack space.
11967 However, when @option{-mbackchain} is also in effect, the topmost word of
11968 the save area is always used to store the backchain, and the return address
11969 register is always saved two words below the backchain.
11971 As long as the stack frame backchain is not used, code generated with
11972 @option{-mpacked-stack} is call-compatible with code generated with
11973 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
11974 S/390 or zSeries generated code that uses the stack frame backchain at run
11975 time, not just for debugging purposes. Such code is not call-compatible
11976 with code compiled with @option{-mpacked-stack}. Also, note that the
11977 combination of @option{-mbackchain},
11978 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11979 to build a linux kernel use @option{-msoft-float}.
11981 The default is to not use the packed stack layout.
11984 @itemx -mno-small-exec
11985 @opindex msmall-exec
11986 @opindex mno-small-exec
11987 Generate (or do not generate) code using the @code{bras} instruction
11988 to do subroutine calls.
11989 This only works reliably if the total executable size does not
11990 exceed 64k. The default is to use the @code{basr} instruction instead,
11991 which does not have this limitation.
11997 When @option{-m31} is specified, generate code compliant to the
11998 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
11999 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
12000 particular to generate 64-bit instructions. For the @samp{s390}
12001 targets, the default is @option{-m31}, while the @samp{s390x}
12002 targets default to @option{-m64}.
12008 When @option{-mzarch} is specified, generate code using the
12009 instructions available on z/Architecture.
12010 When @option{-mesa} is specified, generate code using the
12011 instructions available on ESA/390. Note that @option{-mesa} is
12012 not possible with @option{-m64}.
12013 When generating code compliant to the GNU/Linux for S/390 ABI,
12014 the default is @option{-mesa}. When generating code compliant
12015 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
12021 Generate (or do not generate) code using the @code{mvcle} instruction
12022 to perform block moves. When @option{-mno-mvcle} is specified,
12023 use a @code{mvc} loop instead. This is the default unless optimizing for
12030 Print (or do not print) additional debug information when compiling.
12031 The default is to not print debug information.
12033 @item -march=@var{cpu-type}
12035 Generate code that will run on @var{cpu-type}, which is the name of a system
12036 representing a certain processor type. Possible values for
12037 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
12038 When generating code using the instructions available on z/Architecture,
12039 the default is @option{-march=z900}. Otherwise, the default is
12040 @option{-march=g5}.
12042 @item -mtune=@var{cpu-type}
12044 Tune to @var{cpu-type} everything applicable about the generated code,
12045 except for the ABI and the set of available instructions.
12046 The list of @var{cpu-type} values is the same as for @option{-march}.
12047 The default is the value used for @option{-march}.
12050 @itemx -mno-tpf-trace
12051 @opindex mtpf-trace
12052 @opindex mno-tpf-trace
12053 Generate code that adds (does not add) in TPF OS specific branches to trace
12054 routines in the operating system. This option is off by default, even
12055 when compiling for the TPF OS@.
12058 @itemx -mno-fused-madd
12059 @opindex mfused-madd
12060 @opindex mno-fused-madd
12061 Generate code that uses (does not use) the floating point multiply and
12062 accumulate instructions. These instructions are generated by default if
12063 hardware floating point is used.
12065 @item -mwarn-framesize=@var{framesize}
12066 @opindex mwarn-framesize
12067 Emit a warning if the current function exceeds the given frame size. Because
12068 this is a compile time check it doesn't need to be a real problem when the program
12069 runs. It is intended to identify functions which most probably cause
12070 a stack overflow. It is useful to be used in an environment with limited stack
12071 size e.g.@: the linux kernel.
12073 @item -mwarn-dynamicstack
12074 @opindex mwarn-dynamicstack
12075 Emit a warning if the function calls alloca or uses dynamically
12076 sized arrays. This is generally a bad idea with a limited stack size.
12078 @item -mstack-guard=@var{stack-guard}
12079 @item -mstack-size=@var{stack-size}
12080 @opindex mstack-guard
12081 @opindex mstack-size
12082 These arguments always have to be used in conjunction. If they are present the s390
12083 back end emits additional instructions in the function prologue which trigger a trap
12084 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
12085 (remember that the stack on s390 grows downward). These options are intended to
12086 be used to help debugging stack overflow problems. The additionally emitted code
12087 causes only little overhead and hence can also be used in production like systems
12088 without greater performance degradation. The given values have to be exact
12089 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard} without
12091 In order to be efficient the extra code makes the assumption that the stack starts
12092 at an address aligned to the value given by @var{stack-size}.
12095 @node Score Options
12096 @subsection Score Options
12097 @cindex Score Options
12099 These options are defined for Score implementations:
12104 Compile code for big endian mode. This is the default.
12108 Compile code for little endian mode.
12112 Disable generate bcnz instruction.
12116 Enable generate unaligned load and store instruction.
12120 Enable the use of multiply-accumulate instructions. Disabled by default.
12124 Specify the SCORE5 as the target architecture.
12128 Specify the SCORE5U of the target architecture.
12132 Specify the SCORE7 as the target architecture. This is the default.
12136 Specify the SCORE7D as the target architecture.
12140 @subsection SH Options
12142 These @samp{-m} options are defined for the SH implementations:
12147 Generate code for the SH1.
12151 Generate code for the SH2.
12154 Generate code for the SH2e.
12158 Generate code for the SH3.
12162 Generate code for the SH3e.
12166 Generate code for the SH4 without a floating-point unit.
12168 @item -m4-single-only
12169 @opindex m4-single-only
12170 Generate code for the SH4 with a floating-point unit that only
12171 supports single-precision arithmetic.
12175 Generate code for the SH4 assuming the floating-point unit is in
12176 single-precision mode by default.
12180 Generate code for the SH4.
12184 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
12185 floating-point unit is not used.
12187 @item -m4a-single-only
12188 @opindex m4a-single-only
12189 Generate code for the SH4a, in such a way that no double-precision
12190 floating point operations are used.
12193 @opindex m4a-single
12194 Generate code for the SH4a assuming the floating-point unit is in
12195 single-precision mode by default.
12199 Generate code for the SH4a.
12203 Same as @option{-m4a-nofpu}, except that it implicitly passes
12204 @option{-dsp} to the assembler. GCC doesn't generate any DSP
12205 instructions at the moment.
12209 Compile code for the processor in big endian mode.
12213 Compile code for the processor in little endian mode.
12217 Align doubles at 64-bit boundaries. Note that this changes the calling
12218 conventions, and thus some functions from the standard C library will
12219 not work unless you recompile it first with @option{-mdalign}.
12223 Shorten some address references at link time, when possible; uses the
12224 linker option @option{-relax}.
12228 Use 32-bit offsets in @code{switch} tables. The default is to use
12233 Enable the use of the instruction @code{fmovd}.
12237 Comply with the calling conventions defined by Renesas.
12241 Comply with the calling conventions defined by Renesas.
12245 Comply with the calling conventions defined for GCC before the Renesas
12246 conventions were available. This option is the default for all
12247 targets of the SH toolchain except for @samp{sh-symbianelf}.
12250 @opindex mnomacsave
12251 Mark the @code{MAC} register as call-clobbered, even if
12252 @option{-mhitachi} is given.
12256 Increase IEEE-compliance of floating-point code.
12257 At the moment, this is equivalent to @option{-fno-finite-math-only}.
12258 When generating 16 bit SH opcodes, getting IEEE-conforming results for
12259 comparisons of NANs / infinities incurs extra overhead in every
12260 floating point comparison, therefore the default is set to
12261 @option{-ffinite-math-only}.
12265 Dump instruction size and location in the assembly code.
12268 @opindex mpadstruct
12269 This option is deprecated. It pads structures to multiple of 4 bytes,
12270 which is incompatible with the SH ABI@.
12274 Optimize for space instead of speed. Implied by @option{-Os}.
12277 @opindex mprefergot
12278 When generating position-independent code, emit function calls using
12279 the Global Offset Table instead of the Procedure Linkage Table.
12283 Generate a library function call to invalidate instruction cache
12284 entries, after fixing up a trampoline. This library function call
12285 doesn't assume it can write to the whole memory address space. This
12286 is the default when the target is @code{sh-*-linux*}.
12288 @item -multcost=@var{number}
12289 @opindex multcost=@var{number}
12290 Set the cost to assume for a multiply insn.
12292 @item -mdiv=@var{strategy}
12293 @opindex mdiv=@var{strategy}
12294 Set the division strategy to use for SHmedia code. @var{strategy} must be
12295 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
12296 inv:call2, inv:fp .
