1@c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2@c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012 3@c Free Software Foundation, Inc. 4@c This is part of the GCC manual. 5@c For copying conditions, see the file gcc.texi. 6 7@ignore 8@c man begin INCLUDE 9@include gcc-vers.texi 10@c man end 11 12@c man begin COPYRIGHT 13Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 141999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 152012 16Free Software Foundation, Inc. 17 18Permission is granted to copy, distribute and/or modify this document 19under the terms of the GNU Free Documentation License, Version 1.3 or 20any later version published by the Free Software Foundation; with the 21Invariant Sections being ``GNU General Public License'' and ``Funding 22Free Software'', the Front-Cover texts being (a) (see below), and with 23the Back-Cover Texts being (b) (see below). A copy of the license is 24included in the gfdl(7) man page. 25 26(a) The FSF's Front-Cover Text is: 27 28 A GNU Manual 29 30(b) The FSF's Back-Cover Text is: 31 32 You have freedom to copy and modify this GNU Manual, like GNU 33 software. Copies published by the Free Software Foundation raise 34 funds for GNU development. 35@c man end 36@c Set file name and title for the man page. 37@setfilename gcc 38@settitle GNU project C and C++ compiler 39@c man begin SYNOPSIS 40gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}] 41 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}] 42 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}] 43 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}] 44 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}] 45 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}] 46 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{} 47 48Only the most useful options are listed here; see below for the 49remainder. @samp{g++} accepts mostly the same options as @samp{gcc}. 50@c man end 51@c man begin SEEALSO 52gpl(7), gfdl(7), fsf-funding(7), 53cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1) 54and the Info entries for @file{gcc}, @file{cpp}, @file{as}, 55@file{ld}, @file{binutils} and @file{gdb}. 56@c man end 57@c man begin BUGS 58For instructions on reporting bugs, see 59@w{@value{BUGURL}}. 60@c man end 61@c man begin AUTHOR 62See the Info entry for @command{gcc}, or 63@w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}}, 64for contributors to GCC@. 65@c man end 66@end ignore 67 68@node Invoking GCC 69@chapter GCC Command Options 70@cindex GCC command options 71@cindex command options 72@cindex options, GCC command 73 74@c man begin DESCRIPTION 75When you invoke GCC, it normally does preprocessing, compilation, 76assembly and linking. The ``overall options'' allow you to stop this 77process at an intermediate stage. For example, the @option{-c} option 78says not to run the linker. Then the output consists of object files 79output by the assembler. 80 81Other options are passed on to one stage of processing. Some options 82control the preprocessor and others the compiler itself. Yet other 83options control the assembler and linker; most of these are not 84documented here, since you rarely need to use any of them. 85 86@cindex C compilation options 87Most of the command-line options that you can use with GCC are useful 88for C programs; when an option is only useful with another language 89(usually C++), the explanation says so explicitly. If the description 90for a particular option does not mention a source language, you can use 91that option with all supported languages. 92 93@cindex C++ compilation options 94@xref{Invoking G++,,Compiling C++ Programs}, for a summary of special 95options for compiling C++ programs. 96 97@cindex grouping options 98@cindex options, grouping 99The @command{gcc} program accepts options and file names as operands. Many 100options have multi-letter names; therefore multiple single-letter options 101may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d 102-v}}. 103 104@cindex order of options 105@cindex options, order 106You can mix options and other arguments. For the most part, the order 107you use doesn't matter. Order does matter when you use several 108options of the same kind; for example, if you specify @option{-L} more 109than once, the directories are searched in the order specified. Also, 110the placement of the @option{-l} option is significant. 111 112Many options have long names starting with @samp{-f} or with 113@samp{-W}---for example, 114@option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of 115these have both positive and negative forms; the negative form of 116@option{-ffoo} would be @option{-fno-foo}. This manual documents 117only one of these two forms, whichever one is not the default. 118 119@c man end 120 121@xref{Option Index}, for an index to GCC's options. 122 123@menu 124* Option Summary:: Brief list of all options, without explanations. 125* Overall Options:: Controlling the kind of output: 126 an executable, object files, assembler files, 127 or preprocessed source. 128* Invoking G++:: Compiling C++ programs. 129* C Dialect Options:: Controlling the variant of C language compiled. 130* C++ Dialect Options:: Variations on C++. 131* Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C 132 and Objective-C++. 133* Language Independent Options:: Controlling how diagnostics should be 134 formatted. 135* Warning Options:: How picky should the compiler be? 136* Debugging Options:: Symbol tables, measurements, and debugging dumps. 137* Optimize Options:: How much optimization? 138* Preprocessor Options:: Controlling header files and macro definitions. 139 Also, getting dependency information for Make. 140* Assembler Options:: Passing options to the assembler. 141* Link Options:: Specifying libraries and so on. 142* Directory Options:: Where to find header files and libraries. 143 Where to find the compiler executable files. 144* Spec Files:: How to pass switches to sub-processes. 145* Target Options:: Running a cross-compiler, or an old version of GCC. 146* Submodel Options:: Specifying minor hardware or convention variations, 147 such as 68010 vs 68020. 148* Code Gen Options:: Specifying conventions for function calls, data layout 149 and register usage. 150* Environment Variables:: Env vars that affect GCC. 151* Precompiled Headers:: Compiling a header once, and using it many times. 152@end menu 153 154@c man begin OPTIONS 155 156@node Option Summary 157@section Option Summary 158 159Here is a summary of all the options, grouped by type. Explanations are 160in the following sections. 161 162@table @emph 163@item Overall Options 164@xref{Overall Options,,Options Controlling the Kind of Output}. 165@gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol 166-pipe -pass-exit-codes @gol 167-x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol 168--version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol 169-fdump-ada-spec@r{[}-slim@r{]} -fdump-go-spec=@var{file}} 170 171@item C Language Options 172@xref{C Dialect Options,,Options Controlling C Dialect}. 173@gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol 174-aux-info @var{filename} -fallow-parameterless-variadic-functions @gol 175-fno-asm -fno-builtin -fno-builtin-@var{function} @gol 176-fhosted -ffreestanding -fopenmp -fms-extensions -fplan9-extensions @gol 177-trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol 178-fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol 179-fsigned-bitfields -fsigned-char @gol 180-funsigned-bitfields -funsigned-char} 181 182@item C++ Language Options 183@xref{C++ Dialect Options,,Options Controlling C++ Dialect}. 184@gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol 185-fconserve-space -fconstexpr-depth=@var{n} -ffriend-injection @gol 186-fno-elide-constructors @gol 187-fno-enforce-eh-specs @gol 188-ffor-scope -fno-for-scope -fno-gnu-keywords @gol 189-fno-implicit-templates @gol 190-fno-implicit-inline-templates @gol 191-fno-implement-inlines -fms-extensions @gol 192-fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol 193-fno-optional-diags -fpermissive @gol 194-fno-pretty-templates @gol 195-frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol 196-fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol 197-fno-default-inline -fvisibility-inlines-hidden @gol 198-fvisibility-ms-compat @gol 199-Wabi -Wconversion-null -Wctor-dtor-privacy @gol 200-Wdelete-non-virtual-dtor -Wnarrowing -Wnoexcept @gol 201-Wnon-virtual-dtor -Wreorder @gol 202-Weffc++ -Wstrict-null-sentinel @gol 203-Wno-non-template-friend -Wold-style-cast @gol 204-Woverloaded-virtual -Wno-pmf-conversions @gol 205-Wsign-promo} 206 207@item Objective-C and Objective-C++ Language Options 208@xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling 209Objective-C and Objective-C++ Dialects}. 210@gccoptlist{-fconstant-string-class=@var{class-name} @gol 211-fgnu-runtime -fnext-runtime @gol 212-fno-nil-receivers @gol 213-fobjc-abi-version=@var{n} @gol 214-fobjc-call-cxx-cdtors @gol 215-fobjc-direct-dispatch @gol 216-fobjc-exceptions @gol 217-fobjc-gc @gol 218-fobjc-nilcheck @gol 219-fobjc-std=objc1 @gol 220-freplace-objc-classes @gol 221-fzero-link @gol 222-gen-decls @gol 223-Wassign-intercept @gol 224-Wno-protocol -Wselector @gol 225-Wstrict-selector-match @gol 226-Wundeclared-selector} 227 228@item Language Independent Options 229@xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}. 230@gccoptlist{-fmessage-length=@var{n} @gol 231-fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol 232-fno-diagnostics-show-option} 233 234@item Warning Options 235@xref{Warning Options,,Options to Request or Suppress Warnings}. 236@gccoptlist{-fsyntax-only -fmax-errors=@var{n} -pedantic @gol 237-pedantic-errors @gol 238-w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol 239-Wno-attributes -Wno-builtin-macro-redefined @gol 240-Wc++-compat -Wc++11-compat -Wcast-align -Wcast-qual @gol 241-Wchar-subscripts -Wclobbered -Wcomment @gol 242-Wconversion -Wcoverage-mismatch -Wno-cpp -Wno-deprecated @gol 243-Wno-deprecated-declarations -Wdisabled-optimization @gol 244-Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol 245-Wno-endif-labels -Werror -Werror=* @gol 246-Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol 247-Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol 248-Wformat-security -Wformat-y2k @gol 249-Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol 250-Wignored-qualifiers @gol 251-Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol 252-Winit-self -Winline -Wmaybe-uninitialized @gol 253-Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol 254-Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol 255-Wlogical-op -Wlong-long @gol 256-Wmain -Wmaybe-uninitialized -Wmissing-braces -Wmissing-field-initializers @gol 257-Wmissing-format-attribute -Wmissing-include-dirs @gol 258-Wno-mudflap @gol 259-Wno-multichar -Wnonnull -Wno-overflow @gol 260-Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol 261-Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol 262-Wpointer-arith -Wno-pointer-to-int-cast @gol 263-Wredundant-decls @gol 264-Wreturn-type -Wsequence-point -Wshadow @gol 265-Wsign-compare -Wsign-conversion -Wstack-protector @gol 266-Wstack-usage=@var{len} -Wstrict-aliasing -Wstrict-aliasing=n @gol 267-Wstrict-overflow -Wstrict-overflow=@var{n} @gol 268-Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol 269-Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol 270-Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol 271-Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol 272-Wunsuffixed-float-constants -Wunused -Wunused-function @gol 273-Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol 274-Wno-unused-result -Wunused-value @gol -Wunused-variable @gol 275-Wunused-but-set-parameter -Wunused-but-set-variable @gol 276-Wvariadic-macros -Wvector-operation-performance -Wvla 277-Wvolatile-register-var -Wwrite-strings -Wzero-as-null-pointer-constant} 278 279@item C and Objective-C-only Warning Options 280@gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol 281-Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol 282-Wold-style-declaration -Wold-style-definition @gol 283-Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol 284-Wdeclaration-after-statement -Wpointer-sign} 285 286@item Debugging Options 287@xref{Debugging Options,,Options for Debugging Your Program or GCC}. 288@gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol 289-fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol 290-fdisable-ipa-@var{pass_name} @gol 291-fdisable-rtl-@var{pass_name} @gol 292-fdisable-rtl-@var{pass-name}=@var{range-list} @gol 293-fdisable-tree-@var{pass_name} @gol 294-fdisable-tree-@var{pass-name}=@var{range-list} @gol 295-fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol 296-fdump-translation-unit@r{[}-@var{n}@r{]} @gol 297-fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol 298-fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol 299-fdump-passes @gol 300-fdump-statistics @gol 301-fdump-tree-all @gol 302-fdump-tree-original@r{[}-@var{n}@r{]} @gol 303-fdump-tree-optimized@r{[}-@var{n}@r{]} @gol 304-fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol 305-fdump-tree-ch @gol 306-fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol 307-fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol 308-fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol 309-fdump-tree-dom@r{[}-@var{n}@r{]} @gol 310-fdump-tree-dse@r{[}-@var{n}@r{]} @gol 311-fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol 312-fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol 313-fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol 314-fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol 315-fdump-tree-nrv -fdump-tree-vect @gol 316-fdump-tree-sink @gol 317-fdump-tree-sra@r{[}-@var{n}@r{]} @gol 318-fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol 319-fdump-tree-fre@r{[}-@var{n}@r{]} @gol 320-fdump-tree-vrp@r{[}-@var{n}@r{]} @gol 321-ftree-vectorizer-verbose=@var{n} @gol 322-fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol 323-fdump-final-insns=@var{file} @gol 324-fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol 325-feliminate-dwarf2-dups -feliminate-unused-debug-types @gol 326-feliminate-unused-debug-symbols -femit-class-debug-always @gol 327-fenable-@var{kind}-@var{pass} @gol 328-fenable-@var{kind}-@var{pass}=@var{range-list} @gol 329-fdebug-types-section @gol 330-fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol 331-frandom-seed=@var{string} -fsched-verbose=@var{n} @gol 332-fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol 333-fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol 334-fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol 335-g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol 336-ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol 337-gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol 338-gvms -gxcoff -gxcoff+ @gol 339-fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol 340-fdebug-prefix-map=@var{old}=@var{new} @gol 341-femit-struct-debug-baseonly -femit-struct-debug-reduced @gol 342-femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol 343-p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol 344-print-multi-directory -print-multi-lib -print-multi-os-directory @gol 345-print-prog-name=@var{program} -print-search-dirs -Q @gol 346-print-sysroot -print-sysroot-headers-suffix @gol 347-save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}} 348 349@item Optimization Options 350@xref{Optimize Options,,Options that Control Optimization}. 351@gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol 352-falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol 353-fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol 354-fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol 355-fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol 356-fcompare-elim -fcprop-registers -fcrossjumping @gol 357-fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol 358-fcx-limited-range @gol 359-fdata-sections -fdce -fdelayed-branch @gol 360-fdelete-null-pointer-checks -fdevirtualize -fdse @gol 361-fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol 362-ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol 363-fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol 364-fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol 365-fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol 366-finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol 367-finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol 368-fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol 369-fira-algorithm=@var{algorithm} @gol 370-fira-region=@var{region} @gol 371-fira-loop-pressure -fno-ira-share-save-slots @gol 372-fno-ira-share-spill-slots -fira-verbose=@var{n} @gol 373-fivopts -fkeep-inline-functions -fkeep-static-consts @gol 374-floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol 375-floop-parallelize-all -flto -flto-compression-level @gol 376-flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol 377-fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol 378-fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol 379-fno-default-inline @gol 380-fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol 381-fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol 382-fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol 383-fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol 384-fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol 385-fpartial-inlining -fpeel-loops -fpredictive-commoning @gol 386-fprefetch-loop-arrays @gol 387-fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol 388-fprofile-generate=@var{path} @gol 389-fprofile-use -fprofile-use=@var{path} -fprofile-values @gol 390-freciprocal-math -free -fregmove -frename-registers -freorder-blocks @gol 391-freorder-blocks-and-partition -freorder-functions @gol 392-frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol 393-frounding-math -fsched2-use-superblocks -fsched-pressure @gol 394-fsched-spec-load -fsched-spec-load-dangerous @gol 395-fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol 396-fsched-group-heuristic -fsched-critical-path-heuristic @gol 397-fsched-spec-insn-heuristic -fsched-rank-heuristic @gol 398-fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol 399-fschedule-insns -fschedule-insns2 -fsection-anchors @gol 400-fselective-scheduling -fselective-scheduling2 @gol 401-fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol 402-fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol 403-fsplit-ivs-in-unroller -fsplit-wide-types -fstack-protector @gol 404-fstack-protector-all -fstrict-aliasing -fstrict-overflow @gol 405-fthread-jumps -ftracer -ftree-bit-ccp @gol 406-ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol 407-ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol 408-ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol 409-ftree-loop-if-convert-stores -ftree-loop-im @gol 410-ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol 411-ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol 412-ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol 413-ftree-sink -ftree-sra -ftree-switch-conversion -ftree-tail-merge @gol 414-ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol 415-funit-at-a-time -funroll-all-loops -funroll-loops @gol 416-funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol 417-fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol 418-fwhole-program -fwpa -fuse-linker-plugin @gol 419--param @var{name}=@var{value} 420-O -O0 -O1 -O2 -O3 -Os -Ofast} 421 422@item Preprocessor Options 423@xref{Preprocessor Options,,Options Controlling the Preprocessor}. 424@gccoptlist{-A@var{question}=@var{answer} @gol 425-A-@var{question}@r{[}=@var{answer}@r{]} @gol 426-C -dD -dI -dM -dN @gol 427-D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol 428-idirafter @var{dir} @gol 429-include @var{file} -imacros @var{file} @gol 430-iprefix @var{file} -iwithprefix @var{dir} @gol 431-iwithprefixbefore @var{dir} -isystem @var{dir} @gol 432-imultilib @var{dir} -isysroot @var{dir} @gol 433-M -MM -MF -MG -MP -MQ -MT -nostdinc @gol 434-P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol 435-remap -trigraphs -undef -U@var{macro} @gol 436-Wp,@var{option} -Xpreprocessor @var{option}} 437 438@item Assembler Option 439@xref{Assembler Options,,Passing Options to the Assembler}. 440@gccoptlist{-Wa,@var{option} -Xassembler @var{option}} 441 442@item Linker Options 443@xref{Link Options,,Options for Linking}. 444@gccoptlist{@var{object-file-name} -l@var{library} @gol 445-nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol 446-s -static -static-libgcc -static-libstdc++ -shared @gol 447-shared-libgcc -symbolic @gol 448-T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol 449-u @var{symbol}} 450 451@item Directory Options 452@xref{Directory Options,,Options for Directory Search}. 453@gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol 454-iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol 455--sysroot=@var{dir}} 456 457@item Machine Dependent Options 458@xref{Submodel Options,,Hardware Models and Configurations}. 459@c This list is ordered alphanumerically by subsection name. 460@c Try and put the significant identifier (CPU or system) first, 461@c so users have a clue at guessing where the ones they want will be. 462 463@emph{Adapteva Epiphany Options} 464@gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol 465-mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol 466-msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol 467-mround-nearest -mlong-calls -mshort-calls -msmall16 @gol 468-mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol 469-msplit-vecmove-early -m1reg-@var{reg}} 470 471@emph{ARM Options} 472@gccoptlist{-mapcs-frame -mno-apcs-frame @gol 473-mabi=@var{name} @gol 474-mapcs-stack-check -mno-apcs-stack-check @gol 475-mapcs-float -mno-apcs-float @gol 476-mapcs-reentrant -mno-apcs-reentrant @gol 477-msched-prolog -mno-sched-prolog @gol 478-mlittle-endian -mbig-endian -mwords-little-endian @gol 479-mfloat-abi=@var{name} -mfpe @gol 480-mfp16-format=@var{name} 481-mthumb-interwork -mno-thumb-interwork @gol 482-mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol 483-mstructure-size-boundary=@var{n} @gol 484-mabort-on-noreturn @gol 485-mlong-calls -mno-long-calls @gol 486-msingle-pic-base -mno-single-pic-base @gol 487-mpic-register=@var{reg} @gol 488-mnop-fun-dllimport @gol 489-mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol 490-mpoke-function-name @gol 491-mthumb -marm @gol 492-mtpcs-frame -mtpcs-leaf-frame @gol 493-mcaller-super-interworking -mcallee-super-interworking @gol 494-mtp=@var{name} -mtls-dialect=@var{dialect} @gol 495-mword-relocations @gol 496-mfix-cortex-m3-ldrd @gol 497-munaligned-access} 498 499@emph{AVR Options} 500@gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol 501-mcall-prologues -mint8 -mno-interrupts -mrelax -mshort-calls @gol 502-mstrict-X -mtiny-stack} 503 504@emph{Blackfin Options} 505@gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol 506-msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol 507-mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol 508-mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol 509-mno-id-shared-library -mshared-library-id=@var{n} @gol 510-mleaf-id-shared-library -mno-leaf-id-shared-library @gol 511-msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol 512-mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol 513-micplb} 514 515@emph{C6X Options} 516@gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol 517-msim -msdata=@var{sdata-type}} 518 519@emph{CRIS Options} 520@gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol 521-mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol 522-metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol 523-mstack-align -mdata-align -mconst-align @gol 524-m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol 525-melf -maout -melinux -mlinux -sim -sim2 @gol 526-mmul-bug-workaround -mno-mul-bug-workaround} 527 528@emph{CR16 Options} 529@gccoptlist{-mmac @gol 530-mcr16cplus -mcr16c @gol 531-msim -mint32 -mbit-ops 532-mdata-model=@var{model}} 533 534@emph{Darwin Options} 535@gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol 536-arch_only -bind_at_load -bundle -bundle_loader @gol 537-client_name -compatibility_version -current_version @gol 538-dead_strip @gol 539-dependency-file -dylib_file -dylinker_install_name @gol 540-dynamic -dynamiclib -exported_symbols_list @gol 541-filelist -flat_namespace -force_cpusubtype_ALL @gol 542-force_flat_namespace -headerpad_max_install_names @gol 543-iframework @gol 544-image_base -init -install_name -keep_private_externs @gol 545-multi_module -multiply_defined -multiply_defined_unused @gol 546-noall_load -no_dead_strip_inits_and_terms @gol 547-nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol 548-pagezero_size -prebind -prebind_all_twolevel_modules @gol 549-private_bundle -read_only_relocs -sectalign @gol 550-sectobjectsymbols -whyload -seg1addr @gol 551-sectcreate -sectobjectsymbols -sectorder @gol 552-segaddr -segs_read_only_addr -segs_read_write_addr @gol 553-seg_addr_table -seg_addr_table_filename -seglinkedit @gol 554-segprot -segs_read_only_addr -segs_read_write_addr @gol 555-single_module -static -sub_library -sub_umbrella @gol 556-twolevel_namespace -umbrella -undefined @gol 557-unexported_symbols_list -weak_reference_mismatches @gol 558-whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol 559-mkernel -mone-byte-bool} 560 561@emph{DEC Alpha Options} 562@gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol 563-mieee -mieee-with-inexact -mieee-conformant @gol 564-mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol 565-mtrap-precision=@var{mode} -mbuild-constants @gol 566-mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol 567-mbwx -mmax -mfix -mcix @gol 568-mfloat-vax -mfloat-ieee @gol 569-mexplicit-relocs -msmall-data -mlarge-data @gol 570-msmall-text -mlarge-text @gol 571-mmemory-latency=@var{time}} 572 573@emph{DEC Alpha/VMS Options} 574@gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64} 575 576@emph{FR30 Options} 577@gccoptlist{-msmall-model -mno-lsim} 578 579@emph{FRV Options} 580@gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol 581-mhard-float -msoft-float @gol 582-malloc-cc -mfixed-cc -mdword -mno-dword @gol 583-mdouble -mno-double @gol 584-mmedia -mno-media -mmuladd -mno-muladd @gol 585-mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol 586-mlinked-fp -mlong-calls -malign-labels @gol 587-mlibrary-pic -macc-4 -macc-8 @gol 588-mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol 589-moptimize-membar -mno-optimize-membar @gol 590-mscc -mno-scc -mcond-exec -mno-cond-exec @gol 591-mvliw-branch -mno-vliw-branch @gol 592-mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol 593-mno-nested-cond-exec -mtomcat-stats @gol 594-mTLS -mtls @gol 595-mcpu=@var{cpu}} 596 597@emph{GNU/Linux Options} 598@gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol 599-tno-android-cc -tno-android-ld} 600 601@emph{H8/300 Options} 602@gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300} 603 604@emph{HPPA Options} 605@gccoptlist{-march=@var{architecture-type} @gol 606-mbig-switch -mdisable-fpregs -mdisable-indexing @gol 607-mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol 608-mfixed-range=@var{register-range} @gol 609-mjump-in-delay -mlinker-opt -mlong-calls @gol 610-mlong-load-store -mno-big-switch -mno-disable-fpregs @gol 611-mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol 612-mno-jump-in-delay -mno-long-load-store @gol 613-mno-portable-runtime -mno-soft-float @gol 614-mno-space-regs -msoft-float -mpa-risc-1-0 @gol 615-mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol 616-mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol 617-munix=@var{unix-std} -nolibdld -static -threads} 618 619@emph{i386 and x86-64 Options} 620@gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol 621-mfpmath=@var{unit} @gol 622-masm=@var{dialect} -mno-fancy-math-387 @gol 623-mno-fp-ret-in-387 -msoft-float @gol 624-mno-wide-multiply -mrtd -malign-double @gol 625-mpreferred-stack-boundary=@var{num} @gol 626-mincoming-stack-boundary=@var{num} @gol 627-mcld -mcx16 -msahf -mmovbe -mcrc32 @gol 628-mrecip -mrecip=@var{opt} @gol 629-mvzeroupper @gol 630-mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol 631-mavx2 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol 632-msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol 633-mbmi2 -mlwp -mthreads -mno-align-stringops -minline-all-stringops @gol 634-minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol 635-mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol 636-m96bit-long-double -mregparm=@var{num} -msseregparm @gol 637-mveclibabi=@var{type} -mvect8-ret-in-mem @gol 638-mpc32 -mpc64 -mpc80 -mstackrealign @gol 639-momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol 640-mcmodel=@var{code-model} -mabi=@var{name} @gol 641-m32 -m64 -mx32 -mlarge-data-threshold=@var{num} @gol 642-msse2avx -mfentry -m8bit-idiv @gol 643-mavx256-split-unaligned-load -mavx256-split-unaligned-store} 644 645@emph{i386 and x86-64 Windows Options} 646@gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol 647-mnop-fun-dllimport -mthread @gol 648-municode -mwin32 -mwindows -fno-set-stack-executable} 649 650@emph{IA-64 Options} 651@gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol 652-mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol 653-mconstant-gp -mauto-pic -mfused-madd @gol 654-minline-float-divide-min-latency @gol 655-minline-float-divide-max-throughput @gol 656-mno-inline-float-divide @gol 657-minline-int-divide-min-latency @gol 658-minline-int-divide-max-throughput @gol 659-mno-inline-int-divide @gol 660-minline-sqrt-min-latency -minline-sqrt-max-throughput @gol 661-mno-inline-sqrt @gol 662-mdwarf2-asm -mearly-stop-bits @gol 663-mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol 664-mtune=@var{cpu-type} -milp32 -mlp64 @gol 665-msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol 666-msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol 667-msched-spec-ldc -msched-spec-control-ldc @gol 668-msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol 669-msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol 670-msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol 671-msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}} 672 673@emph{IA-64/VMS Options} 674@gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64} 675 676@emph{LM32 Options} 677@gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol 678-msign-extend-enabled -muser-enabled} 679 680@emph{M32R/D Options} 681@gccoptlist{-m32r2 -m32rx -m32r @gol 682-mdebug @gol 683-malign-loops -mno-align-loops @gol 684-missue-rate=@var{number} @gol 685-mbranch-cost=@var{number} @gol 686-mmodel=@var{code-size-model-type} @gol 687-msdata=@var{sdata-type} @gol 688-mno-flush-func -mflush-func=@var{name} @gol 689-mno-flush-trap -mflush-trap=@var{number} @gol 690-G @var{num}} 691 692@emph{M32C Options} 693@gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}} 694 695@emph{M680x0 Options} 696@gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} 697-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol 698-m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol 699-mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol 700-mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol 701-mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol 702-malign-int -mstrict-align -msep-data -mno-sep-data @gol 703-mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol 704-mxgot -mno-xgot} 705 706@emph{MCore Options} 707@gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol 708-mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol 709-m4byte-functions -mno-4byte-functions -mcallgraph-data @gol 710-mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol 711-mlittle-endian -mbig-endian -m210 -m340 -mstack-increment} 712 713@emph{MeP Options} 714@gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol 715-mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol 716-mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol 717-mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol 718-mtiny=@var{n}} 719 720@emph{MicroBlaze Options} 721@gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol 722-mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol 723-mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol 724-mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol 725-mxl-mode-@var{app-model}} 726 727@emph{MIPS Options} 728@gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol 729-mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol 730-mips64 -mips64r2 @gol 731-mips16 -mno-mips16 -mflip-mips16 @gol 732-minterlink-mips16 -mno-interlink-mips16 @gol 733-mabi=@var{abi} -mabicalls -mno-abicalls @gol 734-mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol 735-mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol 736-msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol 737-mfpu=@var{fpu-type} @gol 738-msmartmips -mno-smartmips @gol 739-mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol 740-mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol 741-mlong64 -mlong32 -msym32 -mno-sym32 @gol 742-G@var{num} -mlocal-sdata -mno-local-sdata @gol 743-mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol 744-membedded-data -mno-embedded-data @gol 745-muninit-const-in-rodata -mno-uninit-const-in-rodata @gol 746-mcode-readable=@var{setting} @gol 747-msplit-addresses -mno-split-addresses @gol 748-mexplicit-relocs -mno-explicit-relocs @gol 749-mcheck-zero-division -mno-check-zero-division @gol 750-mdivide-traps -mdivide-breaks @gol 751-mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol 752-mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol 753-mfix-24k -mno-fix-24k @gol 754-mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol 755-mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol 756-mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol 757-mflush-func=@var{func} -mno-flush-func @gol 758-mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol 759-mfp-exceptions -mno-fp-exceptions @gol 760-mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol 761-mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address} 762 763@emph{MMIX Options} 764@gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol 765-mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol 766-melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol 767-mno-base-addresses -msingle-exit -mno-single-exit} 768 769@emph{MN10300 Options} 770@gccoptlist{-mmult-bug -mno-mult-bug @gol 771-mno-am33 -mam33 -mam33-2 -mam34 @gol 772-mtune=@var{cpu-type} @gol 773-mreturn-pointer-on-d0 @gol 774-mno-crt0 -mrelax -mliw -msetlb} 775 776@emph{PDP-11 Options} 777@gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol 778-mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol 779-mint16 -mno-int32 -mfloat32 -mno-float64 @gol 780-mfloat64 -mno-float32 -mabshi -mno-abshi @gol 781-mbranch-expensive -mbranch-cheap @gol 782-munix-asm -mdec-asm} 783 784@emph{picoChip Options} 785@gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol 786-msymbol-as-address -mno-inefficient-warnings} 787 788@emph{PowerPC Options} 789See RS/6000 and PowerPC Options. 790 791@emph{RL78 Options} 792@gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78} 793 794@emph{RS/6000 and PowerPC Options} 795@gccoptlist{-mcpu=@var{cpu-type} @gol 796-mtune=@var{cpu-type} @gol 797-mcmodel=@var{code-model} @gol 798-mpower -mno-power -mpower2 -mno-power2 @gol 799-mpowerpc -mpowerpc64 -mno-powerpc @gol 800-maltivec -mno-altivec @gol 801-mpowerpc-gpopt -mno-powerpc-gpopt @gol 802-mpowerpc-gfxopt -mno-powerpc-gfxopt @gol 803-mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol 804-mfprnd -mno-fprnd @gol 805-mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol 806-mnew-mnemonics -mold-mnemonics @gol 807-mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol 808-m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol 809-malign-power -malign-natural @gol 810-msoft-float -mhard-float -mmultiple -mno-multiple @gol 811-msingle-float -mdouble-float -msimple-fpu @gol 812-mstring -mno-string -mupdate -mno-update @gol 813-mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol 814-mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol 815-mstrict-align -mno-strict-align -mrelocatable @gol 816-mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol 817-mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol 818-mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol 819-mprioritize-restricted-insns=@var{priority} @gol 820-msched-costly-dep=@var{dependence_type} @gol 821-minsert-sched-nops=@var{scheme} @gol 822-mcall-sysv -mcall-netbsd @gol 823-maix-struct-return -msvr4-struct-return @gol 824-mabi=@var{abi-type} -msecure-plt -mbss-plt @gol 825-mblock-move-inline-limit=@var{num} @gol 826-misel -mno-isel @gol 827-misel=yes -misel=no @gol 828-mspe -mno-spe @gol 829-mspe=yes -mspe=no @gol 830-mpaired @gol 831-mgen-cell-microcode -mwarn-cell-microcode @gol 832-mvrsave -mno-vrsave @gol 833-mmulhw -mno-mulhw @gol 834-mdlmzb -mno-dlmzb @gol 835-mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol 836-mprototype -mno-prototype @gol 837-msim -mmvme -mads -myellowknife -memb -msdata @gol 838-msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol 839-mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol 840-mno-recip-precision @gol 841-mveclibabi=@var{type} -mfriz -mno-friz @gol 842-mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol 843-msave-toc-indirect -mno-save-toc-indirect} 844 845@emph{RX Options} 846@gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol 847-mcpu=@gol 848-mbig-endian-data -mlittle-endian-data @gol 849-msmall-data @gol 850-msim -mno-sim@gol 851-mas100-syntax -mno-as100-syntax@gol 852-mrelax@gol 853-mmax-constant-size=@gol 854-mint-register=@gol 855-mpid@gol 856-msave-acc-in-interrupts} 857 858@emph{S/390 and zSeries Options} 859@gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol 860-mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol 861-mlong-double-64 -mlong-double-128 @gol 862-mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol 863-msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol 864-m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol 865-mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol 866-mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard} 867 868@emph{Score Options} 869@gccoptlist{-meb -mel @gol 870-mnhwloop @gol 871-muls @gol 872-mmac @gol 873-mscore5 -mscore5u -mscore7 -mscore7d} 874 875@emph{SH Options} 876@gccoptlist{-m1 -m2 -m2e @gol 877-m2a-nofpu -m2a-single-only -m2a-single -m2a @gol 878-m3 -m3e @gol 879-m4-nofpu -m4-single-only -m4-single -m4 @gol 880-m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol 881-m5-64media -m5-64media-nofpu @gol 882-m5-32media -m5-32media-nofpu @gol 883-m5-compact -m5-compact-nofpu @gol 884-mb -ml -mdalign -mrelax @gol 885-mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol 886-mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol 887-mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol 888-mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol 889-madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol 890-maccumulate-outgoing-args -minvalid-symbols -msoft-atomic @gol 891-mbranch-cost=@var{num} -mcbranchdi -mcmpeqdi -mfused-madd -mpretend-cmove} 892 893@emph{Solaris 2 Options} 894@gccoptlist{-mimpure-text -mno-impure-text @gol 895-pthreads -pthread} 896 897@emph{SPARC Options} 898@gccoptlist{-mcpu=@var{cpu-type} @gol 899-mtune=@var{cpu-type} @gol 900-mcmodel=@var{code-model} @gol 901-mmemory-model=@var{mem-model} @gol 902-m32 -m64 -mapp-regs -mno-app-regs @gol 903-mfaster-structs -mno-faster-structs -mflat -mno-flat @gol 904-mfpu -mno-fpu -mhard-float -msoft-float @gol 905-mhard-quad-float -msoft-quad-float @gol 906-mlittle-endian @gol 907-mstack-bias -mno-stack-bias @gol 908-munaligned-doubles -mno-unaligned-doubles @gol 909-mv8plus -mno-v8plus -mvis -mno-vis @gol 910-mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol 911-mfmaf -mno-fmaf -mpopc -mno-popc @gol 912-mfix-at697f} 913 914@emph{SPU Options} 915@gccoptlist{-mwarn-reloc -merror-reloc @gol 916-msafe-dma -munsafe-dma @gol 917-mbranch-hints @gol 918-msmall-mem -mlarge-mem -mstdmain @gol 919-mfixed-range=@var{register-range} @gol 920-mea32 -mea64 @gol 921-maddress-space-conversion -mno-address-space-conversion @gol 922-mcache-size=@var{cache-size} @gol 923-matomic-updates -mno-atomic-updates} 924 925@emph{System V Options} 926@gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}} 927 928@emph{TILE-Gx Options} 929@gccoptlist{-mcpu=@var{cpu} -m32 -m64} 930 931@emph{TILEPro Options} 932@gccoptlist{-mcpu=@var{cpu} -m32} 933 934@emph{V850 Options} 935@gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol 936-mprolog-function -mno-prolog-function -mspace @gol 937-mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol 938-mapp-regs -mno-app-regs @gol 939-mdisable-callt -mno-disable-callt @gol 940-mv850e2v3 @gol 941-mv850e2 @gol 942-mv850e1 -mv850es @gol 943-mv850e @gol 944-mv850 -mbig-switch} 945 946@emph{VAX Options} 947@gccoptlist{-mg -mgnu -munix} 948 949@emph{VxWorks Options} 950@gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol 951-Xbind-lazy -Xbind-now} 952 953@emph{x86-64 Options} 954See i386 and x86-64 Options. 955 956@emph{Xstormy16 Options} 957@gccoptlist{-msim} 958 959@emph{Xtensa Options} 960@gccoptlist{-mconst16 -mno-const16 @gol 961-mfused-madd -mno-fused-madd @gol 962-mforce-no-pic @gol 963-mserialize-volatile -mno-serialize-volatile @gol 964-mtext-section-literals -mno-text-section-literals @gol 965-mtarget-align -mno-target-align @gol 966-mlongcalls -mno-longcalls} 967 968@emph{zSeries Options} 969See S/390 and zSeries Options. 970 971@item Code Generation Options 972@xref{Code Gen Options,,Options for Code Generation Conventions}. 973@gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol 974-ffixed-@var{reg} -fexceptions @gol 975-fnon-call-exceptions -funwind-tables @gol 976-fasynchronous-unwind-tables @gol 977-finhibit-size-directive -finstrument-functions @gol 978-finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol 979-finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol 980-fno-common -fno-ident @gol 981-fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol 982-fno-jump-tables @gol 983-frecord-gcc-switches @gol 984-freg-struct-return -fshort-enums @gol 985-fshort-double -fshort-wchar @gol 986-fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol 987-fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol 988-fno-stack-limit -fsplit-stack @gol 989-fleading-underscore -ftls-model=@var{model} @gol 990-ftrapv -fwrapv -fbounds-check @gol 991-fvisibility -fstrict-volatile-bitfields} 992@end table 993 994@menu 995* Overall Options:: Controlling the kind of output: 996 an executable, object files, assembler files, 997 or preprocessed source. 998* C Dialect Options:: Controlling the variant of C language compiled. 999* C++ Dialect Options:: Variations on C++. 1000* Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C 1001 and Objective-C++. 1002* Language Independent Options:: Controlling how diagnostics should be 1003 formatted. 1004* Warning Options:: How picky should the compiler be? 1005* Debugging Options:: Symbol tables, measurements, and debugging dumps. 1006* Optimize Options:: How much optimization? 1007* Preprocessor Options:: Controlling header files and macro definitions. 1008 Also, getting dependency information for Make. 1009* Assembler Options:: Passing options to the assembler. 1010* Link Options:: Specifying libraries and so on. 1011* Directory Options:: Where to find header files and libraries. 1012 Where to find the compiler executable files. 1013* Spec Files:: How to pass switches to sub-processes. 1014* Target Options:: Running a cross-compiler, or an old version of GCC. 1015@end menu 1016 1017@node Overall Options 1018@section Options Controlling the Kind of Output 1019 1020Compilation can involve up to four stages: preprocessing, compilation 1021proper, assembly and linking, always in that order. GCC is capable of 1022preprocessing and compiling several files either into several 1023assembler input files, or into one assembler input file; then each 1024assembler input file produces an object file, and linking combines all 1025the object files (those newly compiled, and those specified as input) 1026into an executable file. 1027 1028@cindex file name suffix 1029For any given input file, the file name suffix determines what kind of 1030compilation is done: 1031 1032@table @gcctabopt 1033@item @var{file}.c 1034C source code that must be preprocessed. 1035 1036@item @var{file}.i 1037C source code that should not be preprocessed. 1038 1039@item @var{file}.ii 1040C++ source code that should not be preprocessed. 1041 1042@item @var{file}.m 1043Objective-C source code. Note that you must link with the @file{libobjc} 1044library to make an Objective-C program work. 1045 1046@item @var{file}.mi 1047Objective-C source code that should not be preprocessed. 1048 1049@item @var{file}.mm 1050@itemx @var{file}.M 1051Objective-C++ source code. Note that you must link with the @file{libobjc} 1052library to make an Objective-C++ program work. Note that @samp{.M} refers 1053to a literal capital M@. 1054 1055@item @var{file}.mii 1056Objective-C++ source code that should not be preprocessed. 1057 1058@item @var{file}.h 1059C, C++, Objective-C or Objective-C++ header file to be turned into a 1060precompiled header (default), or C, C++ header file to be turned into an 1061Ada spec (via the @option{-fdump-ada-spec} switch). 1062 1063@item @var{file}.cc 1064@itemx @var{file}.cp 1065@itemx @var{file}.cxx 1066@itemx @var{file}.cpp 1067@itemx @var{file}.CPP 1068@itemx @var{file}.c++ 1069@itemx @var{file}.C 1070C++ source code that must be preprocessed. Note that in @samp{.cxx}, 1071the last two letters must both be literally @samp{x}. Likewise, 1072@samp{.C} refers to a literal capital C@. 1073 1074@item @var{file}.mm 1075@itemx @var{file}.M 1076Objective-C++ source code that must be preprocessed. 1077 1078@item @var{file}.mii 1079Objective-C++ source code that should not be preprocessed. 1080 1081@item @var{file}.hh 1082@itemx @var{file}.H 1083@itemx @var{file}.hp 1084@itemx @var{file}.hxx 1085@itemx @var{file}.hpp 1086@itemx @var{file}.HPP 1087@itemx @var{file}.h++ 1088@itemx @var{file}.tcc 1089C++ header file to be turned into a precompiled header or Ada spec. 1090 1091@item @var{file}.f 1092@itemx @var{file}.for 1093@itemx @var{file}.ftn 1094Fixed form Fortran source code that should not be preprocessed. 1095 1096@item @var{file}.F 1097@itemx @var{file}.FOR 1098@itemx @var{file}.fpp 1099@itemx @var{file}.FPP 1100@itemx @var{file}.FTN 1101Fixed form Fortran source code that must be preprocessed (with the traditional 1102preprocessor). 1103 1104@item @var{file}.f90 1105@itemx @var{file}.f95 1106@itemx @var{file}.f03 1107@itemx @var{file}.f08 1108Free form Fortran source code that should not be preprocessed. 1109 1110@item @var{file}.F90 1111@itemx @var{file}.F95 1112@itemx @var{file}.F03 1113@itemx @var{file}.F08 1114Free form Fortran source code that must be preprocessed (with the 1115traditional preprocessor). 1116 1117@item @var{file}.go 1118Go source code. 1119 1120@c FIXME: Descriptions of Java file types. 1121@c @var{file}.java 1122@c @var{file}.class 1123@c @var{file}.zip 1124@c @var{file}.jar 1125 1126@item @var{file}.ads 1127Ada source code file that contains a library unit declaration (a 1128declaration of a package, subprogram, or generic, or a generic 1129instantiation), or a library unit renaming declaration (a package, 1130generic, or subprogram renaming declaration). Such files are also 1131called @dfn{specs}. 1132 1133@item @var{file}.adb 1134Ada source code file containing a library unit body (a subprogram or 1135package body). Such files are also called @dfn{bodies}. 1136 1137@c GCC also knows about some suffixes for languages not yet included: 1138@c Pascal: 1139@c @var{file}.p 1140@c @var{file}.pas 1141@c Ratfor: 1142@c @var{file}.r 1143 1144@item @var{file}.s 1145Assembler code. 1146 1147@item @var{file}.S 1148@itemx @var{file}.sx 1149Assembler code that must be preprocessed. 1150 1151@item @var{other} 1152An object file to be fed straight into linking. 1153Any file name with no recognized suffix is treated this way. 1154@end table 1155 1156@opindex x 1157You can specify the input language explicitly with the @option{-x} option: 1158 1159@table @gcctabopt 1160@item -x @var{language} 1161Specify explicitly the @var{language} for the following input files 1162(rather than letting the compiler choose a default based on the file 1163name suffix). This option applies to all following input files until 1164the next @option{-x} option. Possible values for @var{language} are: 1165@smallexample 1166c c-header cpp-output 1167c++ c++-header c++-cpp-output 1168objective-c objective-c-header objective-c-cpp-output 1169objective-c++ objective-c++-header objective-c++-cpp-output 1170assembler assembler-with-cpp 1171ada 1172f77 f77-cpp-input f95 f95-cpp-input 1173go 1174java 1175@end smallexample 1176 1177@item -x none 1178Turn off any specification of a language, so that subsequent files are 1179handled according to their file name suffixes (as they are if @option{-x} 1180has not been used at all). 1181 1182@item -pass-exit-codes 1183@opindex pass-exit-codes 1184Normally the @command{gcc} program will exit with the code of 1 if any 1185phase of the compiler returns a non-success return code. If you specify 1186@option{-pass-exit-codes}, the @command{gcc} program will instead return with 1187numerically highest error produced by any phase that returned an error 1188indication. The C, C++, and Fortran frontends return 4, if an internal 1189compiler error is encountered. 1190@end table 1191 1192If you only want some of the stages of compilation, you can use 1193@option{-x} (or filename suffixes) to tell @command{gcc} where to start, and 1194one of the options @option{-c}, @option{-S}, or @option{-E} to say where 1195@command{gcc} is to stop. Note that some combinations (for example, 1196@samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all. 1197 1198@table @gcctabopt 1199@item -c 1200@opindex c 1201Compile or assemble the source files, but do not link. The linking 1202stage simply is not done. The ultimate output is in the form of an 1203object file for each source file. 1204 1205By default, the object file name for a source file is made by replacing 1206the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}. 1207 1208Unrecognized input files, not requiring compilation or assembly, are 1209ignored. 1210 1211@item -S 1212@opindex S 1213Stop after the stage of compilation proper; do not assemble. The output 1214is in the form of an assembler code file for each non-assembler input 1215file specified. 1216 1217By default, the assembler file name for a source file is made by 1218replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}. 1219 1220Input files that don't require compilation are ignored. 1221 1222@item -E 1223@opindex E 1224Stop after the preprocessing stage; do not run the compiler proper. The 1225output is in the form of preprocessed source code, which is sent to the 1226standard output. 1227 1228Input files that don't require preprocessing are ignored. 1229 1230@cindex output file option 1231@item -o @var{file} 1232@opindex o 1233Place output in file @var{file}. This applies regardless to whatever 1234sort of output is being produced, whether it be an executable file, 1235an object file, an assembler file or preprocessed C code. 1236 1237If @option{-o} is not specified, the default is to put an executable 1238file in @file{a.out}, the object file for 1239@file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its 1240assembler file in @file{@var{source}.s}, a precompiled header file in 1241@file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on 1242standard output. 1243 1244@item -v 1245@opindex v 1246Print (on standard error output) the commands executed to run the stages 1247of compilation. Also print the version number of the compiler driver 1248program and of the preprocessor and the compiler proper. 1249 1250@item -### 1251@opindex ### 1252Like @option{-v} except the commands are not executed and arguments 1253are quoted unless they contain only alphanumeric characters or @code{./-_}. 1254This is useful for shell scripts to capture the driver-generated command lines. 1255 1256@item -pipe 1257@opindex pipe 1258Use pipes rather than temporary files for communication between the 1259various stages of compilation. This fails to work on some systems where 1260the assembler is unable to read from a pipe; but the GNU assembler has 1261no trouble. 1262 1263@item --help 1264@opindex help 1265Print (on the standard output) a description of the command-line options 1266understood by @command{gcc}. If the @option{-v} option is also specified 1267then @option{--help} will also be passed on to the various processes 1268invoked by @command{gcc}, so that they can display the command-line options 1269they accept. If the @option{-Wextra} option has also been specified 1270(prior to the @option{--help} option), then command-line options that 1271have no documentation associated with them will also be displayed. 1272 1273@item --target-help 1274@opindex target-help 1275Print (on the standard output) a description of target-specific command-line 1276options for each tool. For some targets extra target-specific 1277information may also be printed. 1278 1279@item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]} 1280Print (on the standard output) a description of the command-line 1281options understood by the compiler that fit into all specified classes 1282and qualifiers. These are the supported classes: 1283 1284@table @asis 1285@item @samp{optimizers} 1286This will display all of the optimization options supported by the 1287compiler. 1288 1289@item @samp{warnings} 1290This will display all of the options controlling warning messages 1291produced by the compiler. 1292 1293@item @samp{target} 1294This will display target-specific options. Unlike the 1295@option{--target-help} option however, target-specific options of the 1296linker and assembler will not be displayed. This is because those 1297tools do not currently support the extended @option{--help=} syntax. 1298 1299@item @samp{params} 1300This will display the values recognized by the @option{--param} 1301option. 1302 1303@item @var{language} 1304This will display the options supported for @var{language}, where 1305@var{language} is the name of one of the languages supported in this 1306version of GCC. 1307 1308@item @samp{common} 1309This will display the options that are common to all languages. 1310@end table 1311 1312These are the supported qualifiers: 1313 1314@table @asis 1315@item @samp{undocumented} 1316Display only those options that are undocumented. 1317 1318@item @samp{joined} 1319Display options taking an argument that appears after an equal 1320sign in the same continuous piece of text, such as: 1321@samp{--help=target}. 1322 1323@item @samp{separate} 1324Display options taking an argument that appears as a separate word 1325following the original option, such as: @samp{-o output-file}. 1326@end table 1327 1328Thus for example to display all the undocumented target-specific 1329switches supported by the compiler the following can be used: 1330 1331@smallexample 1332--help=target,undocumented 1333@end smallexample 1334 1335The sense of a qualifier can be inverted by prefixing it with the 1336@samp{^} character, so for example to display all binary warning 1337options (i.e., ones that are either on or off and that do not take an 1338argument) that have a description, use: 1339 1340@smallexample 1341--help=warnings,^joined,^undocumented 1342@end smallexample 1343 1344The argument to @option{--help=} should not consist solely of inverted 1345qualifiers. 1346 1347Combining several classes is possible, although this usually 1348restricts the output by so much that there is nothing to display. One 1349case where it does work however is when one of the classes is 1350@var{target}. So for example to display all the target-specific 1351optimization options the following can be used: 1352 1353@smallexample 1354--help=target,optimizers 1355@end smallexample 1356 1357The @option{--help=} option can be repeated on the command line. Each 1358successive use will display its requested class of options, skipping 1359those that have already been displayed. 1360 1361If the @option{-Q} option appears on the command line before the 1362@option{--help=} option, then the descriptive text displayed by 1363@option{--help=} is changed. Instead of describing the displayed 1364options, an indication is given as to whether the option is enabled, 1365disabled or set to a specific value (assuming that the compiler 1366knows this at the point where the @option{--help=} option is used). 1367 1368Here is a truncated example from the ARM port of @command{gcc}: 1369 1370@smallexample 1371 % gcc -Q -mabi=2 --help=target -c 1372 The following options are target specific: 1373 -mabi= 2 1374 -mabort-on-noreturn [disabled] 1375 -mapcs [disabled] 1376@end smallexample 1377 1378The output is sensitive to the effects of previous command-line 1379options, so for example it is possible to find out which optimizations 1380are enabled at @option{-O2} by using: 1381 1382@smallexample 1383-Q -O2 --help=optimizers 1384@end smallexample 1385 1386Alternatively you can discover which binary optimizations are enabled 1387by @option{-O3} by using: 1388 1389@smallexample 1390gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 1391gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 1392diff /tmp/O2-opts /tmp/O3-opts | grep enabled 1393@end smallexample 1394 1395@item -no-canonical-prefixes 1396@opindex no-canonical-prefixes 1397Do not expand any symbolic links, resolve references to @samp{/../} 1398or @samp{/./}, or make the path absolute when generating a relative 1399prefix. 1400 1401@item --version 1402@opindex version 1403Display the version number and copyrights of the invoked GCC@. 1404 1405@item -wrapper 1406@opindex wrapper 1407Invoke all subcommands under a wrapper program. The name of the 1408wrapper program and its parameters are passed as a comma separated 1409list. 1410 1411@smallexample 1412gcc -c t.c -wrapper gdb,--args 1413@end smallexample 1414 1415This will invoke all subprograms of @command{gcc} under 1416@samp{gdb --args}, thus the invocation of @command{cc1} will be 1417@samp{gdb --args cc1 @dots{}}. 1418 1419@item -fplugin=@var{name}.so 1420Load the plugin code in file @var{name}.so, assumed to be a 1421shared object to be dlopen'd by the compiler. The base name of 1422the shared object file is used to identify the plugin for the 1423purposes of argument parsing (See 1424@option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below). 1425Each plugin should define the callback functions specified in the 1426Plugins API. 1427 1428@item -fplugin-arg-@var{name}-@var{key}=@var{value} 1429Define an argument called @var{key} with a value of @var{value} 1430for the plugin called @var{name}. 1431 1432@item -fdump-ada-spec@r{[}-slim@r{]} 1433For C and C++ source and include files, generate corresponding Ada 1434specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn, 1435GNAT User's Guide}, which provides detailed documentation on this feature. 1436 1437@item -fdump-go-spec=@var{file} 1438For input files in any language, generate corresponding Go 1439declarations in @var{file}. This generates Go @code{const}, 1440@code{type}, @code{var}, and @code{func} declarations which may be a 1441useful way to start writing a Go interface to code written in some 1442other language. 1443 1444@include @value{srcdir}/../libiberty/at-file.texi 1445@end table 1446 1447@node Invoking G++ 1448@section Compiling C++ Programs 1449 1450@cindex suffixes for C++ source 1451@cindex C++ source file suffixes 1452C++ source files conventionally use one of the suffixes @samp{.C}, 1453@samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or 1454@samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp}, 1455@samp{.H}, or (for shared template code) @samp{.tcc}; and 1456preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes 1457files with these names and compiles them as C++ programs even if you 1458call the compiler the same way as for compiling C programs (usually 1459with the name @command{gcc}). 1460 1461@findex g++ 1462@findex c++ 1463However, the use of @command{gcc} does not add the C++ library. 1464@command{g++} is a program that calls GCC and treats @samp{.c}, 1465@samp{.h} and @samp{.i} files as C++ source files instead of C source 1466files unless @option{-x} is used, and automatically specifies linking 1467against the C++ library. This program is also useful when 1468precompiling a C header file with a @samp{.h} extension for use in C++ 1469compilations. On many systems, @command{g++} is also installed with 1470the name @command{c++}. 1471 1472@cindex invoking @command{g++} 1473When you compile C++ programs, you may specify many of the same 1474command-line options that you use for compiling programs in any 1475language; or command-line options meaningful for C and related 1476languages; or options that are meaningful only for C++ programs. 1477@xref{C Dialect Options,,Options Controlling C Dialect}, for 1478explanations of options for languages related to C@. 1479@xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for 1480explanations of options that are meaningful only for C++ programs. 1481 1482@node C Dialect Options 1483@section Options Controlling C Dialect 1484@cindex dialect options 1485@cindex language dialect options 1486@cindex options, dialect 1487 1488The following options control the dialect of C (or languages derived 1489from C, such as C++, Objective-C and Objective-C++) that the compiler 1490accepts: 1491 1492@table @gcctabopt 1493@cindex ANSI support 1494@cindex ISO support 1495@item -ansi 1496@opindex ansi 1497In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is 1498equivalent to @samp{-std=c++98}. 1499 1500This turns off certain features of GCC that are incompatible with ISO 1501C90 (when compiling C code), or of standard C++ (when compiling C++ code), 1502such as the @code{asm} and @code{typeof} keywords, and 1503predefined macros such as @code{unix} and @code{vax} that identify the 1504type of system you are using. It also enables the undesirable and 1505rarely used ISO trigraph feature. For the C compiler, 1506it disables recognition of C++ style @samp{//} comments as well as 1507the @code{inline} keyword. 1508 1509The alternate keywords @code{__asm__}, @code{__extension__}, 1510@code{__inline__} and @code{__typeof__} continue to work despite 1511@option{-ansi}. You would not want to use them in an ISO C program, of 1512course, but it is useful to put them in header files that might be included 1513in compilations done with @option{-ansi}. Alternate predefined macros 1514such as @code{__unix__} and @code{__vax__} are also available, with or 1515without @option{-ansi}. 1516 1517The @option{-ansi} option does not cause non-ISO programs to be 1518rejected gratuitously. For that, @option{-pedantic} is required in 1519addition to @option{-ansi}. @xref{Warning Options}. 1520 1521The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi} 1522option is used. Some header files may notice this macro and refrain 1523from declaring certain functions or defining certain macros that the 1524ISO standard doesn't call for; this is to avoid interfering with any 1525programs that might use these names for other things. 1526 1527Functions that would normally be built in but do not have semantics 1528defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in 1529functions when @option{-ansi} is used. @xref{Other Builtins,,Other 1530built-in functions provided by GCC}, for details of the functions 1531affected. 1532 1533@item -std= 1534@opindex std 1535Determine the language standard. @xref{Standards,,Language Standards 1536Supported by GCC}, for details of these standard versions. This option 1537is currently only supported when compiling C or C++. 1538 1539The compiler can accept several base standards, such as @samp{c90} or 1540@samp{c++98}, and GNU dialects of those standards, such as 1541@samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the 1542compiler will accept all programs following that standard and those 1543using GNU extensions that do not contradict it. For example, 1544@samp{-std=c90} turns off certain features of GCC that are 1545incompatible with ISO C90, such as the @code{asm} and @code{typeof} 1546keywords, but not other GNU extensions that do not have a meaning in 1547ISO C90, such as omitting the middle term of a @code{?:} 1548expression. On the other hand, by specifying a GNU dialect of a 1549standard, all features the compiler support are enabled, even when 1550those features change the meaning of the base standard and some 1551strict-conforming programs may be rejected. The particular standard 1552is used by @option{-pedantic} to identify which features are GNU 1553extensions given that version of the standard. For example 1554@samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//} 1555comments, while @samp{-std=gnu99 -pedantic} would not. 1556 1557A value for this option must be provided; possible values are 1558 1559@table @samp 1560@item c90 1561@itemx c89 1562@itemx iso9899:1990 1563Support all ISO C90 programs (certain GNU extensions that conflict 1564with ISO C90 are disabled). Same as @option{-ansi} for C code. 1565 1566@item iso9899:199409 1567ISO C90 as modified in amendment 1. 1568 1569@item c99 1570@itemx c9x 1571@itemx iso9899:1999 1572@itemx iso9899:199x 1573ISO C99. Note that this standard is not yet fully supported; see 1574@w{@uref{http://gcc.gnu.org/gcc-4.7/c99status.html}} for more information. The 1575names @samp{c9x} and @samp{iso9899:199x} are deprecated. 1576 1577@item c11 1578@itemx c1x 1579@itemx iso9899:2011 1580ISO C11, the 2011 revision of the ISO C standard. 1581Support is incomplete and experimental. The name @samp{c1x} is 1582deprecated. 1583 1584@item gnu90 1585@itemx gnu89 1586GNU dialect of ISO C90 (including some C99 features). This 1587is the default for C code. 1588 1589@item gnu99 1590@itemx gnu9x 1591GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC, 1592this will become the default. The name @samp{gnu9x} is deprecated. 1593 1594@item gnu11 1595@item gnu1x 1596GNU dialect of ISO C11. Support is incomplete and experimental. The 1597name @samp{gnu1x} is deprecated. 1598 1599@item c++98 1600The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for 1601C++ code. 1602 1603@item gnu++98 1604GNU dialect of @option{-std=c++98}. This is the default for 1605C++ code. 1606 1607@item c++11 1608The 2011 ISO C++ standard plus amendments. Support for C++11 is still 1609experimental, and may change in incompatible ways in future releases. 1610 1611@item gnu++11 1612GNU dialect of @option{-std=c++11}. Support for C++11 is still 1613experimental, and may change in incompatible ways in future releases. 1614@end table 1615 1616@item -fgnu89-inline 1617@opindex fgnu89-inline 1618The option @option{-fgnu89-inline} tells GCC to use the traditional 1619GNU semantics for @code{inline} functions when in C99 mode. 1620@xref{Inline,,An Inline Function is As Fast As a Macro}. This option 1621is accepted and ignored by GCC versions 4.1.3 up to but not including 16224.3. In GCC versions 4.3 and later it changes the behavior of GCC in 1623C99 mode. Using this option is roughly equivalent to adding the 1624@code{gnu_inline} function attribute to all inline functions 1625(@pxref{Function Attributes}). 1626 1627The option @option{-fno-gnu89-inline} explicitly tells GCC to use the 1628C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it 1629specifies the default behavior). This option was first supported in 1630GCC 4.3. This option is not supported in @option{-std=c90} or 1631@option{-std=gnu90} mode. 1632 1633The preprocessor macros @code{__GNUC_GNU_INLINE__} and 1634@code{__GNUC_STDC_INLINE__} may be used to check which semantics are 1635in effect for @code{inline} functions. @xref{Common Predefined 1636Macros,,,cpp,The C Preprocessor}. 1637 1638@item -aux-info @var{filename} 1639@opindex aux-info 1640Output to the given filename prototyped declarations for all functions 1641declared and/or defined in a translation unit, including those in header 1642files. This option is silently ignored in any language other than C@. 1643 1644Besides declarations, the file indicates, in comments, the origin of 1645each declaration (source file and line), whether the declaration was 1646implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or 1647@samp{O} for old, respectively, in the first character after the line 1648number and the colon), and whether it came from a declaration or a 1649definition (@samp{C} or @samp{F}, respectively, in the following 1650character). In the case of function definitions, a K&R-style list of 1651arguments followed by their declarations is also provided, inside 1652comments, after the declaration. 1653 1654@item -fallow-parameterless-variadic-functions 1655Accept variadic functions without named parameters. 1656 1657Although it is possible to define such a function, this is not very 1658useful as it is not possible to read the arguments. This is only 1659supported for C as this construct is allowed by C++. 1660 1661@item -fno-asm 1662@opindex fno-asm 1663Do not recognize @code{asm}, @code{inline} or @code{typeof} as a 1664keyword, so that code can use these words as identifiers. You can use 1665the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__} 1666instead. @option{-ansi} implies @option{-fno-asm}. 1667 1668In C++, this switch only affects the @code{typeof} keyword, since 1669@code{asm} and @code{inline} are standard keywords. You may want to 1670use the @option{-fno-gnu-keywords} flag instead, which has the same 1671effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this 1672switch only affects the @code{asm} and @code{typeof} keywords, since 1673@code{inline} is a standard keyword in ISO C99. 1674 1675@item -fno-builtin 1676@itemx -fno-builtin-@var{function} 1677@opindex fno-builtin 1678@cindex built-in functions 1679Don't recognize built-in functions that do not begin with 1680@samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in 1681functions provided by GCC}, for details of the functions affected, 1682including those which are not built-in functions when @option{-ansi} or 1683@option{-std} options for strict ISO C conformance are used because they 1684do not have an ISO standard meaning. 1685 1686GCC normally generates special code to handle certain built-in functions 1687more efficiently; for instance, calls to @code{alloca} may become single 1688instructions which adjust the stack directly, and calls to @code{memcpy} 1689may become inline copy loops. The resulting code is often both smaller 1690and faster, but since the function calls no longer appear as such, you 1691cannot set a breakpoint on those calls, nor can you change the behavior 1692of the functions by linking with a different library. In addition, 1693when a function is recognized as a built-in function, GCC may use 1694information about that function to warn about problems with calls to 1695that function, or to generate more efficient code, even if the 1696resulting code still contains calls to that function. For example, 1697warnings are given with @option{-Wformat} for bad calls to 1698@code{printf}, when @code{printf} is built in, and @code{strlen} is 1699known not to modify global memory. 1700 1701With the @option{-fno-builtin-@var{function}} option 1702only the built-in function @var{function} is 1703disabled. @var{function} must not begin with @samp{__builtin_}. If a 1704function is named that is not built-in in this version of GCC, this 1705option is ignored. There is no corresponding 1706@option{-fbuiltin-@var{function}} option; if you wish to enable 1707built-in functions selectively when using @option{-fno-builtin} or 1708@option{-ffreestanding}, you may define macros such as: 1709 1710@smallexample 1711#define abs(n) __builtin_abs ((n)) 1712#define strcpy(d, s) __builtin_strcpy ((d), (s)) 1713@end smallexample 1714 1715@item -fhosted 1716@opindex fhosted 1717@cindex hosted environment 1718 1719Assert that compilation takes place in a hosted environment. This implies 1720@option{-fbuiltin}. A hosted environment is one in which the 1721entire standard library is available, and in which @code{main} has a return 1722type of @code{int}. Examples are nearly everything except a kernel. 1723This is equivalent to @option{-fno-freestanding}. 1724 1725@item -ffreestanding 1726@opindex ffreestanding 1727@cindex hosted environment 1728 1729Assert that compilation takes place in a freestanding environment. This 1730implies @option{-fno-builtin}. A freestanding environment 1731is one in which the standard library may not exist, and program startup may 1732not necessarily be at @code{main}. The most obvious example is an OS kernel. 1733This is equivalent to @option{-fno-hosted}. 1734 1735@xref{Standards,,Language Standards Supported by GCC}, for details of 1736freestanding and hosted environments. 1737 1738@item -fopenmp 1739@opindex fopenmp 1740@cindex OpenMP parallel 1741Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and 1742@code{!$omp} in Fortran. When @option{-fopenmp} is specified, the 1743compiler generates parallel code according to the OpenMP Application 1744Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option 1745implies @option{-pthread}, and thus is only supported on targets that 1746have support for @option{-pthread}. 1747 1748@item -fgnu-tm 1749@opindex fgnu-tm 1750When the option @option{-fgnu-tm} is specified, the compiler will 1751generate code for the Linux variant of Intel's current Transactional 1752Memory ABI specification document (Revision 1.1, May 6 2009). This is 1753an experimental feature whose interface may change in future versions 1754of GCC, as the official specification changes. Please note that not 1755all architectures are supported for this feature. 1756 1757For more information on GCC's support for transactional memory, 1758@xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU 1759Transactional Memory Library}. 1760 1761Note that the transactional memory feature is not supported with 1762non-call exceptions (@option{-fnon-call-exceptions}). 1763 1764@item -fms-extensions 1765@opindex fms-extensions 1766Accept some non-standard constructs used in Microsoft header files. 1767 1768In C++ code, this allows member names in structures to be similar 1769to previous types declarations. 1770 1771@smallexample 1772typedef int UOW; 1773struct ABC @{ 1774 UOW UOW; 1775@}; 1776@end smallexample 1777 1778Some cases of unnamed fields in structures and unions are only 1779accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union 1780fields within structs/unions}, for details. 1781 1782@item -fplan9-extensions 1783Accept some non-standard constructs used in Plan 9 code. 1784 1785This enables @option{-fms-extensions}, permits passing pointers to 1786structures with anonymous fields to functions that expect pointers to 1787elements of the type of the field, and permits referring to anonymous 1788fields declared using a typedef. @xref{Unnamed Fields,,Unnamed 1789struct/union fields within structs/unions}, for details. This is only 1790supported for C, not C++. 1791 1792@item -trigraphs 1793@opindex trigraphs 1794Support ISO C trigraphs. The @option{-ansi} option (and @option{-std} 1795options for strict ISO C conformance) implies @option{-trigraphs}. 1796 1797@item -no-integrated-cpp 1798@opindex no-integrated-cpp 1799Performs a compilation in two passes: preprocessing and compiling. This 1800option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the 1801@option{-B} option. The user supplied compilation step can then add in 1802an additional preprocessing step after normal preprocessing but before 1803compiling. The default is to use the integrated cpp (internal cpp) 1804 1805The semantics of this option will change if "cc1", "cc1plus", and 1806"cc1obj" are merged. 1807 1808@cindex traditional C language 1809@cindex C language, traditional 1810@item -traditional 1811@itemx -traditional-cpp 1812@opindex traditional-cpp 1813@opindex traditional 1814Formerly, these options caused GCC to attempt to emulate a pre-standard 1815C compiler. They are now only supported with the @option{-E} switch. 1816The preprocessor continues to support a pre-standard mode. See the GNU 1817CPP manual for details. 1818 1819@item -fcond-mismatch 1820@opindex fcond-mismatch 1821Allow conditional expressions with mismatched types in the second and 1822third arguments. The value of such an expression is void. This option 1823is not supported for C++. 1824 1825@item -flax-vector-conversions 1826@opindex flax-vector-conversions 1827Allow implicit conversions between vectors with differing numbers of 1828elements and/or incompatible element types. This option should not be 1829used for new code. 1830 1831@item -funsigned-char 1832@opindex funsigned-char 1833Let the type @code{char} be unsigned, like @code{unsigned char}. 1834 1835Each kind of machine has a default for what @code{char} should 1836be. It is either like @code{unsigned char} by default or like 1837@code{signed char} by default. 1838 1839Ideally, a portable program should always use @code{signed char} or 1840@code{unsigned char} when it depends on the signedness of an object. 1841But many programs have been written to use plain @code{char} and 1842expect it to be signed, or expect it to be unsigned, depending on the 1843machines they were written for. This option, and its inverse, let you 1844make such a program work with the opposite default. 1845 1846The type @code{char} is always a distinct type from each of 1847@code{signed char} or @code{unsigned char}, even though its behavior 1848is always just like one of those two. 1849 1850@item -fsigned-char 1851@opindex fsigned-char 1852Let the type @code{char} be signed, like @code{signed char}. 1853 1854Note that this is equivalent to @option{-fno-unsigned-char}, which is 1855the negative form of @option{-funsigned-char}. Likewise, the option 1856@option{-fno-signed-char} is equivalent to @option{-funsigned-char}. 1857 1858@item -fsigned-bitfields 1859@itemx -funsigned-bitfields 1860@itemx -fno-signed-bitfields 1861@itemx -fno-unsigned-bitfields 1862@opindex fsigned-bitfields 1863@opindex funsigned-bitfields 1864@opindex fno-signed-bitfields 1865@opindex fno-unsigned-bitfields 1866These options control whether a bit-field is signed or unsigned, when the 1867declaration does not use either @code{signed} or @code{unsigned}. By 1868default, such a bit-field is signed, because this is consistent: the 1869basic integer types such as @code{int} are signed types. 1870@end table 1871 1872@node C++ Dialect Options 1873@section Options Controlling C++ Dialect 1874 1875@cindex compiler options, C++ 1876@cindex C++ options, command-line 1877@cindex options, C++ 1878This section describes the command-line options that are only meaningful 1879for C++ programs; but you can also use most of the GNU compiler options 1880regardless of what language your program is in. For example, you 1881might compile a file @code{firstClass.C} like this: 1882 1883@smallexample 1884g++ -g -frepo -O -c firstClass.C 1885@end smallexample 1886 1887@noindent 1888In this example, only @option{-frepo} is an option meant 1889only for C++ programs; you can use the other options with any 1890language supported by GCC@. 1891 1892Here is a list of options that are @emph{only} for compiling C++ programs: 1893 1894@table @gcctabopt 1895 1896@item -fabi-version=@var{n} 1897@opindex fabi-version 1898Use version @var{n} of the C++ ABI@. Version 2 is the version of the 1899C++ ABI that first appeared in G++ 3.4. Version 1 is the version of 1900the C++ ABI that first appeared in G++ 3.2. Version 0 will always be 1901the version that conforms most closely to the C++ ABI specification. 1902Therefore, the ABI obtained using version 0 will change as ABI bugs 1903are fixed. 1904 1905The default is version 2. 1906 1907Version 3 corrects an error in mangling a constant address as a 1908template argument. 1909 1910Version 4, which first appeared in G++ 4.5, implements a standard 1911mangling for vector types. 1912 1913Version 5, which first appeared in G++ 4.6, corrects the mangling of 1914attribute const/volatile on function pointer types, decltype of a 1915plain decl, and use of a function parameter in the declaration of 1916another parameter. 1917 1918Version 6, which first appeared in G++ 4.7, corrects the promotion 1919behavior of C++11 scoped enums and the mangling of template argument 1920packs, const/static_cast, prefix ++ and --, and a class scope function 1921used as a template argument. 1922 1923See also @option{-Wabi}. 1924 1925@item -fno-access-control 1926@opindex fno-access-control 1927Turn off all access checking. This switch is mainly useful for working 1928around bugs in the access control code. 1929 1930@item -fcheck-new 1931@opindex fcheck-new 1932Check that the pointer returned by @code{operator new} is non-null 1933before attempting to modify the storage allocated. This check is 1934normally unnecessary because the C++ standard specifies that 1935@code{operator new} will only return @code{0} if it is declared 1936@samp{throw()}, in which case the compiler will always check the 1937return value even without this option. In all other cases, when 1938@code{operator new} has a non-empty exception specification, memory 1939exhaustion is signalled by throwing @code{std::bad_alloc}. See also 1940@samp{new (nothrow)}. 1941 1942@item -fconserve-space 1943@opindex fconserve-space 1944Put uninitialized or run-time-initialized global variables into the 1945common segment, as C does. This saves space in the executable at the 1946cost of not diagnosing duplicate definitions. If you compile with this 1947flag and your program mysteriously crashes after @code{main()} has 1948completed, you may have an object that is being destroyed twice because 1949two definitions were merged. 1950 1951This option is no longer useful on most targets, now that support has 1952been added for putting variables into BSS without making them common. 1953 1954@item -fconstexpr-depth=@var{n} 1955@opindex fconstexpr-depth 1956Set the maximum nested evaluation depth for C++11 constexpr functions 1957to @var{n}. A limit is needed to detect endless recursion during 1958constant expression evaluation. The minimum specified by the standard 1959is 512. 1960 1961@item -fdeduce-init-list 1962@opindex fdeduce-init-list 1963Enable deduction of a template type parameter as 1964std::initializer_list from a brace-enclosed initializer list, i.e. 1965 1966@smallexample 1967template <class T> auto forward(T t) -> decltype (realfn (t)) 1968@{ 1969 return realfn (t); 1970@} 1971 1972void f() 1973@{ 1974 forward(@{1,2@}); // call forward<std::initializer_list<int>> 1975@} 1976@end smallexample 1977 1978This deduction was implemented as a possible extension to the 1979originally proposed semantics for the C++11 standard, but was not part 1980of the final standard, so it is disabled by default. This option is 1981deprecated, and may be removed in a future version of G++. 1982 1983@item -ffriend-injection 1984@opindex ffriend-injection 1985Inject friend functions into the enclosing namespace, so that they are 1986visible outside the scope of the class in which they are declared. 1987Friend functions were documented to work this way in the old Annotated 1988C++ Reference Manual, and versions of G++ before 4.1 always worked 1989that way. However, in ISO C++ a friend function that is not declared 1990in an enclosing scope can only be found using argument dependent 1991lookup. This option causes friends to be injected as they were in 1992earlier releases. 1993 1994This option is for compatibility, and may be removed in a future 1995release of G++. 1996 1997@item -fno-elide-constructors 1998@opindex fno-elide-constructors 1999The C++ standard allows an implementation to omit creating a temporary 2000that is only used to initialize another object of the same type. 2001Specifying this option disables that optimization, and forces G++ to 2002call the copy constructor in all cases. 2003 2004@item -fno-enforce-eh-specs 2005@opindex fno-enforce-eh-specs 2006Don't generate code to check for violation of exception specifications 2007at run time. This option violates the C++ standard, but may be useful 2008for reducing code size in production builds, much like defining 2009@samp{NDEBUG}. This does not give user code permission to throw 2010exceptions in violation of the exception specifications; the compiler 2011will still optimize based on the specifications, so throwing an 2012unexpected exception will result in undefined behavior. 2013 2014@item -ffor-scope 2015@itemx -fno-for-scope 2016@opindex ffor-scope 2017@opindex fno-for-scope 2018If @option{-ffor-scope} is specified, the scope of variables declared in 2019a @i{for-init-statement} is limited to the @samp{for} loop itself, 2020as specified by the C++ standard. 2021If @option{-fno-for-scope} is specified, the scope of variables declared in 2022a @i{for-init-statement} extends to the end of the enclosing scope, 2023as was the case in old versions of G++, and other (traditional) 2024implementations of C++. 2025 2026The default if neither flag is given to follow the standard, 2027but to allow and give a warning for old-style code that would 2028otherwise be invalid, or have different behavior. 2029 2030@item -fno-gnu-keywords 2031@opindex fno-gnu-keywords 2032Do not recognize @code{typeof} as a keyword, so that code can use this 2033word as an identifier. You can use the keyword @code{__typeof__} instead. 2034@option{-ansi} implies @option{-fno-gnu-keywords}. 2035 2036@item -fno-implicit-templates 2037@opindex fno-implicit-templates 2038Never emit code for non-inline templates that are instantiated 2039implicitly (i.e.@: by use); only emit code for explicit instantiations. 2040@xref{Template Instantiation}, for more information. 2041 2042@item -fno-implicit-inline-templates 2043@opindex fno-implicit-inline-templates 2044Don't emit code for implicit instantiations of inline templates, either. 2045The default is to handle inlines differently so that compiles with and 2046without optimization will need the same set of explicit instantiations. 2047 2048@item -fno-implement-inlines 2049@opindex fno-implement-inlines 2050To save space, do not emit out-of-line copies of inline functions 2051controlled by @samp{#pragma implementation}. This will cause linker 2052errors if these functions are not inlined everywhere they are called. 2053 2054@item -fms-extensions 2055@opindex fms-extensions 2056Disable pedantic warnings about constructs used in MFC, such as implicit 2057int and getting a pointer to member function via non-standard syntax. 2058 2059@item -fno-nonansi-builtins 2060@opindex fno-nonansi-builtins 2061Disable built-in declarations of functions that are not mandated by 2062ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit}, 2063@code{index}, @code{bzero}, @code{conjf}, and other related functions. 2064 2065@item -fnothrow-opt 2066@opindex fnothrow-opt 2067Treat a @code{throw()} exception specification as though it were a 2068@code{noexcept} specification to reduce or eliminate the text size 2069overhead relative to a function with no exception specification. If 2070the function has local variables of types with non-trivial 2071destructors, the exception specification will actually make the 2072function smaller because the EH cleanups for those variables can be 2073optimized away. The semantic effect is that an exception thrown out of 2074a function with such an exception specification will result in a call 2075to @code{terminate} rather than @code{unexpected}. 2076 2077@item -fno-operator-names 2078@opindex fno-operator-names 2079Do not treat the operator name keywords @code{and}, @code{bitand}, 2080@code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as 2081synonyms as keywords. 2082 2083@item -fno-optional-diags 2084@opindex fno-optional-diags 2085Disable diagnostics that the standard says a compiler does not need to 2086issue. Currently, the only such diagnostic issued by G++ is the one for 2087a name having multiple meanings within a class. 2088 2089@item -fpermissive 2090@opindex fpermissive 2091Downgrade some diagnostics about nonconformant code from errors to 2092warnings. Thus, using @option{-fpermissive} will allow some 2093nonconforming code to compile. 2094 2095@item -fno-pretty-templates 2096@opindex fno-pretty-templates 2097When an error message refers to a specialization of a function 2098template, the compiler will normally print the signature of the 2099template followed by the template arguments and any typedefs or 2100typenames in the signature (e.g. @code{void f(T) [with T = int]} 2101rather than @code{void f(int)}) so that it's clear which template is 2102involved. When an error message refers to a specialization of a class 2103template, the compiler will omit any template arguments that match 2104the default template arguments for that template. If either of these 2105behaviors make it harder to understand the error message rather than 2106easier, using @option{-fno-pretty-templates} will disable them. 2107 2108@item -frepo 2109@opindex frepo 2110Enable automatic template instantiation at link time. This option also 2111implies @option{-fno-implicit-templates}. @xref{Template 2112Instantiation}, for more information. 2113 2114@item -fno-rtti 2115@opindex fno-rtti 2116Disable generation of information about every class with virtual 2117functions for use by the C++ run-time type identification features 2118(@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts 2119of the language, you can save some space by using this flag. Note that 2120exception handling uses the same information, but it will generate it as 2121needed. The @samp{dynamic_cast} operator can still be used for casts that 2122do not require run-time type information, i.e.@: casts to @code{void *} or to 2123unambiguous base classes. 2124 2125@item -fstats 2126@opindex fstats 2127Emit statistics about front-end processing at the end of the compilation. 2128This information is generally only useful to the G++ development team. 2129 2130@item -fstrict-enums 2131@opindex fstrict-enums 2132Allow the compiler to optimize using the assumption that a value of 2133enumerated type can only be one of the values of the enumeration (as 2134defined in the C++ standard; basically, a value that can be 2135represented in the minimum number of bits needed to represent all the 2136enumerators). This assumption may not be valid if the program uses a 2137cast to convert an arbitrary integer value to the enumerated type. 2138 2139@item -ftemplate-depth=@var{n} 2140@opindex ftemplate-depth 2141Set the maximum instantiation depth for template classes to @var{n}. 2142A limit on the template instantiation depth is needed to detect 2143endless recursions during template class instantiation. ANSI/ISO C++ 2144conforming programs must not rely on a maximum depth greater than 17 2145(changed to 1024 in C++11). The default value is 900, as the compiler 2146can run out of stack space before hitting 1024 in some situations. 2147 2148@item -fno-threadsafe-statics 2149@opindex fno-threadsafe-statics 2150Do not emit the extra code to use the routines specified in the C++ 2151ABI for thread-safe initialization of local statics. You can use this 2152option to reduce code size slightly in code that doesn't need to be 2153thread-safe. 2154 2155@item -fuse-cxa-atexit 2156@opindex fuse-cxa-atexit 2157Register destructors for objects with static storage duration with the 2158@code{__cxa_atexit} function rather than the @code{atexit} function. 2159This option is required for fully standards-compliant handling of static 2160destructors, but will only work if your C library supports 2161@code{__cxa_atexit}. 2162 2163@item -fno-use-cxa-get-exception-ptr 2164@opindex fno-use-cxa-get-exception-ptr 2165Don't use the @code{__cxa_get_exception_ptr} runtime routine. This 2166will cause @code{std::uncaught_exception} to be incorrect, but is necessary 2167if the runtime routine is not available. 2168 2169@item -fvisibility-inlines-hidden 2170@opindex fvisibility-inlines-hidden 2171This switch declares that the user does not attempt to compare 2172pointers to inline functions or methods where the addresses of the two functions 2173were taken in different shared objects. 2174 2175The effect of this is that GCC may, effectively, mark inline methods with 2176@code{__attribute__ ((visibility ("hidden")))} so that they do not 2177appear in the export table of a DSO and do not require a PLT indirection 2178when used within the DSO@. Enabling this option can have a dramatic effect 2179on load and link times of a DSO as it massively reduces the size of the 2180dynamic export table when the library makes heavy use of templates. 2181 2182The behavior of this switch is not quite the same as marking the 2183methods as hidden directly, because it does not affect static variables 2184local to the function or cause the compiler to deduce that 2185the function is defined in only one shared object. 2186 2187You may mark a method as having a visibility explicitly to negate the 2188effect of the switch for that method. For example, if you do want to 2189compare pointers to a particular inline method, you might mark it as 2190having default visibility. Marking the enclosing class with explicit 2191visibility will have no effect. 2192 2193Explicitly instantiated inline methods are unaffected by this option 2194as their linkage might otherwise cross a shared library boundary. 2195@xref{Template Instantiation}. 