12297 "fp" performs the operation in floating point. This has a very high latency,
12298 but needs only a few instructions, so it might be a good choice if
12299 your code has enough easily exploitable ILP to allow the compiler to
12300 schedule the floating point instructions together with other instructions.
12301 Division by zero causes a floating point exception.
12302 "inv" uses integer operations to calculate the inverse of the divisor,
12303 and then multiplies the dividend with the inverse. This strategy allows
12304 cse and hoisting of the inverse calculation. Division by zero calculates
12305 an unspecified result, but does not trap.
12306 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
12307 have been found, or if the entire operation has been hoisted to the same
12308 place, the last stages of the inverse calculation are intertwined with the
12309 final multiply to reduce the overall latency, at the expense of using a few
12310 more instructions, and thus offering fewer scheduling opportunities with
12312 "call" calls a library function that usually implements the inv:minlat
12314 This gives high code density for m5-*media-nofpu compilations.
12315 "call2" uses a different entry point of the same library function, where it
12316 assumes that a pointer to a lookup table has already been set up, which
12317 exposes the pointer load to cse / code hoisting optimizations.
12318 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
12319 code generation, but if the code stays unoptimized, revert to the "call",
12320 "call2", or "fp" strategies, respectively. Note that the
12321 potentially-trapping side effect of division by zero is carried by a
12322 separate instruction, so it is possible that all the integer instructions
12323 are hoisted out, but the marker for the side effect stays where it is.
12324 A recombination to fp operations or a call is not possible in that case.
12325 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
12326 that the inverse calculation was nor separated from the multiply, they speed
12327 up division where the dividend fits into 20 bits (plus sign where applicable),
12328 by inserting a test to skip a number of operations in this case; this test
12329 slows down the case of larger dividends. inv20u assumes the case of a such
12330 a small dividend to be unlikely, and inv20l assumes it to be likely.
12332 @item -mdivsi3_libfunc=@var{name}
12333 @opindex mdivsi3_libfunc=@var{name}
12334 Set the name of the library function used for 32 bit signed division to
12335 @var{name}. This only affect the name used in the call and inv:call
12336 division strategies, and the compiler will still expect the same
12337 sets of input/output/clobbered registers as if this option was not present.
12339 @item -madjust-unroll
12340 @opindex madjust-unroll
12341 Throttle unrolling to avoid thrashing target registers.
12342 This option only has an effect if the gcc code base supports the
12343 TARGET_ADJUST_UNROLL_MAX target hook.
12345 @item -mindexed-addressing
12346 @opindex mindexed-addressing
12347 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
12348 This is only safe if the hardware and/or OS implement 32 bit wrap-around
12349 semantics for the indexed addressing mode. The architecture allows the
12350 implementation of processors with 64 bit MMU, which the OS could use to
12351 get 32 bit addressing, but since no current hardware implementation supports
12352 this or any other way to make the indexed addressing mode safe to use in
12353 the 32 bit ABI, the default is -mno-indexed-addressing.
12355 @item -mgettrcost=@var{number}
12356 @opindex mgettrcost=@var{number}
12357 Set the cost assumed for the gettr instruction to @var{number}.
12358 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
12362 Assume pt* instructions won't trap. This will generally generate better
12363 scheduled code, but is unsafe on current hardware. The current architecture
12364 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
12365 This has the unintentional effect of making it unsafe to schedule ptabs /
12366 ptrel before a branch, or hoist it out of a loop. For example,
12367 __do_global_ctors, a part of libgcc that runs constructors at program
12368 startup, calls functions in a list which is delimited by -1. With the
12369 -mpt-fixed option, the ptabs will be done before testing against -1.
12370 That means that all the constructors will be run a bit quicker, but when
12371 the loop comes to the end of the list, the program crashes because ptabs
12372 loads -1 into a target register. Since this option is unsafe for any
12373 hardware implementing the current architecture specification, the default
12374 is -mno-pt-fixed. Unless the user specifies a specific cost with
12375 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
12376 this deters register allocation using target registers for storing
12379 @item -minvalid-symbols
12380 @opindex minvalid-symbols
12381 Assume symbols might be invalid. Ordinary function symbols generated by
12382 the compiler will always be valid to load with movi/shori/ptabs or
12383 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
12384 to generate symbols that will cause ptabs / ptrel to trap.
12385 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
12386 It will then prevent cross-basic-block cse, hoisting and most scheduling
12387 of symbol loads. The default is @option{-mno-invalid-symbols}.
12390 @node SPARC Options
12391 @subsection SPARC Options
12392 @cindex SPARC options
12394 These @samp{-m} options are supported on the SPARC:
12397 @item -mno-app-regs
12399 @opindex mno-app-regs
12401 Specify @option{-mapp-regs} to generate output using the global registers
12402 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
12405 To be fully SVR4 ABI compliant at the cost of some performance loss,
12406 specify @option{-mno-app-regs}. You should compile libraries and system
12407 software with this option.
12410 @itemx -mhard-float
12412 @opindex mhard-float
12413 Generate output containing floating point instructions. This is the
12417 @itemx -msoft-float
12419 @opindex msoft-float
12420 Generate output containing library calls for floating point.
12421 @strong{Warning:} the requisite libraries are not available for all SPARC
12422 targets. Normally the facilities of the machine's usual C compiler are
12423 used, but this cannot be done directly in cross-compilation. You must make
12424 your own arrangements to provide suitable library functions for
12425 cross-compilation. The embedded targets @samp{sparc-*-aout} and
12426 @samp{sparclite-*-*} do provide software floating point support.
12428 @option{-msoft-float} changes the calling convention in the output file;
12429 therefore, it is only useful if you compile @emph{all} of a program with
12430 this option. In particular, you need to compile @file{libgcc.a}, the
12431 library that comes with GCC, with @option{-msoft-float} in order for
12434 @item -mhard-quad-float
12435 @opindex mhard-quad-float
12436 Generate output containing quad-word (long double) floating point
12439 @item -msoft-quad-float
12440 @opindex msoft-quad-float
12441 Generate output containing library calls for quad-word (long double)
12442 floating point instructions. The functions called are those specified
12443 in the SPARC ABI@. This is the default.
12445 As of this writing, there are no SPARC implementations that have hardware
12446 support for the quad-word floating point instructions. They all invoke
12447 a trap handler for one of these instructions, and then the trap handler
12448 emulates the effect of the instruction. Because of the trap handler overhead,
12449 this is much slower than calling the ABI library routines. Thus the
12450 @option{-msoft-quad-float} option is the default.
12452 @item -mno-unaligned-doubles
12453 @itemx -munaligned-doubles
12454 @opindex mno-unaligned-doubles
12455 @opindex munaligned-doubles
12456 Assume that doubles have 8 byte alignment. This is the default.
12458 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
12459 alignment only if they are contained in another type, or if they have an
12460 absolute address. Otherwise, it assumes they have 4 byte alignment.
12461 Specifying this option avoids some rare compatibility problems with code
12462 generated by other compilers. It is not the default because it results
12463 in a performance loss, especially for floating point code.
12465 @item -mno-faster-structs
12466 @itemx -mfaster-structs
12467 @opindex mno-faster-structs
12468 @opindex mfaster-structs
12469 With @option{-mfaster-structs}, the compiler assumes that structures
12470 should have 8 byte alignment. This enables the use of pairs of
12471 @code{ldd} and @code{std} instructions for copies in structure
12472 assignment, in place of twice as many @code{ld} and @code{st} pairs.
12473 However, the use of this changed alignment directly violates the SPARC
12474 ABI@. Thus, it's intended only for use on targets where the developer
12475 acknowledges that their resulting code will not be directly in line with
12476 the rules of the ABI@.
12478 @item -mimpure-text
12479 @opindex mimpure-text
12480 @option{-mimpure-text}, used in addition to @option{-shared}, tells
12481 the compiler to not pass @option{-z text} to the linker when linking a
12482 shared object. Using this option, you can link position-dependent
12483 code into a shared object.
12485 @option{-mimpure-text} suppresses the ``relocations remain against
12486 allocatable but non-writable sections'' linker error message.
12487 However, the necessary relocations will trigger copy-on-write, and the
12488 shared object is not actually shared across processes. Instead of
12489 using @option{-mimpure-text}, you should compile all source code with
12490 @option{-fpic} or @option{-fPIC}.
12492 This option is only available on SunOS and Solaris.
12494 @item -mcpu=@var{cpu_type}
12496 Set the instruction set, register set, and instruction scheduling parameters
12497 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
12498 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
12499 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
12500 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
12501 @samp{ultrasparc3}, and @samp{niagara}.
12503 Default instruction scheduling parameters are used for values that select
12504 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
12505 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
12507 Here is a list of each supported architecture and their supported
12512 v8: supersparc, hypersparc
12513 sparclite: f930, f934, sparclite86x
12515 v9: ultrasparc, ultrasparc3, niagara
12518 By default (unless configured otherwise), GCC generates code for the V7
12519 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
12520 additionally optimizes it for the Cypress CY7C602 chip, as used in the
12521 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
12522 SPARCStation 1, 2, IPX etc.
12524 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
12525 architecture. The only difference from V7 code is that the compiler emits
12526 the integer multiply and integer divide instructions which exist in SPARC-V8
12527 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
12528 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
12531 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
12532 the SPARC architecture. This adds the integer multiply, integer divide step
12533 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
12534 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
12535 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
12536 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
12537 MB86934 chip, which is the more recent SPARClite with FPU@.
12539 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
12540 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
12541 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
12542 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
12543 optimizes it for the TEMIC SPARClet chip.
12545 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
12546 architecture. This adds 64-bit integer and floating-point move instructions,
12547 3 additional floating-point condition code registers and conditional move
12548 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
12549 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
12550 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
12551 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
12552 @option{-mcpu=niagara}, the compiler additionally optimizes it for
12553 Sun UltraSPARC T1 chips.
12555 @item -mtune=@var{cpu_type}
12557 Set the instruction scheduling parameters for machine type
12558 @var{cpu_type}, but do not set the instruction set or register set that the
12559 option @option{-mcpu=@var{cpu_type}} would.
12561 The same values for @option{-mcpu=@var{cpu_type}} can be used for
12562 @option{-mtune=@var{cpu_type}}, but the only useful values are those
12563 that select a particular cpu implementation. Those are @samp{cypress},
12564 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
12565 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
12566 @samp{ultrasparc3}, and @samp{niagara}.
12571 @opindex mno-v8plus
12572 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
12573 difference from the V8 ABI is that the global and out registers are
12574 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
12575 mode for all SPARC-V9 processors.
12581 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
12582 Visual Instruction Set extensions. The default is @option{-mno-vis}.
12585 These @samp{-m} options are supported in addition to the above
12586 on SPARC-V9 processors in 64-bit environments:
12589 @item -mlittle-endian
12590 @opindex mlittle-endian
12591 Generate code for a processor running in little-endian mode. It is only
12592 available for a few configurations and most notably not on Solaris and Linux.
12598 Generate code for a 32-bit or 64-bit environment.
12599 The 32-bit environment sets int, long and pointer to 32 bits.
12600 The 64-bit environment sets int to 32 bits and long and pointer
12603 @item -mcmodel=medlow
12604 @opindex mcmodel=medlow
12605 Generate code for the Medium/Low code model: 64-bit addresses, programs
12606 must be linked in the low 32 bits of memory. Programs can be statically
12607 or dynamically linked.
12609 @item -mcmodel=medmid
12610 @opindex mcmodel=medmid
12611 Generate code for the Medium/Middle code model: 64-bit addresses, programs
12612 must be linked in the low 44 bits of memory, the text and data segments must
12613 be less than 2GB in size and the data segment must be located within 2GB of
12616 @item -mcmodel=medany
12617 @opindex mcmodel=medany
12618 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
12619 may be linked anywhere in memory, the text and data segments must be less
12620 than 2GB in size and the data segment must be located within 2GB of the
12623 @item -mcmodel=embmedany
12624 @opindex mcmodel=embmedany
12625 Generate code for the Medium/Anywhere code model for embedded systems:
12626 64-bit addresses, the text and data segments must be less than 2GB in
12627 size, both starting anywhere in memory (determined at link time). The
12628 global register %g4 points to the base of the data segment. Programs
12629 are statically linked and PIC is not supported.
12632 @itemx -mno-stack-bias
12633 @opindex mstack-bias
12634 @opindex mno-stack-bias
12635 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
12636 frame pointer if present, are offset by @minus{}2047 which must be added back
12637 when making stack frame references. This is the default in 64-bit mode.
12638 Otherwise, assume no such offset is present.
12641 These switches are supported in addition to the above on Solaris:
12646 Add support for multithreading using the Solaris threads library. This
12647 option sets flags for both the preprocessor and linker. This option does
12648 not affect the thread safety of object code produced by the compiler or
12649 that of libraries supplied with it.
12653 Add support for multithreading using the POSIX threads library. This
12654 option sets flags for both the preprocessor and linker. This option does
12655 not affect the thread safety of object code produced by the compiler or
12656 that of libraries supplied with it.
12660 This is a synonym for @option{-pthreads}.
12663 @node System V Options
12664 @subsection Options for System V
12666 These additional options are available on System V Release 4 for
12667 compatibility with other compilers on those systems:
12672 Create a shared object.
12673 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
12677 Identify the versions of each tool used by the compiler, in a
12678 @code{.ident} assembler directive in the output.
12682 Refrain from adding @code{.ident} directives to the output file (this is
12685 @item -YP,@var{dirs}
12687 Search the directories @var{dirs}, and no others, for libraries
12688 specified with @option{-l}.
12690 @item -Ym,@var{dir}
12692 Look in the directory @var{dir} to find the M4 preprocessor.
12693 The assembler uses this option.
12694 @c This is supposed to go with a -Yd for predefined M4 macro files, but
12695 @c the generic assembler that comes with Solaris takes just -Ym.
12698 @node TMS320C3x/C4x Options
12699 @subsection TMS320C3x/C4x Options
12700 @cindex TMS320C3x/C4x Options
12702 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
12706 @item -mcpu=@var{cpu_type}
12708 Set the instruction set, register set, and instruction scheduling
12709 parameters for machine type @var{cpu_type}. Supported values for
12710 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
12711 @samp{c44}. The default is @samp{c40} to generate code for the
12716 @itemx -msmall-memory
12718 @opindex mbig-memory
12720 @opindex msmall-memory
12722 Generates code for the big or small memory model. The small memory
12723 model assumed that all data fits into one 64K word page. At run-time
12724 the data page (DP) register must be set to point to the 64K page
12725 containing the .bss and .data program sections. The big memory model is
12726 the default and requires reloading of the DP register for every direct
12733 Allow (disallow) allocation of general integer operands into the block
12734 count register BK@.
12740 Enable (disable) generation of code using decrement and branch,
12741 DBcond(D), instructions. This is enabled by default for the C4x. To be
12742 on the safe side, this is disabled for the C3x, since the maximum
12743 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
12744 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
12745 that it can utilize the decrement and branch instruction, but will give
12746 up if there is more than one memory reference in the loop. Thus a loop
12747 where the loop counter is decremented can generate slightly more
12748 efficient code, in cases where the RPTB instruction cannot be utilized.
12750 @item -mdp-isr-reload
12752 @opindex mdp-isr-reload
12754 Force the DP register to be saved on entry to an interrupt service
12755 routine (ISR), reloaded to point to the data section, and restored on
12756 exit from the ISR@. This should not be required unless someone has
12757 violated the small memory model by modifying the DP register, say within
12764 For the C3x use the 24-bit MPYI instruction for integer multiplies
12765 instead of a library call to guarantee 32-bit results. Note that if one
12766 of the operands is a constant, then the multiplication will be performed
12767 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
12768 then squaring operations are performed inline instead of a library call.
12771 @itemx -mno-fast-fix
12773 @opindex mno-fast-fix
12774 The C3x/C4x FIX instruction to convert a floating point value to an
12775 integer value chooses the nearest integer less than or equal to the
12776 floating point value rather than to the nearest integer. Thus if the
12777 floating point number is negative, the result will be incorrectly
12778 truncated an additional code is necessary to detect and correct this
12779 case. This option can be used to disable generation of the additional
12780 code required to correct the result.
12786 Enable (disable) generation of repeat block sequences using the RPTB
12787 instruction for zero overhead looping. The RPTB construct is only used
12788 for innermost loops that do not call functions or jump across the loop
12789 boundaries. There is no advantage having nested RPTB loops due to the
12790 overhead required to save and restore the RC, RS, and RE registers.