2196 2197@item -fvisibility-ms-compat 2198@opindex fvisibility-ms-compat 2199This flag attempts to use visibility settings to make GCC's C++ 2200linkage model compatible with that of Microsoft Visual Studio. 2201 2202The flag makes these changes to GCC's linkage model: 2203 2204@enumerate 2205@item 2206It sets the default visibility to @code{hidden}, like 2207@option{-fvisibility=hidden}. 2208 2209@item 2210Types, but not their members, are not hidden by default. 2211 2212@item 2213The One Definition Rule is relaxed for types without explicit 2214visibility specifications that are defined in more than one different 2215shared object: those declarations are permitted if they would have 2216been permitted when this option was not used. 2217@end enumerate 2218 2219In new code it is better to use @option{-fvisibility=hidden} and 2220export those classes that are intended to be externally visible. 2221Unfortunately it is possible for code to rely, perhaps accidentally, 2222on the Visual Studio behavior. 2223 2224Among the consequences of these changes are that static data members 2225of the same type with the same name but defined in different shared 2226objects will be different, so changing one will not change the other; 2227and that pointers to function members defined in different shared 2228objects may not compare equal. When this flag is given, it is a 2229violation of the ODR to define types with the same name differently. 2230 2231@item -fno-weak 2232@opindex fno-weak 2233Do not use weak symbol support, even if it is provided by the linker. 2234By default, G++ will use weak symbols if they are available. This 2235option exists only for testing, and should not be used by end-users; 2236it will result in inferior code and has no benefits. This option may 2237be removed in a future release of G++. 2238 2239@item -nostdinc++ 2240@opindex nostdinc++ 2241Do not search for header files in the standard directories specific to 2242C++, but do still search the other standard directories. (This option 2243is used when building the C++ library.) 2244@end table 2245 2246In addition, these optimization, warning, and code generation options 2247have meanings only for C++ programs: 2248 2249@table @gcctabopt 2250@item -fno-default-inline 2251@opindex fno-default-inline 2252Do not assume @samp{inline} for functions defined inside a class scope. 2253@xref{Optimize Options,,Options That Control Optimization}. Note that these 2254functions will have linkage like inline functions; they just won't be 2255inlined by default. 2256 2257@item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)} 2258@opindex Wabi 2259@opindex Wno-abi 2260Warn when G++ generates code that is probably not compatible with the 2261vendor-neutral C++ ABI@. Although an effort has been made to warn about 2262all such cases, there are probably some cases that are not warned about, 2263even though G++ is generating incompatible code. There may also be 2264cases where warnings are emitted even though the code that is generated 2265will be compatible. 2266 2267You should rewrite your code to avoid these warnings if you are 2268concerned about the fact that code generated by G++ may not be binary 2269compatible with code generated by other compilers. 2270 2271The known incompatibilities in @option{-fabi-version=2} (the default) include: 2272 2273@itemize @bullet 2274 2275@item 2276A template with a non-type template parameter of reference type is 2277mangled incorrectly: 2278@smallexample 2279extern int N; 2280template <int &> struct S @{@}; 2281void n (S<N>) @{2@} 2282@end smallexample 2283 2284This is fixed in @option{-fabi-version=3}. 2285 2286@item 2287SIMD vector types declared using @code{__attribute ((vector_size))} are 2288mangled in a non-standard way that does not allow for overloading of 2289functions taking vectors of different sizes. 2290 2291The mangling is changed in @option{-fabi-version=4}. 2292@end itemize 2293 2294The known incompatibilities in @option{-fabi-version=1} include: 2295 2296@itemize @bullet 2297 2298@item 2299Incorrect handling of tail-padding for bit-fields. G++ may attempt to 2300pack data into the same byte as a base class. For example: 2301 2302@smallexample 2303struct A @{ virtual void f(); int f1 : 1; @}; 2304struct B : public A @{ int f2 : 1; @}; 2305@end smallexample 2306 2307@noindent 2308In this case, G++ will place @code{B::f2} into the same byte 2309as@code{A::f1}; other compilers will not. You can avoid this problem 2310by explicitly padding @code{A} so that its size is a multiple of the 2311byte size on your platform; that will cause G++ and other compilers to 2312layout @code{B} identically. 2313 2314@item 2315Incorrect handling of tail-padding for virtual bases. G++ does not use 2316tail padding when laying out virtual bases. For example: 2317 2318@smallexample 2319struct A @{ virtual void f(); char c1; @}; 2320struct B @{ B(); char c2; @}; 2321struct C : public A, public virtual B @{@}; 2322@end smallexample 2323 2324@noindent 2325In this case, G++ will not place @code{B} into the tail-padding for 2326@code{A}; other compilers will. You can avoid this problem by 2327explicitly padding @code{A} so that its size is a multiple of its 2328alignment (ignoring virtual base classes); that will cause G++ and other 2329compilers to layout @code{C} identically. 2330 2331@item 2332Incorrect handling of bit-fields with declared widths greater than that 2333of their underlying types, when the bit-fields appear in a union. For 2334example: 2335 2336@smallexample 2337union U @{ int i : 4096; @}; 2338@end smallexample 2339 2340@noindent 2341Assuming that an @code{int} does not have 4096 bits, G++ will make the 2342union too small by the number of bits in an @code{int}. 2343 2344@item 2345Empty classes can be placed at incorrect offsets. For example: 2346 2347@smallexample 2348struct A @{@}; 2349 2350struct B @{ 2351 A a; 2352 virtual void f (); 2353@}; 2354 2355struct C : public B, public A @{@}; 2356@end smallexample 2357 2358@noindent 2359G++ will place the @code{A} base class of @code{C} at a nonzero offset; 2360it should be placed at offset zero. G++ mistakenly believes that the 2361@code{A} data member of @code{B} is already at offset zero. 2362 2363@item 2364Names of template functions whose types involve @code{typename} or 2365template template parameters can be mangled incorrectly. 2366 2367@smallexample 2368template <typename Q> 2369void f(typename Q::X) @{@} 2370 2371template <template <typename> class Q> 2372void f(typename Q<int>::X) @{@} 2373@end smallexample 2374 2375@noindent 2376Instantiations of these templates may be mangled incorrectly. 2377 2378@end itemize 2379 2380It also warns psABI related changes. The known psABI changes at this 2381point include: 2382 2383@itemize @bullet 2384 2385@item 2386For SYSV/x86-64, when passing union with long double, it is changed to 2387pass in memory as specified in psABI. For example: 2388 2389@smallexample 2390union U @{ 2391 long double ld; 2392 int i; 2393@}; 2394@end smallexample 2395 2396@noindent 2397@code{union U} will always be passed in memory. 2398 2399@end itemize 2400 2401@item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)} 2402@opindex Wctor-dtor-privacy 2403@opindex Wno-ctor-dtor-privacy 2404Warn when a class seems unusable because all the constructors or 2405destructors in that class are private, and it has neither friends nor 2406public static member functions. 2407 2408@item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)} 2409@opindex Wdelete-non-virtual-dtor 2410@opindex Wno-delete-non-virtual-dtor 2411Warn when @samp{delete} is used to destroy an instance of a class that 2412has virtual functions and non-virtual destructor. It is unsafe to delete 2413an instance of a derived class through a pointer to a base class if the 2414base class does not have a virtual destructor. This warning is enabled 2415by @option{-Wall}. 2416 2417@item -Wnarrowing @r{(C++ and Objective-C++ only)} 2418@opindex Wnarrowing 2419@opindex Wno-narrowing 2420Warn when a narrowing conversion prohibited by C++11 occurs within 2421@samp{@{ @}}, e.g. 2422 2423@smallexample 2424int i = @{ 2.2 @}; // error: narrowing from double to int 2425@end smallexample 2426 2427This flag is included in @option{-Wall} and @option{-Wc++11-compat}. 2428 2429With -std=c++11, @option{-Wno-narrowing} suppresses the diagnostic 2430required by the standard. Note that this does not affect the meaning 2431of well-formed code; narrowing conversions are still considered 2432ill-formed in SFINAE context. 2433 2434@item -Wnoexcept @r{(C++ and Objective-C++ only)} 2435@opindex Wnoexcept 2436@opindex Wno-noexcept 2437Warn when a noexcept-expression evaluates to false because of a call 2438to a function that does not have a non-throwing exception 2439specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by 2440the compiler to never throw an exception. 2441 2442@item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)} 2443@opindex Wnon-virtual-dtor 2444@opindex Wno-non-virtual-dtor 2445Warn when a class has virtual functions and accessible non-virtual 2446destructor, in which case it would be possible but unsafe to delete 2447an instance of a derived class through a pointer to the base class. 2448This warning is also enabled if @option{-Weffc++} is specified. 2449 2450@item -Wreorder @r{(C++ and Objective-C++ only)} 2451@opindex Wreorder 2452@opindex Wno-reorder 2453@cindex reordering, warning 2454@cindex warning for reordering of member initializers 2455Warn when the order of member initializers given in the code does not 2456match the order in which they must be executed. For instance: 2457 2458@smallexample 2459struct A @{ 2460 int i; 2461 int j; 2462 A(): j (0), i (1) @{ @} 2463@}; 2464@end smallexample 2465 2466The compiler will rearrange the member initializers for @samp{i} 2467and @samp{j} to match the declaration order of the members, emitting 2468a warning to that effect. This warning is enabled by @option{-Wall}. 2469@end table 2470 2471The following @option{-W@dots{}} options are not affected by @option{-Wall}. 2472 2473@table @gcctabopt 2474@item -Weffc++ @r{(C++ and Objective-C++ only)} 2475@opindex Weffc++ 2476@opindex Wno-effc++ 2477Warn about violations of the following style guidelines from Scott Meyers' 2478@cite{Effective C++, Second Edition} book: 2479 2480@itemize @bullet 2481@item 2482Item 11: Define a copy constructor and an assignment operator for classes 2483with dynamically allocated memory. 2484 2485@item 2486Item 12: Prefer initialization to assignment in constructors. 2487 2488@item 2489Item 14: Make destructors virtual in base classes. 2490 2491@item 2492Item 15: Have @code{operator=} return a reference to @code{*this}. 2493 2494@item 2495Item 23: Don't try to return a reference when you must return an object. 2496 2497@end itemize 2498 2499Also warn about violations of the following style guidelines from 2500Scott Meyers' @cite{More Effective C++} book: 2501 2502@itemize @bullet 2503@item 2504Item 6: Distinguish between prefix and postfix forms of increment and 2505decrement operators. 2506 2507@item 2508Item 7: Never overload @code{&&}, @code{||}, or @code{,}. 2509 2510@end itemize 2511 2512When selecting this option, be aware that the standard library 2513headers do not obey all of these guidelines; use @samp{grep -v} 2514to filter out those warnings. 2515 2516@item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)} 2517@opindex Wstrict-null-sentinel 2518@opindex Wno-strict-null-sentinel 2519Warn also about the use of an uncasted @code{NULL} as sentinel. When 2520compiling only with GCC this is a valid sentinel, as @code{NULL} is defined 2521to @code{__null}. Although it is a null pointer constant not a null pointer, 2522it is guaranteed to be of the same size as a pointer. But this use is 2523not portable across different compilers. 2524 2525@item -Wno-non-template-friend @r{(C++ and Objective-C++ only)} 2526@opindex Wno-non-template-friend 2527@opindex Wnon-template-friend 2528Disable warnings when non-templatized friend functions are declared 2529within a template. Since the advent of explicit template specification 2530support in G++, if the name of the friend is an unqualified-id (i.e., 2531@samp{friend foo(int)}), the C++ language specification demands that the 2532friend declare or define an ordinary, nontemplate function. (Section 253314.5.3). Before G++ implemented explicit specification, unqualified-ids 2534could be interpreted as a particular specialization of a templatized 2535function. Because this non-conforming behavior is no longer the default 2536behavior for G++, @option{-Wnon-template-friend} allows the compiler to 2537check existing code for potential trouble spots and is on by default. 2538This new compiler behavior can be turned off with 2539@option{-Wno-non-template-friend}, which keeps the conformant compiler code 2540but disables the helpful warning. 2541 2542@item -Wold-style-cast @r{(C++ and Objective-C++ only)} 2543@opindex Wold-style-cast 2544@opindex Wno-old-style-cast 2545Warn if an old-style (C-style) cast to a non-void type is used within 2546a C++ program. The new-style casts (@samp{dynamic_cast}, 2547@samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are 2548less vulnerable to unintended effects and much easier to search for. 2549 2550@item -Woverloaded-virtual @r{(C++ and Objective-C++ only)} 2551@opindex Woverloaded-virtual 2552@opindex Wno-overloaded-virtual 2553@cindex overloaded virtual function, warning 2554@cindex warning for overloaded virtual function 2555Warn when a function declaration hides virtual functions from a 2556base class. For example, in: 2557 2558@smallexample 2559struct A @{ 2560 virtual void f(); 2561@}; 2562 2563struct B: public A @{ 2564 void f(int); 2565@}; 2566@end smallexample 2567 2568the @code{A} class version of @code{f} is hidden in @code{B}, and code 2569like: 2570 2571@smallexample 2572B* b; 2573b->f(); 2574@end smallexample 2575 2576will fail to compile. 2577 2578@item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)} 2579@opindex Wno-pmf-conversions 2580@opindex Wpmf-conversions 2581Disable the diagnostic for converting a bound pointer to member function 2582to a plain pointer. 2583 2584@item -Wsign-promo @r{(C++ and Objective-C++ only)} 2585@opindex Wsign-promo 2586@opindex Wno-sign-promo 2587Warn when overload resolution chooses a promotion from unsigned or 2588enumerated type to a signed type, over a conversion to an unsigned type of 2589the same size. Previous versions of G++ would try to preserve 2590unsignedness, but the standard mandates the current behavior. 2591 2592@smallexample 2593struct A @{ 2594 operator int (); 2595 A& operator = (int); 2596@}; 2597 2598main () 2599@{ 2600 A a,b; 2601 a = b; 2602@} 2603@end smallexample 2604 2605In this example, G++ will synthesize a default @samp{A& operator = 2606(const A&);}, while cfront will use the user-defined @samp{operator =}. 2607@end table 2608 2609@node Objective-C and Objective-C++ Dialect Options 2610@section Options Controlling Objective-C and Objective-C++ Dialects 2611 2612@cindex compiler options, Objective-C and Objective-C++ 2613@cindex Objective-C and Objective-C++ options, command-line 2614@cindex options, Objective-C and Objective-C++ 2615(NOTE: This manual does not describe the Objective-C and Objective-C++ 2616languages themselves. @xref{Standards,,Language Standards 2617Supported by GCC}, for references.) 2618 2619This section describes the command-line options that are only meaningful 2620for Objective-C and Objective-C++ programs, but you can also use most of 2621the language-independent GNU compiler options. 2622For example, you might compile a file @code{some_class.m} like this: 2623 2624@smallexample 2625gcc -g -fgnu-runtime -O -c some_class.m 2626@end smallexample 2627 2628@noindent 2629In this example, @option{-fgnu-runtime} is an option meant only for 2630Objective-C and Objective-C++ programs; you can use the other options with 2631any language supported by GCC@. 2632 2633Note that since Objective-C is an extension of the C language, Objective-C 2634compilations may also use options specific to the C front-end (e.g., 2635@option{-Wtraditional}). Similarly, Objective-C++ compilations may use 2636C++-specific options (e.g., @option{-Wabi}). 2637 2638Here is a list of options that are @emph{only} for compiling Objective-C 2639and Objective-C++ programs: 2640 2641@table @gcctabopt 2642@item -fconstant-string-class=@var{class-name} 2643@opindex fconstant-string-class 2644Use @var{class-name} as the name of the class to instantiate for each 2645literal string specified with the syntax @code{@@"@dots{}"}. The default 2646class name is @code{NXConstantString} if the GNU runtime is being used, and 2647@code{NSConstantString} if the NeXT runtime is being used (see below). The 2648@option{-fconstant-cfstrings} option, if also present, will override the 2649@option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals 2650to be laid out as constant CoreFoundation strings. 2651 2652@item -fgnu-runtime 2653@opindex fgnu-runtime 2654Generate object code compatible with the standard GNU Objective-C 2655runtime. This is the default for most types of systems. 2656 2657@item -fnext-runtime 2658@opindex fnext-runtime 2659Generate output compatible with the NeXT runtime. This is the default 2660for NeXT-based systems, including Darwin and Mac OS X@. The macro 2661@code{__NEXT_RUNTIME__} is predefined if (and only if) this option is 2662used. 2663 2664@item -fno-nil-receivers 2665@opindex fno-nil-receivers 2666Assume that all Objective-C message dispatches (@code{[receiver 2667message:arg]}) in this translation unit ensure that the receiver is 2668not @code{nil}. This allows for more efficient entry points in the 2669runtime to be used. This option is only available in conjunction with 2670the NeXT runtime and ABI version 0 or 1. 2671 2672@item -fobjc-abi-version=@var{n} 2673@opindex fobjc-abi-version 2674Use version @var{n} of the Objective-C ABI for the selected runtime. 2675This option is currently supported only for the NeXT runtime. In that 2676case, Version 0 is the traditional (32-bit) ABI without support for 2677properties and other Objective-C 2.0 additions. Version 1 is the 2678traditional (32-bit) ABI with support for properties and other 2679Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If 2680nothing is specified, the default is Version 0 on 32-bit target 2681machines, and Version 2 on 64-bit target machines. 2682 2683@item -fobjc-call-cxx-cdtors 2684@opindex fobjc-call-cxx-cdtors 2685For each Objective-C class, check if any of its instance variables is a 2686C++ object with a non-trivial default constructor. If so, synthesize a 2687special @code{- (id) .cxx_construct} instance method which will run 2688non-trivial default constructors on any such instance variables, in order, 2689and then return @code{self}. Similarly, check if any instance variable 2690is a C++ object with a non-trivial destructor, and if so, synthesize a 2691special @code{- (void) .cxx_destruct} method which will run 2692all such default destructors, in reverse order. 2693 2694The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct} 2695methods thusly generated will only operate on instance variables 2696declared in the current Objective-C class, and not those inherited 2697from superclasses. It is the responsibility of the Objective-C 2698runtime to invoke all such methods in an object's inheritance 2699hierarchy. The @code{- (id) .cxx_construct} methods will be invoked 2700by the runtime immediately after a new object instance is allocated; 2701the @code{- (void) .cxx_destruct} methods will be invoked immediately 2702before the runtime deallocates an object instance. 2703 2704As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has 2705support for invoking the @code{- (id) .cxx_construct} and 2706@code{- (void) .cxx_destruct} methods. 2707 2708@item -fobjc-direct-dispatch 2709@opindex fobjc-direct-dispatch 2710Allow fast jumps to the message dispatcher. On Darwin this is 2711accomplished via the comm page. 2712 2713@item -fobjc-exceptions 2714@opindex fobjc-exceptions 2715Enable syntactic support for structured exception handling in 2716Objective-C, similar to what is offered by C++ and Java. This option 2717is required to use the Objective-C keywords @code{@@try}, 2718@code{@@throw}, @code{@@catch}, @code{@@finally} and 2719@code{@@synchronized}. This option is available with both the GNU 2720runtime and the NeXT runtime (but not available in conjunction with 2721the NeXT runtime on Mac OS X 10.2 and earlier). 2722 2723@item -fobjc-gc 2724@opindex fobjc-gc 2725Enable garbage collection (GC) in Objective-C and Objective-C++ 2726programs. This option is only available with the NeXT runtime; the 2727GNU runtime has a different garbage collection implementation that 2728does not require special compiler flags. 2729 2730@item -fobjc-nilcheck 2731@opindex fobjc-nilcheck 2732For the NeXT runtime with version 2 of the ABI, check for a nil 2733receiver in method invocations before doing the actual method call. 2734This is the default and can be disabled using 2735@option{-fno-objc-nilcheck}. Class methods and super calls are never 2736checked for nil in this way no matter what this flag is set to. 2737Currently this flag does nothing when the GNU runtime, or an older 2738version of the NeXT runtime ABI, is used. 2739 2740@item -fobjc-std=objc1 2741@opindex fobjc-std 2742Conform to the language syntax of Objective-C 1.0, the language 2743recognized by GCC 4.0. This only affects the Objective-C additions to 2744the C/C++ language; it does not affect conformance to C/C++ standards, 2745which is controlled by the separate C/C++ dialect option flags. When 2746this option is used with the Objective-C or Objective-C++ compiler, 2747any Objective-C syntax that is not recognized by GCC 4.0 is rejected. 2748This is useful if you need to make sure that your Objective-C code can 2749be compiled with older versions of GCC. 2750 2751@item -freplace-objc-classes 2752@opindex freplace-objc-classes 2753Emit a special marker instructing @command{ld(1)} not to statically link in 2754the resulting object file, and allow @command{dyld(1)} to load it in at 2755run time instead. This is used in conjunction with the Fix-and-Continue 2756debugging mode, where the object file in question may be recompiled and 2757dynamically reloaded in the course of program execution, without the need 2758to restart the program itself. Currently, Fix-and-Continue functionality 2759is only available in conjunction with the NeXT runtime on Mac OS X 10.3 2760and later. 2761 2762@item -fzero-link 2763@opindex fzero-link 2764When compiling for the NeXT runtime, the compiler ordinarily replaces calls 2765to @code{objc_getClass("@dots{}")} (when the name of the class is known at 2766compile time) with static class references that get initialized at load time, 2767which improves run-time performance. Specifying the @option{-fzero-link} flag 2768suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")} 2769to be retained. This is useful in Zero-Link debugging mode, since it allows 2770for individual class implementations to be modified during program execution. 2771The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")} 2772regardless of command-line options. 2773 2774@item -gen-decls 2775@opindex gen-decls 2776Dump interface declarations for all classes seen in the source file to a 2777file named @file{@var{sourcename}.decl}. 2778 2779@item -Wassign-intercept @r{(Objective-C and Objective-C++ only)} 2780@opindex Wassign-intercept 2781@opindex Wno-assign-intercept 2782Warn whenever an Objective-C assignment is being intercepted by the 2783garbage collector. 2784 2785@item -Wno-protocol @r{(Objective-C and Objective-C++ only)} 2786@opindex Wno-protocol 2787@opindex Wprotocol 2788If a class is declared to implement a protocol, a warning is issued for 2789every method in the protocol that is not implemented by the class. The 2790default behavior is to issue a warning for every method not explicitly 2791implemented in the class, even if a method implementation is inherited 2792from the superclass. If you use the @option{-Wno-protocol} option, then 2793methods inherited from the superclass are considered to be implemented, 2794and no warning is issued for them. 2795 2796@item -Wselector @r{(Objective-C and Objective-C++ only)} 2797@opindex Wselector 2798@opindex Wno-selector 2799Warn if multiple methods of different types for the same selector are 2800found during compilation. The check is performed on the list of methods 2801in the final stage of compilation. Additionally, a check is performed 2802for each selector appearing in a @code{@@selector(@dots{})} 2803expression, and a corresponding method for that selector has been found 2804during compilation. Because these checks scan the method table only at 2805the end of compilation, these warnings are not produced if the final 2806stage of compilation is not reached, for example because an error is 2807found during compilation, or because the @option{-fsyntax-only} option is 2808being used. 2809 2810@item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)} 2811@opindex Wstrict-selector-match 2812@opindex Wno-strict-selector-match 2813Warn if multiple methods with differing argument and/or return types are 2814found for a given selector when attempting to send a message using this 2815selector to a receiver of type @code{id} or @code{Class}. When this flag 2816is off (which is the default behavior), the compiler will omit such warnings 2817if any differences found are confined to types that share the same size 2818and alignment. 2819 2820@item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)} 2821@opindex Wundeclared-selector 2822@opindex Wno-undeclared-selector 2823Warn if a @code{@@selector(@dots{})} expression referring to an 2824undeclared selector is found. A selector is considered undeclared if no 2825method with that name has been declared before the 2826@code{@@selector(@dots{})} expression, either explicitly in an 2827@code{@@interface} or @code{@@protocol} declaration, or implicitly in 2828an @code{@@implementation} section. This option always performs its 2829checks as soon as a @code{@@selector(@dots{})} expression is found, 2830while @option{-Wselector} only performs its checks in the final stage of 2831compilation. This also enforces the coding style convention 2832that methods and selectors must be declared before being used. 2833 2834@item -print-objc-runtime-info 2835@opindex print-objc-runtime-info 2836Generate C header describing the largest structure that is passed by 2837value, if any. 2838 2839@end table 2840 2841@node Language Independent Options 2842@section Options to Control Diagnostic Messages Formatting 2843@cindex options to control diagnostics formatting 2844@cindex diagnostic messages 2845@cindex message formatting 2846 2847Traditionally, diagnostic messages have been formatted irrespective of 2848the output device's aspect (e.g.@: its width, @dots{}). The options described 2849below can be used to control the diagnostic messages formatting 2850algorithm, e.g.@: how many characters per line, how often source location 2851information should be reported. Right now, only the C++ front end can 2852honor these options. However it is expected, in the near future, that 2853the remaining front ends would be able to digest them correctly. 2854 2855@table @gcctabopt 2856@item -fmessage-length=@var{n} 2857@opindex fmessage-length 2858Try to format error messages so that they fit on lines of about @var{n} 2859characters. The default is 72 characters for @command{g++} and 0 for the rest of 2860the front ends supported by GCC@. If @var{n} is zero, then no 2861line-wrapping will be done; each error message will appear on a single 2862line. 2863 2864@opindex fdiagnostics-show-location 2865@item -fdiagnostics-show-location=once 2866Only meaningful in line-wrapping mode. Instructs the diagnostic messages 2867reporter to emit @emph{once} source location information; that is, in 2868case the message is too long to fit on a single physical line and has to 2869be wrapped, the source location won't be emitted (as prefix) again, 2870over and over, in subsequent continuation lines. This is the default 2871behavior. 2872 2873@item -fdiagnostics-show-location=every-line 2874Only meaningful in line-wrapping mode. Instructs the diagnostic 2875messages reporter to emit the same source location information (as 2876prefix) for physical lines that result from the process of breaking 2877a message which is too long to fit on a single line. 2878 2879@item -fno-diagnostics-show-option 2880@opindex fno-diagnostics-show-option 2881@opindex fdiagnostics-show-option 2882By default, each diagnostic emitted includes text indicating the 2883command-line option that directly controls the diagnostic (if such an 2884option is known to the diagnostic machinery). Specifying the 2885@option{-fno-diagnostics-show-option} flag suppresses that behavior. 2886 2887@end table 2888 2889@node Warning Options 2890@section Options to Request or Suppress Warnings 2891@cindex options to control warnings 2892@cindex warning messages 2893@cindex messages, warning 2894@cindex suppressing warnings 2895 2896Warnings are diagnostic messages that report constructions that 2897are not inherently erroneous but that are risky or suggest there 2898may have been an error. 2899 2900The following language-independent options do not enable specific 2901warnings but control the kinds of diagnostics produced by GCC. 2902 2903@table @gcctabopt 2904@cindex syntax checking 2905@item -fsyntax-only 2906@opindex fsyntax-only 2907Check the code for syntax errors, but don't do anything beyond that. 2908 2909@item -fmax-errors=@var{n} 2910@opindex fmax-errors 2911Limits the maximum number of error messages to @var{n}, at which point 2912GCC bails out rather than attempting to continue processing the source 2913code. If @var{n} is 0 (the default), there is no limit on the number 2914of error messages produced. If @option{-Wfatal-errors} is also 2915specified, then @option{-Wfatal-errors} takes precedence over this 2916option. 2917 2918@item -w 2919@opindex w 2920Inhibit all warning messages. 2921 2922@item -Werror 2923@opindex Werror 2924@opindex Wno-error 2925Make all warnings into errors. 2926 2927@item -Werror= 2928@opindex Werror= 2929@opindex Wno-error= 2930Make the specified warning into an error. The specifier for a warning 2931is appended, for example @option{-Werror=switch} turns the warnings 2932controlled by @option{-Wswitch} into errors. This switch takes a 2933negative form, to be used to negate @option{-Werror} for specific 2934warnings, for example @option{-Wno-error=switch} makes 2935@option{-Wswitch} warnings not be errors, even when @option{-Werror} 2936is in effect. 2937 2938The warning message for each controllable warning includes the 2939option that controls the warning. That option can then be used with 2940@option{-Werror=} and @option{-Wno-error=} as described above. 2941(Printing of the option in the warning message can be disabled using the 2942@option{-fno-diagnostics-show-option} flag.) 2943 2944Note that specifying @option{-Werror=}@var{foo} automatically implies 2945@option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not 2946imply anything. 2947 2948@item -Wfatal-errors 2949@opindex Wfatal-errors 2950@opindex Wno-fatal-errors 2951This option causes the compiler to abort compilation on the first error 2952occurred rather than trying to keep going and printing further error 2953messages. 2954 2955@end table 2956 2957You can request many specific warnings with options beginning 2958@samp{-W}, for example @option{-Wimplicit} to request warnings on 2959implicit declarations. Each of these specific warning options also 2960has a negative form beginning @samp{-Wno-} to turn off warnings; for 2961example, @option{-Wno-implicit}. This manual lists only one of the 2962two forms, whichever is not the default. For further, 2963language-specific options also refer to @ref{C++ Dialect Options} and 2964@ref{Objective-C and Objective-C++ Dialect Options}. 2965 2966When an unrecognized warning option is requested (e.g., 2967@option{-Wunknown-warning}), GCC will emit a diagnostic stating 2968that the option is not recognized. However, if the @option{-Wno-} form 2969is used, the behavior is slightly different: No diagnostic will be 2970produced for @option{-Wno-unknown-warning} unless other diagnostics 2971are being produced. This allows the use of new @option{-Wno-} options 2972with old compilers, but if something goes wrong, the compiler will 2973warn that an unrecognized option was used. 2974 2975@table @gcctabopt 2976@item -pedantic 2977@opindex pedantic 2978Issue all the warnings demanded by strict ISO C and ISO C++; 2979reject all programs that use forbidden extensions, and some other 2980programs that do not follow ISO C and ISO C++. For ISO C, follows the 2981version of the ISO C standard specified by any @option{-std} option used. 2982 2983Valid ISO C and ISO C++ programs should compile properly with or without 2984this option (though a rare few will require @option{-ansi} or a 2985@option{-std} option specifying the required version of ISO C)@. However, 2986without this option, certain GNU extensions and traditional C and C++ 2987features are supported as well. With this option, they are rejected. 2988 2989@option{-pedantic} does not cause warning messages for use of the 2990alternate keywords whose names begin and end with @samp{__}. Pedantic 2991warnings are also disabled in the expression that follows 2992@code{__extension__}. However, only system header files should use 2993these escape routes; application programs should avoid them. 2994@xref{Alternate Keywords}. 2995 2996Some users try to use @option{-pedantic} to check programs for strict ISO 2997C conformance. They soon find that it does not do quite what they want: 2998it finds some non-ISO practices, but not all---only those for which 2999ISO C @emph{requires} a diagnostic, and some others for which 3000diagnostics have been added. 3001 3002A feature to report any failure to conform to ISO C might be useful in 3003some instances, but would require considerable additional work and would 3004be quite different from @option{-pedantic}. We don't have plans to 3005support such a feature in the near future. 3006 3007Where the standard specified with @option{-std} represents a GNU 3008extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a 3009corresponding @dfn{base standard}, the version of ISO C on which the GNU 3010extended dialect is based. Warnings from @option{-pedantic} are given 3011where they are required by the base standard. (It would not make sense 3012for such warnings to be given only for features not in the specified GNU 3013C dialect, since by definition the GNU dialects of C include all 3014features the compiler supports with the given option, and there would be 3015nothing to warn about.) 3016 3017@item -pedantic-errors 3018@opindex pedantic-errors 3019Like @option{-pedantic}, except that errors are produced rather than 3020warnings. 3021 3022@item -Wall 3023@opindex Wall 3024@opindex Wno-all 3025This enables all the warnings about constructions that some users 3026consider questionable, and that are easy to avoid (or modify to 3027prevent the warning), even in conjunction with macros. This also 3028enables some language-specific warnings described in @ref{C++ Dialect 3029Options} and @ref{Objective-C and Objective-C++ Dialect Options}. 3030 3031@option{-Wall} turns on the following warning flags: 3032 3033@gccoptlist{-Waddress @gol 3034-Warray-bounds @r{(only with} @option{-O2}@r{)} @gol 3035-Wc++11-compat @gol 3036-Wchar-subscripts @gol 3037-Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol 3038-Wimplicit-int @r{(C and Objective-C only)} @gol 3039-Wimplicit-function-declaration @r{(C and Objective-C only)} @gol 3040-Wcomment @gol 3041-Wformat @gol 3042-Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol 3043-Wmaybe-uninitialized @gol 3044-Wmissing-braces @gol 3045-Wnonnull @gol 3046-Wparentheses @gol 3047-Wpointer-sign @gol 3048-Wreorder @gol 3049-Wreturn-type @gol 3050-Wsequence-point @gol 3051-Wsign-compare @r{(only in C++)} @gol 3052-Wstrict-aliasing @gol 3053-Wstrict-overflow=1 @gol 3054-Wswitch @gol 3055-Wtrigraphs @gol 3056-Wuninitialized @gol 3057-Wunknown-pragmas @gol 3058-Wunused-function @gol 3059-Wunused-label @gol 3060-Wunused-value @gol 3061-Wunused-variable @gol 3062-Wvolatile-register-var @gol 3063} 3064 3065Note that some warning flags are not implied by @option{-Wall}. Some of 3066them warn about constructions that users generally do not consider 3067questionable, but which occasionally you might wish to check for; 3068others warn about constructions that are necessary or hard to avoid in 3069some cases, and there is no simple way to modify the code to suppress 3070the warning. Some of them are enabled by @option{-Wextra} but many of 3071them must be enabled individually. 3072 3073@item -Wextra 3074@opindex W 3075@opindex Wextra 3076@opindex Wno-extra 3077This enables some extra warning flags that are not enabled by 3078@option{-Wall}. (This option used to be called @option{-W}. The older 3079name is still supported, but the newer name is more descriptive.) 3080 3081@gccoptlist{-Wclobbered @gol 3082-Wempty-body @gol 3083-Wignored-qualifiers @gol 3084-Wmissing-field-initializers @gol 3085-Wmissing-parameter-type @r{(C only)} @gol 3086-Wold-style-declaration @r{(C only)} @gol 3087-Woverride-init @gol 3088-Wsign-compare @gol 3089-Wtype-limits @gol 3090-Wuninitialized @gol 3091-Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol 3092-Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol 3093} 3094 3095The option @option{-Wextra} also prints warning messages for the 3096following cases: 3097 3098@itemize @bullet 3099 3100@item 3101A pointer is compared against integer zero with @samp{<}, @samp{<=}, 3102@samp{>}, or @samp{>=}. 3103 3104@item 3105(C++ only) An enumerator and a non-enumerator both appear in a 3106conditional expression. 3107 3108@item 3109(C++ only) Ambiguous virtual bases. 3110 3111@item 3112(C++ only) Subscripting an array that has been declared @samp{register}. 3113 3114@item 3115(C++ only) Taking the address of a variable that has been declared 3116@samp{register}. 3117 3118@item 3119(C++ only) A base class is not initialized in a derived class' copy 3120constructor. 3121 3122@end itemize 3123 3124@item -Wchar-subscripts 3125@opindex Wchar-subscripts 3126@opindex Wno-char-subscripts 3127Warn if an array subscript has type @code{char}. This is a common cause 3128of error, as programmers often forget that this type is signed on some 3129machines. 3130This warning is enabled by @option{-Wall}. 3131 3132@item -Wcomment 3133@opindex Wcomment 3134@opindex Wno-comment 3135Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*} 3136comment, or whenever a Backslash-Newline appears in a @samp{//} comment. 3137This warning is enabled by @option{-Wall}. 3138 3139@item -Wno-coverage-mismatch 3140@opindex Wno-coverage-mismatch 3141Warn if feedback profiles do not match when using the 3142@option{-fprofile-use} option. 3143If a source file was changed between @option{-fprofile-gen} and 3144@option{-fprofile-use}, the files with the profile feedback can fail 3145to match the source file and GCC cannot use the profile feedback 3146information. By default, this warning is enabled and is treated as an 3147error. @option{-Wno-coverage-mismatch} can be used to disable the 3148warning or @option{-Wno-error=coverage-mismatch} can be used to 3149disable the error. Disabling the error for this warning can result in 3150poorly optimized code and is useful only in the 3151case of very minor changes such as bug fixes to an existing code-base. 3152Completely disabling the warning is not recommended. 3153 3154@item -Wno-cpp 3155@r{(C, Objective-C, C++, Objective-C++ and Fortran only)} 3156 3157Suppress warning messages emitted by @code{#warning} directives. 3158 3159@item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)} 3160@opindex Wdouble-promotion 3161@opindex Wno-double-promotion 3162Give a warning when a value of type @code{float} is implicitly 3163promoted to @code{double}. CPUs with a 32-bit ``single-precision'' 3164floating-point unit implement @code{float} in hardware, but emulate 3165@code{double} in software. On such a machine, doing computations 3166using @code{double} values is much more expensive because of the 3167overhead required for software emulation. 3168 3169It is easy to accidentally do computations with @code{double} because 3170floating-point literals are implicitly of type @code{double}. For 3171example, in: 3172@smallexample 3173@group 3174float area(float radius) 3175@{ 3176 return 3.14159 * radius * radius; 3177@} 3178@end group 3179@end smallexample 3180the compiler will perform the entire computation with @code{double} 3181because the floating-point literal is a @code{double}. 3182 3183@item -Wformat 3184@opindex Wformat 3185@opindex Wno-format 3186@opindex ffreestanding 3187@opindex fno-builtin 3188Check calls to @code{printf} and @code{scanf}, etc., to make sure that 3189the arguments supplied have types appropriate to the format string 3190specified, and that the conversions specified in the format string make 3191sense. This includes standard functions, and others specified by format 3192attributes (@pxref{Function Attributes}), in the @code{printf}, 3193@code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension, 3194not in the C standard) families (or other target-specific families). 3195Which functions are checked without format attributes having been 3196specified depends on the standard version selected, and such checks of 3197functions without the attribute specified are disabled by 3198@option{-ffreestanding} or @option{-fno-builtin}. 3199 3200The formats are checked against the format features supported by GNU 3201libc version 2.2. These include all ISO C90 and C99 features, as well 3202as features from the Single Unix Specification and some BSD and GNU 3203extensions. Other library implementations may not support all these 3204features; GCC does not support warning about features that go beyond a 3205particular library's limitations. However, if @option{-pedantic} is used 3206with @option{-Wformat}, warnings will be given about format features not 3207in the selected standard version (but not for @code{strfmon} formats, 3208since those are not in any version of the C standard). @xref{C Dialect 3209Options,,Options Controlling C Dialect}. 3210 3211Since @option{-Wformat} also checks for null format arguments for 3212several functions, @option{-Wformat} also implies @option{-Wnonnull}. 3213 3214@option{-Wformat} is included in @option{-Wall}. For more control over some 3215aspects of format checking, the options @option{-Wformat-y2k}, 3216@option{-Wno-format-extra-args}, @option{-Wno-format-zero-length}, 3217@option{-Wformat-nonliteral}, @option{-Wformat-security}, and 3218@option{-Wformat=2} are available, but are not included in @option{-Wall}. 3219 3220@item -Wformat-y2k 3221@opindex Wformat-y2k 3222@opindex Wno-format-y2k 3223If @option{-Wformat} is specified, also warn about @code{strftime} 3224formats that may yield only a two-digit year. 3225 3226@item -Wno-format-contains-nul 3227@opindex Wno-format-contains-nul 3228@opindex Wformat-contains-nul 3229If @option{-Wformat} is specified, do not warn about format strings that 3230contain NUL bytes. 3231 3232@item -Wno-format-extra-args 3233@opindex Wno-format-extra-args 3234@opindex Wformat-extra-args 3235If @option{-Wformat} is specified, do not warn about excess arguments to a 3236@code{printf} or @code{scanf} format function. The C standard specifies 3237that such arguments are ignored. 3238 3239Where the unused arguments lie between used arguments that are 3240specified with @samp{$} operand number specifications, normally 3241warnings are still given, since the implementation could not know what 3242type to pass to @code{va_arg} to skip the unused arguments. However, 3243in the case of @code{scanf} formats, this option will suppress the 3244warning if the unused arguments are all pointers, since the Single 3245Unix Specification says that such unused arguments are allowed. 3246 3247@item -Wno-format-zero-length 3248@opindex Wno-format-zero-length 3249@opindex Wformat-zero-length 3250If @option{-Wformat} is specified, do not warn about zero-length formats. 3251The C standard specifies that zero-length formats are allowed. 3252 3253@item -Wformat-nonliteral 3254@opindex Wformat-nonliteral 3255@opindex Wno-format-nonliteral 3256If @option{-Wformat} is specified, also warn if the format string is not a 3257string literal and so cannot be checked, unless the format function 3258takes its format arguments as a @code{va_list}. 3259 3260@item -Wformat-security 3261@opindex Wformat-security 3262@opindex Wno-format-security 3263If @option{-Wformat} is specified, also warn about uses of format 3264functions that represent possible security problems. At present, this 3265warns about calls to @code{printf} and @code{scanf} functions where the 3266format string is not a string literal and there are no format arguments, 3267as in @code{printf (foo);}. This may be a security hole if the format 3268string came from untrusted input and contains @samp{%n}. (This is 3269currently a subset of what @option{-Wformat-nonliteral} warns about, but 3270in future warnings may be added to @option{-Wformat-security} that are not 3271included in @option{-Wformat-nonliteral}.) 3272 3273@item -Wformat=2 3274@opindex Wformat=2 3275@opindex Wno-format=2 3276Enable @option{-Wformat} plus format checks not included in 3277@option{-Wformat}. Currently equivalent to @samp{-Wformat 3278-Wformat-nonliteral -Wformat-security -Wformat-y2k}. 3279 3280@item -Wnonnull 3281@opindex Wnonnull 3282@opindex Wno-nonnull 3283Warn about passing a null pointer for arguments marked as 3284requiring a non-null value by the @code{nonnull} function attribute. 3285 3286@option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It 3287can be disabled with the @option{-Wno-nonnull} option. 3288 3289@item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)} 3290@opindex Winit-self 3291@opindex Wno-init-self 3292Warn about uninitialized variables that are initialized with themselves. 3293Note this option can only be used with the @option{-Wuninitialized} option. 3294 3295For example, GCC will warn about @code{i} being uninitialized in the 3296following snippet only when @option{-Winit-self} has been specified: 3297@smallexample 3298@group 3299int f() 3300@{ 3301 int i = i; 3302 return i; 3303@} 3304@end group 3305@end smallexample 3306 3307@item -Wimplicit-int @r{(C and Objective-C only)} 3308@opindex Wimplicit-int 3309@opindex Wno-implicit-int 3310Warn when a declaration does not specify a type. 3311This warning is enabled by @option{-Wall}. 3312 3313@item -Wimplicit-function-declaration @r{(C and Objective-C only)} 3314@opindex Wimplicit-function-declaration 3315@opindex Wno-implicit-function-declaration 3316Give a warning whenever a function is used before being declared. In 3317C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is 3318enabled by default and it is made into an error by 3319@option{-pedantic-errors}. This warning is also enabled by 3320@option{-Wall}. 3321 3322@item -Wimplicit @r{(C and Objective-C only)} 3323@opindex Wimplicit 3324@opindex Wno-implicit 3325Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}. 3326This warning is enabled by @option{-Wall}. 3327 3328@item -Wignored-qualifiers @r{(C and C++ only)} 3329@opindex Wignored-qualifiers 3330@opindex Wno-ignored-qualifiers 3331Warn if the return type of a function has a type qualifier 3332such as @code{const}. For ISO C such a type qualifier has no effect, 3333since the value returned by a function is not an lvalue. 3334For C++, the warning is only emitted for scalar types or @code{void}. 3335ISO C prohibits qualified @code{void} return types on function 3336definitions, so such return types always receive a warning 3337even without this option. 3338 3339This warning is also enabled by @option{-Wextra}. 3340 3341@item -Wmain 3342@opindex Wmain 3343@opindex Wno-main 3344Warn if the type of @samp{main} is suspicious. @samp{main} should be 3345a function with external linkage, returning int, taking either zero 3346arguments, two, or three arguments of appropriate types. This warning 3347is enabled by default in C++ and is enabled by either @option{-Wall} 3348or @option{-pedantic}. 3349 3350@item -Wmissing-braces 3351@opindex Wmissing-braces 3352@opindex Wno-missing-braces 3353Warn if an aggregate or union initializer is not fully bracketed. In 3354the following example, the initializer for @samp{a} is not fully 3355bracketed, but that for @samp{b} is fully bracketed. 3356 3357@smallexample 3358int a[2][2] = @{ 0, 1, 2, 3 @}; 3359int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @}; 3360@end smallexample 3361 3362This warning is enabled by @option{-Wall}. 3363 3364@item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)} 3365@opindex Wmissing-include-dirs 3366@opindex Wno-missing-include-dirs 3367Warn if a user-supplied include directory does not exist. 3368 3369@item -Wparentheses 3370@opindex Wparentheses 3371@opindex Wno-parentheses 3372Warn if parentheses are omitted in certain contexts, such 3373as when there is an assignment in a context where a truth value 3374is expected, or when operators are nested whose precedence people 3375often get confused about. 3376 3377Also warn if a comparison like @samp{x<=y<=z} appears; this is 3378equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different 3379interpretation from that of ordinary mathematical notation. 3380 3381Also warn about constructions where there may be confusion to which 3382@code{if} statement an @code{else} branch belongs. Here is an example of 3383such a case: 3384 3385@smallexample 3386@group 3387@{ 3388 if (a) 3389 if (b) 3390 foo (); 3391 else 3392 bar (); 3393@} 3394@end group 3395@end smallexample 3396 3397In C/C++, every @code{else} branch belongs to the innermost possible 3398@code{if} statement, which in this example is @code{if (b)}. This is 3399often not what the programmer expected, as illustrated in the above 3400example by indentation the programmer chose. When there is the 3401potential for this confusion, GCC will issue a warning when this flag 3402is specified. To eliminate the warning, add explicit braces around 3403the innermost @code{if} statement so there is no way the @code{else} 3404could belong to the enclosing @code{if}. The resulting code would 3405look like this: 3406 3407@smallexample 3408@group 3409@{ 3410 if (a) 3411 @{ 3412 if (b) 3413 foo (); 3414 else 3415 bar (); 3416 @} 3417@} 3418@end group 3419@end smallexample 3420 3421Also warn for dangerous uses of the 3422?: with omitted middle operand GNU extension. When the condition 3423in the ?: operator is a boolean expression the omitted value will 3424be always 1. Often the user expects it to be a value computed 3425inside the conditional expression instead. 3426 3427This warning is enabled by @option{-Wall}. 3428 3429@item -Wsequence-point 3430@opindex Wsequence-point 3431@opindex Wno-sequence-point 3432Warn about code that may have undefined semantics because of violations 3433of sequence point rules in the C and C++ standards. 3434 3435The C and C++ standards defines the order in which expressions in a C/C++ 3436program are evaluated in terms of @dfn{sequence points}, which represent 3437a partial ordering between the execution of parts of the program: those 3438executed before the sequence point, and those executed after it. These 3439occur after the evaluation of a full expression (one which is not part 3440of a larger expression), after the evaluation of the first operand of a 3441@code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a 3442function is called (but after the evaluation of its arguments and the 3443expression denoting the called function), and in certain other places. 3444Other than as expressed by the sequence point rules, the order of 3445evaluation of subexpressions of an expression is not specified. All 3446these rules describe only a partial order rather than a total order, 3447since, for example, if two functions are called within one expression 3448with no sequence point between them, the order in which the functions 3449are called is not specified. However, the standards committee have 3450ruled that function calls do not overlap. 3451 3452It is not specified when between sequence points modifications to the 3453values of objects take effect. Programs whose behavior depends on this 3454have undefined behavior; the C and C++ standards specify that ``Between 3455the previous and next sequence point an object shall have its stored 3456value modified at most once by the evaluation of an expression. 3457Furthermore, the prior value shall be read only to determine the value 3458to be stored.''. If a program breaks these rules, the results on any 3459particular implementation are entirely unpredictable. 3460 3461Examples of code with undefined behavior are @code{a = a++;}, @code{a[n] 3462= b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not 3463diagnosed by this option, and it may give an occasional false positive 3464result, but in general it has been found fairly effective at detecting 3465this sort of problem in programs. 3466 3467The standard is worded confusingly, therefore there is some debate 3468over the precise meaning of the sequence point rules in subtle cases. 3469Links to discussions of the problem, including proposed formal 3470definitions, may be found on the GCC readings page, at 3471@uref{http://gcc.gnu.org/@/readings.html}. 3472 3473This warning is enabled by @option{-Wall} for C and C++. 3474 3475@item -Wreturn-type 3476@opindex Wreturn-type 3477@opindex Wno-return-type 3478Warn whenever a function is defined with a return-type that defaults 3479to @code{int}. Also warn about any @code{return} statement with no 3480return-value in a function whose return-type is not @code{void} 3481(falling off the end of the function body is considered returning 3482without a value), and about a @code{return} statement with an 3483expression in a function whose return-type is @code{void}. 3484 3485For C++, a function without return type always produces a diagnostic 3486message, even when @option{-Wno-return-type} is specified. The only 3487exceptions are @samp{main} and functions defined in system headers. 3488 3489This warning is enabled by @option{-Wall}. 3490 3491@item -Wswitch 3492@opindex Wswitch 3493@opindex Wno-switch 3494Warn whenever a @code{switch} statement has an index of enumerated type 3495and lacks a @code{case} for one or more of the named codes of that 3496enumeration. (The presence of a @code{default} label prevents this 3497warning.) @code{case} labels outside the enumeration range also 3498provoke warnings when this option is used (even if there is a 3499@code{default} label). 3500This warning is enabled by @option{-Wall}. 3501 3502@item -Wswitch-default 3503@opindex Wswitch-default 3504@opindex Wno-switch-default 3505Warn whenever a @code{switch} statement does not have a @code{default} 3506case. 3507 3508@item -Wswitch-enum 3509@opindex Wswitch-enum 3510@opindex Wno-switch-enum 3511Warn whenever a @code{switch} statement has an index of enumerated type 3512and lacks a @code{case} for one or more of the named codes of that 3513enumeration. @code{case} labels outside the enumeration range also 3514provoke warnings when this option is used. The only difference 3515between @option{-Wswitch} and this option is that this option gives a 3516warning about an omitted enumeration code even if there is a 3517@code{default} label. 3518 3519@item -Wsync-nand @r{(C and C++ only)} 3520@opindex Wsync-nand 3521@opindex Wno-sync-nand 3522Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch} 3523built-in functions are used. These functions changed semantics in GCC 4.4. 3524 3525@item -Wtrigraphs 3526@opindex Wtrigraphs 3527@opindex Wno-trigraphs 3528Warn if any trigraphs are encountered that might change the meaning of 3529the program (trigraphs within comments are not warned about). 3530This warning is enabled by @option{-Wall}. 3531 3532@item -Wunused-but-set-parameter 3533@opindex Wunused-but-set-parameter 3534@opindex Wno-unused-but-set-parameter 3535Warn whenever a function parameter is assigned to, but otherwise unused 3536(aside from its declaration). 3537 3538To suppress this warning use the @samp{unused} attribute 3539(@pxref{Variable Attributes}). 3540 3541This warning is also enabled by @option{-Wunused} together with 3542@option{-Wextra}. 3543 3544@item -Wunused-but-set-variable 3545@opindex Wunused-but-set-variable 3546@opindex Wno-unused-but-set-variable 3547Warn whenever a local variable is assigned to, but otherwise unused 3548(aside from its declaration). 3549This warning is enabled by @option{-Wall}. 3550 3551To suppress this warning use the @samp{unused} attribute 3552(@pxref{Variable Attributes}). 3553 3554This warning is also enabled by @option{-Wunused}, which is enabled 3555by @option{-Wall}. 3556 3557@item -Wunused-function 3558@opindex Wunused-function 3559@opindex Wno-unused-function 3560Warn whenever a static function is declared but not defined or a 3561non-inline static function is unused. 3562This warning is enabled by @option{-Wall}. 3563 3564@item -Wunused-label 3565@opindex Wunused-label 3566@opindex Wno-unused-label 3567Warn whenever a label is declared but not used. 3568This warning is enabled by @option{-Wall}. 3569 3570To suppress this warning use the @samp{unused} attribute 3571(@pxref{Variable Attributes}). 3572 3573@item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)} 3574@opindex Wunused-local-typedefs 3575Warn when a typedef locally defined in a function is not used. 3576 3577@item -Wunused-parameter 3578@opindex Wunused-parameter 3579@opindex Wno-unused-parameter 3580Warn whenever a function parameter is unused aside from its declaration. 3581 3582To suppress this warning use the @samp{unused} attribute 3583(@pxref{Variable Attributes}). 3584 3585@item -Wno-unused-result 3586@opindex Wunused-result 3587@opindex Wno-unused-result 3588Do not warn if a caller of a function marked with attribute 3589@code{warn_unused_result} (@pxref{Function Attributes}) does not use 3590its return value. The default is @option{-Wunused-result}. 3591 3592@item -Wunused-variable 3593@opindex Wunused-variable 3594@opindex Wno-unused-variable 3595Warn whenever a local variable or non-constant static variable is unused 3596aside from its declaration. 3597This warning is enabled by @option{-Wall}. 3598 3599To suppress this warning use the @samp{unused} attribute 3600(@pxref{Variable Attributes}). 3601 3602@item -Wunused-value 3603@opindex Wunused-value 3604@opindex Wno-unused-value 3605Warn whenever a statement computes a result that is explicitly not 3606used. To suppress this warning cast the unused expression to 3607@samp{void}. This includes an expression-statement or the left-hand 3608side of a comma expression that contains no side effects. For example, 3609an expression such as @samp{x[i,j]} will cause a warning, while 3610@samp{x[(void)i,j]} will not. 3611 3612This warning is enabled by @option{-Wall}. 3613 3614@item -Wunused 3615@opindex Wunused 3616@opindex Wno-unused 3617All the above @option{-Wunused} options combined. 3618 3619In order to get a warning about an unused function parameter, you must 3620either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies 3621@samp{-Wunused}), or separately specify @option{-Wunused-parameter}. 3622 3623@item -Wuninitialized 3624@opindex Wuninitialized 3625@opindex Wno-uninitialized 3626Warn if an automatic variable is used without first being initialized 3627or if a variable may be clobbered by a @code{setjmp} call. In C++, 3628warn if a non-static reference or non-static @samp{const} member 3629appears in a class without constructors. 3630 3631If you want to warn about code that uses the uninitialized value of the 3632variable in its own initializer, use the @option{-Winit-self} option. 3633 3634These warnings occur for individual uninitialized or clobbered 3635elements of structure, union or array variables as well as for 3636variables that are uninitialized or clobbered as a whole. They do 3637not occur for variables or elements declared @code{volatile}. Because 3638these warnings depend on optimization, the exact variables or elements 3639for which there are warnings will depend on the precise optimization 3640options and version of GCC used. 3641 3642Note that there may be no warning about a variable that is used only 3643to compute a value that itself is never used, because such 3644computations may be deleted by data flow analysis before the warnings 3645are printed. 3646 3647@item -Wmaybe-uninitialized 3648@opindex Wmaybe-uninitialized 3649@opindex Wno-maybe-uninitialized 3650For an automatic variable, if there exists a path from the function 3651entry to a use of the variable that is initialized, but there exist 3652some other paths the variable is not initialized, the compiler will 3653emit a warning if it can not prove the uninitialized paths do not 3654happen at run time. These warnings are made optional because GCC is 3655not smart enough to see all the reasons why the code might be correct 3656despite appearing to have an error. Here is one example of how 3657this can happen: 3658 3659@smallexample 3660@group 3661@{ 3662 int x; 3663 switch (y) 3664 @{ 3665 case 1: x = 1; 3666 break; 3667 case 2: x = 4; 3668 break; 3669 case 3: x = 5; 3670 @} 3671 foo (x); 3672@} 3673@end group 3674@end smallexample 3675 3676@noindent 3677If the value of @code{y} is always 1, 2 or 3, then @code{x} is 3678always initialized, but GCC doesn't know this. To suppress the 3679warning, the user needs to provide a default case with assert(0) or 3680similar code. 3681 3682@cindex @code{longjmp} warnings 3683This option also warns when a non-volatile automatic variable might be 3684changed by a call to @code{longjmp}. These warnings as well are possible 3685only in optimizing compilation. 3686 3687The compiler sees only the calls to @code{setjmp}. It cannot know 3688where @code{longjmp} will be called; in fact, a signal handler could 3689call it at any point in the code. As a result, you may get a warning 3690even when there is in fact no problem because @code{longjmp} cannot 3691in fact be called at the place that would cause a problem. 3692 3693Some spurious warnings can be avoided if you declare all the functions 3694you use that never return as @code{noreturn}. @xref{Function 3695Attributes}. 3696 3697This warning is enabled by @option{-Wall} or @option{-Wextra}. 3698 3699@item -Wunknown-pragmas 3700@opindex Wunknown-pragmas 3701@opindex Wno-unknown-pragmas 3702@cindex warning for unknown pragmas 3703@cindex unknown pragmas, warning 3704@cindex pragmas, warning of unknown 3705Warn when a @code{#pragma} directive is encountered that is not understood by 3706GCC@. If this command-line option is used, warnings will even be issued 3707for unknown pragmas in system header files. This is not the case if 3708the warnings were only enabled by the @option{-Wall} command-line option. 3709 3710@item -Wno-pragmas 3711@opindex Wno-pragmas 3712@opindex Wpragmas 3713Do not warn about misuses of pragmas, such as incorrect parameters, 3714invalid syntax, or conflicts between pragmas. See also 3715@samp{-Wunknown-pragmas}. 3716 3717@item -Wstrict-aliasing 3718@opindex Wstrict-aliasing 3719@opindex Wno-strict-aliasing 3720This option is only active when @option{-fstrict-aliasing} is active. 3721It warns about code that might break the strict aliasing rules that the 3722compiler is using for optimization. The warning does not catch all 3723cases, but does attempt to catch the more common pitfalls. It is 3724included in @option{-Wall}. 3725It is equivalent to @option{-Wstrict-aliasing=3} 3726 3727@item -Wstrict-aliasing=n 3728@opindex Wstrict-aliasing=n 3729@opindex Wno-strict-aliasing=n 3730This option is only active when @option{-fstrict-aliasing} is active. 3731It warns about code that might break the strict aliasing rules that the 3732compiler is using for optimization. 3733Higher levels correspond to higher accuracy (fewer false positives). 3734Higher levels also correspond to more effort, similar to the way -O works. 3735@option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n}, 3736with n=3. 3737 3738Level 1: Most aggressive, quick, least accurate. 3739Possibly useful when higher levels 3740do not warn but -fstrict-aliasing still breaks the code, as it has very few 3741false negatives. However, it has many false positives. 3742Warns for all pointer conversions between possibly incompatible types, 3743even if never dereferenced. Runs in the front end only. 3744 3745Level 2: Aggressive, quick, not too precise. 3746May still have many false positives (not as many as level 1 though), 3747and few false negatives (but possibly more than level 1). 3748Unlike level 1, it only warns when an address is taken. Warns about 3749incomplete types. Runs in the front end only. 3750 3751Level 3 (default for @option{-Wstrict-aliasing}): 3752Should have very few false positives and few false 3753negatives. Slightly slower than levels 1 or 2 when optimization is enabled. 3754Takes care of the common pun+dereference pattern in the front end: 3755@code{*(int*)&some_float}. 3756If optimization is enabled, it also runs in the back end, where it deals 3757with multiple statement cases using flow-sensitive points-to information. 3758Only warns when the converted pointer is dereferenced. 3759Does not warn about incomplete types. 3760 3761@item -Wstrict-overflow 3762@itemx -Wstrict-overflow=@var{n} 3763@opindex Wstrict-overflow 3764@opindex Wno-strict-overflow 3765This option is only active when @option{-fstrict-overflow} is active. 3766It warns about cases where the compiler optimizes based on the 3767assumption that signed overflow does not occur. Note that it does not 3768warn about all cases where the code might overflow: it only warns 3769about cases where the compiler implements some optimization. Thus 3770this warning depends on the optimization level. 3771 3772An optimization that assumes that signed overflow does not occur is 3773perfectly safe if the values of the variables involved are such that 3774overflow never does, in fact, occur. Therefore this warning can 3775easily give a false positive: a warning about code that is not 3776actually a problem. To help focus on important issues, several 3777warning levels are defined. No warnings are issued for the use of 3778undefined signed overflow when estimating how many iterations a loop 3779will require, in particular when determining whether a loop will be 3780executed at all. 3781 3782@table @gcctabopt 3783@item -Wstrict-overflow=1 3784Warn about cases that are both questionable and easy to avoid. For 3785example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the 3786compiler will simplify this to @code{1}. This level of 3787@option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels 3788are not, and must be explicitly requested. 3789 3790@item -Wstrict-overflow=2 3791Also warn about other cases where a comparison is simplified to a 3792constant. For example: @code{abs (x) >= 0}. This can only be 3793simplified when @option{-fstrict-overflow} is in effect, because 3794@code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than 3795zero. @option{-Wstrict-overflow} (with no level) is the same as 3796@option{-Wstrict-overflow=2}. 3797 3798@item -Wstrict-overflow=3 3799Also warn about other cases where a comparison is simplified. For 3800example: @code{x + 1 > 1} will be simplified to @code{x > 0}. 3801 3802@item -Wstrict-overflow=4 3803Also warn about other simplifications not covered by the above cases. 3804For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}. 3805 3806@item -Wstrict-overflow=5 3807Also warn about cases where the compiler reduces the magnitude of a 3808constant involved in a comparison. For example: @code{x + 2 > y} will 3809be simplified to @code{x + 1 >= y}. This is reported only at the 3810highest warning level because this simplification applies to many 3811comparisons, so this warning level will give a very large number of 3812false positives. 3813@end table 3814 3815@item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} 3816@opindex Wsuggest-attribute= 3817@opindex Wno-suggest-attribute= 3818Warn for cases where adding an attribute may be beneficial. The 3819attributes currently supported are listed below. 3820 3821@table @gcctabopt 3822@item -Wsuggest-attribute=pure 3823@itemx -Wsuggest-attribute=const 3824@itemx -Wsuggest-attribute=noreturn 3825@opindex Wsuggest-attribute=pure 3826@opindex Wno-suggest-attribute=pure 3827@opindex Wsuggest-attribute=const 3828@opindex Wno-suggest-attribute=const 3829@opindex Wsuggest-attribute=noreturn 3830@opindex Wno-suggest-attribute=noreturn 3831 3832Warn about functions that might be candidates for attributes 3833@code{pure}, @code{const} or @code{noreturn}. The compiler only warns for 3834functions visible in other compilation units or (in the case of @code{pure} and 3835@code{const}) if it cannot prove that the function returns normally. A function 3836returns normally if it doesn't contain an infinite loop nor returns abnormally 3837by throwing, calling @code{abort()} or trapping. This analysis requires option 3838@option{-fipa-pure-const}, which is enabled by default at @option{-O} and 3839higher. Higher optimization levels improve the accuracy of the analysis. 3840@end table 3841 3842@item -Warray-bounds 3843@opindex Wno-array-bounds 3844@opindex Warray-bounds 3845This option is only active when @option{-ftree-vrp} is active 3846(default for @option{-O2} and above). It warns about subscripts to arrays 3847that are always out of bounds. This warning is enabled by @option{-Wall}. 3848 3849@item -Wno-div-by-zero 3850@opindex Wno-div-by-zero 3851@opindex Wdiv-by-zero 3852Do not warn about compile-time integer division by zero. Floating-point 3853division by zero is not warned about, as it can be a legitimate way of 3854obtaining infinities and NaNs. 3855 3856@item -Wsystem-headers 3857@opindex Wsystem-headers 3858@opindex Wno-system-headers 3859@cindex warnings from system headers 3860@cindex system headers, warnings from 3861Print warning messages for constructs found in system header files. 3862Warnings from system headers are normally suppressed, on the assumption 3863that they usually do not indicate real problems and would only make the 3864compiler output harder to read. Using this command-line option tells 3865GCC to emit warnings from system headers as if they occurred in user 3866code. However, note that using @option{-Wall} in conjunction with this 3867option will @emph{not} warn about unknown pragmas in system 3868headers---for that, @option{-Wunknown-pragmas} must also be used. 3869 3870@item -Wtrampolines 3871@opindex Wtrampolines 3872@opindex Wno-trampolines 3873 Warn about trampolines generated for pointers to nested functions. 3874 3875 A trampoline is a small piece of data or code that is created at run 3876 time on the stack when the address of a nested function is taken, and 3877 is used to call the nested function indirectly. For some targets, it 3878 is made up of data only and thus requires no special treatment. But, 3879 for most targets, it is made up of code and thus requires the stack 3880 to be made executable in order for the program to work properly. 3881 3882@item -Wfloat-equal 3883@opindex Wfloat-equal 3884@opindex Wno-float-equal 3885Warn if floating-point values are used in equality comparisons. 3886 3887The idea behind this is that sometimes it is convenient (for the 3888programmer) to consider floating-point values as approximations to 3889infinitely precise real numbers. If you are doing this, then you need 3890to compute (by analyzing the code, or in some other way) the maximum or 3891likely maximum error that the computation introduces, and allow for it 3892when performing comparisons (and when producing output, but that's a 3893different problem). In particular, instead of testing for equality, you 3894would check to see whether the two values have ranges that overlap; and 3895this is done with the relational operators, so equality comparisons are 3896probably mistaken. 3897 3898@item -Wtraditional @r{(C and Objective-C only)} 3899@opindex Wtraditional 3900@opindex Wno-traditional 3901Warn about certain constructs that behave differently in traditional and 3902ISO C@. Also warn about ISO C constructs that have no traditional C 3903equivalent, and/or problematic constructs that should be avoided. 3904 3905@itemize @bullet 3906@item 3907Macro parameters that appear within string literals in the macro body. 3908In traditional C macro replacement takes place within string literals, 3909but does not in ISO C@. 3910 3911@item 3912In traditional C, some preprocessor directives did not exist. 3913Traditional preprocessors would only consider a line to be a directive 3914if the @samp{#} appeared in column 1 on the line. Therefore 3915@option{-Wtraditional} warns about directives that traditional C 3916understands but would ignore because the @samp{#} does not appear as the 3917first character on the line. It also suggests you hide directives like 3918@samp{#pragma} not understood by traditional C by indenting them. Some 3919traditional implementations would not recognize @samp{#elif}, so it 3920suggests avoiding it altogether. 3921 3922@item 3923A function-like macro that appears without arguments. 3924 3925@item 3926The unary plus operator. 3927 3928@item 3929The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point 3930constant suffixes. (Traditional C does support the @samp{L} suffix on integer 3931constants.) Note, these suffixes appear in macros defined in the system 3932headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}. 3933Use of these macros in user code might normally lead to spurious 3934warnings, however GCC's integrated preprocessor has enough context to 3935avoid warning in these cases. 3936 3937@item 3938A function declared external in one block and then used after the end of 3939the block. 3940 3941@item 3942A @code{switch} statement has an operand of type @code{long}. 3943 3944@item 3945A non-@code{static} function declaration follows a @code{static} one. 3946This construct is not accepted by some traditional C compilers. 3947 3948@item 3949The ISO type of an integer constant has a different width or 3950signedness from its traditional type. This warning is only issued if 3951the base of the constant is ten. I.e.@: hexadecimal or octal values, which 3952typically represent bit patterns, are not warned about. 3953 3954@item 3955Usage of ISO string concatenation is detected. 3956 3957@item 3958Initialization of automatic aggregates. 3959 3960@item 3961Identifier conflicts with labels. Traditional C lacks a separate 3962namespace for labels. 3963 3964@item 3965Initialization of unions. If the initializer is zero, the warning is 3966omitted. This is done under the assumption that the zero initializer in 3967user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing 3968initializer warnings and relies on default initialization to zero in the 3969traditional C case. 3970 3971@item 3972Conversions by prototypes between fixed/floating-point values and vice 3973versa. The absence of these prototypes when compiling with traditional 3974C would cause serious problems. This is a subset of the possible 3975conversion warnings, for the full set use @option{-Wtraditional-conversion}. 3976 3977@item 3978Use of ISO C style function definitions. This warning intentionally is 3979@emph{not} issued for prototype declarations or variadic functions 3980because these ISO C features will appear in your code when using 3981libiberty's traditional C compatibility macros, @code{PARAMS} and 3982@code{VPARAMS}. This warning is also bypassed for nested functions 3983because that feature is already a GCC extension and thus not relevant to 3984traditional C compatibility. 3985@end itemize 3986 3987@item -Wtraditional-conversion @r{(C and Objective-C only)} 3988@opindex Wtraditional-conversion 3989@opindex Wno-traditional-conversion 3990Warn if a prototype causes a type conversion that is different from what 3991would happen to the same argument in the absence of a prototype. This 3992includes conversions of fixed point to floating and vice versa, and 3993conversions changing the width or signedness of a fixed-point argument 3994except when the same as the default promotion. 3995 3996@item -Wdeclaration-after-statement @r{(C and Objective-C only)} 3997@opindex Wdeclaration-after-statement 3998@opindex Wno-declaration-after-statement 3999Warn when a declaration is found after a statement in a block. This 4000construct, known from C++, was introduced with ISO C99 and is by default 4001allowed in GCC@. It is not supported by ISO C90 and was not supported by 4002GCC versions before GCC 3.0. @xref{Mixed Declarations}. 4003 4004@item -Wundef 4005@opindex Wundef 4006@opindex Wno-undef 4007Warn if an undefined identifier is evaluated in an @samp{#if} directive. 4008 4009@item -Wno-endif-labels 4010@opindex Wno-endif-labels 4011@opindex Wendif-labels 4012Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text. 4013 4014@item -Wshadow 4015@opindex Wshadow 4016@opindex Wno-shadow 4017Warn whenever a local variable or type declaration shadows another variable, 4018parameter, type, or class member (in C++), or whenever a built-in function 4019is shadowed. Note that in C++, the compiler will not warn if a local variable 4020shadows a struct/class/enum, but will warn if it shadows an explicit typedef. 4021 4022@item -Wlarger-than=@var{len} 4023@opindex Wlarger-than=@var{len} 4024@opindex Wlarger-than-@var{len} 4025Warn whenever an object of larger than @var{len} bytes is defined. 4026 4027@item -Wframe-larger-than=@var{len} 4028@opindex Wframe-larger-than 4029Warn if the size of a function frame is larger than @var{len} bytes. 4030The computation done to determine the stack frame size is approximate 4031and not conservative. 4032The actual requirements may be somewhat greater than @var{len} 4033even if you do not get a warning. In addition, any space allocated 4034via @code{alloca}, variable-length arrays, or related constructs 4035is not included by the compiler when determining 4036whether or not to issue a warning. 4037 4038@item -Wno-free-nonheap-object 4039@opindex Wno-free-nonheap-object 4040@opindex Wfree-nonheap-object 4041Do not warn when attempting to free an object that was not allocated 4042on the heap. 4043 4044@item -Wstack-usage=@var{len} 4045@opindex Wstack-usage 4046Warn if the stack usage of a function might be larger than @var{len} bytes. 4047The computation done to determine the stack usage is conservative. 4048Any space allocated via @code{alloca}, variable-length arrays, or related 4049constructs is included by the compiler when determining whether or not to 4050issue a warning. 4051 4052The message is in keeping with the output of @option{-fstack-usage}. 4053 4054@itemize 4055@item 4056If the stack usage is fully static but exceeds the specified amount, it's: 4057 4058@smallexample 4059 warning: stack usage is 1120 bytes 4060@end smallexample 4061@item 4062If the stack usage is (partly) dynamic but bounded, it's: 4063 4064@smallexample 4065 warning: stack usage might be 1648 bytes 4066@end smallexample 4067@item 4068If the stack usage is (partly) dynamic and not bounded, it's: 4069 4070@smallexample 4071 warning: stack usage might be unbounded 4072@end smallexample 4073@end itemize 4074 4075@item -Wunsafe-loop-optimizations 4076@opindex Wunsafe-loop-optimizations 4077@opindex Wno-unsafe-loop-optimizations 4078Warn if the loop cannot be optimized because the compiler could not 4079assume anything on the bounds of the loop indices. With 4080@option{-funsafe-loop-optimizations} warn if the compiler made 4081such assumptions. 4082 4083@item -Wno-pedantic-ms-format @r{(MinGW targets only)} 4084@opindex Wno-pedantic-ms-format 4085@opindex Wpedantic-ms-format 4086Disables the warnings about non-ISO @code{printf} / @code{scanf} format 4087width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets 4088depending on the MS runtime, when you are using the options @option{-Wformat} 4089and @option{-pedantic} without gnu-extensions. 4090 4091@item -Wpointer-arith 4092@opindex Wpointer-arith 4093@opindex Wno-pointer-arith 4094Warn about anything that depends on the ``size of'' a function type or 4095of @code{void}. GNU C assigns these types a size of 1, for 4096convenience in calculations with @code{void *} pointers and pointers 4097to functions. In C++, warn also when an arithmetic operation involves 4098@code{NULL}. This warning is also enabled by @option{-pedantic}. 4099 4100@item -Wtype-limits 4101@opindex Wtype-limits 4102@opindex Wno-type-limits 4103Warn if a comparison is always true or always false due to the limited 4104range of the data type, but do not warn for constant expressions. For 4105example, warn if an unsigned variable is compared against zero with 4106@samp{<} or @samp{>=}. This warning is also enabled by 4107@option{-Wextra}. 4108 4109@item -Wbad-function-cast @r{(C and Objective-C only)} 4110@opindex Wbad-function-cast 4111@opindex Wno-bad-function-cast 4112Warn whenever a function call is cast to a non-matching type. 4113For example, warn if @code{int malloc()} is cast to @code{anything *}. 4114 4115@item -Wc++-compat @r{(C and Objective-C only)} 4116Warn about ISO C constructs that are outside of the common subset of 4117ISO C and ISO C++, e.g.@: request for implicit conversion from 4118@code{void *} to a pointer to non-@code{void} type. 4119 4120@item -Wc++11-compat @r{(C++ and Objective-C++ only)} 4121Warn about C++ constructs whose meaning differs between ISO C++ 1998 4122and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords 4123in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is 4124enabled by @option{-Wall}. 4125 4126@item -Wcast-qual 4127@opindex Wcast-qual 4128@opindex Wno-cast-qual 4129Warn whenever a pointer is cast so as to remove a type qualifier from 4130the target type. For example, warn if a @code{const char *} is cast 4131to an ordinary @code{char *}. 4132 4133Also warn when making a cast that introduces a type qualifier in an 4134unsafe way. For example, casting @code{char **} to @code{const char **} 4135is unsafe, as in this example: 4136 4137@smallexample 4138 /* p is char ** value. */ 4139 const char **q = (const char **) p; 4140 /* Assignment of readonly string to const char * is OK. */ 4141 *q = "string"; 4142 /* Now char** pointer points to read-only memory. */ 4143 **p = 'b'; 4144@end smallexample 4145 4146@item -Wcast-align 4147@opindex Wcast-align 4148@opindex Wno-cast-align 4149Warn whenever a pointer is cast such that the required alignment of the 4150target is increased. For example, warn if a @code{char *} is cast to 4151an @code{int *} on machines where integers can only be accessed at 4152two- or four-byte boundaries. 4153 4154@item -Wwrite-strings 4155@opindex Wwrite-strings 4156@opindex Wno-write-strings 4157When compiling C, give string constants the type @code{const 4158char[@var{length}]} so that copying the address of one into a 4159non-@code{const} @code{char *} pointer will get a warning. These 4160warnings will help you find at compile time code that can try to write 4161into a string constant, but only if you have been very careful about 4162using @code{const} in declarations and prototypes. Otherwise, it will 4163just be a nuisance. This is why we did not make @option{-Wall} request 4164these warnings. 4165 4166When compiling C++, warn about the deprecated conversion from string 4167literals to @code{char *}. This warning is enabled by default for C++ 4168programs. 4169 4170@item -Wclobbered 4171@opindex Wclobbered 4172@opindex Wno-clobbered 4173Warn for variables that might be changed by @samp{longjmp} or 4174@samp{vfork}. This warning is also enabled by @option{-Wextra}. 4175 4176@item -Wconversion 4177@opindex Wconversion 4178@opindex Wno-conversion 4179Warn for implicit conversions that may alter a value. This includes 4180conversions between real and integer, like @code{abs (x)} when 4181@code{x} is @code{double}; conversions between signed and unsigned, 4182like @code{unsigned ui = -1}; and conversions to smaller types, like 4183@code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs 4184((int) x)} and @code{ui = (unsigned) -1}, or if the value is not 4185changed by the conversion like in @code{abs (2.0)}. Warnings about 4186conversions between signed and unsigned integers can be disabled by 4187using @option{-Wno-sign-conversion}. 4188 4189For C++, also warn for confusing overload resolution for user-defined 4190conversions; and conversions that will never use a type conversion 4191operator: conversions to @code{void}, the same type, a base class or a 4192reference to them. Warnings about conversions between signed and 4193unsigned integers are disabled by default in C++ unless 4194@option{-Wsign-conversion} is explicitly enabled. 4195 4196@item -Wno-conversion-null @r{(C++ and Objective-C++ only)} 4197@opindex Wconversion-null 4198@opindex Wno-conversion-null 4199Do not warn for conversions between @code{NULL} and non-pointer 4200types. @option{-Wconversion-null} is enabled by default. 4201 4202@item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)} 4203@opindex Wzero-as-null-pointer-constant 4204@opindex Wno-zero-as-null-pointer-constant 4205Warn when a literal '0' is used as null pointer constant. This can 4206be useful to facilitate the conversion to @code{nullptr} in C++11. 4207 4208@item -Wempty-body 4209@opindex Wempty-body 4210@opindex Wno-empty-body 4211Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do 4212while} statement. This warning is also enabled by @option{-Wextra}. 4213 4214@item -Wenum-compare 4215@opindex Wenum-compare 4216@opindex Wno-enum-compare 4217Warn about a comparison between values of different enumerated types. 4218In C++ enumeral mismatches in conditional expressions are also 4219diagnosed and the warning is enabled by default. In C this warning is 4220enabled by @option{-Wall}. 4221 4222@item -Wjump-misses-init @r{(C, Objective-C only)} 4223@opindex Wjump-misses-init 4224@opindex Wno-jump-misses-init 4225Warn if a @code{goto} statement or a @code{switch} statement jumps 4226forward across the initialization of a variable, or jumps backward to a 4227label after the variable has been initialized. This only warns about 4228variables that are initialized when they are declared. This warning is 4229only supported for C and Objective-C; in C++ this sort of branch is an 4230error in any case. 4231 4232@option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It 4233can be disabled with the @option{-Wno-jump-misses-init} option. 4234 4235@item -Wsign-compare 4236@opindex Wsign-compare 4237@opindex Wno-sign-compare 4238@cindex warning for comparison of signed and unsigned values 4239@cindex comparison of signed and unsigned values, warning 4240@cindex signed and unsigned values, comparison warning 4241Warn when a comparison between signed and unsigned values could produce 4242an incorrect result when the signed value is converted to unsigned. 4243This warning is also enabled by @option{-Wextra}; to get the other warnings 4244of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}. 4245 4246@item -Wsign-conversion 4247@opindex Wsign-conversion 4248@opindex Wno-sign-conversion 4249Warn for implicit conversions that may change the sign of an integer 4250value, like assigning a signed integer expression to an unsigned 4251integer variable. An explicit cast silences the warning. In C, this 4252option is enabled also by @option{-Wconversion}. 4253 4254@item -Waddress 4255@opindex Waddress 4256@opindex Wno-address 4257Warn about suspicious uses of memory addresses. These include using 4258the address of a function in a conditional expression, such as 4259@code{void func(void); if (func)}, and comparisons against the memory 4260address of a string literal, such as @code{if (x == "abc")}. Such 4261uses typically indicate a programmer error: the address of a function 4262always evaluates to true, so their use in a conditional usually 4263indicate that the programmer forgot the parentheses in a function 4264call; and comparisons against string literals result in unspecified 4265behavior and are not portable in C, so they usually indicate that the 4266programmer intended to use @code{strcmp}. This warning is enabled by 4267@option{-Wall}. 4268 4269@item -Wlogical-op 4270@opindex Wlogical-op 4271@opindex Wno-logical-op 4272Warn about suspicious uses of logical operators in expressions. 4273This includes using logical operators in contexts where a 4274bit-wise operator is likely to be expected. 4275 4276@item -Waggregate-return 4277@opindex Waggregate-return 4278@opindex Wno-aggregate-return 4279Warn if any functions that return structures or unions are defined or 4280called. (In languages where you can return an array, this also elicits 4281a warning.) 4282 4283@item -Wno-attributes 4284@opindex Wno-attributes 4285@opindex Wattributes 4286Do not warn if an unexpected @code{__attribute__} is used, such as 4287unrecognized attributes, function attributes applied to variables, 4288etc. This will not stop errors for incorrect use of supported 4289attributes. 4290 4291@item -Wno-builtin-macro-redefined 4292@opindex Wno-builtin-macro-redefined 4293@opindex Wbuiltin-macro-redefined 4294Do not warn if certain built-in macros are redefined. This suppresses 4295warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__}, 4296@code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}. 4297 4298@item -Wstrict-prototypes @r{(C and Objective-C only)} 4299@opindex Wstrict-prototypes 4300@opindex Wno-strict-prototypes 4301Warn if a function is declared or defined without specifying the 4302argument types. (An old-style function definition is permitted without 4303a warning if preceded by a declaration that specifies the argument 4304types.) 4305 4306@item -Wold-style-declaration @r{(C and Objective-C only)} 4307@opindex Wold-style-declaration 4308@opindex Wno-old-style-declaration 4309Warn for obsolescent usages, according to the C Standard, in a 4310declaration. For example, warn if storage-class specifiers like 4311@code{static} are not the first things in a declaration. This warning 4312is also enabled by @option{-Wextra}. 4313 4314@item -Wold-style-definition @r{(C and Objective-C only)} 4315@opindex Wold-style-definition 4316@opindex Wno-old-style-definition 4317Warn if an old-style function definition is used. A warning is given 4318even if there is a previous prototype. 4319 4320@item -Wmissing-parameter-type @r{(C and Objective-C only)} 4321@opindex Wmissing-parameter-type 4322@opindex Wno-missing-parameter-type 4323A function parameter is declared without a type specifier in K&R-style 4324functions: 4325 4326@smallexample 4327void foo(bar) @{ @} 4328@end smallexample 4329 4330This warning is also enabled by @option{-Wextra}. 4331 4332@item -Wmissing-prototypes @r{(C and Objective-C only)} 4333@opindex Wmissing-prototypes 4334@opindex Wno-missing-prototypes 4335Warn if a global function is defined without a previous prototype 4336declaration. This warning is issued even if the definition itself 4337provides a prototype. The aim is to detect global functions that 4338are not declared in header files. 4339 4340@item -Wmissing-declarations 4341@opindex Wmissing-declarations 4342@opindex Wno-missing-declarations 4343Warn if a global function is defined without a previous declaration. 4344Do so even if the definition itself provides a prototype. 4345Use this option to detect global functions that are not declared in 4346header files. In C++, no warnings are issued for function templates, 4347or for inline functions, or for functions in anonymous namespaces. 4348 4349@item -Wmissing-field-initializers 4350@opindex Wmissing-field-initializers 4351@opindex Wno-missing-field-initializers 4352@opindex W 4353@opindex Wextra 4354@opindex Wno-extra 4355Warn if a structure's initializer has some fields missing. For 4356example, the following code would cause such a warning, because 4357@code{x.h} is implicitly zero: 4358 4359@smallexample 4360struct s @{ int f, g, h; @}; 4361struct s x = @{ 3, 4 @}; 4362@end smallexample 4363 4364This option does not warn about designated initializers, so the following 4365modification would not trigger a warning: 4366 4367@smallexample 4368struct s @{ int f, g, h; @}; 4369struct s x = @{ .f = 3, .g = 4 @}; 4370@end smallexample 4371 4372This warning is included in @option{-Wextra}. To get other @option{-Wextra} 4373warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}. 4374 4375@item -Wmissing-format-attribute 4376@opindex Wmissing-format-attribute 4377@opindex Wno-missing-format-attribute 4378@opindex Wformat 4379@opindex Wno-format 4380Warn about function pointers that might be candidates for @code{format} 4381attributes. Note these are only possible candidates, not absolute ones. 4382GCC will guess that function pointers with @code{format} attributes that 4383are used in assignment, initialization, parameter passing or return 4384statements should have a corresponding @code{format} attribute in the 4385resulting type. I.e.@: the left-hand side of the assignment or 4386initialization, the type of the parameter variable, or the return type 4387of the containing function respectively should also have a @code{format} 4388attribute to avoid the warning. 4389 4390GCC will also warn about function definitions that might be 4391candidates for @code{format} attributes. Again, these are only 4392possible candidates. GCC will guess that @code{format} attributes 4393might be appropriate for any function that calls a function like 4394@code{vprintf} or @code{vscanf}, but this might not always be the 4395case, and some functions for which @code{format} attributes are 4396appropriate may not be detected. 