12791 This is enabled by default with @option{-O2}.
12793 @item -mrpts=@var{count}
12797 Enable (disable) the use of the single instruction repeat instruction
12798 RPTS@. If a repeat block contains a single instruction, and the loop
12799 count can be guaranteed to be less than the value @var{count}, GCC will
12800 emit a RPTS instruction instead of a RPTB@. If no value is specified,
12801 then a RPTS will be emitted even if the loop count cannot be determined
12802 at compile time. Note that the repeated instruction following RPTS does
12803 not have to be reloaded from memory each iteration, thus freeing up the
12804 CPU buses for operands. However, since interrupts are blocked by this
12805 instruction, it is disabled by default.
12807 @item -mloop-unsigned
12808 @itemx -mno-loop-unsigned
12809 @opindex mloop-unsigned
12810 @opindex mno-loop-unsigned
12811 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
12812 is @math{2^{31} + 1} since these instructions test if the iteration count is
12813 negative to terminate the loop. If the iteration count is unsigned
12814 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
12815 exceeded. This switch allows an unsigned iteration count.
12819 Try to emit an assembler syntax that the TI assembler (asm30) is happy
12820 with. This also enforces compatibility with the API employed by the TI
12821 C3x C compiler. For example, long doubles are passed as structures
12822 rather than in floating point registers.
12828 Generate code that uses registers (stack) for passing arguments to functions.
12829 By default, arguments are passed in registers where possible rather
12830 than by pushing arguments on to the stack.
12832 @item -mparallel-insns
12833 @itemx -mno-parallel-insns
12834 @opindex mparallel-insns
12835 @opindex mno-parallel-insns
12836 Allow the generation of parallel instructions. This is enabled by
12837 default with @option{-O2}.
12839 @item -mparallel-mpy
12840 @itemx -mno-parallel-mpy
12841 @opindex mparallel-mpy
12842 @opindex mno-parallel-mpy
12843 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
12844 provided @option{-mparallel-insns} is also specified. These instructions have
12845 tight register constraints which can pessimize the code generation
12846 of large functions.
12851 @subsection V850 Options
12852 @cindex V850 Options
12854 These @samp{-m} options are defined for V850 implementations:
12858 @itemx -mno-long-calls
12859 @opindex mlong-calls
12860 @opindex mno-long-calls
12861 Treat all calls as being far away (near). If calls are assumed to be
12862 far away, the compiler will always load the functions address up into a
12863 register, and call indirect through the pointer.
12869 Do not optimize (do optimize) basic blocks that use the same index
12870 pointer 4 or more times to copy pointer into the @code{ep} register, and
12871 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
12872 option is on by default if you optimize.
12874 @item -mno-prolog-function
12875 @itemx -mprolog-function
12876 @opindex mno-prolog-function
12877 @opindex mprolog-function
12878 Do not use (do use) external functions to save and restore registers
12879 at the prologue and epilogue of a function. The external functions
12880 are slower, but use less code space if more than one function saves
12881 the same number of registers. The @option{-mprolog-function} option
12882 is on by default if you optimize.
12886 Try to make the code as small as possible. At present, this just turns
12887 on the @option{-mep} and @option{-mprolog-function} options.
12889 @item -mtda=@var{n}
12891 Put static or global variables whose size is @var{n} bytes or less into
12892 the tiny data area that register @code{ep} points to. The tiny data
12893 area can hold up to 256 bytes in total (128 bytes for byte references).
12895 @item -msda=@var{n}
12897 Put static or global variables whose size is @var{n} bytes or less into
12898 the small data area that register @code{gp} points to. The small data
12899 area can hold up to 64 kilobytes.
12901 @item -mzda=@var{n}
12903 Put static or global variables whose size is @var{n} bytes or less into
12904 the first 32 kilobytes of memory.
12908 Specify that the target processor is the V850.
12911 @opindex mbig-switch
12912 Generate code suitable for big switch tables. Use this option only if
12913 the assembler/linker complain about out of range branches within a switch
12918 This option will cause r2 and r5 to be used in the code generated by
12919 the compiler. This setting is the default.
12921 @item -mno-app-regs
12922 @opindex mno-app-regs
12923 This option will cause r2 and r5 to be treated as fixed registers.
12927 Specify that the target processor is the V850E1. The preprocessor
12928 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
12929 this option is used.
12933 Specify that the target processor is the V850E@. The preprocessor
12934 constant @samp{__v850e__} will be defined if this option is used.
12936 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
12937 are defined then a default target processor will be chosen and the
12938 relevant @samp{__v850*__} preprocessor constant will be defined.
12940 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
12941 defined, regardless of which processor variant is the target.
12943 @item -mdisable-callt
12944 @opindex mdisable-callt
12945 This option will suppress generation of the CALLT instruction for the
12946 v850e and v850e1 flavors of the v850 architecture. The default is
12947 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
12952 @subsection VAX Options
12953 @cindex VAX options
12955 These @samp{-m} options are defined for the VAX:
12960 Do not output certain jump instructions (@code{aobleq} and so on)
12961 that the Unix assembler for the VAX cannot handle across long
12966 Do output those jump instructions, on the assumption that you
12967 will assemble with the GNU assembler.
12971 Output code for g-format floating point numbers instead of d-format.
12974 @node x86-64 Options
12975 @subsection x86-64 Options
12976 @cindex x86-64 options
12978 These are listed under @xref{i386 and x86-64 Options}.
12980 @node Xstormy16 Options
12981 @subsection Xstormy16 Options
12982 @cindex Xstormy16 Options
12984 These options are defined for Xstormy16:
12989 Choose startup files and linker script suitable for the simulator.
12992 @node Xtensa Options
12993 @subsection Xtensa Options
12994 @cindex Xtensa Options
12996 These options are supported for Xtensa targets:
13000 @itemx -mno-const16
13002 @opindex mno-const16
13003 Enable or disable use of @code{CONST16} instructions for loading
13004 constant values. The @code{CONST16} instruction is currently not a
13005 standard option from Tensilica. When enabled, @code{CONST16}
13006 instructions are always used in place of the standard @code{L32R}
13007 instructions. The use of @code{CONST16} is enabled by default only if
13008 the @code{L32R} instruction is not available.
13011 @itemx -mno-fused-madd
13012 @opindex mfused-madd
13013 @opindex mno-fused-madd
13014 Enable or disable use of fused multiply/add and multiply/subtract
13015 instructions in the floating-point option. This has no effect if the
13016 floating-point option is not also enabled. Disabling fused multiply/add
13017 and multiply/subtract instructions forces the compiler to use separate
13018 instructions for the multiply and add/subtract operations. This may be
13019 desirable in some cases where strict IEEE 754-compliant results are
13020 required: the fused multiply add/subtract instructions do not round the
13021 intermediate result, thereby producing results with @emph{more} bits of
13022 precision than specified by the IEEE standard. Disabling fused multiply
13023 add/subtract instructions also ensures that the program output is not
13024 sensitive to the compiler's ability to combine multiply and add/subtract
13027 @item -mtext-section-literals
13028 @itemx -mno-text-section-literals
13029 @opindex mtext-section-literals
13030 @opindex mno-text-section-literals
13031 Control the treatment of literal pools. The default is
13032 @option{-mno-text-section-literals}, which places literals in a separate
13033 section in the output file. This allows the literal pool to be placed
13034 in a data RAM/ROM, and it also allows the linker to combine literal
13035 pools from separate object files to remove redundant literals and
13036 improve code size. With @option{-mtext-section-literals}, the literals
13037 are interspersed in the text section in order to keep them as close as
13038 possible to their references. This may be necessary for large assembly
13041 @item -mtarget-align
13042 @itemx -mno-target-align
13043 @opindex mtarget-align
13044 @opindex mno-target-align
13045 When this option is enabled, GCC instructs the assembler to
13046 automatically align instructions to reduce branch penalties at the
13047 expense of some code density. The assembler attempts to widen density
13048 instructions to align branch targets and the instructions following call
13049 instructions. If there are not enough preceding safe density
13050 instructions to align a target, no widening will be performed. The
13051 default is @option{-mtarget-align}. These options do not affect the
13052 treatment of auto-aligned instructions like @code{LOOP}, which the
13053 assembler will always align, either by widening density instructions or
13054 by inserting no-op instructions.
13057 @itemx -mno-longcalls
13058 @opindex mlongcalls
13059 @opindex mno-longcalls
13060 When this option is enabled, GCC instructs the assembler to translate
13061 direct calls to indirect calls unless it can determine that the target
13062 of a direct call is in the range allowed by the call instruction. This
13063 translation typically occurs for calls to functions in other source
13064 files. Specifically, the assembler translates a direct @code{CALL}
13065 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
13066 The default is @option{-mno-longcalls}. This option should be used in
13067 programs where the call target can potentially be out of range. This
13068 option is implemented in the assembler, not the compiler, so the
13069 assembly code generated by GCC will still show direct call
13070 instructions---look at the disassembled object code to see the actual
13071 instructions. Note that the assembler will use an indirect call for
13072 every cross-file call, not just those that really will be out of range.