4397 4398@item -Wno-multichar 4399@opindex Wno-multichar 4400@opindex Wmultichar 4401Do not warn if a multicharacter constant (@samp{'FOOF'}) is used. 4402Usually they indicate a typo in the user's code, as they have 4403implementation-defined values, and should not be used in portable code. 4404 4405@item -Wnormalized=<none|id|nfc|nfkc> 4406@opindex Wnormalized= 4407@cindex NFC 4408@cindex NFKC 4409@cindex character set, input normalization 4410In ISO C and ISO C++, two identifiers are different if they are 4411different sequences of characters. However, sometimes when characters 4412outside the basic ASCII character set are used, you can have two 4413different character sequences that look the same. To avoid confusion, 4414the ISO 10646 standard sets out some @dfn{normalization rules} which 4415when applied ensure that two sequences that look the same are turned into 4416the same sequence. GCC can warn you if you are using identifiers that 4417have not been normalized; this option controls that warning. 4418 4419There are four levels of warning supported by GCC. The default is 4420@option{-Wnormalized=nfc}, which warns about any identifier that is 4421not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the 4422recommended form for most uses. 4423 4424Unfortunately, there are some characters allowed in identifiers by 4425ISO C and ISO C++ that, when turned into NFC, are not allowed in 4426identifiers. That is, there's no way to use these symbols in portable 4427ISO C or C++ and have all your identifiers in NFC@. 4428@option{-Wnormalized=id} suppresses the warning for these characters. 4429It is hoped that future versions of the standards involved will correct 4430this, which is why this option is not the default. 4431 4432You can switch the warning off for all characters by writing 4433@option{-Wnormalized=none}. You would only want to do this if you 4434were using some other normalization scheme (like ``D''), because 4435otherwise you can easily create bugs that are literally impossible to see. 4436 4437Some characters in ISO 10646 have distinct meanings but look identical 4438in some fonts or display methodologies, especially once formatting has 4439been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL 4440LETTER N'', will display just like a regular @code{n} that has been 4441placed in a superscript. ISO 10646 defines the @dfn{NFKC} 4442normalization scheme to convert all these into a standard form as 4443well, and GCC will warn if your code is not in NFKC if you use 4444@option{-Wnormalized=nfkc}. This warning is comparable to warning 4445about every identifier that contains the letter O because it might be 4446confused with the digit 0, and so is not the default, but may be 4447useful as a local coding convention if the programming environment is 4448unable to be fixed to display these characters distinctly. 4449 4450@item -Wno-deprecated 4451@opindex Wno-deprecated 4452@opindex Wdeprecated 4453Do not warn about usage of deprecated features. @xref{Deprecated Features}. 4454 4455@item -Wno-deprecated-declarations 4456@opindex Wno-deprecated-declarations 4457@opindex Wdeprecated-declarations 4458Do not warn about uses of functions (@pxref{Function Attributes}), 4459variables (@pxref{Variable Attributes}), and types (@pxref{Type 4460Attributes}) marked as deprecated by using the @code{deprecated} 4461attribute. 4462 4463@item -Wno-overflow 4464@opindex Wno-overflow 4465@opindex Woverflow 4466Do not warn about compile-time overflow in constant expressions. 4467 4468@item -Woverride-init @r{(C and Objective-C only)} 4469@opindex Woverride-init 4470@opindex Wno-override-init 4471@opindex W 4472@opindex Wextra 4473@opindex Wno-extra 4474Warn if an initialized field without side effects is overridden when 4475using designated initializers (@pxref{Designated Inits, , Designated 4476Initializers}). 4477 4478This warning is included in @option{-Wextra}. To get other 4479@option{-Wextra} warnings without this one, use @samp{-Wextra 4480-Wno-override-init}. 4481 4482@item -Wpacked 4483@opindex Wpacked 4484@opindex Wno-packed 4485Warn if a structure is given the packed attribute, but the packed 4486attribute has no effect on the layout or size of the structure. 4487Such structures may be mis-aligned for little benefit. For 4488instance, in this code, the variable @code{f.x} in @code{struct bar} 4489will be misaligned even though @code{struct bar} does not itself 4490have the packed attribute: 4491 4492@smallexample 4493@group 4494struct foo @{ 4495 int x; 4496 char a, b, c, d; 4497@} __attribute__((packed)); 4498struct bar @{ 4499 char z; 4500 struct foo f; 4501@}; 4502@end group 4503@end smallexample 4504 4505@item -Wpacked-bitfield-compat 4506@opindex Wpacked-bitfield-compat 4507@opindex Wno-packed-bitfield-compat 4508The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute 4509on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but 4510the change can lead to differences in the structure layout. GCC 4511informs you when the offset of such a field has changed in GCC 4.4. 4512For example there is no longer a 4-bit padding between field @code{a} 4513and @code{b} in this structure: 4514 4515@smallexample 4516struct foo 4517@{ 4518 char a:4; 4519 char b:8; 4520@} __attribute__ ((packed)); 4521@end smallexample 4522 4523This warning is enabled by default. Use 4524@option{-Wno-packed-bitfield-compat} to disable this warning. 4525 4526@item -Wpadded 4527@opindex Wpadded 4528@opindex Wno-padded 4529Warn if padding is included in a structure, either to align an element 4530of the structure or to align the whole structure. Sometimes when this 4531happens it is possible to rearrange the fields of the structure to 4532reduce the padding and so make the structure smaller. 4533 4534@item -Wredundant-decls 4535@opindex Wredundant-decls 4536@opindex Wno-redundant-decls 4537Warn if anything is declared more than once in the same scope, even in 4538cases where multiple declaration is valid and changes nothing. 4539 4540@item -Wnested-externs @r{(C and Objective-C only)} 4541@opindex Wnested-externs 4542@opindex Wno-nested-externs 4543Warn if an @code{extern} declaration is encountered within a function. 4544 4545@item -Winline 4546@opindex Winline 4547@opindex Wno-inline 4548Warn if a function can not be inlined and it was declared as inline. 4549Even with this option, the compiler will not warn about failures to 4550inline functions declared in system headers. 4551 4552The compiler uses a variety of heuristics to determine whether or not 4553to inline a function. For example, the compiler takes into account 4554the size of the function being inlined and the amount of inlining 4555that has already been done in the current function. Therefore, 4556seemingly insignificant changes in the source program can cause the 4557warnings produced by @option{-Winline} to appear or disappear. 4558 4559@item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)} 4560@opindex Wno-invalid-offsetof 4561@opindex Winvalid-offsetof 4562Suppress warnings from applying the @samp{offsetof} macro to a non-POD 4563type. According to the 1998 ISO C++ standard, applying @samp{offsetof} 4564to a non-POD type is undefined. In existing C++ implementations, 4565however, @samp{offsetof} typically gives meaningful results even when 4566applied to certain kinds of non-POD types. (Such as a simple 4567@samp{struct} that fails to be a POD type only by virtue of having a 4568constructor.) This flag is for users who are aware that they are 4569writing nonportable code and who have deliberately chosen to ignore the 4570warning about it. 4571 4572The restrictions on @samp{offsetof} may be relaxed in a future version 4573of the C++ standard. 4574 4575@item -Wno-int-to-pointer-cast 4576@opindex Wno-int-to-pointer-cast 4577@opindex Wint-to-pointer-cast 4578Suppress warnings from casts to pointer type of an integer of a 4579different size. In C++, casting to a pointer type of smaller size is 4580an error. @option{Wint-to-pointer-cast} is enabled by default. 4581 4582 4583@item -Wno-pointer-to-int-cast @r{(C and Objective-C only)} 4584@opindex Wno-pointer-to-int-cast 4585@opindex Wpointer-to-int-cast 4586Suppress warnings from casts from a pointer to an integer type of a 4587different size. 4588 4589@item -Winvalid-pch 4590@opindex Winvalid-pch 4591@opindex Wno-invalid-pch 4592Warn if a precompiled header (@pxref{Precompiled Headers}) is found in 4593the search path but can't be used. 4594 4595@item -Wlong-long 4596@opindex Wlong-long 4597@opindex Wno-long-long 4598Warn if @samp{long long} type is used. This is enabled by either 4599@option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98 4600modes. To inhibit the warning messages, use @option{-Wno-long-long}. 4601 4602@item -Wvariadic-macros 4603@opindex Wvariadic-macros 4604@opindex Wno-variadic-macros 4605Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU 4606alternate syntax when in pedantic ISO C99 mode. This is default. 4607To inhibit the warning messages, use @option{-Wno-variadic-macros}. 4608 4609@item -Wvector-operation-performance 4610@opindex Wvector-operation-performance 4611@opindex Wno-vector-operation-performance 4612Warn if vector operation is not implemented via SIMD capabilities of the 4613architecture. Mainly useful for the performance tuning. 4614Vector operation can be implemented @code{piecewise}, which means that the 4615scalar operation is performed on every vector element; 4616@code{in parallel}, which means that the vector operation is implemented 4617using scalars of wider type, which normally is more performance efficient; 4618and @code{as a single scalar}, which means that vector fits into a 4619scalar type. 4620 4621@item -Wvla 4622@opindex Wvla 4623@opindex Wno-vla 4624Warn if variable length array is used in the code. 4625@option{-Wno-vla} will prevent the @option{-pedantic} warning of 4626the variable length array. 4627 4628@item -Wvolatile-register-var 4629@opindex Wvolatile-register-var 4630@opindex Wno-volatile-register-var 4631Warn if a register variable is declared volatile. The volatile 4632modifier does not inhibit all optimizations that may eliminate reads 4633and/or writes to register variables. This warning is enabled by 4634@option{-Wall}. 4635 4636@item -Wdisabled-optimization 4637@opindex Wdisabled-optimization 4638@opindex Wno-disabled-optimization 4639Warn if a requested optimization pass is disabled. This warning does 4640not generally indicate that there is anything wrong with your code; it 4641merely indicates that GCC's optimizers were unable to handle the code 4642effectively. Often, the problem is that your code is too big or too 4643complex; GCC will refuse to optimize programs when the optimization 4644itself is likely to take inordinate amounts of time. 4645 4646@item -Wpointer-sign @r{(C and Objective-C only)} 4647@opindex Wpointer-sign 4648@opindex Wno-pointer-sign 4649Warn for pointer argument passing or assignment with different signedness. 4650This option is only supported for C and Objective-C@. It is implied by 4651@option{-Wall} and by @option{-pedantic}, which can be disabled with 4652@option{-Wno-pointer-sign}. 4653 4654@item -Wstack-protector 4655@opindex Wstack-protector 4656@opindex Wno-stack-protector 4657This option is only active when @option{-fstack-protector} is active. It 4658warns about functions that will not be protected against stack smashing. 4659 4660@item -Wno-mudflap 4661@opindex Wno-mudflap 4662Suppress warnings about constructs that cannot be instrumented by 4663@option{-fmudflap}. 4664 4665@item -Woverlength-strings 4666@opindex Woverlength-strings 4667@opindex Wno-overlength-strings 4668Warn about string constants that are longer than the ``minimum 4669maximum'' length specified in the C standard. Modern compilers 4670generally allow string constants that are much longer than the 4671standard's minimum limit, but very portable programs should avoid 4672using longer strings. 4673 4674The limit applies @emph{after} string constant concatenation, and does 4675not count the trailing NUL@. In C90, the limit was 509 characters; in 4676C99, it was raised to 4095. C++98 does not specify a normative 4677minimum maximum, so we do not diagnose overlength strings in C++@. 4678 4679This option is implied by @option{-pedantic}, and can be disabled with 4680@option{-Wno-overlength-strings}. 4681 4682@item -Wunsuffixed-float-constants @r{(C and Objective-C only)} 4683@opindex Wunsuffixed-float-constants 4684 4685GCC will issue a warning for any floating constant that does not have 4686a suffix. When used together with @option{-Wsystem-headers} it will 4687warn about such constants in system header files. This can be useful 4688when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma 4689from the decimal floating-point extension to C99. 4690@end table 4691 4692@node Debugging Options 4693@section Options for Debugging Your Program or GCC 4694@cindex options, debugging 4695@cindex debugging information options 4696 4697GCC has various special options that are used for debugging 4698either your program or GCC: 4699 4700@table @gcctabopt 4701@item -g 4702@opindex g 4703Produce debugging information in the operating system's native format 4704(stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging 4705information. 4706 4707On most systems that use stabs format, @option{-g} enables use of extra 4708debugging information that only GDB can use; this extra information 4709makes debugging work better in GDB but will probably make other debuggers 4710crash or 4711refuse to read the program. If you want to control for certain whether 4712to generate the extra information, use @option{-gstabs+}, @option{-gstabs}, 4713@option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below). 4714 4715GCC allows you to use @option{-g} with 4716@option{-O}. The shortcuts taken by optimized code may occasionally 4717produce surprising results: some variables you declared may not exist 4718at all; flow of control may briefly move where you did not expect it; 4719some statements may not be executed because they compute constant 4720results or their values were already at hand; some statements may 4721execute in different places because they were moved out of loops. 4722 4723Nevertheless it proves possible to debug optimized output. This makes 4724it reasonable to use the optimizer for programs that might have bugs. 4725 4726The following options are useful when GCC is generated with the 4727capability for more than one debugging format. 4728 4729@item -ggdb 4730@opindex ggdb 4731Produce debugging information for use by GDB@. This means to use the 4732most expressive format available (DWARF 2, stabs, or the native format 4733if neither of those are supported), including GDB extensions if at all 4734possible. 4735 4736@item -gstabs 4737@opindex gstabs 4738Produce debugging information in stabs format (if that is supported), 4739without GDB extensions. This is the format used by DBX on most BSD 4740systems. On MIPS, Alpha and System V Release 4 systems this option 4741produces stabs debugging output that is not understood by DBX or SDB@. 4742On System V Release 4 systems this option requires the GNU assembler. 4743 4744@item -feliminate-unused-debug-symbols 4745@opindex feliminate-unused-debug-symbols 4746Produce debugging information in stabs format (if that is supported), 4747for only symbols that are actually used. 4748 4749@item -femit-class-debug-always 4750Instead of emitting debugging information for a C++ class in only one 4751object file, emit it in all object files using the class. This option 4752should be used only with debuggers that are unable to handle the way GCC 4753normally emits debugging information for classes because using this 4754option will increase the size of debugging information by as much as a 4755factor of two. 4756 4757@item -fno-debug-types-section 4758@opindex fno-debug-types-section 4759@opindex fdebug-types-section 4760By default when using DWARF v4 or higher type DIEs will be put into 4761their own .debug_types section instead of making them part of the 4762.debug_info section. It is more efficient to put them in a separate 4763comdat sections since the linker will then be able to remove duplicates. 4764But not all DWARF consumers support .debug_types sections yet. 4765 4766@item -gstabs+ 4767@opindex gstabs+ 4768Produce debugging information in stabs format (if that is supported), 4769using GNU extensions understood only by the GNU debugger (GDB)@. The 4770use of these extensions is likely to make other debuggers crash or 4771refuse to read the program. 4772 4773@item -gcoff 4774@opindex gcoff 4775Produce debugging information in COFF format (if that is supported). 4776This is the format used by SDB on most System V systems prior to 4777System V Release 4. 4778 4779@item -gxcoff 4780@opindex gxcoff 4781Produce debugging information in XCOFF format (if that is supported). 4782This is the format used by the DBX debugger on IBM RS/6000 systems. 4783 4784@item -gxcoff+ 4785@opindex gxcoff+ 4786Produce debugging information in XCOFF format (if that is supported), 4787using GNU extensions understood only by the GNU debugger (GDB)@. The 4788use of these extensions is likely to make other debuggers crash or 4789refuse to read the program, and may cause assemblers other than the GNU 4790assembler (GAS) to fail with an error. 4791 4792@item -gdwarf-@var{version} 4793@opindex gdwarf-@var{version} 4794Produce debugging information in DWARF format (if that is 4795supported). This is the format used by DBX on IRIX 6. The value 4796of @var{version} may be either 2, 3 or 4; the default version is 2. 4797 4798Note that with DWARF version 2 some ports require, and will always 4799use, some non-conflicting DWARF 3 extensions in the unwind tables. 4800 4801Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments} 4802for maximum benefit. 4803 4804@item -grecord-gcc-switches 4805@opindex grecord-gcc-switches 4806This switch causes the command-line options used to invoke the 4807compiler that may affect code generation to be appended to the 4808DW_AT_producer attribute in DWARF debugging information. The options 4809are concatenated with spaces separating them from each other and from 4810the compiler version. See also @option{-frecord-gcc-switches} for another 4811way of storing compiler options into the object file. 4812 4813@item -gno-record-gcc-switches 4814@opindex gno-record-gcc-switches 4815Disallow appending command-line options to the DW_AT_producer attribute 4816in DWARF debugging information. This is the default. 4817 4818@item -gstrict-dwarf 4819@opindex gstrict-dwarf 4820Disallow using extensions of later DWARF standard version than selected 4821with @option{-gdwarf-@var{version}}. On most targets using non-conflicting 4822DWARF extensions from later standard versions is allowed. 4823 4824@item -gno-strict-dwarf 4825@opindex gno-strict-dwarf 4826Allow using extensions of later DWARF standard version than selected with 4827@option{-gdwarf-@var{version}}. 4828 4829@item -gvms 4830@opindex gvms 4831Produce debugging information in VMS debug format (if that is 4832supported). This is the format used by DEBUG on VMS systems. 4833 4834@item -g@var{level} 4835@itemx -ggdb@var{level} 4836@itemx -gstabs@var{level} 4837@itemx -gcoff@var{level} 4838@itemx -gxcoff@var{level} 4839@itemx -gvms@var{level} 4840Request debugging information and also use @var{level} to specify how 4841much information. The default level is 2. 4842 4843Level 0 produces no debug information at all. Thus, @option{-g0} negates 4844@option{-g}. 4845 4846Level 1 produces minimal information, enough for making backtraces in 4847parts of the program that you don't plan to debug. This includes 4848descriptions of functions and external variables, but no information 4849about local variables and no line numbers. 4850 4851Level 3 includes extra information, such as all the macro definitions 4852present in the program. Some debuggers support macro expansion when 4853you use @option{-g3}. 4854 4855@option{-gdwarf-2} does not accept a concatenated debug level, because 4856GCC used to support an option @option{-gdwarf} that meant to generate 4857debug information in version 1 of the DWARF format (which is very 4858different from version 2), and it would have been too confusing. That 4859debug format is long obsolete, but the option cannot be changed now. 4860Instead use an additional @option{-g@var{level}} option to change the 4861debug level for DWARF. 4862 4863@item -gtoggle 4864@opindex gtoggle 4865Turn off generation of debug info, if leaving out this option would have 4866generated it, or turn it on at level 2 otherwise. The position of this 4867argument in the command line does not matter, it takes effect after all 4868other options are processed, and it does so only once, no matter how 4869many times it is given. This is mainly intended to be used with 4870@option{-fcompare-debug}. 4871 4872@item -fdump-final-insns@r{[}=@var{file}@r{]} 4873@opindex fdump-final-insns 4874Dump the final internal representation (RTL) to @var{file}. If the 4875optional argument is omitted (or if @var{file} is @code{.}), the name 4876of the dump file will be determined by appending @code{.gkd} to the 4877compilation output file name. 4878 4879@item -fcompare-debug@r{[}=@var{opts}@r{]} 4880@opindex fcompare-debug 4881@opindex fno-compare-debug 4882If no error occurs during compilation, run the compiler a second time, 4883adding @var{opts} and @option{-fcompare-debug-second} to the arguments 4884passed to the second compilation. Dump the final internal 4885representation in both compilations, and print an error if they differ. 4886 4887If the equal sign is omitted, the default @option{-gtoggle} is used. 4888 4889The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty 4890and nonzero, implicitly enables @option{-fcompare-debug}. If 4891@env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash, 4892then it is used for @var{opts}, otherwise the default @option{-gtoggle} 4893is used. 4894 4895@option{-fcompare-debug=}, with the equal sign but without @var{opts}, 4896is equivalent to @option{-fno-compare-debug}, which disables the dumping 4897of the final representation and the second compilation, preventing even 4898@env{GCC_COMPARE_DEBUG} from taking effect. 4899 4900To verify full coverage during @option{-fcompare-debug} testing, set 4901@env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden}, 4902which GCC will reject as an invalid option in any actual compilation 4903(rather than preprocessing, assembly or linking). To get just a 4904warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug 4905not overridden} will do. 4906 4907@item -fcompare-debug-second 4908@opindex fcompare-debug-second 4909This option is implicitly passed to the compiler for the second 4910compilation requested by @option{-fcompare-debug}, along with options to 4911silence warnings, and omitting other options that would cause 4912side-effect compiler outputs to files or to the standard output. Dump 4913files and preserved temporary files are renamed so as to contain the 4914@code{.gk} additional extension during the second compilation, to avoid 4915overwriting those generated by the first. 4916 4917When this option is passed to the compiler driver, it causes the 4918@emph{first} compilation to be skipped, which makes it useful for little 4919other than debugging the compiler proper. 4920 4921@item -feliminate-dwarf2-dups 4922@opindex feliminate-dwarf2-dups 4923Compress DWARF2 debugging information by eliminating duplicated 4924information about each symbol. This option only makes sense when 4925generating DWARF2 debugging information with @option{-gdwarf-2}. 4926 4927@item -femit-struct-debug-baseonly 4928Emit debug information for struct-like types 4929only when the base name of the compilation source file 4930matches the base name of file in which the struct was defined. 4931 4932This option substantially reduces the size of debugging information, 4933but at significant potential loss in type information to the debugger. 4934See @option{-femit-struct-debug-reduced} for a less aggressive option. 4935See @option{-femit-struct-debug-detailed} for more detailed control. 4936 4937This option works only with DWARF 2. 4938 4939@item -femit-struct-debug-reduced 4940Emit debug information for struct-like types 4941only when the base name of the compilation source file 4942matches the base name of file in which the type was defined, 4943unless the struct is a template or defined in a system header. 4944 4945This option significantly reduces the size of debugging information, 4946with some potential loss in type information to the debugger. 4947See @option{-femit-struct-debug-baseonly} for a more aggressive option. 4948See @option{-femit-struct-debug-detailed} for more detailed control. 4949 4950This option works only with DWARF 2. 4951 4952@item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} 4953Specify the struct-like types 4954for which the compiler will generate debug information. 4955The intent is to reduce duplicate struct debug information 4956between different object files within the same program. 4957 4958This option is a detailed version of 4959@option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly}, 4960which will serve for most needs. 4961 4962A specification has the syntax@* 4963[@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none}) 4964 4965The optional first word limits the specification to 4966structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}). 4967A struct type is used directly when it is the type of a variable, member. 4968Indirect uses arise through pointers to structs. 4969That is, when use of an incomplete struct would be legal, the use is indirect. 4970An example is 4971@samp{struct one direct; struct two * indirect;}. 4972 4973The optional second word limits the specification to 4974ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}). 4975Generic structs are a bit complicated to explain. 4976For C++, these are non-explicit specializations of template classes, 4977or non-template classes within the above. 4978Other programming languages have generics, 4979but @samp{-femit-struct-debug-detailed} does not yet implement them. 4980 4981The third word specifies the source files for those 4982structs for which the compiler will emit debug information. 4983The values @samp{none} and @samp{any} have the normal meaning. 4984The value @samp{base} means that 4985the base of name of the file in which the type declaration appears 4986must match the base of the name of the main compilation file. 4987In practice, this means that 4988types declared in @file{foo.c} and @file{foo.h} will have debug information, 4989but types declared in other header will not. 4990The value @samp{sys} means those types satisfying @samp{base} 4991or declared in system or compiler headers. 4992 4993You may need to experiment to determine the best settings for your application. 4994 4995The default is @samp{-femit-struct-debug-detailed=all}. 4996 4997This option works only with DWARF 2. 4998 4999@item -fno-merge-debug-strings 5000@opindex fmerge-debug-strings 5001@opindex fno-merge-debug-strings 5002Direct the linker to not merge together strings in the debugging 5003information that are identical in different object files. Merging is 5004not supported by all assemblers or linkers. Merging decreases the size 5005of the debug information in the output file at the cost of increasing 5006link processing time. Merging is enabled by default. 5007 5008@item -fdebug-prefix-map=@var{old}=@var{new} 5009@opindex fdebug-prefix-map 5010When compiling files in directory @file{@var{old}}, record debugging 5011information describing them as in @file{@var{new}} instead. 5012 5013@item -fno-dwarf2-cfi-asm 5014@opindex fdwarf2-cfi-asm 5015@opindex fno-dwarf2-cfi-asm 5016Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section 5017instead of using GAS @code{.cfi_*} directives. 5018 5019@cindex @command{prof} 5020@item -p 5021@opindex p 5022Generate extra code to write profile information suitable for the 5023analysis program @command{prof}. You must use this option when compiling 5024the source files you want data about, and you must also use it when 5025linking. 5026 5027@cindex @command{gprof} 5028@item -pg 5029@opindex pg 5030Generate extra code to write profile information suitable for the 5031analysis program @command{gprof}. You must use this option when compiling 5032the source files you want data about, and you must also use it when 5033linking. 5034 5035@item -Q 5036@opindex Q 5037Makes the compiler print out each function name as it is compiled, and 5038print some statistics about each pass when it finishes. 5039 5040@item -ftime-report 5041@opindex ftime-report 5042Makes the compiler print some statistics about the time consumed by each 5043pass when it finishes. 5044 5045@item -fmem-report 5046@opindex fmem-report 5047Makes the compiler print some statistics about permanent memory 5048allocation when it finishes. 5049 5050@item -fpre-ipa-mem-report 5051@opindex fpre-ipa-mem-report 5052@item -fpost-ipa-mem-report 5053@opindex fpost-ipa-mem-report 5054Makes the compiler print some statistics about permanent memory 5055allocation before or after interprocedural optimization. 5056 5057@item -fstack-usage 5058@opindex fstack-usage 5059Makes the compiler output stack usage information for the program, on a 5060per-function basis. The filename for the dump is made by appending 5061@file{.su} to the @var{auxname}. @var{auxname} is generated from the name of 5062the output file, if explicitly specified and it is not an executable, 5063otherwise it is the basename of the source file. An entry is made up 5064of three fields: 5065 5066@itemize 5067@item 5068The name of the function. 5069@item 5070A number of bytes. 5071@item 5072One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}. 5073@end itemize 5074 5075The qualifier @code{static} means that the function manipulates the stack 5076statically: a fixed number of bytes are allocated for the frame on function 5077entry and released on function exit; no stack adjustments are otherwise made 5078in the function. The second field is this fixed number of bytes. 5079 5080The qualifier @code{dynamic} means that the function manipulates the stack 5081dynamically: in addition to the static allocation described above, stack 5082adjustments are made in the body of the function, for example to push/pop 5083arguments around function calls. If the qualifier @code{bounded} is also 5084present, the amount of these adjustments is bounded at compile time and 5085the second field is an upper bound of the total amount of stack used by 5086the function. If it is not present, the amount of these adjustments is 5087not bounded at compile time and the second field only represents the 5088bounded part. 5089 5090@item -fprofile-arcs 5091@opindex fprofile-arcs 5092Add code so that program flow @dfn{arcs} are instrumented. During 5093execution the program records how many times each branch and call is 5094executed and how many times it is taken or returns. When the compiled 5095program exits it saves this data to a file called 5096@file{@var{auxname}.gcda} for each source file. The data may be used for 5097profile-directed optimizations (@option{-fbranch-probabilities}), or for 5098test coverage analysis (@option{-ftest-coverage}). Each object file's 5099@var{auxname} is generated from the name of the output file, if 5100explicitly specified and it is not the final executable, otherwise it is 5101the basename of the source file. In both cases any suffix is removed 5102(e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or 5103@file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}). 5104@xref{Cross-profiling}. 5105 5106@cindex @command{gcov} 5107@item --coverage 5108@opindex coverage 5109 5110This option is used to compile and link code instrumented for coverage 5111analysis. The option is a synonym for @option{-fprofile-arcs} 5112@option{-ftest-coverage} (when compiling) and @option{-lgcov} (when 5113linking). See the documentation for those options for more details. 5114 5115@itemize 5116 5117@item 5118Compile the source files with @option{-fprofile-arcs} plus optimization 5119and code generation options. For test coverage analysis, use the 5120additional @option{-ftest-coverage} option. You do not need to profile 5121every source file in a program. 5122 5123@item 5124Link your object files with @option{-lgcov} or @option{-fprofile-arcs} 5125(the latter implies the former). 5126 5127@item 5128Run the program on a representative workload to generate the arc profile 5129information. This may be repeated any number of times. You can run 5130concurrent instances of your program, and provided that the file system 5131supports locking, the data files will be correctly updated. Also 5132@code{fork} calls are detected and correctly handled (double counting 5133will not happen). 5134 5135@item 5136For profile-directed optimizations, compile the source files again with 5137the same optimization and code generation options plus 5138@option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that 5139Control Optimization}). 5140 5141@item 5142For test coverage analysis, use @command{gcov} to produce human readable 5143information from the @file{.gcno} and @file{.gcda} files. Refer to the 5144@command{gcov} documentation for further information. 5145 5146@end itemize 5147 5148With @option{-fprofile-arcs}, for each function of your program GCC 5149creates a program flow graph, then finds a spanning tree for the graph. 5150Only arcs that are not on the spanning tree have to be instrumented: the 5151compiler adds code to count the number of times that these arcs are 5152executed. When an arc is the only exit or only entrance to a block, the 5153instrumentation code can be added to the block; otherwise, a new basic 5154block must be created to hold the instrumentation code. 5155 5156@need 2000 5157@item -ftest-coverage 5158@opindex ftest-coverage 5159Produce a notes file that the @command{gcov} code-coverage utility 5160(@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to 5161show program coverage. Each source file's note file is called 5162@file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option 5163above for a description of @var{auxname} and instructions on how to 5164generate test coverage data. Coverage data will match the source files 5165more closely, if you do not optimize. 5166 5167@item -fdbg-cnt-list 5168@opindex fdbg-cnt-list 5169Print the name and the counter upper bound for all debug counters. 5170 5171 5172@item -fdbg-cnt=@var{counter-value-list} 5173@opindex fdbg-cnt 5174Set the internal debug counter upper bound. @var{counter-value-list} 5175is a comma-separated list of @var{name}:@var{value} pairs 5176which sets the upper bound of each debug counter @var{name} to @var{value}. 5177All debug counters have the initial upper bound of @var{UINT_MAX}, 5178thus dbg_cnt() returns true always unless the upper bound is set by this option. 5179e.g. With -fdbg-cnt=dce:10,tail_call:0 5180dbg_cnt(dce) will return true only for first 10 invocations 5181 5182@itemx -fenable-@var{kind}-@var{pass} 5183@itemx -fdisable-@var{kind}-@var{pass}=@var{range-list} 5184@opindex fdisable- 5185@opindex fenable- 5186 5187This is a set of debugging options that are used to explicitly disable/enable 5188optimization passes. For compiler users, regular options for enabling/disabling 5189passes should be used instead. 5190 5191@itemize 5192 5193@item -fdisable-ipa-@var{pass} 5194Disable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is 5195statically invoked in the compiler multiple times, the pass name should be 5196appended with a sequential number starting from 1. 5197 5198@item -fdisable-rtl-@var{pass} 5199@item -fdisable-rtl-@var{pass}=@var{range-list} 5200Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is 5201statically invoked in the compiler multiple times, the pass name should be 5202appended with a sequential number starting from 1. @var{range-list} is a comma 5203seperated list of function ranges or assembler names. Each range is a number 5204pair seperated by a colon. The range is inclusive in both ends. If the range 5205is trivial, the number pair can be simplified as a single number. If the 5206function's cgraph node's @var{uid} is falling within one of the specified ranges, 5207the @var{pass} is disabled for that function. The @var{uid} is shown in the 5208function header of a dump file, and the pass names can be dumped by using 5209option @option{-fdump-passes}. 5210 5211@item -fdisable-tree-@var{pass} 5212@item -fdisable-tree-@var{pass}=@var{range-list} 5213Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of 5214option arguments. 5215 5216@item -fenable-ipa-@var{pass} 5217Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is 5218statically invoked in the compiler multiple times, the pass name should be 5219appended with a sequential number starting from 1. 5220 5221@item -fenable-rtl-@var{pass} 5222@item -fenable-rtl-@var{pass}=@var{range-list} 5223Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument 5224description and examples. 5225 5226@item -fenable-tree-@var{pass} 5227@item -fenable-tree-@var{pass}=@var{range-list} 5228Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description 5229of option arguments. 5230 5231@smallexample 5232 5233# disable ccp1 for all functions 5234 -fdisable-tree-ccp1 5235# disable complete unroll for function whose cgraph node uid is 1 5236 -fenable-tree-cunroll=1 5237# disable gcse2 for functions at the following ranges [1,1], 5238# [300,400], and [400,1000] 5239# disable gcse2 for functions foo and foo2 5240 -fdisable-rtl-gcse2=foo,foo2 5241# disable early inlining 5242 -fdisable-tree-einline 5243# disable ipa inlining 5244 -fdisable-ipa-inline 5245# enable tree full unroll 5246 -fenable-tree-unroll 5247 5248@end smallexample 5249 5250@end itemize 5251 5252@item -d@var{letters} 5253@itemx -fdump-rtl-@var{pass} 5254@opindex d 5255Says to make debugging dumps during compilation at times specified by 5256@var{letters}. This is used for debugging the RTL-based passes of the 5257compiler. The file names for most of the dumps are made by appending 5258a pass number and a word to the @var{dumpname}, and the files are 5259created in the directory of the output file. Note that the pass 5260number is computed statically as passes get registered into the pass 5261manager. Thus the numbering is not related to the dynamic order of 5262execution of passes. In particular, a pass installed by a plugin 5263could have a number over 200 even if it executed quite early. 5264@var{dumpname} is generated from the name of the output file, if 5265explicitly specified and it is not an executable, otherwise it is the 5266basename of the source file. These switches may have different effects 5267when @option{-E} is used for preprocessing. 5268 5269Debug dumps can be enabled with a @option{-fdump-rtl} switch or some 5270@option{-d} option @var{letters}. Here are the possible 5271letters for use in @var{pass} and @var{letters}, and their meanings: 5272 5273@table @gcctabopt 5274 5275@item -fdump-rtl-alignments 5276@opindex fdump-rtl-alignments 5277Dump after branch alignments have been computed. 5278 5279@item -fdump-rtl-asmcons 5280@opindex fdump-rtl-asmcons 5281Dump after fixing rtl statements that have unsatisfied in/out constraints. 5282 5283@item -fdump-rtl-auto_inc_dec 5284@opindex fdump-rtl-auto_inc_dec 5285Dump after auto-inc-dec discovery. This pass is only run on 5286architectures that have auto inc or auto dec instructions. 5287 5288@item -fdump-rtl-barriers 5289@opindex fdump-rtl-barriers 5290Dump after cleaning up the barrier instructions. 5291 5292@item -fdump-rtl-bbpart 5293@opindex fdump-rtl-bbpart 5294Dump after partitioning hot and cold basic blocks. 5295 5296@item -fdump-rtl-bbro 5297@opindex fdump-rtl-bbro 5298Dump after block reordering. 5299 5300@item -fdump-rtl-btl1 5301@itemx -fdump-rtl-btl2 5302@opindex fdump-rtl-btl2 5303@opindex fdump-rtl-btl2 5304@option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping 5305after the two branch 5306target load optimization passes. 5307 5308@item -fdump-rtl-bypass 5309@opindex fdump-rtl-bypass 5310Dump after jump bypassing and control flow optimizations. 5311 5312@item -fdump-rtl-combine 5313@opindex fdump-rtl-combine 5314Dump after the RTL instruction combination pass. 5315 5316@item -fdump-rtl-compgotos 5317@opindex fdump-rtl-compgotos 5318Dump after duplicating the computed gotos. 5319 5320@item -fdump-rtl-ce1 5321@itemx -fdump-rtl-ce2 5322@itemx -fdump-rtl-ce3 5323@opindex fdump-rtl-ce1 5324@opindex fdump-rtl-ce2 5325@opindex fdump-rtl-ce3 5326@option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and 5327@option{-fdump-rtl-ce3} enable dumping after the three 5328if conversion passes. 5329 5330@itemx -fdump-rtl-cprop_hardreg 5331@opindex fdump-rtl-cprop_hardreg 5332Dump after hard register copy propagation. 5333 5334@itemx -fdump-rtl-csa 5335@opindex fdump-rtl-csa 5336Dump after combining stack adjustments. 5337 5338@item -fdump-rtl-cse1 5339@itemx -fdump-rtl-cse2 5340@opindex fdump-rtl-cse1 5341@opindex fdump-rtl-cse2 5342@option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after 5343the two common sub-expression elimination passes. 5344 5345@itemx -fdump-rtl-dce 5346@opindex fdump-rtl-dce 5347Dump after the standalone dead code elimination passes. 5348 5349@itemx -fdump-rtl-dbr 5350@opindex fdump-rtl-dbr 5351Dump after delayed branch scheduling. 5352 5353@item -fdump-rtl-dce1 5354@itemx -fdump-rtl-dce2 5355@opindex fdump-rtl-dce1 5356@opindex fdump-rtl-dce2 5357@option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after 5358the two dead store elimination passes. 5359 5360@item -fdump-rtl-eh 5361@opindex fdump-rtl-eh 5362Dump after finalization of EH handling code. 5363 5364@item -fdump-rtl-eh_ranges 5365@opindex fdump-rtl-eh_ranges 5366Dump after conversion of EH handling range regions. 5367 5368@item -fdump-rtl-expand 5369@opindex fdump-rtl-expand 5370Dump after RTL generation. 5371 5372@item -fdump-rtl-fwprop1 5373@itemx -fdump-rtl-fwprop2 5374@opindex fdump-rtl-fwprop1 5375@opindex fdump-rtl-fwprop2 5376@option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable 5377dumping after the two forward propagation passes. 5378 5379@item -fdump-rtl-gcse1 5380@itemx -fdump-rtl-gcse2 5381@opindex fdump-rtl-gcse1 5382@opindex fdump-rtl-gcse2 5383@option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping 5384after global common subexpression elimination. 5385 5386@item -fdump-rtl-init-regs 5387@opindex fdump-rtl-init-regs 5388Dump after the initialization of the registers. 5389 5390@item -fdump-rtl-initvals 5391@opindex fdump-rtl-initvals 5392Dump after the computation of the initial value sets. 5393 5394@itemx -fdump-rtl-into_cfglayout 5395@opindex fdump-rtl-into_cfglayout 5396Dump after converting to cfglayout mode. 5397 5398@item -fdump-rtl-ira 5399@opindex fdump-rtl-ira 5400Dump after iterated register allocation. 5401 5402@item -fdump-rtl-jump 5403@opindex fdump-rtl-jump 5404Dump after the second jump optimization. 5405 5406@item -fdump-rtl-loop2 5407@opindex fdump-rtl-loop2 5408@option{-fdump-rtl-loop2} enables dumping after the rtl 5409loop optimization passes. 5410 5411@item -fdump-rtl-mach 5412@opindex fdump-rtl-mach 5413Dump after performing the machine dependent reorganization pass, if that 5414pass exists. 5415 5416@item -fdump-rtl-mode_sw 5417@opindex fdump-rtl-mode_sw 5418Dump after removing redundant mode switches. 5419 5420@item -fdump-rtl-rnreg 5421@opindex fdump-rtl-rnreg 5422Dump after register renumbering. 5423 5424@itemx -fdump-rtl-outof_cfglayout 5425@opindex fdump-rtl-outof_cfglayout 5426Dump after converting from cfglayout mode. 5427 5428@item -fdump-rtl-peephole2 5429@opindex fdump-rtl-peephole2 5430Dump after the peephole pass. 5431 5432@item -fdump-rtl-postreload 5433@opindex fdump-rtl-postreload 5434Dump after post-reload optimizations. 5435 5436@itemx -fdump-rtl-pro_and_epilogue 5437@opindex fdump-rtl-pro_and_epilogue 5438Dump after generating the function prologues and epilogues. 5439 5440@item -fdump-rtl-regmove 5441@opindex fdump-rtl-regmove 5442Dump after the register move pass. 5443 5444@item -fdump-rtl-sched1 5445@itemx -fdump-rtl-sched2 5446@opindex fdump-rtl-sched1 5447@opindex fdump-rtl-sched2 5448@option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping 5449after the basic block scheduling passes. 5450 5451@item -fdump-rtl-see 5452@opindex fdump-rtl-see 5453Dump after sign extension elimination. 5454 5455@item -fdump-rtl-seqabstr 5456@opindex fdump-rtl-seqabstr 5457Dump after common sequence discovery. 5458 5459@item -fdump-rtl-shorten 5460@opindex fdump-rtl-shorten 5461Dump after shortening branches. 5462 5463@item -fdump-rtl-sibling 5464@opindex fdump-rtl-sibling 5465Dump after sibling call optimizations. 5466 5467@item -fdump-rtl-split1 5468@itemx -fdump-rtl-split2 5469@itemx -fdump-rtl-split3 5470@itemx -fdump-rtl-split4 5471@itemx -fdump-rtl-split5 5472@opindex fdump-rtl-split1 5473@opindex fdump-rtl-split2 5474@opindex fdump-rtl-split3 5475@opindex fdump-rtl-split4 5476@opindex fdump-rtl-split5 5477@option{-fdump-rtl-split1}, @option{-fdump-rtl-split2}, 5478@option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and 5479@option{-fdump-rtl-split5} enable dumping after five rounds of 5480instruction splitting. 5481 5482@item -fdump-rtl-sms 5483@opindex fdump-rtl-sms 5484Dump after modulo scheduling. This pass is only run on some 5485architectures. 5486 5487@item -fdump-rtl-stack 5488@opindex fdump-rtl-stack 5489Dump after conversion from GCC's "flat register file" registers to the 5490x87's stack-like registers. This pass is only run on x86 variants. 5491 5492@item -fdump-rtl-subreg1 5493@itemx -fdump-rtl-subreg2 5494@opindex fdump-rtl-subreg1 5495@opindex fdump-rtl-subreg2 5496@option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after 5497the two subreg expansion passes. 5498 5499@item -fdump-rtl-unshare 5500@opindex fdump-rtl-unshare 5501Dump after all rtl has been unshared. 5502 5503@item -fdump-rtl-vartrack 5504@opindex fdump-rtl-vartrack 5505Dump after variable tracking. 5506 5507@item -fdump-rtl-vregs 5508@opindex fdump-rtl-vregs 5509Dump after converting virtual registers to hard registers. 5510 5511@item -fdump-rtl-web 5512@opindex fdump-rtl-web 5513Dump after live range splitting. 5514 5515@item -fdump-rtl-regclass 5516@itemx -fdump-rtl-subregs_of_mode_init 5517@itemx -fdump-rtl-subregs_of_mode_finish 5518@itemx -fdump-rtl-dfinit 5519@itemx -fdump-rtl-dfinish 5520@opindex fdump-rtl-regclass 5521@opindex fdump-rtl-subregs_of_mode_init 5522@opindex fdump-rtl-subregs_of_mode_finish 5523@opindex fdump-rtl-dfinit 5524@opindex fdump-rtl-dfinish 5525These dumps are defined but always produce empty files. 5526 5527@item -da 5528@itemx -fdump-rtl-all 5529@opindex da 5530@opindex fdump-rtl-all 5531Produce all the dumps listed above. 5532 5533@item -dA 5534@opindex dA 5535Annotate the assembler output with miscellaneous debugging information. 5536 5537@item -dD 5538@opindex dD 5539Dump all macro definitions, at the end of preprocessing, in addition to 5540normal output. 5541 5542@item -dH 5543@opindex dH 5544Produce a core dump whenever an error occurs. 5545 5546@item -dp 5547@opindex dp 5548Annotate the assembler output with a comment indicating which 5549pattern and alternative was used. The length of each instruction is 5550also printed. 5551 5552@item -dP 5553@opindex dP 5554Dump the RTL in the assembler output as a comment before each instruction. 5555Also turns on @option{-dp} annotation. 5556 5557@item -dv 5558@opindex dv 5559For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}), 5560dump a representation of the control flow graph suitable for viewing with VCG 5561to @file{@var{file}.@var{pass}.vcg}. 5562 5563@item -dx 5564@opindex dx 5565Just generate RTL for a function instead of compiling it. Usually used 5566with @option{-fdump-rtl-expand}. 5567@end table 5568 5569@item -fdump-noaddr 5570@opindex fdump-noaddr 5571When doing debugging dumps, suppress address output. This makes it more 5572feasible to use diff on debugging dumps for compiler invocations with 5573different compiler binaries and/or different 5574text / bss / data / heap / stack / dso start locations. 5575 5576@item -fdump-unnumbered 5577@opindex fdump-unnumbered 5578When doing debugging dumps, suppress instruction numbers and address output. 5579This makes it more feasible to use diff on debugging dumps for compiler 5580invocations with different options, in particular with and without 5581@option{-g}. 5582 5583@item -fdump-unnumbered-links 5584@opindex fdump-unnumbered-links 5585When doing debugging dumps (see @option{-d} option above), suppress 5586instruction numbers for the links to the previous and next instructions 5587in a sequence. 5588 5589@item -fdump-translation-unit @r{(C++ only)} 5590@itemx -fdump-translation-unit-@var{options} @r{(C++ only)} 5591@opindex fdump-translation-unit 5592Dump a representation of the tree structure for the entire translation 5593unit to a file. The file name is made by appending @file{.tu} to the 5594source file name, and the file is created in the same directory as the 5595output file. If the @samp{-@var{options}} form is used, @var{options} 5596controls the details of the dump as described for the 5597@option{-fdump-tree} options. 5598 5599@item -fdump-class-hierarchy @r{(C++ only)} 5600@itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)} 5601@opindex fdump-class-hierarchy 5602Dump a representation of each class's hierarchy and virtual function 5603table layout to a file. The file name is made by appending 5604@file{.class} to the source file name, and the file is created in the 5605same directory as the output file. If the @samp{-@var{options}} form 5606is used, @var{options} controls the details of the dump as described 5607for the @option{-fdump-tree} options. 5608 5609@item -fdump-ipa-@var{switch} 5610@opindex fdump-ipa 5611Control the dumping at various stages of inter-procedural analysis 5612language tree to a file. The file name is generated by appending a 5613switch specific suffix to the source file name, and the file is created 5614in the same directory as the output file. The following dumps are 5615possible: 5616 5617@table @samp 5618@item all 5619Enables all inter-procedural analysis dumps. 5620 5621@item cgraph 5622Dumps information about call-graph optimization, unused function removal, 5623and inlining decisions. 5624 5625@item inline 5626Dump after function inlining. 5627 5628@end table 5629 5630@item -fdump-passes 5631@opindex fdump-passes 5632Dump the list of optimization passes that are turned on and off by 5633the current command-line options. 5634 5635@item -fdump-statistics-@var{option} 5636@opindex fdump-statistics 5637Enable and control dumping of pass statistics in a separate file. The 5638file name is generated by appending a suffix ending in 5639@samp{.statistics} to the source file name, and the file is created in 5640the same directory as the output file. If the @samp{-@var{option}} 5641form is used, @samp{-stats} will cause counters to be summed over the 5642whole compilation unit while @samp{-details} will dump every event as 5643the passes generate them. The default with no option is to sum 5644counters for each function compiled. 5645 5646@item -fdump-tree-@var{switch} 5647@itemx -fdump-tree-@var{switch}-@var{options} 5648@opindex fdump-tree 5649Control the dumping at various stages of processing the intermediate 5650language tree to a file. The file name is generated by appending a 5651switch specific suffix to the source file name, and the file is 5652created in the same directory as the output file. If the 5653@samp{-@var{options}} form is used, @var{options} is a list of 5654@samp{-} separated options which control the details of the dump. Not 5655all options are applicable to all dumps; those that are not 5656meaningful will be ignored. The following options are available 5657 5658@table @samp 5659@item address 5660Print the address of each node. Usually this is not meaningful as it 5661changes according to the environment and source file. Its primary use 5662is for tying up a dump file with a debug environment. 5663@item asmname 5664If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that 5665in the dump instead of @code{DECL_NAME}. Its primary use is ease of 5666use working backward from mangled names in the assembly file. 5667@item slim 5668Inhibit dumping of members of a scope or body of a function merely 5669because that scope has been reached. Only dump such items when they 5670are directly reachable by some other path. When dumping pretty-printed 5671trees, this option inhibits dumping the bodies of control structures. 5672@item raw 5673Print a raw representation of the tree. By default, trees are 5674pretty-printed into a C-like representation. 5675@item details 5676Enable more detailed dumps (not honored by every dump option). 5677@item stats 5678Enable dumping various statistics about the pass (not honored by every dump 5679option). 5680@item blocks 5681Enable showing basic block boundaries (disabled in raw dumps). 5682@item vops 5683Enable showing virtual operands for every statement. 5684@item lineno 5685Enable showing line numbers for statements. 5686@item uid 5687Enable showing the unique ID (@code{DECL_UID}) for each variable. 5688@item verbose 5689Enable showing the tree dump for each statement. 5690@item eh 5691Enable showing the EH region number holding each statement. 5692@item scev 5693Enable showing scalar evolution analysis details. 5694@item all 5695Turn on all options, except @option{raw}, @option{slim}, @option{verbose} 5696and @option{lineno}. 5697@end table 5698 5699The following tree dumps are possible: 5700@table @samp 5701 5702@item original 5703@opindex fdump-tree-original 5704Dump before any tree based optimization, to @file{@var{file}.original}. 5705 5706@item optimized 5707@opindex fdump-tree-optimized 5708Dump after all tree based optimization, to @file{@var{file}.optimized}. 5709 5710@item gimple 5711@opindex fdump-tree-gimple 5712Dump each function before and after the gimplification pass to a file. The 5713file name is made by appending @file{.gimple} to the source file name. 5714 5715@item cfg 5716@opindex fdump-tree-cfg 5717Dump the control flow graph of each function to a file. The file name is 5718made by appending @file{.cfg} to the source file name. 5719 5720@item vcg 5721@opindex fdump-tree-vcg 5722Dump the control flow graph of each function to a file in VCG format. The 5723file name is made by appending @file{.vcg} to the source file name. Note 5724that if the file contains more than one function, the generated file cannot 5725be used directly by VCG@. You will need to cut and paste each function's 5726graph into its own separate file first. 5727 5728@item ch 5729@opindex fdump-tree-ch 5730Dump each function after copying loop headers. The file name is made by 5731appending @file{.ch} to the source file name. 5732 5733@item ssa 5734@opindex fdump-tree-ssa 5735Dump SSA related information to a file. The file name is made by appending 5736@file{.ssa} to the source file name. 5737 5738@item alias 5739@opindex fdump-tree-alias 5740Dump aliasing information for each function. The file name is made by 5741appending @file{.alias} to the source file name. 5742 5743@item ccp 5744@opindex fdump-tree-ccp 5745Dump each function after CCP@. The file name is made by appending 5746@file{.ccp} to the source file name. 5747 5748@item storeccp 5749@opindex fdump-tree-storeccp 5750Dump each function after STORE-CCP@. The file name is made by appending 5751@file{.storeccp} to the source file name. 5752 5753@item pre 5754@opindex fdump-tree-pre 5755Dump trees after partial redundancy elimination. The file name is made 5756by appending @file{.pre} to the source file name. 5757 5758@item fre 5759@opindex fdump-tree-fre 5760Dump trees after full redundancy elimination. The file name is made 5761by appending @file{.fre} to the source file name. 5762 5763@item copyprop 5764@opindex fdump-tree-copyprop 5765Dump trees after copy propagation. The file name is made 5766by appending @file{.copyprop} to the source file name. 5767 5768@item store_copyprop 5769@opindex fdump-tree-store_copyprop 5770Dump trees after store copy-propagation. The file name is made 5771by appending @file{.store_copyprop} to the source file name. 5772 5773@item dce 5774@opindex fdump-tree-dce 5775Dump each function after dead code elimination. The file name is made by 5776appending @file{.dce} to the source file name. 5777 5778@item mudflap 5779@opindex fdump-tree-mudflap 5780Dump each function after adding mudflap instrumentation. The file name is 5781made by appending @file{.mudflap} to the source file name. 5782 5783@item sra 5784@opindex fdump-tree-sra 5785Dump each function after performing scalar replacement of aggregates. The 5786file name is made by appending @file{.sra} to the source file name. 5787 5788@item sink 5789@opindex fdump-tree-sink 5790Dump each function after performing code sinking. The file name is made 5791by appending @file{.sink} to the source file name. 5792 5793@item dom 5794@opindex fdump-tree-dom 5795Dump each function after applying dominator tree optimizations. The file 5796name is made by appending @file{.dom} to the source file name. 5797 5798@item dse 5799@opindex fdump-tree-dse 5800Dump each function after applying dead store elimination. The file 5801name is made by appending @file{.dse} to the source file name. 5802 5803@item phiopt 5804@opindex fdump-tree-phiopt 5805Dump each function after optimizing PHI nodes into straightline code. The file 5806name is made by appending @file{.phiopt} to the source file name. 5807 5808@item forwprop 5809@opindex fdump-tree-forwprop 5810Dump each function after forward propagating single use variables. The file 5811name is made by appending @file{.forwprop} to the source file name. 5812 5813@item copyrename 5814@opindex fdump-tree-copyrename 5815Dump each function after applying the copy rename optimization. The file 5816name is made by appending @file{.copyrename} to the source file name. 5817 5818@item nrv 5819@opindex fdump-tree-nrv 5820Dump each function after applying the named return value optimization on 5821generic trees. The file name is made by appending @file{.nrv} to the source 5822file name. 5823 5824@item vect 5825@opindex fdump-tree-vect 5826Dump each function after applying vectorization of loops. The file name is 5827made by appending @file{.vect} to the source file name. 5828 5829@item slp 5830@opindex fdump-tree-slp 5831Dump each function after applying vectorization of basic blocks. The file name 5832is made by appending @file{.slp} to the source file name. 5833 5834@item vrp 5835@opindex fdump-tree-vrp 5836Dump each function after Value Range Propagation (VRP). The file name 5837is made by appending @file{.vrp} to the source file name. 5838 5839@item all 5840@opindex fdump-tree-all 5841Enable all the available tree dumps with the flags provided in this option. 5842@end table 5843 5844@item -ftree-vectorizer-verbose=@var{n} 5845@opindex ftree-vectorizer-verbose 5846This option controls the amount of debugging output the vectorizer prints. 5847This information is written to standard error, unless 5848@option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified, 5849in which case it is output to the usual dump listing file, @file{.vect}. 5850For @var{n}=0 no diagnostic information is reported. 5851If @var{n}=1 the vectorizer reports each loop that got vectorized, 5852and the total number of loops that got vectorized. 5853If @var{n}=2 the vectorizer also reports non-vectorized loops that passed 5854the first analysis phase (vect_analyze_loop_form) - i.e.@: countable, 5855inner-most, single-bb, single-entry/exit loops. This is the same verbosity 5856level that @option{-fdump-tree-vect-stats} uses. 5857Higher verbosity levels mean either more information dumped for each 5858reported loop, or same amount of information reported for more loops: 5859if @var{n}=3, vectorizer cost model information is reported. 5860If @var{n}=4, alignment related information is added to the reports. 5861If @var{n}=5, data-references related information (e.g.@: memory dependences, 5862memory access-patterns) is added to the reports. 5863If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops 5864that did not pass the first analysis phase (i.e., may not be countable, or 5865may have complicated control-flow). 5866If @var{n}=7, the vectorizer reports also non-vectorized nested loops. 5867If @var{n}=8, SLP related information is added to the reports. 5868For @var{n}=9, all the information the vectorizer generates during its 5869analysis and transformation is reported. This is the same verbosity level 5870that @option{-fdump-tree-vect-details} uses. 5871 5872@item -frandom-seed=@var{string} 5873@opindex frandom-seed 5874This option provides a seed that GCC uses when it would otherwise use 5875random numbers. It is used to generate certain symbol names 5876that have to be different in every compiled file. It is also used to 5877place unique stamps in coverage data files and the object files that 5878produce them. You can use the @option{-frandom-seed} option to produce 5879reproducibly identical object files. 5880 5881The @var{string} should be different for every file you compile. 5882 5883@item -fsched-verbose=@var{n} 5884@opindex fsched-verbose 5885On targets that use instruction scheduling, this option controls the 5886amount of debugging output the scheduler prints. This information is 5887written to standard error, unless @option{-fdump-rtl-sched1} or 5888@option{-fdump-rtl-sched2} is specified, in which case it is output 5889to the usual dump listing file, @file{.sched1} or @file{.sched2} 5890respectively. However for @var{n} greater than nine, the output is 5891always printed to standard error. 5892 5893For @var{n} greater than zero, @option{-fsched-verbose} outputs the 5894same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}. 5895For @var{n} greater than one, it also output basic block probabilities, 5896detailed ready list information and unit/insn info. For @var{n} greater 5897than two, it includes RTL at abort point, control-flow and regions info. 5898And for @var{n} over four, @option{-fsched-verbose} also includes 5899dependence info. 5900 5901@item -save-temps 5902@itemx -save-temps=cwd 5903@opindex save-temps 5904Store the usual ``temporary'' intermediate files permanently; place them 5905in the current directory and name them based on the source file. Thus, 5906compiling @file{foo.c} with @samp{-c -save-temps} would produce files 5907@file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a 5908preprocessed @file{foo.i} output file even though the compiler now 5909normally uses an integrated preprocessor. 5910 5911When used in combination with the @option{-x} command-line option, 5912@option{-save-temps} is sensible enough to avoid over writing an 5913input source file with the same extension as an intermediate file. 5914The corresponding intermediate file may be obtained by renaming the 5915source file before using @option{-save-temps}. 5916 5917If you invoke GCC in parallel, compiling several different source 5918files that share a common base name in different subdirectories or the 5919same source file compiled for multiple output destinations, it is 5920likely that the different parallel compilers will interfere with each 5921other, and overwrite the temporary files. For instance: 5922 5923@smallexample 5924gcc -save-temps -o outdir1/foo.o indir1/foo.c& 5925gcc -save-temps -o outdir2/foo.o indir2/foo.c& 5926@end smallexample 5927 5928may result in @file{foo.i} and @file{foo.o} being written to 5929simultaneously by both compilers. 5930 5931@item -save-temps=obj 5932@opindex save-temps=obj 5933Store the usual ``temporary'' intermediate files permanently. If the 5934@option{-o} option is used, the temporary files are based on the 5935object file. If the @option{-o} option is not used, the 5936@option{-save-temps=obj} switch behaves like @option{-save-temps}. 5937 5938For example: 5939 5940@smallexample 5941gcc -save-temps=obj -c foo.c 5942gcc -save-temps=obj -c bar.c -o dir/xbar.o 5943gcc -save-temps=obj foobar.c -o dir2/yfoobar 5944@end smallexample 5945 5946would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i}, 5947@file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and 5948@file{dir2/yfoobar.o}. 5949 5950@item -time@r{[}=@var{file}@r{]} 5951@opindex time 5952Report the CPU time taken by each subprocess in the compilation 5953sequence. For C source files, this is the compiler proper and assembler 5954(plus the linker if linking is done). 5955 5956Without the specification of an output file, the output looks like this: 5957 5958@smallexample 5959# cc1 0.12 0.01 5960# as 0.00 0.01 5961@end smallexample 5962 5963The first number on each line is the ``user time'', that is time spent 5964executing the program itself. The second number is ``system time'', 5965time spent executing operating system routines on behalf of the program. 5966Both numbers are in seconds. 5967 5968With the specification of an output file, the output is appended to the 5969named file, and it looks like this: 5970 5971@smallexample 59720.12 0.01 cc1 @var{options} 59730.00 0.01 as @var{options} 5974@end smallexample 5975 5976The ``user time'' and the ``system time'' are moved before the program 5977name, and the options passed to the program are displayed, so that one 5978can later tell what file was being compiled, and with which options. 5979 5980@item -fvar-tracking 5981@opindex fvar-tracking 5982Run variable tracking pass. It computes where variables are stored at each 5983position in code. Better debugging information is then generated 5984(if the debugging information format supports this information). 5985 5986It is enabled by default when compiling with optimization (@option{-Os}, 5987@option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and 5988the debug info format supports it. 5989 5990@item -fvar-tracking-assignments 5991@opindex fvar-tracking-assignments 5992@opindex fno-var-tracking-assignments 5993Annotate assignments to user variables early in the compilation and 5994attempt to carry the annotations over throughout the compilation all the 5995way to the end, in an attempt to improve debug information while 5996optimizing. Use of @option{-gdwarf-4} is recommended along with it. 5997 5998It can be enabled even if var-tracking is disabled, in which case 5999annotations will be created and maintained, but discarded at the end. 6000 6001@item -fvar-tracking-assignments-toggle 6002@opindex fvar-tracking-assignments-toggle 6003@opindex fno-var-tracking-assignments-toggle 6004Toggle @option{-fvar-tracking-assignments}, in the same way that 6005@option{-gtoggle} toggles @option{-g}. 6006 6007@item -print-file-name=@var{library} 6008@opindex print-file-name 6009Print the full absolute name of the library file @var{library} that 6010would be used when linking---and don't do anything else. With this 6011option, GCC does not compile or link anything; it just prints the 6012file name. 6013 6014@item -print-multi-directory 6015@opindex print-multi-directory 6016Print the directory name corresponding to the multilib selected by any 6017other switches present in the command line. This directory is supposed 6018to exist in @env{GCC_EXEC_PREFIX}. 6019 6020@item -print-multi-lib 6021@opindex print-multi-lib 6022Print the mapping from multilib directory names to compiler switches 6023that enable them. The directory name is separated from the switches by 6024@samp{;}, and each switch starts with an @samp{@@} instead of the 6025@samp{-}, without spaces between multiple switches. This is supposed to 6026ease shell-processing. 6027 6028@item -print-multi-os-directory 6029@opindex print-multi-os-directory 6030Print the path to OS libraries for the selected 6031multilib, relative to some @file{lib} subdirectory. If OS libraries are 6032present in the @file{lib} subdirectory and no multilibs are used, this is 6033usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}} 6034sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or 6035@file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}} 6036subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}. 6037 6038@item -print-prog-name=@var{program} 6039@opindex print-prog-name 6040Like @option{-print-file-name}, but searches for a program such as @samp{cpp}. 6041 6042@item -print-libgcc-file-name 6043@opindex print-libgcc-file-name 6044Same as @option{-print-file-name=libgcc.a}. 6045 6046This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs} 6047but you do want to link with @file{libgcc.a}. You can do 6048 6049@smallexample 6050gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name` 6051@end smallexample 6052 6053@item -print-search-dirs 6054@opindex print-search-dirs 6055Print the name of the configured installation directory and a list of 6056program and library directories @command{gcc} will search---and don't do anything else. 6057 6058This is useful when @command{gcc} prints the error message 6059@samp{installation problem, cannot exec cpp0: No such file or directory}. 6060To resolve this you either need to put @file{cpp0} and the other compiler 6061components where @command{gcc} expects to find them, or you can set the environment 6062variable @env{GCC_EXEC_PREFIX} to the directory where you installed them. 6063Don't forget the trailing @samp{/}. 6064@xref{Environment Variables}. 6065 6066@item -print-sysroot 6067@opindex print-sysroot 6068Print the target sysroot directory that will be used during 6069compilation. This is the target sysroot specified either at configure 6070time or using the @option{--sysroot} option, possibly with an extra 6071suffix that depends on compilation options. If no target sysroot is 6072specified, the option prints nothing. 6073 6074@item -print-sysroot-headers-suffix 6075@opindex print-sysroot-headers-suffix 6076Print the suffix added to the target sysroot when searching for 6077headers, or give an error if the compiler is not configured with such 6078a suffix---and don't do anything else. 6079 6080@item -dumpmachine 6081@opindex dumpmachine 6082Print the compiler's target machine (for example, 6083@samp{i686-pc-linux-gnu})---and don't do anything else. 6084 6085@item -dumpversion 6086@opindex dumpversion 6087Print the compiler version (for example, @samp{3.0})---and don't do 6088anything else. 6089 6090@item -dumpspecs 6091@opindex dumpspecs 6092Print the compiler's built-in specs---and don't do anything else. (This 6093is used when GCC itself is being built.) @xref{Spec Files}. 6094 6095@item -feliminate-unused-debug-types 6096@opindex feliminate-unused-debug-types 6097Normally, when producing DWARF2 output, GCC will emit debugging 6098information for all types declared in a compilation 6099unit, regardless of whether or not they are actually used 6100in that compilation unit. Sometimes this is useful, such as 6101if, in the debugger, you want to cast a value to a type that is 6102not actually used in your program (but is declared). More often, 6103however, this results in a significant amount of wasted space. 6104With this option, GCC will avoid producing debug symbol output 6105for types that are nowhere used in the source file being compiled. 6106@end table 6107 6108@node Optimize Options 6109@section Options That Control Optimization 6110@cindex optimize options 6111@cindex options, optimization 6112 6113These options control various sorts of optimizations. 6114 6115Without any optimization option, the compiler's goal is to reduce the 6116cost of compilation and to make debugging produce the expected 6117results. Statements are independent: if you stop the program with a 6118breakpoint between statements, you can then assign a new value to any 6119variable or change the program counter to any other statement in the 6120function and get exactly the results you would expect from the source 6121code. 6122 6123Turning on optimization flags makes the compiler attempt to improve 6124the performance and/or code size at the expense of compilation time 6125and possibly the ability to debug the program. 6126 6127The compiler performs optimization based on the knowledge it has of the 6128program. Compiling multiple files at once to a single output file mode allows 6129the compiler to use information gained from all of the files when compiling 6130each of them. 6131 6132Not all optimizations are controlled directly by a flag. Only 6133optimizations that have a flag are listed in this section. 6134 6135Most optimizations are only enabled if an @option{-O} level is set on 6136the command line. Otherwise they are disabled, even if individual 6137optimization flags are specified. 6138 6139Depending on the target and how GCC was configured, a slightly different 6140set of optimizations may be enabled at each @option{-O} level than 6141those listed here. You can invoke GCC with @samp{-Q --help=optimizers} 6142to find out the exact set of optimizations that are enabled at each level. 6143@xref{Overall Options}, for examples. 6144 6145@table @gcctabopt 6146@item -O 6147@itemx -O1 6148@opindex O 6149@opindex O1 6150Optimize. Optimizing compilation takes somewhat more time, and a lot 6151more memory for a large function. 6152 6153With @option{-O}, the compiler tries to reduce code size and execution 6154time, without performing any optimizations that take a great deal of 6155compilation time. 6156 6157@option{-O} turns on the following optimization flags: 6158@gccoptlist{ 6159-fauto-inc-dec @gol 6160-fcompare-elim @gol 6161-fcprop-registers @gol 6162-fdce @gol 6163-fdefer-pop @gol 6164-fdelayed-branch @gol 6165-fdse @gol 6166-fguess-branch-probability @gol 6167-fif-conversion2 @gol 6168-fif-conversion @gol 6169-fipa-pure-const @gol 6170-fipa-profile @gol 6171-fipa-reference @gol 6172-fmerge-constants 6173-fsplit-wide-types @gol 6174-ftree-bit-ccp @gol 6175-ftree-builtin-call-dce @gol 6176-ftree-ccp @gol 6177-ftree-ch @gol 6178-ftree-copyrename @gol 6179-ftree-dce @gol 6180-ftree-dominator-opts @gol 6181-ftree-dse @gol 6182-ftree-forwprop @gol 6183-ftree-fre @gol 6184-ftree-phiprop @gol 6185-ftree-sra @gol 6186-ftree-pta @gol 6187-ftree-ter @gol 6188-funit-at-a-time} 6189 6190@option{-O} also turns on @option{-fomit-frame-pointer} on machines 6191where doing so does not interfere with debugging. 6192 6193@item -O2 6194@opindex O2 6195Optimize even more. GCC performs nearly all supported optimizations 6196that do not involve a space-speed tradeoff. 6197As compared to @option{-O}, this option increases both compilation time 6198and the performance of the generated code. 6199 6200@option{-O2} turns on all optimization flags specified by @option{-O}. It 6201also turns on the following optimization flags: 6202@gccoptlist{-fthread-jumps @gol 6203-falign-functions -falign-jumps @gol 6204-falign-loops -falign-labels @gol 6205-fcaller-saves @gol 6206-fcrossjumping @gol 6207-fcse-follow-jumps -fcse-skip-blocks @gol 6208-fdelete-null-pointer-checks @gol 6209-fdevirtualize @gol 6210-fexpensive-optimizations @gol 6211-fgcse -fgcse-lm @gol 6212-finline-small-functions @gol 6213-findirect-inlining @gol 6214-fipa-sra @gol 6215-foptimize-sibling-calls @gol 6216-fpartial-inlining @gol 6217-fpeephole2 @gol 6218-fregmove @gol 6219-freorder-blocks -freorder-functions @gol 6220-frerun-cse-after-loop @gol 6221-fsched-interblock -fsched-spec @gol 6222-fschedule-insns -fschedule-insns2 @gol 6223-fstrict-aliasing -fstrict-overflow @gol 6224-ftree-switch-conversion -ftree-tail-merge @gol 6225-ftree-pre @gol 6226-ftree-vrp} 6227 6228Please note the warning under @option{-fgcse} about 6229invoking @option{-O2} on programs that use computed gotos. 6230 6231@item -O3 6232@opindex O3 6233Optimize yet more. @option{-O3} turns on all optimizations specified 6234by @option{-O2} and also turns on the @option{-finline-functions}, 6235@option{-funswitch-loops}, @option{-fpredictive-commoning}, 6236@option{-fgcse-after-reload}, @option{-ftree-vectorize} and 6237@option{-fipa-cp-clone} options. 6238 6239@item -O0 6240@opindex O0 6241Reduce compilation time and make debugging produce the expected 6242results. This is the default. 6243 6244@item -Os 6245@opindex Os 6246Optimize for size. @option{-Os} enables all @option{-O2} optimizations that 6247do not typically increase code size. It also performs further 6248optimizations designed to reduce code size. 6249 6250@option{-Os} disables the following optimization flags: 6251@gccoptlist{-falign-functions -falign-jumps -falign-loops @gol 6252-falign-labels -freorder-blocks -freorder-blocks-and-partition @gol 6253-fprefetch-loop-arrays -ftree-vect-loop-version} 6254 6255@item -Ofast 6256@opindex Ofast 6257Disregard strict standards compliance. @option{-Ofast} enables all 6258@option{-O3} optimizations. It also enables optimizations that are not 6259valid for all standard compliant programs. 6260It turns on @option{-ffast-math} and the Fortran-specific 6261@option{-fno-protect-parens} and @option{-fstack-arrays}. 6262 6263If you use multiple @option{-O} options, with or without level numbers, 6264the last such option is the one that is effective. 6265@end table 6266 6267Options of the form @option{-f@var{flag}} specify machine-independent 6268flags. Most flags have both positive and negative forms; the negative 6269form of @option{-ffoo} would be @option{-fno-foo}. In the table 6270below, only one of the forms is listed---the one you typically will 6271use. You can figure out the other form by either removing @samp{no-} 6272or adding it. 6273 6274The following options control specific optimizations. They are either 6275activated by @option{-O} options or are related to ones that are. You 6276can use the following flags in the rare cases when ``fine-tuning'' of 6277optimizations to be performed is desired. 6278 6279@table @gcctabopt 6280@item -fno-default-inline 6281@opindex fno-default-inline 6282Do not make member functions inline by default merely because they are 6283defined inside the class scope (C++ only). Otherwise, when you specify 6284@w{@option{-O}}, member functions defined inside class scope are compiled 6285inline by default; i.e., you don't need to add @samp{inline} in front of 6286the member function name. 6287 6288@item -fno-defer-pop 6289@opindex fno-defer-pop 6290Always pop the arguments to each function call as soon as that function 6291returns. For machines that must pop arguments after a function call, 6292the compiler normally lets arguments accumulate on the stack for several 6293function calls and pops them all at once. 6294 6295Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 6296 6297@item -fforward-propagate 6298@opindex fforward-propagate 6299Perform a forward propagation pass on RTL@. The pass tries to combine two 6300instructions and checks if the result can be simplified. If loop unrolling 6301is active, two passes are performed and the second is scheduled after 6302loop unrolling. 6303 6304This option is enabled by default at optimization levels @option{-O}, 6305@option{-O2}, @option{-O3}, @option{-Os}. 6306 6307@item -ffp-contract=@var{style} 6308@opindex ffp-contract 6309@option{-ffp-contract=off} disables floating-point expression contraction. 6310@option{-ffp-contract=fast} enables floating-point expression contraction 6311such as forming of fused multiply-add operations if the target has 6312native support for them. 6313@option{-ffp-contract=on} enables floating-point expression contraction 6314if allowed by the language standard. This is currently not implemented 6315and treated equal to @option{-ffp-contract=off}. 6316 6317The default is @option{-ffp-contract=fast}. 6318 6319@item -fomit-frame-pointer 6320@opindex fomit-frame-pointer 6321Don't keep the frame pointer in a register for functions that 6322don't need one. This avoids the instructions to save, set up and 6323restore frame pointers; it also makes an extra register available 6324in many functions. @strong{It also makes debugging impossible on 6325some machines.} 6326 6327On some machines, such as the VAX, this flag has no effect, because 6328the standard calling sequence automatically handles the frame pointer 6329and nothing is saved by pretending it doesn't exist. The 6330machine-description macro @code{FRAME_POINTER_REQUIRED} controls 6331whether a target machine supports this flag. @xref{Registers,,Register 6332Usage, gccint, GNU Compiler Collection (GCC) Internals}. 6333 6334Starting with GCC version 4.6, the default setting (when not optimizing for 6335size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to 6336@option{-fomit-frame-pointer}. The default can be reverted to 6337@option{-fno-omit-frame-pointer} by configuring GCC with the 6338@option{--enable-frame-pointer} configure option. 6339 6340Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 6341 6342@item -foptimize-sibling-calls 6343@opindex foptimize-sibling-calls 6344Optimize sibling and tail recursive calls. 6345 6346Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6347 6348@item -fno-inline 6349@opindex fno-inline 6350Do not expand any functions inline apart from those marked with 6351the @code{always_inline} attribute. This is the default when not 6352optimizing. 6353 6354Single functions can be exempted from inlining by marking them 6355with the @code{noinline} attribute. 6356 6357@item -finline-small-functions 6358@opindex finline-small-functions 6359Integrate functions into their callers when their body is smaller than expected 6360function call code (so overall size of program gets smaller). The compiler 6361heuristically decides which functions are simple enough to be worth integrating 6362in this way. This inlining applies to all functions, even those not declared 6363inline. 6364 6365Enabled at level @option{-O2}. 6366 6367@item -findirect-inlining 6368@opindex findirect-inlining 6369Inline also indirect calls that are discovered to be known at compile 6370time thanks to previous inlining. This option has any effect only 6371when inlining itself is turned on by the @option{-finline-functions} 6372or @option{-finline-small-functions} options. 6373 6374Enabled at level @option{-O2}. 6375 6376@item -finline-functions 6377@opindex finline-functions 6378Consider all functions for inlining, even if they are not declared inline. 6379The compiler heuristically decides which functions are worth integrating 6380in this way. 6381 6382If all calls to a given function are integrated, and the function is 6383declared @code{static}, then the function is normally not output as 6384assembler code in its own right. 6385 6386Enabled at level @option{-O3}. 6387 6388@item -finline-functions-called-once 6389@opindex finline-functions-called-once 6390Consider all @code{static} functions called once for inlining into their 6391caller even if they are not marked @code{inline}. If a call to a given 6392function is integrated, then the function is not output as assembler code 6393in its own right. 6394 6395Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}. 6396 6397@item -fearly-inlining 6398@opindex fearly-inlining 6399Inline functions marked by @code{always_inline} and functions whose body seems 6400smaller than the function call overhead early before doing 6401@option{-fprofile-generate} instrumentation and real inlining pass. Doing so 6402makes profiling significantly cheaper and usually inlining faster on programs 6403having large chains of nested wrapper functions. 6404 6405Enabled by default. 6406 6407@item -fipa-sra 6408@opindex fipa-sra 6409Perform interprocedural scalar replacement of aggregates, removal of 6410unused parameters and replacement of parameters passed by reference 6411by parameters passed by value. 6412 6413Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}. 6414 6415@item -finline-limit=@var{n} 6416@opindex finline-limit 6417By default, GCC limits the size of functions that can be inlined. This flag 6418allows coarse control of this limit. @var{n} is the size of functions that 6419can be inlined in number of pseudo instructions. 6420 6421Inlining is actually controlled by a number of parameters, which may be 6422specified individually by using @option{--param @var{name}=@var{value}}. 6423The @option{-finline-limit=@var{n}} option sets some of these parameters 6424as follows: 6425 6426@table @gcctabopt 6427@item max-inline-insns-single 6428is set to @var{n}/2. 6429@item max-inline-insns-auto 6430is set to @var{n}/2. 6431@end table 6432 6433See below for a documentation of the individual 6434parameters controlling inlining and for the defaults of these parameters. 6435 6436@emph{Note:} there may be no value to @option{-finline-limit} that results 6437in default behavior. 6438 6439@emph{Note:} pseudo instruction represents, in this particular context, an 6440abstract measurement of function's size. In no way does it represent a count 6441of assembly instructions and as such its exact meaning might change from one 6442release to an another. 6443 6444@item -fno-keep-inline-dllexport 6445@opindex -fno-keep-inline-dllexport 6446This is a more fine-grained version of @option{-fkeep-inline-functions}, 6447which applies only to functions that are declared using the @code{dllexport} 6448attribute or declspec (@xref{Function Attributes,,Declaring Attributes of 6449Functions}.) 6450 6451@item -fkeep-inline-functions 6452@opindex fkeep-inline-functions 6453In C, emit @code{static} functions that are declared @code{inline} 6454into the object file, even if the function has been inlined into all 6455of its callers. This switch does not affect functions using the 6456@code{extern inline} extension in GNU C90@. In C++, emit any and all 6457inline functions into the object file. 6458 6459@item -fkeep-static-consts 6460@opindex fkeep-static-consts 6461Emit variables declared @code{static const} when optimization isn't turned 6462on, even if the variables aren't referenced. 6463 6464GCC enables this option by default. If you want to force the compiler to 6465check if the variable was referenced, regardless of whether or not 6466optimization is turned on, use the @option{-fno-keep-static-consts} option. 6467 6468@item -fmerge-constants 6469@opindex fmerge-constants 6470Attempt to merge identical constants (string constants and floating-point 6471constants) across compilation units. 6472 6473This option is the default for optimized compilation if the assembler and 6474linker support it. Use @option{-fno-merge-constants} to inhibit this 6475behavior. 6476 6477Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 6478 6479@item -fmerge-all-constants 6480@opindex fmerge-all-constants 6481Attempt to merge identical constants and identical variables. 6482 6483This option implies @option{-fmerge-constants}. In addition to 6484@option{-fmerge-constants} this considers e.g.@: even constant initialized 6485arrays or initialized constant variables with integral or floating-point 6486types. Languages like C or C++ require each variable, including multiple 6487instances of the same variable in recursive calls, to have distinct locations, 6488so using this option will result in non-conforming 6489behavior. 6490 6491@item -fmodulo-sched 6492@opindex fmodulo-sched 6493Perform swing modulo scheduling immediately before the first scheduling 6494pass. This pass looks at innermost loops and reorders their 6495instructions by overlapping different iterations. 6496 6497@item -fmodulo-sched-allow-regmoves 6498@opindex fmodulo-sched-allow-regmoves 6499Perform more aggressive SMS based modulo scheduling with register moves 6500allowed. By setting this flag certain anti-dependences edges will be 6501deleted which will trigger the generation of reg-moves based on the 6502life-range analysis. This option is effective only with 6503@option{-fmodulo-sched} enabled. 6504 6505@item -fno-branch-count-reg 6506@opindex fno-branch-count-reg 6507Do not use ``decrement and branch'' instructions on a count register, 6508but instead generate a sequence of instructions that decrement a 6509register, compare it against zero, then branch based upon the result. 6510This option is only meaningful on architectures that support such 6511instructions, which include x86, PowerPC, IA-64 and S/390. 6512 6513The default is @option{-fbranch-count-reg}. 6514 6515@item -fno-function-cse 6516@opindex fno-function-cse 6517Do not put function addresses in registers; make each instruction that 6518calls a constant function contain the function's address explicitly. 6519 6520This option results in less efficient code, but some strange hacks 6521that alter the assembler output may be confused by the optimizations 6522performed when this option is not used. 6523 6524The default is @option{-ffunction-cse} 6525 6526@item -fno-zero-initialized-in-bss 6527@opindex fno-zero-initialized-in-bss 6528If the target supports a BSS section, GCC by default puts variables that 6529are initialized to zero into BSS@. This can save space in the resulting 6530code. 6531 6532This option turns off this behavior because some programs explicitly 6533rely on variables going to the data section. E.g., so that the 6534resulting executable can find the beginning of that section and/or make 6535assumptions based on that. 6536 6537The default is @option{-fzero-initialized-in-bss}. 6538 6539@item -fmudflap -fmudflapth -fmudflapir 6540@opindex fmudflap 6541@opindex fmudflapth 6542@opindex fmudflapir 6543@cindex bounds checking 6544@cindex mudflap 6545For front-ends that support it (C and C++), instrument all risky 6546pointer/array dereferencing operations, some standard library 6547string/heap functions, and some other associated constructs with 6548range/validity tests. Modules so instrumented should be immune to 6549buffer overflows, invalid heap use, and some other classes of C/C++ 6550programming errors. The instrumentation relies on a separate runtime 6551library (@file{libmudflap}), which will be linked into a program if 6552@option{-fmudflap} is given at link time. Run-time behavior of the 6553instrumented program is controlled by the @env{MUDFLAP_OPTIONS} 6554environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out} 6555for its options. 6556 6557Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to 6558link if your program is multi-threaded. Use @option{-fmudflapir}, in 6559addition to @option{-fmudflap} or @option{-fmudflapth}, if 6560instrumentation should ignore pointer reads. This produces less 6561instrumentation (and therefore faster execution) and still provides 6562some protection against outright memory corrupting writes, but allows 6563erroneously read data to propagate within a program. 6564 6565@item -fthread-jumps 6566@opindex fthread-jumps 6567Perform optimizations where we check to see if a jump branches to a 6568location where another comparison subsumed by the first is found. If 6569so, the first branch is redirected to either the destination of the 6570second branch or a point immediately following it, depending on whether 6571the condition is known to be true or false. 6572 6573Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6574 6575@item -fsplit-wide-types 6576@opindex fsplit-wide-types 6577When using a type that occupies multiple registers, such as @code{long 6578long} on a 32-bit system, split the registers apart and allocate them 6579independently. This normally generates better code for those types, 6580but may make debugging more difficult. 6581 6582Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, 6583@option{-Os}. 6584 6585@item -fcse-follow-jumps 6586@opindex fcse-follow-jumps 6587In common subexpression elimination (CSE), scan through jump instructions 6588when the target of the jump is not reached by any other path. For 6589example, when CSE encounters an @code{if} statement with an 6590@code{else} clause, CSE will follow the jump when the condition 6591tested is false. 6592 6593Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6594 6595@item -fcse-skip-blocks 6596@opindex fcse-skip-blocks 6597This is similar to @option{-fcse-follow-jumps}, but causes CSE to 6598follow jumps that conditionally skip over blocks. When CSE 6599encounters a simple @code{if} statement with no else clause, 6600@option{-fcse-skip-blocks} causes CSE to follow the jump around the 6601body of the @code{if}. 6602 6603Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6604 6605@item -frerun-cse-after-loop 6606@opindex frerun-cse-after-loop 6607Re-run common subexpression elimination after loop optimizations has been 6608performed. 6609 6610Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6611 6612@item -fgcse 6613@opindex fgcse 6614Perform a global common subexpression elimination pass. 6615This pass also performs global constant and copy propagation. 6616 6617@emph{Note:} When compiling a program using computed gotos, a GCC 6618extension, you may get better run-time performance if you disable 6619the global common subexpression elimination pass by adding 6620@option{-fno-gcse} to the command line. 6621 6622Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6623 6624@item -fgcse-lm 6625@opindex fgcse-lm 6626When @option{-fgcse-lm} is enabled, global common subexpression elimination will 6627attempt to move loads that are only killed by stores into themselves. This 6628allows a loop containing a load/store sequence to be changed to a load outside 6629the loop, and a copy/store within the loop. 6630 6631Enabled by default when gcse is enabled. 6632 6633@item -fgcse-sm 6634@opindex fgcse-sm 6635When @option{-fgcse-sm} is enabled, a store motion pass is run after 6636global common subexpression elimination. This pass will attempt to move 6637stores out of loops. When used in conjunction with @option{-fgcse-lm}, 6638loops containing a load/store sequence can be changed to a load before 6639the loop and a store after the loop. 6640 6641Not enabled at any optimization level. 6642 6643@item -fgcse-las 6644@opindex fgcse-las 6645When @option{-fgcse-las} is enabled, the global common subexpression 6646elimination pass eliminates redundant loads that come after stores to the 6647same memory location (both partial and full redundancies). 6648 6649Not enabled at any optimization level. 6650 6651@item -fgcse-after-reload 6652@opindex fgcse-after-reload 6653When @option{-fgcse-after-reload} is enabled, a redundant load elimination 6654pass is performed after reload. The purpose of this pass is to cleanup 6655redundant spilling. 6656 6657@item -funsafe-loop-optimizations 6658@opindex funsafe-loop-optimizations 6659If given, the loop optimizer will assume that loop indices do not 6660overflow, and that the loops with nontrivial exit condition are not 6661infinite. This enables a wider range of loop optimizations even if 6662the loop optimizer itself cannot prove that these assumptions are valid. 6663Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you 6664if it finds this kind of loop. 6665 6666@item -fcrossjumping 6667@opindex fcrossjumping 6668Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The 6669resulting code may or may not perform better than without cross-jumping. 6670 6671Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6672 6673@item -fauto-inc-dec 6674@opindex fauto-inc-dec 6675Combine increments or decrements of addresses with memory accesses. 6676This pass is always skipped on architectures that do not have 6677instructions to support this. Enabled by default at @option{-O} and 6678higher on architectures that support this. 6679 6680@item -fdce 6681@opindex fdce 6682Perform dead code elimination (DCE) on RTL@. 6683Enabled by default at @option{-O} and higher. 6684 6685@item -fdse 6686@opindex fdse 6687Perform dead store elimination (DSE) on RTL@. 6688Enabled by default at @option{-O} and higher. 6689 6690@item -fif-conversion 6691@opindex fif-conversion 6692Attempt to transform conditional jumps into branch-less equivalents. This 6693include use of conditional moves, min, max, set flags and abs instructions, and 6694some tricks doable by standard arithmetics. The use of conditional execution 6695on chips where it is available is controlled by @code{if-conversion2}. 6696 6697Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 6698 6699@item -fif-conversion2 6700@opindex fif-conversion2 6701Use conditional execution (where available) to transform conditional jumps into 6702branch-less equivalents. 6703 6704Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 6705 6706@item -fdelete-null-pointer-checks 6707@opindex fdelete-null-pointer-checks 6708Assume that programs cannot safely dereference null pointers, and that 6709no code or data element resides there. This enables simple constant 6710folding optimizations at all optimization levels. In addition, other 6711optimization passes in GCC use this flag to control global dataflow 6712analyses that eliminate useless checks for null pointers; these assume 6713that if a pointer is checked after it has already been dereferenced, 6714it cannot be null. 6715 6716Note however that in some environments this assumption is not true. 6717Use @option{-fno-delete-null-pointer-checks} to disable this optimization 6718for programs that depend on that behavior. 6719 6720Some targets, especially embedded ones, disable this option at all levels. 6721Otherwise it is enabled at all levels: @option{-O0}, @option{-O1}, 6722@option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information 6723are enabled independently at different optimization levels. 6724 6725@item -fdevirtualize 6726@opindex fdevirtualize 6727Attempt to convert calls to virtual functions to direct calls. This 6728is done both within a procedure and interprocedurally as part of 6729indirect inlining (@code{-findirect-inlining}) and interprocedural constant 6730propagation (@option{-fipa-cp}). 6731Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6732 6733@item -fexpensive-optimizations 6734@opindex fexpensive-optimizations 6735Perform a number of minor optimizations that are relatively expensive. 6736 6737Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6738 6739@item -free 6740@opindex free 6741Attempt to remove redundant extension instructions. This is especially 6742helpful for the x86-64 architecture which implicitly zero-extends in 64-bit 6743registers after writing to their lower 32-bit half. 6744 6745Enabled for x86 at levels @option{-O2}, @option{-O3}. 6746 6747@item -foptimize-register-move 6748@itemx -fregmove 6749@opindex foptimize-register-move 6750@opindex fregmove 6751Attempt to reassign register numbers in move instructions and as 6752operands of other simple instructions in order to maximize the amount of 6753register tying. This is especially helpful on machines with two-operand 6754instructions. 6755 6756Note @option{-fregmove} and @option{-foptimize-register-move} are the same 6757optimization. 6758 6759Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6760 6761@item -fira-algorithm=@var{algorithm} 6762Use the specified coloring algorithm for the integrated register 6763allocator. The @var{algorithm} argument can be @samp{priority}, which 6764specifies Chow's priority coloring, or @samp{CB}, which specifies 6765Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented 6766for all architectures, but for those targets that do support it, it is 6767the default because it generates better code. 6768 6769@item -fira-region=@var{region} 6770Use specified regions for the integrated register allocator. The 6771@var{region} argument should be one of the following: 6772 6773@table @samp 6774 6775@item all 6776Use all loops as register allocation regions. 6777This can give the best results for machines with a small and/or 6778irregular register set. 6779 6780@item mixed 6781Use all loops except for loops with small register pressure 6782as the regions. This value usually gives 6783the best results in most cases and for most architectures, 6784and is enabled by default when compiling with optimization for speed 6785(@option{-O}, @option{-O2}, @dots{}). 6786 6787@item one 6788Use all functions as a single region. 6789This typically results in the smallest code size, and is enabled by default for 6790@option{-Os} or @option{-O0}. 6791 6792@end table 6793 6794@item -fira-loop-pressure 6795@opindex fira-loop-pressure 6796Use IRA to evaluate register pressure in loops for decisions to move 6797loop invariants. This option usually results in generation 6798of faster and smaller code on machines with large register files (>= 32 6799registers), but it can slow the compiler down. 6800 6801This option is enabled at level @option{-O3} for some targets. 6802 6803@item -fno-ira-share-save-slots 6804@opindex fno-ira-share-save-slots 6805Disable sharing of stack slots used for saving call-used hard 6806registers living through a call. Each hard register gets a 6807separate stack slot, and as a result function stack frames are 6808larger. 6809 6810@item -fno-ira-share-spill-slots 6811@opindex fno-ira-share-spill-slots 6812Disable sharing of stack slots allocated for pseudo-registers. Each 6813pseudo-register that does not get a hard register gets a separate 6814stack slot, and as a result function stack frames are larger. 6815 6816@item -fira-verbose=@var{n} 6817@opindex fira-verbose 6818Control the verbosity of the dump file for the integrated register allocator. 6819The default value is 5. If the value @var{n} is greater or equal to 10, 6820the dump output is sent to stderr using the same format as @var{n} minus 10. 6821 6822@item -fdelayed-branch 6823@opindex fdelayed-branch 6824If supported for the target machine, attempt to reorder instructions 6825to exploit instruction slots available after delayed branch 6826instructions. 6827 6828Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 6829 6830@item -fschedule-insns 6831@opindex fschedule-insns 6832If supported for the target machine, attempt to reorder instructions to 6833eliminate execution stalls due to required data being unavailable. This 6834helps machines that have slow floating point or memory load instructions 6835by allowing other instructions to be issued until the result of the load 6836or floating-point instruction is required. 6837 6838Enabled at levels @option{-O2}, @option{-O3}. 6839 6840@item -fschedule-insns2 6841@opindex fschedule-insns2 6842Similar to @option{-fschedule-insns}, but requests an additional pass of 6843instruction scheduling after register allocation has been done. This is 6844especially useful on machines with a relatively small number of 6845registers and where memory load instructions take more than one cycle. 6846 6847Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6848 6849@item -fno-sched-interblock 6850@opindex fno-sched-interblock 6851Don't schedule instructions across basic blocks. This is normally 6852enabled by default when scheduling before register allocation, i.e.@: 6853with @option{-fschedule-insns} or at @option{-O2} or higher. 6854 6855@item -fno-sched-spec 6856@opindex fno-sched-spec 6857Don't allow speculative motion of non-load instructions. This is normally 6858enabled by default when scheduling before register allocation, i.e.@: 6859with @option{-fschedule-insns} or at @option{-O2} or higher. 6860 6861@item -fsched-pressure 6862@opindex fsched-pressure 6863Enable register pressure sensitive insn scheduling before the register 6864allocation. This only makes sense when scheduling before register 6865allocation is enabled, i.e.@: with @option{-fschedule-insns} or at 6866@option{-O2} or higher. Usage of this option can improve the 6867generated code and decrease its size by preventing register pressure 6868increase above the number of available hard registers and as a 6869consequence register spills in the register allocation. 6870 6871@item -fsched-spec-load 6872@opindex fsched-spec-load 6873Allow speculative motion of some load instructions. This only makes 6874sense when scheduling before register allocation, i.e.@: with 6875@option{-fschedule-insns} or at @option{-O2} or higher. 6876 6877@item -fsched-spec-load-dangerous 6878@opindex fsched-spec-load-dangerous 6879Allow speculative motion of more load instructions. This only makes 6880sense when scheduling before register allocation, i.e.@: with 6881@option{-fschedule-insns} or at @option{-O2} or higher. 6882 6883@item -fsched-stalled-insns 6884@itemx -fsched-stalled-insns=@var{n} 6885@opindex fsched-stalled-insns 6886Define how many insns (if any) can be moved prematurely from the queue 6887of stalled insns into the ready list, during the second scheduling pass. 6888@option{-fno-sched-stalled-insns} means that no insns will be moved 6889prematurely, @option{-fsched-stalled-insns=0} means there is no limit 6890on how many queued insns can be moved prematurely. 6891@option{-fsched-stalled-insns} without a value is equivalent to 6892@option{-fsched-stalled-insns=1}. 6893 6894@item -fsched-stalled-insns-dep 6895@itemx -fsched-stalled-insns-dep=@var{n} 6896@opindex fsched-stalled-insns-dep 6897Define how many insn groups (cycles) will be examined for a dependency 6898on a stalled insn that is candidate for premature removal from the queue 6899of stalled insns. This has an effect only during the second scheduling pass, 6900and only if @option{-fsched-stalled-insns} is used. 6901@option{-fno-sched-stalled-insns-dep} is equivalent to 6902@option{-fsched-stalled-insns-dep=0}. 6903@option{-fsched-stalled-insns-dep} without a value is equivalent to 6904@option{-fsched-stalled-insns-dep=1}. 6905 6906@item -fsched2-use-superblocks 6907@opindex fsched2-use-superblocks 6908When scheduling after register allocation, do use superblock scheduling 6909algorithm. Superblock scheduling allows motion across basic block boundaries 6910resulting on faster schedules. This option is experimental, as not all machine 6911descriptions used by GCC model the CPU closely enough to avoid unreliable 6912results from the algorithm. 6913 6914This only makes sense when scheduling after register allocation, i.e.@: with 6915@option{-fschedule-insns2} or at @option{-O2} or higher. 6916 6917@item -fsched-group-heuristic 6918@opindex fsched-group-heuristic 6919Enable the group heuristic in the scheduler. This heuristic favors 6920the instruction that belongs to a schedule group. This is enabled 6921by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns} 6922or @option{-fschedule-insns2} or at @option{-O2} or higher. 6923 6924@item -fsched-critical-path-heuristic 6925@opindex fsched-critical-path-heuristic 6926Enable the critical-path heuristic in the scheduler. This heuristic favors 6927instructions on the critical path. This is enabled by default when 6928scheduling is enabled, i.e.@: with @option{-fschedule-insns} 6929or @option{-fschedule-insns2} or at @option{-O2} or higher. 6930 6931@item -fsched-spec-insn-heuristic 6932@opindex fsched-spec-insn-heuristic 6933Enable the speculative instruction heuristic in the scheduler. This 6934heuristic favors speculative instructions with greater dependency weakness. 6935This is enabled by default when scheduling is enabled, i.e.@: 6936with @option{-fschedule-insns} or @option{-fschedule-insns2} 6937or at @option{-O2} or higher. 6938 6939@item -fsched-rank-heuristic 6940@opindex fsched-rank-heuristic 6941Enable the rank heuristic in the scheduler. This heuristic favors 6942the instruction belonging to a basic block with greater size or frequency. 6943This is enabled by default when scheduling is enabled, i.e.@: 6944with @option{-fschedule-insns} or @option{-fschedule-insns2} or 6945at @option{-O2} or higher. 6946 6947@item -fsched-last-insn-heuristic 6948@opindex fsched-last-insn-heuristic 6949Enable the last-instruction heuristic in the scheduler. This heuristic 6950favors the instruction that is less dependent on the last instruction 6951scheduled. This is enabled by default when scheduling is enabled, 6952i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or 6953at @option{-O2} or higher. 6954 6955@item -fsched-dep-count-heuristic 6956@opindex fsched-dep-count-heuristic 6957Enable the dependent-count heuristic in the scheduler. This heuristic 6958favors the instruction that has more instructions depending on it. 6959This is enabled by default when scheduling is enabled, i.e.@: 6960with @option{-fschedule-insns} or @option{-fschedule-insns2} or 6961at @option{-O2} or higher. 6962 6963@item -freschedule-modulo-scheduled-loops 6964@opindex freschedule-modulo-scheduled-loops 6965The modulo scheduling comes before the traditional scheduling, if a loop 6966was modulo scheduled we may want to prevent the later scheduling passes 6967from changing its schedule, we use this option to control that. 6968 6969@item -fselective-scheduling 6970@opindex fselective-scheduling 6971Schedule instructions using selective scheduling algorithm. Selective 6972scheduling runs instead of the first scheduler pass. 6973 6974@item -fselective-scheduling2 6975@opindex fselective-scheduling2 6976Schedule instructions using selective scheduling algorithm. Selective 6977scheduling runs instead of the second scheduler pass. 6978 6979@item -fsel-sched-pipelining 6980@opindex fsel-sched-pipelining 6981Enable software pipelining of innermost loops during selective scheduling. 6982This option has no effect until one of @option{-fselective-scheduling} or 6983@option{-fselective-scheduling2} is turned on. 6984 6985@item -fsel-sched-pipelining-outer-loops 6986@opindex fsel-sched-pipelining-outer-loops 6987When pipelining loops during selective scheduling, also pipeline outer loops. 6988This option has no effect until @option{-fsel-sched-pipelining} is turned on. 6989 6990@item -fshrink-wrap 6991@opindex fshrink-wrap 6992Emit function prologues only before parts of the function that need it, 6993rather than at the top of the function. This flag is enabled by default at 6994@option{-O} and higher. 6995 6996@item -fcaller-saves 6997@opindex fcaller-saves 6998Enable values to be allocated in registers that will be clobbered by 6999function calls, by emitting extra instructions to save and restore the 7000registers around such calls. Such allocation is done only when it 7001seems to result in better code than would otherwise be produced. 7002 7003This option is always enabled by default on certain machines, usually 7004those which have no call-preserved registers to use instead. 7005 7006Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 7007 7008@item -fcombine-stack-adjustments 7009@opindex fcombine-stack-adjustments 7010Tracks stack adjustments (pushes and pops) and stack memory references 7011and then tries to find ways to combine them. 7012 7013Enabled by default at @option{-O1} and higher. 7014 7015@item -fconserve-stack 7016@opindex fconserve-stack 7017Attempt to minimize stack usage. The compiler will attempt to use less 7018stack space, even if that makes the program slower. This option 7019implies setting the @option{large-stack-frame} parameter to 100 7020and the @option{large-stack-frame-growth} parameter to 400. 7021 7022@item -ftree-reassoc 7023@opindex ftree-reassoc 7024Perform reassociation on trees. This flag is enabled by default 7025at @option{-O} and higher. 7026 7027@item -ftree-pre 7028@opindex ftree-pre 7029Perform partial redundancy elimination (PRE) on trees. This flag is 7030enabled by default at @option{-O2} and @option{-O3}. 7031 7032@item -ftree-forwprop 7033@opindex ftree-forwprop 7034Perform forward propagation on trees. This flag is enabled by default 7035at @option{-O} and higher. 7036 7037@item -ftree-fre 7038@opindex ftree-fre 7039Perform full redundancy elimination (FRE) on trees. The difference 7040between FRE and PRE is that FRE only considers expressions 7041that are computed on all paths leading to the redundant computation. 7042This analysis is faster than PRE, though it exposes fewer redundancies. 7043This flag is enabled by default at @option{-O} and higher. 7044 7045@item -ftree-phiprop 7046@opindex ftree-phiprop 7047Perform hoisting of loads from conditional pointers on trees. This 7048pass is enabled by default at @option{-O} and higher. 7049 7050@item -ftree-copy-prop 7051@opindex ftree-copy-prop 7052Perform copy propagation on trees. This pass eliminates unnecessary 7053copy operations. This flag is enabled by default at @option{-O} and 7054higher. 7055 7056@item -fipa-pure-const 7057@opindex fipa-pure-const 7058Discover which functions are pure or constant. 7059Enabled by default at @option{-O} and higher. 7060 7061@item -fipa-reference 7062@opindex fipa-reference 7063Discover which static variables do not escape cannot escape the 7064compilation unit. 7065Enabled by default at @option{-O} and higher. 7066 7067@item -fipa-pta 7068@opindex fipa-pta 7069Perform interprocedural pointer analysis and interprocedural modification 7070and reference analysis. This option can cause excessive memory and 7071compile-time usage on large compilation units. It is not enabled by 7072default at any optimization level. 7073 7074@item -fipa-profile 7075@opindex fipa-profile 7076Perform interprocedural profile propagation. The functions called only from 7077cold functions are marked as cold. Also functions executed once (such as 7078@code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold 7079functions and loop less parts of functions executed once are then optimized for 7080size. 7081Enabled by default at @option{-O} and higher. 7082 7083@item -fipa-cp 7084@opindex fipa-cp 7085Perform interprocedural constant propagation. 7086This optimization analyzes the program to determine when values passed 7087to functions are constants and then optimizes accordingly. 7088This optimization can substantially increase performance 7089if the application has constants passed to functions. 7090This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}. 7091 7092@item -fipa-cp-clone 7093@opindex fipa-cp-clone 7094Perform function cloning to make interprocedural constant propagation stronger. 7095When enabled, interprocedural constant propagation will perform function cloning 7096when externally visible function can be called with constant arguments. 7097Because this optimization can create multiple copies of functions, 7098it may significantly increase code size 7099(see @option{--param ipcp-unit-growth=@var{value}}). 7100This flag is enabled by default at @option{-O3}. 7101 7102@item -fipa-matrix-reorg 7103@opindex fipa-matrix-reorg 7104Perform matrix flattening and transposing. 7105Matrix flattening tries to replace an @math{m}-dimensional matrix 7106with its equivalent @math{n}-dimensional matrix, where @math{n < m}. 7107This reduces the level of indirection needed for accessing the elements 7108of the matrix. The second optimization is matrix transposing, which 7109attempts to change the order of the matrix's dimensions in order to 7110improve cache locality. 7111Both optimizations need the @option{-fwhole-program} flag. 7112Transposing is enabled only if profiling information is available. 7113 7114@item -ftree-sink 7115@opindex ftree-sink 7116Perform forward store motion on trees. This flag is 7117enabled by default at @option{-O} and higher. 7118 7119@item -ftree-bit-ccp 7120@opindex ftree-bit-ccp 7121Perform sparse conditional bit constant propagation on trees and propagate 7122pointer alignment information. 7123This pass only operates on local scalar variables and is enabled by default 7124at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled. 7125 7126@item -ftree-ccp 7127@opindex ftree-ccp 7128Perform sparse conditional constant propagation (CCP) on trees. This 7129pass only operates on local scalar variables and is enabled by default 7130at @option{-O} and higher. 7131 7132@item -ftree-switch-conversion 7133Perform conversion of simple initializations in a switch to 7134initializations from a scalar array. This flag is enabled by default 7135at @option{-O2} and higher. 7136 7137@item -ftree-tail-merge 7138Look for identical code sequences. When found, replace one with a jump to the 7139other. This optimization is known as tail merging or cross jumping. This flag 7140is enabled by default at @option{-O2} and higher. The compilation time 7141in this pass can 7142be limited using @option{max-tail-merge-comparisons} parameter and 7143@option{max-tail-merge-iterations} parameter. 7144 7145@item -ftree-dce 7146@opindex ftree-dce 7147Perform dead code elimination (DCE) on trees. This flag is enabled by 7148default at @option{-O} and higher. 7149 7150@item -ftree-builtin-call-dce 7151@opindex ftree-builtin-call-dce 7152Perform conditional dead code elimination (DCE) for calls to builtin functions 7153that may set @code{errno} but are otherwise side-effect free. This flag is 7154enabled by default at @option{-O2} and higher if @option{-Os} is not also 7155specified. 7156 7157@item -ftree-dominator-opts 7158@opindex ftree-dominator-opts 7159Perform a variety of simple scalar cleanups (constant/copy 7160propagation, redundancy elimination, range propagation and expression 7161simplification) based on a dominator tree traversal. This also 7162performs jump threading (to reduce jumps to jumps). This flag is 7163enabled by default at @option{-O} and higher. 7164 7165@item -ftree-dse 7166@opindex ftree-dse 7167Perform dead store elimination (DSE) on trees. A dead store is a store into 7168a memory location that is later overwritten by another store without 7169any intervening loads. In this case the earlier store can be deleted. This 7170flag is enabled by default at @option{-O} and higher. 7171 7172@item -ftree-ch 7173@opindex ftree-ch 7174Perform loop header copying on trees. This is beneficial since it increases 7175effectiveness of code motion optimizations. It also saves one jump. This flag 7176is enabled by default at @option{-O} and higher. It is not enabled 7177for @option{-Os}, since it usually increases code size. 7178 7179@item -ftree-loop-optimize 7180@opindex ftree-loop-optimize 7181Perform loop optimizations on trees. This flag is enabled by default 7182at @option{-O} and higher. 7183 7184@item -ftree-loop-linear 7185@opindex ftree-loop-linear 7186Perform loop interchange transformations on tree. Same as 7187@option{-floop-interchange}. To use this code transformation, GCC has 7188to be configured with @option{--with-ppl} and @option{--with-cloog} to 7189enable the Graphite loop transformation infrastructure. 7190 7191@item -floop-interchange 7192@opindex floop-interchange 7193Perform loop interchange transformations on loops. Interchanging two 7194nested loops switches the inner and outer loops. For example, given a 7195loop like: 7196@smallexample 7197DO J = 1, M 7198 DO I = 1, N 7199 A(J, I) = A(J, I) * C 7200 ENDDO 7201ENDDO 7202@end smallexample 7203loop interchange will transform the loop as if the user had written: 7204@smallexample 7205DO I = 1, N 7206 DO J = 1, M 7207 A(J, I) = A(J, I) * C 7208 ENDDO 7209ENDDO 7210@end smallexample 7211which can be beneficial when @code{N} is larger than the caches, 7212because in Fortran, the elements of an array are stored in memory 7213contiguously by column, and the original loop iterates over rows, 7214potentially creating at each access a cache miss. This optimization 7215applies to all the languages supported by GCC and is not limited to 7216Fortran. To use this code transformation, GCC has to be configured 7217with @option{--with-ppl} and @option{--with-cloog} to enable the 7218Graphite loop transformation infrastructure. 7219 7220@item -floop-strip-mine 7221@opindex floop-strip-mine 7222Perform loop strip mining transformations on loops. Strip mining 7223splits a loop into two nested loops. The outer loop has strides 7224equal to the strip size and the inner loop has strides of the 7225original loop within a strip. The strip length can be changed 7226using the @option{loop-block-tile-size} parameter. For example, 7227given a loop like: 7228@smallexample 7229DO I = 1, N 7230 A(I) = A(I) + C 7231ENDDO 7232@end smallexample 7233loop strip mining will transform the loop as if the user had written: 7234@smallexample 7235DO II = 1, N, 51 7236 DO I = II, min (II + 50, N) 7237 A(I) = A(I) + C 7238 ENDDO 7239ENDDO 7240@end smallexample 7241This optimization applies to all the languages supported by GCC and is 7242not limited to Fortran. To use this code transformation, GCC has to 7243be configured with @option{--with-ppl} and @option{--with-cloog} to 7244enable the Graphite loop transformation infrastructure. 7245 7246@item -floop-block 7247@opindex floop-block 7248Perform loop blocking transformations on loops. Blocking strip mines 7249each loop in the loop nest such that the memory accesses of the 7250element loops fit inside caches. The strip length can be changed 7251using the @option{loop-block-tile-size} parameter. For example, given 7252a loop like: 7253@smallexample 7254DO I = 1, N 7255 DO J = 1, M 7256 A(J, I) = B(I) + C(J) 7257 ENDDO 7258ENDDO 7259@end smallexample 7260loop blocking will transform the loop as if the user had written: 7261@smallexample 7262DO II = 1, N, 51 7263 DO JJ = 1, M, 51 7264 DO I = II, min (II + 50, N) 7265 DO J = JJ, min (JJ + 50, M) 7266 A(J, I) = B(I) + C(J) 7267 ENDDO 7268 ENDDO 7269 ENDDO 7270ENDDO 7271@end smallexample 7272which can be beneficial when @code{M} is larger than the caches, 7273because the innermost loop will iterate over a smaller amount of data 7274which can be kept in the caches. This optimization applies to all the 7275languages supported by GCC and is not limited to Fortran. To use this 7276code transformation, GCC has to be configured with @option{--with-ppl} 7277and @option{--with-cloog} to enable the Graphite loop transformation 7278infrastructure. 7279 7280@item -fgraphite-identity 7281@opindex fgraphite-identity 7282Enable the identity transformation for graphite. For every SCoP we generate 7283the polyhedral representation and transform it back to gimple. Using 7284@option{-fgraphite-identity} we can check the costs or benefits of the 7285GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations 7286are also performed by the code generator CLooG, like index splitting and 7287dead code elimination in loops. 7288 7289@item -floop-flatten 7290@opindex floop-flatten 7291Removes the loop nesting structure: transforms the loop nest into a 7292single loop. This transformation can be useful as an enablement 7293transform for vectorization and parallelization. This feature 7294is experimental. 7295To use this code transformation, GCC has to be configured 7296with @option{--with-ppl} and @option{--with-cloog} to enable the 7297Graphite loop transformation infrastructure. 7298 7299@item -floop-parallelize-all 7300@opindex floop-parallelize-all 7301Use the Graphite data dependence analysis to identify loops that can 7302be parallelized. Parallelize all the loops that can be analyzed to 7303not contain loop carried dependences without checking that it is 7304profitable to parallelize the loops. 7305 7306@item -fcheck-data-deps 7307@opindex fcheck-data-deps 7308Compare the results of several data dependence analyzers. This option 7309is used for debugging the data dependence analyzers. 7310 7311@item -ftree-loop-if-convert 7312Attempt to transform conditional jumps in the innermost loops to 7313branch-less equivalents. The intent is to remove control-flow from 7314the innermost loops in order to improve the ability of the 7315vectorization pass to handle these loops. This is enabled by default 7316if vectorization is enabled. 7317 7318@item -ftree-loop-if-convert-stores 7319Attempt to also if-convert conditional jumps containing memory writes. 7320This transformation can be unsafe for multi-threaded programs as it 7321transforms conditional memory writes into unconditional memory writes. 7322For example, 7323@smallexample 7324for (i = 0; i < N; i++) 7325 if (cond) 7326 A[i] = expr; 7327@end smallexample 7328would be transformed to 7329@smallexample 7330for (i = 0; i < N; i++) 7331 A[i] = cond ? expr : A[i]; 7332@end smallexample 7333potentially producing data races. 7334 7335@item -ftree-loop-distribution 7336Perform loop distribution. This flag can improve cache performance on 7337big loop bodies and allow further loop optimizations, like 7338parallelization or vectorization, to take place. For example, the loop 7339@smallexample 7340DO I = 1, N 7341 A(I) = B(I) + C 7342 D(I) = E(I) * F 7343ENDDO 7344@end smallexample 7345is transformed to 7346@smallexample 7347DO I = 1, N 7348 A(I) = B(I) + C 7349ENDDO 7350DO I = 1, N 7351 D(I) = E(I) * F 7352ENDDO 7353@end smallexample 7354 7355@item -ftree-loop-distribute-patterns 7356Perform loop distribution of patterns that can be code generated with 7357calls to a library. This flag is enabled by default at @option{-O3}. 7358 7359This pass distributes the initialization loops and generates a call to 7360memset zero. For example, the loop 7361@smallexample 7362DO I = 1, N 7363 A(I) = 0 7364 B(I) = A(I) + I 7365ENDDO 7366@end smallexample 7367is transformed to 7368@smallexample 7369DO I = 1, N 7370 A(I) = 0 7371ENDDO 7372DO I = 1, N 7373 B(I) = A(I) + I 7374ENDDO 7375@end smallexample 7376and the initialization loop is transformed into a call to memset zero. 7377 7378@item -ftree-loop-im 7379@opindex ftree-loop-im 7380Perform loop invariant motion on trees. This pass moves only invariants that 7381would be hard to handle at RTL level (function calls, operations that expand to 7382nontrivial sequences of insns). With @option{-funswitch-loops} it also moves 7383operands of conditions that are invariant out of the loop, so that we can use 7384just trivial invariantness analysis in loop unswitching. The pass also includes 7385store motion. 7386 7387@item -ftree-loop-ivcanon 7388@opindex ftree-loop-ivcanon 7389Create a canonical counter for number of iterations in loops for which 7390determining number of iterations requires complicated analysis. Later 7391optimizations then may determine the number easily. Useful especially 7392in connection with unrolling. 7393 7394@item -fivopts 7395@opindex fivopts 7396Perform induction variable optimizations (strength reduction, induction 7397variable merging and induction variable elimination) on trees. 7398 7399@item -ftree-parallelize-loops=n 7400@opindex ftree-parallelize-loops 7401Parallelize loops, i.e., split their iteration space to run in n threads. 7402This is only possible for loops whose iterations are independent 7403and can be arbitrarily reordered. The optimization is only 7404profitable on multiprocessor machines, for loops that are CPU-intensive, 7405rather than constrained e.g.@: by memory bandwidth. This option 7406implies @option{-pthread}, and thus is only supported on targets 7407that have support for @option{-pthread}. 7408 7409@item -ftree-pta 7410@opindex ftree-pta 7411Perform function-local points-to analysis on trees. This flag is 7412enabled by default at @option{-O} and higher. 7413 7414@item -ftree-sra 7415@opindex ftree-sra 7416Perform scalar replacement of aggregates. This pass replaces structure 7417references with scalars to prevent committing structures to memory too 7418early. This flag is enabled by default at @option{-O} and higher. 7419 7420@item -ftree-copyrename 7421@opindex ftree-copyrename 7422Perform copy renaming on trees. This pass attempts to rename compiler 7423temporaries to other variables at copy locations, usually resulting in 7424variable names which more closely resemble the original variables. This flag 7425is enabled by default at @option{-O} and higher. 7426 7427@item -ftree-ter 7428@opindex ftree-ter 7429Perform temporary expression replacement during the SSA->normal phase. Single 7430use/single def temporaries are replaced at their use location with their 7431defining expression. This results in non-GIMPLE code, but gives the expanders 7432much more complex trees to work on resulting in better RTL generation. This is 7433enabled by default at @option{-O} and higher. 7434 7435@item -ftree-vectorize 7436@opindex ftree-vectorize 7437Perform loop vectorization on trees. This flag is enabled by default at 7438@option{-O3}. 7439 7440@item -ftree-slp-vectorize 7441@opindex ftree-slp-vectorize 7442Perform basic block vectorization on trees. This flag is enabled by default at 7443@option{-O3} and when @option{-ftree-vectorize} is enabled. 7444 7445@item -ftree-vect-loop-version 7446@opindex ftree-vect-loop-version 7447Perform loop versioning when doing loop vectorization on trees. When a loop 7448appears to be vectorizable except that data alignment or data dependence cannot 7449be determined at compile time, then vectorized and non-vectorized versions of 7450the loop are generated along with run-time checks for alignment or dependence 7451to control which version is executed. This option is enabled by default 7452except at level @option{-Os} where it is disabled. 7453 7454@item -fvect-cost-model 7455@opindex fvect-cost-model 7456Enable cost model for vectorization. 7457 7458@item -ftree-vrp 7459@opindex ftree-vrp 7460Perform Value Range Propagation on trees. This is similar to the 7461constant propagation pass, but instead of values, ranges of values are 7462propagated. This allows the optimizers to remove unnecessary range 7463checks like array bound checks and null pointer checks. This is 7464enabled by default at @option{-O2} and higher. Null pointer check 7465elimination is only done if @option{-fdelete-null-pointer-checks} is 7466enabled. 7467 7468@item -ftracer 7469@opindex ftracer 7470Perform tail duplication to enlarge superblock size. This transformation 7471simplifies the control flow of the function allowing other optimizations to do 7472better job. 7473 7474@item -funroll-loops 7475@opindex funroll-loops 7476Unroll loops whose number of iterations can be determined at compile 7477time or upon entry to the loop. @option{-funroll-loops} implies 7478@option{-frerun-cse-after-loop}. This option makes code larger, 7479and may or may not make it run faster. 7480 7481@item -funroll-all-loops 7482@opindex funroll-all-loops 7483Unroll all loops, even if their number of iterations is uncertain when 7484the loop is entered. This usually makes programs run more slowly. 7485@option{-funroll-all-loops} implies the same options as 7486@option{-funroll-loops}, 7487 7488@item -fsplit-ivs-in-unroller 7489@opindex fsplit-ivs-in-unroller 7490Enables expressing of values of induction variables in later iterations 7491of the unrolled loop using the value in the first iteration. This breaks 7492long dependency chains, thus improving efficiency of the scheduling passes. 7493 7494Combination of @option{-fweb} and CSE is often sufficient to obtain the 7495same effect. However in cases the loop body is more complicated than 7496a single basic block, this is not reliable. It also does not work at all 7497on some of the architectures due to restrictions in the CSE pass. 7498 7499This optimization is enabled by default. 7500 7501@item -fvariable-expansion-in-unroller 7502@opindex fvariable-expansion-in-unroller 7503With this option, the compiler will create multiple copies of some 7504local variables when unrolling a loop which can result in superior code. 7505 7506@item -fpartial-inlining 7507@opindex fpartial-inlining 7508Inline parts of functions. This option has any effect only 7509when inlining itself is turned on by the @option{-finline-functions} 7510or @option{-finline-small-functions} options. 7511 7512Enabled at level @option{-O2}. 7513 7514@item -fpredictive-commoning 7515@opindex fpredictive-commoning 7516Perform predictive commoning optimization, i.e., reusing computations 7517(especially memory loads and stores) performed in previous 7518iterations of loops. 7519 7520This option is enabled at level @option{-O3}. 7521 7522@item -fprefetch-loop-arrays 7523@opindex fprefetch-loop-arrays 7524If supported by the target machine, generate instructions to prefetch 7525memory to improve the performance of loops that access large arrays. 7526 7527This option may generate better or worse code; results are highly 7528dependent on the structure of loops within the source code. 7529 7530Disabled at level @option{-Os}. 7531 7532@item -fno-peephole 7533@itemx -fno-peephole2 7534@opindex fno-peephole 7535@opindex fno-peephole2 7536Disable any machine-specific peephole optimizations. The difference 7537between @option{-fno-peephole} and @option{-fno-peephole2} is in how they 7538are implemented in the compiler; some targets use one, some use the 7539other, a few use both. 7540 7541@option{-fpeephole} is enabled by default. 7542@option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 7543 7544@item -fno-guess-branch-probability 7545@opindex fno-guess-branch-probability 7546Do not guess branch probabilities using heuristics. 7547 7548GCC will use heuristics to guess branch probabilities if they are 7549not provided by profiling feedback (@option{-fprofile-arcs}). These 7550heuristics are based on the control flow graph. If some branch probabilities 7551are specified by @samp{__builtin_expect}, then the heuristics will be 7552used to guess branch probabilities for the rest of the control flow graph, 7553taking the @samp{__builtin_expect} info into account. The interactions 7554between the heuristics and @samp{__builtin_expect} can be complex, and in 7555some cases, it may be useful to disable the heuristics so that the effects 7556of @samp{__builtin_expect} are easier to understand. 7557 7558The default is @option{-fguess-branch-probability} at levels 7559@option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 7560 7561@item -freorder-blocks 7562@opindex freorder-blocks 7563Reorder basic blocks in the compiled function in order to reduce number of 7564taken branches and improve code locality. 7565 7566Enabled at levels @option{-O2}, @option{-O3}. 7567 7568@item -freorder-blocks-and-partition 7569@opindex freorder-blocks-and-partition 7570In addition to reordering basic blocks in the compiled function, in order 7571to reduce number of taken branches, partitions hot and cold basic blocks 7572into separate sections of the assembly and .o files, to improve 7573paging and cache locality performance. 7574 7575This optimization is automatically turned off in the presence of 7576exception handling, for linkonce sections, for functions with a user-defined 7577section attribute and on any architecture that does not support named 7578sections. 7579 7580@item -freorder-functions 7581@opindex freorder-functions 7582Reorder functions in the object file in order to 7583improve code locality. This is implemented by using special 7584subsections @code{.text.hot} for most frequently executed functions and 7585@code{.text.unlikely} for unlikely executed functions. Reordering is done by 7586the linker so object file format must support named sections and linker must 7587place them in a reasonable way. 7588 7589Also profile feedback must be available in to make this option effective. See 7590@option{-fprofile-arcs} for details. 7591 7592Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 7593 7594@item -fstrict-aliasing 7595@opindex fstrict-aliasing 7596Allow the compiler to assume the strictest aliasing rules applicable to 7597the language being compiled. For C (and C++), this activates 7598optimizations based on the type of expressions. In particular, an 7599object of one type is assumed never to reside at the same address as an 7600object of a different type, unless the types are almost the same. For 7601example, an @code{unsigned int} can alias an @code{int}, but not a 7602@code{void*} or a @code{double}. A character type may alias any other 7603type. 7604 7605@anchor{Type-punning}Pay special attention to code like this: 7606@smallexample 7607union a_union @{ 7608 int i; 7609 double d; 7610@}; 7611 7612int f() @{ 7613 union a_union t; 7614 t.d = 3.0; 7615 return t.i; 7616@} 7617@end smallexample 7618The practice of reading from a different union member than the one most 7619recently written to (called ``type-punning'') is common. Even with 7620@option{-fstrict-aliasing}, type-punning is allowed, provided the memory 7621is accessed through the union type. So, the code above will work as 7622expected. @xref{Structures unions enumerations and bit-fields 7623implementation}. However, this code might not: 7624@smallexample 7625int f() @{ 7626 union a_union t; 7627 int* ip; 7628 t.d = 3.0; 7629 ip = &t.i; 7630 return *ip; 7631@} 7632@end smallexample 7633 7634Similarly, access by taking the address, casting the resulting pointer 7635and dereferencing the result has undefined behavior, even if the cast 7636uses a union type, e.g.: 7637@smallexample 7638int f() @{ 7639 double d = 3.0; 7640 return ((union a_union *) &d)->i; 7641@} 7642@end smallexample 7643 7644The @option{-fstrict-aliasing} option is enabled at levels 7645@option{-O2}, @option{-O3}, @option{-Os}. 7646 7647@item -fstrict-overflow 7648@opindex fstrict-overflow 7649Allow the compiler to assume strict signed overflow rules, depending 7650on the language being compiled. For C (and C++) this means that 7651overflow when doing arithmetic with signed numbers is undefined, which 7652means that the compiler may assume that it will not happen. This 7653permits various optimizations. For example, the compiler will assume 7654that an expression like @code{i + 10 > i} will always be true for 7655signed @code{i}. This assumption is only valid if signed overflow is 7656undefined, as the expression is false if @code{i + 10} overflows when 7657using twos complement arithmetic. When this option is in effect any 7658attempt to determine whether an operation on signed numbers will 7659overflow must be written carefully to not actually involve overflow. 7660 7661This option also allows the compiler to assume strict pointer 7662semantics: given a pointer to an object, if adding an offset to that 7663pointer does not produce a pointer to the same object, the addition is 7664undefined. This permits the compiler to conclude that @code{p + u > 7665p} is always true for a pointer @code{p} and unsigned integer 7666@code{u}. This assumption is only valid because pointer wraparound is 7667undefined, as the expression is false if @code{p + u} overflows using 7668twos complement arithmetic. 7669 7670See also the @option{-fwrapv} option. Using @option{-fwrapv} means 7671that integer signed overflow is fully defined: it wraps. When 7672@option{-fwrapv} is used, there is no difference between 7673@option{-fstrict-overflow} and @option{-fno-strict-overflow} for 7674integers. With @option{-fwrapv} certain types of overflow are 7675permitted. For example, if the compiler gets an overflow when doing 7676arithmetic on constants, the overflowed value can still be used with 7677@option{-fwrapv}, but not otherwise. 7678 7679The @option{-fstrict-overflow} option is enabled at levels 7680@option{-O2}, @option{-O3}, @option{-Os}. 7681 7682@item -falign-functions 7683@itemx -falign-functions=@var{n} 7684@opindex falign-functions 7685Align the start of functions to the next power-of-two greater than 7686@var{n}, skipping up to @var{n} bytes. For instance, 7687@option{-falign-functions=32} aligns functions to the next 32-byte 7688boundary, but @option{-falign-functions=24} would align to the next 768932-byte boundary only if this can be done by skipping 23 bytes or less. 7690 7691@option{-fno-align-functions} and @option{-falign-functions=1} are 7692equivalent and mean that functions will not be aligned. 7693 7694Some assemblers only support this flag when @var{n} is a power of two; 7695in that case, it is rounded up. 7696 7697If @var{n} is not specified or is zero, use a machine-dependent default. 7698 7699Enabled at levels @option{-O2}, @option{-O3}. 7700 7701@item -falign-labels 7702@itemx -falign-labels=@var{n} 7703@opindex falign-labels 7704Align all branch targets to a power-of-two boundary, skipping up to 7705@var{n} bytes like @option{-falign-functions}. This option can easily 7706make code slower, because it must insert dummy operations for when the 7707branch target is reached in the usual flow of the code. 7708 7709@option{-fno-align-labels} and @option{-falign-labels=1} are 7710equivalent and mean that labels will not be aligned. 7711 7712If @option{-falign-loops} or @option{-falign-jumps} are applicable and 7713are greater than this value, then their values are used instead. 7714 7715If @var{n} is not specified or is zero, use a machine-dependent default 7716which is very likely to be @samp{1}, meaning no alignment. 7717 7718Enabled at levels @option{-O2}, @option{-O3}. 7719 7720@item -falign-loops 7721@itemx -falign-loops=@var{n} 7722@opindex falign-loops 7723Align loops to a power-of-two boundary, skipping up to @var{n} bytes 7724like @option{-falign-functions}. The hope is that the loop will be 7725executed many times, which will make up for any execution of the dummy 7726operations. 7727 7728@option{-fno-align-loops} and @option{-falign-loops=1} are 7729equivalent and mean that loops will not be aligned. 7730 7731If @var{n} is not specified or is zero, use a machine-dependent default. 7732 7733Enabled at levels @option{-O2}, @option{-O3}. 7734 7735@item -falign-jumps 7736@itemx -falign-jumps=@var{n} 7737@opindex falign-jumps 7738Align branch targets to a power-of-two boundary, for branch targets 7739where the targets can only be reached by jumping, skipping up to @var{n} 7740bytes like @option{-falign-functions}. In this case, no dummy operations 7741need be executed. 7742 7743@option{-fno-align-jumps} and @option{-falign-jumps=1} are 7744equivalent and mean that loops will not be aligned. 7745 7746If @var{n} is not specified or is zero, use a machine-dependent default. 7747 7748Enabled at levels @option{-O2}, @option{-O3}. 7749 7750@item -funit-at-a-time 7751@opindex funit-at-a-time 7752This option is left for compatibility reasons. @option{-funit-at-a-time} 7753has no effect, while @option{-fno-unit-at-a-time} implies 7754@option{-fno-toplevel-reorder} and @option{-fno-section-anchors}. 7755 7756Enabled by default. 7757 7758@item -fno-toplevel-reorder 7759@opindex fno-toplevel-reorder 7760Do not reorder top-level functions, variables, and @code{asm} 7761statements. Output them in the same order that they appear in the 7762input file. When this option is used, unreferenced static variables 7763will not be removed. This option is intended to support existing code 7764that relies on a particular ordering. For new code, it is better to 7765use attributes. 7766 7767Enabled at level @option{-O0}. When disabled explicitly, it also implies 7768@option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some 7769targets. 7770 7771@item -fweb 7772@opindex fweb 7773Constructs webs as commonly used for register allocation purposes and assign 7774each web individual pseudo register. This allows the register allocation pass 7775to operate on pseudos directly, but also strengthens several other optimization 7776passes, such as CSE, loop optimizer and trivial dead code remover. It can, 7777however, make debugging impossible, since variables will no longer stay in a 7778``home register''. 7779 7780Enabled by default with @option{-funroll-loops}. 7781 7782@item -fwhole-program 7783@opindex fwhole-program 7784Assume that the current compilation unit represents the whole program being 7785compiled. All public functions and variables with the exception of @code{main} 7786and those merged by attribute @code{externally_visible} become static functions 7787and in effect are optimized more aggressively by interprocedural optimizers. If @command{gold} is used as the linker plugin, @code{externally_visible} attributes are automatically added to functions (not variable yet due to a current @command{gold} issue) that are accessed outside of LTO objects according to resolution file produced by @command{gold}. For other linkers that cannot generate resolution file, explicit @code{externally_visible} attributes are still necessary. 7788While this option is equivalent to proper use of the @code{static} keyword for 7789programs consisting of a single file, in combination with option 7790@option{-flto} this flag can be used to 7791compile many smaller scale programs since the functions and variables become 7792local for the whole combined compilation unit, not for the single source file 7793itself. 7794 7795This option implies @option{-fwhole-file} for Fortran programs. 7796 7797@item -flto[=@var{n}] 7798@opindex flto 7799This option runs the standard link-time optimizer. When invoked 7800with source code, it generates GIMPLE (one of GCC's internal 7801representations) and writes it to special ELF sections in the object 7802file. When the object files are linked together, all the function 7803bodies are read from these ELF sections and instantiated as if they 7804had been part of the same translation unit. 7805 7806To use the link-time optimizer, @option{-flto} needs to be specified at 7807compile time and during the final link. For example: 7808 7809@smallexample 7810gcc -c -O2 -flto foo.c 7811gcc -c -O2 -flto bar.c 7812gcc -o myprog -flto -O2 foo.o bar.o 7813@end smallexample 7814 7815The first two invocations to GCC save a bytecode representation 7816of GIMPLE into special ELF sections inside @file{foo.o} and 7817@file{bar.o}. The final invocation reads the GIMPLE bytecode from 7818@file{foo.o} and @file{bar.o}, merges the two files into a single 7819internal image, and compiles the result as usual. Since both 7820@file{foo.o} and @file{bar.o} are merged into a single image, this 7821causes all the interprocedural analyses and optimizations in GCC to 7822work across the two files as if they were a single one. This means, 7823for example, that the inliner is able to inline functions in 7824@file{bar.o} into functions in @file{foo.o} and vice-versa. 7825 7826Another (simpler) way to enable link-time optimization is: 7827 7828@smallexample 7829gcc -o myprog -flto -O2 foo.c bar.c 7830@end smallexample 7831 7832The above generates bytecode for @file{foo.c} and @file{bar.c}, 7833merges them together into a single GIMPLE representation and optimizes 7834them as usual to produce @file{myprog}. 7835 7836The only important thing to keep in mind is that to enable link-time 7837optimizations the @option{-flto} flag needs to be passed to both the 7838compile and the link commands. 7839 7840To make whole program optimization effective, it is necessary to make 7841certain whole program assumptions. The compiler needs to know 7842what functions and variables can be accessed by libraries and runtime 7843outside of the link-time optimized unit. When supported by the linker, 7844the linker plugin (see @option{-fuse-linker-plugin}) passes information 7845to the compiler about used and externally visible symbols. When 7846the linker plugin is not available, @option{-fwhole-program} should be 7847used to allow the compiler to make these assumptions, which leads 7848to more aggressive optimization decisions. 7849 7850Note that when a file is compiled with @option{-flto}, the generated 7851object file is larger than a regular object file because it 7852contains GIMPLE bytecodes and the usual final code. This means that 7853object files with LTO information can be linked as normal object 7854files; if @option{-flto} is not passed to the linker, no 7855interprocedural optimizations are applied. 7856 7857Additionally, the optimization flags used to compile individual files 7858are not necessarily related to those used at link time. For instance, 7859 7860@smallexample 7861gcc -c -O0 -flto foo.c 7862gcc -c -O0 -flto bar.c 7863gcc -o myprog -flto -O3 foo.o bar.o 7864@end smallexample 7865 7866This produces individual object files with unoptimized assembler 7867code, but the resulting binary @file{myprog} is optimized at 7868@option{-O3}. If, instead, the final binary is generated without 7869@option{-flto}, then @file{myprog} is not optimized. 7870 7871When producing the final binary with @option{-flto}, GCC only 7872applies link-time optimizations to those files that contain bytecode. 7873Therefore, you can mix and match object files and libraries with 7874GIMPLE bytecodes and final object code. GCC automatically selects 7875which files to optimize in LTO mode and which files to link without 7876further processing. 7877 7878There are some code generation flags preserved by GCC when 7879generating bytecodes, as they need to be used during the final link 7880stage. Currently, the following options are saved into the GIMPLE 7881bytecode files: @option{-fPIC}, @option{-fcommon} and all the 7882@option{-m} target flags. 7883 7884At link time, these options are read in and reapplied. Note that the 7885current implementation makes no attempt to recognize conflicting 7886values for these options. If different files have conflicting option 7887values (e.g., one file is compiled with @option{-fPIC} and another 7888isn't), the compiler simply uses the last value read from the 7889bytecode files. It is recommended, then, that you compile all the files 7890participating in the same link with the same options. 7891 7892If LTO encounters objects with C linkage declared with incompatible 7893types in separate translation units to be linked together (undefined 7894behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be 7895issued. The behavior is still undefined at run time. 7896 7897Another feature of LTO is that it is possible to apply interprocedural 7898optimizations on files written in different languages. This requires 7899support in the language front end. Currently, the C, C++ and 7900Fortran front ends are capable of emitting GIMPLE bytecodes, so 7901something like this should work: 7902 7903@smallexample 7904gcc -c -flto foo.c 7905g++ -c -flto bar.cc 7906gfortran -c -flto baz.f90 7907g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran 7908@end smallexample 7909 7910Notice that the final link is done with @command{g++} to get the C++ 7911runtime libraries and @option{-lgfortran} is added to get the Fortran 7912runtime libraries. In general, when mixing languages in LTO mode, you 7913should use the same link command options as when mixing languages in a 7914regular (non-LTO) compilation; all you need to add is @option{-flto} to 7915all the compile and link commands. 7916 7917If object files containing GIMPLE bytecode are stored in a library archive, say 7918@file{libfoo.a}, it is possible to extract and use them in an LTO link if you 7919are using a linker with plugin support. To enable this feature, use 7920the flag @option{-fuse-linker-plugin} at link time: 7921 7922@smallexample 7923gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo 7924@end smallexample 7925 7926With the linker plugin enabled, the linker extracts the needed 7927GIMPLE files from @file{libfoo.a} and passes them on to the running GCC 7928to make them part of the aggregated GIMPLE image to be optimized. 7929 7930If you are not using a linker with plugin support and/or do not 7931enable the linker plugin, then the objects inside @file{libfoo.a} 7932are extracted and linked as usual, but they do not participate 7933in the LTO optimization process. 7934 7935Link-time optimizations do not require the presence of the whole program to 7936operate. If the program does not require any symbols to be exported, it is 7937possible to combine @option{-flto} and @option{-fwhole-program} to allow 7938the interprocedural optimizers to use more aggressive assumptions which may 7939lead to improved optimization opportunities. 7940Use of @option{-fwhole-program} is not needed when linker plugin is 7941active (see @option{-fuse-linker-plugin}). 7942 7943The current implementation of LTO makes no 7944attempt to generate bytecode that is portable between different 7945types of hosts. The bytecode files are versioned and there is a 7946strict version check, so bytecode files generated in one version of 7947GCC will not work with an older/newer version of GCC. 7948 7949Link-time optimization does not work well with generation of debugging 7950information. Combining @option{-flto} with 7951@option{-g} is currently experimental and expected to produce wrong 7952results. 7953 7954If you specify the optional @var{n}, the optimization and code 7955generation done at link time is executed in parallel using @var{n} 7956parallel jobs by utilizing an installed @command{make} program. The 7957environment variable @env{MAKE} may be used to override the program 7958used. The default value for @var{n} is 1. 7959 7960You can also specify @option{-flto=jobserver} to use GNU make's 7961job server mode to determine the number of parallel jobs. This 7962is useful when the Makefile calling GCC is already executing in parallel. 7963You must prepend a @samp{+} to the command recipe in the parent Makefile 7964for this to work. This option likely only works if @env{MAKE} is 7965GNU make. 7966 7967This option is disabled by default 7968 7969@item -flto-partition=@var{alg} 7970@opindex flto-partition 7971Specify the partitioning algorithm used by the link-time optimizer. 7972The value is either @code{1to1} to specify a partitioning mirroring 7973the original source files or @code{balanced} to specify partitioning 7974into equally sized chunks (whenever possible). Specifying @code{none} 7975as an algorithm disables partitioning and streaming completely. The 7976default value is @code{balanced}. 7977 7978@item -flto-compression-level=@var{n} 7979This option specifies the level of compression used for intermediate 7980language written to LTO object files, and is only meaningful in 7981conjunction with LTO mode (@option{-flto}). Valid 7982values are 0 (no compression) to 9 (maximum compression). Values 7983outside this range are clamped to either 0 or 9. If the option is not 7984given, a default balanced compression setting is used. 7985 7986@item -flto-report 7987Prints a report with internal details on the workings of the link-time 7988optimizer. The contents of this report vary from version to version. 7989It is meant to be useful to GCC developers when processing object 7990files in LTO mode (via @option{-flto}). 7991 7992Disabled by default. 7993 7994@item -fuse-linker-plugin 7995Enables the use of a linker plugin during link-time optimization. This 7996option relies on plugin support in the linker, which is available in gold 7997or in GNU ld 2.21 or newer. 7998 7999This option enables the extraction of object files with GIMPLE bytecode out 8000of library archives. This improves the quality of optimization by exposing 8001more code to the link-time optimizer. This information specifies what 8002symbols can be accessed externally (by non-LTO object or during dynamic 8003linking). Resulting code quality improvements on binaries (and shared 8004libraries that use hidden visibility) are similar to @code{-fwhole-program}. 8005See @option{-flto} for a description of the effect of this flag and how to 8006use it. 8007 8008This option is enabled by default when LTO support in GCC is enabled 8009and GCC was configured for use with 8010a linker supporting plugins (GNU ld 2.21 or newer or gold). 8011 8012@item -ffat-lto-objects 8013@opindex ffat-lto-objects 8014Fat LTO objects are object files that contain both the intermediate language 8015and the object code. This makes them usable for both LTO linking and normal 8016linking. This option is effective only when compiling with @option{-flto} 8017and is ignored at link time. 8018 8019@option{-fno-fat-lto-objects} improves compilation time over plain LTO, but 8020requires the complete toolchain to be aware of LTO. It requires a linker with 8021linker plugin support for basic functionality. Additionally, nm, ar and ranlib 8022need to support linker plugins to allow a full-featured build environment 8023(capable of building static libraries etc). 8024 8025The default is @option{-ffat-lto-objects} but this default is intended to 8026change in future releases when linker plugin enabled environments become more 8027common. 8028 8029@item -fcompare-elim 8030@opindex fcompare-elim 8031After register allocation and post-register allocation instruction splitting, 8032identify arithmetic instructions that compute processor flags similar to a 8033comparison operation based on that arithmetic. If possible, eliminate the 8034explicit comparison operation. 8035 8036This pass only applies to certain targets that cannot explicitly represent 8037the comparison operation before register allocation is complete. 8038 8039Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 8040 8041@item -fcprop-registers 8042@opindex fcprop-registers 8043After register allocation and post-register allocation instruction splitting, 8044we perform a copy-propagation pass to try to reduce scheduling dependencies 8045and occasionally eliminate the copy. 8046 8047Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 8048 8049@item -fprofile-correction 8050@opindex fprofile-correction 8051Profiles collected using an instrumented binary for multi-threaded programs may 8052be inconsistent due to missed counter updates. When this option is specified, 8053GCC will use heuristics to correct or smooth out such inconsistencies. By 8054default, GCC will emit an error message when an inconsistent profile is detected. 8055 8056@item -fprofile-dir=@var{path} 8057@opindex fprofile-dir 8058 8059Set the directory to search for the profile data files in to @var{path}. 8060This option affects only the profile data generated by 8061@option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs} 8062and used by @option{-fprofile-use} and @option{-fbranch-probabilities} 8063and its related options. Both absolute and relative paths can be used. 8064By default, GCC will use the current directory as @var{path}, thus the 8065profile data file will appear in the same directory as the object file. 8066 8067@item -fprofile-generate 8068@itemx -fprofile-generate=@var{path} 8069@opindex fprofile-generate 8070 8071Enable options usually used for instrumenting application to produce 8072profile useful for later recompilation with profile feedback based 8073optimization. You must use @option{-fprofile-generate} both when 8074compiling and when linking your program. 8075 8076The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}. 8077 8078If @var{path} is specified, GCC will look at the @var{path} to find 8079the profile feedback data files. See @option{-fprofile-dir}. 8080 8081@item -fprofile-use 8082@itemx -fprofile-use=@var{path} 8083@opindex fprofile-use 8084Enable profile feedback directed optimizations, and optimizations 8085generally profitable only with profile feedback available. 8086 8087The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt}, 8088@code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer} 8089 8090By default, GCC emits an error message if the feedback profiles do not 8091match the source code. This error can be turned into a warning by using 8092@option{-Wcoverage-mismatch}. Note this may result in poorly optimized 8093code. 8094 8095If @var{path} is specified, GCC will look at the @var{path} to find 8096the profile feedback data files. See @option{-fprofile-dir}. 8097@end table 8098 8099The following options control compiler behavior regarding floating-point 8100arithmetic. These options trade off between speed and 8101correctness. All must be specifically enabled. 8102 8103@table @gcctabopt 8104@item -ffloat-store 8105@opindex ffloat-store 8106Do not store floating-point variables in registers, and inhibit other 8107options that might change whether a floating-point value is taken from a 8108register or memory. 8109 8110@cindex floating-point precision 8111This option prevents undesirable excess precision on machines such as 8112the 68000 where the floating registers (of the 68881) keep more 8113precision than a @code{double} is supposed to have. Similarly for the 8114x86 architecture. For most programs, the excess precision does only 8115good, but a few programs rely on the precise definition of IEEE floating 8116point. Use @option{-ffloat-store} for such programs, after modifying 8117them to store all pertinent intermediate computations into variables. 8118 8119@item -fexcess-precision=@var{style} 8120@opindex fexcess-precision 8121This option allows further control over excess precision on machines 8122where floating-point registers have more precision than the IEEE 8123@code{float} and @code{double} types and the processor does not 8124support operations rounding to those types. By default, 8125@option{-fexcess-precision=fast} is in effect; this means that 8126operations are carried out in the precision of the registers and that 8127it is unpredictable when rounding to the types specified in the source 8128code takes place. When compiling C, if 8129@option{-fexcess-precision=standard} is specified then excess 8130precision will follow the rules specified in ISO C99; in particular, 8131both casts and assignments cause values to be rounded to their 8132semantic types (whereas @option{-ffloat-store} only affects 8133assignments). This option is enabled by default for C if a strict 8134conformance option such as @option{-std=c99} is used. 8135 8136@opindex mfpmath 8137@option{-fexcess-precision=standard} is not implemented for languages 8138other than C, and has no effect if 8139@option{-funsafe-math-optimizations} or @option{-ffast-math} is 8140specified. On the x86, it also has no effect if @option{-mfpmath=sse} 8141or @option{-mfpmath=sse+387} is specified; in the former case, IEEE 8142semantics apply without excess precision, and in the latter, rounding 8143is unpredictable. 8144 8145@item -ffast-math 8146@opindex ffast-math 8147Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, 8148@option{-ffinite-math-only}, @option{-fno-rounding-math}, 8149@option{-fno-signaling-nans} and @option{-fcx-limited-range}. 8150 8151This option causes the preprocessor macro @code{__FAST_MATH__} to be defined. 8152 8153This option is not turned on by any @option{-O} option besides 8154@option{-Ofast} since it can result in incorrect output for programs 8155that depend on an exact implementation of IEEE or ISO rules/specifications 8156for math functions. It may, however, yield faster code for programs 8157that do not require the guarantees of these specifications. 8158 8159@item -fno-math-errno 8160@opindex fno-math-errno 8161Do not set ERRNO after calling math functions that are executed 8162with a single instruction, e.g., sqrt. A program that relies on 8163IEEE exceptions for math error handling may want to use this flag 8164for speed while maintaining IEEE arithmetic compatibility. 8165 8166This option is not turned on by any @option{-O} option since 8167it can result in incorrect output for programs that depend on 8168an exact implementation of IEEE or ISO rules/specifications for 8169math functions. It may, however, yield faster code for programs 8170that do not require the guarantees of these specifications. 8171 8172The default is @option{-fmath-errno}. 8173 8174On Darwin systems, the math library never sets @code{errno}. There is 8175therefore no reason for the compiler to consider the possibility that 8176it might, and @option{-fno-math-errno} is the default. 8177 8178@item -funsafe-math-optimizations 8179@opindex funsafe-math-optimizations 8180 8181Allow optimizations for floating-point arithmetic that (a) assume 8182that arguments and results are valid and (b) may violate IEEE or 8183ANSI standards. When used at link-time, it may include libraries 8184or startup files that change the default FPU control word or other 8185similar optimizations. 8186 8187This option is not turned on by any @option{-O} option since 8188it can result in incorrect output for programs that depend on 8189an exact implementation of IEEE or ISO rules/specifications for 8190math functions. It may, however, yield faster code for programs 8191that do not require the guarantees of these specifications. 8192Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math}, 8193@option{-fassociative-math} and @option{-freciprocal-math}. 8194 8195The default is @option{-fno-unsafe-math-optimizations}. 8196 8197@item -fassociative-math 8198@opindex fassociative-math 8199 8200Allow re-association of operands in series of floating-point operations. 8201This violates the ISO C and C++ language standard by possibly changing 8202computation result. NOTE: re-ordering may change the sign of zero as 8203well as ignore NaNs and inhibit or create underflow or overflow (and 8204thus cannot be used on code that relies on rounding behavior like 8205@code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons 8206and thus may not be used when ordered comparisons are required. 8207This option requires that both @option{-fno-signed-zeros} and 8208@option{-fno-trapping-math} be in effect. Moreover, it doesn't make 8209much sense with @option{-frounding-math}. For Fortran the option 8210is automatically enabled when both @option{-fno-signed-zeros} and 8211@option{-fno-trapping-math} are in effect. 8212 8213The default is @option{-fno-associative-math}. 8214 8215@item -freciprocal-math 8216@opindex freciprocal-math 8217 8218Allow the reciprocal of a value to be used instead of dividing by 8219the value if this enables optimizations. For example @code{x / y} 8220can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)} 8221is subject to common subexpression elimination. Note that this loses 8222precision and increases the number of flops operating on the value. 8223 8224The default is @option{-fno-reciprocal-math}. 8225 8226@item -ffinite-math-only 8227@opindex ffinite-math-only 8228Allow optimizations for floating-point arithmetic that assume 8229that arguments and results are not NaNs or +-Infs. 8230 8231This option is not turned on by any @option{-O} option since 8232it can result in incorrect output for programs that depend on 8233an exact implementation of IEEE or ISO rules/specifications for 8234math functions. It may, however, yield faster code for programs 8235that do not require the guarantees of these specifications. 8236 8237The default is @option{-fno-finite-math-only}. 8238 8239@item -fno-signed-zeros 8240@opindex fno-signed-zeros 8241Allow optimizations for floating-point arithmetic that ignore the 8242signedness of zero. IEEE arithmetic specifies the behavior of 8243distinct +0.0 and @minus{}0.0 values, which then prohibits simplification 8244of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}). 8245This option implies that the sign of a zero result isn't significant. 8246 8247The default is @option{-fsigned-zeros}. 8248 8249@item -fno-trapping-math 8250@opindex fno-trapping-math 8251Compile code assuming that floating-point operations cannot generate 8252user-visible traps. These traps include division by zero, overflow, 8253underflow, inexact result and invalid operation. This option requires 8254that @option{-fno-signaling-nans} be in effect. Setting this option may 8255allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example. 8256 8257This option should never be turned on by any @option{-O} option since 8258it can result in incorrect output for programs that depend on 8259an exact implementation of IEEE or ISO rules/specifications for 8260math functions. 8261 8262The default is @option{-ftrapping-math}. 8263 8264@item -frounding-math 8265@opindex frounding-math 8266Disable transformations and optimizations that assume default floating-point 8267rounding behavior. This is round-to-zero for all floating point 8268to integer conversions, and round-to-nearest for all other arithmetic 8269truncations. This option should be specified for programs that change 8270the FP rounding mode dynamically, or that may be executed with a 8271non-default rounding mode. This option disables constant folding of 8272floating-point expressions at compile time (which may be affected by 8273rounding mode) and arithmetic transformations that are unsafe in the 8274presence of sign-dependent rounding modes. 8275 8276The default is @option{-fno-rounding-math}. 8277 8278This option is experimental and does not currently guarantee to 8279disable all GCC optimizations that are affected by rounding mode. 8280Future versions of GCC may provide finer control of this setting 8281using C99's @code{FENV_ACCESS} pragma. This command-line option 8282will be used to specify the default state for @code{FENV_ACCESS}. 8283 8284@item -fsignaling-nans 8285@opindex fsignaling-nans 8286Compile code assuming that IEEE signaling NaNs may generate user-visible 8287traps during floating-point operations. Setting this option disables 8288optimizations that may change the number of exceptions visible with 8289signaling NaNs. This option implies @option{-ftrapping-math}. 8290 8291This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to 8292be defined. 8293 8294The default is @option{-fno-signaling-nans}. 8295 8296This option is experimental and does not currently guarantee to 8297disable all GCC optimizations that affect signaling NaN behavior. 8298 8299@item -fsingle-precision-constant 8300@opindex fsingle-precision-constant 8301Treat floating-point constants as single precision instead of 8302implicitly converting them to double-precision constants. 8303 8304@item -fcx-limited-range 8305@opindex fcx-limited-range 8306When enabled, this option states that a range reduction step is not 8307needed when performing complex division. Also, there is no checking 8308whether the result of a complex multiplication or division is @code{NaN 8309+ I*NaN}, with an attempt to rescue the situation in that case. The 8310default is @option{-fno-cx-limited-range}, but is enabled by 8311@option{-ffast-math}. 8312 8313This option controls the default setting of the ISO C99 8314@code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to 8315all languages. 8316 8317@item -fcx-fortran-rules 8318@opindex fcx-fortran-rules 8319Complex multiplication and division follow Fortran rules. Range 8320reduction is done as part of complex division, but there is no checking 8321whether the result of a complex multiplication or division is @code{NaN 8322+ I*NaN}, with an attempt to rescue the situation in that case. 8323 8324The default is @option{-fno-cx-fortran-rules}. 8325 8326@end table 8327 8328The following options control optimizations that may improve 8329performance, but are not enabled by any @option{-O} options. This 8330section includes experimental options that may produce broken code. 8331 8332@table @gcctabopt 8333@item -fbranch-probabilities 8334@opindex fbranch-probabilities 8335After running a program compiled with @option{-fprofile-arcs} 8336(@pxref{Debugging Options,, Options for Debugging Your Program or 8337@command{gcc}}), you can compile it a second time using 8338@option{-fbranch-probabilities}, to improve optimizations based on 8339the number of times each branch was taken. When the program 8340compiled with @option{-fprofile-arcs} exits it saves arc execution 8341counts to a file called @file{@var{sourcename}.gcda} for each source 8342file. The information in this data file is very dependent on the 8343structure of the generated code, so you must use the same source code 8344and the same optimization options for both compilations. 8345 8346With @option{-fbranch-probabilities}, GCC puts a 8347@samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}. 8348These can be used to improve optimization. Currently, they are only 8349used in one place: in @file{reorg.c}, instead of guessing which path a 8350branch is most likely to take, the @samp{REG_BR_PROB} values are used to 8351exactly determine which path is taken more often. 8352 8353@item -fprofile-values 8354@opindex fprofile-values 8355If combined with @option{-fprofile-arcs}, it adds code so that some 8356data about values of expressions in the program is gathered. 8357 8358With @option{-fbranch-probabilities}, it reads back the data gathered 8359from profiling values of expressions for usage in optimizations. 8360 8361Enabled with @option{-fprofile-generate} and @option{-fprofile-use}. 8362 8363@item -fvpt 8364@opindex fvpt 8365If combined with @option{-fprofile-arcs}, it instructs the compiler to add 8366a code to gather information about values of expressions. 8367 8368With @option{-fbranch-probabilities}, it reads back the data gathered 8369and actually performs the optimizations based on them. 8370Currently the optimizations include specialization of division operation 8371using the knowledge about the value of the denominator. 8372 8373@item -frename-registers 8374@opindex frename-registers 8375Attempt to avoid false dependencies in scheduled code by making use 8376of registers left over after register allocation. This optimization 8377will most benefit processors with lots of registers. Depending on the 8378debug information format adopted by the target, however, it can 8379make debugging impossible, since variables will no longer stay in 8380a ``home register''. 8381 8382Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}. 8383 8384@item -ftracer 8385@opindex ftracer 8386Perform tail duplication to enlarge superblock size. This transformation 8387simplifies the control flow of the function allowing other optimizations to do 8388better job. 8389 8390Enabled with @option{-fprofile-use}. 8391 8392@item -funroll-loops 8393@opindex funroll-loops 8394Unroll loops whose number of iterations can be determined at compile time or 8395upon entry to the loop. @option{-funroll-loops} implies 8396@option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}. 8397It also turns on complete loop peeling (i.e.@: complete removal of loops with 8398small constant number of iterations). This option makes code larger, and may 8399or may not make it run faster. 8400 8401Enabled with @option{-fprofile-use}. 8402 8403@item -funroll-all-loops 8404@opindex funroll-all-loops 8405Unroll all loops, even if their number of iterations is uncertain when 8406the loop is entered. This usually makes programs run more slowly. 8407@option{-funroll-all-loops} implies the same options as 8408@option{-funroll-loops}. 8409 8410@item -fpeel-loops 8411@opindex fpeel-loops 8412Peels loops for which there is enough information that they do not 8413roll much (from profile feedback). It also turns on complete loop peeling 8414(i.e.@: complete removal of loops with small constant number of iterations). 8415 8416Enabled with @option{-fprofile-use}. 8417 8418@item -fmove-loop-invariants 8419@opindex fmove-loop-invariants 8420Enables the loop invariant motion pass in the RTL loop optimizer. Enabled 8421at level @option{-O1} 8422 8423@item -funswitch-loops 8424@opindex funswitch-loops 8425Move branches with loop invariant conditions out of the loop, with duplicates 8426of the loop on both branches (modified according to result of the condition). 8427 8428@item -ffunction-sections 8429@itemx -fdata-sections 8430@opindex ffunction-sections 8431@opindex fdata-sections 8432Place each function or data item into its own section in the output 8433file if the target supports arbitrary sections. The name of the 8434function or the name of the data item determines the section's name 8435in the output file. 8436 8437Use these options on systems where the linker can perform optimizations 8438to improve locality of reference in the instruction space. Most systems 8439using the ELF object format and SPARC processors running Solaris 2 have 8440linkers with such optimizations. AIX may have these optimizations in 8441the future. 8442 8443Only use these options when there are significant benefits from doing 8444so. When you specify these options, the assembler and linker will 8445create larger object and executable files and will also be slower. 8446You will not be able to use @code{gprof} on all systems if you 8447specify this option and you may have problems with debugging if 8448you specify both this option and @option{-g}. 8449 8450@item -fbranch-target-load-optimize 8451@opindex fbranch-target-load-optimize 8452Perform branch target register load optimization before prologue / epilogue 8453threading. 8454The use of target registers can typically be exposed only during reload, 8455thus hoisting loads out of loops and doing inter-block scheduling needs 8456a separate optimization pass. 8457 8458@item -fbranch-target-load-optimize2 8459@opindex fbranch-target-load-optimize2 8460Perform branch target register load optimization after prologue / epilogue 8461threading. 8462 8463@item -fbtr-bb-exclusive 8464@opindex fbtr-bb-exclusive 8465When performing branch target register load optimization, don't reuse 8466branch target registers in within any basic block. 8467 8468@item -fstack-protector 8469@opindex fstack-protector 8470Emit extra code to check for buffer overflows, such as stack smashing 8471attacks. This is done by adding a guard variable to functions with 8472vulnerable objects. This includes functions that call alloca, and 8473functions with buffers larger than 8 bytes. The guards are initialized 8474when a function is entered and then checked when the function exits. 8475If a guard check fails, an error message is printed and the program exits. 8476 8477@item -fstack-protector-all 8478@opindex fstack-protector-all 8479Like @option{-fstack-protector} except that all functions are protected. 8480 8481@item -fsection-anchors 8482@opindex fsection-anchors 8483Try to reduce the number of symbolic address calculations by using 8484shared ``anchor'' symbols to address nearby objects. This transformation 8485can help to reduce the number of GOT entries and GOT accesses on some 8486targets. 8487 8488For example, the implementation of the following function @code{foo}: 8489 8490@smallexample 8491static int a, b, c; 8492int foo (void) @{ return a + b + c; @} 8493@end smallexample 8494 8495would usually calculate the addresses of all three variables, but if you 8496compile it with @option{-fsection-anchors}, it will access the variables 8497from a common anchor point instead. The effect is similar to the 8498following pseudocode (which isn't valid C): 8499 8500@smallexample 8501int foo (void) 8502@{ 8503 register int *xr = &x; 8504 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x]; 8505@} 8506@end smallexample 8507 8508Not all targets support this option. 8509 8510@item --param @var{name}=@var{value} 8511@opindex param 8512In some places, GCC uses various constants to control the amount of 8513optimization that is done. For example, GCC will not inline functions 8514that contain more than a certain number of instructions. You can 8515control some of these constants on the command line using the 8516@option{--param} option. 8517 8518The names of specific parameters, and the meaning of the values, are 8519tied to the internals of the compiler, and are subject to change 8520without notice in future releases. 8521 8522In each case, the @var{value} is an integer. The allowable choices for 8523@var{name} are given in the following table: 8524 8525@table @gcctabopt 8526@item predictable-branch-outcome 8527When branch is predicted to be taken with probability lower than this threshold 8528(in percent), then it is considered well predictable. The default is 10. 8529 8530@item max-crossjump-edges 8531The maximum number of incoming edges to consider for crossjumping. 8532The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in 8533the number of edges incoming to each block. Increasing values mean 8534more aggressive optimization, making the compilation time increase with 8535probably small improvement in executable size. 8536 8537@item min-crossjump-insns 8538The minimum number of instructions that must be matched at the end 8539of two blocks before crossjumping will be performed on them. This 8540value is ignored in the case where all instructions in the block being 8541crossjumped from are matched. The default value is 5. 8542 8543@item max-grow-copy-bb-insns 8544The maximum code size expansion factor when copying basic blocks 8545instead of jumping. The expansion is relative to a jump instruction. 8546The default value is 8. 8547 8548@item max-goto-duplication-insns 8549The maximum number of instructions to duplicate to a block that jumps 8550to a computed goto. To avoid @math{O(N^2)} behavior in a number of 8551passes, GCC factors computed gotos early in the compilation process, 8552and unfactors them as late as possible. Only computed jumps at the 8553end of a basic blocks with no more than max-goto-duplication-insns are 8554unfactored. The default value is 8. 8555 8556@item max-delay-slot-insn-search 8557The maximum number of instructions to consider when looking for an 8558instruction to fill a delay slot. If more than this arbitrary number of 8559instructions is searched, the time savings from filling the delay slot 8560will be minimal so stop searching. Increasing values mean more 8561aggressive optimization, making the compilation time increase with probably 8562small improvement in execution time. 8563 8564@item max-delay-slot-live-search 8565When trying to fill delay slots, the maximum number of instructions to 8566consider when searching for a block with valid live register 8567information. Increasing this arbitrarily chosen value means more 8568aggressive optimization, increasing the compilation time. This parameter 8569should be removed when the delay slot code is rewritten to maintain the 8570control-flow graph. 8571 8572@item max-gcse-memory 8573The approximate maximum amount of memory that will be allocated in 8574order to perform the global common subexpression elimination 8575optimization. If more memory than specified is required, the 8576optimization will not be done. 8577 8578@item max-gcse-insertion-ratio 8579If the ratio of expression insertions to deletions is larger than this value 8580for any expression, then RTL PRE will insert or remove the expression and thus 8581leave partially redundant computations in the instruction stream. The default value is 20. 8582 8583@item max-pending-list-length 8584The maximum number of pending dependencies scheduling will allow 8585before flushing the current state and starting over. Large functions 8586with few branches or calls can create excessively large lists which 8587needlessly consume memory and resources. 8588 8589@item max-modulo-backtrack-attempts 8590The maximum number of backtrack attempts the scheduler should make 8591when modulo scheduling a loop. Larger values can exponentially increase 8592compilation time. 8593 8594@item max-inline-insns-single 8595Several parameters control the tree inliner used in gcc. 8596This number sets the maximum number of instructions (counted in GCC's 8597internal representation) in a single function that the tree inliner 8598will consider for inlining. This only affects functions declared 8599inline and methods implemented in a class declaration (C++). 8600The default value is 400. 8601 8602@item max-inline-insns-auto 8603When you use @option{-finline-functions} (included in @option{-O3}), 8604a lot of functions that would otherwise not be considered for inlining 8605by the compiler will be investigated. To those functions, a different 8606(more restrictive) limit compared to functions declared inline can 8607be applied. 8608The default value is 40. 8609 8610@item large-function-insns 8611The limit specifying really large functions. For functions larger than this 8612limit after inlining, inlining is constrained by 8613@option{--param large-function-growth}. This parameter is useful primarily 8614to avoid extreme compilation time caused by non-linear algorithms used by the 8615back end. 8616The default value is 2700. 8617 8618@item large-function-growth 8619Specifies maximal growth of large function caused by inlining in percents. 8620The default value is 100 which limits large function growth to 2.0 times 8621the original size. 8622 8623@item large-unit-insns 8624The limit specifying large translation unit. Growth caused by inlining of 8625units larger than this limit is limited by @option{--param inline-unit-growth}. 8626For small units this might be too tight (consider unit consisting of function A 8627that is inline and B that just calls A three time. If B is small relative to 8628A, the growth of unit is 300\% and yet such inlining is very sane. For very 8629large units consisting of small inlineable functions however the overall unit 8630growth limit is needed to avoid exponential explosion of code size. Thus for 8631smaller units, the size is increased to @option{--param large-unit-insns} 8632before applying @option{--param inline-unit-growth}. The default is 10000 8633 8634@item inline-unit-growth 8635Specifies maximal overall growth of the compilation unit caused by inlining. 8636The default value is 30 which limits unit growth to 1.3 times the original 8637size. 8638 8639@item ipcp-unit-growth 8640Specifies maximal overall growth of the compilation unit caused by 8641interprocedural constant propagation. The default value is 10 which limits 8642unit growth to 1.1 times the original size. 8643 8644@item large-stack-frame 8645The limit specifying large stack frames. While inlining the algorithm is trying 8646to not grow past this limit too much. Default value is 256 bytes. 8647 8648@item large-stack-frame-growth 8649Specifies maximal growth of large stack frames caused by inlining in percents. 8650The default value is 1000 which limits large stack frame growth to 11 times 8651the original size. 8652 8653@item max-inline-insns-recursive 8654@itemx max-inline-insns-recursive-auto 8655Specifies maximum number of instructions out-of-line copy of self recursive inline 8656function can grow into by performing recursive inlining. 8657 8658For functions declared inline @option{--param max-inline-insns-recursive} is 8659taken into account. For function not declared inline, recursive inlining 8660happens only when @option{-finline-functions} (included in @option{-O3}) is 8661enabled and @option{--param max-inline-insns-recursive-auto} is used. The 8662default value is 450. 8663 8664@item max-inline-recursive-depth 8665@itemx max-inline-recursive-depth-auto 8666Specifies maximum recursion depth used by the recursive inlining. 8667 8668For functions declared inline @option{--param max-inline-recursive-depth} is 8669taken into account. For function not declared inline, recursive inlining 8670happens only when @option{-finline-functions} (included in @option{-O3}) is 8671enabled and @option{--param max-inline-recursive-depth-auto} is used. The 8672default value is 8. 8673 8674@item min-inline-recursive-probability 8675Recursive inlining is profitable only for function having deep recursion 8676in average and can hurt for function having little recursion depth by 8677increasing the prologue size or complexity of function body to other 8678optimizers. 8679 8680When profile feedback is available (see @option{-fprofile-generate}) the actual 8681recursion depth can be guessed from probability that function will recurse via 8682given call expression. This parameter limits inlining only to call expression 8683whose probability exceeds given threshold (in percents). The default value is 868410. 8685 8686@item early-inlining-insns 8687Specify growth that early inliner can make. In effect it increases amount of 8688inlining for code having large abstraction penalty. The default value is 10. 8689 8690@item max-early-inliner-iterations 8691@itemx max-early-inliner-iterations 8692Limit of iterations of early inliner. This basically bounds number of nested 8693indirect calls early inliner can resolve. Deeper chains are still handled by 8694late inlining. 8695 8696@item comdat-sharing-probability 8697@itemx comdat-sharing-probability 8698Probability (in percent) that C++ inline function with comdat visibility 8699will be shared across multiple compilation units. The default value is 20. 8700 8701@item min-vect-loop-bound 8702The minimum number of iterations under which a loop will not get vectorized 8703when @option{-ftree-vectorize} is used. The number of iterations after 8704vectorization needs to be greater than the value specified by this option 8705to allow vectorization. The default value is 0. 8706 8707@item gcse-cost-distance-ratio 8708Scaling factor in calculation of maximum distance an expression 8709can be moved by GCSE optimizations. This is currently supported only in the 8710code hoisting pass. The bigger the ratio, the more aggressive code hoisting 8711will be with simple expressions, i.e., the expressions that have cost 8712less than @option{gcse-unrestricted-cost}. Specifying 0 will disable 8713hoisting of simple expressions. The default value is 10. 