13075 @node zSeries Options
13076 @subsection zSeries Options
13077 @cindex zSeries options
13079 These are listed under @xref{S/390 and zSeries Options}.
13081 @node Code Gen Options
13082 @section Options for Code Generation Conventions
13083 @cindex code generation conventions
13084 @cindex options, code generation
13085 @cindex run-time options
13087 These machine-independent options control the interface conventions
13088 used in code generation.
13090 Most of them have both positive and negative forms; the negative form
13091 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
13092 one of the forms is listed---the one which is not the default. You
13093 can figure out the other form by either removing @samp{no-} or adding
13097 @item -fbounds-check
13098 @opindex fbounds-check
13099 For front-ends that support it, generate additional code to check that
13100 indices used to access arrays are within the declared range. This is
13101 currently only supported by the Java and Fortran front-ends, where
13102 this option defaults to true and false respectively.
13106 This option generates traps for signed overflow on addition, subtraction,
13107 multiplication operations.
13111 This option instructs the compiler to assume that signed arithmetic
13112 overflow of addition, subtraction and multiplication wraps around
13113 using twos-complement representation. This flag enables some optimizations
13114 and disables others. This option is enabled by default for the Java
13115 front-end, as required by the Java language specification.
13118 @opindex fexceptions
13119 Enable exception handling. Generates extra code needed to propagate
13120 exceptions. For some targets, this implies GCC will generate frame
13121 unwind information for all functions, which can produce significant data
13122 size overhead, although it does not affect execution. If you do not
13123 specify this option, GCC will enable it by default for languages like
13124 C++ which normally require exception handling, and disable it for
13125 languages like C that do not normally require it. However, you may need
13126 to enable this option when compiling C code that needs to interoperate
13127 properly with exception handlers written in C++. You may also wish to
13128 disable this option if you are compiling older C++ programs that don't
13129 use exception handling.
13131 @item -fnon-call-exceptions
13132 @opindex fnon-call-exceptions
13133 Generate code that allows trapping instructions to throw exceptions.
13134 Note that this requires platform-specific runtime support that does
13135 not exist everywhere. Moreover, it only allows @emph{trapping}
13136 instructions to throw exceptions, i.e.@: memory references or floating
13137 point instructions. It does not allow exceptions to be thrown from
13138 arbitrary signal handlers such as @code{SIGALRM}.
13140 @item -funwind-tables
13141 @opindex funwind-tables
13142 Similar to @option{-fexceptions}, except that it will just generate any needed
13143 static data, but will not affect the generated code in any other way.
13144 You will normally not enable this option; instead, a language processor
13145 that needs this handling would enable it on your behalf.
13147 @item -fasynchronous-unwind-tables
13148 @opindex fasynchronous-unwind-tables
13149 Generate unwind table in dwarf2 format, if supported by target machine. The
13150 table is exact at each instruction boundary, so it can be used for stack
13151 unwinding from asynchronous events (such as debugger or garbage collector).
13153 @item -fpcc-struct-return
13154 @opindex fpcc-struct-return
13155 Return ``short'' @code{struct} and @code{union} values in memory like
13156 longer ones, rather than in registers. This convention is less
13157 efficient, but it has the advantage of allowing intercallability between
13158 GCC-compiled files and files compiled with other compilers, particularly
13159 the Portable C Compiler (pcc).
13161 The precise convention for returning structures in memory depends
13162 on the target configuration macros.
13164 Short structures and unions are those whose size and alignment match
13165 that of some integer type.
13167 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
13168 switch is not binary compatible with code compiled with the
13169 @option{-freg-struct-return} switch.
13170 Use it to conform to a non-default application binary interface.
13172 @item -freg-struct-return
13173 @opindex freg-struct-return
13174 Return @code{struct} and @code{union} values in registers when possible.
13175 This is more efficient for small structures than
13176 @option{-fpcc-struct-return}.
13178 If you specify neither @option{-fpcc-struct-return} nor
13179 @option{-freg-struct-return}, GCC defaults to whichever convention is
13180 standard for the target. If there is no standard convention, GCC
13181 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
13182 the principal compiler. In those cases, we can choose the standard, and
13183 we chose the more efficient register return alternative.
13185 @strong{Warning:} code compiled with the @option{-freg-struct-return}
13186 switch is not binary compatible with code compiled with the
13187 @option{-fpcc-struct-return} switch.
13188 Use it to conform to a non-default application binary interface.
13190 @item -fshort-enums
13191 @opindex fshort-enums
13192 Allocate to an @code{enum} type only as many bytes as it needs for the
13193 declared range of possible values. Specifically, the @code{enum} type
13194 will be equivalent to the smallest integer type which has enough room.
13196 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
13197 code that is not binary compatible with code generated without that switch.
13198 Use it to conform to a non-default application binary interface.
13200 @item -fshort-double
13201 @opindex fshort-double
13202 Use the same size for @code{double} as for @code{float}.
13204 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
13205 code that is not binary compatible with code generated without that switch.
13206 Use it to conform to a non-default application binary interface.
13208 @item -fshort-wchar
13209 @opindex fshort-wchar
13210 Override the underlying type for @samp{wchar_t} to be @samp{short
13211 unsigned int} instead of the default for the target. This option is
13212 useful for building programs to run under WINE@.
13214 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
13215 code that is not binary compatible with code generated without that switch.
13216 Use it to conform to a non-default application binary interface.
13219 @opindex fno-common
13220 In C, allocate even uninitialized global variables in the data section of the
13221 object file, rather than generating them as common blocks. This has the
13222 effect that if the same variable is declared (without @code{extern}) in
13223 two different compilations, you will get an error when you link them.
13224 The only reason this might be useful is if you wish to verify that the
13225 program will work on other systems which always work this way.
13229 Ignore the @samp{#ident} directive.
13231 @item -finhibit-size-directive
13232 @opindex finhibit-size-directive
13233 Don't output a @code{.size} assembler directive, or anything else that
13234 would cause trouble if the function is split in the middle, and the
13235 two halves are placed at locations far apart in memory. This option is
13236 used when compiling @file{crtstuff.c}; you should not need to use it
13239 @item -fverbose-asm
13240 @opindex fverbose-asm
13241 Put extra commentary information in the generated assembly code to
13242 make it more readable. This option is generally only of use to those
13243 who actually need to read the generated assembly code (perhaps while
13244 debugging the compiler itself).
13246 @option{-fno-verbose-asm}, the default, causes the
13247 extra information to be omitted and is useful when comparing two assembler
13252 @cindex global offset table
13254 Generate position-independent code (PIC) suitable for use in a shared
13255 library, if supported for the target machine. Such code accesses all
13256 constant addresses through a global offset table (GOT)@. The dynamic
13257 loader resolves the GOT entries when the program starts (the dynamic
13258 loader is not part of GCC; it is part of the operating system). If
13259 the GOT size for the linked executable exceeds a machine-specific
13260 maximum size, you get an error message from the linker indicating that
13261 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
13262 instead. (These maximums are 8k on the SPARC and 32k
13263 on the m68k and RS/6000. The 386 has no such limit.)
13265 Position-independent code requires special support, and therefore works
13266 only on certain machines. For the 386, GCC supports PIC for System V
13267 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
13268 position-independent.
13270 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13275 If supported for the target machine, emit position-independent code,
13276 suitable for dynamic linking and avoiding any limit on the size of the
13277 global offset table. This option makes a difference on the m68k,
13278 PowerPC and SPARC@.
13280 Position-independent code requires special support, and therefore works
13281 only on certain machines.
13283 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13290 These options are similar to @option{-fpic} and @option{-fPIC}, but
13291 generated position independent code can be only linked into executables.
13292 Usually these options are used when @option{-pie} GCC option will be
13293 used during linking.
13295 @item -fno-jump-tables
13296 @opindex fno-jump-tables
13297 Do not use jump tables for switch statements even where it would be
13298 more efficient than other code generation strategies. This option is
13299 of use in conjunction with @option{-fpic} or @option{-fPIC} for
13300 building code which forms part of a dynamic linker and cannot
13301 reference the address of a jump table. On some targets, jump tables
13302 do not require a GOT and this option is not needed.