8714 8715@item gcse-unrestricted-cost 8716Cost, roughly measured as the cost of a single typical machine 8717instruction, at which GCSE optimizations will not constrain 8718the distance an expression can travel. This is currently 8719supported only in the code hoisting pass. The lesser the cost, 8720the more aggressive code hoisting will be. Specifying 0 will 8721allow all expressions to travel unrestricted distances. 8722The default value is 3. 8723 8724@item max-hoist-depth 8725The depth of search in the dominator tree for expressions to hoist. 8726This is used to avoid quadratic behavior in hoisting algorithm. 8727The value of 0 will avoid limiting the search, but may slow down compilation 8728of huge functions. The default value is 30. 8729 8730@item max-tail-merge-comparisons 8731The maximum amount of similar bbs to compare a bb with. This is used to 8732avoid quadratic behavior in tree tail merging. The default value is 10. 8733 8734@item max-tail-merge-iterations 8735The maximum amount of iterations of the pass over the function. This is used to 8736limit compilation time in tree tail merging. The default value is 2. 8737 8738@item max-unrolled-insns 8739The maximum number of instructions that a loop should have if that loop 8740is unrolled, and if the loop is unrolled, it determines how many times 8741the loop code is unrolled. 8742 8743@item max-average-unrolled-insns 8744The maximum number of instructions biased by probabilities of their execution 8745that a loop should have if that loop is unrolled, and if the loop is unrolled, 8746it determines how many times the loop code is unrolled. 8747 8748@item max-unroll-times 8749The maximum number of unrollings of a single loop. 8750 8751@item max-peeled-insns 8752The maximum number of instructions that a loop should have if that loop 8753is peeled, and if the loop is peeled, it determines how many times 8754the loop code is peeled. 8755 8756@item max-peel-times 8757The maximum number of peelings of a single loop. 8758 8759@item max-completely-peeled-insns 8760The maximum number of insns of a completely peeled loop. 8761 8762@item max-completely-peel-times 8763The maximum number of iterations of a loop to be suitable for complete peeling. 8764 8765@item max-completely-peel-loop-nest-depth 8766The maximum depth of a loop nest suitable for complete peeling. 8767 8768@item max-unswitch-insns 8769The maximum number of insns of an unswitched loop. 8770 8771@item max-unswitch-level 8772The maximum number of branches unswitched in a single loop. 8773 8774@item lim-expensive 8775The minimum cost of an expensive expression in the loop invariant motion. 8776 8777@item iv-consider-all-candidates-bound 8778Bound on number of candidates for induction variables below that 8779all candidates are considered for each use in induction variable 8780optimizations. Only the most relevant candidates are considered 8781if there are more candidates, to avoid quadratic time complexity. 8782 8783@item iv-max-considered-uses 8784The induction variable optimizations give up on loops that contain more 8785induction variable uses. 8786 8787@item iv-always-prune-cand-set-bound 8788If number of candidates in the set is smaller than this value, 8789we always try to remove unnecessary ivs from the set during its 8790optimization when a new iv is added to the set. 8791 8792@item scev-max-expr-size 8793Bound on size of expressions used in the scalar evolutions analyzer. 8794Large expressions slow the analyzer. 8795 8796@item scev-max-expr-complexity 8797Bound on the complexity of the expressions in the scalar evolutions analyzer. 8798Complex expressions slow the analyzer. 8799 8800@item omega-max-vars 8801The maximum number of variables in an Omega constraint system. 8802The default value is 128. 8803 8804@item omega-max-geqs 8805The maximum number of inequalities in an Omega constraint system. 8806The default value is 256. 8807 8808@item omega-max-eqs 8809The maximum number of equalities in an Omega constraint system. 8810The default value is 128. 8811 8812@item omega-max-wild-cards 8813The maximum number of wildcard variables that the Omega solver will 8814be able to insert. The default value is 18. 8815 8816@item omega-hash-table-size 8817The size of the hash table in the Omega solver. The default value is 8818550. 8819 8820@item omega-max-keys 8821The maximal number of keys used by the Omega solver. The default 8822value is 500. 8823 8824@item omega-eliminate-redundant-constraints 8825When set to 1, use expensive methods to eliminate all redundant 8826constraints. The default value is 0. 8827 8828@item vect-max-version-for-alignment-checks 8829The maximum number of run-time checks that can be performed when 8830doing loop versioning for alignment in the vectorizer. See option 8831ftree-vect-loop-version for more information. 8832 8833@item vect-max-version-for-alias-checks 8834The maximum number of run-time checks that can be performed when 8835doing loop versioning for alias in the vectorizer. See option 8836ftree-vect-loop-version for more information. 8837 8838@item max-iterations-to-track 8839 8840The maximum number of iterations of a loop the brute force algorithm 8841for analysis of # of iterations of the loop tries to evaluate. 8842 8843@item hot-bb-count-fraction 8844Select fraction of the maximal count of repetitions of basic block in program 8845given basic block needs to have to be considered hot. 8846 8847@item hot-bb-frequency-fraction 8848Select fraction of the entry block frequency of executions of basic block in 8849function given basic block needs to have to be considered hot. 8850 8851@item max-predicted-iterations 8852The maximum number of loop iterations we predict statically. This is useful 8853in cases where function contain single loop with known bound and other loop 8854with unknown. We predict the known number of iterations correctly, while 8855the unknown number of iterations average to roughly 10. This means that the 8856loop without bounds would appear artificially cold relative to the other one. 8857 8858@item align-threshold 8859 8860Select fraction of the maximal frequency of executions of basic block in 8861function given basic block will get aligned. 8862 8863@item align-loop-iterations 8864 8865A loop expected to iterate at lest the selected number of iterations will get 8866aligned. 8867 8868@item tracer-dynamic-coverage 8869@itemx tracer-dynamic-coverage-feedback 8870 8871This value is used to limit superblock formation once the given percentage of 8872executed instructions is covered. This limits unnecessary code size 8873expansion. 8874 8875The @option{tracer-dynamic-coverage-feedback} is used only when profile 8876feedback is available. The real profiles (as opposed to statically estimated 8877ones) are much less balanced allowing the threshold to be larger value. 8878 8879@item tracer-max-code-growth 8880Stop tail duplication once code growth has reached given percentage. This is 8881rather hokey argument, as most of the duplicates will be eliminated later in 8882cross jumping, so it may be set to much higher values than is the desired code 8883growth. 8884 8885@item tracer-min-branch-ratio 8886 8887Stop reverse growth when the reverse probability of best edge is less than this 8888threshold (in percent). 8889 8890@item tracer-min-branch-ratio 8891@itemx tracer-min-branch-ratio-feedback 8892 8893Stop forward growth if the best edge do have probability lower than this 8894threshold. 8895 8896Similarly to @option{tracer-dynamic-coverage} two values are present, one for 8897compilation for profile feedback and one for compilation without. The value 8898for compilation with profile feedback needs to be more conservative (higher) in 8899order to make tracer effective. 8900 8901@item max-cse-path-length 8902 8903Maximum number of basic blocks on path that cse considers. The default is 10. 8904 8905@item max-cse-insns 8906The maximum instructions CSE process before flushing. The default is 1000. 8907 8908@item ggc-min-expand 8909 8910GCC uses a garbage collector to manage its own memory allocation. This 8911parameter specifies the minimum percentage by which the garbage 8912collector's heap should be allowed to expand between collections. 8913Tuning this may improve compilation speed; it has no effect on code 8914generation. 8915 8916The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when 8917RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is 8918the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If 8919GCC is not able to calculate RAM on a particular platform, the lower 8920bound of 30% is used. Setting this parameter and 8921@option{ggc-min-heapsize} to zero causes a full collection to occur at 8922every opportunity. This is extremely slow, but can be useful for 8923debugging. 8924 8925@item ggc-min-heapsize 8926 8927Minimum size of the garbage collector's heap before it begins bothering 8928to collect garbage. The first collection occurs after the heap expands 8929by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again, 8930tuning this may improve compilation speed, and has no effect on code 8931generation. 8932 8933The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that 8934tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but 8935with a lower bound of 4096 (four megabytes) and an upper bound of 8936131072 (128 megabytes). If GCC is not able to calculate RAM on a 8937particular platform, the lower bound is used. Setting this parameter 8938very large effectively disables garbage collection. Setting this 8939parameter and @option{ggc-min-expand} to zero causes a full collection 8940to occur at every opportunity. 8941 8942@item max-reload-search-insns 8943The maximum number of instruction reload should look backward for equivalent 8944register. Increasing values mean more aggressive optimization, making the 8945compilation time increase with probably slightly better performance. 8946The default value is 100. 8947 8948@item max-cselib-memory-locations 8949The maximum number of memory locations cselib should take into account. 8950Increasing values mean more aggressive optimization, making the compilation time 8951increase with probably slightly better performance. The default value is 500. 8952 8953@item reorder-blocks-duplicate 8954@itemx reorder-blocks-duplicate-feedback 8955 8956Used by basic block reordering pass to decide whether to use unconditional 8957branch or duplicate the code on its destination. Code is duplicated when its 8958estimated size is smaller than this value multiplied by the estimated size of 8959unconditional jump in the hot spots of the program. 8960 8961The @option{reorder-block-duplicate-feedback} is used only when profile 8962feedback is available and may be set to higher values than 8963@option{reorder-block-duplicate} since information about the hot spots is more 8964accurate. 8965 8966@item max-sched-ready-insns 8967The maximum number of instructions ready to be issued the scheduler should 8968consider at any given time during the first scheduling pass. Increasing 8969values mean more thorough searches, making the compilation time increase 8970with probably little benefit. The default value is 100. 8971 8972@item max-sched-region-blocks 8973The maximum number of blocks in a region to be considered for 8974interblock scheduling. The default value is 10. 8975 8976@item max-pipeline-region-blocks 8977The maximum number of blocks in a region to be considered for 8978pipelining in the selective scheduler. The default value is 15. 8979 8980@item max-sched-region-insns 8981The maximum number of insns in a region to be considered for 8982interblock scheduling. The default value is 100. 8983 8984@item max-pipeline-region-insns 8985The maximum number of insns in a region to be considered for 8986pipelining in the selective scheduler. The default value is 200. 8987 8988@item min-spec-prob 8989The minimum probability (in percents) of reaching a source block 8990for interblock speculative scheduling. The default value is 40. 8991 8992@item max-sched-extend-regions-iters 8993The maximum number of iterations through CFG to extend regions. 89940 - disable region extension, 8995N - do at most N iterations. 8996The default value is 0. 8997 8998@item max-sched-insn-conflict-delay 8999The maximum conflict delay for an insn to be considered for speculative motion. 9000The default value is 3. 9001 9002@item sched-spec-prob-cutoff 9003The minimal probability of speculation success (in percents), so that 9004speculative insn will be scheduled. 9005The default value is 40. 9006 9007@item sched-mem-true-dep-cost 9008Minimal distance (in CPU cycles) between store and load targeting same 9009memory locations. The default value is 1. 9010 9011@item selsched-max-lookahead 9012The maximum size of the lookahead window of selective scheduling. It is a 9013depth of search for available instructions. 9014The default value is 50. 9015 9016@item selsched-max-sched-times 9017The maximum number of times that an instruction will be scheduled during 9018selective scheduling. This is the limit on the number of iterations 9019through which the instruction may be pipelined. The default value is 2. 9020 9021@item selsched-max-insns-to-rename 9022The maximum number of best instructions in the ready list that are considered 9023for renaming in the selective scheduler. The default value is 2. 9024 9025@item sms-min-sc 9026The minimum value of stage count that swing modulo scheduler will 9027generate. The default value is 2. 9028 9029@item max-last-value-rtl 9030The maximum size measured as number of RTLs that can be recorded in an expression 9031in combiner for a pseudo register as last known value of that register. The default 9032is 10000. 9033 9034@item integer-share-limit 9035Small integer constants can use a shared data structure, reducing the 9036compiler's memory usage and increasing its speed. This sets the maximum 9037value of a shared integer constant. The default value is 256. 9038 9039@item min-virtual-mappings 9040Specifies the minimum number of virtual mappings in the incremental 9041SSA updater that should be registered to trigger the virtual mappings 9042heuristic defined by virtual-mappings-ratio. The default value is 9043100. 9044 9045@item virtual-mappings-ratio 9046If the number of virtual mappings is virtual-mappings-ratio bigger 9047than the number of virtual symbols to be updated, then the incremental 9048SSA updater switches to a full update for those symbols. The default 9049ratio is 3. 9050 9051@item ssp-buffer-size 9052The minimum size of buffers (i.e.@: arrays) that will receive stack smashing 9053protection when @option{-fstack-protection} is used. 9054 9055@item max-jump-thread-duplication-stmts 9056Maximum number of statements allowed in a block that needs to be 9057duplicated when threading jumps. 9058 9059@item max-fields-for-field-sensitive 9060Maximum number of fields in a structure we will treat in 9061a field sensitive manner during pointer analysis. The default is zero 9062for -O0, and -O1 and 100 for -Os, -O2, and -O3. 9063 9064@item prefetch-latency 9065Estimate on average number of instructions that are executed before 9066prefetch finishes. The distance we prefetch ahead is proportional 9067to this constant. Increasing this number may also lead to less 9068streams being prefetched (see @option{simultaneous-prefetches}). 9069 9070@item simultaneous-prefetches 9071Maximum number of prefetches that can run at the same time. 9072 9073@item l1-cache-line-size 9074The size of cache line in L1 cache, in bytes. 9075 9076@item l1-cache-size 9077The size of L1 cache, in kilobytes. 9078 9079@item l2-cache-size 9080The size of L2 cache, in kilobytes. 9081 9082@item min-insn-to-prefetch-ratio 9083The minimum ratio between the number of instructions and the 9084number of prefetches to enable prefetching in a loop. 9085 9086@item prefetch-min-insn-to-mem-ratio 9087The minimum ratio between the number of instructions and the 9088number of memory references to enable prefetching in a loop. 9089 9090@item use-canonical-types 9091Whether the compiler should use the ``canonical'' type system. By 9092default, this should always be 1, which uses a more efficient internal 9093mechanism for comparing types in C++ and Objective-C++. However, if 9094bugs in the canonical type system are causing compilation failures, 9095set this value to 0 to disable canonical types. 9096 9097@item switch-conversion-max-branch-ratio 9098Switch initialization conversion will refuse to create arrays that are 9099bigger than @option{switch-conversion-max-branch-ratio} times the number of 9100branches in the switch. 9101 9102@item max-partial-antic-length 9103Maximum length of the partial antic set computed during the tree 9104partial redundancy elimination optimization (@option{-ftree-pre}) when 9105optimizing at @option{-O3} and above. For some sorts of source code 9106the enhanced partial redundancy elimination optimization can run away, 9107consuming all of the memory available on the host machine. This 9108parameter sets a limit on the length of the sets that are computed, 9109which prevents the runaway behavior. Setting a value of 0 for 9110this parameter will allow an unlimited set length. 9111 9112@item sccvn-max-scc-size 9113Maximum size of a strongly connected component (SCC) during SCCVN 9114processing. If this limit is hit, SCCVN processing for the whole 9115function will not be done and optimizations depending on it will 9116be disabled. The default maximum SCC size is 10000. 9117 9118@item ira-max-loops-num 9119IRA uses regional register allocation by default. If a function 9120contains more loops than the number given by this parameter, only at most 9121the given number of the most frequently-executed loops form regions 9122for regional register allocation. The default value of the 9123parameter is 100. 9124 9125@item ira-max-conflict-table-size 9126Although IRA uses a sophisticated algorithm to compress the conflict 9127table, the table can still require excessive amounts of memory for 9128huge functions. If the conflict table for a function could be more 9129than the size in MB given by this parameter, the register allocator 9130instead uses a faster, simpler, and lower-quality 9131algorithm that does not require building a pseudo-register conflict table. 9132The default value of the parameter is 2000. 9133 9134@item ira-loop-reserved-regs 9135IRA can be used to evaluate more accurate register pressure in loops 9136for decisions to move loop invariants (see @option{-O3}). The number 9137of available registers reserved for some other purposes is given 9138by this parameter. The default value of the parameter is 2, which is 9139the minimal number of registers needed by typical instructions. 9140This value is the best found from numerous experiments. 9141 9142@item loop-invariant-max-bbs-in-loop 9143Loop invariant motion can be very expensive, both in compilation time and 9144in amount of needed compile-time memory, with very large loops. Loops 9145with more basic blocks than this parameter won't have loop invariant 9146motion optimization performed on them. The default value of the 9147parameter is 1000 for -O1 and 10000 for -O2 and above. 9148 9149@item loop-max-datarefs-for-datadeps 9150Building data dapendencies is expensive for very large loops. This 9151parameter limits the number of data references in loops that are 9152considered for data dependence analysis. These large loops will not 9153be handled then by the optimizations using loop data dependencies. 9154The default value is 1000. 9155 9156@item max-vartrack-size 9157Sets a maximum number of hash table slots to use during variable 9158tracking dataflow analysis of any function. If this limit is exceeded 9159with variable tracking at assignments enabled, analysis for that 9160function is retried without it, after removing all debug insns from 9161the function. If the limit is exceeded even without debug insns, var 9162tracking analysis is completely disabled for the function. Setting 9163the parameter to zero makes it unlimited. 9164 9165@item max-vartrack-expr-depth 9166Sets a maximum number of recursion levels when attempting to map 9167variable names or debug temporaries to value expressions. This trades 9168compilation time for more complete debug information. If this is set too 9169low, value expressions that are available and could be represented in 9170debug information may end up not being used; setting this higher may 9171enable the compiler to find more complex debug expressions, but compile 9172time and memory use may grow. The default is 12. 9173 9174@item min-nondebug-insn-uid 9175Use uids starting at this parameter for nondebug insns. The range below 9176the parameter is reserved exclusively for debug insns created by 9177@option{-fvar-tracking-assignments}, but debug insns may get 9178(non-overlapping) uids above it if the reserved range is exhausted. 9179 9180@item ipa-sra-ptr-growth-factor 9181IPA-SRA will replace a pointer to an aggregate with one or more new 9182parameters only when their cumulative size is less or equal to 9183@option{ipa-sra-ptr-growth-factor} times the size of the original 9184pointer parameter. 9185 9186@item tm-max-aggregate-size 9187When making copies of thread-local variables in a transaction, this 9188parameter specifies the size in bytes after which variables will be 9189saved with the logging functions as opposed to save/restore code 9190sequence pairs. This option only applies when using 9191@option{-fgnu-tm}. 9192 9193@item graphite-max-nb-scop-params 9194To avoid exponential effects in the Graphite loop transforms, the 9195number of parameters in a Static Control Part (SCoP) is bounded. The 9196default value is 10 parameters. A variable whose value is unknown at 9197compilation time and defined outside a SCoP is a parameter of the SCoP. 9198 9199@item graphite-max-bbs-per-function 9200To avoid exponential effects in the detection of SCoPs, the size of 9201the functions analyzed by Graphite is bounded. The default value is 9202100 basic blocks. 9203 9204@item loop-block-tile-size 9205Loop blocking or strip mining transforms, enabled with 9206@option{-floop-block} or @option{-floop-strip-mine}, strip mine each 9207loop in the loop nest by a given number of iterations. The strip 9208length can be changed using the @option{loop-block-tile-size} 9209parameter. The default value is 51 iterations. 9210 9211@item ipa-cp-value-list-size 9212IPA-CP attempts to track all possible values and types passed to a function's 9213parameter in order to propagate them and perform devirtualization. 9214@option{ipa-cp-value-list-size} is the maximum number of values and types it 9215stores per one formal parameter of a function. 9216 9217@item lto-partitions 9218Specify desired number of partitions produced during WHOPR compilation. 9219The number of partitions should exceed the number of CPUs used for compilation. 9220The default value is 32. 9221 9222@item lto-minpartition 9223Size of minimal partition for WHOPR (in estimated instructions). 9224This prevents expenses of splitting very small programs into too many 9225partitions. 9226 9227@item cxx-max-namespaces-for-diagnostic-help 9228The maximum number of namespaces to consult for suggestions when C++ 9229name lookup fails for an identifier. The default is 1000. 9230 9231@item sink-frequency-threshold 9232The maximum relative execution frequency (in percents) of the target block 9233relative to a statement's original block to allow statement sinking of a 9234statement. Larger numbers result in more aggressive statement sinking. 9235The default value is 75. A small positive adjustment is applied for 9236statements with memory operands as those are even more profitable so sink. 9237 9238@item max-stores-to-sink 9239The maximum number of conditional stores paires that can be sunk. Set to 0 9240if either vectorization (@option{-ftree-vectorize}) or if-conversion 9241(@option{-ftree-loop-if-convert}) is disabled. The default is 2. 9242 9243@item allow-load-data-races 9244Allow optimizers to introduce new data races on loads. 9245Set to 1 to allow, otherwise to 0. This option is enabled by default 9246unless implicitly set by the @option{-fmemory-model=} option. 9247 9248@item allow-store-data-races 9249Allow optimizers to introduce new data races on stores. 9250Set to 1 to allow, otherwise to 0. This option is enabled by default 9251unless implicitly set by the @option{-fmemory-model=} option. 9252 9253@item allow-packed-load-data-races 9254Allow optimizers to introduce new data races on packed data loads. 9255Set to 1 to allow, otherwise to 0. This option is enabled by default 9256unless implicitly set by the @option{-fmemory-model=} option. 9257 9258@item allow-packed-store-data-races 9259Allow optimizers to introduce new data races on packed data stores. 9260Set to 1 to allow, otherwise to 0. This option is enabled by default 9261unless implicitly set by the @option{-fmemory-model=} option. 9262 9263@item case-values-threshold 9264The smallest number of different values for which it is best to use a 9265jump-table instead of a tree of conditional branches. If the value is 92660, use the default for the machine. The default is 0. 9267 9268@item tree-reassoc-width 9269Set the maximum number of instructions executed in parallel in 9270reassociated tree. This parameter overrides target dependent 9271heuristics used by default if has non zero value. 9272 9273@end table 9274@end table 9275 9276@node Preprocessor Options 9277@section Options Controlling the Preprocessor 9278@cindex preprocessor options 9279@cindex options, preprocessor 9280 9281These options control the C preprocessor, which is run on each C source 9282file before actual compilation. 9283 9284If you use the @option{-E} option, nothing is done except preprocessing. 9285Some of these options make sense only together with @option{-E} because 9286they cause the preprocessor output to be unsuitable for actual 9287compilation. 9288 9289@table @gcctabopt 9290@item -Wp,@var{option} 9291@opindex Wp 9292You can use @option{-Wp,@var{option}} to bypass the compiler driver 9293and pass @var{option} directly through to the preprocessor. If 9294@var{option} contains commas, it is split into multiple options at the 9295commas. However, many options are modified, translated or interpreted 9296by the compiler driver before being passed to the preprocessor, and 9297@option{-Wp} forcibly bypasses this phase. The preprocessor's direct 9298interface is undocumented and subject to change, so whenever possible 9299you should avoid using @option{-Wp} and let the driver handle the 9300options instead. 9301 9302@item -Xpreprocessor @var{option} 9303@opindex Xpreprocessor 9304Pass @var{option} as an option to the preprocessor. You can use this to 9305supply system-specific preprocessor options that GCC does not know how to 9306recognize. 9307 9308If you want to pass an option that takes an argument, you must use 9309@option{-Xpreprocessor} twice, once for the option and once for the argument. 9310@end table 9311 9312@include cppopts.texi 9313 9314@node Assembler Options 9315@section Passing Options to the Assembler 9316 9317@c prevent bad page break with this line 9318You can pass options to the assembler. 9319 9320@table @gcctabopt 9321@item -Wa,@var{option} 9322@opindex Wa 9323Pass @var{option} as an option to the assembler. If @var{option} 9324contains commas, it is split into multiple options at the commas. 9325 9326@item -Xassembler @var{option} 9327@opindex Xassembler 9328Pass @var{option} as an option to the assembler. You can use this to 9329supply system-specific assembler options that GCC does not know how to 9330recognize. 9331 9332If you want to pass an option that takes an argument, you must use 9333@option{-Xassembler} twice, once for the option and once for the argument. 9334 9335@end table 9336 9337@node Link Options 9338@section Options for Linking 9339@cindex link options 9340@cindex options, linking 9341 9342These options come into play when the compiler links object files into 9343an executable output file. They are meaningless if the compiler is 9344not doing a link step. 9345 9346@table @gcctabopt 9347@cindex file names 9348@item @var{object-file-name} 9349A file name that does not end in a special recognized suffix is 9350considered to name an object file or library. (Object files are 9351distinguished from libraries by the linker according to the file 9352contents.) If linking is done, these object files are used as input 9353to the linker. 9354 9355@item -c 9356@itemx -S 9357@itemx -E 9358@opindex c 9359@opindex S 9360@opindex E 9361If any of these options is used, then the linker is not run, and 9362object file names should not be used as arguments. @xref{Overall 9363Options}. 9364 9365@cindex Libraries 9366@item -l@var{library} 9367@itemx -l @var{library} 9368@opindex l 9369Search the library named @var{library} when linking. (The second 9370alternative with the library as a separate argument is only for 9371POSIX compliance and is not recommended.) 9372 9373It makes a difference where in the command you write this option; the 9374linker searches and processes libraries and object files in the order they 9375are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z} 9376after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers 9377to functions in @samp{z}, those functions may not be loaded. 9378 9379The linker searches a standard list of directories for the library, 9380which is actually a file named @file{lib@var{library}.a}. The linker 9381then uses this file as if it had been specified precisely by name. 9382 9383The directories searched include several standard system directories 9384plus any that you specify with @option{-L}. 9385 9386Normally the files found this way are library files---archive files 9387whose members are object files. The linker handles an archive file by 9388scanning through it for members which define symbols that have so far 9389been referenced but not defined. But if the file that is found is an 9390ordinary object file, it is linked in the usual fashion. The only 9391difference between using an @option{-l} option and specifying a file name 9392is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a} 9393and searches several directories. 9394 9395@item -lobjc 9396@opindex lobjc 9397You need this special case of the @option{-l} option in order to 9398link an Objective-C or Objective-C++ program. 9399 9400@item -nostartfiles 9401@opindex nostartfiles 9402Do not use the standard system startup files when linking. 9403The standard system libraries are used normally, unless @option{-nostdlib} 9404or @option{-nodefaultlibs} is used. 9405 9406@item -nodefaultlibs 9407@opindex nodefaultlibs 9408Do not use the standard system libraries when linking. 9409Only the libraries you specify will be passed to the linker, options 9410specifying linkage of the system libraries, such as @code{-static-libgcc} 9411or @code{-shared-libgcc}, will be ignored. 9412The standard startup files are used normally, unless @option{-nostartfiles} 9413is used. The compiler may generate calls to @code{memcmp}, 9414@code{memset}, @code{memcpy} and @code{memmove}. 9415These entries are usually resolved by entries in 9416libc. These entry points should be supplied through some other 9417mechanism when this option is specified. 9418 9419@item -nostdlib 9420@opindex nostdlib 9421Do not use the standard system startup files or libraries when linking. 9422No startup files and only the libraries you specify will be passed to 9423the linker, options specifying linkage of the system libraries, such as 9424@code{-static-libgcc} or @code{-shared-libgcc}, will be ignored. 9425The compiler may generate calls to @code{memcmp}, @code{memset}, 9426@code{memcpy} and @code{memmove}. 9427These entries are usually resolved by entries in 9428libc. These entry points should be supplied through some other 9429mechanism when this option is specified. 9430 9431@cindex @option{-lgcc}, use with @option{-nostdlib} 9432@cindex @option{-nostdlib} and unresolved references 9433@cindex unresolved references and @option{-nostdlib} 9434@cindex @option{-lgcc}, use with @option{-nodefaultlibs} 9435@cindex @option{-nodefaultlibs} and unresolved references 9436@cindex unresolved references and @option{-nodefaultlibs} 9437One of the standard libraries bypassed by @option{-nostdlib} and 9438@option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines 9439which GCC uses to overcome shortcomings of particular machines, or special 9440needs for some languages. 9441(@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler 9442Collection (GCC) Internals}, 9443for more discussion of @file{libgcc.a}.) 9444In most cases, you need @file{libgcc.a} even when you want to avoid 9445other standard libraries. In other words, when you specify @option{-nostdlib} 9446or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well. 9447This ensures that you have no unresolved references to internal GCC 9448library subroutines. (For example, @samp{__main}, used to ensure C++ 9449constructors will be called; @pxref{Collect2,,@code{collect2}, gccint, 9450GNU Compiler Collection (GCC) Internals}.) 9451 9452@item -pie 9453@opindex pie 9454Produce a position independent executable on targets that support it. 9455For predictable results, you must also specify the same set of options 9456that were used to generate code (@option{-fpie}, @option{-fPIE}, 9457or model suboptions) when you specify this option. 9458 9459@item -rdynamic 9460@opindex rdynamic 9461Pass the flag @option{-export-dynamic} to the ELF linker, on targets 9462that support it. This instructs the linker to add all symbols, not 9463only used ones, to the dynamic symbol table. This option is needed 9464for some uses of @code{dlopen} or to allow obtaining backtraces 9465from within a program. 9466 9467@item -s 9468@opindex s 9469Remove all symbol table and relocation information from the executable. 9470 9471@item -static 9472@opindex static 9473On systems that support dynamic linking, this prevents linking with the shared 9474libraries. On other systems, this option has no effect. 9475 9476@item -shared 9477@opindex shared 9478Produce a shared object which can then be linked with other objects to 9479form an executable. Not all systems support this option. For predictable 9480results, you must also specify the same set of options that were used to 9481generate code (@option{-fpic}, @option{-fPIC}, or model suboptions) 9482when you specify this option.@footnote{On some systems, @samp{gcc -shared} 9483needs to build supplementary stub code for constructors to work. On 9484multi-libbed systems, @samp{gcc -shared} must select the correct support 9485libraries to link against. Failing to supply the correct flags may lead 9486to subtle defects. Supplying them in cases where they are not necessary 9487is innocuous.} 9488 9489@item -shared-libgcc 9490@itemx -static-libgcc 9491@opindex shared-libgcc 9492@opindex static-libgcc 9493On systems that provide @file{libgcc} as a shared library, these options 9494force the use of either the shared or static version respectively. 9495If no shared version of @file{libgcc} was built when the compiler was 9496configured, these options have no effect. 9497 9498There are several situations in which an application should use the 9499shared @file{libgcc} instead of the static version. The most common 9500of these is when the application wishes to throw and catch exceptions 9501across different shared libraries. In that case, each of the libraries 9502as well as the application itself should use the shared @file{libgcc}. 9503 9504Therefore, the G++ and GCJ drivers automatically add 9505@option{-shared-libgcc} whenever you build a shared library or a main 9506executable, because C++ and Java programs typically use exceptions, so 9507this is the right thing to do. 9508 9509If, instead, you use the GCC driver to create shared libraries, you may 9510find that they will not always be linked with the shared @file{libgcc}. 9511If GCC finds, at its configuration time, that you have a non-GNU linker 9512or a GNU linker that does not support option @option{--eh-frame-hdr}, 9513it will link the shared version of @file{libgcc} into shared libraries 9514by default. Otherwise, it will take advantage of the linker and optimize 9515away the linking with the shared version of @file{libgcc}, linking with 9516the static version of libgcc by default. This allows exceptions to 9517propagate through such shared libraries, without incurring relocation 9518costs at library load time. 9519 9520However, if a library or main executable is supposed to throw or catch 9521exceptions, you must link it using the G++ or GCJ driver, as appropriate 9522for the languages used in the program, or using the option 9523@option{-shared-libgcc}, such that it is linked with the shared 9524@file{libgcc}. 9525 9526@item -static-libstdc++ 9527When the @command{g++} program is used to link a C++ program, it will 9528normally automatically link against @option{libstdc++}. If 9529@file{libstdc++} is available as a shared library, and the 9530@option{-static} option is not used, then this will link against the 9531shared version of @file{libstdc++}. That is normally fine. However, it 9532is sometimes useful to freeze the version of @file{libstdc++} used by 9533the program without going all the way to a fully static link. The 9534@option{-static-libstdc++} option directs the @command{g++} driver to 9535link @file{libstdc++} statically, without necessarily linking other 9536libraries statically. 9537 9538@item -symbolic 9539@opindex symbolic 9540Bind references to global symbols when building a shared object. Warn 9541about any unresolved references (unless overridden by the link editor 9542option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support 9543this option. 9544 9545@item -T @var{script} 9546@opindex T 9547@cindex linker script 9548Use @var{script} as the linker script. This option is supported by most 9549systems using the GNU linker. On some targets, such as bare-board 9550targets without an operating system, the @option{-T} option may be required 9551when linking to avoid references to undefined symbols. 9552 9553@item -Xlinker @var{option} 9554@opindex Xlinker 9555Pass @var{option} as an option to the linker. You can use this to 9556supply system-specific linker options that GCC does not recognize. 9557 9558If you want to pass an option that takes a separate argument, you must use 9559@option{-Xlinker} twice, once for the option and once for the argument. 9560For example, to pass @option{-assert definitions}, you must write 9561@samp{-Xlinker -assert -Xlinker definitions}. It does not work to write 9562@option{-Xlinker "-assert definitions"}, because this passes the entire 9563string as a single argument, which is not what the linker expects. 9564 9565When using the GNU linker, it is usually more convenient to pass 9566arguments to linker options using the @option{@var{option}=@var{value}} 9567syntax than as separate arguments. For example, you can specify 9568@samp{-Xlinker -Map=output.map} rather than 9569@samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support 9570this syntax for command-line options. 9571 9572@item -Wl,@var{option} 9573@opindex Wl 9574Pass @var{option} as an option to the linker. If @var{option} contains 9575commas, it is split into multiple options at the commas. You can use this 9576syntax to pass an argument to the option. 9577For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the 9578linker. When using the GNU linker, you can also get the same effect with 9579@samp{-Wl,-Map=output.map}. 9580 9581@item -u @var{symbol} 9582@opindex u 9583Pretend the symbol @var{symbol} is undefined, to force linking of 9584library modules to define it. You can use @option{-u} multiple times with 9585different symbols to force loading of additional library modules. 9586@end table 9587 9588@node Directory Options 9589@section Options for Directory Search 9590@cindex directory options 9591@cindex options, directory search 9592@cindex search path 9593 9594These options specify directories to search for header files, for 9595libraries and for parts of the compiler: 9596 9597@table @gcctabopt 9598@item -I@var{dir} 9599@opindex I 9600Add the directory @var{dir} to the head of the list of directories to be 9601searched for header files. This can be used to override a system header 9602file, substituting your own version, since these directories are 9603searched before the system header file directories. However, you should 9604not use this option to add directories that contain vendor-supplied 9605system header files (use @option{-isystem} for that). If you use more than 9606one @option{-I} option, the directories are scanned in left-to-right 9607order; the standard system directories come after. 9608 9609If a standard system include directory, or a directory specified with 9610@option{-isystem}, is also specified with @option{-I}, the @option{-I} 9611option will be ignored. The directory will still be searched but as a 9612system directory at its normal position in the system include chain. 9613This is to ensure that GCC's procedure to fix buggy system headers and 9614the ordering for the include_next directive are not inadvertently changed. 9615If you really need to change the search order for system directories, 9616use the @option{-nostdinc} and/or @option{-isystem} options. 9617 9618@item -iplugindir=@var{dir} 9619Set the directory to search for plugins that are passed 9620by @option{-fplugin=@var{name}} instead of 9621@option{-fplugin=@var{path}/@var{name}.so}. This option is not meant 9622to be used by the user, but only passed by the driver. 9623 9624@item -iquote@var{dir} 9625@opindex iquote 9626Add the directory @var{dir} to the head of the list of directories to 9627be searched for header files only for the case of @samp{#include 9628"@var{file}"}; they are not searched for @samp{#include <@var{file}>}, 9629otherwise just like @option{-I}. 9630 9631@item -L@var{dir} 9632@opindex L 9633Add directory @var{dir} to the list of directories to be searched 9634for @option{-l}. 9635 9636@item -B@var{prefix} 9637@opindex B 9638This option specifies where to find the executables, libraries, 9639include files, and data files of the compiler itself. 9640 9641The compiler driver program runs one or more of the subprograms 9642@file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries 9643@var{prefix} as a prefix for each program it tries to run, both with and 9644without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}). 9645 9646For each subprogram to be run, the compiler driver first tries the 9647@option{-B} prefix, if any. If that name is not found, or if @option{-B} 9648was not specified, the driver tries two standard prefixes, 9649@file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of 9650those results in a file name that is found, the unmodified program 9651name is searched for using the directories specified in your 9652@env{PATH} environment variable. 9653 9654The compiler will check to see if the path provided by the @option{-B} 9655refers to a directory, and if necessary it will add a directory 9656separator character at the end of the path. 9657 9658@option{-B} prefixes that effectively specify directory names also apply 9659to libraries in the linker, because the compiler translates these 9660options into @option{-L} options for the linker. They also apply to 9661includes files in the preprocessor, because the compiler translates these 9662options into @option{-isystem} options for the preprocessor. In this case, 9663the compiler appends @samp{include} to the prefix. 9664 9665The runtime support file @file{libgcc.a} can also be searched for using 9666the @option{-B} prefix, if needed. If it is not found there, the two 9667standard prefixes above are tried, and that is all. The file is left 9668out of the link if it is not found by those means. 9669 9670Another way to specify a prefix much like the @option{-B} prefix is to use 9671the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment 9672Variables}. 9673 9674As a special kludge, if the path provided by @option{-B} is 9675@file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to 96769, then it will be replaced by @file{[dir/]include}. This is to help 9677with boot-strapping the compiler. 9678 9679@item -specs=@var{file} 9680@opindex specs 9681Process @var{file} after the compiler reads in the standard @file{specs} 9682file, in order to override the defaults which the @file{gcc} driver 9683program uses when determining what switches to pass to @file{cc1}, 9684@file{cc1plus}, @file{as}, @file{ld}, etc. More than one 9685@option{-specs=@var{file}} can be specified on the command line, and they 9686are processed in order, from left to right. 9687 9688@item --sysroot=@var{dir} 9689@opindex sysroot 9690Use @var{dir} as the logical root directory for headers and libraries. 9691For example, if the compiler would normally search for headers in 9692@file{/usr/include} and libraries in @file{/usr/lib}, it will instead 9693search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}. 9694 9695If you use both this option and the @option{-isysroot} option, then 9696the @option{--sysroot} option will apply to libraries, but the 9697@option{-isysroot} option will apply to header files. 9698 9699The GNU linker (beginning with version 2.16) has the necessary support 9700for this option. If your linker does not support this option, the 9701header file aspect of @option{--sysroot} will still work, but the 9702library aspect will not. 9703 9704@item -I- 9705@opindex I- 9706This option has been deprecated. Please use @option{-iquote} instead for 9707@option{-I} directories before the @option{-I-} and remove the @option{-I-}. 9708Any directories you specify with @option{-I} options before the @option{-I-} 9709option are searched only for the case of @samp{#include "@var{file}"}; 9710they are not searched for @samp{#include <@var{file}>}. 9711 9712If additional directories are specified with @option{-I} options after 9713the @option{-I-}, these directories are searched for all @samp{#include} 9714directives. (Ordinarily @emph{all} @option{-I} directories are used 9715this way.) 9716 9717In addition, the @option{-I-} option inhibits the use of the current 9718directory (where the current input file came from) as the first search 9719directory for @samp{#include "@var{file}"}. There is no way to 9720override this effect of @option{-I-}. With @option{-I.} you can specify 9721searching the directory that was current when the compiler was 9722invoked. That is not exactly the same as what the preprocessor does 9723by default, but it is often satisfactory. 9724 9725@option{-I-} does not inhibit the use of the standard system directories 9726for header files. Thus, @option{-I-} and @option{-nostdinc} are 9727independent. 9728@end table 9729 9730@c man end 9731 9732@node Spec Files 9733@section Specifying subprocesses and the switches to pass to them 9734@cindex Spec Files 9735 9736@command{gcc} is a driver program. It performs its job by invoking a 9737sequence of other programs to do the work of compiling, assembling and 9738linking. GCC interprets its command-line parameters and uses these to 9739deduce which programs it should invoke, and which command-line options 9740it ought to place on their command lines. This behavior is controlled 9741by @dfn{spec strings}. In most cases there is one spec string for each 9742program that GCC can invoke, but a few programs have multiple spec 9743strings to control their behavior. The spec strings built into GCC can 9744be overridden by using the @option{-specs=} command-line switch to specify 9745a spec file. 9746 9747@dfn{Spec files} are plaintext files that are used to construct spec 9748strings. They consist of a sequence of directives separated by blank 9749lines. The type of directive is determined by the first non-whitespace 9750character on the line, which can be one of the following: 9751 9752@table @code 9753@item %@var{command} 9754Issues a @var{command} to the spec file processor. The commands that can 9755appear here are: 9756 9757@table @code 9758@item %include <@var{file}> 9759@cindex @code{%include} 9760Search for @var{file} and insert its text at the current point in the 9761specs file. 9762 9763@item %include_noerr <@var{file}> 9764@cindex @code{%include_noerr} 9765Just like @samp{%include}, but do not generate an error message if the include 9766file cannot be found. 9767 9768@item %rename @var{old_name} @var{new_name} 9769@cindex @code{%rename} 9770Rename the spec string @var{old_name} to @var{new_name}. 9771 9772@end table 9773 9774@item *[@var{spec_name}]: 9775This tells the compiler to create, override or delete the named spec 9776string. All lines after this directive up to the next directive or 9777blank line are considered to be the text for the spec string. If this 9778results in an empty string then the spec will be deleted. (Or, if the 9779spec did not exist, then nothing will happen.) Otherwise, if the spec 9780does not currently exist a new spec will be created. If the spec does 9781exist then its contents will be overridden by the text of this 9782directive, unless the first character of that text is the @samp{+} 9783character, in which case the text will be appended to the spec. 9784 9785@item [@var{suffix}]: 9786Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive 9787and up to the next directive or blank line are considered to make up the 9788spec string for the indicated suffix. When the compiler encounters an 9789input file with the named suffix, it will processes the spec string in 9790order to work out how to compile that file. For example: 9791 9792@smallexample 9793.ZZ: 9794z-compile -input %i 9795@end smallexample 9796 9797This says that any input file whose name ends in @samp{.ZZ} should be 9798passed to the program @samp{z-compile}, which should be invoked with the 9799command-line switch @option{-input} and with the result of performing the 9800@samp{%i} substitution. (See below.) 9801 9802As an alternative to providing a spec string, the text that follows a 9803suffix directive can be one of the following: 9804 9805@table @code 9806@item @@@var{language} 9807This says that the suffix is an alias for a known @var{language}. This is 9808similar to using the @option{-x} command-line switch to GCC to specify a 9809language explicitly. For example: 9810 9811@smallexample 9812.ZZ: 9813@@c++ 9814@end smallexample 9815 9816Says that .ZZ files are, in fact, C++ source files. 9817 9818@item #@var{name} 9819This causes an error messages saying: 9820 9821@smallexample 9822@var{name} compiler not installed on this system. 9823@end smallexample 9824@end table 9825 9826GCC already has an extensive list of suffixes built into it. 9827This directive will add an entry to the end of the list of suffixes, but 9828since the list is searched from the end backwards, it is effectively 9829possible to override earlier entries using this technique. 9830 9831@end table 9832 9833GCC has the following spec strings built into it. Spec files can 9834override these strings or create their own. Note that individual 9835targets can also add their own spec strings to this list. 9836 9837@smallexample 9838asm Options to pass to the assembler 9839asm_final Options to pass to the assembler post-processor 9840cpp Options to pass to the C preprocessor 9841cc1 Options to pass to the C compiler 9842cc1plus Options to pass to the C++ compiler 9843endfile Object files to include at the end of the link 9844link Options to pass to the linker 9845lib Libraries to include on the command line to the linker 9846libgcc Decides which GCC support library to pass to the linker 9847linker Sets the name of the linker 9848predefines Defines to be passed to the C preprocessor 9849signed_char Defines to pass to CPP to say whether @code{char} is signed 9850 by default 9851startfile Object files to include at the start of the link 9852@end smallexample 9853 9854Here is a small example of a spec file: 9855 9856@smallexample 9857%rename lib old_lib 9858 9859*lib: 9860--start-group -lgcc -lc -leval1 --end-group %(old_lib) 9861@end smallexample 9862 9863This example renames the spec called @samp{lib} to @samp{old_lib} and 9864then overrides the previous definition of @samp{lib} with a new one. 9865The new definition adds in some extra command-line options before 9866including the text of the old definition. 9867 9868@dfn{Spec strings} are a list of command-line options to be passed to their 9869corresponding program. In addition, the spec strings can contain 9870@samp{%}-prefixed sequences to substitute variable text or to 9871conditionally insert text into the command line. Using these constructs 9872it is possible to generate quite complex command lines. 9873 9874Here is a table of all defined @samp{%}-sequences for spec 9875strings. Note that spaces are not generated automatically around the 9876results of expanding these sequences. Therefore you can concatenate them 9877together or combine them with constant text in a single argument. 9878 9879@table @code 9880@item %% 9881Substitute one @samp{%} into the program name or argument. 9882 9883@item %i 9884Substitute the name of the input file being processed. 9885 9886@item %b 9887Substitute the basename of the input file being processed. 9888This is the substring up to (and not including) the last period 9889and not including the directory. 9890 9891@item %B 9892This is the same as @samp{%b}, but include the file suffix (text after 9893the last period). 9894 9895@item %d 9896Marks the argument containing or following the @samp{%d} as a 9897temporary file name, so that that file will be deleted if GCC exits 9898successfully. Unlike @samp{%g}, this contributes no text to the 9899argument. 9900 9901@item %g@var{suffix} 9902Substitute a file name that has suffix @var{suffix} and is chosen 9903once per compilation, and mark the argument in the same way as 9904@samp{%d}. To reduce exposure to denial-of-service attacks, the file 9905name is now chosen in a way that is hard to predict even when previously 9906chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s} 9907might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches 9908the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is 9909treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g} 9910was simply substituted with a file name chosen once per compilation, 9911without regard to any appended suffix (which was therefore treated 9912just like ordinary text), making such attacks more likely to succeed. 9913 9914@item %u@var{suffix} 9915Like @samp{%g}, but generates a new temporary file name even if 9916@samp{%u@var{suffix}} was already seen. 9917 9918@item %U@var{suffix} 9919Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a 9920new one if there is no such last file name. In the absence of any 9921@samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share 9922the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s} 9923would involve the generation of two distinct file names, one 9924for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was 9925simply substituted with a file name chosen for the previous @samp{%u}, 9926without regard to any appended suffix. 9927 9928@item %j@var{suffix} 9929Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is 9930writable, and if save-temps is off; otherwise, substitute the name 9931of a temporary file, just like @samp{%u}. This temporary file is not 9932meant for communication between processes, but rather as a junk 9933disposal mechanism. 9934 9935@item %|@var{suffix} 9936@itemx %m@var{suffix} 9937Like @samp{%g}, except if @option{-pipe} is in effect. In that case 9938@samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at 9939all. These are the two most common ways to instruct a program that it 9940should read from standard input or write to standard output. If you 9941need something more elaborate you can use an @samp{%@{pipe:@code{X}@}} 9942construct: see for example @file{f/lang-specs.h}. 9943 9944@item %.@var{SUFFIX} 9945Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args 9946when it is subsequently output with @samp{%*}. @var{SUFFIX} is 9947terminated by the next space or %. 9948 9949@item %w 9950Marks the argument containing or following the @samp{%w} as the 9951designated output file of this compilation. This puts the argument 9952into the sequence of arguments that @samp{%o} will substitute later. 9953 9954@item %o 9955Substitutes the names of all the output files, with spaces 9956automatically placed around them. You should write spaces 9957around the @samp{%o} as well or the results are undefined. 9958@samp{%o} is for use in the specs for running the linker. 9959Input files whose names have no recognized suffix are not compiled 9960at all, but they are included among the output files, so they will 9961be linked. 9962 9963@item %O 9964Substitutes the suffix for object files. Note that this is 9965handled specially when it immediately follows @samp{%g, %u, or %U}, 9966because of the need for those to form complete file names. The 9967handling is such that @samp{%O} is treated exactly as if it had already 9968been substituted, except that @samp{%g, %u, and %U} do not currently 9969support additional @var{suffix} characters following @samp{%O} as they would 9970following, for example, @samp{.o}. 9971 9972@item %p 9973Substitutes the standard macro predefinitions for the 9974current target machine. Use this when running @code{cpp}. 9975 9976@item %P 9977Like @samp{%p}, but puts @samp{__} before and after the name of each 9978predefined macro, except for macros that start with @samp{__} or with 9979@samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO 9980C@. 9981 9982@item %I 9983Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}), 9984@option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), 9985@option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options) 9986and @option{-imultilib} as necessary. 9987 9988@item %s 9989Current argument is the name of a library or startup file of some sort. 9990Search for that file in a standard list of directories and substitute 9991the full name found. The current working directory is included in the 9992list of directories scanned. 9993 9994@item %T 9995Current argument is the name of a linker script. Search for that file 9996in the current list of directories to scan for libraries. If the file 9997is located insert a @option{--script} option into the command line 9998followed by the full path name found. If the file is not found then 9999generate an error message. Note: the current working directory is not 10000searched. 10001 10002@item %e@var{str} 10003Print @var{str} as an error message. @var{str} is terminated by a newline. 10004Use this when inconsistent options are detected. 10005 10006@item %(@var{name}) 10007Substitute the contents of spec string @var{name} at this point. 10008 10009@item %x@{@var{option}@} 10010Accumulate an option for @samp{%X}. 10011 10012@item %X 10013Output the accumulated linker options specified by @option{-Wl} or a @samp{%x} 10014spec string. 10015 10016@item %Y 10017Output the accumulated assembler options specified by @option{-Wa}. 10018 10019@item %Z 10020Output the accumulated preprocessor options specified by @option{-Wp}. 10021 10022@item %a 10023Process the @code{asm} spec. This is used to compute the 10024switches to be passed to the assembler. 10025 10026@item %A 10027Process the @code{asm_final} spec. This is a spec string for 10028passing switches to an assembler post-processor, if such a program is 10029needed. 10030 10031@item %l 10032Process the @code{link} spec. This is the spec for computing the 10033command line passed to the linker. Typically it will make use of the 10034@samp{%L %G %S %D and %E} sequences. 10035 10036@item %D 10037Dump out a @option{-L} option for each directory that GCC believes might 10038contain startup files. If the target supports multilibs then the 10039current multilib directory will be prepended to each of these paths. 10040 10041@item %L 10042Process the @code{lib} spec. This is a spec string for deciding which 10043libraries should be included on the command line to the linker. 10044 10045@item %G 10046Process the @code{libgcc} spec. This is a spec string for deciding 10047which GCC support library should be included on the command line to the linker. 10048 10049@item %S 10050Process the @code{startfile} spec. This is a spec for deciding which 10051object files should be the first ones passed to the linker. Typically 10052this might be a file named @file{crt0.o}. 10053 10054@item %E 10055Process the @code{endfile} spec. This is a spec string that specifies 10056the last object files that will be passed to the linker. 10057 10058@item %C 10059Process the @code{cpp} spec. This is used to construct the arguments 10060to be passed to the C preprocessor. 10061 10062@item %1 10063Process the @code{cc1} spec. This is used to construct the options to be 10064passed to the actual C compiler (@samp{cc1}). 10065 10066@item %2 10067Process the @code{cc1plus} spec. This is used to construct the options to be 10068passed to the actual C++ compiler (@samp{cc1plus}). 10069 10070@item %* 10071Substitute the variable part of a matched option. See below. 10072Note that each comma in the substituted string is replaced by 10073a single space. 10074 10075@item %<@code{S} 10076Remove all occurrences of @code{-S} from the command line. Note---this 10077command is position dependent. @samp{%} commands in the spec string 10078before this one will see @code{-S}, @samp{%} commands in the spec string 10079after this one will not. 10080 10081@item %:@var{function}(@var{args}) 10082Call the named function @var{function}, passing it @var{args}. 10083@var{args} is first processed as a nested spec string, then split 10084into an argument vector in the usual fashion. The function returns 10085a string which is processed as if it had appeared literally as part 10086of the current spec. 10087 10088The following built-in spec functions are provided: 10089 10090@table @code 10091@item @code{getenv} 10092The @code{getenv} spec function takes two arguments: an environment 10093variable name and a string. If the environment variable is not 10094defined, a fatal error is issued. Otherwise, the return value is the 10095value of the environment variable concatenated with the string. For 10096example, if @env{TOPDIR} is defined as @file{/path/to/top}, then: 10097 10098@smallexample 10099%:getenv(TOPDIR /include) 10100@end smallexample 10101 10102expands to @file{/path/to/top/include}. 10103 10104@item @code{if-exists} 10105The @code{if-exists} spec function takes one argument, an absolute 10106pathname to a file. If the file exists, @code{if-exists} returns the 10107pathname. Here is a small example of its usage: 10108 10109@smallexample 10110*startfile: 10111crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s 10112@end smallexample 10113 10114@item @code{if-exists-else} 10115The @code{if-exists-else} spec function is similar to the @code{if-exists} 10116spec function, except that it takes two arguments. The first argument is 10117an absolute pathname to a file. If the file exists, @code{if-exists-else} 10118returns the pathname. If it does not exist, it returns the second argument. 10119This way, @code{if-exists-else} can be used to select one file or another, 10120based on the existence of the first. Here is a small example of its usage: 10121 10122@smallexample 10123*startfile: 10124crt0%O%s %:if-exists(crti%O%s) \ 10125%:if-exists-else(crtbeginT%O%s crtbegin%O%s) 10126@end smallexample 10127 10128@item @code{replace-outfile} 10129The @code{replace-outfile} spec function takes two arguments. It looks for the 10130first argument in the outfiles array and replaces it with the second argument. Here 10131is a small example of its usage: 10132 10133@smallexample 10134%@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@} 10135@end smallexample 10136 10137@item @code{remove-outfile} 10138The @code{remove-outfile} spec function takes one argument. It looks for the 10139first argument in the outfiles array and removes it. Here is a small example 10140its usage: 10141 10142@smallexample 10143%:remove-outfile(-lm) 10144@end smallexample 10145 10146@item @code{pass-through-libs} 10147The @code{pass-through-libs} spec function takes any number of arguments. It 10148finds any @option{-l} options and any non-options ending in ".a" (which it 10149assumes are the names of linker input library archive files) and returns a 10150result containing all the found arguments each prepended by 10151@option{-plugin-opt=-pass-through=} and joined by spaces. This list is 10152intended to be passed to the LTO linker plugin. 10153 10154@smallexample 10155%:pass-through-libs(%G %L %G) 10156@end smallexample 10157 10158@item @code{print-asm-header} 10159The @code{print-asm-header} function takes no arguments and simply 10160prints a banner like: 10161 10162@smallexample 10163Assembler options 10164================= 10165 10166Use "-Wa,OPTION" to pass "OPTION" to the assembler. 10167@end smallexample 10168 10169It is used to separate compiler options from assembler options 10170in the @option{--target-help} output. 10171@end table 10172 10173@item %@{@code{S}@} 10174Substitutes the @code{-S} switch, if that switch was given to GCC@. 10175If that switch was not specified, this substitutes nothing. Note that 10176the leading dash is omitted when specifying this option, and it is 10177automatically inserted if the substitution is performed. Thus the spec 10178string @samp{%@{foo@}} would match the command-line option @option{-foo} 10179and would output the command-line option @option{-foo}. 10180 10181@item %W@{@code{S}@} 10182Like %@{@code{S}@} but mark last argument supplied within as a file to be 10183deleted on failure. 10184 10185@item %@{@code{S}*@} 10186Substitutes all the switches specified to GCC whose names start 10187with @code{-S}, but which also take an argument. This is used for 10188switches like @option{-o}, @option{-D}, @option{-I}, etc. 10189GCC considers @option{-o foo} as being 10190one switch whose names starts with @samp{o}. %@{o*@} would substitute this 10191text, including the space. Thus two arguments would be generated. 10192 10193@item %@{@code{S}*&@code{T}*@} 10194Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options 10195(the order of @code{S} and @code{T} in the spec is not significant). 10196There can be any number of ampersand-separated variables; for each the 10197wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}. 10198 10199@item %@{@code{S}:@code{X}@} 10200Substitutes @code{X}, if the @samp{-S} switch was given to GCC@. 10201 10202@item %@{!@code{S}:@code{X}@} 10203Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@. 10204 10205@item %@{@code{S}*:@code{X}@} 10206Substitutes @code{X} if one or more switches whose names start with 10207@code{-S} are specified to GCC@. Normally @code{X} is substituted only 10208once, no matter how many such switches appeared. However, if @code{%*} 10209appears somewhere in @code{X}, then @code{X} will be substituted once 10210for each matching switch, with the @code{%*} replaced by the part of 10211that switch that matched the @code{*}. 10212 10213@item %@{.@code{S}:@code{X}@} 10214Substitutes @code{X}, if processing a file with suffix @code{S}. 10215 10216@item %@{!.@code{S}:@code{X}@} 10217Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}. 10218 10219@item %@{,@code{S}:@code{X}@} 10220Substitutes @code{X}, if processing a file for language @code{S}. 10221 10222@item %@{!,@code{S}:@code{X}@} 10223Substitutes @code{X}, if not processing a file for language @code{S}. 10224 10225@item %@{@code{S}|@code{P}:@code{X}@} 10226Substitutes @code{X} if either @code{-S} or @code{-P} was given to 10227GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and 10228@code{*} sequences as well, although they have a stronger binding than 10229the @samp{|}. If @code{%*} appears in @code{X}, all of the 10230alternatives must be starred, and only the first matching alternative 10231is substituted. 10232 10233For example, a spec string like this: 10234 10235@smallexample 10236%@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@} 10237@end smallexample 10238 10239will output the following command-line options from the following input 10240command-line options: 10241 10242@smallexample 10243fred.c -foo -baz 10244jim.d -bar -boggle 10245-d fred.c -foo -baz -boggle 10246-d jim.d -bar -baz -boggle 10247@end smallexample 10248 10249@item %@{S:X; T:Y; :D@} 10250 10251If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was 10252given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can 10253be as many clauses as you need. This may be combined with @code{.}, 10254@code{,}, @code{!}, @code{|}, and @code{*} as needed. 10255 10256 10257@end table 10258 10259The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar 10260construct may contain other nested @samp{%} constructs or spaces, or 10261even newlines. They are processed as usual, as described above. 10262Trailing white space in @code{X} is ignored. White space may also 10263appear anywhere on the left side of the colon in these constructs, 10264except between @code{.} or @code{*} and the corresponding word. 10265 10266The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are 10267handled specifically in these constructs. If another value of 10268@option{-O} or the negated form of a @option{-f}, @option{-m}, or 10269@option{-W} switch is found later in the command line, the earlier 10270switch value is ignored, except with @{@code{S}*@} where @code{S} is 10271just one letter, which passes all matching options. 10272 10273The character @samp{|} at the beginning of the predicate text is used to 10274indicate that a command should be piped to the following command, but 10275only if @option{-pipe} is specified. 10276 10277It is built into GCC which switches take arguments and which do not. 10278(You might think it would be useful to generalize this to allow each 10279compiler's spec to say which switches take arguments. But this cannot 10280be done in a consistent fashion. GCC cannot even decide which input 10281files have been specified without knowing which switches take arguments, 10282and it must know which input files to compile in order to tell which 10283compilers to run). 10284 10285GCC also knows implicitly that arguments starting in @option{-l} are to be 10286treated as compiler output files, and passed to the linker in their 10287proper position among the other output files. 10288 10289@c man begin OPTIONS 10290 10291@node Target Options 10292@section Specifying Target Machine and Compiler Version 10293@cindex target options 10294@cindex cross compiling 10295@cindex specifying machine version 10296@cindex specifying compiler version and target machine 10297@cindex compiler version, specifying 10298@cindex target machine, specifying 10299 10300The usual way to run GCC is to run the executable called @command{gcc}, or 10301@command{@var{machine}-gcc} when cross-compiling, or 10302@command{@var{machine}-gcc-@var{version}} to run a version other than the 10303one that was installed last. 10304 10305@node Submodel Options 10306@section Hardware Models and Configurations 10307@cindex submodel options 10308@cindex specifying hardware config 10309@cindex hardware models and configurations, specifying 10310@cindex machine dependent options 10311 10312Each target machine types can have its own 10313special options, starting with @samp{-m}, to choose among various 10314hardware models or configurations---for example, 68010 vs 68020, 10315floating coprocessor or none. A single installed version of the 10316compiler can compile for any model or configuration, according to the 10317options specified. 10318 10319Some configurations of the compiler also support additional special 10320options, usually for compatibility with other compilers on the same 10321platform. 10322 10323@c This list is ordered alphanumerically by subsection name. 10324@c It should be the same order and spelling as these options are listed 10325@c in Machine Dependent Options 10326 10327@menu 10328* Adapteva Epiphany Options:: 10329* ARM Options:: 10330* AVR Options:: 10331* Blackfin Options:: 10332* C6X Options:: 10333* CRIS Options:: 10334* CR16 Options:: 10335* Darwin Options:: 10336* DEC Alpha Options:: 10337* DEC Alpha/VMS Options:: 10338* FR30 Options:: 10339* FRV Options:: 10340* GNU/Linux Options:: 10341* H8/300 Options:: 10342* HPPA Options:: 10343* i386 and x86-64 Options:: 10344* i386 and x86-64 Windows Options:: 10345* IA-64 Options:: 10346* IA-64/VMS Options:: 10347* LM32 Options:: 10348* M32C Options:: 10349* M32R/D Options:: 10350* M680x0 Options:: 10351* MCore Options:: 10352* MeP Options:: 10353* MicroBlaze Options:: 10354* MIPS Options:: 10355* MMIX Options:: 10356* MN10300 Options:: 10357* PDP-11 Options:: 10358* picoChip Options:: 10359* PowerPC Options:: 10360* RL78 Options:: 10361* RS/6000 and PowerPC Options:: 10362* RX Options:: 10363* S/390 and zSeries Options:: 10364* Score Options:: 10365* SH Options:: 10366* Solaris 2 Options:: 10367* SPARC Options:: 10368* SPU Options:: 10369* System V Options:: 10370* TILE-Gx Options:: 10371* TILEPro Options:: 10372* V850 Options:: 10373* VAX Options:: 10374* VxWorks Options:: 10375* x86-64 Options:: 10376* Xstormy16 Options:: 10377* Xtensa Options:: 10378* zSeries Options:: 10379@end menu 10380 10381@node Adapteva Epiphany Options 10382@subsection Adapteva Epiphany Options 10383 10384These @samp{-m} options are defined for Adapteva Epiphany: 10385 10386@table @gcctabopt 10387@item -mhalf-reg-file 10388@opindex mhalf-reg-file 10389Don't allocate any register in the range @code{r32}@dots{}@code{r63}. 10390That allows code to run on hardware variants that lack these registers. 10391 10392@item -mprefer-short-insn-regs 10393@opindex mprefer-short-insn-regs 10394Preferrentially allocate registers that allow short instruction generation. 10395This can result in increasesd instruction count, so if this reduces or 10396increases code size might vary from case to case. 10397 10398@item -mbranch-cost=@var{num} 10399@opindex mbranch-cost 10400Set the cost of branches to roughly @var{num} ``simple'' instructions. 10401This cost is only a heuristic and is not guaranteed to produce 10402consistent results across releases. 10403 10404@item -mcmove 10405@opindex mcmove 10406Enable the generation of conditional moves. 10407 10408@item -mnops=@var{num} 10409@opindex mnops 10410Emit @var{num} nops before every other generated instruction. 10411 10412@item -mno-soft-cmpsf 10413@opindex mno-soft-cmpsf 10414For single-precision floating-point comparisons, emit an fsub instruction 10415and test the flags. This is faster than a software comparison, but can 10416get incorrect results in the presence of NaNs, or when two different small 10417numbers are compared such that their difference is calculated as zero. 10418The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant, 10419software comparisons. 10420 10421@item -mstack-offset=@var{num} 10422@opindex mstack-offset 10423Set the offset between the top of the stack and the stack pointer. 10424E.g., a value of 8 means that the eight bytes in the range sp+0@dots{}sp+7 10425can be used by leaf functions without stack allocation. 10426Values other than @samp{8} or @samp{16} are untested and unlikely to work. 10427Note also that this option changes the ABI, compiling a program with a 10428different stack offset than the libraries have been compiled with 10429will generally not work. 10430This option can be useful if you want to evaluate if a different stack 10431offset would give you better code, but to actually use a different stack 10432offset to build working programs, it is recommended to configure the 10433toolchain with the appropriate @samp{--with-stack-offset=@var{num}} option. 10434 10435@item -mno-round-nearest 10436@opindex mno-round-nearest 10437Make the scheduler assume that the rounding mode has been set to 10438truncating. The default is @option{-mround-nearest}. 10439 10440@item -mlong-calls 10441@opindex mlong-calls 10442If not otherwise specified by an attribute, assume all calls might be beyond 10443the offset range of the b / bl instructions, and therefore load the 10444function address into a register before performing a (otherwise direct) call. 10445This is the default. 10446 10447@item -mshort-calls 10448@opindex short-calls 10449If not otherwise specified by an attribute, assume all direct calls are 10450in the range of the b / bl instructions, so use these instructions 10451for direct calls. The default is @option{-mlong-calls}. 10452 10453@item -msmall16 10454@opindex msmall16 10455Assume addresses can be loaded as 16-bit unsigned values. This does not 10456apply to function addresses for which @option{-mlong-calls} semantics 10457are in effect. 10458 10459@item -mfp-mode=@var{mode} 10460@opindex mfp-mode 10461Set the prevailing mode of the floating-point unit. 10462This determines the floating-point mode that is provided and expected 10463at function call and return time. Making this mode match the mode you 10464predominantly need at function start can make your programs smaller and 10465faster by avoiding unnecessary mode switches. 10466 10467@var{mode} can be set to one the following values: 10468 10469@table @samp 10470@item caller 10471Any mode at function entry is valid, and retained or restored when 10472the function returns, and when it calls other functions. 10473This mode is useful for compiling libraries or other compilation units 10474you might want to incorporate into different programs with different 10475prevailing FPU modes, and the convenience of being able to use a single 10476object file outweighs the size and speed overhead for any extra 10477mode switching that might be needed, compared with what would be needed 10478with a more specific choice of prevailing FPU mode. 10479 10480@item truncate 10481This is the mode used for floating-point calculations with 10482truncating (i.e.@: round towards zero) rounding mode. That includes 10483conversion from floating point to integer. 10484 10485@item round-nearest 10486This is the mode used for floating-point calculations with 10487round-to-nearest-or-even rounding mode. 10488 10489@item int 10490This is the mode used to perform integer calculations in the FPU, e.g.@: 10491integer multiply, or integer multiply-and-accumulate. 10492@end table 10493 10494The default is @option{-mfp-mode=caller} 10495 10496@item -mnosplit-lohi 10497@opindex mnosplit-lohi 10498@item -mno-postinc 10499@opindex mno-postinc 10500@item -mno-postmodify 10501@opindex mno-postmodify 10502Code generation tweaks that disable, respectively, splitting of 32-bit 10503loads, generation of post-increment addresses, and generation of 10504post-modify addresses. The defaults are @option{msplit-lohi}, 10505@option{-mpost-inc}, and @option{-mpost-modify}. 10506 10507@item -mnovect-double 10508@opindex mno-vect-double 10509Change the preferred SIMD mode to SImode. The default is 10510@option{-mvect-double}, which uses DImode as preferred SIMD mode. 10511 10512@item -max-vect-align=@var{num} 10513@opindex max-vect-align 10514The maximum alignment for SIMD vector mode types. 10515@var{num} may be 4 or 8. The default is 8. 10516Note that this is an ABI change, even though many library function 10517interfaces will be unaffected, if they don't use SIMD vector modes 10518in places where they affect size and/or alignment of relevant types. 10519 10520@item -msplit-vecmove-early 10521@opindex msplit-vecmove-early 10522Split vector moves into single word moves before reload. In theory this 10523could give better register allocation, but so far the reverse seems to be 10524generally the case. 10525 10526@item -m1reg-@var{reg} 10527@opindex m1reg- 10528Specify a register to hold the constant @minus{}1, which makes loading small negative 10529constants and certain bitmasks faster. 10530Allowable values for reg are r43 and r63, which specify to use that register 10531as a fixed register, and none, which means that no register is used for this 10532purpose. The default is @option{-m1reg-none}. 10533 10534@end table 10535 10536@node ARM Options 10537@subsection ARM Options 10538@cindex ARM options 10539 10540These @samp{-m} options are defined for Advanced RISC Machines (ARM) 10541architectures: 10542 10543@table @gcctabopt 10544@item -mabi=@var{name} 10545@opindex mabi 10546Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu}, 10547@samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}. 10548 10549@item -mapcs-frame 10550@opindex mapcs-frame 10551Generate a stack frame that is compliant with the ARM Procedure Call 10552Standard for all functions, even if this is not strictly necessary for 10553correct execution of the code. Specifying @option{-fomit-frame-pointer} 10554with this option will cause the stack frames not to be generated for 10555leaf functions. The default is @option{-mno-apcs-frame}. 10556 10557@item -mapcs 10558@opindex mapcs 10559This is a synonym for @option{-mapcs-frame}. 10560 10561@ignore 10562@c not currently implemented 10563@item -mapcs-stack-check 10564@opindex mapcs-stack-check 10565Generate code to check the amount of stack space available upon entry to 10566every function (that actually uses some stack space). If there is 10567insufficient space available then either the function 10568@samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be 10569called, depending upon the amount of stack space required. The runtime 10570system is required to provide these functions. The default is 10571@option{-mno-apcs-stack-check}, since this produces smaller code. 10572 10573@c not currently implemented 10574@item -mapcs-float 10575@opindex mapcs-float 10576Pass floating-point arguments using the floating-point registers. This is 10577one of the variants of the APCS@. This option is recommended if the 10578target hardware has a floating-point unit or if a lot of floating-point 10579arithmetic is going to be performed by the code. The default is 10580@option{-mno-apcs-float}, since integer only code is slightly increased in 10581size if @option{-mapcs-float} is used. 10582 10583@c not currently implemented 10584@item -mapcs-reentrant 10585@opindex mapcs-reentrant 10586Generate reentrant, position independent code. The default is 10587@option{-mno-apcs-reentrant}. 10588@end ignore 10589 10590@item -mthumb-interwork 10591@opindex mthumb-interwork 10592Generate code that supports calling between the ARM and Thumb 10593instruction sets. Without this option, on pre-v5 architectures, the 10594two instruction sets cannot be reliably used inside one program. The 10595default is @option{-mno-thumb-interwork}, since slightly larger code 10596is generated when @option{-mthumb-interwork} is specified. In AAPCS 10597configurations this option is meaningless. 10598 10599@item -mno-sched-prolog 10600@opindex mno-sched-prolog 10601Prevent the reordering of instructions in the function prologue, or the 10602merging of those instruction with the instructions in the function's 10603body. This means that all functions will start with a recognizable set 10604of instructions (or in fact one of a choice from a small set of 10605different function prologues), and this information can be used to 10606locate the start if functions inside an executable piece of code. The 10607default is @option{-msched-prolog}. 10608 10609@item -mfloat-abi=@var{name} 10610@opindex mfloat-abi 10611Specifies which floating-point ABI to use. Permissible values 10612are: @samp{soft}, @samp{softfp} and @samp{hard}. 10613 10614Specifying @samp{soft} causes GCC to generate output containing 10615library calls for floating-point operations. 10616@samp{softfp} allows the generation of code using hardware floating-point 10617instructions, but still uses the soft-float calling conventions. 10618@samp{hard} allows generation of floating-point instructions 10619and uses FPU-specific calling conventions. 10620 10621The default depends on the specific target configuration. Note that 10622the hard-float and soft-float ABIs are not link-compatible; you must 10623compile your entire program with the same ABI, and link with a 10624compatible set of libraries. 10625 10626@item -mlittle-endian 10627@opindex mlittle-endian 10628Generate code for a processor running in little-endian mode. This is 10629the default for all standard configurations. 10630 10631@item -mbig-endian 10632@opindex mbig-endian 10633Generate code for a processor running in big-endian mode; the default is 10634to compile code for a little-endian processor. 10635 10636@item -mwords-little-endian 10637@opindex mwords-little-endian 10638This option only applies when generating code for big-endian processors. 10639Generate code for a little-endian word order but a big-endian byte 10640order. That is, a byte order of the form @samp{32107654}. Note: this 10641option should only be used if you require compatibility with code for 10642big-endian ARM processors generated by versions of the compiler prior to 106432.8. This option is now deprecated. 10644 10645@item -mcpu=@var{name} 10646@opindex mcpu 10647This specifies the name of the target ARM processor. GCC uses this name 10648to determine what kind of instructions it can emit when generating 10649assembly code. Permissible names are: @samp{arm2}, @samp{arm250}, 10650@samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610}, 10651@samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm}, 10652@samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700}, 10653@samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100}, 10654@samp{arm720}, 10655@samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s}, 10656@samp{arm710t}, @samp{arm720t}, @samp{arm740t}, 10657@samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100}, 10658@samp{strongarm1110}, 10659@samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920}, 10660@samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s}, 10661@samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi}, 10662@samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s}, 10663@samp{arm10e}, @samp{arm1020e}, @samp{arm1022e}, 10664@samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp}, 10665@samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s}, 10666@samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9}, 10667@samp{cortex-a15}, @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5}, 10668@samp{cortex-m4}, @samp{cortex-m3}, 10669@samp{cortex-m1}, 10670@samp{cortex-m0}, 10671@samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}, 10672@samp{fa526}, @samp{fa626}, 10673@samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}. 10674 10675 10676@option{-mcpu=generic-@var{arch}} is also permissible, and is 10677equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}. 10678See @option{-mtune} for more information. 10679 10680@option{-mcpu=native} causes the compiler to auto-detect the CPU 10681of the build computer. At present, this feature is only supported on 10682Linux, and not all architectures are recognized. If the auto-detect is 10683unsuccessful the option has no effect. 10684 10685@item -mtune=@var{name} 10686@opindex mtune 10687This option is very similar to the @option{-mcpu=} option, except that 10688instead of specifying the actual target processor type, and hence 10689restricting which instructions can be used, it specifies that GCC should 10690tune the performance of the code as if the target were of the type 10691specified in this option, but still choosing the instructions that it 10692will generate based on the CPU specified by a @option{-mcpu=} option. 10693For some ARM implementations better performance can be obtained by using 10694this option. 10695 10696@option{-mtune=generic-@var{arch}} specifies that GCC should tune the 10697performance for a blend of processors within architecture @var{arch}. 10698The aim is to generate code that run well on the current most popular 10699processors, balancing between optimizations that benefit some CPUs in the 10700range, and avoiding performance pitfalls of other CPUs. The effects of 10701this option may change in future GCC versions as CPU models come and go. 10702 10703@option{-mtune=native} causes the compiler to auto-detect the CPU 10704of the build computer. At present, this feature is only supported on 10705Linux, and not all architectures are recognized. If the auto-detect is 10706unsuccessful the option has no effect. 10707 10708@item -march=@var{name} 10709@opindex march 10710This specifies the name of the target ARM architecture. GCC uses this 10711name to determine what kind of instructions it can emit when generating 10712assembly code. This option can be used in conjunction with or instead 10713of the @option{-mcpu=} option. Permissible names are: @samp{armv2}, 10714@samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t}, 10715@samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te}, 10716@samp{armv6}, @samp{armv6j}, 10717@samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m}, 10718@samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, 10719@samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}. 10720 10721@option{-march=native} causes the compiler to auto-detect the architecture 10722of the build computer. At present, this feature is only supported on 10723Linux, and not all architectures are recognized. If the auto-detect is 10724unsuccessful the option has no effect. 10725 10726@item -mfpu=@var{name} 10727@itemx -mfpe=@var{number} 10728@itemx -mfp=@var{number} 10729@opindex mfpu 10730@opindex mfpe 10731@opindex mfp 10732This specifies what floating-point hardware (or hardware emulation) is 10733available on the target. Permissible names are: @samp{fpa}, @samp{fpe2}, 10734@samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16}, 10735@samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16}, 10736@samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16}, 10737@samp{fpv4-sp-d16} and @samp{neon-vfpv4}. 10738@option{-mfp} and @option{-mfpe} are synonyms for 10739@option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions 10740of GCC@. 10741 10742If @option{-msoft-float} is specified this specifies the format of 10743floating-point values. 10744 10745If the selected floating-point hardware includes the NEON extension 10746(e.g. @option{-mfpu}=@samp{neon}), note that floating-point 10747operations will not be used by GCC's auto-vectorization pass unless 10748@option{-funsafe-math-optimizations} is also specified. This is 10749because NEON hardware does not fully implement the IEEE 754 standard for 10750floating-point arithmetic (in particular denormal values are treated as 10751zero), so the use of NEON instructions may lead to a loss of precision. 10752 10753@item -mfp16-format=@var{name} 10754@opindex mfp16-format 10755Specify the format of the @code{__fp16} half-precision floating-point type. 10756Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative}; 10757the default is @samp{none}, in which case the @code{__fp16} type is not 10758defined. @xref{Half-Precision}, for more information. 10759 10760@item -mstructure-size-boundary=@var{n} 10761@opindex mstructure-size-boundary 10762The size of all structures and unions will be rounded up to a multiple 10763of the number of bits set by this option. Permissible values are 8, 32 10764and 64. The default value varies for different toolchains. For the COFF 10765targeted toolchain the default value is 8. A value of 64 is only allowed 10766if the underlying ABI supports it. 10767 10768Specifying the larger number can produce faster, more efficient code, but 10769can also increase the size of the program. Different values are potentially 10770incompatible. Code compiled with one value cannot necessarily expect to 10771work with code or libraries compiled with another value, if they exchange 10772information using structures or unions. 10773 10774@item -mabort-on-noreturn 10775@opindex mabort-on-noreturn 10776Generate a call to the function @code{abort} at the end of a 10777@code{noreturn} function. It will be executed if the function tries to 10778return. 10779 10780@item -mlong-calls 10781@itemx -mno-long-calls 10782@opindex mlong-calls 10783@opindex mno-long-calls 10784Tells the compiler to perform function calls by first loading the 10785address of the function into a register and then performing a subroutine 10786call on this register. This switch is needed if the target function 10787will lie outside of the 64 megabyte addressing range of the offset based 10788version of subroutine call instruction. 10789 10790Even if this switch is enabled, not all function calls will be turned 10791into long calls. The heuristic is that static functions, functions 10792that have the @samp{short-call} attribute, functions that are inside 10793the scope of a @samp{#pragma no_long_calls} directive and functions whose 10794definitions have already been compiled within the current compilation 10795unit, will not be turned into long calls. The exception to this rule is 10796that weak function definitions, functions with the @samp{long-call} 10797attribute or the @samp{section} attribute, and functions that are within 10798the scope of a @samp{#pragma long_calls} directive, will always be 10799turned into long calls. 10800 10801This feature is not enabled by default. Specifying 10802@option{-mno-long-calls} will restore the default behavior, as will 10803placing the function calls within the scope of a @samp{#pragma 10804long_calls_off} directive. Note these switches have no effect on how 10805the compiler generates code to handle function calls via function 10806pointers. 10807 10808@item -msingle-pic-base 10809@opindex msingle-pic-base 10810Treat the register used for PIC addressing as read-only, rather than 10811loading it in the prologue for each function. The runtime system is 10812responsible for initializing this register with an appropriate value 10813before execution begins. 