13304 @item -ffixed-@var{reg}
13306 Treat the register named @var{reg} as a fixed register; generated code
13307 should never refer to it (except perhaps as a stack pointer, frame
13308 pointer or in some other fixed role).
13310 @var{reg} must be the name of a register. The register names accepted
13311 are machine-specific and are defined in the @code{REGISTER_NAMES}
13312 macro in the machine description macro file.
13314 This flag does not have a negative form, because it specifies a
13317 @item -fcall-used-@var{reg}
13318 @opindex fcall-used
13319 Treat the register named @var{reg} as an allocable register that is
13320 clobbered by function calls. It may be allocated for temporaries or
13321 variables that do not live across a call. Functions compiled this way
13322 will not save and restore the register @var{reg}.
13324 It is an error to used this flag with the frame pointer or stack pointer.
13325 Use of this flag for other registers that have fixed pervasive roles in
13326 the machine's execution model will produce disastrous results.
13328 This flag does not have a negative form, because it specifies a
13331 @item -fcall-saved-@var{reg}
13332 @opindex fcall-saved
13333 Treat the register named @var{reg} as an allocable register saved by
13334 functions. It may be allocated even for temporaries or variables that
13335 live across a call. Functions compiled this way will save and restore
13336 the register @var{reg} if they use it.
13338 It is an error to used this flag with the frame pointer or stack pointer.
13339 Use of this flag for other registers that have fixed pervasive roles in
13340 the machine's execution model will produce disastrous results.
13342 A different sort of disaster will result from the use of this flag for
13343 a register in which function values may be returned.
13345 This flag does not have a negative form, because it specifies a
13348 @item -fpack-struct[=@var{n}]
13349 @opindex fpack-struct
13350 Without a value specified, pack all structure members together without
13351 holes. When a value is specified (which must be a small power of two), pack
13352 structure members according to this value, representing the maximum
13353 alignment (that is, objects with default alignment requirements larger than
13354 this will be output potentially unaligned at the next fitting location.
13356 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
13357 code that is not binary compatible with code generated without that switch.
13358 Additionally, it makes the code suboptimal.
13359 Use it to conform to a non-default application binary interface.
13361 @item -finstrument-functions
13362 @opindex finstrument-functions
13363 Generate instrumentation calls for entry and exit to functions. Just
13364 after function entry and just before function exit, the following
13365 profiling functions will be called with the address of the current
13366 function and its call site. (On some platforms,
13367 @code{__builtin_return_address} does not work beyond the current
13368 function, so the call site information may not be available to the
13369 profiling functions otherwise.)
13372 void __cyg_profile_func_enter (void *this_fn,
13374 void __cyg_profile_func_exit (void *this_fn,
13378 The first argument is the address of the start of the current function,
13379 which may be looked up exactly in the symbol table.
13381 This instrumentation is also done for functions expanded inline in other
13382 functions. The profiling calls will indicate where, conceptually, the
13383 inline function is entered and exited. This means that addressable
13384 versions of such functions must be available. If all your uses of a
13385 function are expanded inline, this may mean an additional expansion of
13386 code size. If you use @samp{extern inline} in your C code, an
13387 addressable version of such functions must be provided. (This is
13388 normally the case anyways, but if you get lucky and the optimizer always
13389 expands the functions inline, you might have gotten away without
13390 providing static copies.)
13392 A function may be given the attribute @code{no_instrument_function}, in
13393 which case this instrumentation will not be done. This can be used, for
13394 example, for the profiling functions listed above, high-priority
13395 interrupt routines, and any functions from which the profiling functions
13396 cannot safely be called (perhaps signal handlers, if the profiling
13397 routines generate output or allocate memory).
13399 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
13400 @opindex finstrument-functions-exclude-file-list
13402 Set the list of functions that are excluded from instrumentation (see
13403 the description of @code{-finstrument-functions}). If the file that
13404 contains a function definition matches with one of @var{file}, then
13405 that function is not instrumented. The match is done on substrings:
13406 if the @var{file} parameter is a substring of the file name, it is
13407 considered to be a match.
13410 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
13411 will exclude any inline function defined in files whose pathnames
13412 contain @code{/bits/stl} or @code{include/sys}.
13414 If, for some reason, you want to include letter @code{','} in one of
13415 @var{sym}, write @code{'\,'}. For example,
13416 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
13417 (note the single quote surrounding the option).
13419 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
13420 @opindex finstrument-functions-exclude-function-list
13422 This is similar to @code{-finstrument-functions-exclude-file-list},
13423 but this option sets the list of function names to be excluded from
13424 instrumentation. The function name to be matched is its user-visible
13425 name, such as @code{vector<int> blah(const vector<int> &)}, not the
13426 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
13427 match is done on substrings: if the @var{sym} parameter is a substring
13428 of the function name, it is considered to be a match.
13430 @item -fstack-check
13431 @opindex fstack-check
13432 Generate code to verify that you do not go beyond the boundary of the
13433 stack. You should specify this flag if you are running in an
13434 environment with multiple threads, but only rarely need to specify it in
13435 a single-threaded environment since stack overflow is automatically
13436 detected on nearly all systems if there is only one stack.
13438 Note that this switch does not actually cause checking to be done; the
13439 operating system must do that. The switch causes generation of code
13440 to ensure that the operating system sees the stack being extended.
13442 @item -fstack-limit-register=@var{reg}
13443 @itemx -fstack-limit-symbol=@var{sym}
13444 @itemx -fno-stack-limit
13445 @opindex fstack-limit-register
13446 @opindex fstack-limit-symbol
13447 @opindex fno-stack-limit
13448 Generate code to ensure that the stack does not grow beyond a certain value,
13449 either the value of a register or the address of a symbol. If the stack
13450 would grow beyond the value, a signal is raised. For most targets,
13451 the signal is raised before the stack overruns the boundary, so
13452 it is possible to catch the signal without taking special precautions.
13454 For instance, if the stack starts at absolute address @samp{0x80000000}
13455 and grows downwards, you can use the flags
13456 @option{-fstack-limit-symbol=__stack_limit} and
13457 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
13458 of 128KB@. Note that this may only work with the GNU linker.
13460 @cindex aliasing of parameters
13461 @cindex parameters, aliased
13462 @item -fargument-alias
13463 @itemx -fargument-noalias
13464 @itemx -fargument-noalias-global
13465 @itemx -fargument-noalias-anything
13466 @opindex fargument-alias
13467 @opindex fargument-noalias
13468 @opindex fargument-noalias-global
13469 @opindex fargument-noalias-anything
13470 Specify the possible relationships among parameters and between
13471 parameters and global data.
13473 @option{-fargument-alias} specifies that arguments (parameters) may
13474 alias each other and may alias global storage.@*
13475 @option{-fargument-noalias} specifies that arguments do not alias
13476 each other, but may alias global storage.@*
13477 @option{-fargument-noalias-global} specifies that arguments do not
13478 alias each other and do not alias global storage.
13479 @option{-fargument-noalias-anything} specifies that arguments do not
13480 alias any other storage.
13482 Each language will automatically use whatever option is required by
13483 the language standard. You should not need to use these options yourself.
13485 @item -fleading-underscore
13486 @opindex fleading-underscore
13487 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
13488 change the way C symbols are represented in the object file. One use
13489 is to help link with legacy assembly code.
13491 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
13492 generate code that is not binary compatible with code generated without that
13493 switch. Use it to conform to a non-default application binary interface.
13494 Not all targets provide complete support for this switch.
13496 @item -ftls-model=@var{model}
13497 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
13498 The @var{model} argument should be one of @code{global-dynamic},
13499 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
13501 The default without @option{-fpic} is @code{initial-exec}; with
13502 @option{-fpic} the default is @code{global-dynamic}.
13504 @item -fvisibility=@var{default|internal|hidden|protected}
13505 @opindex fvisibility
13506 Set the default ELF image symbol visibility to the specified option---all
13507 symbols will be marked with this unless overridden within the code.
13508 Using this feature can very substantially improve linking and
13509 load times of shared object libraries, produce more optimized
13510 code, provide near-perfect API export and prevent symbol clashes.
13511 It is @strong{strongly} recommended that you use this in any shared objects
13514 Despite the nomenclature, @code{default} always means public ie;
13515 available to be linked against from outside the shared object.
13516 @code{protected} and @code{internal} are pretty useless in real-world
13517 usage so the only other commonly used option will be @code{hidden}.