10814 10815@item -mpic-register=@var{reg} 10816@opindex mpic-register 10817Specify the register to be used for PIC addressing. The default is R10 10818unless stack-checking is enabled, when R9 is used. 10819 10820@item -mcirrus-fix-invalid-insns 10821@opindex mcirrus-fix-invalid-insns 10822@opindex mno-cirrus-fix-invalid-insns 10823Insert NOPs into the instruction stream to in order to work around 10824problems with invalid Maverick instruction combinations. This option 10825is only valid if the @option{-mcpu=ep9312} option has been used to 10826enable generation of instructions for the Cirrus Maverick floating-point 10827co-processor. This option is not enabled by default, since the 10828problem is only present in older Maverick implementations. The default 10829can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns} 10830switch. 10831 10832@item -mpoke-function-name 10833@opindex mpoke-function-name 10834Write the name of each function into the text section, directly 10835preceding the function prologue. The generated code is similar to this: 10836 10837@smallexample 10838 t0 10839 .ascii "arm_poke_function_name", 0 10840 .align 10841 t1 10842 .word 0xff000000 + (t1 - t0) 10843 arm_poke_function_name 10844 mov ip, sp 10845 stmfd sp!, @{fp, ip, lr, pc@} 10846 sub fp, ip, #4 10847@end smallexample 10848 10849When performing a stack backtrace, code can inspect the value of 10850@code{pc} stored at @code{fp + 0}. If the trace function then looks at 10851location @code{pc - 12} and the top 8 bits are set, then we know that 10852there is a function name embedded immediately preceding this location 10853and has length @code{((pc[-3]) & 0xff000000)}. 10854 10855@item -mthumb 10856@itemx -marm 10857@opindex marm 10858@opindex mthumb 10859 10860Select between generating code that executes in ARM and Thumb 10861states. The default for most configurations is to generate code 10862that executes in ARM state, but the default can be changed by 10863configuring GCC with the @option{--with-mode=}@var{state} 10864configure option. 10865 10866@item -mtpcs-frame 10867@opindex mtpcs-frame 10868Generate a stack frame that is compliant with the Thumb Procedure Call 10869Standard for all non-leaf functions. (A leaf function is one that does 10870not call any other functions.) The default is @option{-mno-tpcs-frame}. 10871 10872@item -mtpcs-leaf-frame 10873@opindex mtpcs-leaf-frame 10874Generate a stack frame that is compliant with the Thumb Procedure Call 10875Standard for all leaf functions. (A leaf function is one that does 10876not call any other functions.) The default is @option{-mno-apcs-leaf-frame}. 10877 10878@item -mcallee-super-interworking 10879@opindex mcallee-super-interworking 10880Gives all externally visible functions in the file being compiled an ARM 10881instruction set header which switches to Thumb mode before executing the 10882rest of the function. This allows these functions to be called from 10883non-interworking code. This option is not valid in AAPCS configurations 10884because interworking is enabled by default. 10885 10886@item -mcaller-super-interworking 10887@opindex mcaller-super-interworking 10888Allows calls via function pointers (including virtual functions) to 10889execute correctly regardless of whether the target code has been 10890compiled for interworking or not. There is a small overhead in the cost 10891of executing a function pointer if this option is enabled. This option 10892is not valid in AAPCS configurations because interworking is enabled 10893by default. 10894 10895@item -mtp=@var{name} 10896@opindex mtp 10897Specify the access model for the thread local storage pointer. The valid 10898models are @option{soft}, which generates calls to @code{__aeabi_read_tp}, 10899@option{cp15}, which fetches the thread pointer from @code{cp15} directly 10900(supported in the arm6k architecture), and @option{auto}, which uses the 10901best available method for the selected processor. The default setting is 10902@option{auto}. 10903 10904@item -mtls-dialect=@var{dialect} 10905@opindex mtls-dialect 10906Specify the dialect to use for accessing thread local storage. Two 10907dialects are supported --- @option{gnu} and @option{gnu2}. The 10908@option{gnu} dialect selects the original GNU scheme for supporting 10909local and global dynamic TLS models. The @option{gnu2} dialect 10910selects the GNU descriptor scheme, which provides better performance 10911for shared libraries. The GNU descriptor scheme is compatible with 10912the original scheme, but does require new assembler, linker and 10913library support. Initial and local exec TLS models are unaffected by 10914this option and always use the original scheme. 10915 10916@item -mword-relocations 10917@opindex mword-relocations 10918Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32). 10919This is enabled by default on targets (uClinux, SymbianOS) where the runtime 10920loader imposes this restriction, and when @option{-fpic} or @option{-fPIC} 10921is specified. 10922 10923@item -mfix-cortex-m3-ldrd 10924@opindex mfix-cortex-m3-ldrd 10925Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions 10926with overlapping destination and base registers are used. This option avoids 10927generating these instructions. This option is enabled by default when 10928@option{-mcpu=cortex-m3} is specified. 10929 10930@item -munaligned-access 10931@itemx -mno-unaligned-access 10932@opindex munaligned-access 10933@opindex mno-unaligned-access 10934Enables (or disables) reading and writing of 16- and 32- bit values 10935from addresses that are not 16- or 32- bit aligned. By default 10936unaligned access is disabled for all pre-ARMv6 and all ARMv6-M 10937architectures, and enabled for all other architectures. If unaligned 10938access is not enabled then words in packed data structures will be 10939accessed a byte at a time. 10940 10941The ARM attribute @code{Tag_CPU_unaligned_access} will be set in the 10942generated object file to either true or false, depending upon the 10943setting of this option. If unaligned access is enabled then the 10944preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} will also be 10945defined. 10946 10947@end table 10948 10949@node AVR Options 10950@subsection AVR Options 10951@cindex AVR Options 10952 10953@table @gcctabopt 10954@item -mmcu=@var{mcu} 10955@opindex mmcu 10956Specify Atmel AVR instruction set architectures (ISA) or MCU type. 10957 10958For a complete list of @var{mcu} values that are supported by @command{avr-gcc}, 10959see the compiler output when called with the @option{--help=target} 10960command line option. 10961The default for this option is@tie{}@code{avr2}. 10962 10963GCC supports the following AVR devices and ISAs: 10964 10965@table @code 10966 10967@item avr2 10968``Classic'' devices with up to 8@tie{}KiB of program memory. 10969@*@var{mcu}@tie{}= @code{at90c8534}, @code{at90s2313}, 10970@code{at90s2323}, @code{at90s2333}, @code{at90s2343}, 10971@code{at90s4414}, @code{at90s4433}, @code{at90s4434}, 10972@code{at90s8515}, @code{at90s8535}, @code{attiny22}, @code{attiny26}. 10973 10974@item avr25 10975``Classic'' devices with up to 8@tie{}KiB of program memory and with 10976the @code{MOVW} instruction. 10977@*@var{mcu}@tie{}= @code{at86rf401}, @code{ata6289}, @code{attiny13}, 10978@code{attiny13a}, @code{attiny2313}, @code{attiny2313a}, 10979@code{attiny24}, @code{attiny24a}, @code{attiny25}, @code{attiny261}, 10980@code{attiny261a}, @code{attiny4313}, @code{attiny43u}, 10981@code{attiny44}, @code{attiny44a}, @code{attiny45}, @code{attiny461}, 10982@code{attiny461a}, @code{attiny48}, @code{attiny84}, @code{attiny84a}, 10983@code{attiny85}, @code{attiny861}, @code{attiny861a}, @code{attiny87}, 10984@code{attiny88}. 10985 10986@item avr3 10987``Classic'' devices with 16@tie{}KiB up to 64@tie{}KiB of program memory. 10988@*@var{mcu}@tie{}= @code{at43usb355}, @code{at76c711}. 10989 10990@item avr31 10991``Classic'' devices with 128@tie{}KiB of program memory. 10992@*@var{mcu}@tie{}= @code{at43usb320}, @code{atmega103}. 10993 10994@item avr35 10995``Classic'' devices with 16@tie{}KiB up to 64@tie{}KiB of program 10996memory and with the @code{MOVW} instruction. 10997@*@var{mcu}@tie{}= @code{at90usb162}, @code{at90usb82}, 10998@code{atmega16u2}, @code{atmega32u2}, @code{atmega8u2}, 10999@code{attiny167}. 11000 11001@item avr4 11002``Enhanced'' devices with up to 8@tie{}KiB of program memory. 11003@*@var{mcu}@tie{}= @code{at90pwm1}, @code{at90pwm2}, @code{at90pwm2b}, 11004@code{at90pwm3}, @code{at90pwm3b}, @code{at90pwm81}, @code{atmega48}, 11005@code{atmega48a}, @code{atmega48p}, @code{atmega8}, @code{atmega8515}, 11006@code{atmega8535}, @code{atmega88}, @code{atmega88a}, 11007@code{atmega88p}, @code{atmega88pa}, @code{atmega8hva}. 11008 11009@item avr5 11010``Enhanced'' devices with 16@tie{}KiB up to 64@tie{}KiB of program memory. 11011@*@var{mcu}@tie{}= @code{at90can32}, @code{at90can64}, 11012@code{at90pwm216}, @code{at90pwm316}, @code{at90scr100}, 11013@code{at90usb646}, @code{at90usb647}, @code{at94k}, @code{atmega16}, 11014@code{atmega161}, @code{atmega162}, @code{atmega163}, 11015@code{atmega164a}, @code{atmega164p}, @code{atmega165}, 11016@code{atmega165a}, @code{atmega165p}, @code{atmega168}, 11017@code{atmega168a}, @code{atmega168p}, @code{atmega169}, 11018@code{atmega169a}, @code{atmega169p}, @code{atmega169pa}, 11019@code{atmega16a}, @code{atmega16hva}, @code{atmega16hva2}, 11020@code{atmega16hvb}, @code{atmega16m1}, @code{atmega16u4}, 11021@code{atmega32}, @code{atmega323}, @code{atmega324a}, 11022@code{atmega324p}, @code{atmega324pa}, @code{atmega325}, 11023@code{atmega3250}, @code{atmega3250a}, @code{atmega3250p}, 11024@code{atmega325a}, @code{atmega325p}, @code{atmega328}, 11025@code{atmega328p}, @code{atmega329}, @code{atmega3290}, 11026@code{atmega3290a}, @code{atmega3290p}, @code{atmega329a}, 11027@code{atmega329p}, @code{atmega329pa}, @code{atmega32c1}, 11028@code{atmega32hvb}, @code{atmega32m1}, @code{atmega32u4}, 11029@code{atmega32u6}, @code{atmega406}, @code{atmega64}, 11030@code{atmega640}, @code{atmega644}, @code{atmega644a}, 11031@code{atmega644p}, @code{atmega644pa}, @code{atmega645}, 11032@code{atmega6450}, @code{atmega6450a}, @code{atmega6450p}, 11033@code{atmega645a}, @code{atmega645p}, @code{atmega649}, 11034@code{atmega6490}, @code{atmega649a}, @code{atmega649p}, 11035@code{atmega64c1}, @code{atmega64hve}, @code{atmega64m1}, 11036@code{m3000}. 11037 11038@item avr51 11039``Enhanced'' devices with 128@tie{}KiB of program memory. 11040@*@var{mcu}@tie{}= @code{at90can128}, @code{at90usb1286}, 11041@code{at90usb1287}, @code{atmega128}, @code{atmega1280}, 11042@code{atmega1281}, @code{atmega1284p}, @code{atmega128rfa1}. 11043 11044@item avr6 11045``Enhanced'' devices with 3-byte PC, i.e.@: with more than 11046128@tie{}KiB of program memory. 11047@*@var{mcu}@tie{}= @code{atmega2560}, @code{atmega2561}. 11048 11049@item avrxmega2 11050``XMEGA'' devices with more than 8@tie{}KiB and up to 64@tie{}KiB of 11051program memory. 11052@*@var{mcu}@tie{}= @code{atxmega16a4}, @code{atxmega16d4}, 11053@code{atxmega16x1}, @code{atxmega32a4}, @code{atxmega32d4}, 11054@code{atxmega32x1}. 11055 11056@item avrxmega4 11057``XMEGA'' devices with more than 64@tie{}KiB and up to 128@tie{}KiB of 11058program memory. 11059@*@var{mcu}@tie{}= @code{atxmega64a3}, @code{atxmega64d3}. 11060 11061@item avrxmega5 11062``XMEGA'' devices with more than 64@tie{}KiB and up to 128@tie{}KiB of 11063program memory and more than 64@tie{}KiB of RAM. 11064@*@var{mcu}@tie{}= @code{atxmega64a1}, @code{atxmega64a1u}. 11065 11066@item avrxmega6 11067``XMEGA'' devices with more than 128@tie{}KiB of program memory. 11068@*@var{mcu}@tie{}= @code{atxmega128a3}, @code{atxmega128d3}, 11069@code{atxmega192a3}, @code{atxmega192d3}, @code{atxmega256a3}, 11070@code{atxmega256a3b}, @code{atxmega256a3bu}, @code{atxmega256d3}. 11071 11072@item avrxmega7 11073``XMEGA'' devices with more than 128@tie{}KiB of program memory and 11074more than 64@tie{}KiB of RAM. 11075@*@var{mcu}@tie{}= @code{atxmega128a1}, @code{atxmega128a1u}. 11076 11077@item avr1 11078This ISA is implemented by the minimal AVR core and supported for 11079assembler only. 11080@*@var{mcu}@tie{}= @code{at90s1200}, @code{attiny11}, @code{attiny12}, 11081@code{attiny15}, @code{attiny28}. 11082 11083@end table 11084 11085@item -maccumulate-args 11086@opindex maccumulate-args 11087Accumulate outgoing function arguments and acquire/release the needed 11088stack space for outgoing function arguments once in function 11089prologue/epilogue. Without this option, outgoing arguments are pushed 11090before calling a function and popped afterwards. 11091 11092Popping the arguments after the function call can be expensive on 11093AVR so that accumulating the stack space might lead to smaller 11094executables because arguments need not to be removed from the 11095stack after such a function call. 11096 11097This option can lead to reduced code size for functions that perform 11098several calls to functions that get their arguments on the stack like 11099calls to printf-like functions. 11100 11101@item -mbranch-cost=@var{cost} 11102@opindex mbranch-cost 11103Set the branch costs for conditional branch instructions to 11104@var{cost}. Reasonable values for @var{cost} are small, non-negative 11105integers. The default branch cost is 0. 11106 11107@item -mcall-prologues 11108@opindex mcall-prologues 11109Functions prologues/epilogues are expanded as calls to appropriate 11110subroutines. Code size is smaller. 11111 11112@item -mint8 11113@opindex mint8 11114Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a 11115@code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes, 11116and @code{long long} is 4 bytes. Please note that this option does not 11117conform to the C standards, but it results in smaller code 11118size. 11119 11120@item -mno-interrupts 11121@opindex mno-interrupts 11122Generated code is not compatible with hardware interrupts. 11123Code size is smaller. 11124 11125@item -mrelax 11126@opindex mrelax 11127Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter 11128@code{RCALL} resp.@: @code{RJMP} instruction if applicable. 11129Setting @code{-mrelax} just adds the @code{--relax} option to the 11130linker command line when the linker is called. 11131 11132Jump relaxing is performed by the linker because jump offsets are not 11133known before code is located. Therefore, the assembler code generated by the 11134compiler is the same, but the instructions in the executable may 11135differ from instructions in the assembler code. 11136 11137Relaxing must be turned on if linker stubs are needed, see the 11138section on @code{EIND} and linker stubs below. 11139 11140@item -mshort-calls 11141@opindex mshort-calls 11142Use @code{RCALL}/@code{RJMP} instructions even on devices with 1114316@tie{}KiB or more of program memory, i.e.@: on devices that 11144have the @code{CALL} and @code{JMP} instructions. 11145See also the @code{-mrelax} command line option. 11146 11147@item -msp8 11148@opindex msp8 11149Treat the stack pointer register as an 8-bit register, 11150i.e.@: assume the high byte of the stack pointer is zero. 11151In general, you don't need to set this option by hand. 11152 11153This option is used internally by the compiler to select and 11154build multilibs for architectures @code{avr2} and @code{avr25}. 11155These architectures mix devices with and without @code{SPH}. 11156For any setting other than @code{-mmcu=avr2} or @code{-mmcu=avr25} 11157the compiler driver will add or remove this option from the compiler 11158proper's command line, because the compiler then knows if the device 11159or architecture has an 8-bit stack pointer and thus no @code{SPH} 11160register or not. 11161 11162@item -mstrict-X 11163@opindex mstrict-X 11164Use address register @code{X} in a way proposed by the hardware. This means 11165that @code{X} is only used in indirect, post-increment or 11166pre-decrement addressing. 11167 11168Without this option, the @code{X} register may be used in the same way 11169as @code{Y} or @code{Z} which then is emulated by additional 11170instructions. 11171For example, loading a value with @code{X+const} addressing with a 11172small non-negative @code{const < 64} to a register @var{Rn} is 11173performed as 11174 11175@example 11176adiw r26, const ; X += const 11177ld @var{Rn}, X ; @var{Rn} = *X 11178sbiw r26, const ; X -= const 11179@end example 11180 11181@item -mtiny-stack 11182@opindex mtiny-stack 11183Only change the lower 8@tie{}bits of the stack pointer. 11184@end table 11185 11186@subsubsection @code{EIND} and Devices with more than 128 Ki Bytes of Flash 11187@cindex @code{EIND} 11188Pointers in the implementation are 16@tie{}bits wide. 11189The address of a function or label is represented as word address so 11190that indirect jumps and calls can target any code address in the 11191range of 64@tie{}Ki words. 11192 11193In order to facilitate indirect jump on devices with more than 128@tie{}Ki 11194bytes of program memory space, there is a special function register called 11195@code{EIND} that serves as most significant part of the target address 11196when @code{EICALL} or @code{EIJMP} instructions are used. 11197 11198Indirect jumps and calls on these devices are handled as follows by 11199the compiler and are subject to some limitations: 11200 11201@itemize @bullet 11202 11203@item 11204The compiler never sets @code{EIND}. 11205 11206@item 11207The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP} 11208instructions or might read @code{EIND} directly in order to emulate an 11209indirect call/jump by means of a @code{RET} instruction. 11210 11211@item 11212The compiler assumes that @code{EIND} never changes during the startup 11213code or during the application. In particular, @code{EIND} is not 11214saved/restored in function or interrupt service routine 11215prologue/epilogue. 11216 11217@item 11218For indirect calls to functions and computed goto, the linker 11219generates @emph{stubs}. Stubs are jump pads sometimes also called 11220@emph{trampolines}. Thus, the indirect call/jump jumps to such a stub. 11221The stub contains a direct jump to the desired address. 11222 11223@item 11224Linker relaxation must be turned on so that the linker will generate 11225the stubs correctly an all situaltion. See the compiler option 11226@code{-mrelax} and the linler option @code{--relax}. 11227There are corner cases where the linker is supposed to generate stubs 11228but aborts without relaxation and without a helpful error message. 11229 11230@item 11231The default linker script is arranged for code with @code{EIND = 0}. 11232If code is supposed to work for a setup with @code{EIND != 0}, a custom 11233linker script has to be used in order to place the sections whose 11234name start with @code{.trampolines} into the segment where @code{EIND} 11235points to. 11236 11237@item 11238The startup code from libgcc never sets @code{EIND}. 11239Notice that startup code is a blend of code from libgcc and AVR-LibC. 11240For the impact of AVR-LibC on @code{EIND}, see the 11241@w{@uref{http://nongnu.org/avr-libc/user-manual,AVR-LibC user manual}}. 11242 11243@item 11244It is legitimate for user-specific startup code to set up @code{EIND} 11245early, for example by means of initialization code located in 11246section @code{.init3}. Such code runs prior to general startup code 11247that initializes RAM and calls constructors, but after the bit 11248of startup code from AVR-LibC that sets @code{EIND} to the segment 11249where the vector table is located. 11250@example 11251#include <avr/io.h> 11252 11253static void 11254__attribute__((section(".init3"),naked,used,no_instrument_function)) 11255init3_set_eind (void) 11256@{ 11257 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t" 11258 "out %i0,r24" :: "n" (&EIND) : "r24","memory"); 11259@} 11260@end example 11261 11262@noindent 11263The @code{__trampolines_start} symbol is defined in the linker script. 11264 11265@item 11266Stubs are generated automatically by the linker if 11267the following two conditions are met: 11268@itemize @minus 11269 11270@item The address of a label is taken by means of the @code{gs} modifier 11271(short for @emph{generate stubs}) like so: 11272@example 11273LDI r24, lo8(gs(@var{func})) 11274LDI r25, hi8(gs(@var{func})) 11275@end example 11276@item The final location of that label is in a code segment 11277@emph{outside} the segment where the stubs are located. 11278@end itemize 11279 11280@item 11281The compiler emits such @code{gs} modifiers for code labels in the 11282following situations: 11283@itemize @minus 11284@item Taking address of a function or code label. 11285@item Computed goto. 11286@item If prologue-save function is used, see @option{-mcall-prologues} 11287command-line option. 11288@item Switch/case dispatch tables. If you do not want such dispatch 11289tables you can specify the @option{-fno-jump-tables} command-line option. 11290@item C and C++ constructors/destructors called during startup/shutdown. 11291@item If the tools hit a @code{gs()} modifier explained above. 11292@end itemize 11293 11294@item 11295Jumping to non-symbolic addresses like so is @emph{not} supported: 11296 11297@example 11298int main (void) 11299@{ 11300 /* Call function at word address 0x2 */ 11301 return ((int(*)(void)) 0x2)(); 11302@} 11303@end example 11304 11305Instead, a stub has to be set up, i.e.@: the function has to be called 11306through a symbol (@code{func_4} in the example): 11307 11308@example 11309int main (void) 11310@{ 11311 extern int func_4 (void); 11312 11313 /* Call function at byte address 0x4 */ 11314 return func_4(); 11315@} 11316@end example 11317 11318and the application be linked with @code{-Wl,--defsym,func_4=0x4}. 11319Alternatively, @code{func_4} can be defined in the linker script. 11320@end itemize 11321 11322@subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers 11323@cindex @code{RAMPD} 11324@cindex @code{RAMPX} 11325@cindex @code{RAMPY} 11326@cindex @code{RAMPZ} 11327Some AVR devices support memories larger than the 64@tie{}KiB range 11328that can be accessed with 16-bit pointers. To access memory locations 11329outside this 64@tie{}KiB range, the contentent of a @code{RAMP} 11330register is used as high part of the address: 11331The @code{X}, @code{Y}, @code{Z} address register is concatenated 11332with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function 11333register, respectively, to get a wide address. Similarly, 11334@code{RAMPD} is used together with direct addressing. 11335 11336@itemize 11337@item 11338The startup code initializes the @code{RAMP} special function 11339registers with zero. 11340 11341@item 11342If a @ref{AVR Named Address Spaces,named address space} other than 11343generic or @code{__flash} is used, then @code{RAMPZ} is set 11344as needed before the operation. 11345 11346@item 11347If the device supports RAM larger than 64@tie{KiB} and the compiler 11348needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ} 11349is reset to zero after the operation. 11350 11351@item 11352If the device comes with a specific @code{RAMP} register, the ISR 11353prologue/epilogue saves/restores that SFR and initializes it with 11354zero in case the ISR code might (implicitly) use it. 11355 11356@item 11357RAM larger than 64@tie{KiB} is not supported by GCC for AVR targets. 11358If you use inline assembler to read from locations outside the 1135916-bit address range and change one of the @code{RAMP} registers, 11360you must reset it to zero after the access. 11361 11362@end itemize 11363 11364@subsubsection AVR Built-in Macros 11365 11366GCC defines several built-in macros so that the user code can test 11367for the presence or absence of features. Almost any of the following 11368built-in macros are deduced from device capabilities and thus 11369triggered by the @code{-mmcu=} command-line option. 11370 11371For even more AVR-specific built-in macros see 11372@ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}. 11373 11374@table @code 11375 11376@item __AVR_@var{Device}__ 11377Setting @code{-mmcu=@var{device}} defines this built-in macro which reflects 11378the device's name. For example, @code{-mmcu=atmega8} defines the 11379built-in macro @code{__AVR_ATmega8__}, @code{-mmcu=attiny261a} defines 11380@code{__AVR_ATtiny261A__}, etc. 11381 11382The built-in macros' names follow 11383the scheme @code{__AVR_@var{Device}__} where @var{Device} is 11384the device name as from the AVR user manual. The difference between 11385@var{Device} in the built-in macro and @var{device} in 11386@code{-mmcu=@var{device}} is that the latter is always lowercase. 11387 11388@item __AVR_HAVE_ELPM__ 11389The device has the the @code{ELPM} instruction. 11390 11391@item __AVR_HAVE_ELPMX__ 11392The device has the @code{ELPM R@var{n},Z} and @code{ELPM 11393R@var{n},Z+} instructions. 11394 11395@item __AVR_HAVE_MOVW__ 11396The device has the @code{MOVW} instruction to perform 16-bit 11397register-register moves. 11398 11399@item __AVR_HAVE_LPMX__ 11400The device has the @code{LPM R@var{n},Z} and 11401@code{LPM R@var{n},Z+} instructions. 11402 11403@item __AVR_HAVE_MUL__ 11404The device has a hardware multiplier. 11405 11406@item __AVR_HAVE_JMP_CALL__ 11407The device has the @code{JMP} and @code{CALL} instructions. 11408This is the case for devices with at least 16@tie{}KiB of program 11409memory and if @code{-mshort-calls} is not set. 11410 11411@item __AVR_HAVE_EIJMP_EICALL__ 11412@item __AVR_3_BYTE_PC__ 11413The device has the @code{EIJMP} and @code{EICALL} instructions. 11414This is the case for devices with more than 128@tie{}KiB of program memory. 11415This also means that the program counter 11416(PC) is 3@tie{}bytes wide. 11417 11418@item __AVR_2_BYTE_PC__ 11419The program counter (PC) is 2@tie{}bytes wide. This is the case for devices 11420with up to 128@tie{}KiB of program memory. 11421 11422@item __AVR_HAVE_8BIT_SP__ 11423@item __AVR_HAVE_16BIT_SP__ 11424The stack pointer (SP) register is treated as 8-bit respectively 1142516-bit register by the compiler. 11426The definition of these macros is affected by @code{-mtiny-stack}. 11427 11428@item __AVR_HAVE_SPH__ 11429@item __AVR_SP8__ 11430The device has the SPH (high part of stack pointer) special function 11431register or has an 8-bit stack pointer, respectively. 11432The definition of these macros is affected by @code{-mmcu=} and 11433in the cases of @code{-mmcu=avr2} and @code{-mmcu=avr25} also 11434by @code{-msp8}. 11435 11436@item __AVR_HAVE_RAMPD__ 11437@item __AVR_HAVE_RAMPX__ 11438@item __AVR_HAVE_RAMPY__ 11439@item __AVR_HAVE_RAMPZ__ 11440The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY}, 11441@code{RAMPZ} special function register, respectively. 11442 11443@item __NO_INTERRUPTS__ 11444This macro reflects the @code{-mno-interrupts} command line option. 11445 11446@item __AVR_ERRATA_SKIP__ 11447@item __AVR_ERRATA_SKIP_JMP_CALL__ 11448Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit 11449instructions because of a hardware erratum. Skip instructions are 11450@code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}. 11451The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also 11452set. 11453 11454@item __AVR_SFR_OFFSET__=@var{offset} 11455Instructions that can address I/O special function registers directly 11456like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different 11457address as if addressed by an instruction to access RAM like @code{LD} 11458or @code{STS}. This offset depends on the device architecture and has 11459to be subtracted from the RAM address in order to get the 11460respective I/O@tie{}address. 11461 11462@item __WITH_AVRLIBC__ 11463The compiler is configured to be used together with AVR-Libc. 11464See the @code{--with-avrlibc} configure option. 11465 11466@end table 11467 11468@node Blackfin Options 11469@subsection Blackfin Options 11470@cindex Blackfin Options 11471 11472@table @gcctabopt 11473@item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} 11474@opindex mcpu= 11475Specifies the name of the target Blackfin processor. Currently, @var{cpu} 11476can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518}, 11477@samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526}, 11478@samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533}, 11479@samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539}, 11480@samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549}, 11481@samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m}, 11482@samp{bf561}, @samp{bf592}. 11483The optional @var{sirevision} specifies the silicon revision of the target 11484Blackfin processor. Any workarounds available for the targeted silicon revision 11485will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled. 11486If @var{sirevision} is @samp{any}, all workarounds for the targeted processor 11487will be enabled. The @code{__SILICON_REVISION__} macro is defined to two 11488hexadecimal digits representing the major and minor numbers in the silicon 11489revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__} 11490is not defined. If @var{sirevision} is @samp{any}, the 11491@code{__SILICON_REVISION__} is defined to be @code{0xffff}. 11492If this optional @var{sirevision} is not used, GCC assumes the latest known 11493silicon revision of the targeted Blackfin processor. 11494 11495Support for @samp{bf561} is incomplete. For @samp{bf561}, 11496Only the processor macro is defined. 11497Without this option, @samp{bf532} is used as the processor by default. 11498The corresponding predefined processor macros for @var{cpu} is to 11499be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP 11500provided by libgloss to be linked in if @option{-msim} is not given. 11501 11502@item -msim 11503@opindex msim 11504Specifies that the program will be run on the simulator. This causes 11505the simulator BSP provided by libgloss to be linked in. This option 11506has effect only for @samp{bfin-elf} toolchain. 11507Certain other options, such as @option{-mid-shared-library} and 11508@option{-mfdpic}, imply @option{-msim}. 11509 11510@item -momit-leaf-frame-pointer 11511@opindex momit-leaf-frame-pointer 11512Don't keep the frame pointer in a register for leaf functions. This 11513avoids the instructions to save, set up and restore frame pointers and 11514makes an extra register available in leaf functions. The option 11515@option{-fomit-frame-pointer} removes the frame pointer for all functions, 11516which might make debugging harder. 11517 11518@item -mspecld-anomaly 11519@opindex mspecld-anomaly 11520When enabled, the compiler will ensure that the generated code does not 11521contain speculative loads after jump instructions. If this option is used, 11522@code{__WORKAROUND_SPECULATIVE_LOADS} is defined. 11523 11524@item -mno-specld-anomaly 11525@opindex mno-specld-anomaly 11526Don't generate extra code to prevent speculative loads from occurring. 11527 11528@item -mcsync-anomaly 11529@opindex mcsync-anomaly 11530When enabled, the compiler will ensure that the generated code does not 11531contain CSYNC or SSYNC instructions too soon after conditional branches. 11532If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined. 11533 11534@item -mno-csync-anomaly 11535@opindex mno-csync-anomaly 11536Don't generate extra code to prevent CSYNC or SSYNC instructions from 11537occurring too soon after a conditional branch. 11538 11539@item -mlow-64k 11540@opindex mlow-64k 11541When enabled, the compiler is free to take advantage of the knowledge that 11542the entire program fits into the low 64k of memory. 11543 11544@item -mno-low-64k 11545@opindex mno-low-64k 11546Assume that the program is arbitrarily large. This is the default. 11547 11548@item -mstack-check-l1 11549@opindex mstack-check-l1 11550Do stack checking using information placed into L1 scratchpad memory by the 11551uClinux kernel. 11552 11553@item -mid-shared-library 11554@opindex mid-shared-library 11555Generate code that supports shared libraries via the library ID method. 11556This allows for execute in place and shared libraries in an environment 11557without virtual memory management. This option implies @option{-fPIC}. 11558With a @samp{bfin-elf} target, this option implies @option{-msim}. 11559 11560@item -mno-id-shared-library 11561@opindex mno-id-shared-library 11562Generate code that doesn't assume ID based shared libraries are being used. 11563This is the default. 11564 11565@item -mleaf-id-shared-library 11566@opindex mleaf-id-shared-library 11567Generate code that supports shared libraries via the library ID method, 11568but assumes that this library or executable won't link against any other 11569ID shared libraries. That allows the compiler to use faster code for jumps 11570and calls. 11571 11572@item -mno-leaf-id-shared-library 11573@opindex mno-leaf-id-shared-library 11574Do not assume that the code being compiled won't link against any ID shared 11575libraries. Slower code will be generated for jump and call insns. 11576 11577@item -mshared-library-id=n 11578@opindex mshared-library-id 11579Specified the identification number of the ID based shared library being 11580compiled. Specifying a value of 0 will generate more compact code, specifying 11581other values will force the allocation of that number to the current 11582library but is no more space or time efficient than omitting this option. 11583 11584@item -msep-data 11585@opindex msep-data 11586Generate code that allows the data segment to be located in a different 11587area of memory from the text segment. This allows for execute in place in 11588an environment without virtual memory management by eliminating relocations 11589against the text section. 11590 11591@item -mno-sep-data 11592@opindex mno-sep-data 11593Generate code that assumes that the data segment follows the text segment. 11594This is the default. 11595 11596@item -mlong-calls 11597@itemx -mno-long-calls 11598@opindex mlong-calls 11599@opindex mno-long-calls 11600Tells the compiler to perform function calls by first loading the 11601address of the function into a register and then performing a subroutine 11602call on this register. This switch is needed if the target function 11603lies outside of the 24-bit addressing range of the offset-based 11604version of subroutine call instruction. 11605 11606This feature is not enabled by default. Specifying 11607@option{-mno-long-calls} will restore the default behavior. Note these 11608switches have no effect on how the compiler generates code to handle 11609function calls via function pointers. 11610 11611@item -mfast-fp 11612@opindex mfast-fp 11613Link with the fast floating-point library. This library relaxes some of 11614the IEEE floating-point standard's rules for checking inputs against 11615Not-a-Number (NAN), in the interest of performance. 11616 11617@item -minline-plt 11618@opindex minline-plt 11619Enable inlining of PLT entries in function calls to functions that are 11620not known to bind locally. It has no effect without @option{-mfdpic}. 11621 11622@item -mmulticore 11623@opindex mmulticore 11624Build standalone application for multicore Blackfin processor. Proper 11625start files and link scripts will be used to support multicore. 11626This option defines @code{__BFIN_MULTICORE}. It can only be used with 11627@option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with 11628@option{-mcorea} or @option{-mcoreb}. If it's used without 11629@option{-mcorea} or @option{-mcoreb}, single application/dual core 11630programming model is used. In this model, the main function of Core B 11631should be named as coreb_main. If it's used with @option{-mcorea} or 11632@option{-mcoreb}, one application per core programming model is used. 11633If this option is not used, single core application programming 11634model is used. 11635 11636@item -mcorea 11637@opindex mcorea 11638Build standalone application for Core A of BF561 when using 11639one application per core programming model. Proper start files 11640and link scripts will be used to support Core A. This option 11641defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}. 11642 11643@item -mcoreb 11644@opindex mcoreb 11645Build standalone application for Core B of BF561 when using 11646one application per core programming model. Proper start files 11647and link scripts will be used to support Core B. This option 11648defines @code{__BFIN_COREB}. When this option is used, coreb_main 11649should be used instead of main. It must be used with 11650@option{-mmulticore}. 11651 11652@item -msdram 11653@opindex msdram 11654Build standalone application for SDRAM. Proper start files and 11655link scripts will be used to put the application into SDRAM. 11656Loader should initialize SDRAM before loading the application 11657into SDRAM. This option defines @code{__BFIN_SDRAM}. 11658 11659@item -micplb 11660@opindex micplb 11661Assume that ICPLBs are enabled at run time. This has an effect on certain 11662anomaly workarounds. For Linux targets, the default is to assume ICPLBs 11663are enabled; for standalone applications the default is off. 11664@end table 11665 11666@node C6X Options 11667@subsection C6X Options 11668@cindex C6X Options 11669 11670@table @gcctabopt 11671@item -march=@var{name} 11672@opindex march 11673This specifies the name of the target architecture. GCC uses this 11674name to determine what kind of instructions it can emit when generating 11675assembly code. Permissible names are: @samp{c62x}, 11676@samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}. 11677 11678@item -mbig-endian 11679@opindex mbig-endian 11680Generate code for a big-endian target. 11681 11682@item -mlittle-endian 11683@opindex mlittle-endian 11684Generate code for a little-endian target. This is the default. 11685 11686@item -msim 11687@opindex msim 11688Choose startup files and linker script suitable for the simulator. 11689 11690@item -msdata=default 11691@opindex msdata=default 11692Put small global and static data in the @samp{.neardata} section, 11693which is pointed to by register @code{B14}. Put small uninitialized 11694global and static data in the @samp{.bss} section, which is adjacent 11695to the @samp{.neardata} section. Put small read-only data into the 11696@samp{.rodata} section. The corresponding sections used for large 11697pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}. 11698 11699@item -msdata=all 11700@opindex msdata=all 11701Put all data, not just small objets, into the sections reserved for 11702small data, and use addressing relative to the @code{B14} register to 11703access them. 11704 11705@item -msdata=none 11706@opindex msdata=none 11707Make no use of the sections reserved for small data, and use absolute 11708addresses to access all data. Put all initialized global and static 11709data in the @samp{.fardata} section, and all uninitialized data in the 11710@samp{.far} section. Put all constant data into the @samp{.const} 11711section. 11712@end table 11713 11714@node CRIS Options 11715@subsection CRIS Options 11716@cindex CRIS Options 11717 11718These options are defined specifically for the CRIS ports. 11719 11720@table @gcctabopt 11721@item -march=@var{architecture-type} 11722@itemx -mcpu=@var{architecture-type} 11723@opindex march 11724@opindex mcpu 11725Generate code for the specified architecture. The choices for 11726@var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for 11727respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@. 11728Default is @samp{v0} except for cris-axis-linux-gnu, where the default is 11729@samp{v10}. 11730 11731@item -mtune=@var{architecture-type} 11732@opindex mtune 11733Tune to @var{architecture-type} everything applicable about the generated 11734code, except for the ABI and the set of available instructions. The 11735choices for @var{architecture-type} are the same as for 11736@option{-march=@var{architecture-type}}. 11737 11738@item -mmax-stack-frame=@var{n} 11739@opindex mmax-stack-frame 11740Warn when the stack frame of a function exceeds @var{n} bytes. 11741 11742@item -metrax4 11743@itemx -metrax100 11744@opindex metrax4 11745@opindex metrax100 11746The options @option{-metrax4} and @option{-metrax100} are synonyms for 11747@option{-march=v3} and @option{-march=v8} respectively. 11748 11749@item -mmul-bug-workaround 11750@itemx -mno-mul-bug-workaround 11751@opindex mmul-bug-workaround 11752@opindex mno-mul-bug-workaround 11753Work around a bug in the @code{muls} and @code{mulu} instructions for CPU 11754models where it applies. This option is active by default. 11755 11756@item -mpdebug 11757@opindex mpdebug 11758Enable CRIS-specific verbose debug-related information in the assembly 11759code. This option also has the effect to turn off the @samp{#NO_APP} 11760formatted-code indicator to the assembler at the beginning of the 11761assembly file. 11762 11763@item -mcc-init 11764@opindex mcc-init 11765Do not use condition-code results from previous instruction; always emit 11766compare and test instructions before use of condition codes. 11767 11768@item -mno-side-effects 11769@opindex mno-side-effects 11770Do not emit instructions with side-effects in addressing modes other than 11771post-increment. 11772 11773@item -mstack-align 11774@itemx -mno-stack-align 11775@itemx -mdata-align 11776@itemx -mno-data-align 11777@itemx -mconst-align 11778@itemx -mno-const-align 11779@opindex mstack-align 11780@opindex mno-stack-align 11781@opindex mdata-align 11782@opindex mno-data-align 11783@opindex mconst-align 11784@opindex mno-const-align 11785These options (no-options) arranges (eliminate arrangements) for the 11786stack-frame, individual data and constants to be aligned for the maximum 11787single data access size for the chosen CPU model. The default is to 11788arrange for 32-bit alignment. ABI details such as structure layout are 11789not affected by these options. 11790 11791@item -m32-bit 11792@itemx -m16-bit 11793@itemx -m8-bit 11794@opindex m32-bit 11795@opindex m16-bit 11796@opindex m8-bit 11797Similar to the stack- data- and const-align options above, these options 11798arrange for stack-frame, writable data and constants to all be 32-bit, 1179916-bit or 8-bit aligned. The default is 32-bit alignment. 11800 11801@item -mno-prologue-epilogue 11802@itemx -mprologue-epilogue 11803@opindex mno-prologue-epilogue 11804@opindex mprologue-epilogue 11805With @option{-mno-prologue-epilogue}, the normal function prologue and 11806epilogue which set up the stack frame are omitted and no return 11807instructions or return sequences are generated in the code. Use this 11808option only together with visual inspection of the compiled code: no 11809warnings or errors are generated when call-saved registers must be saved, 11810or storage for local variable needs to be allocated. 11811 11812@item -mno-gotplt 11813@itemx -mgotplt 11814@opindex mno-gotplt 11815@opindex mgotplt 11816With @option{-fpic} and @option{-fPIC}, don't generate (do generate) 11817instruction sequences that load addresses for functions from the PLT part 11818of the GOT rather than (traditional on other architectures) calls to the 11819PLT@. The default is @option{-mgotplt}. 11820 11821@item -melf 11822@opindex melf 11823Legacy no-op option only recognized with the cris-axis-elf and 11824cris-axis-linux-gnu targets. 11825 11826@item -mlinux 11827@opindex mlinux 11828Legacy no-op option only recognized with the cris-axis-linux-gnu target. 11829 11830@item -sim 11831@opindex sim 11832This option, recognized for the cris-axis-elf arranges 11833to link with input-output functions from a simulator library. Code, 11834initialized data and zero-initialized data are allocated consecutively. 11835 11836@item -sim2 11837@opindex sim2 11838Like @option{-sim}, but pass linker options to locate initialized data at 118390x40000000 and zero-initialized data at 0x80000000. 11840@end table 11841 11842@node CR16 Options 11843@subsection CR16 Options 11844@cindex CR16 Options 11845 11846These options are defined specifically for the CR16 ports. 11847 11848@table @gcctabopt 11849 11850@item -mmac 11851@opindex mmac 11852Enable the use of multiply-accumulate instructions. Disabled by default. 11853 11854@item -mcr16cplus 11855@itemx -mcr16c 11856@opindex mcr16cplus 11857@opindex mcr16c 11858Generate code for CR16C or CR16C+ architecture. CR16C+ architecture 11859is default. 11860 11861@item -msim 11862@opindex msim 11863Links the library libsim.a which is in compatible with simulator. Applicable 11864to elf compiler only. 11865 11866@item -mint32 11867@opindex mint32 11868Choose integer type as 32-bit wide. 11869 11870@item -mbit-ops 11871@opindex mbit-ops 11872Generates sbit/cbit instructions for bit manipulations. 11873 11874@item -mdata-model=@var{model} 11875@opindex mdata-model 11876Choose a data model. The choices for @var{model} are @samp{near}, 11877@samp{far} or @samp{medium}. @samp{medium} is default. 11878However, @samp{far} is not valid when -mcr16c option is chosen as 11879CR16C architecture does not support far data model. 11880@end table 11881 11882@node Darwin Options 11883@subsection Darwin Options 11884@cindex Darwin options 11885 11886These options are defined for all architectures running the Darwin operating 11887system. 11888 11889FSF GCC on Darwin does not create ``fat'' object files; it will create 11890an object file for the single architecture that it was built to 11891target. Apple's GCC on Darwin does create ``fat'' files if multiple 11892@option{-arch} options are used; it does so by running the compiler or 11893linker multiple times and joining the results together with 11894@file{lipo}. 11895 11896The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or 11897@samp{i686}) is determined by the flags that specify the ISA 11898that GCC is targetting, like @option{-mcpu} or @option{-march}. The 11899@option{-force_cpusubtype_ALL} option can be used to override this. 11900 11901The Darwin tools vary in their behavior when presented with an ISA 11902mismatch. The assembler, @file{as}, will only permit instructions to 11903be used that are valid for the subtype of the file it is generating, 11904so you cannot put 64-bit instructions in a @samp{ppc750} object file. 11905The linker for shared libraries, @file{/usr/bin/libtool}, will fail 11906and print an error if asked to create a shared library with a less 11907restrictive subtype than its input files (for instance, trying to put 11908a @samp{ppc970} object file in a @samp{ppc7400} library). The linker 11909for executables, @file{ld}, will quietly give the executable the most 11910restrictive subtype of any of its input files. 11911 11912@table @gcctabopt 11913@item -F@var{dir} 11914@opindex F 11915Add the framework directory @var{dir} to the head of the list of 11916directories to be searched for header files. These directories are 11917interleaved with those specified by @option{-I} options and are 11918scanned in a left-to-right order. 11919 11920A framework directory is a directory with frameworks in it. A 11921framework is a directory with a @samp{"Headers"} and/or 11922@samp{"PrivateHeaders"} directory contained directly in it that ends 11923in @samp{".framework"}. The name of a framework is the name of this 11924directory excluding the @samp{".framework"}. Headers associated with 11925the framework are found in one of those two directories, with 11926@samp{"Headers"} being searched first. A subframework is a framework 11927directory that is in a framework's @samp{"Frameworks"} directory. 11928Includes of subframework headers can only appear in a header of a 11929framework that contains the subframework, or in a sibling subframework 11930header. Two subframeworks are siblings if they occur in the same 11931framework. A subframework should not have the same name as a 11932framework, a warning will be issued if this is violated. Currently a 11933subframework cannot have subframeworks, in the future, the mechanism 11934may be extended to support this. The standard frameworks can be found 11935in @samp{"/System/Library/Frameworks"} and 11936@samp{"/Library/Frameworks"}. An example include looks like 11937@code{#include <Framework/header.h>}, where @samp{Framework} denotes 11938the name of the framework and header.h is found in the 11939@samp{"PrivateHeaders"} or @samp{"Headers"} directory. 11940 11941@item -iframework@var{dir} 11942@opindex iframework 11943Like @option{-F} except the directory is a treated as a system 11944directory. The main difference between this @option{-iframework} and 11945@option{-F} is that with @option{-iframework} the compiler does not 11946warn about constructs contained within header files found via 11947@var{dir}. This option is valid only for the C family of languages. 11948 11949@item -gused 11950@opindex gused 11951Emit debugging information for symbols that are used. For STABS 11952debugging format, this enables @option{-feliminate-unused-debug-symbols}. 11953This is by default ON@. 11954 11955@item -gfull 11956@opindex gfull 11957Emit debugging information for all symbols and types. 11958 11959@item -mmacosx-version-min=@var{version} 11960The earliest version of MacOS X that this executable will run on 11961is @var{version}. Typical values of @var{version} include @code{10.1}, 11962@code{10.2}, and @code{10.3.9}. 11963 11964If the compiler was built to use the system's headers by default, 11965then the default for this option is the system version on which the 11966compiler is running, otherwise the default is to make choices that 11967are compatible with as many systems and code bases as possible. 11968 11969@item -mkernel 11970@opindex mkernel 11971Enable kernel development mode. The @option{-mkernel} option sets 11972@option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit}, 11973@option{-fno-exceptions}, @option{-fno-non-call-exceptions}, 11974@option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where 11975applicable. This mode also sets @option{-mno-altivec}, 11976@option{-msoft-float}, @option{-fno-builtin} and 11977@option{-mlong-branch} for PowerPC targets. 11978 11979@item -mone-byte-bool 11980@opindex mone-byte-bool 11981Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}. 11982By default @samp{sizeof(bool)} is @samp{4} when compiling for 11983Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this 11984option has no effect on x86. 11985 11986@strong{Warning:} The @option{-mone-byte-bool} switch causes GCC 11987to generate code that is not binary compatible with code generated 11988without that switch. Using this switch may require recompiling all 11989other modules in a program, including system libraries. Use this 11990switch to conform to a non-default data model. 11991 11992@item -mfix-and-continue 11993@itemx -ffix-and-continue 11994@itemx -findirect-data 11995@opindex mfix-and-continue 11996@opindex ffix-and-continue 11997@opindex findirect-data 11998Generate code suitable for fast turn around development. Needed to 11999enable gdb to dynamically load @code{.o} files into already running 12000programs. @option{-findirect-data} and @option{-ffix-and-continue} 12001are provided for backwards compatibility. 12002 12003@item -all_load 12004@opindex all_load 12005Loads all members of static archive libraries. 12006See man ld(1) for more information. 12007 12008@item -arch_errors_fatal 12009@opindex arch_errors_fatal 12010Cause the errors having to do with files that have the wrong architecture 12011to be fatal. 12012 12013@item -bind_at_load 12014@opindex bind_at_load 12015Causes the output file to be marked such that the dynamic linker will 12016bind all undefined references when the file is loaded or launched. 12017 12018@item -bundle 12019@opindex bundle 12020Produce a Mach-o bundle format file. 12021See man ld(1) for more information. 12022 12023@item -bundle_loader @var{executable} 12024@opindex bundle_loader 12025This option specifies the @var{executable} that will be loading the build 12026output file being linked. See man ld(1) for more information. 12027 12028@item -dynamiclib 12029@opindex dynamiclib 12030When passed this option, GCC will produce a dynamic library instead of 12031an executable when linking, using the Darwin @file{libtool} command. 12032 12033@item -force_cpusubtype_ALL 12034@opindex force_cpusubtype_ALL 12035This causes GCC's output file to have the @var{ALL} subtype, instead of 12036one controlled by the @option{-mcpu} or @option{-march} option. 12037 12038@item -allowable_client @var{client_name} 12039@itemx -client_name 12040@itemx -compatibility_version 12041@itemx -current_version 12042@itemx -dead_strip 12043@itemx -dependency-file 12044@itemx -dylib_file 12045@itemx -dylinker_install_name 12046@itemx -dynamic 12047@itemx -exported_symbols_list 12048@itemx -filelist 12049@need 800 12050@itemx -flat_namespace 12051@itemx -force_flat_namespace 12052@itemx -headerpad_max_install_names 12053@itemx -image_base 12054@itemx -init 12055@itemx -install_name 12056@itemx -keep_private_externs 12057@itemx -multi_module 12058@itemx -multiply_defined 12059@itemx -multiply_defined_unused 12060@need 800 12061@itemx -noall_load 12062@itemx -no_dead_strip_inits_and_terms 12063@itemx -nofixprebinding 12064@itemx -nomultidefs 12065@itemx -noprebind 12066@itemx -noseglinkedit 12067@itemx -pagezero_size 12068@itemx -prebind 12069@itemx -prebind_all_twolevel_modules 12070@itemx -private_bundle 12071@need 800 12072@itemx -read_only_relocs 12073@itemx -sectalign 12074@itemx -sectobjectsymbols 12075@itemx -whyload 12076@itemx -seg1addr 12077@itemx -sectcreate 12078@itemx -sectobjectsymbols 12079@itemx -sectorder 12080@itemx -segaddr 12081@itemx -segs_read_only_addr 12082@need 800 12083@itemx -segs_read_write_addr 12084@itemx -seg_addr_table 12085@itemx -seg_addr_table_filename 12086@itemx -seglinkedit 12087@itemx -segprot 12088@itemx -segs_read_only_addr 12089@itemx -segs_read_write_addr 12090@itemx -single_module 12091@itemx -static 12092@itemx -sub_library 12093@need 800 12094@itemx -sub_umbrella 12095@itemx -twolevel_namespace 12096@itemx -umbrella 12097@itemx -undefined 12098@itemx -unexported_symbols_list 12099@itemx -weak_reference_mismatches 12100@itemx -whatsloaded 12101@opindex allowable_client 12102@opindex client_name 12103@opindex compatibility_version 12104@opindex current_version 12105@opindex dead_strip 12106@opindex dependency-file 12107@opindex dylib_file 12108@opindex dylinker_install_name 12109@opindex dynamic 12110@opindex exported_symbols_list 12111@opindex filelist 12112@opindex flat_namespace 12113@opindex force_flat_namespace 12114@opindex headerpad_max_install_names 12115@opindex image_base 12116@opindex init 12117@opindex install_name 12118@opindex keep_private_externs 12119@opindex multi_module 12120@opindex multiply_defined 12121@opindex multiply_defined_unused 12122@opindex noall_load 12123@opindex no_dead_strip_inits_and_terms 12124@opindex nofixprebinding 12125@opindex nomultidefs 12126@opindex noprebind 12127@opindex noseglinkedit 12128@opindex pagezero_size 12129@opindex prebind 12130@opindex prebind_all_twolevel_modules 12131@opindex private_bundle 12132@opindex read_only_relocs 12133@opindex sectalign 12134@opindex sectobjectsymbols 12135@opindex whyload 12136@opindex seg1addr 12137@opindex sectcreate 12138@opindex sectobjectsymbols 12139@opindex sectorder 12140@opindex segaddr 12141@opindex segs_read_only_addr 12142@opindex segs_read_write_addr 12143@opindex seg_addr_table 12144@opindex seg_addr_table_filename 12145@opindex seglinkedit 12146@opindex segprot 12147@opindex segs_read_only_addr 12148@opindex segs_read_write_addr 12149@opindex single_module 12150@opindex static 12151@opindex sub_library 12152@opindex sub_umbrella 12153@opindex twolevel_namespace 12154@opindex umbrella 12155@opindex undefined 12156@opindex unexported_symbols_list 12157@opindex weak_reference_mismatches 12158@opindex whatsloaded 12159These options are passed to the Darwin linker. The Darwin linker man page 12160describes them in detail. 12161@end table 12162 12163@node DEC Alpha Options 12164@subsection DEC Alpha Options 12165 12166These @samp{-m} options are defined for the DEC Alpha implementations: 12167 12168@table @gcctabopt 12169@item -mno-soft-float 12170@itemx -msoft-float 12171@opindex mno-soft-float 12172@opindex msoft-float 12173Use (do not use) the hardware floating-point instructions for 12174floating-point operations. When @option{-msoft-float} is specified, 12175functions in @file{libgcc.a} will be used to perform floating-point 12176operations. Unless they are replaced by routines that emulate the 12177floating-point operations, or compiled in such a way as to call such 12178emulations routines, these routines will issue floating-point 12179operations. If you are compiling for an Alpha without floating-point 12180operations, you must ensure that the library is built so as not to call 12181them. 12182 12183Note that Alpha implementations without floating-point operations are 12184required to have floating-point registers. 12185 12186@item -mfp-reg 12187@itemx -mno-fp-regs 12188@opindex mfp-reg 12189@opindex mno-fp-regs 12190Generate code that uses (does not use) the floating-point register set. 12191@option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point 12192register set is not used, floating-point operands are passed in integer 12193registers as if they were integers and floating-point results are passed 12194in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence, 12195so any function with a floating-point argument or return value called by code 12196compiled with @option{-mno-fp-regs} must also be compiled with that 12197option. 12198 12199A typical use of this option is building a kernel that does not use, 12200and hence need not save and restore, any floating-point registers. 12201 12202@item -mieee 12203@opindex mieee 12204The Alpha architecture implements floating-point hardware optimized for 12205maximum performance. It is mostly compliant with the IEEE floating-point 12206standard. However, for full compliance, software assistance is 12207required. This option generates code fully IEEE-compliant code 12208@emph{except} that the @var{inexact-flag} is not maintained (see below). 12209If this option is turned on, the preprocessor macro @code{_IEEE_FP} is 12210defined during compilation. The resulting code is less efficient but is 12211able to correctly support denormalized numbers and exceptional IEEE 12212values such as not-a-number and plus/minus infinity. Other Alpha 12213compilers call this option @option{-ieee_with_no_inexact}. 12214 12215@item -mieee-with-inexact 12216@opindex mieee-with-inexact 12217This is like @option{-mieee} except the generated code also maintains 12218the IEEE @var{inexact-flag}. Turning on this option causes the 12219generated code to implement fully-compliant IEEE math. In addition to 12220@code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor 12221macro. On some Alpha implementations the resulting code may execute 12222significantly slower than the code generated by default. Since there is 12223very little code that depends on the @var{inexact-flag}, you should 12224normally not specify this option. Other Alpha compilers call this 12225option @option{-ieee_with_inexact}. 12226 12227@item -mfp-trap-mode=@var{trap-mode} 12228@opindex mfp-trap-mode 12229This option controls what floating-point related traps are enabled. 12230Other Alpha compilers call this option @option{-fptm @var{trap-mode}}. 12231The trap mode can be set to one of four values: 12232 12233@table @samp 12234@item n 12235This is the default (normal) setting. The only traps that are enabled 12236are the ones that cannot be disabled in software (e.g., division by zero 12237trap). 12238 12239@item u 12240In addition to the traps enabled by @samp{n}, underflow traps are enabled 12241as well. 12242 12243@item su 12244Like @samp{u}, but the instructions are marked to be safe for software 12245completion (see Alpha architecture manual for details). 12246 12247@item sui 12248Like @samp{su}, but inexact traps are enabled as well. 12249@end table 12250 12251@item -mfp-rounding-mode=@var{rounding-mode} 12252@opindex mfp-rounding-mode 12253Selects the IEEE rounding mode. Other Alpha compilers call this option 12254@option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one 12255of: 12256 12257@table @samp 12258@item n 12259Normal IEEE rounding mode. Floating-point numbers are rounded towards 12260the nearest machine number or towards the even machine number in case 12261of a tie. 12262 12263@item m 12264Round towards minus infinity. 12265 12266@item c 12267Chopped rounding mode. Floating-point numbers are rounded towards zero. 12268 12269@item d 12270Dynamic rounding mode. A field in the floating-point control register 12271(@var{fpcr}, see Alpha architecture reference manual) controls the 12272rounding mode in effect. The C library initializes this register for 12273rounding towards plus infinity. Thus, unless your program modifies the 12274@var{fpcr}, @samp{d} corresponds to round towards plus infinity. 12275@end table 12276 12277@item -mtrap-precision=@var{trap-precision} 12278@opindex mtrap-precision 12279In the Alpha architecture, floating-point traps are imprecise. This 12280means without software assistance it is impossible to recover from a 12281floating trap and program execution normally needs to be terminated. 12282GCC can generate code that can assist operating system trap handlers 12283in determining the exact location that caused a floating-point trap. 12284Depending on the requirements of an application, different levels of 12285precisions can be selected: 12286 12287@table @samp 12288@item p 12289Program precision. This option is the default and means a trap handler 12290can only identify which program caused a floating-point exception. 12291 12292@item f 12293Function precision. The trap handler can determine the function that 12294caused a floating-point exception. 12295 12296@item i 12297Instruction precision. The trap handler can determine the exact 12298instruction that caused a floating-point exception. 12299@end table 12300 12301Other Alpha compilers provide the equivalent options called 12302@option{-scope_safe} and @option{-resumption_safe}. 12303 12304@item -mieee-conformant 12305@opindex mieee-conformant 12306This option marks the generated code as IEEE conformant. You must not 12307use this option unless you also specify @option{-mtrap-precision=i} and either 12308@option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect 12309is to emit the line @samp{.eflag 48} in the function prologue of the 12310generated assembly file. Under DEC Unix, this has the effect that 12311IEEE-conformant math library routines will be linked in. 12312 12313@item -mbuild-constants 12314@opindex mbuild-constants 12315Normally GCC examines a 32- or 64-bit integer constant to 12316see if it can construct it from smaller constants in two or three 12317instructions. If it cannot, it will output the constant as a literal and 12318generate code to load it from the data segment at run time. 12319 12320Use this option to require GCC to construct @emph{all} integer constants 12321using code, even if it takes more instructions (the maximum is six). 12322 12323You would typically use this option to build a shared library dynamic 12324loader. Itself a shared library, it must relocate itself in memory 12325before it can find the variables and constants in its own data segment. 12326 12327@item -malpha-as 12328@itemx -mgas 12329@opindex malpha-as 12330@opindex mgas 12331Select whether to generate code to be assembled by the vendor-supplied 12332assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}. 12333 12334@item -mbwx 12335@itemx -mno-bwx 12336@itemx -mcix 12337@itemx -mno-cix 12338@itemx -mfix 12339@itemx -mno-fix 12340@itemx -mmax 12341@itemx -mno-max 12342@opindex mbwx 12343@opindex mno-bwx 12344@opindex mcix 12345@opindex mno-cix 12346@opindex mfix 12347@opindex mno-fix 12348@opindex mmax 12349@opindex mno-max 12350Indicate whether GCC should generate code to use the optional BWX, 12351CIX, FIX and MAX instruction sets. The default is to use the instruction 12352sets supported by the CPU type specified via @option{-mcpu=} option or that 12353of the CPU on which GCC was built if none was specified. 12354 12355@item -mfloat-vax 12356@itemx -mfloat-ieee 12357@opindex mfloat-vax 12358@opindex mfloat-ieee 12359Generate code that uses (does not use) VAX F and G floating-point 12360arithmetic instead of IEEE single and double precision. 12361 12362@item -mexplicit-relocs 12363@itemx -mno-explicit-relocs 12364@opindex mexplicit-relocs 12365@opindex mno-explicit-relocs 12366Older Alpha assemblers provided no way to generate symbol relocations 12367except via assembler macros. Use of these macros does not allow 12368optimal instruction scheduling. GNU binutils as of version 2.12 12369supports a new syntax that allows the compiler to explicitly mark 12370which relocations should apply to which instructions. This option 12371is mostly useful for debugging, as GCC detects the capabilities of 12372the assembler when it is built and sets the default accordingly. 12373 12374@item -msmall-data 12375@itemx -mlarge-data 12376@opindex msmall-data 12377@opindex mlarge-data 12378When @option{-mexplicit-relocs} is in effect, static data is 12379accessed via @dfn{gp-relative} relocations. When @option{-msmall-data} 12380is used, objects 8 bytes long or smaller are placed in a @dfn{small data area} 12381(the @code{.sdata} and @code{.sbss} sections) and are accessed via 1238216-bit relocations off of the @code{$gp} register. This limits the 12383size of the small data area to 64KB, but allows the variables to be 12384directly accessed via a single instruction. 12385 12386The default is @option{-mlarge-data}. With this option the data area 12387is limited to just below 2GB@. Programs that require more than 2GB of 12388data must use @code{malloc} or @code{mmap} to allocate the data in the 12389heap instead of in the program's data segment. 12390 12391When generating code for shared libraries, @option{-fpic} implies 12392@option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}. 12393 12394@item -msmall-text 12395@itemx -mlarge-text 12396@opindex msmall-text 12397@opindex mlarge-text 12398When @option{-msmall-text} is used, the compiler assumes that the 12399code of the entire program (or shared library) fits in 4MB, and is 12400thus reachable with a branch instruction. When @option{-msmall-data} 12401is used, the compiler can assume that all local symbols share the 12402same @code{$gp} value, and thus reduce the number of instructions 12403required for a function call from 4 to 1. 12404 12405The default is @option{-mlarge-text}. 12406 12407@item -mcpu=@var{cpu_type} 12408@opindex mcpu 12409Set the instruction set and instruction scheduling parameters for 12410machine type @var{cpu_type}. You can specify either the @samp{EV} 12411style name or the corresponding chip number. GCC supports scheduling 12412parameters for the EV4, EV5 and EV6 family of processors and will 12413choose the default values for the instruction set from the processor 12414you specify. If you do not specify a processor type, GCC will default 12415to the processor on which the compiler was built. 12416 12417Supported values for @var{cpu_type} are 12418 12419@table @samp 12420@item ev4 12421@itemx ev45 12422@itemx 21064 12423Schedules as an EV4 and has no instruction set extensions. 12424 12425@item ev5 12426@itemx 21164 12427Schedules as an EV5 and has no instruction set extensions. 12428 12429@item ev56 12430@itemx 21164a 12431Schedules as an EV5 and supports the BWX extension. 12432 12433@item pca56 12434@itemx 21164pc 12435@itemx 21164PC 12436Schedules as an EV5 and supports the BWX and MAX extensions. 12437 12438@item ev6 12439@itemx 21264 12440Schedules as an EV6 and supports the BWX, FIX, and MAX extensions. 12441 12442@item ev67 12443@itemx 21264a 12444Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions. 12445@end table 12446 12447Native toolchains also support the value @samp{native}, 12448which selects the best architecture option for the host processor. 12449@option{-mcpu=native} has no effect if GCC does not recognize 12450the processor. 12451 12452@item -mtune=@var{cpu_type} 12453@opindex mtune 12454Set only the instruction scheduling parameters for machine type 12455@var{cpu_type}. The instruction set is not changed. 12456 12457Native toolchains also support the value @samp{native}, 12458which selects the best architecture option for the host processor. 12459@option{-mtune=native} has no effect if GCC does not recognize 12460the processor. 12461 12462@item -mmemory-latency=@var{time} 12463@opindex mmemory-latency 12464Sets the latency the scheduler should assume for typical memory 12465references as seen by the application. This number is highly 12466dependent on the memory access patterns used by the application 12467and the size of the external cache on the machine. 12468 12469Valid options for @var{time} are 12470 12471@table @samp 12472@item @var{number} 12473A decimal number representing clock cycles. 12474 12475@item L1 12476@itemx L2 12477@itemx L3 12478@itemx main 12479The compiler contains estimates of the number of clock cycles for 12480``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches 12481(also called Dcache, Scache, and Bcache), as well as to main memory. 12482Note that L3 is only valid for EV5. 12483 12484@end table 12485@end table 12486 12487@node DEC Alpha/VMS Options 12488@subsection DEC Alpha/VMS Options 12489 12490These @samp{-m} options are defined for the DEC Alpha/VMS implementations: 12491 12492@table @gcctabopt 12493@item -mvms-return-codes 12494@opindex mvms-return-codes 12495Return VMS condition codes from main. The default is to return POSIX 12496style condition (e.g.@: error) codes. 12497 12498@item -mdebug-main=@var{prefix} 12499@opindex mdebug-main=@var{prefix} 12500Flag the first routine whose name starts with @var{prefix} as the main 12501routine for the debugger. 12502 12503@item -mmalloc64 12504@opindex mmalloc64 12505Default to 64-bit memory allocation routines. 12506@end table 12507 12508@node FR30 Options 12509@subsection FR30 Options 12510@cindex FR30 Options 12511 12512These options are defined specifically for the FR30 port. 12513 12514@table @gcctabopt 12515 12516@item -msmall-model 12517@opindex msmall-model 12518Use the small address space model. This can produce smaller code, but 12519it does assume that all symbolic values and addresses will fit into a 1252020-bit range. 12521 12522@item -mno-lsim 12523@opindex mno-lsim 12524Assume that runtime support has been provided and so there is no need 12525to include the simulator library (@file{libsim.a}) on the linker 12526command line. 12527 12528@end table 12529 12530@node FRV Options 12531@subsection FRV Options 12532@cindex FRV Options 12533 12534@table @gcctabopt 12535@item -mgpr-32 12536@opindex mgpr-32 12537 12538Only use the first 32 general-purpose registers. 12539 12540@item -mgpr-64 12541@opindex mgpr-64 12542 12543Use all 64 general-purpose registers. 12544 12545@item -mfpr-32 12546@opindex mfpr-32 12547 12548Use only the first 32 floating-point registers. 12549 12550@item -mfpr-64 12551@opindex mfpr-64 12552 12553Use all 64 floating-point registers. 12554 12555@item -mhard-float 12556@opindex mhard-float 12557 12558Use hardware instructions for floating-point operations. 12559 12560@item -msoft-float 12561@opindex msoft-float 12562 12563Use library routines for floating-point operations. 12564 12565@item -malloc-cc 12566@opindex malloc-cc 12567 12568Dynamically allocate condition code registers. 12569 12570@item -mfixed-cc 12571@opindex mfixed-cc 12572 12573Do not try to dynamically allocate condition code registers, only 12574use @code{icc0} and @code{fcc0}. 12575 12576@item -mdword 12577@opindex mdword 12578 12579Change ABI to use double word insns. 12580 12581@item -mno-dword 12582@opindex mno-dword 12583 12584Do not use double word instructions. 12585 12586@item -mdouble 12587@opindex mdouble 12588 12589Use floating-point double instructions. 12590 12591@item -mno-double 12592@opindex mno-double 12593 12594Do not use floating-point double instructions. 12595 12596@item -mmedia 12597@opindex mmedia 12598 12599Use media instructions. 12600 12601@item -mno-media 12602@opindex mno-media 12603 12604Do not use media instructions. 12605 12606@item -mmuladd 12607@opindex mmuladd 12608 12609Use multiply and add/subtract instructions. 12610 12611@item -mno-muladd 12612@opindex mno-muladd 12613 12614Do not use multiply and add/subtract instructions. 12615 12616@item -mfdpic 12617@opindex mfdpic 12618 12619Select the FDPIC ABI, which uses function descriptors to represent 12620pointers to functions. Without any PIC/PIE-related options, it 12621implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it 12622assumes GOT entries and small data are within a 12-bit range from the 12623GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets 12624are computed with 32 bits. 12625With a @samp{bfin-elf} target, this option implies @option{-msim}. 12626 12627@item -minline-plt 12628@opindex minline-plt 12629 12630Enable inlining of PLT entries in function calls to functions that are 12631not known to bind locally. It has no effect without @option{-mfdpic}. 12632It's enabled by default if optimizing for speed and compiling for 12633shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an 12634optimization option such as @option{-O3} or above is present in the 12635command line. 12636 12637@item -mTLS 12638@opindex mTLS 12639 12640Assume a large TLS segment when generating thread-local code. 12641 12642@item -mtls 12643@opindex mtls 12644 12645Do not assume a large TLS segment when generating thread-local code. 12646 12647@item -mgprel-ro 12648@opindex mgprel-ro 12649 12650Enable the use of @code{GPREL} relocations in the FDPIC ABI for data 12651that is known to be in read-only sections. It's enabled by default, 12652except for @option{-fpic} or @option{-fpie}: even though it may help 12653make the global offset table smaller, it trades 1 instruction for 4. 12654With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4, 12655one of which may be shared by multiple symbols, and it avoids the need 12656for a GOT entry for the referenced symbol, so it's more likely to be a 12657win. If it is not, @option{-mno-gprel-ro} can be used to disable it. 12658 12659@item -multilib-library-pic 12660@opindex multilib-library-pic 12661 12662Link with the (library, not FD) pic libraries. It's implied by 12663@option{-mlibrary-pic}, as well as by @option{-fPIC} and 12664@option{-fpic} without @option{-mfdpic}. You should never have to use 12665it explicitly. 12666 12667@item -mlinked-fp 12668@opindex mlinked-fp 12669 12670Follow the EABI requirement of always creating a frame pointer whenever 12671a stack frame is allocated. This option is enabled by default and can 12672be disabled with @option{-mno-linked-fp}. 12673 12674@item -mlong-calls 12675@opindex mlong-calls 12676 12677Use indirect addressing to call functions outside the current 12678compilation unit. This allows the functions to be placed anywhere 12679within the 32-bit address space. 12680 12681@item -malign-labels 12682@opindex malign-labels 12683 12684Try to align labels to an 8-byte boundary by inserting nops into the 12685previous packet. This option only has an effect when VLIW packing 12686is enabled. It doesn't create new packets; it merely adds nops to 12687existing ones. 12688 12689@item -mlibrary-pic 12690@opindex mlibrary-pic 12691 12692Generate position-independent EABI code. 12693 12694@item -macc-4 12695@opindex macc-4 12696 12697Use only the first four media accumulator registers. 12698 12699@item -macc-8 12700@opindex macc-8 12701 12702Use all eight media accumulator registers. 12703 12704@item -mpack 12705@opindex mpack 12706 12707Pack VLIW instructions. 12708 12709@item -mno-pack 12710@opindex mno-pack 12711 12712Do not pack VLIW instructions. 12713 12714@item -mno-eflags 12715@opindex mno-eflags 12716 12717Do not mark ABI switches in e_flags. 12718 12719@item -mcond-move 12720@opindex mcond-move 12721 12722Enable the use of conditional-move instructions (default). 12723 12724This switch is mainly for debugging the compiler and will likely be removed 12725in a future version. 12726 12727@item -mno-cond-move 12728@opindex mno-cond-move 12729 12730Disable the use of conditional-move instructions. 12731 12732This switch is mainly for debugging the compiler and will likely be removed 12733in a future version. 12734 12735@item -mscc 12736@opindex mscc 12737 12738Enable the use of conditional set instructions (default). 12739 12740This switch is mainly for debugging the compiler and will likely be removed 12741in a future version. 12742 12743@item -mno-scc 12744@opindex mno-scc 12745 12746Disable the use of conditional set instructions. 12747 12748This switch is mainly for debugging the compiler and will likely be removed 12749in a future version. 12750 12751@item -mcond-exec 12752@opindex mcond-exec 12753 12754Enable the use of conditional execution (default). 12755 12756This switch is mainly for debugging the compiler and will likely be removed 12757in a future version. 12758 12759@item -mno-cond-exec 12760@opindex mno-cond-exec 12761 12762Disable the use of conditional execution. 12763 12764This switch is mainly for debugging the compiler and will likely be removed 12765in a future version. 12766 12767@item -mvliw-branch 12768@opindex mvliw-branch 12769 12770Run a pass to pack branches into VLIW instructions (default). 12771 12772This switch is mainly for debugging the compiler and will likely be removed 12773in a future version. 12774 12775@item -mno-vliw-branch 12776@opindex mno-vliw-branch 12777 12778Do not run a pass to pack branches into VLIW instructions. 12779 12780This switch is mainly for debugging the compiler and will likely be removed 12781in a future version. 12782 12783@item -mmulti-cond-exec 12784@opindex mmulti-cond-exec 12785 12786Enable optimization of @code{&&} and @code{||} in conditional execution 12787(default). 12788 12789This switch is mainly for debugging the compiler and will likely be removed 12790in a future version. 12791 12792@item -mno-multi-cond-exec 12793@opindex mno-multi-cond-exec 12794 12795Disable optimization of @code{&&} and @code{||} in conditional execution. 12796 12797This switch is mainly for debugging the compiler and will likely be removed 12798in a future version. 12799 12800@item -mnested-cond-exec 12801@opindex mnested-cond-exec 12802 12803Enable nested conditional execution optimizations (default). 12804 12805This switch is mainly for debugging the compiler and will likely be removed 12806in a future version. 12807 12808@item -mno-nested-cond-exec 12809@opindex mno-nested-cond-exec 12810 12811Disable nested conditional execution optimizations. 12812 12813This switch is mainly for debugging the compiler and will likely be removed 12814in a future version. 12815 12816@item -moptimize-membar 12817@opindex moptimize-membar 12818 12819This switch removes redundant @code{membar} instructions from the 12820compiler generated code. It is enabled by default. 12821 12822@item -mno-optimize-membar 12823@opindex mno-optimize-membar 12824 12825This switch disables the automatic removal of redundant @code{membar} 12826instructions from the generated code. 12827 12828@item -mtomcat-stats 12829@opindex mtomcat-stats 12830 12831Cause gas to print out tomcat statistics. 12832 12833@item -mcpu=@var{cpu} 12834@opindex mcpu 12835 12836Select the processor type for which to generate code. Possible values are 12837@samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450}, 12838@samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}. 12839 12840@end table 12841 12842@node GNU/Linux Options 12843@subsection GNU/Linux Options 12844 12845These @samp{-m} options are defined for GNU/Linux targets: 12846 12847@table @gcctabopt 12848@item -mglibc 12849@opindex mglibc 12850Use the GNU C library. This is the default except 12851on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets. 12852 12853@item -muclibc 12854@opindex muclibc 12855Use uClibc C library. This is the default on 12856@samp{*-*-linux-*uclibc*} targets. 12857 12858@item -mbionic 12859@opindex mbionic 12860Use Bionic C library. This is the default on 12861@samp{*-*-linux-*android*} targets. 12862 12863@item -mandroid 12864@opindex mandroid 12865Compile code compatible with Android platform. This is the default on 12866@samp{*-*-linux-*android*} targets. 12867 12868When compiling, this option enables @option{-mbionic}, @option{-fPIC}, 12869@option{-fno-exceptions} and @option{-fno-rtti} by default. When linking, 12870this option makes the GCC driver pass Android-specific options to the linker. 12871Finally, this option causes the preprocessor macro @code{__ANDROID__} 12872to be defined. 12873 12874@item -tno-android-cc 12875@opindex tno-android-cc 12876Disable compilation effects of @option{-mandroid}, i.e., do not enable 12877@option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and 12878@option{-fno-rtti} by default. 12879 12880@item -tno-android-ld 12881@opindex tno-android-ld 12882Disable linking effects of @option{-mandroid}, i.e., pass standard Linux 12883linking options to the linker. 12884 12885@end table 12886 12887@node H8/300 Options 12888@subsection H8/300 Options 12889 12890These @samp{-m} options are defined for the H8/300 implementations: 12891 12892@table @gcctabopt 12893@item -mrelax 12894@opindex mrelax 12895Shorten some address references at link time, when possible; uses the 12896linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300, 12897ld, Using ld}, for a fuller description. 12898 12899@item -mh 12900@opindex mh 12901Generate code for the H8/300H@. 12902 12903@item -ms 12904@opindex ms 12905Generate code for the H8S@. 12906 12907@item -mn 12908@opindex mn 12909Generate code for the H8S and H8/300H in the normal mode. This switch 12910must be used either with @option{-mh} or @option{-ms}. 12911 12912@item -ms2600 12913@opindex ms2600 12914Generate code for the H8S/2600. This switch must be used with @option{-ms}. 12915 12916@item -mint32 12917@opindex mint32 12918Make @code{int} data 32 bits by default. 12919 12920@item -malign-300 12921@opindex malign-300 12922On the H8/300H and H8S, use the same alignment rules as for the H8/300. 12923The default for the H8/300H and H8S is to align longs and floats on 129244-byte boundaries. 12925@option{-malign-300} causes them to be aligned on 2-byte boundaries. 12926This option has no effect on the H8/300. 12927@end table 12928 12929@node HPPA Options 12930@subsection HPPA Options 12931@cindex HPPA Options 12932 12933These @samp{-m} options are defined for the HPPA family of computers: 12934 12935@table @gcctabopt 12936@item -march=@var{architecture-type} 12937@opindex march 12938Generate code for the specified architecture. The choices for 12939@var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA 129401.1, and @samp{2.0} for PA 2.0 processors. Refer to 12941@file{/usr/lib/sched.models} on an HP-UX system to determine the proper 12942architecture option for your machine. Code compiled for lower numbered 12943architectures will run on higher numbered architectures, but not the 12944other way around. 12945 12946@item -mpa-risc-1-0 12947@itemx -mpa-risc-1-1 12948@itemx -mpa-risc-2-0 12949@opindex mpa-risc-1-0 12950@opindex mpa-risc-1-1 12951@opindex mpa-risc-2-0 12952Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively. 12953 12954@item -mbig-switch 12955@opindex mbig-switch 12956Generate code suitable for big switch tables. Use this option only if 12957the assembler/linker complain about out of range branches within a switch 12958table. 12959 12960@item -mjump-in-delay 12961@opindex mjump-in-delay 12962Fill delay slots of function calls with unconditional jump instructions 12963by modifying the return pointer for the function call to be the target 12964of the conditional jump. 12965 12966@item -mdisable-fpregs 12967@opindex mdisable-fpregs 12968Prevent floating-point registers from being used in any manner. This is 12969necessary for compiling kernels that perform lazy context switching of 12970floating-point registers. If you use this option and attempt to perform 12971floating-point operations, the compiler aborts. 12972 12973@item -mdisable-indexing 12974@opindex mdisable-indexing 12975Prevent the compiler from using indexing address modes. This avoids some 12976rather obscure problems when compiling MIG generated code under MACH@. 12977 12978@item -mno-space-regs 12979@opindex mno-space-regs 12980Generate code that assumes the target has no space registers. This allows 12981GCC to generate faster indirect calls and use unscaled index address modes. 12982 12983Such code is suitable for level 0 PA systems and kernels. 12984 12985@item -mfast-indirect-calls 12986@opindex mfast-indirect-calls 12987Generate code that assumes calls never cross space boundaries. This 12988allows GCC to emit code that performs faster indirect calls. 12989 12990This option will not work in the presence of shared libraries or nested 12991functions. 12992 12993@item -mfixed-range=@var{register-range} 12994@opindex mfixed-range 12995Generate code treating the given register range as fixed registers. 12996A fixed register is one that the register allocator can not use. This is 12997useful when compiling kernel code. A register range is specified as 12998two registers separated by a dash. Multiple register ranges can be 12999specified separated by a comma. 13000 13001@item -mlong-load-store 13002@opindex mlong-load-store 13003Generate 3-instruction load and store sequences as sometimes required by 13004the HP-UX 10 linker. This is equivalent to the @samp{+k} option to 13005the HP compilers. 13006 13007@item -mportable-runtime 13008@opindex mportable-runtime 13009Use the portable calling conventions proposed by HP for ELF systems. 13010 13011@item -mgas 13012@opindex mgas 13013Enable the use of assembler directives only GAS understands. 13014 13015@item -mschedule=@var{cpu-type} 13016@opindex mschedule 13017Schedule code according to the constraints for the machine type 13018@var{cpu-type}. The choices for @var{cpu-type} are @samp{700} 13019@samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer 13020to @file{/usr/lib/sched.models} on an HP-UX system to determine the 13021proper scheduling option for your machine. The default scheduling is 13022@samp{8000}. 13023 13024@item -mlinker-opt 13025@opindex mlinker-opt 13026Enable the optimization pass in the HP-UX linker. Note this makes symbolic 13027debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9 13028linkers in which they give bogus error messages when linking some programs. 13029 13030@item -msoft-float 13031@opindex msoft-float 13032Generate output containing library calls for floating point. 13033@strong{Warning:} the requisite libraries are not available for all HPPA 13034targets. Normally the facilities of the machine's usual C compiler are 13035used, but this cannot be done directly in cross-compilation. You must make 13036your own arrangements to provide suitable library functions for 13037cross-compilation. 13038 13039@option{-msoft-float} changes the calling convention in the output file; 13040therefore, it is only useful if you compile @emph{all} of a program with 13041this option. In particular, you need to compile @file{libgcc.a}, the 13042library that comes with GCC, with @option{-msoft-float} in order for 13043this to work. 13044 13045@item -msio 13046@opindex msio 13047Generate the predefine, @code{_SIO}, for server IO@. The default is 13048@option{-mwsio}. This generates the predefines, @code{__hp9000s700}, 13049@code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These 13050options are available under HP-UX and HI-UX@. 13051 13052@item -mgnu-ld 13053@opindex mgnu-ld 13054Use GNU ld specific options. This passes @option{-shared} to ld when 13055building a shared library. It is the default when GCC is configured, 13056explicitly or implicitly, with the GNU linker. This option does not 13057have any affect on which ld is called, it only changes what parameters 13058are passed to that ld. The ld that is called is determined by the 13059@option{--with-ld} configure option, GCC's program search path, and 13060finally by the user's @env{PATH}. The linker used by GCC can be printed 13061using @samp{which `gcc -print-prog-name=ld`}. This option is only available 13062on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}. 13063 13064@item -mhp-ld 13065@opindex mhp-ld 13066Use HP ld specific options. This passes @option{-b} to ld when building 13067a shared library and passes @option{+Accept TypeMismatch} to ld on all 13068links. It is the default when GCC is configured, explicitly or 13069implicitly, with the HP linker. This option does not have any affect on 13070which ld is called, it only changes what parameters are passed to that 13071ld. The ld that is called is determined by the @option{--with-ld} 13072configure option, GCC's program search path, and finally by the user's 13073@env{PATH}. The linker used by GCC can be printed using @samp{which 13074`gcc -print-prog-name=ld`}. This option is only available on the 64-bit 13075HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}. 13076 13077@item -mlong-calls 13078@opindex mno-long-calls 13079Generate code that uses long call sequences. This ensures that a call 13080is always able to reach linker generated stubs. The default is to generate 13081long calls only when the distance from the call site to the beginning 13082of the function or translation unit, as the case may be, exceeds a 13083predefined limit set by the branch type being used. The limits for 13084normal calls are 7,600,000 and 240,000 bytes, respectively for the 13085PA 2.0 and PA 1.X architectures. Sibcalls are always limited at 13086240,000 bytes. 13087 13088Distances are measured from the beginning of functions when using the 13089@option{-ffunction-sections} option, or when using the @option{-mgas} 13090and @option{-mno-portable-runtime} options together under HP-UX with 13091the SOM linker. 13092 13093It is normally not desirable to use this option as it will degrade 13094performance. However, it may be useful in large applications, 13095particularly when partial linking is used to build the application. 13096 13097The types of long calls used depends on the capabilities of the 13098assembler and linker, and the type of code being generated. The 13099impact on systems that support long absolute calls, and long pic 13100symbol-difference or pc-relative calls should be relatively small. 13101However, an indirect call is used on 32-bit ELF systems in pic code 13102and it is quite long. 13103 13104@item -munix=@var{unix-std} 13105@opindex march 13106Generate compiler predefines and select a startfile for the specified 13107UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95} 13108and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95} 13109is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX 1311011.11 and later. The default values are @samp{93} for HP-UX 10.00, 13111@samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11 13112and later. 13113 13114@option{-munix=93} provides the same predefines as GCC 3.3 and 3.4. 13115@option{-munix=95} provides additional predefines for @code{XOPEN_UNIX} 13116and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}. 13117@option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX}, 13118@code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and 13119@code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}. 13120 13121It is @emph{important} to note that this option changes the interfaces 13122for various library routines. It also affects the operational behavior 13123of the C library. Thus, @emph{extreme} care is needed in using this 13124option. 13125 13126Library code that is intended to operate with more than one UNIX 13127standard must test, set and restore the variable @var{__xpg4_extended_mask} 13128as appropriate. Most GNU software doesn't provide this capability. 13129 13130@item -nolibdld 13131@opindex nolibdld 13132Suppress the generation of link options to search libdld.sl when the 13133@option{-static} option is specified on HP-UX 10 and later. 13134 13135@item -static 13136@opindex static 13137The HP-UX implementation of setlocale in libc has a dependency on 13138libdld.sl. There isn't an archive version of libdld.sl. Thus, 13139when the @option{-static} option is specified, special link options 13140are needed to resolve this dependency. 13141 13142On HP-UX 10 and later, the GCC driver adds the necessary options to 13143link with libdld.sl when the @option{-static} option is specified. 13144This causes the resulting binary to be dynamic. On the 64-bit port, 13145the linkers generate dynamic binaries by default in any case. The 13146@option{-nolibdld} option can be used to prevent the GCC driver from 13147adding these link options. 13148 13149@item -threads 13150@opindex threads 13151Add support for multithreading with the @dfn{dce thread} library 13152under HP-UX@. This option sets flags for both the preprocessor and 13153linker. 13154@end table 13155 13156@node i386 and x86-64 Options 13157@subsection Intel 386 and AMD x86-64 Options 13158@cindex i386 Options 13159@cindex x86-64 Options 13160@cindex Intel 386 Options 13161@cindex AMD x86-64 Options 13162 13163These @samp{-m} options are defined for the i386 and x86-64 family of 13164computers: 13165 13166@table @gcctabopt 13167@item -mtune=@var{cpu-type} 13168@opindex mtune 13169Tune to @var{cpu-type} everything applicable about the generated code, except 13170for the ABI and the set of available instructions. The choices for 13171@var{cpu-type} are: 13172@table @emph 13173@item generic 13174Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors. 13175If you know the CPU on which your code will run, then you should use 13176the corresponding @option{-mtune} option instead of 13177@option{-mtune=generic}. But, if you do not know exactly what CPU users 13178of your application will have, then you should use this option. 13179 13180As new processors are deployed in the marketplace, the behavior of this 13181option will change. Therefore, if you upgrade to a newer version of 13182GCC, the code generated option will change to reflect the processors 13183that were most common when that version of GCC was released. 13184 13185There is no @option{-march=generic} option because @option{-march} 13186indicates the instruction set the compiler can use, and there is no 13187generic instruction set applicable to all processors. In contrast, 13188@option{-mtune} indicates the processor (or, in this case, collection of 13189processors) for which the code is optimized. 13190@item native 13191This selects the CPU to tune for at compilation time by determining 13192the processor type of the compiling machine. Using @option{-mtune=native} 13193will produce code optimized for the local machine under the constraints 13194of the selected instruction set. Using @option{-march=native} will 13195enable all instruction subsets supported by the local machine (hence 13196the result might not run on different machines). 13197@item i386 13198Original Intel's i386 CPU@. 13199@item i486 13200Intel's i486 CPU@. (No scheduling is implemented for this chip.) 13201@item i586, pentium 13202Intel Pentium CPU with no MMX support. 13203@item pentium-mmx 13204Intel PentiumMMX CPU based on Pentium core with MMX instruction set support. 13205@item pentiumpro 13206Intel PentiumPro CPU@. 13207@item i686 13208Same as @code{generic}, but when used as @code{march} option, PentiumPro 13209instruction set will be used, so the code will run on all i686 family chips. 13210@item pentium2 13211Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support. 13212@item pentium3, pentium3m 13213Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set 13214support. 13215@item pentium-m 13216Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set 13217support. Used by Centrino notebooks. 13218@item pentium4, pentium4m 13219Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support. 13220@item prescott 13221Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction 13222set support. 13223@item nocona 13224Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE, 13225SSE2 and SSE3 instruction set support. 13226@item core2 13227Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3 13228instruction set support. 13229@item corei7 13230Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1 13231and SSE4.2 instruction set support. 13232@item corei7-avx 13233Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, 13234SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support. 13235@item core-avx-i 13236Intel Core CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, 13237SSE4.1, SSE4.2, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C instruction 13238set support. 13239@item atom 13240Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3 13241instruction set support. 13242@item k6 13243AMD K6 CPU with MMX instruction set support. 13244@item k6-2, k6-3 13245Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support. 13246@item athlon, athlon-tbird 13247AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions 13248support. 13249@item athlon-4, athlon-xp, athlon-mp 13250Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE 13251instruction set support. 13252@item k8, opteron, athlon64, athlon-fx 13253AMD K8 core based CPUs with x86-64 instruction set support. (This supersets 13254MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.) 13255@item k8-sse3, opteron-sse3, athlon64-sse3 13256Improved versions of k8, opteron and athlon64 with SSE3 instruction set support. 13257@item amdfam10, barcelona 13258AMD Family 10h core based CPUs with x86-64 instruction set support. (This 13259supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit 13260instruction set extensions.) 13261@item bdver1 13262AMD Family 15h core based CPUs with x86-64 instruction set support. (This 13263supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, 13264SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.) 13265@item bdver2 13266AMD Family 15h core based CPUs with x86-64 instruction set support. (This 13267supersets BMI, TBM, F16C, FMA, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, 13268SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set 13269extensions.) 13270@item btver1 13271AMD Family 14h core based CPUs with x86-64 instruction set support. (This 13272supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit 13273instruction set extensions.) 13274@item winchip-c6 13275IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction 13276set support. 13277@item winchip2 13278IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@: 13279instruction set support. 13280@item c3 13281Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is 13282implemented for this chip.) 13283@item c3-2 13284Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is 13285implemented for this chip.) 13286@item geode 13287Embedded AMD CPU with MMX and 3DNow!@: instruction set support. 13288@end table 13289 13290While picking a specific @var{cpu-type} will schedule things appropriately 13291for that particular chip, the compiler will not generate any code that 13292does not run on the default machine type without the @option{-march=@var{cpu-type}} 13293option being used. For example, if GCC is configured for i686-pc-linux-gnu 13294then @option{-mtune=pentium4} will generate code that is tuned for Pentium4 13295but will still run on i686 machines. 13296 13297@item -march=@var{cpu-type} 13298@opindex march 13299Generate instructions for the machine type @var{cpu-type}. The choices 13300for @var{cpu-type} are the same as for @option{-mtune}. Moreover, 13301specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}. 13302 13303@item -mcpu=@var{cpu-type} 13304@opindex mcpu 13305A deprecated synonym for @option{-mtune}. 13306 13307@item -mfpmath=@var{unit} 13308@opindex mfpmath 13309Generate floating-point arithmetic for selected unit @var{unit}. The choices 13310for @var{unit} are: 13311 13312@table @samp 13313@item 387 13314Use the standard 387 floating-point coprocessor present on the majority of chips and 13315emulated otherwise. Code compiled with this option runs almost everywhere. 13316The temporary results are computed in 80-bit precision instead of the precision 13317specified by the type, resulting in slightly different results compared to most 13318of other chips. See @option{-ffloat-store} for more detailed description. 13319 13320This is the default choice for i386 compiler. 13321 13322@item sse 13323Use scalar floating-point instructions present in the SSE instruction set. 13324This instruction set is supported by Pentium3 and newer chips, in the AMD line 13325by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE 13326instruction set supports only single-precision arithmetic, thus the double and 13327extended-precision arithmetic are still done using 387. A later version, present 13328only in Pentium4 and the future AMD x86-64 chips, supports double-precision 13329arithmetic too. 13330 13331For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse} 13332or @option{-msse2} switches to enable SSE extensions and make this option 13333effective. For the x86-64 compiler, these extensions are enabled by default. 13334 13335The resulting code should be considerably faster in the majority of cases and avoid 13336the numerical instability problems of 387 code, but may break some existing 13337code that expects temporaries to be 80 bits. 13338 13339This is the default choice for the x86-64 compiler. 13340 13341@item sse,387 13342@itemx sse+387 13343@itemx both 13344Attempt to utilize both instruction sets at once. This effectively double the 13345amount of available registers and on chips with separate execution units for 13346387 and SSE the execution resources too. Use this option with care, as it is 13347still experimental, because the GCC register allocator does not model separate 13348functional units well resulting in instable performance. 13349@end table 13350 13351@item -masm=@var{dialect} 13352@opindex masm=@var{dialect} 13353Output asm instructions using selected @var{dialect}. Supported 13354choices are @samp{intel} or @samp{att} (the default one). Darwin does 13355not support @samp{intel}. 13356 13357@item -mieee-fp 13358@itemx -mno-ieee-fp 13359@opindex mieee-fp 13360@opindex mno-ieee-fp 13361Control whether or not the compiler uses IEEE floating-point 13362comparisons. These handle correctly the case where the result of a 13363comparison is unordered. 13364 13365@item -msoft-float 13366@opindex msoft-float 13367Generate output containing library calls for floating point. 13368@strong{Warning:} the requisite libraries are not part of GCC@. 13369Normally the facilities of the machine's usual C compiler are used, but 13370this can't be done directly in cross-compilation. You must make your 13371own arrangements to provide suitable library functions for 13372cross-compilation. 13373 13374On machines where a function returns floating-point results in the 80387 13375register stack, some floating-point opcodes may be emitted even if 13376@option{-msoft-float} is used. 13377 13378@item -mno-fp-ret-in-387 13379@opindex mno-fp-ret-in-387 13380Do not use the FPU registers for return values of functions. 13381 13382The usual calling convention has functions return values of types 13383@code{float} and @code{double} in an FPU register, even if there 13384is no FPU@. The idea is that the operating system should emulate 13385an FPU@. 13386 13387The option @option{-mno-fp-ret-in-387} causes such values to be returned 13388in ordinary CPU registers instead. 13389 13390@item -mno-fancy-math-387 13391@opindex mno-fancy-math-387 13392Some 387 emulators do not support the @code{sin}, @code{cos} and 13393@code{sqrt} instructions for the 387. Specify this option to avoid 13394generating those instructions. This option is the default on FreeBSD, 13395OpenBSD and NetBSD@. This option is overridden when @option{-march} 13396indicates that the target CPU will always have an FPU and so the 13397instruction will not need emulation. As of revision 2.6.1, these 13398instructions are not generated unless you also use the 13399@option{-funsafe-math-optimizations} switch. 13400 13401@item -malign-double 13402@itemx -mno-align-double 13403@opindex malign-double 13404@opindex mno-align-double 13405Control whether GCC aligns @code{double}, @code{long double}, and 13406@code{long long} variables on a two-word boundary or a one-word 13407boundary. Aligning @code{double} variables on a two-word boundary 13408produces code that runs somewhat faster on a @samp{Pentium} at the 13409expense of more memory. 13410 13411On x86-64, @option{-malign-double} is enabled by default. 13412 13413@strong{Warning:} if you use the @option{-malign-double} switch, 13414structures containing the above types will be aligned differently than 13415the published application binary interface specifications for the 386 13416and will not be binary compatible with structures in code compiled 13417without that switch. 13418 13419@item -m96bit-long-double 13420@itemx -m128bit-long-double 13421@opindex m96bit-long-double 13422@opindex m128bit-long-double 13423These switches control the size of @code{long double} type. The i386 13424application binary interface specifies the size to be 96 bits, 13425so @option{-m96bit-long-double} is the default in 32-bit mode. 13426 13427Modern architectures (Pentium and newer) prefer @code{long double} 13428to be aligned to an 8- or 16-byte boundary. In arrays or structures 13429conforming to the ABI, this is not possible. So specifying 13430@option{-m128bit-long-double} aligns @code{long double} 13431to a 16-byte boundary by padding the @code{long double} with an additional 1343232-bit zero. 13433 13434In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as 13435its ABI specifies that @code{long double} is to be aligned on 16-byte boundary. 13436 13437Notice that neither of these options enable any extra precision over the x87 13438standard of 80 bits for a @code{long double}. 13439 13440@strong{Warning:} if you override the default value for your target ABI, the 13441structures and arrays containing @code{long double} variables will change 13442their size as well as function calling convention for function taking 13443@code{long double} will be modified. Hence they will not be binary 13444compatible with arrays or structures in code compiled without that switch. 13445 13446@item -mlarge-data-threshold=@var{number} 13447@opindex mlarge-data-threshold=@var{number} 13448When @option{-mcmodel=medium} is specified, the data greater than 13449@var{threshold} are placed in large data section. This value must be the 13450same across all object linked into the binary and defaults to 65535. 13451 13452@item -mrtd 13453@opindex mrtd 13454Use a different function-calling convention, in which functions that 13455take a fixed number of arguments return with the @code{ret} @var{num} 13456instruction, which pops their arguments while returning. This saves one 13457instruction in the caller since there is no need to pop the arguments 13458there. 13459 13460You can specify that an individual function is called with this calling 13461sequence with the function attribute @samp{stdcall}. You can also 13462override the @option{-mrtd} option by using the function attribute 13463@samp{cdecl}. @xref{Function Attributes}. 13464 13465@strong{Warning:} this calling convention is incompatible with the one 13466normally used on Unix, so you cannot use it if you need to call 13467libraries compiled with the Unix compiler. 13468 13469Also, you must provide function prototypes for all functions that 13470take variable numbers of arguments (including @code{printf}); 13471otherwise incorrect code will be generated for calls to those 13472functions. 13473 13474In addition, seriously incorrect code will result if you call a 13475function with too many arguments. (Normally, extra arguments are 13476harmlessly ignored.) 13477 13478@item -mregparm=@var{num} 13479@opindex mregparm 13480Control how many registers are used to pass integer arguments. By 13481default, no registers are used to pass arguments, and at most 3 13482registers can be used. You can control this behavior for a specific 13483function by using the function attribute @samp{regparm}. 13484@xref{Function Attributes}. 13485 13486@strong{Warning:} if you use this switch, and 13487@var{num} is nonzero, then you must build all modules with the same 13488value, including any libraries. This includes the system libraries and 13489startup modules. 13490 13491@item -msseregparm 13492@opindex msseregparm 13493Use SSE register passing conventions for float and double arguments 13494and return values. You can control this behavior for a specific 13495function by using the function attribute @samp{sseregparm}. 13496@xref{Function Attributes}. 13497 13498@strong{Warning:} if you use this switch then you must build all 13499modules with the same value, including any libraries. This includes 13500the system libraries and startup modules. 13501 13502@item -mvect8-ret-in-mem 13503@opindex mvect8-ret-in-mem 13504Return 8-byte vectors in memory instead of MMX registers. This is the 13505default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun 13506Studio compilers until version 12. Later compiler versions (starting 13507with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which 13508is the default on Solaris@tie{}10 and later. @emph{Only} use this option if 13509you need to remain compatible with existing code produced by those 13510previous compiler versions or older versions of GCC. 13511 13512@item -mpc32 13513@itemx -mpc64 13514@itemx -mpc80 13515@opindex mpc32 13516@opindex mpc64 13517@opindex mpc80 13518 13519Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32} 13520is specified, the significands of results of floating-point operations are 13521rounded to 24 bits (single precision); @option{-mpc64} rounds the 13522significands of results of floating-point operations to 53 bits (double 13523precision) and @option{-mpc80} rounds the significands of results of 13524floating-point operations to 64 bits (extended double precision), which is 13525the default. When this option is used, floating-point operations in higher 13526precisions are not available to the programmer without setting the FPU 13527control word explicitly. 13528 13529Setting the rounding of floating-point operations to less than the default 1353080 bits can speed some programs by 2% or more. Note that some mathematical 13531libraries assume that extended-precision (80-bit) floating-point operations 13532are enabled by default; routines in such libraries could suffer significant 13533loss of accuracy, typically through so-called "catastrophic cancellation", 13534when this option is used to set the precision to less than extended precision. 13535 13536@item -mstackrealign 13537@opindex mstackrealign 13538Realign the stack at entry. On the Intel x86, the @option{-mstackrealign} 13539option will generate an alternate prologue and epilogue that realigns the 13540run-time stack if necessary. This supports mixing legacy codes that keep 13541a 4-byte aligned stack with modern codes that keep a 16-byte stack for 13542SSE compatibility. See also the attribute @code{force_align_arg_pointer}, 13543applicable to individual functions. 13544 13545@item -mpreferred-stack-boundary=@var{num} 13546@opindex mpreferred-stack-boundary 13547Attempt to keep the stack boundary aligned to a 2 raised to @var{num} 13548byte boundary. If @option{-mpreferred-stack-boundary} is not specified, 13549the default is 4 (16 bytes or 128 bits). 13550 13551@item -mincoming-stack-boundary=@var{num} 13552@opindex mincoming-stack-boundary 13553Assume the incoming stack is aligned to a 2 raised to @var{num} byte 13554boundary. If @option{-mincoming-stack-boundary} is not specified, 13555the one specified by @option{-mpreferred-stack-boundary} will be used. 13556 13557On Pentium and PentiumPro, @code{double} and @code{long double} values 13558should be aligned to an 8-byte boundary (see @option{-malign-double}) or 13559suffer significant run time performance penalties. On Pentium III, the 13560Streaming SIMD Extension (SSE) data type @code{__m128} may not work 13561properly if it is not 16-byte aligned. 13562 13563To ensure proper alignment of this values on the stack, the stack boundary 13564must be as aligned as that required by any value stored on the stack. 13565Further, every function must be generated such that it keeps the stack 13566aligned. Thus calling a function compiled with a higher preferred 13567stack boundary from a function compiled with a lower preferred stack 13568boundary will most likely misalign the stack. It is recommended that 13569libraries that use callbacks always use the default setting. 13570 13571This extra alignment does consume extra stack space, and generally 13572increases code size. Code that is sensitive to stack space usage, such 13573as embedded systems and operating system kernels, may want to reduce the 13574preferred alignment to @option{-mpreferred-stack-boundary=2}. 13575 13576@item -mmmx 13577@itemx -mno-mmx 13578@itemx -msse 13579@itemx -mno-sse 13580@itemx -msse2 13581@itemx -mno-sse2 13582@itemx -msse3 13583@itemx -mno-sse3 13584@itemx -mssse3 13585@itemx -mno-ssse3 13586@itemx -msse4.1 13587@need 800 13588@itemx -mno-sse4.1 13589@itemx -msse4.2 13590@itemx -mno-sse4.2 13591@itemx -msse4 13592@itemx -mno-sse4 13593@itemx -mavx 13594@itemx -mno-avx 13595@itemx -mavx2 13596@itemx -mno-avx2 13597@itemx -maes 13598@itemx -mno-aes 13599@itemx -mpclmul 13600@need 800 13601@itemx -mno-pclmul 13602@itemx -mfsgsbase 13603@itemx -mno-fsgsbase 13604@itemx -mrdrnd 13605@itemx -mno-rdrnd 13606@itemx -mf16c 13607@itemx -mno-f16c 13608@itemx -mfma 13609@itemx -mno-fma 13610@itemx -msse4a 13611@itemx -mno-sse4a 13612@itemx -mfma4 13613@need 800 13614@itemx -mno-fma4 13615@itemx -mxop 13616@itemx -mno-xop 13617@itemx -mlwp 13618@itemx -mno-lwp 13619@itemx -m3dnow 13620@itemx -mno-3dnow 13621@itemx -mpopcnt 13622@itemx -mno-popcnt 13623@itemx -mabm 13624@itemx -mno-abm 13625@itemx -mbmi 13626@itemx -mbmi2 13627@itemx -mno-bmi 13628@itemx -mno-bmi2 13629@itemx -mlzcnt 13630@itemx -mno-lzcnt 13631@itemx -mtbm 13632@itemx -mno-tbm 13633@opindex mmmx 13634@opindex mno-mmx 13635@opindex msse 13636@opindex mno-sse 13637@opindex m3dnow 13638@opindex mno-3dnow 13639These switches enable or disable the use of instructions in the MMX, SSE, 13640SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, F16C, 13641FMA, SSE4A, FMA4, XOP, LWP, ABM, BMI, BMI2, LZCNT or 3DNow! 13642@: extended instruction sets. 13643These extensions are also available as built-in functions: see 13644@ref{X86 Built-in Functions}, for details of the functions enabled and 13645disabled by these switches. 13646 13647To have SSE/SSE2 instructions generated automatically from floating-point 13648code (as opposed to 387 instructions), see @option{-mfpmath=sse}. 13649 13650GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it 13651generates new AVX instructions or AVX equivalence for all SSEx instructions 13652when needed. 13653 13654These options will enable GCC to use these extended instructions in 13655generated code, even without @option{-mfpmath=sse}. Applications that 13656perform run-time CPU detection must compile separate files for each 13657supported architecture, using the appropriate flags. In particular, 13658the file containing the CPU detection code should be compiled without 13659these options. 13660 13661@item -mcld 13662@opindex mcld 13663This option instructs GCC to emit a @code{cld} instruction in the prologue 13664of functions that use string instructions. String instructions depend on 13665the DF flag to select between autoincrement or autodecrement mode. While the 13666ABI specifies the DF flag to be cleared on function entry, some operating 13667systems violate this specification by not clearing the DF flag in their 13668exception dispatchers. The exception handler can be invoked with the DF flag 13669set, which leads to wrong direction mode when string instructions are used. 13670This option can be enabled by default on 32-bit x86 targets by configuring 13671GCC with the @option{--enable-cld} configure option. Generation of @code{cld} 13672instructions can be suppressed with the @option{-mno-cld} compiler option 13673in this case. 13674 13675@item -mvzeroupper 13676@opindex mvzeroupper 13677This option instructs GCC to emit a @code{vzeroupper} instruction 13678before a transfer of control flow out of the function to minimize 13679AVX to SSE transition penalty as well as remove unnecessary zeroupper 13680intrinsics. 13681 13682@item -mcx16 13683@opindex mcx16 13684This option will enable GCC to use CMPXCHG16B instruction in generated code. 13685CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword) 13686data types. This is useful for high resolution counters that could be updated 13687by multiple processors (or cores). This instruction is generated as part of 13688atomic built-in functions: see @ref{__sync Builtins} or 13689@ref{__atomic Builtins} for details. 13690 13691@item -msahf 13692@opindex msahf 13693This option will enable GCC to use SAHF instruction in generated 64-bit code. 13694Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported 13695by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and 13696SAHF are load and store instructions, respectively, for certain status flags. 13697In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem} 13698or @code{remainder} built-in functions: see @ref{Other Builtins} for details. 13699 13700@item -mmovbe 13701@opindex mmovbe 13702This option will enable GCC to use movbe instruction to implement 13703@code{__builtin_bswap32} and @code{__builtin_bswap64}. 13704 13705@item -mcrc32 13706@opindex mcrc32 13707This option will enable built-in functions, @code{__builtin_ia32_crc32qi}, 13708@code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and 13709@code{__builtin_ia32_crc32di} to generate the crc32 machine instruction. 13710 13711@item -mrecip 13712@opindex mrecip 13713This option will enable GCC to use RCPSS and RSQRTSS instructions (and their 13714vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step 13715to increase precision instead of DIVSS and SQRTSS (and their vectorized 13716variants) for single-precision floating-point arguments. These instructions 13717are generated only when @option{-funsafe-math-optimizations} is enabled 13718together with @option{-finite-math-only} and @option{-fno-trapping-math}. 13719Note that while the throughput of the sequence is higher than the throughput 13720of the non-reciprocal instruction, the precision of the sequence can be 13721decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994). 13722 13723Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of RSQRTSS 13724(or RSQRTPS) already with @option{-ffast-math} (or the above option 13725combination), and doesn't need @option{-mrecip}. 13726 13727Also note that GCC emits the above sequence with additional Newton-Raphson step 13728for vectorized single-float division and vectorized @code{sqrtf(@var{x})} 13729already with @option{-ffast-math} (or the above option combination), and 13730doesn't need @option{-mrecip}. 13731 13732@item -mrecip=@var{opt} 13733@opindex mrecip=opt 13734This option allows to control which reciprocal estimate instructions 13735may be used. @var{opt} is a comma separated list of options, which may 13736be preceded by a @code{!} to invert the option: 13737@code{all}: enable all estimate instructions, 13738@code{default}: enable the default instructions, equivalent to @option{-mrecip}, 13739@code{none}: disable all estimate instructions, equivalent to @option{-mno-recip}, 13740@code{div}: enable the approximation for scalar division, 13741@code{vec-div}: enable the approximation for vectorized division, 13742@code{sqrt}: enable the approximation for scalar square root, 13743@code{vec-sqrt}: enable the approximation for vectorized square root. 13744 13745So for example, @option{-mrecip=all,!sqrt} would enable 13746all of the reciprocal approximations, except for square root. 13747 13748@item -mveclibabi=@var{type} 13749@opindex mveclibabi 13750Specifies the ABI type to use for vectorizing intrinsics using an 13751external library. Supported types are @code{svml} for the Intel short 13752vector math library and @code{acml} for the AMD math core library style 13753of interfacing. GCC will currently emit calls to @code{vmldExp2}, 13754@code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2}, 13755@code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2}, 13756@code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2}, 13757@code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2}, 13758@code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104}, 13759@code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4}, 13760@code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4}, 13761@code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4}, 13762@code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding 13763function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin}, 13764@code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2}, 13765@code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf}, 13766@code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f}, 13767@code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type 13768when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and 13769@option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI 13770compatible library will have to be specified at link time. 13771 13772@item -mabi=@var{name} 13773@opindex mabi 13774Generate code for the specified calling convention. Permissible values 13775are: @samp{sysv} for the ABI used on GNU/Linux and other systems and 13776@samp{ms} for the Microsoft ABI. The default is to use the Microsoft 13777ABI when targeting Windows. On all other systems, the default is the 13778SYSV ABI. You can control this behavior for a specific function by 13779using the function attribute @samp{ms_abi}/@samp{sysv_abi}. 13780@xref{Function Attributes}. 13781 13782@item -mtls-dialect=@var{type} 13783@opindex mtls-dialect 13784Generate code to access thread-local storage using the @samp{gnu} or 13785@samp{gnu2} conventions. @samp{gnu} is the conservative default; 13786@samp{gnu2} is more efficient, but it may add compile- and run-time 13787requirements that cannot be satisfied on all systems. 13788 13789@item -mpush-args 13790@itemx -mno-push-args 13791@opindex mpush-args 13792@opindex mno-push-args 13793Use PUSH operations to store outgoing parameters. This method is shorter 13794and usually equally fast as method using SUB/MOV operations and is enabled 13795by default. In some cases disabling it may improve performance because of 13796improved scheduling and reduced dependencies. 13797 13798@item -maccumulate-outgoing-args 13799@opindex maccumulate-outgoing-args 13800If enabled, the maximum amount of space required for outgoing arguments will be 13801computed in the function prologue. This is faster on most modern CPUs 13802because of reduced dependencies, improved scheduling and reduced stack usage 13803when preferred stack boundary is not equal to 2. The drawback is a notable 13804increase in code size. This switch implies @option{-mno-push-args}. 13805 13806@item -mthreads 13807@opindex mthreads 13808Support thread-safe exception handling on @samp{Mingw32}. Code that relies 13809on thread-safe exception handling must compile and link all code with the 13810@option{-mthreads} option. When compiling, @option{-mthreads} defines 13811@option{-D_MT}; when linking, it links in a special thread helper library 13812@option{-lmingwthrd} which cleans up per thread exception handling data. 13813 13814@item -mno-align-stringops 13815@opindex mno-align-stringops 13816Do not align destination of inlined string operations. This switch reduces 13817code size and improves performance in case the destination is already aligned, 13818but GCC doesn't know about it. 13819 13820@item -minline-all-stringops 13821@opindex minline-all-stringops 13822By default GCC inlines string operations only when the destination is 13823known to be aligned to least a 4-byte boundary. 13824This enables more inlining, increase code 13825size, but may improve performance of code that depends on fast memcpy, strlen 13826and memset for short lengths. 13827 13828@item -minline-stringops-dynamically 13829@opindex minline-stringops-dynamically 13830For string operations of unknown size, use run-time checks with 13831inline code for small blocks and a library call for large blocks. 13832 13833@item -mstringop-strategy=@var{alg} 13834@opindex mstringop-strategy=@var{alg} 13835Overwrite internal decision heuristic about particular algorithm to inline 13836string operation with. The allowed values are @code{rep_byte}, 13837@code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix 13838of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for 13839expanding inline loop, @code{libcall} for always expanding library call. 13840 13841@item -momit-leaf-frame-pointer 13842@opindex momit-leaf-frame-pointer 13843Don't keep the frame pointer in a register for leaf functions. This 13844avoids the instructions to save, set up and restore frame pointers and 13845makes an extra register available in leaf functions. The option 13846@option{-fomit-frame-pointer} removes the frame pointer for all functions, 13847which might make debugging harder. 13848 13849@item -mtls-direct-seg-refs 13850@itemx -mno-tls-direct-seg-refs 13851@opindex mtls-direct-seg-refs 13852Controls whether TLS variables may be accessed with offsets from the 13853TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit), 13854or whether the thread base pointer must be added. Whether or not this 13855is legal depends on the operating system, and whether it maps the 13856segment to cover the entire TLS area. 13857 13858For systems that use GNU libc, the default is on. 13859 13860@item -msse2avx 13861@itemx -mno-sse2avx 13862@opindex msse2avx 13863Specify that the assembler should encode SSE instructions with VEX 13864prefix. The option @option{-mavx} turns this on by default. 13865 13866@item -mfentry 13867@itemx -mno-fentry 13868@opindex mfentry 13869If profiling is active @option{-pg} put the profiling 13870counter call before prologue. 13871Note: On x86 architectures the attribute @code{ms_hook_prologue} 13872isn't possible at the moment for @option{-mfentry} and @option{-pg}. 13873 13874@item -m8bit-idiv 13875@itemx -mno-8bit-idiv 13876@opindex 8bit-idiv 13877On some processors, like Intel Atom, 8-bit unsigned integer divide is 13878much faster than 32-bit/64-bit integer divide. This option generates a 13879run-time check. If both dividend and divisor are within range of 0 13880to 255, 8-bit unsigned integer divide is used instead of 1388132-bit/64-bit integer divide. 13882 13883@item -mavx256-split-unaligned-load 13884@item -mavx256-split-unaligned-store 13885@opindex avx256-split-unaligned-load 13886@opindex avx256-split-unaligned-store 13887Split 32-byte AVX unaligned load and store. 13888 13889@end table 13890 13891These @samp{-m} switches are supported in addition to the above 13892on AMD x86-64 processors in 64-bit environments. 13893 13894@table @gcctabopt 13895@item -m32 13896@itemx -m64 13897@itemx -mx32 13898@opindex m32 13899@opindex m64 13900@opindex mx32 13901Generate code for a 32-bit or 64-bit environment. 13902The @option{-m32} option sets int, long and pointer to 32 bits and 13903generates code that runs on any i386 system. 13904The @option{-m64} option sets int to 32 bits and long and pointer 13905to 64 bits and generates code for AMD's x86-64 architecture. 13906The @option{-mx32} option sets int, long and pointer to 32 bits and 13907generates code for AMD's x86-64 architecture. 13908For darwin only the @option{-m64} option turns off the @option{-fno-pic} 13909and @option{-mdynamic-no-pic} options. 13910 13911@item -mno-red-zone 13912@opindex mno-red-zone 13913Do not use a so called red zone for x86-64 code. The red zone is mandated 13914by the x86-64 ABI, it is a 128-byte area beyond the location of the 13915stack pointer that will not be modified by signal or interrupt handlers 13916and therefore can be used for temporary data without adjusting the stack 13917pointer. The flag @option{-mno-red-zone} disables this red zone. 13918 13919@item -mcmodel=small 13920@opindex mcmodel=small 13921Generate code for the small code model: the program and its symbols must 13922be linked in the lower 2 GB of the address space. Pointers are 64 bits. 13923Programs can be statically or dynamically linked. This is the default 13924code model. 13925 13926@item -mcmodel=kernel 13927@opindex mcmodel=kernel 13928Generate code for the kernel code model. The kernel runs in the 13929negative 2 GB of the address space. 13930This model has to be used for Linux kernel code. 13931 13932@item -mcmodel=medium 13933@opindex mcmodel=medium 13934Generate code for the medium model: The program is linked in the lower 2 13935GB of the address space. Small symbols are also placed there. Symbols 13936with sizes larger than @option{-mlarge-data-threshold} are put into 13937large data or bss sections and can be located above 2GB. Programs can 13938be statically or dynamically linked. 13939 13940@item -mcmodel=large 13941@opindex mcmodel=large 13942Generate code for the large model: This model makes no assumptions 13943about addresses and sizes of sections. 13944@end table 13945 13946@node i386 and x86-64 Windows Options 13947@subsection i386 and x86-64 Windows Options 13948@cindex i386 and x86-64 Windows Options 13949 13950These additional options are available for Windows targets: 13951 13952@table @gcctabopt 13953@item -mconsole 13954@opindex mconsole 13955This option is available for Cygwin and MinGW targets. It 13956specifies that a console application is to be generated, by 13957instructing the linker to set the PE header subsystem type 13958required for console applications. 13959This is the default behavior for Cygwin and MinGW targets. 13960 13961@item -mdll 13962@opindex mdll 13963This option is available for Cygwin and MinGW targets. It 13964specifies that a DLL - a dynamic link library - is to be 13965generated, enabling the selection of the required runtime 13966startup object and entry point. 13967 13968@item -mnop-fun-dllimport 13969@opindex mnop-fun-dllimport 13970This option is available for Cygwin and MinGW targets. It 13971specifies that the dllimport attribute should be ignored. 13972 13973@item -mthread 13974@opindex mthread 13975This option is available for MinGW targets. It specifies 13976that MinGW-specific thread support is to be used. 13977 13978@item -municode 13979@opindex municode 13980This option is available for mingw-w64 targets. It specifies 13981that the UNICODE macro is getting pre-defined and that the 13982unicode capable runtime startup code is chosen. 13983 13984@item -mwin32 13985@opindex mwin32 13986This option is available for Cygwin and MinGW targets. It 13987specifies that the typical Windows pre-defined macros are to 13988be set in the pre-processor, but does not influence the choice 13989of runtime library/startup code. 13990 13991@item -mwindows 13992@opindex mwindows 13993This option is available for Cygwin and MinGW targets. It 13994specifies that a GUI application is to be generated by 13995instructing the linker to set the PE header subsystem type 13996appropriately. 13997 13998@item -fno-set-stack-executable 13999@opindex fno-set-stack-executable 14000This option is available for MinGW targets. It specifies that 14001the executable flag for stack used by nested functions isn't 14002set. This is necessary for binaries running in kernel mode of 14003Windows, as there the user32 API, which is used to set executable 14004privileges, isn't available. 14005 14006@item -mpe-aligned-commons 14007@opindex mpe-aligned-commons 14008This option is available for Cygwin and MinGW targets. It 14009specifies that the GNU extension to the PE file format that 14010permits the correct alignment of COMMON variables should be 14011used when generating code. It will be enabled by default if 14012GCC detects that the target assembler found during configuration 14013supports the feature. 14014@end table 14015 14016See also under @ref{i386 and x86-64 Options} for standard options. 14017 14018@node IA-64 Options 14019@subsection IA-64 Options 14020@cindex IA-64 Options 14021 14022These are the @samp{-m} options defined for the Intel IA-64 architecture. 14023 14024@table @gcctabopt 14025@item -mbig-endian 14026@opindex mbig-endian 14027Generate code for a big-endian target. This is the default for HP-UX@. 14028 14029@item -mlittle-endian 14030@opindex mlittle-endian 14031Generate code for a little-endian target. This is the default for AIX5 14032and GNU/Linux. 14033 14034@item -mgnu-as 14035@itemx -mno-gnu-as 14036@opindex mgnu-as 14037@opindex mno-gnu-as 14038Generate (or don't) code for the GNU assembler. This is the default. 14039@c Also, this is the default if the configure option @option{--with-gnu-as} 14040@c is used. 14041 14042@item -mgnu-ld 14043@itemx -mno-gnu-ld 14044@opindex mgnu-ld 14045@opindex mno-gnu-ld 14046Generate (or don't) code for the GNU linker. This is the default. 14047@c Also, this is the default if the configure option @option{--with-gnu-ld} 14048@c is used. 14049 14050@item -mno-pic 14051@opindex mno-pic 14052Generate code that does not use a global pointer register. The result 14053is not position independent code, and violates the IA-64 ABI@. 14054 14055@item -mvolatile-asm-stop 14056@itemx -mno-volatile-asm-stop 14057@opindex mvolatile-asm-stop 14058@opindex mno-volatile-asm-stop 14059Generate (or don't) a stop bit immediately before and after volatile asm 14060statements. 14061 14062@item -mregister-names 14063@itemx -mno-register-names 14064@opindex mregister-names 14065@opindex mno-register-names 14066Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for 14067the stacked registers. This may make assembler output more readable. 14068 14069@item -mno-sdata 14070@itemx -msdata 14071@opindex mno-sdata 14072@opindex msdata 14073Disable (or enable) optimizations that use the small data section. This may 14074be useful for working around optimizer bugs. 14075 14076@item -mconstant-gp 14077@opindex mconstant-gp 14078Generate code that uses a single constant global pointer value. This is 14079useful when compiling kernel code. 14080 14081@item -mauto-pic 14082@opindex mauto-pic 14083Generate code that is self-relocatable. This implies @option{-mconstant-gp}. 14084This is useful when compiling firmware code. 14085 14086@item -minline-float-divide-min-latency 14087@opindex minline-float-divide-min-latency 14088Generate code for inline divides of floating-point values 14089using the minimum latency algorithm. 14090 14091@item -minline-float-divide-max-throughput 14092@opindex minline-float-divide-max-throughput 14093Generate code for inline divides of floating-point values 14094using the maximum throughput algorithm. 14095 14096@item -mno-inline-float-divide 14097@opindex mno-inline-float-divide 14098Do not generate inline code for divides of floating-point values. 14099 14100@item -minline-int-divide-min-latency 14101@opindex minline-int-divide-min-latency 14102Generate code for inline divides of integer values 14103using the minimum latency algorithm. 14104 14105@item -minline-int-divide-max-throughput 14106@opindex minline-int-divide-max-throughput 14107Generate code for inline divides of integer values 14108using the maximum throughput algorithm. 14109 14110@item -mno-inline-int-divide 14111@opindex mno-inline-int-divide 14112Do not generate inline code for divides of integer values. 14113 14114@item -minline-sqrt-min-latency 14115@opindex minline-sqrt-min-latency 14116Generate code for inline square roots 14117using the minimum latency algorithm. 14118 14119@item -minline-sqrt-max-throughput 14120@opindex minline-sqrt-max-throughput 14121Generate code for inline square roots 14122using the maximum throughput algorithm. 14123 14124@item -mno-inline-sqrt 14125@opindex mno-inline-sqrt 14126Do not generate inline code for sqrt. 14127 14128@item -mfused-madd 14129@itemx -mno-fused-madd 14130@opindex mfused-madd 14131@opindex mno-fused-madd 14132Do (don't) generate code that uses the fused multiply/add or multiply/subtract 14133instructions. The default is to use these instructions. 14134 14135@item -mno-dwarf2-asm 14136@itemx -mdwarf2-asm 14137@opindex mno-dwarf2-asm 14138@opindex mdwarf2-asm 14139Don't (or do) generate assembler code for the DWARF2 line number debugging 14140info. This may be useful when not using the GNU assembler. 14141 14142@item -mearly-stop-bits 14143@itemx -mno-early-stop-bits 14144@opindex mearly-stop-bits 14145@opindex mno-early-stop-bits 14146Allow stop bits to be placed earlier than immediately preceding the 14147instruction that triggered the stop bit. This can improve instruction 14148scheduling, but does not always do so. 14149 14150@item -mfixed-range=@var{register-range} 14151@opindex mfixed-range 14152Generate code treating the given register range as fixed registers. 14153A fixed register is one that the register allocator can not use. This is 14154useful when compiling kernel code. A register range is specified as 14155two registers separated by a dash. Multiple register ranges can be 14156specified separated by a comma. 14157 14158@item -mtls-size=@var{tls-size} 14159@opindex mtls-size 14160Specify bit size of immediate TLS offsets. Valid values are 14, 22, and 1416164. 14162 14163@item -mtune=@var{cpu-type} 14164@opindex mtune 14165Tune the instruction scheduling for a particular CPU, Valid values are 14166itanium, itanium1, merced, itanium2, and mckinley. 14167 14168@item -milp32 14169@itemx -mlp64 14170@opindex milp32 14171@opindex mlp64 14172Generate code for a 32-bit or 64-bit environment. 14173The 32-bit environment sets int, long and pointer to 32 bits. 14174The 64-bit environment sets int to 32 bits and long and pointer 14175to 64 bits. These are HP-UX specific flags. 14176 14177@item -mno-sched-br-data-spec 14178@itemx -msched-br-data-spec 14179@opindex mno-sched-br-data-spec 14180@opindex msched-br-data-spec 14181(Dis/En)able data speculative scheduling before reload. 14182This will result in generation of the ld.a instructions and 14183the corresponding check instructions (ld.c / chk.a). 14184The default is 'disable'. 14185 14186@item -msched-ar-data-spec 14187@itemx -mno-sched-ar-data-spec 14188@opindex msched-ar-data-spec 14189@opindex mno-sched-ar-data-spec 14190(En/Dis)able data speculative scheduling after reload. 14191This will result in generation of the ld.a instructions and 14192the corresponding check instructions (ld.c / chk.a). 14193The default is 'enable'. 14194 14195@item -mno-sched-control-spec 14196@itemx -msched-control-spec 14197@opindex mno-sched-control-spec 14198@opindex msched-control-spec 14199(Dis/En)able control speculative scheduling. This feature is 14200available only during region scheduling (i.e.@: before reload). 14201This will result in generation of the ld.s instructions and 14202the corresponding check instructions chk.s . 14203The default is 'disable'. 14204 14205@item -msched-br-in-data-spec 14206@itemx -mno-sched-br-in-data-spec 14207@opindex msched-br-in-data-spec 14208@opindex mno-sched-br-in-data-spec 14209(En/Dis)able speculative scheduling of the instructions that 14210are dependent on the data speculative loads before reload. 14211This is effective only with @option{-msched-br-data-spec} enabled. 14212The default is 'enable'. 14213 14214@item -msched-ar-in-data-spec 14215@itemx -mno-sched-ar-in-data-spec 14216@opindex msched-ar-in-data-spec 14217@opindex mno-sched-ar-in-data-spec 14218(En/Dis)able speculative scheduling of the instructions that 14219are dependent on the data speculative loads after reload. 14220This is effective only with @option{-msched-ar-data-spec} enabled. 14221The default is 'enable'. 14222 14223@item -msched-in-control-spec 14224@itemx -mno-sched-in-control-spec 14225@opindex msched-in-control-spec 14226@opindex mno-sched-in-control-spec 14227(En/Dis)able speculative scheduling of the instructions that 14228are dependent on the control speculative loads. 14229This is effective only with @option{-msched-control-spec} enabled. 14230The default is 'enable'. 14231 14232@item -mno-sched-prefer-non-data-spec-insns 14233@itemx -msched-prefer-non-data-spec-insns 14234@opindex mno-sched-prefer-non-data-spec-insns 14235@opindex msched-prefer-non-data-spec-insns 14236If enabled, data speculative instructions will be chosen for schedule 14237only if there are no other choices at the moment. This will make 14238the use of the data speculation much more conservative. 14239The default is 'disable'. 14240 14241@item -mno-sched-prefer-non-control-spec-insns 14242@itemx -msched-prefer-non-control-spec-insns 14243@opindex mno-sched-prefer-non-control-spec-insns 14244@opindex msched-prefer-non-control-spec-insns 14245If enabled, control speculative instructions will be chosen for schedule 14246only if there are no other choices at the moment. This will make 14247the use of the control speculation much more conservative. 14248The default is 'disable'. 14249 14250@item -mno-sched-count-spec-in-critical-path 14251@itemx -msched-count-spec-in-critical-path 14252@opindex mno-sched-count-spec-in-critical-path 14253@opindex msched-count-spec-in-critical-path 14254If enabled, speculative dependencies will be considered during 14255computation of the instructions priorities. This will make the use of the 14256speculation a bit more conservative. 14257The default is 'disable'. 14258 14259@item -msched-spec-ldc 14260@opindex msched-spec-ldc 14261Use a simple data speculation check. This option is on by default. 14262 14263@item -msched-control-spec-ldc 14264@opindex msched-spec-ldc 14265Use a simple check for control speculation. This option is on by default. 14266 14267@item -msched-stop-bits-after-every-cycle 14268@opindex msched-stop-bits-after-every-cycle 14269Place a stop bit after every cycle when scheduling. This option is on 14270by default. 14271 14272@item -msched-fp-mem-deps-zero-cost 14273@opindex msched-fp-mem-deps-zero-cost 14274Assume that floating-point stores and loads are not likely to cause a conflict 14275when placed into the same instruction group. This option is disabled by 14276default. 14277 14278@item -msel-sched-dont-check-control-spec 14279@opindex msel-sched-dont-check-control-spec 14280Generate checks for control speculation in selective scheduling. 14281This flag is disabled by default. 14282 14283@item -msched-max-memory-insns=@var{max-insns} 14284@opindex msched-max-memory-insns 14285Limit on the number of memory insns per instruction group, giving lower 14286priority to subsequent memory insns attempting to schedule in the same 14287instruction group. Frequently useful to prevent cache bank conflicts. 14288The default value is 1. 14289 14290@item -msched-max-memory-insns-hard-limit 14291@opindex msched-max-memory-insns-hard-limit 14292Disallow more than `msched-max-memory-insns' in instruction group. 14293Otherwise, limit is `soft' meaning that we would prefer non-memory operations 14294when limit is reached but may still schedule memory operations. 14295 14296@end table 14297 14298@node IA-64/VMS Options 14299@subsection IA-64/VMS Options 14300 14301These @samp{-m} options are defined for the IA-64/VMS implementations: 14302 14303@table @gcctabopt 14304@item -mvms-return-codes 14305@opindex mvms-return-codes 14306Return VMS condition codes from main. The default is to return POSIX 14307style condition (e.g.@ error) codes. 14308 14309@item -mdebug-main=@var{prefix} 14310@opindex mdebug-main=@var{prefix} 14311Flag the first routine whose name starts with @var{prefix} as the main 14312routine for the debugger. 14313 14314@item -mmalloc64 14315@opindex mmalloc64 14316Default to 64-bit memory allocation routines. 14317@end table 14318 14319@node LM32 Options 14320@subsection LM32 Options 14321@cindex LM32 options 14322 14323These @option{-m} options are defined for the Lattice Mico32 architecture: 14324 14325@table @gcctabopt 14326@item -mbarrel-shift-enabled 14327@opindex mbarrel-shift-enabled 14328Enable barrel-shift instructions. 14329 14330@item -mdivide-enabled 14331@opindex mdivide-enabled 14332Enable divide and modulus instructions. 14333 14334@item -mmultiply-enabled 14335@opindex multiply-enabled 14336Enable multiply instructions. 14337 14338@item -msign-extend-enabled 14339@opindex msign-extend-enabled 14340Enable sign extend instructions. 14341 14342@item -muser-enabled 14343@opindex muser-enabled 14344Enable user-defined instructions. 14345 14346@end table 14347 14348@node M32C Options 14349@subsection M32C Options 14350@cindex M32C options 14351 14352@table @gcctabopt 14353@item -mcpu=@var{name} 14354@opindex mcpu= 14355Select the CPU for which code is generated. @var{name} may be one of 14356@samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to 14357/60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for 14358the M32C/80 series. 14359 14360@item -msim 14361@opindex msim 14362Specifies that the program will be run on the simulator. This causes 14363an alternate runtime library to be linked in which supports, for 14364example, file I/O@. You must not use this option when generating 14365programs that will run on real hardware; you must provide your own 14366runtime library for whatever I/O functions are needed. 14367 14368@item -memregs=@var{number} 14369@opindex memregs= 14370Specifies the number of memory-based pseudo-registers GCC will use 14371during code generation. These pseudo-registers will be used like real 14372registers, so there is a tradeoff between GCC's ability to fit the 14373code into available registers, and the performance penalty of using 14374memory instead of registers. Note that all modules in a program must 14375be compiled with the same value for this option. Because of that, you 14376must not use this option with the default runtime libraries gcc 14377builds. 14378 14379@end table 14380 14381@node M32R/D Options 14382@subsection M32R/D Options 14383@cindex M32R/D options 14384 14385These @option{-m} options are defined for Renesas M32R/D architectures: 14386 14387@table @gcctabopt 14388@item -m32r2 14389@opindex m32r2 14390Generate code for the M32R/2@. 14391 14392@item -m32rx 14393@opindex m32rx 14394Generate code for the M32R/X@. 14395 14396@item -m32r 14397@opindex m32r 14398Generate code for the M32R@. This is the default. 14399 14400@item -mmodel=small 14401@opindex mmodel=small 14402Assume all objects live in the lower 16MB of memory (so that their addresses 14403can be loaded with the @code{ld24} instruction), and assume all subroutines 14404are reachable with the @code{bl} instruction. 14405This is the default. 14406 14407The addressability of a particular object can be set with the 14408@code{model} attribute. 14409 14410@item -mmodel=medium 14411@opindex mmodel=medium 14412Assume objects may be anywhere in the 32-bit address space (the compiler 14413will generate @code{seth/add3} instructions to load their addresses), and 14414assume all subroutines are reachable with the @code{bl} instruction. 14415 14416@item -mmodel=large 14417@opindex mmodel=large 14418Assume objects may be anywhere in the 32-bit address space (the compiler 14419will generate @code{seth/add3} instructions to load their addresses), and 14420assume subroutines may not be reachable with the @code{bl} instruction 14421(the compiler will generate the much slower @code{seth/add3/jl} 14422instruction sequence). 14423 14424@item -msdata=none 14425@opindex msdata=none 14426Disable use of the small data area. Variables will be put into 14427one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the 14428@code{section} attribute has been specified). 14429This is the default. 14430 14431The small data area consists of sections @samp{.sdata} and @samp{.sbss}. 14432Objects may be explicitly put in the small data area with the 14433@code{section} attribute using one of these sections. 14434 14435@item -msdata=sdata 14436@opindex msdata=sdata 14437Put small global and static data in the small data area, but do not 14438generate special code to reference them. 14439 14440@item -msdata=use 14441@opindex msdata=use 14442Put small global and static data in the small data area, and generate 14443special instructions to reference them. 14444 14445@item -G @var{num} 14446@opindex G 14447@cindex smaller data references 14448Put global and static objects less than or equal to @var{num} bytes 14449into the small data or bss sections instead of the normal data or bss 14450sections. The default value of @var{num} is 8. 14451The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use} 14452for this option to have any effect. 14453 14454All modules should be compiled with the same @option{-G @var{num}} value. 14455Compiling with different values of @var{num} may or may not work; if it 14456doesn't the linker will give an error message---incorrect code will not be 14457generated. 14458 14459@item -mdebug 14460@opindex mdebug 14461Makes the M32R specific code in the compiler display some statistics 14462that might help in debugging programs. 14463 14464@item -malign-loops 14465@opindex malign-loops 14466Align all loops to a 32-byte boundary. 14467 14468@item -mno-align-loops 14469@opindex mno-align-loops 14470Do not enforce a 32-byte alignment for loops. This is the default. 14471 14472@item -missue-rate=@var{number} 14473@opindex missue-rate=@var{number} 14474Issue @var{number} instructions per cycle. @var{number} can only be 1 14475or 2. 14476 14477@item -mbranch-cost=@var{number} 14478@opindex mbranch-cost=@var{number} 14479@var{number} can only be 1 or 2. If it is 1 then branches will be 14480preferred over conditional code, if it is 2, then the opposite will 14481apply. 14482 14483@item -mflush-trap=@var{number} 14484@opindex mflush-trap=@var{number} 14485Specifies the trap number to use to flush the cache. The default is 1448612. Valid numbers are between 0 and 15 inclusive. 14487 14488@item -mno-flush-trap 14489@opindex mno-flush-trap 14490Specifies that the cache cannot be flushed by using a trap. 14491 14492@item -mflush-func=@var{name} 14493@opindex mflush-func=@var{name} 14494Specifies the name of the operating system function to call to flush 14495the cache. The default is @emph{_flush_cache}, but a function call 14496will only be used if a trap is not available. 14497 14498@item -mno-flush-func 14499@opindex mno-flush-func 14500Indicates that there is no OS function for flushing the cache. 14501 14502@end table 14503 14504@node M680x0 Options 14505@subsection M680x0 Options 14506@cindex M680x0 options 14507 14508These are the @samp{-m} options defined for M680x0 and ColdFire processors. 14509The default settings depend on which architecture was selected when 14510the compiler was configured; the defaults for the most common choices 14511are given below. 14512 14513@table @gcctabopt 14514@item -march=@var{arch} 14515@opindex march 14516Generate code for a specific M680x0 or ColdFire instruction set 14517architecture. Permissible values of @var{arch} for M680x0 14518architectures are: @samp{68000}, @samp{68010}, @samp{68020}, 14519@samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire 14520architectures are selected according to Freescale's ISA classification 14521and the permissible values are: @samp{isaa}, @samp{isaaplus}, 14522@samp{isab} and @samp{isac}. 14523 14524gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating 14525code for a ColdFire target. The @var{arch} in this macro is one of the 14526@option{-march} arguments given above. 14527 14528When used together, @option{-march} and @option{-mtune} select code 14529that runs on a family of similar processors but that is optimized 14530for a particular microarchitecture. 14531 14532@item -mcpu=@var{cpu} 14533@opindex mcpu 14534Generate code for a specific M680x0 or ColdFire processor. 14535The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020}, 14536@samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332} 14537and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table 14538below, which also classifies the CPUs into families: 14539 14540@multitable @columnfractions 0.20 0.80 14541@item @strong{Family} @tab @strong{@samp{-mcpu} arguments} 14542@item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe} 14543@item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206} 14544@item @samp{5206e} @tab @samp{5206e} 14545@item @samp{5208} @tab @samp{5207} @samp{5208} 14546@item @samp{5211a} @tab @samp{5210a} @samp{5211a} 14547@item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213} 14548@item @samp{5216} @tab @samp{5214} @samp{5216} 14549@item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235} 14550@item @samp{5225} @tab @samp{5224} @samp{5225} 14551@item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259} 14552@item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x} 14553@item @samp{5249} @tab @samp{5249} 14554@item @samp{5250} @tab @samp{5250} 14555@item @samp{5271} @tab @samp{5270} @samp{5271} 14556@item @samp{5272} @tab @samp{5272} 14557@item @samp{5275} @tab @samp{5274} @samp{5275} 14558@item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x} 14559@item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017} 14560@item @samp{5307} @tab @samp{5307} 14561@item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x} 14562@item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x} 14563@item @samp{5407} @tab @samp{5407} 14564@item @samp{5475} @tab @samp{5470} @samp{5471} @samp{5472} @samp{5473} @samp{5474} @samp{5475} @samp{547x} @samp{5480} @samp{5481} @samp{5482} @samp{5483} @samp{5484} @samp{5485} 14565@end multitable 14566 14567@option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if 14568@var{arch} is compatible with @var{cpu}. Other combinations of 14569@option{-mcpu} and @option{-march} are rejected. 14570 14571gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target 14572@var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}}, 14573where the value of @var{family} is given by the table above. 14574 14575@item -mtune=@var{tune} 14576@opindex mtune 14577Tune the code for a particular microarchitecture, within the 14578constraints set by @option{-march} and @option{-mcpu}. 14579The M680x0 microarchitectures are: @samp{68000}, @samp{68010}, 14580@samp{68020}, @samp{68030}, @samp{68040}, @samp{68060} 14581and @samp{cpu32}. The ColdFire microarchitectures 14582are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}. 14583 14584You can also use @option{-mtune=68020-40} for code that needs 14585to run relatively well on 68020, 68030 and 68040 targets. 14586@option{-mtune=68020-60} is similar but includes 68060 targets 14587as well. These two options select the same tuning decisions as 14588@option{-m68020-40} and @option{-m68020-60} respectively. 14589 14590gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__} 14591when tuning for 680x0 architecture @var{arch}. It also defines 14592@samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std} 14593option is used. If gcc is tuning for a range of architectures, 14594as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60}, 14595it defines the macros for every architecture in the range. 14596 14597gcc also defines the macro @samp{__m@var{uarch}__} when tuning for 14598ColdFire microarchitecture @var{uarch}, where @var{uarch} is one 14599of the arguments given above. 14600 14601@item -m68000 14602@itemx -mc68000 14603@opindex m68000 14604@opindex mc68000 14605Generate output for a 68000. This is the default 14606when the compiler is configured for 68000-based systems. 14607It is equivalent to @option{-march=68000}. 14608 14609Use this option for microcontrollers with a 68000 or EC000 core, 14610including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356. 14611 14612@item -m68010 14613@opindex m68010 14614Generate output for a 68010. This is the default 14615when the compiler is configured for 68010-based systems. 14616It is equivalent to @option{-march=68010}. 14617 14618@item -m68020 14619@itemx -mc68020 14620@opindex m68020 14621@opindex mc68020 14622Generate output for a 68020. This is the default 14623when the compiler is configured for 68020-based systems. 14624It is equivalent to @option{-march=68020}. 14625 14626@item -m68030 14627@opindex m68030 14628Generate output for a 68030. This is the default when the compiler is 14629configured for 68030-based systems. It is equivalent to 14630@option{-march=68030}. 14631 14632@item -m68040 14633@opindex m68040 14634Generate output for a 68040. This is the default when the compiler is 14635configured for 68040-based systems. It is equivalent to 14636@option{-march=68040}. 14637 14638This option inhibits the use of 68881/68882 instructions that have to be 14639emulated by software on the 68040. Use this option if your 68040 does not 14640have code to emulate those instructions. 14641 14642@item -m68060 14643@opindex m68060 14644Generate output for a 68060. This is the default when the compiler is 14645configured for 68060-based systems. It is equivalent to 14646@option{-march=68060}. 14647 14648This option inhibits the use of 68020 and 68881/68882 instructions that 14649have to be emulated by software on the 68060. Use this option if your 68060 14650does not have code to emulate those instructions. 14651 14652@item -mcpu32 14653@opindex mcpu32 14654Generate output for a CPU32. This is the default 14655when the compiler is configured for CPU32-based systems. 14656It is equivalent to @option{-march=cpu32}. 14657 14658Use this option for microcontrollers with a 14659CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334, 1466068336, 68340, 68341, 68349 and 68360. 14661 14662@item -m5200 14663@opindex m5200 14664Generate output for a 520X ColdFire CPU@. This is the default 14665when the compiler is configured for 520X-based systems. 14666It is equivalent to @option{-mcpu=5206}, and is now deprecated 14667in favor of that option. 14668 14669Use this option for microcontroller with a 5200 core, including 14670the MCF5202, MCF5203, MCF5204 and MCF5206. 14671 14672@item -m5206e 14673@opindex m5206e 14674Generate output for a 5206e ColdFire CPU@. The option is now 14675deprecated in favor of the equivalent @option{-mcpu=5206e}. 14676 14677@item -m528x 14678@opindex m528x 14679Generate output for a member of the ColdFire 528X family. 14680The option is now deprecated in favor of the equivalent 14681@option{-mcpu=528x}. 14682 14683@item -m5307 14684@opindex m5307 14685Generate output for a ColdFire 5307 CPU@. The option is now deprecated 14686in favor of the equivalent @option{-mcpu=5307}. 14687 14688@item -m5407 14689@opindex m5407 14690Generate output for a ColdFire 5407 CPU@. The option is now deprecated 14691in favor of the equivalent @option{-mcpu=5407}. 14692 14693@item -mcfv4e 14694@opindex mcfv4e 14695Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x). 14696This includes use of hardware floating-point instructions. 14697The option is equivalent to @option{-mcpu=547x}, and is now 14698deprecated in favor of that option. 14699 14700@item -m68020-40 14701@opindex m68020-40 14702Generate output for a 68040, without using any of the new instructions. 14703This results in code that can run relatively efficiently on either a 1470468020/68881 or a 68030 or a 68040. The generated code does use the 1470568881 instructions that are emulated on the 68040. 14706 14707The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}. 14708 14709@item -m68020-60 14710@opindex m68020-60 14711Generate output for a 68060, without using any of the new instructions. 14712This results in code that can run relatively efficiently on either a 1471368020/68881 or a 68030 or a 68040. The generated code does use the 1471468881 instructions that are emulated on the 68060. 14715 14716The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}. 14717 14718@item -mhard-float 14719@itemx -m68881 14720@opindex mhard-float 14721@opindex m68881 14722Generate floating-point instructions. This is the default for 68020 14723and above, and for ColdFire devices that have an FPU@. It defines the 14724macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__} 14725on ColdFire targets. 14726 14727@item -msoft-float 14728@opindex msoft-float 14729Do not generate floating-point instructions; use library calls instead. 14730This is the default for 68000, 68010, and 68832 targets. It is also 14731the default for ColdFire devices that have no FPU. 14732 14733@item -mdiv 14734@itemx -mno-div 14735@opindex mdiv 14736@opindex mno-div 14737Generate (do not generate) ColdFire hardware divide and remainder 14738instructions. If @option{-march} is used without @option{-mcpu}, 14739the default is ``on'' for ColdFire architectures and ``off'' for M680x0 14740architectures. Otherwise, the default is taken from the target CPU 14741(either the default CPU, or the one specified by @option{-mcpu}). For 14742example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for 14743@option{-mcpu=5206e}. 14744 14745gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled. 14746 14747@item -mshort 14748@opindex mshort 14749Consider type @code{int} to be 16 bits wide, like @code{short int}. 14750Additionally, parameters passed on the stack are also aligned to a 1475116-bit boundary even on targets whose API mandates promotion to 32-bit. 14752 14753@item -mno-short 14754@opindex mno-short 14755Do not consider type @code{int} to be 16 bits wide. This is the default. 14756 14757@item -mnobitfield 14758@itemx -mno-bitfield 14759@opindex mnobitfield 14760@opindex mno-bitfield 14761Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32} 14762and @option{-m5200} options imply @w{@option{-mnobitfield}}. 14763 14764@item -mbitfield 14765@opindex mbitfield 14766Do use the bit-field instructions. The @option{-m68020} option implies 14767@option{-mbitfield}. This is the default if you use a configuration 14768designed for a 68020. 14769 14770@item -mrtd 14771@opindex mrtd 14772Use a different function-calling convention, in which functions 14773that take a fixed number of arguments return with the @code{rtd} 14774instruction, which pops their arguments while returning. This 14775saves one instruction in the caller since there is no need to pop 14776the arguments there. 14777 14778This calling convention is incompatible with the one normally 14779used on Unix, so you cannot use it if you need to call libraries 14780compiled with the Unix compiler. 14781 14782Also, you must provide function prototypes for all functions that 14783take variable numbers of arguments (including @code{printf}); 14784otherwise incorrect code will be generated for calls to those 14785functions. 14786 14787In addition, seriously incorrect code will result if you call a 14788function with too many arguments. (Normally, extra arguments are 14789harmlessly ignored.) 14790 14791The @code{rtd} instruction is supported by the 68010, 68020, 68030, 1479268040, 68060 and CPU32 processors, but not by the 68000 or 5200. 14793 14794@item -mno-rtd 14795@opindex mno-rtd 14796Do not use the calling conventions selected by @option{-mrtd}. 14797This is the default. 14798 14799@item -malign-int 14800@itemx -mno-align-int 14801@opindex malign-int 14802@opindex mno-align-int 14803Control whether GCC aligns @code{int}, @code{long}, @code{long long}, 14804@code{float}, @code{double}, and @code{long double} variables on a 32-bit 14805boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}). 14806Aligning variables on 32-bit boundaries produces code that runs somewhat 14807faster on processors with 32-bit busses at the expense of more memory. 14808 14809@strong{Warning:} if you use the @option{-malign-int} switch, GCC will 14810align structures containing the above types differently than 14811most published application binary interface specifications for the m68k. 14812 14813@item -mpcrel 14814@opindex mpcrel 14815Use the pc-relative addressing mode of the 68000 directly, instead of 14816using a global offset table. At present, this option implies @option{-fpic}, 14817allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is 14818not presently supported with @option{-mpcrel}, though this could be supported for 1481968020 and higher processors. 14820 14821@item -mno-strict-align 14822@itemx -mstrict-align 14823@opindex mno-strict-align 14824@opindex mstrict-align 14825Do not (do) assume that unaligned memory references will be handled by 14826the system. 14827 14828@item -msep-data 14829Generate code that allows the data segment to be located in a different 14830area of memory from the text segment. This allows for execute in place in 14831an environment without virtual memory management. This option implies 14832@option{-fPIC}. 14833 14834@item -mno-sep-data 14835Generate code that assumes that the data segment follows the text segment. 14836This is the default. 14837 14838@item -mid-shared-library 14839Generate code that supports shared libraries via the library ID method. 14840This allows for execute in place and shared libraries in an environment 14841without virtual memory management. This option implies @option{-fPIC}. 14842 14843@item -mno-id-shared-library 14844Generate code that doesn't assume ID based shared libraries are being used. 14845This is the default. 14846 14847@item -mshared-library-id=n 14848Specified the identification number of the ID based shared library being 14849compiled. Specifying a value of 0 will generate more compact code, specifying 14850other values will force the allocation of that number to the current 14851library but is no more space or time efficient than omitting this option. 14852 14853@item -mxgot 14854@itemx -mno-xgot 14855@opindex mxgot 14856@opindex mno-xgot 14857When generating position-independent code for ColdFire, generate code 14858that works if the GOT has more than 8192 entries. This code is 14859larger and slower than code generated without this option. On M680x0 14860processors, this option is not needed; @option{-fPIC} suffices. 14861 14862GCC normally uses a single instruction to load values from the GOT@. 14863While this is relatively efficient, it only works if the GOT 14864is smaller than about 64k. Anything larger causes the linker 14865to report an error such as: 14866 14867@cindex relocation truncated to fit (ColdFire) 14868@smallexample 14869relocation truncated to fit: R_68K_GOT16O foobar 14870@end smallexample 14871 14872If this happens, you should recompile your code with @option{-mxgot}. 14873It should then work with very large GOTs. However, code generated with 14874@option{-mxgot} is less efficient, since it takes 4 instructions to fetch 14875the value of a global symbol. 14876 14877Note that some linkers, including newer versions of the GNU linker, 14878can create multiple GOTs and sort GOT entries. If you have such a linker, 14879you should only need to use @option{-mxgot} when compiling a single 14880object file that accesses more than 8192 GOT entries. Very few do. 14881 14882These options have no effect unless GCC is generating 14883position-independent code. 14884 14885@end table 14886 14887@node MCore Options 14888@subsection MCore Options 14889@cindex MCore options 14890 14891These are the @samp{-m} options defined for the Motorola M*Core 14892processors. 14893 14894@table @gcctabopt 14895 14896@item -mhardlit 14897@itemx -mno-hardlit 14898@opindex mhardlit 14899@opindex mno-hardlit 14900Inline constants into the code stream if it can be done in two 14901instructions or less. 14902 14903@item -mdiv 14904@itemx -mno-div 14905@opindex mdiv 14906@opindex mno-div 14907Use the divide instruction. (Enabled by default). 14908 14909@item -mrelax-immediate 14910@itemx -mno-relax-immediate 14911@opindex mrelax-immediate 14912@opindex mno-relax-immediate 14913Allow arbitrary sized immediates in bit operations. 14914 14915@item -mwide-bitfields 14916@itemx -mno-wide-bitfields 14917@opindex mwide-bitfields 14918@opindex mno-wide-bitfields 14919Always treat bit-fields as int-sized. 14920 14921@item -m4byte-functions 14922@itemx -mno-4byte-functions 14923@opindex m4byte-functions 14924@opindex mno-4byte-functions 14925Force all functions to be aligned to a 4-byte boundary. 14926 14927@item -mcallgraph-data 14928@itemx -mno-callgraph-data 14929@opindex mcallgraph-data 14930@opindex mno-callgraph-data 14931Emit callgraph information. 14932 14933@item -mslow-bytes 14934@itemx -mno-slow-bytes 14935@opindex mslow-bytes 14936@opindex mno-slow-bytes 14937Prefer word access when reading byte quantities. 14938 14939@item -mlittle-endian 14940@itemx -mbig-endian 14941@opindex mlittle-endian 14942@opindex mbig-endian 14943Generate code for a little-endian target. 14944 14945@item -m210 14946@itemx -m340 14947@opindex m210 14948@opindex m340 14949Generate code for the 210 processor. 14950 14951@item -mno-lsim 14952@opindex mno-lsim 14953Assume that runtime support has been provided and so omit the 14954simulator library (@file{libsim.a)} from the linker command line. 14955 14956@item -mstack-increment=@var{size} 14957@opindex mstack-increment 14958Set the maximum amount for a single stack increment operation. Large 14959values can increase the speed of programs that contain functions 14960that need a large amount of stack space, but they can also trigger a 14961segmentation fault if the stack is extended too much. The default 14962value is 0x1000. 14963 14964@end table 14965 14966@node MeP Options 14967@subsection MeP Options 14968@cindex MeP options 14969 14970@table @gcctabopt 14971 14972@item -mabsdiff 14973@opindex mabsdiff 14974Enables the @code{abs} instruction, which is the absolute difference 14975between two registers. 14976 14977@item -mall-opts 14978@opindex mall-opts 14979Enables all the optional instructions - average, multiply, divide, bit 14980operations, leading zero, absolute difference, min/max, clip, and 14981saturation. 14982 14983 14984@item -maverage 14985@opindex maverage 14986Enables the @code{ave} instruction, which computes the average of two 14987registers. 14988 14989@item -mbased=@var{n} 14990@opindex mbased= 14991Variables of size @var{n} bytes or smaller will be placed in the 14992@code{.based} section by default. Based variables use the @code{$tp} 14993register as a base register, and there is a 128-byte limit to the 14994@code{.based} section. 14995 14996@item -mbitops 14997@opindex mbitops 14998Enables the bit operation instructions - bit test (@code{btstm}), set 14999(@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and 15000test-and-set (@code{tas}). 15001 15002@item -mc=@var{name} 15003@opindex mc= 15004Selects which section constant data will be placed in. @var{name} may 15005be @code{tiny}, @code{near}, or @code{far}. 15006 15007@item -mclip 15008@opindex mclip 15009Enables the @code{clip} instruction. Note that @code{-mclip} is not 15010useful unless you also provide @code{-mminmax}. 15011 15012@item -mconfig=@var{name} 15013@opindex mconfig= 15014Selects one of the build-in core configurations. Each MeP chip has 15015one or more modules in it; each module has a core CPU and a variety of 15016coprocessors, optional instructions, and peripherals. The 15017@code{MeP-Integrator} tool, not part of GCC, provides these 15018configurations through this option; using this option is the same as 15019using all the corresponding command-line options. The default 15020configuration is @code{default}. 15021 15022@item -mcop 15023@opindex mcop 15024Enables the coprocessor instructions. By default, this is a 32-bit 15025coprocessor. Note that the coprocessor is normally enabled via the 15026@code{-mconfig=} option. 15027 15028@item -mcop32 15029@opindex mcop32 15030Enables the 32-bit coprocessor's instructions. 15031 15032@item -mcop64 15033@opindex mcop64 15034Enables the 64-bit coprocessor's instructions. 15035 15036@item -mivc2 15037@opindex mivc2 15038Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor. 15039 15040@item -mdc 15041@opindex mdc 15042Causes constant variables to be placed in the @code{.near} section. 15043 15044@item -mdiv 15045@opindex mdiv 15046Enables the @code{div} and @code{divu} instructions. 15047 15048@item -meb 15049@opindex meb 15050Generate big-endian code. 15051 15052@item -mel 15053@opindex mel 15054Generate little-endian code. 15055 15056@item -mio-volatile 15057@opindex mio-volatile 15058Tells the compiler that any variable marked with the @code{io} 15059attribute is to be considered volatile. 15060 15061@item -ml 15062@opindex ml 15063Causes variables to be assigned to the @code{.far} section by default. 15064 15065@item -mleadz 15066@opindex mleadz 15067Enables the @code{leadz} (leading zero) instruction. 15068 15069@item -mm 15070@opindex mm 15071Causes variables to be assigned to the @code{.near} section by default. 15072 15073@item -mminmax 15074@opindex mminmax 15075Enables the @code{min} and @code{max} instructions. 15076 15077@item -mmult 15078@opindex mmult 15079Enables the multiplication and multiply-accumulate instructions. 15080 15081@item -mno-opts 15082@opindex mno-opts 15083Disables all the optional instructions enabled by @code{-mall-opts}. 15084 15085@item -mrepeat 15086@opindex mrepeat 15087Enables the @code{repeat} and @code{erepeat} instructions, used for 15088low-overhead looping. 15089 15090@item -ms 15091@opindex ms 15092Causes all variables to default to the @code{.tiny} section. Note 15093that there is a 65536-byte limit to this section. Accesses to these 15094variables use the @code{%gp} base register. 15095 15096@item -msatur 15097@opindex msatur 15098Enables the saturation instructions. Note that the compiler does not 15099currently generate these itself, but this option is included for 15100compatibility with other tools, like @code{as}. 15101 15102@item -msdram 15103@opindex msdram 15104Link the SDRAM-based runtime instead of the default ROM-based runtime. 15105 15106@item -msim 15107@opindex msim 15108Link the simulator runtime libraries. 15109 15110@item -msimnovec 15111@opindex msimnovec 15112Link the simulator runtime libraries, excluding built-in support 15113for reset and exception vectors and tables. 15114 15115@item -mtf 15116@opindex mtf 15117Causes all functions to default to the @code{.far} section. Without 15118this option, functions default to the @code{.near} section. 15119 15120@item -mtiny=@var{n} 15121@opindex mtiny= 15122Variables that are @var{n} bytes or smaller will be allocated to the 15123@code{.tiny} section. These variables use the @code{$gp} base 15124register. The default for this option is 4, but note that there's a 1512565536-byte limit to the @code{.tiny} section. 15126 15127@end table 15128 15129@node MicroBlaze Options 15130@subsection MicroBlaze Options 15131@cindex MicroBlaze Options 15132 15133@table @gcctabopt 15134 15135@item -msoft-float 15136@opindex msoft-float 15137Use software emulation for floating point (default). 15138 15139@item -mhard-float 15140@opindex mhard-float 15141Use hardware floating-point instructions. 15142 15143@item -mmemcpy 15144@opindex mmemcpy 15145Do not optimize block moves, use @code{memcpy}. 15146 15147@item -mno-clearbss 15148@opindex mno-clearbss 15149This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead. 15150 15151@item -mcpu=@var{cpu-type} 15152@opindex mcpu= 15153Use features of and schedule code for given CPU. 15154Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}}, 15155where @var{X} is a major version, @var{YY} is the minor version, and 15156@var{Z} is compatibility code. Example values are @samp{v3.00.a}, 15157@samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}. 15158 15159@item -mxl-soft-mul 15160@opindex mxl-soft-mul 15161Use software multiply emulation (default). 15162 15163@item -mxl-soft-div 15164@opindex mxl-soft-div 15165Use software emulation for divides (default). 15166 15167@item -mxl-barrel-shift 15168@opindex mxl-barrel-shift 15169Use the hardware barrel shifter. 15170 15171@item -mxl-pattern-compare 15172@opindex mxl-pattern-compare 15173Use pattern compare instructions. 15174 15175@item -msmall-divides 15176@opindex msmall-divides 15177Use table lookup optimization for small signed integer divisions. 15178 15179@item -mxl-stack-check 15180@opindex mxl-stack-check 15181This option is deprecated. Use -fstack-check instead. 15182 15183@item -mxl-gp-opt 15184@opindex mxl-gp-opt 15185Use GP relative sdata/sbss sections. 15186 15187@item -mxl-multiply-high 15188@opindex mxl-multiply-high 15189Use multiply high instructions for high part of 32x32 multiply. 15190 15191@item -mxl-float-convert 15192@opindex mxl-float-convert 15193Use hardware floating-point conversion instructions. 15194 15195@item -mxl-float-sqrt 15196@opindex mxl-float-sqrt 15197Use hardware floating-point square root instruction. 15198 15199@item -mxl-mode-@var{app-model} 15200Select application model @var{app-model}. Valid models are 15201@table @samp 15202@item executable 15203normal executable (default), uses startup code @file{crt0.o}. 15204 15205@item xmdstub 15206for use with Xilinx Microprocessor Debugger (XMD) based 15207software intrusive debug agent called xmdstub. This uses startup file 15208@file{crt1.o} and sets the start address of the program to be 0x800. 15209 15210@item bootstrap 15211for applications that are loaded using a bootloader. 15212This model uses startup file @file{crt2.o} which does not contain a processor 15213reset vector handler. This is suitable for transferring control on a 15214processor reset to the bootloader rather than the application. 15215 15216@item novectors 15217for applications that do not require any of the 15218MicroBlaze vectors. This option may be useful for applications running 15219within a monitoring application. This model uses @file{crt3.o} as a startup file. 15220@end table 15221 15222Option @option{-xl-mode-@var{app-model}} is a deprecated alias for 15223@option{-mxl-mode-@var{app-model}}. 15224 15225@end table 15226 15227@node MIPS Options 15228@subsection MIPS Options 15229@cindex MIPS options 15230 15231@table @gcctabopt 15232 15233@item -EB 15234@opindex EB 15235Generate big-endian code. 15236 15237@item -EL 15238@opindex EL 15239Generate little-endian code. This is the default for @samp{mips*el-*-*} 15240configurations. 15241 15242@item -march=@var{arch} 15243@opindex march 15244Generate code that will run on @var{arch}, which can be the name of a 15245generic MIPS ISA, or the name of a particular processor. 15246The ISA names are: 15247@samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4}, 15248@samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}. 15249The processor names are: 15250@samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc}, 15251@samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd}, 15252@samp{5kc}, @samp{5kf}, 15253@samp{20kc}, 15254@samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1}, 15255@samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1}, 15256@samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, 15257@samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2}, 15258@samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1}, 15259@samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a}, 15260@samp{m4k}, 15261@samp{octeon}, @samp{octeon+}, @samp{octeon2}, 15262@samp{orion}, 15263@samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400}, 15264@samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, 15265@samp{rm7000}, @samp{rm9000}, 15266@samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000}, 15267@samp{sb1}, 15268@samp{sr71000}, 15269@samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300}, 15270@samp{vr5000}, @samp{vr5400}, @samp{vr5500} 15271and @samp{xlr}. 15272The special value @samp{from-abi} selects the 15273most compatible architecture for the selected ABI (that is, 15274@samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@. 15275 15276Native Linux/GNU and IRIX toolchains also support the value @samp{native}, 15277which selects the best architecture option for the host processor. 15278@option{-march=native} has no effect if GCC does not recognize 15279the processor. 15280 15281In processor names, a final @samp{000} can be abbreviated as @samp{k} 15282(for example, @samp{-march=r2k}). Prefixes are optional, and 15283@samp{vr} may be written @samp{r}. 15284 15285Names of the form @samp{@var{n}f2_1} refer to processors with 15286FPUs clocked at half the rate of the core, names of the form 15287@samp{@var{n}f1_1} refer to processors with FPUs clocked at the same 15288rate as the core, and names of the form @samp{@var{n}f3_2} refer to 15289processors with FPUs clocked a ratio of 3:2 with respect to the core. 15290For compatibility reasons, @samp{@var{n}f} is accepted as a synonym 15291for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are 15292accepted as synonyms for @samp{@var{n}f1_1}. 15293 15294GCC defines two macros based on the value of this option. The first 15295is @samp{_MIPS_ARCH}, which gives the name of target architecture, as 15296a string. The second has the form @samp{_MIPS_ARCH_@var{foo}}, 15297where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@. 15298For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH} 15299to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}. 15300 15301Note that the @samp{_MIPS_ARCH} macro uses the processor names given 15302above. In other words, it will have the full prefix and will not 15303abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi}, 15304the macro names the resolved architecture (either @samp{"mips1"} or 15305@samp{"mips3"}). It names the default architecture when no 15306@option{-march} option is given. 15307 15308@item -mtune=@var{arch} 15309@opindex mtune 15310Optimize for @var{arch}. Among other things, this option controls 15311the way instructions are scheduled, and the perceived cost of arithmetic 15312operations. The list of @var{arch} values is the same as for 15313@option{-march}. 15314 15315When this option is not used, GCC will optimize for the processor 15316specified by @option{-march}. By using @option{-march} and 15317@option{-mtune} together, it is possible to generate code that will 15318run on a family of processors, but optimize the code for one 15319particular member of that family. 15320 15321@samp{-mtune} defines the macros @samp{_MIPS_TUNE} and 15322@samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the 15323@samp{-march} ones described above. 15324 15325@item -mips1 15326@opindex mips1 15327Equivalent to @samp{-march=mips1}. 15328 15329@item -mips2 15330@opindex mips2 15331Equivalent to @samp{-march=mips2}. 15332 15333@item -mips3 15334@opindex mips3 15335Equivalent to @samp{-march=mips3}. 15336 15337@item -mips4 15338@opindex mips4 15339Equivalent to @samp{-march=mips4}. 15340 15341@item -mips32 15342@opindex mips32 15343Equivalent to @samp{-march=mips32}. 15344 15345@item -mips32r2 15346@opindex mips32r2 15347Equivalent to @samp{-march=mips32r2}. 15348 15349@item -mips64 15350@opindex mips64 15351Equivalent to @samp{-march=mips64}. 15352 15353@item -mips64r2 15354@opindex mips64r2 15355Equivalent to @samp{-march=mips64r2}. 15356 15357@item -mips16 15358@itemx -mno-mips16 15359@opindex mips16 15360@opindex mno-mips16 15361Generate (do not generate) MIPS16 code. If GCC is targetting a 15362MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@. 15363 15364MIPS16 code generation can also be controlled on a per-function basis 15365by means of @code{mips16} and @code{nomips16} attributes. 15366@xref{Function Attributes}, for more information. 15367 15368@item -mflip-mips16 15369@opindex mflip-mips16 15370Generate MIPS16 code on alternating functions. This option is provided 15371for regression testing of mixed MIPS16/non-MIPS16 code generation, and is 15372not intended for ordinary use in compiling user code. 15373 15374@item -minterlink-mips16 15375@itemx -mno-interlink-mips16 15376@opindex minterlink-mips16 15377@opindex mno-interlink-mips16 15378Require (do not require) that non-MIPS16 code be link-compatible with 15379MIPS16 code. 15380 15381For example, non-MIPS16 code cannot jump directly to MIPS16 code; 15382it must either use a call or an indirect jump. @option{-minterlink-mips16} 15383therefore disables direct jumps unless GCC knows that the target of the 15384jump is not MIPS16. 15385 15386@item -mabi=32 15387@itemx -mabi=o64 15388@itemx -mabi=n32 15389@itemx -mabi=64 15390@itemx -mabi=eabi 15391@opindex mabi=32 15392@opindex mabi=o64 15393@opindex mabi=n32 15394@opindex mabi=64 15395@opindex mabi=eabi 15396Generate code for the given ABI@. 15397 15398Note that the EABI has a 32-bit and a 64-bit variant. GCC normally 15399generates 64-bit code when you select a 64-bit architecture, but you 15400can use @option{-mgp32} to get 32-bit code instead. 15401 15402For information about the O64 ABI, see 15403@uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}. 15404 15405GCC supports a variant of the o32 ABI in which floating-point registers 15406are 64 rather than 32 bits wide. You can select this combination with 15407@option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1} 15408and @samp{mfhc1} instructions and is therefore only supported for 15409MIPS32R2 processors. 15410 15411The register assignments for arguments and return values remain the 15412same, but each scalar value is passed in a single 64-bit register 15413rather than a pair of 32-bit registers. For example, scalar 15414floating-point values are returned in @samp{$f0} only, not a 15415@samp{$f0}/@samp{$f1} pair. The set of call-saved registers also 15416remains the same, but all 64 bits are saved. 15417 15418@item -mabicalls 15419@itemx -mno-abicalls 15420@opindex mabicalls 15421@opindex mno-abicalls 15422Generate (do not generate) code that is suitable for SVR4-style 15423dynamic objects. @option{-mabicalls} is the default for SVR4-based 15424systems. 15425 15426@item -mshared 15427@itemx -mno-shared 15428Generate (do not generate) code that is fully position-independent, 15429and that can therefore be linked into shared libraries. This option 15430only affects @option{-mabicalls}. 15431 15432All @option{-mabicalls} code has traditionally been position-independent, 15433regardless of options like @option{-fPIC} and @option{-fpic}. However, 15434as an extension, the GNU toolchain allows executables to use absolute 15435accesses for locally-binding symbols. It can also use shorter GP 15436initialization sequences and generate direct calls to locally-defined 15437functions. This mode is selected by @option{-mno-shared}. 15438 15439@option{-mno-shared} depends on binutils 2.16 or higher and generates 15440objects that can only be linked by the GNU linker. However, the option 15441does not affect the ABI of the final executable; it only affects the ABI 15442of relocatable objects. Using @option{-mno-shared} will generally make 15443executables both smaller and quicker. 15444 15445@option{-mshared} is the default. 15446 15447@item -mplt 15448@itemx -mno-plt 15449@opindex mplt 15450@opindex mno-plt 15451Assume (do not assume) that the static and dynamic linkers 15452support PLTs and copy relocations. This option only affects 15453@samp{-mno-shared -mabicalls}. For the n64 ABI, this option 15454has no effect without @samp{-msym32}. 15455 15456You can make @option{-mplt} the default by configuring 15457GCC with @option{--with-mips-plt}. The default is 15458@option{-mno-plt} otherwise. 15459 15460@item -mxgot 15461@itemx -mno-xgot 15462@opindex mxgot 15463@opindex mno-xgot 15464Lift (do not lift) the usual restrictions on the size of the global 15465offset table. 15466 15467GCC normally uses a single instruction to load values from the GOT@. 15468While this is relatively efficient, it will only work if the GOT 15469is smaller than about 64k. Anything larger will cause the linker 15470to report an error such as: 15471 15472@cindex relocation truncated to fit (MIPS) 15473@smallexample 15474relocation truncated to fit: R_MIPS_GOT16 foobar 15475@end smallexample 15476 15477If this happens, you should recompile your code with @option{-mxgot}. 15478It should then work with very large GOTs, although it will also be 15479less efficient, since it will take three instructions to fetch the 15480value of a global symbol. 