13518 The default if @option{-fvisibility} isn't specified is
13519 @code{default}, i.e., make every
13520 symbol public---this causes the same behavior as previous versions of
13523 A good explanation of the benefits offered by ensuring ELF
13524 symbols have the correct visibility is given by ``How To Write
13525 Shared Libraries'' by Ulrich Drepper (which can be found at
13526 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
13527 solution made possible by this option to marking things hidden when
13528 the default is public is to make the default hidden and mark things
13529 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
13530 and @code{__attribute__ ((visibility("default")))} instead of
13531 @code{__declspec(dllexport)} you get almost identical semantics with
13532 identical syntax. This is a great boon to those working with
13533 cross-platform projects.
13535 For those adding visibility support to existing code, you may find
13536 @samp{#pragma GCC visibility} of use. This works by you enclosing
13537 the declarations you wish to set visibility for with (for example)
13538 @samp{#pragma GCC visibility push(hidden)} and
13539 @samp{#pragma GCC visibility pop}.
13540 Bear in mind that symbol visibility should be viewed @strong{as
13541 part of the API interface contract} and thus all new code should
13542 always specify visibility when it is not the default ie; declarations
13543 only for use within the local DSO should @strong{always} be marked explicitly
13544 as hidden as so to avoid PLT indirection overheads---making this
13545 abundantly clear also aids readability and self-documentation of the code.
13546 Note that due to ISO C++ specification requirements, operator new and
13547 operator delete must always be of default visibility.
13549 Be aware that headers from outside your project, in particular system
13550 headers and headers from any other library you use, may not be
13551 expecting to be compiled with visibility other than the default. You
13552 may need to explicitly say @samp{#pragma GCC visibility push(default)}
13553 before including any such headers.
13555 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
13556 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
13557 no modifications. However, this means that calls to @samp{extern}
13558 functions with no explicit visibility will use the PLT, so it is more
13559 effective to use @samp{__attribute ((visibility))} and/or
13560 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
13561 declarations should be treated as hidden.
13563 Note that @samp{-fvisibility} does affect C++ vague linkage
13564 entities. This means that, for instance, an exception class that will
13565 be thrown between DSOs must be explicitly marked with default
13566 visibility so that the @samp{type_info} nodes will be unified between
13569 An overview of these techniques, their benefits and how to use them
13570 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
13576 @node Environment Variables
13577 @section Environment Variables Affecting GCC
13578 @cindex environment variables
13580 @c man begin ENVIRONMENT
13581 This section describes several environment variables that affect how GCC
13582 operates. Some of them work by specifying directories or prefixes to use
13583 when searching for various kinds of files. Some are used to specify other
13584 aspects of the compilation environment.
13586 Note that you can also specify places to search using options such as
13587 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
13588 take precedence over places specified using environment variables, which
13589 in turn take precedence over those specified by the configuration of GCC@.
13590 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
13591 GNU Compiler Collection (GCC) Internals}.
13596 @c @itemx LC_COLLATE
13598 @c @itemx LC_MONETARY
13599 @c @itemx LC_NUMERIC
13604 @c @findex LC_COLLATE
13605 @findex LC_MESSAGES
13606 @c @findex LC_MONETARY
13607 @c @findex LC_NUMERIC
13611 These environment variables control the way that GCC uses
13612 localization information that allow GCC to work with different
13613 national conventions. GCC inspects the locale categories
13614 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
13615 so. These locale categories can be set to any value supported by your
13616 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
13617 Kingdom encoded in UTF-8.
13619 The @env{LC_CTYPE} environment variable specifies character
13620 classification. GCC uses it to determine the character boundaries in
13621 a string; this is needed for some multibyte encodings that contain quote
13622 and escape characters that would otherwise be interpreted as a string
13625 The @env{LC_MESSAGES} environment variable specifies the language to
13626 use in diagnostic messages.
13628 If the @env{LC_ALL} environment variable is set, it overrides the value
13629 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
13630 and @env{LC_MESSAGES} default to the value of the @env{LANG}
13631 environment variable. If none of these variables are set, GCC
13632 defaults to traditional C English behavior.
13636 If @env{TMPDIR} is set, it specifies the directory to use for temporary
13637 files. GCC uses temporary files to hold the output of one stage of
13638 compilation which is to be used as input to the next stage: for example,
13639 the output of the preprocessor, which is the input to the compiler
13642 @item GCC_EXEC_PREFIX
13643 @findex GCC_EXEC_PREFIX
13644 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
13645 names of the subprograms executed by the compiler. No slash is added
13646 when this prefix is combined with the name of a subprogram, but you can
13647 specify a prefix that ends with a slash if you wish.
13649 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
13650 an appropriate prefix to use based on the pathname it was invoked with.
13652 If GCC cannot find the subprogram using the specified prefix, it
13653 tries looking in the usual places for the subprogram.
13655 The default value of @env{GCC_EXEC_PREFIX} is
13656 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
13657 of @code{prefix} when you ran the @file{configure} script.
13659 Other prefixes specified with @option{-B} take precedence over this prefix.
13661 This prefix is also used for finding files such as @file{crt0.o} that are
13664 In addition, the prefix is used in an unusual way in finding the
13665 directories to search for header files. For each of the standard
13666 directories whose name normally begins with @samp{/usr/local/lib/gcc}
13667 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
13668 replacing that beginning with the specified prefix to produce an
13669 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
13670 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
13671 These alternate directories are searched first; the standard directories
13674 @item COMPILER_PATH
13675 @findex COMPILER_PATH
13676 The value of @env{COMPILER_PATH} is a colon-separated list of
13677 directories, much like @env{PATH}. GCC tries the directories thus
13678 specified when searching for subprograms, if it can't find the
13679 subprograms using @env{GCC_EXEC_PREFIX}.
13682 @findex LIBRARY_PATH
13683 The value of @env{LIBRARY_PATH} is a colon-separated list of
13684 directories, much like @env{PATH}. When configured as a native compiler,
13685 GCC tries the directories thus specified when searching for special
13686 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
13687 using GCC also uses these directories when searching for ordinary
13688 libraries for the @option{-l} option (but directories specified with
13689 @option{-L} come first).
13693 @cindex locale definition
13694 This variable is used to pass locale information to the compiler. One way in
13695 which this information is used is to determine the character set to be used
13696 when character literals, string literals and comments are parsed in C and C++.
13697 When the compiler is configured to allow multibyte characters,
13698 the following values for @env{LANG} are recognized:
13702 Recognize JIS characters.
13704 Recognize SJIS characters.
13706 Recognize EUCJP characters.
13709 If @env{LANG} is not defined, or if it has some other value, then the
13710 compiler will use mblen and mbtowc as defined by the default locale to
13711 recognize and translate multibyte characters.
13715 Some additional environments variables affect the behavior of the
13718 @include cppenv.texi
13722 @node Precompiled Headers
13723 @section Using Precompiled Headers
13724 @cindex precompiled headers
13725 @cindex speed of compilation
13727 Often large projects have many header files that are included in every
13728 source file. The time the compiler takes to process these header files
13729 over and over again can account for nearly all of the time required to
13730 build the project. To make builds faster, GCC allows users to
13731 `precompile' a header file; then, if builds can use the precompiled
13732 header file they will be much faster.
13734 To create a precompiled header file, simply compile it as you would any
13735 other file, if necessary using the @option{-x} option to make the driver
13736 treat it as a C or C++ header file. You will probably want to use a
13737 tool like @command{make} to keep the precompiled header up-to-date when
13738 the headers it contains change.
13740 A precompiled header file will be searched for when @code{#include} is
13741 seen in the compilation. As it searches for the included file
13742 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
13743 compiler looks for a precompiled header in each directory just before it
13744 looks for the include file in that directory. The name searched for is
13745 the name specified in the @code{#include} with @samp{.gch} appended. If
13746 the precompiled header file can't be used, it is ignored.
13748 For instance, if you have @code{#include "all.h"}, and you have
13749 @file{all.h.gch} in the same directory as @file{all.h}, then the
13750 precompiled header file will be used if possible, and the original
13751 header will be used otherwise.
13753 Alternatively, you might decide to put the precompiled header file in a
13754 directory and use @option{-I} to ensure that directory is searched
13755 before (or instead of) the directory containing the original header.
13756 Then, if you want to check that the precompiled header file is always
13757 used, you can put a file of the same name as the original header in this
13758 directory containing an @code{#error} command.
13760 This also works with @option{-include}. So yet another way to use
13761 precompiled headers, good for projects not designed with precompiled
13762 header files in mind, is to simply take most of the header files used by
13763 a project, include them from another header file, precompile that header
13764 file, and @option{-include} the precompiled header. If the header files
13765 have guards against multiple inclusion, they will be skipped because
13766 they've already been included (in the precompiled header).
13768 If you need to precompile the same header file for different
13769 languages, targets, or compiler options, you can instead make a
13770 @emph{directory} named like @file{all.h.gch}, and put each precompiled
13771 header in the directory, perhaps using @option{-o}. It doesn't matter
13772 what you call the files in the directory, every precompiled header in
13773 the directory will be considered. The first precompiled header
13774 encountered in the directory that is valid for this compilation will
13775 be used; they're searched in no particular order.
13777 There are many other possibilities, limited only by your imagination,
13778 good sense, and the constraints of your build system.
13780 A precompiled header file can be used only when these conditions apply:
13784 Only one precompiled header can be used in a particular compilation.
13787 A precompiled header can't be used once the first C token is seen. You
13788 can have preprocessor directives before a precompiled header; you can
13789 even include a precompiled header from inside another header, so long as
13790 there are no C tokens before the @code{#include}.
13793 The precompiled header file must be produced for the same language as
13794 the current compilation. You can't use a C precompiled header for a C++
13798 The precompiled header file must have been produced by the same compiler
13799 binary as the current compilation is using.
13802 Any macros defined before the precompiled header is included must
13803 either be defined in the same way as when the precompiled header was
13804 generated, or must not affect the precompiled header, which usually
13805 means that they don't appear in the precompiled header at all.
13807 The @option{-D} option is one way to define a macro before a
13808 precompiled header is included; using a @code{#define} can also do it.
13809 There are also some options that define macros implicitly, like
13810 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
13813 @item If debugging information is output when using the precompiled
13814 header, using @option{-g} or similar, the same kind of debugging information
13815 must have been output when building the precompiled header. However,
13816 a precompiled header built using @option{-g} can be used in a compilation
13817 when no debugging information is being output.
13819 @item The same @option{-m} options must generally be used when building
13820 and using the precompiled header. @xref{Submodel Options},
13821 for any cases where this rule is relaxed.
13823 @item Each of the following options must be the same when building and using
13824 the precompiled header:
13826 @gccoptlist{-fexceptions -funit-at-a-time}
13829 Some other command-line options starting with @option{-f},
13830 @option{-p}, or @option{-O} must be defined in the same way as when
13831 the precompiled header was generated. At present, it's not clear
13832 which options are safe to change and which are not; the safest choice
13833 is to use exactly the same options when generating and using the
13834 precompiled header. The following are known to be safe:
13836 @gccoptlist{-fmessage-length= -fpreprocessed
13837 -fsched-interblock -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
13838 -fsched-verbose=<number> -fschedule-insns -fvisibility=
13843 For all of these except the last, the compiler will automatically
13844 ignore the precompiled header if the conditions aren't met. If you
13845 find an option combination that doesn't work and doesn't cause the
13846 precompiled header to be ignored, please consider filing a bug report,
13849 If you do use differing options when generating and using the
13850 precompiled header, the actual behavior will be a mixture of the
13851 behavior for the options. For instance, if you use @option{-g} to
13852 generate the precompiled header but not when using it, you may or may
13853 not get debugging information for routines in the precompiled header.
13855 @node Running Protoize
13856 @section Running Protoize
13858 The program @code{protoize} is an optional part of GCC@. You can use
13859 it to add prototypes to a program, thus converting the program to ISO
13860 C in one respect. The companion program @code{unprotoize} does the
13861 reverse: it removes argument types from any prototypes that are found.
13863 When you run these programs, you must specify a set of source files as
13864 command line arguments. The conversion programs start out by compiling
13865 these files to see what functions they define. The information gathered
13866 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
13868 After scanning comes actual conversion. The specified files are all
13869 eligible to be converted; any files they include (whether sources or
13870 just headers) are eligible as well.
13872 But not all the eligible files are converted. By default,
13873 @code{protoize} and @code{unprotoize} convert only source and header
13874 files in the current directory. You can specify additional directories
13875 whose files should be converted with the @option{-d @var{directory}}
13876 option. You can also specify particular files to exclude with the
13877 @option{-x @var{file}} option. A file is converted if it is eligible, its
13878 directory name matches one of the specified directory names, and its
13879 name within the directory has not been excluded.
13881 Basic conversion with @code{protoize} consists of rewriting most
13882 function definitions and function declarations to specify the types of
13883 the arguments. The only ones not rewritten are those for varargs
13886 @code{protoize} optionally inserts prototype declarations at the
13887 beginning of the source file, to make them available for any calls that
13888 precede the function's definition. Or it can insert prototype
13889 declarations with block scope in the blocks where undeclared functions
13892 Basic conversion with @code{unprotoize} consists of rewriting most
13893 function declarations to remove any argument types, and rewriting
13894 function definitions to the old-style pre-ISO form.
13896 Both conversion programs print a warning for any function declaration or
13897 definition that they can't convert. You can suppress these warnings
13900 The output from @code{protoize} or @code{unprotoize} replaces the
13901 original source file. The original file is renamed to a name ending
13902 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
13903 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
13904 for DOS) file already exists, then the source file is simply discarded.
13906 @code{protoize} and @code{unprotoize} both depend on GCC itself to
13907 scan the program and collect information about the functions it uses.
13908 So neither of these programs will work until GCC is installed.
13910 Here is a table of the options you can use with @code{protoize} and
13911 @code{unprotoize}. Each option works with both programs unless
13915 @item -B @var{directory}
13916 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
13917 usual directory (normally @file{/usr/local/lib}). This file contains
13918 prototype information about standard system functions. This option
13919 applies only to @code{protoize}.
13921 @item -c @var{compilation-options}
13922 Use @var{compilation-options} as the options when running @command{gcc} to
13923 produce the @samp{.X} files. The special option @option{-aux-info} is
13924 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
13926 Note that the compilation options must be given as a single argument to
13927 @code{protoize} or @code{unprotoize}. If you want to specify several
13928 @command{gcc} options, you must quote the entire set of compilation options
13929 to make them a single word in the shell.
13931 There are certain @command{gcc} arguments that you cannot use, because they
13932 would produce the wrong kind of output. These include @option{-g},
13933 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
13934 the @var{compilation-options}, they are ignored.
13937 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
13938 systems) instead of @samp{.c}. This is convenient if you are converting
13939 a C program to C++. This option applies only to @code{protoize}.
13942 Add explicit global declarations. This means inserting explicit
13943 declarations at the beginning of each source file for each function
13944 that is called in the file and was not declared. These declarations
13945 precede the first function definition that contains a call to an
13946 undeclared function. This option applies only to @code{protoize}.
13948 @item -i @var{string}
13949 Indent old-style parameter declarations with the string @var{string}.
13950 This option applies only to @code{protoize}.
13952 @code{unprotoize} converts prototyped function definitions to old-style
13953 function definitions, where the arguments are declared between the
13954 argument list and the initial @samp{@{}. By default, @code{unprotoize}
13955 uses five spaces as the indentation. If you want to indent with just
13956 one space instead, use @option{-i " "}.
13959 Keep the @samp{.X} files. Normally, they are deleted after conversion
13963 Add explicit local declarations. @code{protoize} with @option{-l} inserts
13964 a prototype declaration for each function in each block which calls the
13965 function without any declaration. This option applies only to
13969 Make no real changes. This mode just prints information about the conversions
13970 that would have been done without @option{-n}.
13973 Make no @samp{.save} files. The original files are simply deleted.
13974 Use this option with caution.
13976 @item -p @var{program}
13977 Use the program @var{program} as the compiler. Normally, the name
13978 @file{gcc} is used.
13981 Work quietly. Most warnings are suppressed.
13984 Print the version number, just like @option{-v} for @command{gcc}.
13987 If you need special compiler options to compile one of your program's
13988 source files, then you should generate that file's @samp{.X} file
13989 specially, by running @command{gcc} on that source file with the
13990 appropriate options and the option @option{-aux-info}. Then run
13991 @code{protoize} on the entire set of files. @code{protoize} will use
13992 the existing @samp{.X} file because it is newer than the source file.
13996 gcc -Dfoo=bar file1.c -aux-info file1.X
14001 You need to include the special files along with the rest in the
14002 @code{protoize} command, even though their @samp{.X} files already
14003 exist, because otherwise they won't get converted.
14005 @xref{Protoize Caveats}, for more information on how to use
14006 @code{protoize} successfully.