15481 15482Note that some linkers can create multiple GOTs. If you have such a 15483linker, you should only need to use @option{-mxgot} when a single object 15484file accesses more than 64k's worth of GOT entries. Very few do. 15485 15486These options have no effect unless GCC is generating position 15487independent code. 15488 15489@item -mgp32 15490@opindex mgp32 15491Assume that general-purpose registers are 32 bits wide. 15492 15493@item -mgp64 15494@opindex mgp64 15495Assume that general-purpose registers are 64 bits wide. 15496 15497@item -mfp32 15498@opindex mfp32 15499Assume that floating-point registers are 32 bits wide. 15500 15501@item -mfp64 15502@opindex mfp64 15503Assume that floating-point registers are 64 bits wide. 15504 15505@item -mhard-float 15506@opindex mhard-float 15507Use floating-point coprocessor instructions. 15508 15509@item -msoft-float 15510@opindex msoft-float 15511Do not use floating-point coprocessor instructions. Implement 15512floating-point calculations using library calls instead. 15513 15514@item -msingle-float 15515@opindex msingle-float 15516Assume that the floating-point coprocessor only supports single-precision 15517operations. 15518 15519@item -mdouble-float 15520@opindex mdouble-float 15521Assume that the floating-point coprocessor supports double-precision 15522operations. This is the default. 15523 15524@item -mllsc 15525@itemx -mno-llsc 15526@opindex mllsc 15527@opindex mno-llsc 15528Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to 15529implement atomic memory built-in functions. When neither option is 15530specified, GCC will use the instructions if the target architecture 15531supports them. 15532 15533@option{-mllsc} is useful if the runtime environment can emulate the 15534instructions and @option{-mno-llsc} can be useful when compiling for 15535nonstandard ISAs. You can make either option the default by 15536configuring GCC with @option{--with-llsc} and @option{--without-llsc} 15537respectively. @option{--with-llsc} is the default for some 15538configurations; see the installation documentation for details. 15539 15540@item -mdsp 15541@itemx -mno-dsp 15542@opindex mdsp 15543@opindex mno-dsp 15544Use (do not use) revision 1 of the MIPS DSP ASE@. 15545@xref{MIPS DSP Built-in Functions}. This option defines the 15546preprocessor macro @samp{__mips_dsp}. It also defines 15547@samp{__mips_dsp_rev} to 1. 15548 15549@item -mdspr2 15550@itemx -mno-dspr2 15551@opindex mdspr2 15552@opindex mno-dspr2 15553Use (do not use) revision 2 of the MIPS DSP ASE@. 15554@xref{MIPS DSP Built-in Functions}. This option defines the 15555preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}. 15556It also defines @samp{__mips_dsp_rev} to 2. 15557 15558@item -msmartmips 15559@itemx -mno-smartmips 15560@opindex msmartmips 15561@opindex mno-smartmips 15562Use (do not use) the MIPS SmartMIPS ASE. 15563 15564@item -mpaired-single 15565@itemx -mno-paired-single 15566@opindex mpaired-single 15567@opindex mno-paired-single 15568Use (do not use) paired-single floating-point instructions. 15569@xref{MIPS Paired-Single Support}. This option requires 15570hardware floating-point support to be enabled. 15571 15572@item -mdmx 15573@itemx -mno-mdmx 15574@opindex mdmx 15575@opindex mno-mdmx 15576Use (do not use) MIPS Digital Media Extension instructions. 15577This option can only be used when generating 64-bit code and requires 15578hardware floating-point support to be enabled. 15579 15580@item -mips3d 15581@itemx -mno-mips3d 15582@opindex mips3d 15583@opindex mno-mips3d 15584Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}. 15585The option @option{-mips3d} implies @option{-mpaired-single}. 15586 15587@item -mmt 15588@itemx -mno-mt 15589@opindex mmt 15590@opindex mno-mt 15591Use (do not use) MT Multithreading instructions. 15592 15593@item -mlong64 15594@opindex mlong64 15595Force @code{long} types to be 64 bits wide. See @option{-mlong32} for 15596an explanation of the default and the way that the pointer size is 15597determined. 15598 15599@item -mlong32 15600@opindex mlong32 15601Force @code{long}, @code{int}, and pointer types to be 32 bits wide. 15602 15603The default size of @code{int}s, @code{long}s and pointers depends on 15604the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI 15605uses 64-bit @code{long}s, as does the 64-bit EABI; the others use 1560632-bit @code{long}s. Pointers are the same size as @code{long}s, 15607or the same size as integer registers, whichever is smaller. 15608 15609@item -msym32 15610@itemx -mno-sym32 15611@opindex msym32 15612@opindex mno-sym32 15613Assume (do not assume) that all symbols have 32-bit values, regardless 15614of the selected ABI@. This option is useful in combination with 15615@option{-mabi=64} and @option{-mno-abicalls} because it allows GCC 15616to generate shorter and faster references to symbolic addresses. 15617 15618@item -G @var{num} 15619@opindex G 15620Put definitions of externally-visible data in a small data section 15621if that data is no bigger than @var{num} bytes. GCC can then access 15622the data more efficiently; see @option{-mgpopt} for details. 15623 15624The default @option{-G} option depends on the configuration. 15625 15626@item -mlocal-sdata 15627@itemx -mno-local-sdata 15628@opindex mlocal-sdata 15629@opindex mno-local-sdata 15630Extend (do not extend) the @option{-G} behavior to local data too, 15631such as to static variables in C@. @option{-mlocal-sdata} is the 15632default for all configurations. 15633 15634If the linker complains that an application is using too much small data, 15635you might want to try rebuilding the less performance-critical parts with 15636@option{-mno-local-sdata}. You might also want to build large 15637libraries with @option{-mno-local-sdata}, so that the libraries leave 15638more room for the main program. 15639 15640@item -mextern-sdata 15641@itemx -mno-extern-sdata 15642@opindex mextern-sdata 15643@opindex mno-extern-sdata 15644Assume (do not assume) that externally-defined data will be in 15645a small data section if that data is within the @option{-G} limit. 15646@option{-mextern-sdata} is the default for all configurations. 15647 15648If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G 15649@var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var} 15650that is no bigger than @var{num} bytes, you must make sure that @var{Var} 15651is placed in a small data section. If @var{Var} is defined by another 15652module, you must either compile that module with a high-enough 15653@option{-G} setting or attach a @code{section} attribute to @var{Var}'s 15654definition. If @var{Var} is common, you must link the application 15655with a high-enough @option{-G} setting. 15656 15657The easiest way of satisfying these restrictions is to compile 15658and link every module with the same @option{-G} option. However, 15659you may wish to build a library that supports several different 15660small data limits. You can do this by compiling the library with 15661the highest supported @option{-G} setting and additionally using 15662@option{-mno-extern-sdata} to stop the library from making assumptions 15663about externally-defined data. 15664 15665@item -mgpopt 15666@itemx -mno-gpopt 15667@opindex mgpopt 15668@opindex mno-gpopt 15669Use (do not use) GP-relative accesses for symbols that are known to be 15670in a small data section; see @option{-G}, @option{-mlocal-sdata} and 15671@option{-mextern-sdata}. @option{-mgpopt} is the default for all 15672configurations. 15673 15674@option{-mno-gpopt} is useful for cases where the @code{$gp} register 15675might not hold the value of @code{_gp}. For example, if the code is 15676part of a library that might be used in a boot monitor, programs that 15677call boot monitor routines will pass an unknown value in @code{$gp}. 15678(In such situations, the boot monitor itself would usually be compiled 15679with @option{-G0}.) 15680 15681@option{-mno-gpopt} implies @option{-mno-local-sdata} and 15682@option{-mno-extern-sdata}. 15683 15684@item -membedded-data 15685@itemx -mno-embedded-data 15686@opindex membedded-data 15687@opindex mno-embedded-data 15688Allocate variables to the read-only data section first if possible, then 15689next in the small data section if possible, otherwise in data. This gives 15690slightly slower code than the default, but reduces the amount of RAM required 15691when executing, and thus may be preferred for some embedded systems. 15692 15693@item -muninit-const-in-rodata 15694@itemx -mno-uninit-const-in-rodata 15695@opindex muninit-const-in-rodata 15696@opindex mno-uninit-const-in-rodata 15697Put uninitialized @code{const} variables in the read-only data section. 15698This option is only meaningful in conjunction with @option{-membedded-data}. 15699 15700@item -mcode-readable=@var{setting} 15701@opindex mcode-readable 15702Specify whether GCC may generate code that reads from executable sections. 15703There are three possible settings: 15704 15705@table @gcctabopt 15706@item -mcode-readable=yes 15707Instructions may freely access executable sections. This is the 15708default setting. 15709 15710@item -mcode-readable=pcrel 15711MIPS16 PC-relative load instructions can access executable sections, 15712but other instructions must not do so. This option is useful on 4KSc 15713and 4KSd processors when the code TLBs have the Read Inhibit bit set. 15714It is also useful on processors that can be configured to have a dual 15715instruction/data SRAM interface and that, like the M4K, automatically 15716redirect PC-relative loads to the instruction RAM. 15717 15718@item -mcode-readable=no 15719Instructions must not access executable sections. This option can be 15720useful on targets that are configured to have a dual instruction/data 15721SRAM interface but that (unlike the M4K) do not automatically redirect 15722PC-relative loads to the instruction RAM. 15723@end table 15724 15725@item -msplit-addresses 15726@itemx -mno-split-addresses 15727@opindex msplit-addresses 15728@opindex mno-split-addresses 15729Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler 15730relocation operators. This option has been superseded by 15731@option{-mexplicit-relocs} but is retained for backwards compatibility. 15732 15733@item -mexplicit-relocs 15734@itemx -mno-explicit-relocs 15735@opindex mexplicit-relocs 15736@opindex mno-explicit-relocs 15737Use (do not use) assembler relocation operators when dealing with symbolic 15738addresses. The alternative, selected by @option{-mno-explicit-relocs}, 15739is to use assembler macros instead. 15740 15741@option{-mexplicit-relocs} is the default if GCC was configured 15742to use an assembler that supports relocation operators. 15743 15744@item -mcheck-zero-division 15745@itemx -mno-check-zero-division 15746@opindex mcheck-zero-division 15747@opindex mno-check-zero-division 15748Trap (do not trap) on integer division by zero. 15749 15750The default is @option{-mcheck-zero-division}. 15751 15752@item -mdivide-traps 15753@itemx -mdivide-breaks 15754@opindex mdivide-traps 15755@opindex mdivide-breaks 15756MIPS systems check for division by zero by generating either a 15757conditional trap or a break instruction. Using traps results in 15758smaller code, but is only supported on MIPS II and later. Also, some 15759versions of the Linux kernel have a bug that prevents trap from 15760generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to 15761allow conditional traps on architectures that support them and 15762@option{-mdivide-breaks} to force the use of breaks. 15763 15764The default is usually @option{-mdivide-traps}, but this can be 15765overridden at configure time using @option{--with-divide=breaks}. 15766Divide-by-zero checks can be completely disabled using 15767@option{-mno-check-zero-division}. 15768 15769@item -mmemcpy 15770@itemx -mno-memcpy 15771@opindex mmemcpy 15772@opindex mno-memcpy 15773Force (do not force) the use of @code{memcpy()} for non-trivial block 15774moves. The default is @option{-mno-memcpy}, which allows GCC to inline 15775most constant-sized copies. 15776 15777@item -mlong-calls 15778@itemx -mno-long-calls 15779@opindex mlong-calls 15780@opindex mno-long-calls 15781Disable (do not disable) use of the @code{jal} instruction. Calling 15782functions using @code{jal} is more efficient but requires the caller 15783and callee to be in the same 256 megabyte segment. 15784 15785This option has no effect on abicalls code. The default is 15786@option{-mno-long-calls}. 15787 15788@item -mmad 15789@itemx -mno-mad 15790@opindex mmad 15791@opindex mno-mad 15792Enable (disable) use of the @code{mad}, @code{madu} and @code{mul} 15793instructions, as provided by the R4650 ISA@. 15794 15795@item -mfused-madd 15796@itemx -mno-fused-madd 15797@opindex mfused-madd 15798@opindex mno-fused-madd 15799Enable (disable) use of the floating-point multiply-accumulate 15800instructions, when they are available. The default is 15801@option{-mfused-madd}. 15802 15803When multiply-accumulate instructions are used, the intermediate 15804product is calculated to infinite precision and is not subject to 15805the FCSR Flush to Zero bit. This may be undesirable in some 15806circumstances. 15807 15808@item -nocpp 15809@opindex nocpp 15810Tell the MIPS assembler to not run its preprocessor over user 15811assembler files (with a @samp{.s} suffix) when assembling them. 15812 15813@item -mfix-24k 15814@item -mno-fix-24k 15815@opindex mfix-24k 15816@opindex mno-fix-24k 15817Work around the 24K E48 (lost data on stores during refill) errata. 15818The workarounds are implemented by the assembler rather than by GCC. 15819 15820@item -mfix-r4000 15821@itemx -mno-fix-r4000 15822@opindex mfix-r4000 15823@opindex mno-fix-r4000 15824Work around certain R4000 CPU errata: 15825@itemize @minus 15826@item 15827A double-word or a variable shift may give an incorrect result if executed 15828immediately after starting an integer division. 15829@item 15830A double-word or a variable shift may give an incorrect result if executed 15831while an integer multiplication is in progress. 15832@item 15833An integer division may give an incorrect result if started in a delay slot 15834of a taken branch or a jump. 15835@end itemize 15836 15837@item -mfix-r4400 15838@itemx -mno-fix-r4400 15839@opindex mfix-r4400 15840@opindex mno-fix-r4400 15841Work around certain R4400 CPU errata: 15842@itemize @minus 15843@item 15844A double-word or a variable shift may give an incorrect result if executed 15845immediately after starting an integer division. 15846@end itemize 15847 15848@item -mfix-r10000 15849@itemx -mno-fix-r10000 15850@opindex mfix-r10000 15851@opindex mno-fix-r10000 15852Work around certain R10000 errata: 15853@itemize @minus 15854@item 15855@code{ll}/@code{sc} sequences may not behave atomically on revisions 15856prior to 3.0. They may deadlock on revisions 2.6 and earlier. 15857@end itemize 15858 15859This option can only be used if the target architecture supports 15860branch-likely instructions. @option{-mfix-r10000} is the default when 15861@option{-march=r10000} is used; @option{-mno-fix-r10000} is the default 15862otherwise. 15863 15864@item -mfix-vr4120 15865@itemx -mno-fix-vr4120 15866@opindex mfix-vr4120 15867Work around certain VR4120 errata: 15868@itemize @minus 15869@item 15870@code{dmultu} does not always produce the correct result. 15871@item 15872@code{div} and @code{ddiv} do not always produce the correct result if one 15873of the operands is negative. 15874@end itemize 15875The workarounds for the division errata rely on special functions in 15876@file{libgcc.a}. At present, these functions are only provided by 15877the @code{mips64vr*-elf} configurations. 15878 15879Other VR4120 errata require a nop to be inserted between certain pairs of 15880instructions. These errata are handled by the assembler, not by GCC itself. 15881 15882@item -mfix-vr4130 15883@opindex mfix-vr4130 15884Work around the VR4130 @code{mflo}/@code{mfhi} errata. The 15885workarounds are implemented by the assembler rather than by GCC, 15886although GCC will avoid using @code{mflo} and @code{mfhi} if the 15887VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi} 15888instructions are available instead. 15889 15890@item -mfix-sb1 15891@itemx -mno-fix-sb1 15892@opindex mfix-sb1 15893Work around certain SB-1 CPU core errata. 15894(This flag currently works around the SB-1 revision 2 15895``F1'' and ``F2'' floating-point errata.) 15896 15897@item -mr10k-cache-barrier=@var{setting} 15898@opindex mr10k-cache-barrier 15899Specify whether GCC should insert cache barriers to avoid the 15900side-effects of speculation on R10K processors. 15901 15902In common with many processors, the R10K tries to predict the outcome 15903of a conditional branch and speculatively executes instructions from 15904the ``taken'' branch. It later aborts these instructions if the 15905predicted outcome was wrong. However, on the R10K, even aborted 15906instructions can have side effects. 15907 15908This problem only affects kernel stores and, depending on the system, 15909kernel loads. As an example, a speculatively-executed store may load 15910the target memory into cache and mark the cache line as dirty, even if 15911the store itself is later aborted. If a DMA operation writes to the 15912same area of memory before the ``dirty'' line is flushed, the cached 15913data will overwrite the DMA-ed data. See the R10K processor manual 15914for a full description, including other potential problems. 15915 15916One workaround is to insert cache barrier instructions before every memory 15917access that might be speculatively executed and that might have side 15918effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}} 15919controls GCC's implementation of this workaround. It assumes that 15920aborted accesses to any byte in the following regions will not have 15921side effects: 15922 15923@enumerate 15924@item 15925the memory occupied by the current function's stack frame; 15926 15927@item 15928the memory occupied by an incoming stack argument; 15929 15930@item 15931the memory occupied by an object with a link-time-constant address. 15932@end enumerate 15933 15934It is the kernel's responsibility to ensure that speculative 15935accesses to these regions are indeed safe. 15936 15937If the input program contains a function declaration such as: 15938 15939@smallexample 15940void foo (void); 15941@end smallexample 15942 15943then the implementation of @code{foo} must allow @code{j foo} and 15944@code{jal foo} to be executed speculatively. GCC honors this 15945restriction for functions it compiles itself. It expects non-GCC 15946functions (such as hand-written assembly code) to do the same. 15947 15948The option has three forms: 15949 15950@table @gcctabopt 15951@item -mr10k-cache-barrier=load-store 15952Insert a cache barrier before a load or store that might be 15953speculatively executed and that might have side effects even 15954if aborted. 15955 15956@item -mr10k-cache-barrier=store 15957Insert a cache barrier before a store that might be speculatively 15958executed and that might have side effects even if aborted. 15959 15960@item -mr10k-cache-barrier=none 15961Disable the insertion of cache barriers. This is the default setting. 15962@end table 15963 15964@item -mflush-func=@var{func} 15965@itemx -mno-flush-func 15966@opindex mflush-func 15967Specifies the function to call to flush the I and D caches, or to not 15968call any such function. If called, the function must take the same 15969arguments as the common @code{_flush_func()}, that is, the address of the 15970memory range for which the cache is being flushed, the size of the 15971memory range, and the number 3 (to flush both caches). The default 15972depends on the target GCC was configured for, but commonly is either 15973@samp{_flush_func} or @samp{__cpu_flush}. 15974 15975@item mbranch-cost=@var{num} 15976@opindex mbranch-cost 15977Set the cost of branches to roughly @var{num} ``simple'' instructions. 15978This cost is only a heuristic and is not guaranteed to produce 15979consistent results across releases. A zero cost redundantly selects 15980the default, which is based on the @option{-mtune} setting. 15981 15982@item -mbranch-likely 15983@itemx -mno-branch-likely 15984@opindex mbranch-likely 15985@opindex mno-branch-likely 15986Enable or disable use of Branch Likely instructions, regardless of the 15987default for the selected architecture. By default, Branch Likely 15988instructions may be generated if they are supported by the selected 15989architecture. An exception is for the MIPS32 and MIPS64 architectures 15990and processors that implement those architectures; for those, Branch 15991Likely instructions will not be generated by default because the MIPS32 15992and MIPS64 architectures specifically deprecate their use. 15993 15994@item -mfp-exceptions 15995@itemx -mno-fp-exceptions 15996@opindex mfp-exceptions 15997Specifies whether FP exceptions are enabled. This affects how we schedule 15998FP instructions for some processors. The default is that FP exceptions are 15999enabled. 16000 16001For instance, on the SB-1, if FP exceptions are disabled, and we are emitting 1600264-bit code, then we can use both FP pipes. Otherwise, we can only use one 16003FP pipe. 16004 16005@item -mvr4130-align 16006@itemx -mno-vr4130-align 16007@opindex mvr4130-align 16008The VR4130 pipeline is two-way superscalar, but can only issue two 16009instructions together if the first one is 8-byte aligned. When this 16010option is enabled, GCC will align pairs of instructions that it 16011thinks should execute in parallel. 16012 16013This option only has an effect when optimizing for the VR4130. 16014It normally makes code faster, but at the expense of making it bigger. 16015It is enabled by default at optimization level @option{-O3}. 16016 16017@item -msynci 16018@itemx -mno-synci 16019@opindex msynci 16020Enable (disable) generation of @code{synci} instructions on 16021architectures that support it. The @code{synci} instructions (if 16022enabled) will be generated when @code{__builtin___clear_cache()} is 16023compiled. 16024 16025This option defaults to @code{-mno-synci}, but the default can be 16026overridden by configuring with @code{--with-synci}. 16027 16028When compiling code for single processor systems, it is generally safe 16029to use @code{synci}. However, on many multi-core (SMP) systems, it 16030will not invalidate the instruction caches on all cores and may lead 16031to undefined behavior. 16032 16033@item -mrelax-pic-calls 16034@itemx -mno-relax-pic-calls 16035@opindex mrelax-pic-calls 16036Try to turn PIC calls that are normally dispatched via register 16037@code{$25} into direct calls. This is only possible if the linker can 16038resolve the destination at link-time and if the destination is within 16039range for a direct call. 16040 16041@option{-mrelax-pic-calls} is the default if GCC was configured to use 16042an assembler and a linker that supports the @code{.reloc} assembly 16043directive and @code{-mexplicit-relocs} is in effect. With 16044@code{-mno-explicit-relocs}, this optimization can be performed by the 16045assembler and the linker alone without help from the compiler. 16046 16047@item -mmcount-ra-address 16048@itemx -mno-mcount-ra-address 16049@opindex mmcount-ra-address 16050@opindex mno-mcount-ra-address 16051Emit (do not emit) code that allows @code{_mcount} to modify the 16052calling function's return address. When enabled, this option extends 16053the usual @code{_mcount} interface with a new @var{ra-address} 16054parameter, which has type @code{intptr_t *} and is passed in register 16055@code{$12}. @code{_mcount} can then modify the return address by 16056doing both of the following: 16057@itemize 16058@item 16059Returning the new address in register @code{$31}. 16060@item 16061Storing the new address in @code{*@var{ra-address}}, 16062if @var{ra-address} is nonnull. 16063@end itemize 16064 16065The default is @option{-mno-mcount-ra-address}. 16066 16067@end table 16068 16069@node MMIX Options 16070@subsection MMIX Options 16071@cindex MMIX Options 16072 16073These options are defined for the MMIX: 16074 16075@table @gcctabopt 16076@item -mlibfuncs 16077@itemx -mno-libfuncs 16078@opindex mlibfuncs 16079@opindex mno-libfuncs 16080Specify that intrinsic library functions are being compiled, passing all 16081values in registers, no matter the size. 16082 16083@item -mepsilon 16084@itemx -mno-epsilon 16085@opindex mepsilon 16086@opindex mno-epsilon 16087Generate floating-point comparison instructions that compare with respect 16088to the @code{rE} epsilon register. 16089 16090@item -mabi=mmixware 16091@itemx -mabi=gnu 16092@opindex mabi=mmixware 16093@opindex mabi=gnu 16094Generate code that passes function parameters and return values that (in 16095the called function) are seen as registers @code{$0} and up, as opposed to 16096the GNU ABI which uses global registers @code{$231} and up. 16097 16098@item -mzero-extend 16099@itemx -mno-zero-extend 16100@opindex mzero-extend 16101@opindex mno-zero-extend 16102When reading data from memory in sizes shorter than 64 bits, use (do not 16103use) zero-extending load instructions by default, rather than 16104sign-extending ones. 16105 16106@item -mknuthdiv 16107@itemx -mno-knuthdiv 16108@opindex mknuthdiv 16109@opindex mno-knuthdiv 16110Make the result of a division yielding a remainder have the same sign as 16111the divisor. With the default, @option{-mno-knuthdiv}, the sign of the 16112remainder follows the sign of the dividend. Both methods are 16113arithmetically valid, the latter being almost exclusively used. 16114 16115@item -mtoplevel-symbols 16116@itemx -mno-toplevel-symbols 16117@opindex mtoplevel-symbols 16118@opindex mno-toplevel-symbols 16119Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly 16120code can be used with the @code{PREFIX} assembly directive. 16121 16122@item -melf 16123@opindex melf 16124Generate an executable in the ELF format, rather than the default 16125@samp{mmo} format used by the @command{mmix} simulator. 16126 16127@item -mbranch-predict 16128@itemx -mno-branch-predict 16129@opindex mbranch-predict 16130@opindex mno-branch-predict 16131Use (do not use) the probable-branch instructions, when static branch 16132prediction indicates a probable branch. 16133 16134@item -mbase-addresses 16135@itemx -mno-base-addresses 16136@opindex mbase-addresses 16137@opindex mno-base-addresses 16138Generate (do not generate) code that uses @emph{base addresses}. Using a 16139base address automatically generates a request (handled by the assembler 16140and the linker) for a constant to be set up in a global register. The 16141register is used for one or more base address requests within the range 0 16142to 255 from the value held in the register. The generally leads to short 16143and fast code, but the number of different data items that can be 16144addressed is limited. This means that a program that uses lots of static 16145data may require @option{-mno-base-addresses}. 16146 16147@item -msingle-exit 16148@itemx -mno-single-exit 16149@opindex msingle-exit 16150@opindex mno-single-exit 16151Force (do not force) generated code to have a single exit point in each 16152function. 16153@end table 16154 16155@node MN10300 Options 16156@subsection MN10300 Options 16157@cindex MN10300 options 16158 16159These @option{-m} options are defined for Matsushita MN10300 architectures: 16160 16161@table @gcctabopt 16162@item -mmult-bug 16163@opindex mmult-bug 16164Generate code to avoid bugs in the multiply instructions for the MN10300 16165processors. This is the default. 16166 16167@item -mno-mult-bug 16168@opindex mno-mult-bug 16169Do not generate code to avoid bugs in the multiply instructions for the 16170MN10300 processors. 16171 16172@item -mam33 16173@opindex mam33 16174Generate code using features specific to the AM33 processor. 16175 16176@item -mno-am33 16177@opindex mno-am33 16178Do not generate code using features specific to the AM33 processor. This 16179is the default. 16180 16181@item -mam33-2 16182@opindex mam33-2 16183Generate code using features specific to the AM33/2.0 processor. 16184 16185@item -mam34 16186@opindex mam34 16187Generate code using features specific to the AM34 processor. 16188 16189@item -mtune=@var{cpu-type} 16190@opindex mtune 16191Use the timing characteristics of the indicated CPU type when 16192scheduling instructions. This does not change the targeted processor 16193type. The CPU type must be one of @samp{mn10300}, @samp{am33}, 16194@samp{am33-2} or @samp{am34}. 16195 16196@item -mreturn-pointer-on-d0 16197@opindex mreturn-pointer-on-d0 16198When generating a function that returns a pointer, return the pointer 16199in both @code{a0} and @code{d0}. Otherwise, the pointer is returned 16200only in a0, and attempts to call such functions without a prototype 16201would result in errors. Note that this option is on by default; use 16202@option{-mno-return-pointer-on-d0} to disable it. 16203 16204@item -mno-crt0 16205@opindex mno-crt0 16206Do not link in the C run-time initialization object file. 16207 16208@item -mrelax 16209@opindex mrelax 16210Indicate to the linker that it should perform a relaxation optimization pass 16211to shorten branches, calls and absolute memory addresses. This option only 16212has an effect when used on the command line for the final link step. 16213 16214This option makes symbolic debugging impossible. 16215 16216@item -mliw 16217@opindex mliw 16218Allow the compiler to generate @emph{Long Instruction Word} 16219instructions if the target is the @samp{AM33} or later. This is the 16220default. This option defines the preprocessor macro @samp{__LIW__}. 16221 16222@item -mnoliw 16223@opindex mnoliw 16224Do not allow the compiler to generate @emph{Long Instruction Word} 16225instructions. This option defines the preprocessor macro 16226@samp{__NO_LIW__}. 16227 16228@item -msetlb 16229@opindex msetlb 16230Allow the compiler to generate the @emph{SETLB} and @emph{Lcc} 16231instructions if the target is the @samp{AM33} or later. This is the 16232default. This option defines the preprocessor macro @samp{__SETLB__}. 16233 16234@item -mnosetlb 16235@opindex mnosetlb 16236Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc} 16237instructions. This option defines the preprocessor macro 16238@samp{__NO_SETLB__}. 16239 16240@end table 16241 16242@node PDP-11 Options 16243@subsection PDP-11 Options 16244@cindex PDP-11 Options 16245 16246These options are defined for the PDP-11: 16247 16248@table @gcctabopt 16249@item -mfpu 16250@opindex mfpu 16251Use hardware FPP floating point. This is the default. (FIS floating 16252point on the PDP-11/40 is not supported.) 16253 16254@item -msoft-float 16255@opindex msoft-float 16256Do not use hardware floating point. 16257 16258@item -mac0 16259@opindex mac0 16260Return floating-point results in ac0 (fr0 in Unix assembler syntax). 16261 16262@item -mno-ac0 16263@opindex mno-ac0 16264Return floating-point results in memory. This is the default. 16265 16266@item -m40 16267@opindex m40 16268Generate code for a PDP-11/40. 16269 16270@item -m45 16271@opindex m45 16272Generate code for a PDP-11/45. This is the default. 16273 16274@item -m10 16275@opindex m10 16276Generate code for a PDP-11/10. 16277 16278@item -mbcopy-builtin 16279@opindex mbcopy-builtin 16280Use inline @code{movmemhi} patterns for copying memory. This is the 16281default. 16282 16283@item -mbcopy 16284@opindex mbcopy 16285Do not use inline @code{movmemhi} patterns for copying memory. 16286 16287@item -mint16 16288@itemx -mno-int32 16289@opindex mint16 16290@opindex mno-int32 16291Use 16-bit @code{int}. This is the default. 16292 16293@item -mint32 16294@itemx -mno-int16 16295@opindex mint32 16296@opindex mno-int16 16297Use 32-bit @code{int}. 16298 16299@item -mfloat64 16300@itemx -mno-float32 16301@opindex mfloat64 16302@opindex mno-float32 16303Use 64-bit @code{float}. This is the default. 16304 16305@item -mfloat32 16306@itemx -mno-float64 16307@opindex mfloat32 16308@opindex mno-float64 16309Use 32-bit @code{float}. 16310 16311@item -mabshi 16312@opindex mabshi 16313Use @code{abshi2} pattern. This is the default. 16314 16315@item -mno-abshi 16316@opindex mno-abshi 16317Do not use @code{abshi2} pattern. 16318 16319@item -mbranch-expensive 16320@opindex mbranch-expensive 16321Pretend that branches are expensive. This is for experimenting with 16322code generation only. 16323 16324@item -mbranch-cheap 16325@opindex mbranch-cheap 16326Do not pretend that branches are expensive. This is the default. 16327 16328@item -munix-asm 16329@opindex munix-asm 16330Use Unix assembler syntax. This is the default when configured for 16331@samp{pdp11-*-bsd}. 16332 16333@item -mdec-asm 16334@opindex mdec-asm 16335Use DEC assembler syntax. This is the default when configured for any 16336PDP-11 target other than @samp{pdp11-*-bsd}. 16337@end table 16338 16339@node picoChip Options 16340@subsection picoChip Options 16341@cindex picoChip options 16342 16343These @samp{-m} options are defined for picoChip implementations: 16344 16345@table @gcctabopt 16346 16347@item -mae=@var{ae_type} 16348@opindex mcpu 16349Set the instruction set, register set, and instruction scheduling 16350parameters for array element type @var{ae_type}. Supported values 16351for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}. 16352 16353@option{-mae=ANY} selects a completely generic AE type. Code 16354generated with this option will run on any of the other AE types. The 16355code will not be as efficient as it would be if compiled for a specific 16356AE type, and some types of operation (e.g., multiplication) will not 16357work properly on all types of AE. 16358 16359@option{-mae=MUL} selects a MUL AE type. This is the most useful AE type 16360for compiled code, and is the default. 16361 16362@option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this 16363option may suffer from poor performance of byte (char) manipulation, 16364since the DSP AE does not provide hardware support for byte load/stores. 16365 16366@item -msymbol-as-address 16367Enable the compiler to directly use a symbol name as an address in a 16368load/store instruction, without first loading it into a 16369register. Typically, the use of this option will generate larger 16370programs, which run faster than when the option isn't used. However, the 16371results vary from program to program, so it is left as a user option, 16372rather than being permanently enabled. 16373 16374@item -mno-inefficient-warnings 16375Disables warnings about the generation of inefficient code. These 16376warnings can be generated, for example, when compiling code that 16377performs byte-level memory operations on the MAC AE type. The MAC AE has 16378no hardware support for byte-level memory operations, so all byte 16379load/stores must be synthesized from word load/store operations. This is 16380inefficient and a warning will be generated indicating to the programmer 16381that they should rewrite the code to avoid byte operations, or to target 16382an AE type that has the necessary hardware support. This option enables 16383the warning to be turned off. 16384 16385@end table 16386 16387@node PowerPC Options 16388@subsection PowerPC Options 16389@cindex PowerPC options 16390 16391These are listed under @xref{RS/6000 and PowerPC Options}. 16392 16393@node RL78 Options 16394@subsection RL78 Options 16395@cindex RL78 Options 16396 16397@table @gcctabopt 16398 16399@item -msim 16400@opindex msim 16401Links in additional target libraries to support operation within a 16402simulator. 16403 16404@item -mmul=none 16405@itemx -mmul=g13 16406@itemx -mmul=rl78 16407@opindex mmul 16408Specifies the type of hardware multiplication support to be used. The 16409default is @code{none}, which uses software multiplication functions. 16410The @code{g13} option is for the hardware multiply/divide peripheral 16411only on the RL78/G13 targets. The @code{rl78} option is for the 16412standard hardware multiplication defined in the RL78 software manual. 16413 16414@end table 16415 16416@node RS/6000 and PowerPC Options 16417@subsection IBM RS/6000 and PowerPC Options 16418@cindex RS/6000 and PowerPC Options 16419@cindex IBM RS/6000 and PowerPC Options 16420 16421These @samp{-m} options are defined for the IBM RS/6000 and PowerPC: 16422@table @gcctabopt 16423@item -mpower 16424@itemx -mno-power 16425@itemx -mpower2 16426@itemx -mno-power2 16427@itemx -mpowerpc 16428@itemx -mno-powerpc 16429@itemx -mpowerpc-gpopt 16430@itemx -mno-powerpc-gpopt 16431@itemx -mpowerpc-gfxopt 16432@itemx -mno-powerpc-gfxopt 16433@need 800 16434@itemx -mpowerpc64 16435@itemx -mno-powerpc64 16436@itemx -mmfcrf 16437@itemx -mno-mfcrf 16438@itemx -mpopcntb 16439@itemx -mno-popcntb 16440@itemx -mpopcntd 16441@itemx -mno-popcntd 16442@itemx -mfprnd 16443@itemx -mno-fprnd 16444@need 800 16445@itemx -mcmpb 16446@itemx -mno-cmpb 16447@itemx -mmfpgpr 16448@itemx -mno-mfpgpr 16449@itemx -mhard-dfp 16450@itemx -mno-hard-dfp 16451@opindex mpower 16452@opindex mno-power 16453@opindex mpower2 16454@opindex mno-power2 16455@opindex mpowerpc 16456@opindex mno-powerpc 16457@opindex mpowerpc-gpopt 16458@opindex mno-powerpc-gpopt 16459@opindex mpowerpc-gfxopt 16460@opindex mno-powerpc-gfxopt 16461@opindex mpowerpc64 16462@opindex mno-powerpc64 16463@opindex mmfcrf 16464@opindex mno-mfcrf 16465@opindex mpopcntb 16466@opindex mno-popcntb 16467@opindex mpopcntd 16468@opindex mno-popcntd 16469@opindex mfprnd 16470@opindex mno-fprnd 16471@opindex mcmpb 16472@opindex mno-cmpb 16473@opindex mmfpgpr 16474@opindex mno-mfpgpr 16475@opindex mhard-dfp 16476@opindex mno-hard-dfp 16477GCC supports two related instruction set architectures for the 16478RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those 16479instructions supported by the @samp{rios} chip set used in the original 16480RS/6000 systems and the @dfn{PowerPC} instruction set is the 16481architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and 16482the IBM 4xx, 6xx, and follow-on microprocessors. 16483 16484Neither architecture is a subset of the other. However there is a 16485large common subset of instructions supported by both. An MQ 16486register is included in processors supporting the POWER architecture. 16487 16488You use these options to specify which instructions are available on the 16489processor you are using. The default value of these options is 16490determined when configuring GCC@. Specifying the 16491@option{-mcpu=@var{cpu_type}} overrides the specification of these 16492options. We recommend you use the @option{-mcpu=@var{cpu_type}} option 16493rather than the options listed above. 16494 16495The @option{-mpower} option allows GCC to generate instructions that 16496are found only in the POWER architecture and to use the MQ register. 16497Specifying @option{-mpower2} implies @option{-power} and also allows GCC 16498to generate instructions that are present in the POWER2 architecture but 16499not the original POWER architecture. 16500 16501The @option{-mpowerpc} option allows GCC to generate instructions that 16502are found only in the 32-bit subset of the PowerPC architecture. 16503Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows 16504GCC to use the optional PowerPC architecture instructions in the 16505General Purpose group, including floating-point square root. Specifying 16506@option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to 16507use the optional PowerPC architecture instructions in the Graphics 16508group, including floating-point select. 16509 16510The @option{-mmfcrf} option allows GCC to generate the move from 16511condition register field instruction implemented on the POWER4 16512processor and other processors that support the PowerPC V2.01 16513architecture. 16514The @option{-mpopcntb} option allows GCC to generate the popcount and 16515double-precision FP reciprocal estimate instruction implemented on the 16516POWER5 processor and other processors that support the PowerPC V2.02 16517architecture. 16518The @option{-mpopcntd} option allows GCC to generate the popcount 16519instruction implemented on the POWER7 processor and other processors 16520that support the PowerPC V2.06 architecture. 16521The @option{-mfprnd} option allows GCC to generate the FP round to 16522integer instructions implemented on the POWER5+ processor and other 16523processors that support the PowerPC V2.03 architecture. 16524The @option{-mcmpb} option allows GCC to generate the compare bytes 16525instruction implemented on the POWER6 processor and other processors 16526that support the PowerPC V2.05 architecture. 16527The @option{-mmfpgpr} option allows GCC to generate the FP move to/from 16528general-purpose register instructions implemented on the POWER6X 16529processor and other processors that support the extended PowerPC V2.05 16530architecture. 16531The @option{-mhard-dfp} option allows GCC to generate the decimal 16532floating-point instructions implemented on some POWER processors. 16533 16534The @option{-mpowerpc64} option allows GCC to generate the additional 1653564-bit instructions that are found in the full PowerPC64 architecture 16536and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to 16537@option{-mno-powerpc64}. 16538 16539If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC 16540will use only the instructions in the common subset of both 16541architectures plus some special AIX common-mode calls, and will not use 16542the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc} 16543permits GCC to use any instruction from either architecture and to 16544allow use of the MQ register; specify this for the Motorola MPC601. 16545 16546@item -mnew-mnemonics 16547@itemx -mold-mnemonics 16548@opindex mnew-mnemonics 16549@opindex mold-mnemonics 16550Select which mnemonics to use in the generated assembler code. With 16551@option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for 16552the PowerPC architecture. With @option{-mold-mnemonics} it uses the 16553assembler mnemonics defined for the POWER architecture. Instructions 16554defined in only one architecture have only one mnemonic; GCC uses that 16555mnemonic irrespective of which of these options is specified. 16556 16557GCC defaults to the mnemonics appropriate for the architecture in 16558use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the 16559value of these option. Unless you are building a cross-compiler, you 16560should normally not specify either @option{-mnew-mnemonics} or 16561@option{-mold-mnemonics}, but should instead accept the default. 16562 16563@item -mcpu=@var{cpu_type} 16564@opindex mcpu 16565Set architecture type, register usage, choice of mnemonics, and 16566instruction scheduling parameters for machine type @var{cpu_type}. 16567Supported values for @var{cpu_type} are @samp{401}, @samp{403}, 16568@samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp}, 16569@samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603}, 16570@samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740}, 16571@samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823}, 16572@samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2}, 16573@samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3}, 16574@samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3}, 16575@samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, 16576@samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios}, 16577@samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}. 16578 16579@option{-mcpu=common} selects a completely generic processor. Code 16580generated under this option will run on any POWER or PowerPC processor. 16581GCC will use only the instructions in the common subset of both 16582architectures, and will not use the MQ register. GCC assumes a generic 16583processor model for scheduling purposes. 16584 16585@option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and 16586@option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit 16587PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine 16588types, with an appropriate, generic processor model assumed for 16589scheduling purposes. 16590 16591The other options specify a specific processor. Code generated under 16592those options will run best on that processor, and may not run at all on 16593others. 16594 16595The @option{-mcpu} options automatically enable or disable the 16596following options: 16597 16598@gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol 16599-mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol 16600-mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol 16601-msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx} 16602 16603The particular options set for any particular CPU will vary between 16604compiler versions, depending on what setting seems to produce optimal 16605code for that CPU; it doesn't necessarily reflect the actual hardware's 16606capabilities. If you wish to set an individual option to a particular 16607value, you may specify it after the @option{-mcpu} option, like 16608@samp{-mcpu=970 -mno-altivec}. 16609 16610On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are 16611not enabled or disabled by the @option{-mcpu} option at present because 16612AIX does not have full support for these options. You may still 16613enable or disable them individually if you're sure it'll work in your 16614environment. 16615 16616@item -mtune=@var{cpu_type} 16617@opindex mtune 16618Set the instruction scheduling parameters for machine type 16619@var{cpu_type}, but do not set the architecture type, register usage, or 16620choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same 16621values for @var{cpu_type} are used for @option{-mtune} as for 16622@option{-mcpu}. If both are specified, the code generated will use the 16623architecture, registers, and mnemonics set by @option{-mcpu}, but the 16624scheduling parameters set by @option{-mtune}. 16625 16626@item -mcmodel=small 16627@opindex mcmodel=small 16628Generate PowerPC64 code for the small model: The TOC is limited to 1662964k. 16630 16631@item -mcmodel=medium 16632@opindex mcmodel=medium 16633Generate PowerPC64 code for the medium model: The TOC and other static 16634data may be up to a total of 4G in size. 16635 16636@item -mcmodel=large 16637@opindex mcmodel=large 16638Generate PowerPC64 code for the large model: The TOC may be up to 4G 16639in size. Other data and code is only limited by the 64-bit address 16640space. 16641 16642@item -maltivec 16643@itemx -mno-altivec 16644@opindex maltivec 16645@opindex mno-altivec 16646Generate code that uses (does not use) AltiVec instructions, and also 16647enable the use of built-in functions that allow more direct access to 16648the AltiVec instruction set. You may also need to set 16649@option{-mabi=altivec} to adjust the current ABI with AltiVec ABI 16650enhancements. 16651 16652@item -mvrsave 16653@itemx -mno-vrsave 16654@opindex mvrsave 16655@opindex mno-vrsave 16656Generate VRSAVE instructions when generating AltiVec code. 16657 16658@item -mgen-cell-microcode 16659@opindex mgen-cell-microcode 16660Generate Cell microcode instructions 16661 16662@item -mwarn-cell-microcode 16663@opindex mwarn-cell-microcode 16664Warning when a Cell microcode instruction is going to emitted. An example 16665of a Cell microcode instruction is a variable shift. 16666 16667@item -msecure-plt 16668@opindex msecure-plt 16669Generate code that allows ld and ld.so to build executables and shared 16670libraries with non-exec .plt and .got sections. This is a PowerPC 1667132-bit SYSV ABI option. 16672 16673@item -mbss-plt 16674@opindex mbss-plt 16675Generate code that uses a BSS .plt section that ld.so fills in, and 16676requires .plt and .got sections that are both writable and executable. 16677This is a PowerPC 32-bit SYSV ABI option. 16678 16679@item -misel 16680@itemx -mno-isel 16681@opindex misel 16682@opindex mno-isel 16683This switch enables or disables the generation of ISEL instructions. 16684 16685@item -misel=@var{yes/no} 16686This switch has been deprecated. Use @option{-misel} and 16687@option{-mno-isel} instead. 16688 16689@item -mspe 16690@itemx -mno-spe 16691@opindex mspe 16692@opindex mno-spe 16693This switch enables or disables the generation of SPE simd 16694instructions. 16695 16696@item -mpaired 16697@itemx -mno-paired 16698@opindex mpaired 16699@opindex mno-paired 16700This switch enables or disables the generation of PAIRED simd 16701instructions. 16702 16703@item -mspe=@var{yes/no} 16704This option has been deprecated. Use @option{-mspe} and 16705@option{-mno-spe} instead. 16706 16707@item -mvsx 16708@itemx -mno-vsx 16709@opindex mvsx 16710@opindex mno-vsx 16711Generate code that uses (does not use) vector/scalar (VSX) 16712instructions, and also enable the use of built-in functions that allow 16713more direct access to the VSX instruction set. 16714 16715@item -mfloat-gprs=@var{yes/single/double/no} 16716@itemx -mfloat-gprs 16717@opindex mfloat-gprs 16718This switch enables or disables the generation of floating-point 16719operations on the general-purpose registers for architectures that 16720support it. 16721 16722The argument @var{yes} or @var{single} enables the use of 16723single-precision floating-point operations. 16724 16725The argument @var{double} enables the use of single and 16726double-precision floating-point operations. 16727 16728The argument @var{no} disables floating-point operations on the 16729general-purpose registers. 16730 16731This option is currently only available on the MPC854x. 16732 16733@item -m32 16734@itemx -m64 16735@opindex m32 16736@opindex m64 16737Generate code for 32-bit or 64-bit environments of Darwin and SVR4 16738targets (including GNU/Linux). The 32-bit environment sets int, long 16739and pointer to 32 bits and generates code that runs on any PowerPC 16740variant. The 64-bit environment sets int to 32 bits and long and 16741pointer to 64 bits, and generates code for PowerPC64, as for 16742@option{-mpowerpc64}. 16743 16744@item -mfull-toc 16745@itemx -mno-fp-in-toc 16746@itemx -mno-sum-in-toc 16747@itemx -mminimal-toc 16748@opindex mfull-toc 16749@opindex mno-fp-in-toc 16750@opindex mno-sum-in-toc 16751@opindex mminimal-toc 16752Modify generation of the TOC (Table Of Contents), which is created for 16753every executable file. The @option{-mfull-toc} option is selected by 16754default. In that case, GCC will allocate at least one TOC entry for 16755each unique non-automatic variable reference in your program. GCC 16756will also place floating-point constants in the TOC@. However, only 1675716,384 entries are available in the TOC@. 16758 16759If you receive a linker error message that saying you have overflowed 16760the available TOC space, you can reduce the amount of TOC space used 16761with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options. 16762@option{-mno-fp-in-toc} prevents GCC from putting floating-point 16763constants in the TOC and @option{-mno-sum-in-toc} forces GCC to 16764generate code to calculate the sum of an address and a constant at 16765run time instead of putting that sum into the TOC@. You may specify one 16766or both of these options. Each causes GCC to produce very slightly 16767slower and larger code at the expense of conserving TOC space. 16768 16769If you still run out of space in the TOC even when you specify both of 16770these options, specify @option{-mminimal-toc} instead. This option causes 16771GCC to make only one TOC entry for every file. When you specify this 16772option, GCC will produce code that is slower and larger but which 16773uses extremely little TOC space. You may wish to use this option 16774only on files that contain less frequently executed code. 16775 16776@item -maix64 16777@itemx -maix32 16778@opindex maix64 16779@opindex maix32 16780Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit 16781@code{long} type, and the infrastructure needed to support them. 16782Specifying @option{-maix64} implies @option{-mpowerpc64} and 16783@option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and 16784implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}. 16785 16786@item -mxl-compat 16787@itemx -mno-xl-compat 16788@opindex mxl-compat 16789@opindex mno-xl-compat 16790Produce code that conforms more closely to IBM XL compiler semantics 16791when using AIX-compatible ABI@. Pass floating-point arguments to 16792prototyped functions beyond the register save area (RSA) on the stack 16793in addition to argument FPRs. Do not assume that most significant 16794double in 128-bit long double value is properly rounded when comparing 16795values and converting to double. Use XL symbol names for long double 16796support routines. 16797 16798The AIX calling convention was extended but not initially documented to 16799handle an obscure K&R C case of calling a function that takes the 16800address of its arguments with fewer arguments than declared. IBM XL 16801compilers access floating-point arguments that do not fit in the 16802RSA from the stack when a subroutine is compiled without 16803optimization. Because always storing floating-point arguments on the 16804stack is inefficient and rarely needed, this option is not enabled by 16805default and only is necessary when calling subroutines compiled by IBM 16806XL compilers without optimization. 16807 16808@item -mpe 16809@opindex mpe 16810Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an 16811application written to use message passing with special startup code to 16812enable the application to run. The system must have PE installed in the 16813standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file 16814must be overridden with the @option{-specs=} option to specify the 16815appropriate directory location. The Parallel Environment does not 16816support threads, so the @option{-mpe} option and the @option{-pthread} 16817option are incompatible. 16818 16819@item -malign-natural 16820@itemx -malign-power 16821@opindex malign-natural 16822@opindex malign-power 16823On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option 16824@option{-malign-natural} overrides the ABI-defined alignment of larger 16825types, such as floating-point doubles, on their natural size-based boundary. 16826The option @option{-malign-power} instructs GCC to follow the ABI-specified 16827alignment rules. GCC defaults to the standard alignment defined in the ABI@. 16828 16829On 64-bit Darwin, natural alignment is the default, and @option{-malign-power} 16830is not supported. 16831 16832@item -msoft-float 16833@itemx -mhard-float 16834@opindex msoft-float 16835@opindex mhard-float 16836Generate code that does not use (uses) the floating-point register set. 16837Software floating-point emulation is provided if you use the 16838@option{-msoft-float} option, and pass the option to GCC when linking. 16839 16840@item -msingle-float 16841@itemx -mdouble-float 16842@opindex msingle-float 16843@opindex mdouble-float 16844Generate code for single- or double-precision floating-point operations. 16845@option{-mdouble-float} implies @option{-msingle-float}. 16846 16847@item -msimple-fpu 16848@opindex msimple-fpu 16849Do not generate sqrt and div instructions for hardware floating-point unit. 16850 16851@item -mfpu 16852@opindex mfpu 16853Specify type of floating-point unit. Valid values are @var{sp_lite} 16854(equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent 16855to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float), 16856and @var{dp_full} (equivalent to -mdouble-float). 16857 16858@item -mxilinx-fpu 16859@opindex mxilinx-fpu 16860Perform optimizations for the floating-point unit on Xilinx PPC 405/440. 16861 16862@item -mmultiple 16863@itemx -mno-multiple 16864@opindex mmultiple 16865@opindex mno-multiple 16866Generate code that uses (does not use) the load multiple word 16867instructions and the store multiple word instructions. These 16868instructions are generated by default on POWER systems, and not 16869generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian 16870PowerPC systems, since those instructions do not work when the 16871processor is in little-endian mode. The exceptions are PPC740 and 16872PPC750 which permit these instructions in little-endian mode. 16873 16874@item -mstring 16875@itemx -mno-string 16876@opindex mstring 16877@opindex mno-string 16878Generate code that uses (does not use) the load string instructions 16879and the store string word instructions to save multiple registers and 16880do small block moves. These instructions are generated by default on 16881POWER systems, and not generated on PowerPC systems. Do not use 16882@option{-mstring} on little-endian PowerPC systems, since those 16883instructions do not work when the processor is in little-endian mode. 16884The exceptions are PPC740 and PPC750 which permit these instructions 16885in little-endian mode. 16886 16887@item -mupdate 16888@itemx -mno-update 16889@opindex mupdate 16890@opindex mno-update 16891Generate code that uses (does not use) the load or store instructions 16892that update the base register to the address of the calculated memory 16893location. These instructions are generated by default. If you use 16894@option{-mno-update}, there is a small window between the time that the 16895stack pointer is updated and the address of the previous frame is 16896stored, which means code that walks the stack frame across interrupts or 16897signals may get corrupted data. 16898 16899@item -mavoid-indexed-addresses 16900@itemx -mno-avoid-indexed-addresses 16901@opindex mavoid-indexed-addresses 16902@opindex mno-avoid-indexed-addresses 16903Generate code that tries to avoid (not avoid) the use of indexed load 16904or store instructions. These instructions can incur a performance 16905penalty on Power6 processors in certain situations, such as when 16906stepping through large arrays that cross a 16M boundary. This option 16907is enabled by default when targetting Power6 and disabled otherwise. 16908 16909@item -mfused-madd 16910@itemx -mno-fused-madd 16911@opindex mfused-madd 16912@opindex mno-fused-madd 16913Generate code that uses (does not use) the floating-point multiply and 16914accumulate instructions. These instructions are generated by default 16915if hardware floating point is used. The machine-dependent 16916@option{-mfused-madd} option is now mapped to the machine-independent 16917@option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is 16918mapped to @option{-ffp-contract=off}. 16919 16920@item -mmulhw 16921@itemx -mno-mulhw 16922@opindex mmulhw 16923@opindex mno-mulhw 16924Generate code that uses (does not use) the half-word multiply and 16925multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors. 16926These instructions are generated by default when targetting those 16927processors. 16928 16929@item -mdlmzb 16930@itemx -mno-dlmzb 16931@opindex mdlmzb 16932@opindex mno-dlmzb 16933Generate code that uses (does not use) the string-search @samp{dlmzb} 16934instruction on the IBM 405, 440, 464 and 476 processors. This instruction is 16935generated by default when targetting those processors. 16936 16937@item -mno-bit-align 16938@itemx -mbit-align 16939@opindex mno-bit-align 16940@opindex mbit-align 16941On System V.4 and embedded PowerPC systems do not (do) force structures 16942and unions that contain bit-fields to be aligned to the base type of the 16943bit-field. 16944 16945For example, by default a structure containing nothing but 8 16946@code{unsigned} bit-fields of length 1 is aligned to a 4-byte 16947boundary and has a size of 4 bytes. By using @option{-mno-bit-align}, 16948the structure is aligned to a 1-byte boundary and is 1 byte in 16949size. 16950 16951@item -mno-strict-align 16952@itemx -mstrict-align 16953@opindex mno-strict-align 16954@opindex mstrict-align 16955On System V.4 and embedded PowerPC systems do not (do) assume that 16956unaligned memory references will be handled by the system. 16957 16958@item -mrelocatable 16959@itemx -mno-relocatable 16960@opindex mrelocatable 16961@opindex mno-relocatable 16962Generate code that allows (does not allow) a static executable to be 16963relocated to a different address at run time. A simple embedded 16964PowerPC system loader should relocate the entire contents of 16965@code{.got2} and 4-byte locations listed in the @code{.fixup} section, 16966a table of 32-bit addresses generated by this option. For this to 16967work, all objects linked together must be compiled with 16968@option{-mrelocatable} or @option{-mrelocatable-lib}. 16969@option{-mrelocatable} code aligns the stack to an 8-byte boundary. 16970 16971@item -mrelocatable-lib 16972@itemx -mno-relocatable-lib 16973@opindex mrelocatable-lib 16974@opindex mno-relocatable-lib 16975Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a 16976@code{.fixup} section to allow static executables to be relocated at 16977run time, but @option{-mrelocatable-lib} does not use the smaller stack 16978alignment of @option{-mrelocatable}. Objects compiled with 16979@option{-mrelocatable-lib} may be linked with objects compiled with 16980any combination of the @option{-mrelocatable} options. 16981 16982@item -mno-toc 16983@itemx -mtoc 16984@opindex mno-toc 16985@opindex mtoc 16986On System V.4 and embedded PowerPC systems do not (do) assume that 16987register 2 contains a pointer to a global area pointing to the addresses 16988used in the program. 16989 16990@item -mlittle 16991@itemx -mlittle-endian 16992@opindex mlittle 16993@opindex mlittle-endian 16994On System V.4 and embedded PowerPC systems compile code for the 16995processor in little-endian mode. The @option{-mlittle-endian} option is 16996the same as @option{-mlittle}. 16997 16998@item -mbig 16999@itemx -mbig-endian 17000@opindex mbig 17001@opindex mbig-endian 17002On System V.4 and embedded PowerPC systems compile code for the 17003processor in big-endian mode. The @option{-mbig-endian} option is 17004the same as @option{-mbig}. 17005 17006@item -mdynamic-no-pic 17007@opindex mdynamic-no-pic 17008On Darwin and Mac OS X systems, compile code so that it is not 17009relocatable, but that its external references are relocatable. The 17010resulting code is suitable for applications, but not shared 17011libraries. 17012 17013@item -msingle-pic-base 17014@opindex msingle-pic-base 17015Treat the register used for PIC addressing as read-only, rather than 17016loading it in the prologue for each function. The runtime system is 17017responsible for initializing this register with an appropriate value 17018before execution begins. 17019 17020@item -mprioritize-restricted-insns=@var{priority} 17021@opindex mprioritize-restricted-insns 17022This option controls the priority that is assigned to 17023dispatch-slot restricted instructions during the second scheduling 17024pass. The argument @var{priority} takes the value @var{0/1/2} to assign 17025@var{no/highest/second-highest} priority to dispatch slot restricted 17026instructions. 17027 17028@item -msched-costly-dep=@var{dependence_type} 17029@opindex msched-costly-dep 17030This option controls which dependences are considered costly 17031by the target during instruction scheduling. The argument 17032@var{dependence_type} takes one of the following values: 17033@var{no}: no dependence is costly, 17034@var{all}: all dependences are costly, 17035@var{true_store_to_load}: a true dependence from store to load is costly, 17036@var{store_to_load}: any dependence from store to load is costly, 17037@var{number}: any dependence for which latency >= @var{number} is costly. 17038 17039@item -minsert-sched-nops=@var{scheme} 17040@opindex minsert-sched-nops 17041This option controls which nop insertion scheme will be used during 17042the second scheduling pass. The argument @var{scheme} takes one of the 17043following values: 17044@var{no}: Don't insert nops. 17045@var{pad}: Pad with nops any dispatch group that has vacant issue slots, 17046according to the scheduler's grouping. 17047@var{regroup_exact}: Insert nops to force costly dependent insns into 17048separate groups. Insert exactly as many nops as needed to force an insn 17049to a new group, according to the estimated processor grouping. 17050@var{number}: Insert nops to force costly dependent insns into 17051separate groups. Insert @var{number} nops to force an insn to a new group. 17052 17053@item -mcall-sysv 17054@opindex mcall-sysv 17055On System V.4 and embedded PowerPC systems compile code using calling 17056conventions that adheres to the March 1995 draft of the System V 17057Application Binary Interface, PowerPC processor supplement. This is the 17058default unless you configured GCC using @samp{powerpc-*-eabiaix}. 17059 17060@item -mcall-sysv-eabi 17061@itemx -mcall-eabi 17062@opindex mcall-sysv-eabi 17063@opindex mcall-eabi 17064Specify both @option{-mcall-sysv} and @option{-meabi} options. 17065 17066@item -mcall-sysv-noeabi 17067@opindex mcall-sysv-noeabi 17068Specify both @option{-mcall-sysv} and @option{-mno-eabi} options. 17069 17070@item -mcall-aixdesc 17071@opindex m 17072On System V.4 and embedded PowerPC systems compile code for the AIX 17073operating system. 17074 17075@item -mcall-linux 17076@opindex mcall-linux 17077On System V.4 and embedded PowerPC systems compile code for the 17078Linux-based GNU system. 17079 17080@item -mcall-freebsd 17081@opindex mcall-freebsd 17082On System V.4 and embedded PowerPC systems compile code for the 17083FreeBSD operating system. 17084 17085@item -mcall-netbsd 17086@opindex mcall-netbsd 17087On System V.4 and embedded PowerPC systems compile code for the 17088NetBSD operating system. 17089 17090@item -mcall-openbsd 17091@opindex mcall-netbsd 17092On System V.4 and embedded PowerPC systems compile code for the 17093OpenBSD operating system. 17094 17095@item -maix-struct-return 17096@opindex maix-struct-return 17097Return all structures in memory (as specified by the AIX ABI)@. 17098 17099@item -msvr4-struct-return 17100@opindex msvr4-struct-return 17101Return structures smaller than 8 bytes in registers (as specified by the 17102SVR4 ABI)@. 17103 17104@item -mabi=@var{abi-type} 17105@opindex mabi 17106Extend the current ABI with a particular extension, or remove such extension. 17107Valid values are @var{altivec}, @var{no-altivec}, @var{spe}, 17108@var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@. 17109 17110@item -mabi=spe 17111@opindex mabi=spe 17112Extend the current ABI with SPE ABI extensions. This does not change 17113the default ABI, instead it adds the SPE ABI extensions to the current 17114ABI@. 17115 17116@item -mabi=no-spe 17117@opindex mabi=no-spe 17118Disable Booke SPE ABI extensions for the current ABI@. 17119 17120@item -mabi=ibmlongdouble 17121@opindex mabi=ibmlongdouble 17122Change the current ABI to use IBM extended-precision long double. 17123This is a PowerPC 32-bit SYSV ABI option. 17124 17125@item -mabi=ieeelongdouble 17126@opindex mabi=ieeelongdouble 17127Change the current ABI to use IEEE extended-precision long double. 17128This is a PowerPC 32-bit Linux ABI option. 17129 17130@item -mprototype 17131@itemx -mno-prototype 17132@opindex mprototype 17133@opindex mno-prototype 17134On System V.4 and embedded PowerPC systems assume that all calls to 17135variable argument functions are properly prototyped. Otherwise, the 17136compiler must insert an instruction before every non prototyped call to 17137set or clear bit 6 of the condition code register (@var{CR}) to 17138indicate whether floating-point values were passed in the floating-point 17139registers in case the function takes variable arguments. With 17140@option{-mprototype}, only calls to prototyped variable argument functions 17141will set or clear the bit. 17142 17143@item -msim 17144@opindex msim 17145On embedded PowerPC systems, assume that the startup module is called 17146@file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and 17147@file{libc.a}. This is the default for @samp{powerpc-*-eabisim} 17148configurations. 17149 17150@item -mmvme 17151@opindex mmvme 17152On embedded PowerPC systems, assume that the startup module is called 17153@file{crt0.o} and the standard C libraries are @file{libmvme.a} and 17154@file{libc.a}. 17155 17156@item -mads 17157@opindex mads 17158On embedded PowerPC systems, assume that the startup module is called 17159@file{crt0.o} and the standard C libraries are @file{libads.a} and 17160@file{libc.a}. 17161 17162@item -myellowknife 17163@opindex myellowknife 17164On embedded PowerPC systems, assume that the startup module is called 17165@file{crt0.o} and the standard C libraries are @file{libyk.a} and 17166@file{libc.a}. 17167 17168@item -mvxworks 17169@opindex mvxworks 17170On System V.4 and embedded PowerPC systems, specify that you are 17171compiling for a VxWorks system. 17172 17173@item -memb 17174@opindex memb 17175On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags 17176header to indicate that @samp{eabi} extended relocations are used. 17177 17178@item -meabi 17179@itemx -mno-eabi 17180@opindex meabi 17181@opindex mno-eabi 17182On System V.4 and embedded PowerPC systems do (do not) adhere to the 17183Embedded Applications Binary Interface (eabi) which is a set of 17184modifications to the System V.4 specifications. Selecting @option{-meabi} 17185means that the stack is aligned to an 8-byte boundary, a function 17186@code{__eabi} is called to from @code{main} to set up the eabi 17187environment, and the @option{-msdata} option can use both @code{r2} and 17188@code{r13} to point to two separate small data areas. Selecting 17189@option{-mno-eabi} means that the stack is aligned to a 16-byte boundary, 17190do not call an initialization function from @code{main}, and the 17191@option{-msdata} option will only use @code{r13} to point to a single 17192small data area. The @option{-meabi} option is on by default if you 17193configured GCC using one of the @samp{powerpc*-*-eabi*} options. 17194 17195@item -msdata=eabi 17196@opindex msdata=eabi 17197On System V.4 and embedded PowerPC systems, put small initialized 17198@code{const} global and static data in the @samp{.sdata2} section, which 17199is pointed to by register @code{r2}. Put small initialized 17200non-@code{const} global and static data in the @samp{.sdata} section, 17201which is pointed to by register @code{r13}. Put small uninitialized 17202global and static data in the @samp{.sbss} section, which is adjacent to 17203the @samp{.sdata} section. The @option{-msdata=eabi} option is 17204incompatible with the @option{-mrelocatable} option. The 17205@option{-msdata=eabi} option also sets the @option{-memb} option. 17206 17207@item -msdata=sysv 17208@opindex msdata=sysv 17209On System V.4 and embedded PowerPC systems, put small global and static 17210data in the @samp{.sdata} section, which is pointed to by register 17211@code{r13}. Put small uninitialized global and static data in the 17212@samp{.sbss} section, which is adjacent to the @samp{.sdata} section. 17213The @option{-msdata=sysv} option is incompatible with the 17214@option{-mrelocatable} option. 17215 17216@item -msdata=default 17217@itemx -msdata 17218@opindex msdata=default 17219@opindex msdata 17220On System V.4 and embedded PowerPC systems, if @option{-meabi} is used, 17221compile code the same as @option{-msdata=eabi}, otherwise compile code the 17222same as @option{-msdata=sysv}. 17223 17224@item -msdata=data 17225@opindex msdata=data 17226On System V.4 and embedded PowerPC systems, put small global 17227data in the @samp{.sdata} section. Put small uninitialized global 17228data in the @samp{.sbss} section. Do not use register @code{r13} 17229to address small data however. This is the default behavior unless 17230other @option{-msdata} options are used. 17231 17232@item -msdata=none 17233@itemx -mno-sdata 17234@opindex msdata=none 17235@opindex mno-sdata 17236On embedded PowerPC systems, put all initialized global and static data 17237in the @samp{.data} section, and all uninitialized data in the 17238@samp{.bss} section. 17239 17240@item -mblock-move-inline-limit=@var{num} 17241@opindex mblock-move-inline-limit 17242Inline all block moves (such as calls to @code{memcpy} or structure 17243copies) less than or equal to @var{num} bytes. The minimum value for 17244@var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit 17245targets. The default value is target-specific. 17246 17247@item -G @var{num} 17248@opindex G 17249@cindex smaller data references (PowerPC) 17250@cindex .sdata/.sdata2 references (PowerPC) 17251On embedded PowerPC systems, put global and static items less than or 17252equal to @var{num} bytes into the small data or bss sections instead of 17253the normal data or bss section. By default, @var{num} is 8. The 17254@option{-G @var{num}} switch is also passed to the linker. 17255All modules should be compiled with the same @option{-G @var{num}} value. 17256 17257@item -mregnames 17258@itemx -mno-regnames 17259@opindex mregnames 17260@opindex mno-regnames 17261On System V.4 and embedded PowerPC systems do (do not) emit register 17262names in the assembly language output using symbolic forms. 17263 17264@item -mlongcall 17265@itemx -mno-longcall 17266@opindex mlongcall 17267@opindex mno-longcall 17268By default assume that all calls are far away so that a longer more 17269expensive calling sequence is required. This is required for calls 17270further than 32 megabytes (33,554,432 bytes) from the current location. 17271A short call will be generated if the compiler knows 17272the call cannot be that far away. This setting can be overridden by 17273the @code{shortcall} function attribute, or by @code{#pragma 17274longcall(0)}. 17275 17276Some linkers are capable of detecting out-of-range calls and generating 17277glue code on the fly. On these systems, long calls are unnecessary and 17278generate slower code. As of this writing, the AIX linker can do this, 17279as can the GNU linker for PowerPC/64. It is planned to add this feature 17280to the GNU linker for 32-bit PowerPC systems as well. 17281 17282On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr 17283callee, L42'', plus a ``branch island'' (glue code). The two target 17284addresses represent the callee and the ``branch island''. The 17285Darwin/PPC linker will prefer the first address and generate a ``bl 17286callee'' if the PPC ``bl'' instruction will reach the callee directly; 17287otherwise, the linker will generate ``bl L42'' to call the ``branch 17288island''. The ``branch island'' is appended to the body of the 17289calling function; it computes the full 32-bit address of the callee 17290and jumps to it. 17291 17292On Mach-O (Darwin) systems, this option directs the compiler emit to 17293the glue for every direct call, and the Darwin linker decides whether 17294to use or discard it. 17295 17296In the future, we may cause GCC to ignore all longcall specifications 17297when the linker is known to generate glue. 17298 17299@item -mtls-markers 17300@itemx -mno-tls-markers 17301@opindex mtls-markers 17302@opindex mno-tls-markers 17303Mark (do not mark) calls to @code{__tls_get_addr} with a relocation 17304specifying the function argument. The relocation allows ld to 17305reliably associate function call with argument setup instructions for 17306TLS optimization, which in turn allows gcc to better schedule the 17307sequence. 17308 17309@item -pthread 17310@opindex pthread 17311Adds support for multithreading with the @dfn{pthreads} library. 17312This option sets flags for both the preprocessor and linker. 17313 17314@item -mrecip 17315@itemx -mno-recip 17316@opindex mrecip 17317This option will enable GCC to use the reciprocal estimate and 17318reciprocal square root estimate instructions with additional 17319Newton-Raphson steps to increase precision instead of doing a divide or 17320square root and divide for floating-point arguments. You should use 17321the @option{-ffast-math} option when using @option{-mrecip} (or at 17322least @option{-funsafe-math-optimizations}, 17323@option{-finite-math-only}, @option{-freciprocal-math} and 17324@option{-fno-trapping-math}). Note that while the throughput of the 17325sequence is generally higher than the throughput of the non-reciprocal 17326instruction, the precision of the sequence can be decreased by up to 2 17327ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square 17328roots. 17329 17330@item -mrecip=@var{opt} 17331@opindex mrecip=opt 17332This option allows to control which reciprocal estimate instructions 17333may be used. @var{opt} is a comma separated list of options, which may 17334be preceded by a @code{!} to invert the option: 17335@code{all}: enable all estimate instructions, 17336@code{default}: enable the default instructions, equivalent to @option{-mrecip}, 17337@code{none}: disable all estimate instructions, equivalent to @option{-mno-recip}; 17338@code{div}: enable the reciprocal approximation instructions for both single and double precision; 17339@code{divf}: enable the single-precision reciprocal approximation instructions; 17340@code{divd}: enable the double-precision reciprocal approximation instructions; 17341@code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision; 17342@code{rsqrtf}: enable the single-precision reciprocal square root approximation instructions; 17343@code{rsqrtd}: enable the double-precision reciprocal square root approximation instructions; 17344 17345So for example, @option{-mrecip=all,!rsqrtd} would enable the 17346all of the reciprocal estimate instructions, except for the 17347@code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions 17348which handle the double-precision reciprocal square root calculations. 17349 17350@item -mrecip-precision 17351@itemx -mno-recip-precision 17352@opindex mrecip-precision 17353Assume (do not assume) that the reciprocal estimate instructions 17354provide higher-precision estimates than is mandated by the PowerPC 17355ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7} 17356automatically selects @option{-mrecip-precision}. The double-precision 17357square root estimate instructions are not generated by 17358default on low-precision machines, since they do not provide an 17359estimate that converges after three steps. 17360 17361@item -mveclibabi=@var{type} 17362@opindex mveclibabi 17363Specifies the ABI type to use for vectorizing intrinsics using an 17364external library. The only type supported at present is @code{mass}, 17365which specifies to use IBM's Mathematical Acceleration Subsystem 17366(MASS) libraries for vectorizing intrinsics using external libraries. 17367GCC will currently emit calls to @code{acosd2}, @code{acosf4}, 17368@code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4}, 17369@code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4}, 17370@code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4}, 17371@code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4}, 17372@code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4}, 17373@code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4}, 17374@code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4}, 17375@code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4}, 17376@code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4}, 17377@code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4}, 17378@code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2}, 17379@code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2}, 17380@code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code 17381for power7. Both @option{-ftree-vectorize} and 17382@option{-funsafe-math-optimizations} have to be enabled. The MASS 17383libraries will have to be specified at link time. 17384 17385@item -mfriz 17386@itemx -mno-friz 17387@opindex mfriz 17388Generate (do not generate) the @code{friz} instruction when the 17389@option{-funsafe-math-optimizations} option is used to optimize 17390rounding of floating-point values to 64-bit integer and back to floating 17391point. The @code{friz} instruction does not return the same value if 17392the floating-point number is too large to fit in an integer. 17393 17394@item -mpointers-to-nested-functions 17395@itemx -mno-pointers-to-nested-functions 17396@opindex mpointers-to-nested-functions 17397Generate (do not generate) code to load up the static chain register 17398(@var{r11}) when calling through a pointer on AIX and 64-bit Linux 17399systems where a function pointer points to a 3-word descriptor giving 17400the function address, TOC value to be loaded in register @var{r2}, and 17401static chain value to be loaded in register @var{r11}. The 17402@option{-mpointers-to-nested-functions} is on by default. You will 17403not be able to call through pointers to nested functions or pointers 17404to functions compiled in other languages that use the static chain if 17405you use the @option{-mno-pointers-to-nested-functions}. 17406 17407@item -msave-toc-indirect 17408@itemx -mno-save-toc-indirect 17409@opindex msave-toc-indirect 17410Generate (do not generate) code to save the TOC value in the reserved 17411stack location in the function prologue if the function calls through 17412a pointer on AIX and 64-bit Linux systems. If the TOC value is not 17413saved in the prologue, it is saved just before the call through the 17414pointer. The @option{-mno-save-toc-indirect} option is the default. 17415@end table 17416 17417@node RX Options 17418@subsection RX Options 17419@cindex RX Options 17420 17421These command-line options are defined for RX targets: 17422 17423@table @gcctabopt 17424@item -m64bit-doubles 17425@itemx -m32bit-doubles 17426@opindex m64bit-doubles 17427@opindex m32bit-doubles 17428Make the @code{double} data type be 64 bits (@option{-m64bit-doubles}) 17429or 32 bits (@option{-m32bit-doubles}) in size. The default is 17430@option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only 17431works on 32-bit values, which is why the default is 17432@option{-m32bit-doubles}. 17433 17434@item -fpu 17435@itemx -nofpu 17436@opindex fpu 17437@opindex nofpu 17438Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX 17439floating-point hardware. The default is enabled for the @var{RX600} 17440series and disabled for the @var{RX200} series. 17441 17442Floating-point instructions will only be generated for 32-bit floating-point 17443values however, so if the @option{-m64bit-doubles} option is in 17444use then the FPU hardware will not be used for doubles. 17445 17446@emph{Note} If the @option{-fpu} option is enabled then 17447@option{-funsafe-math-optimizations} is also enabled automatically. 17448This is because the RX FPU instructions are themselves unsafe. 17449 17450@item -mcpu=@var{name} 17451@opindex -mcpu 17452Selects the type of RX CPU to be targeted. Currently three types are 17453supported, the generic @var{RX600} and @var{RX200} series hardware and 17454the specific @var{RX610} CPU. The default is @var{RX600}. 17455 17456The only difference between @var{RX600} and @var{RX610} is that the 17457@var{RX610} does not support the @code{MVTIPL} instruction. 17458 17459The @var{RX200} series does not have a hardware floating-point unit 17460and so @option{-nofpu} is enabled by default when this type is 17461selected. 17462 17463@item -mbig-endian-data 17464@itemx -mlittle-endian-data 17465@opindex mbig-endian-data 17466@opindex mlittle-endian-data 17467Store data (but not code) in the big-endian format. The default is 17468@option{-mlittle-endian-data}, i.e.@: to store data in the little-endian 17469format. 17470 17471@item -msmall-data-limit=@var{N} 17472@opindex msmall-data-limit 17473Specifies the maximum size in bytes of global and static variables 17474which can be placed into the small data area. Using the small data 17475area can lead to smaller and faster code, but the size of area is 17476limited and it is up to the programmer to ensure that the area does 17477not overflow. Also when the small data area is used one of the RX's 17478registers (usually @code{r13}) is reserved for use pointing to this 17479area, so it is no longer available for use by the compiler. This 17480could result in slower and/or larger code if variables which once 17481could have been held in the reserved register are now pushed onto the 17482stack. 17483 17484Note, common variables (variables that have not been initialized) and 17485constants are not placed into the small data area as they are assigned 17486to other sections in the output executable. 17487 17488The default value is zero, which disables this feature. Note, this 17489feature is not enabled by default with higher optimization levels 17490(@option{-O2} etc) because of the potentially detrimental effects of 17491reserving a register. It is up to the programmer to experiment and 17492discover whether this feature is of benefit to their program. See the 17493description of the @option{-mpid} option for a description of how the 17494actual register to hold the small data area pointer is chosen. 17495 17496@item -msim 17497@itemx -mno-sim 17498@opindex msim 17499@opindex mno-sim 17500Use the simulator runtime. The default is to use the libgloss board 17501specific runtime. 17502 17503@item -mas100-syntax 17504@itemx -mno-as100-syntax 17505@opindex mas100-syntax 17506@opindex mno-as100-syntax 17507When generating assembler output use a syntax that is compatible with 17508Renesas's AS100 assembler. This syntax can also be handled by the GAS 17509assembler but it has some restrictions so generating it is not the 17510default option. 17511 17512@item -mmax-constant-size=@var{N} 17513@opindex mmax-constant-size 17514Specifies the maximum size, in bytes, of a constant that can be used as 17515an operand in a RX instruction. Although the RX instruction set does 17516allow constants of up to 4 bytes in length to be used in instructions, 17517a longer value equates to a longer instruction. Thus in some 17518circumstances it can be beneficial to restrict the size of constants 17519that are used in instructions. Constants that are too big are instead 17520placed into a constant pool and referenced via register indirection. 17521 17522The value @var{N} can be between 0 and 4. A value of 0 (the default) 17523or 4 means that constants of any size are allowed. 17524 17525@item -mrelax 17526@opindex mrelax 17527Enable linker relaxation. Linker relaxation is a process whereby the 17528linker will attempt to reduce the size of a program by finding shorter 17529versions of various instructions. Disabled by default. 17530 17531@item -mint-register=@var{N} 17532@opindex mint-register 17533Specify the number of registers to reserve for fast interrupt handler 17534functions. The value @var{N} can be between 0 and 4. A value of 1 17535means that register @code{r13} will be reserved for the exclusive use 17536of fast interrupt handlers. A value of 2 reserves @code{r13} and 17537@code{r12}. A value of 3 reserves @code{r13}, @code{r12} and 17538@code{r11}, and a value of 4 reserves @code{r13} through @code{r10}. 17539A value of 0, the default, does not reserve any registers. 17540 17541@item -msave-acc-in-interrupts 17542@opindex msave-acc-in-interrupts 17543Specifies that interrupt handler functions should preserve the 17544accumulator register. This is only necessary if normal code might use 17545the accumulator register, for example because it performs 64-bit 17546multiplications. The default is to ignore the accumulator as this 17547makes the interrupt handlers faster. 17548 17549@item -mpid 17550@itemx -mno-pid 17551@opindex mpid 17552@opindex mno-pid 17553Enables the generation of position independent data. When enabled any 17554access to constant data will done via an offset from a base address 17555held in a register. This allows the location of constant data to be 17556determined at run time without requiring the executable to be 17557relocated, which is a benefit to embedded applications with tight 17558memory constraints. Data that can be modified is not affected by this 17559option. 17560 17561Note, using this feature reserves a register, usually @code{r13}, for 17562the constant data base address. This can result in slower and/or 17563larger code, especially in complicated functions. 17564 17565The actual register chosen to hold the constant data base address 17566depends upon whether the @option{-msmall-data-limit} and/or the 17567@option{-mint-register} command-line options are enabled. Starting 17568with register @code{r13} and proceeding downwards, registers are 17569allocated first to satisfy the requirements of @option{-mint-register}, 17570then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it 17571is possible for the small data area register to be @code{r8} if both 17572@option{-mint-register=4} and @option{-mpid} are specified on the 17573command line. 17574 17575By default this feature is not enabled. The default can be restored 17576via the @option{-mno-pid} command-line option. 17577 17578@end table 17579 17580@emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}} 17581has special significance to the RX port when used with the 17582@code{interrupt} function attribute. This attribute indicates a 17583function intended to process fast interrupts. GCC will will ensure 17584that it only uses the registers @code{r10}, @code{r11}, @code{r12} 17585and/or @code{r13} and only provided that the normal use of the 17586corresponding registers have been restricted via the 17587@option{-ffixed-@var{reg}} or @option{-mint-register} command-line 17588options. 17589 17590@node S/390 and zSeries Options 17591@subsection S/390 and zSeries Options 17592@cindex S/390 and zSeries Options 17593 17594These are the @samp{-m} options defined for the S/390 and zSeries architecture. 17595 17596@table @gcctabopt 17597@item -mhard-float 17598@itemx -msoft-float 17599@opindex mhard-float 17600@opindex msoft-float 17601Use (do not use) the hardware floating-point instructions and registers 17602for floating-point operations. When @option{-msoft-float} is specified, 17603functions in @file{libgcc.a} will be used to perform floating-point 17604operations. When @option{-mhard-float} is specified, the compiler 17605generates IEEE floating-point instructions. This is the default. 17606 17607@item -mhard-dfp 17608@itemx -mno-hard-dfp 17609@opindex mhard-dfp 17610@opindex mno-hard-dfp 17611Use (do not use) the hardware decimal-floating-point instructions for 17612decimal-floating-point operations. When @option{-mno-hard-dfp} is 17613specified, functions in @file{libgcc.a} will be used to perform 17614decimal-floating-point operations. When @option{-mhard-dfp} is 17615specified, the compiler generates decimal-floating-point hardware 17616instructions. This is the default for @option{-march=z9-ec} or higher. 17617 17618@item -mlong-double-64 17619@itemx -mlong-double-128 17620@opindex mlong-double-64 17621@opindex mlong-double-128 17622These switches control the size of @code{long double} type. A size 17623of 64 bits makes the @code{long double} type equivalent to the @code{double} 17624type. This is the default. 17625 17626@item -mbackchain 17627@itemx -mno-backchain 17628@opindex mbackchain 17629@opindex mno-backchain 17630Store (do not store) the address of the caller's frame as backchain pointer 17631into the callee's stack frame. 17632A backchain may be needed to allow debugging using tools that do not understand 17633DWARF-2 call frame information. 17634When @option{-mno-packed-stack} is in effect, the backchain pointer is stored 17635at the bottom of the stack frame; when @option{-mpacked-stack} is in effect, 17636the backchain is placed into the topmost word of the 96/160 byte register 17637save area. 17638 17639In general, code compiled with @option{-mbackchain} is call-compatible with 17640code compiled with @option{-mmo-backchain}; however, use of the backchain 17641for debugging purposes usually requires that the whole binary is built with 17642@option{-mbackchain}. Note that the combination of @option{-mbackchain}, 17643@option{-mpacked-stack} and @option{-mhard-float} is not supported. In order 17644to build a linux kernel use @option{-msoft-float}. 17645 17646The default is to not maintain the backchain. 17647 17648@item -mpacked-stack 17649@itemx -mno-packed-stack 17650@opindex mpacked-stack 17651@opindex mno-packed-stack 17652Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is 17653specified, the compiler uses the all fields of the 96/160 byte register save 17654area only for their default purpose; unused fields still take up stack space. 17655When @option{-mpacked-stack} is specified, register save slots are densely 17656packed at the top of the register save area; unused space is reused for other 17657purposes, allowing for more efficient use of the available stack space. 17658However, when @option{-mbackchain} is also in effect, the topmost word of 17659the save area is always used to store the backchain, and the return address 17660register is always saved two words below the backchain. 17661 17662As long as the stack frame backchain is not used, code generated with 17663@option{-mpacked-stack} is call-compatible with code generated with 17664@option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for 17665S/390 or zSeries generated code that uses the stack frame backchain at run 17666time, not just for debugging purposes. Such code is not call-compatible 17667with code compiled with @option{-mpacked-stack}. Also, note that the 17668combination of @option{-mbackchain}, 17669@option{-mpacked-stack} and @option{-mhard-float} is not supported. In order 17670to build a linux kernel use @option{-msoft-float}. 17671 17672The default is to not use the packed stack layout. 17673 17674@item -msmall-exec 17675@itemx -mno-small-exec 17676@opindex msmall-exec 17677@opindex mno-small-exec 17678Generate (or do not generate) code using the @code{bras} instruction 17679to do subroutine calls. 17680This only works reliably if the total executable size does not 17681exceed 64k. The default is to use the @code{basr} instruction instead, 17682which does not have this limitation. 17683 17684@item -m64 17685@itemx -m31 17686@opindex m64 17687@opindex m31 17688When @option{-m31} is specified, generate code compliant to the 17689GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate 17690code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in 17691particular to generate 64-bit instructions. For the @samp{s390} 17692targets, the default is @option{-m31}, while the @samp{s390x} 17693targets default to @option{-m64}. 17694 17695@item -mzarch 17696@itemx -mesa 17697@opindex mzarch 17698@opindex mesa 17699When @option{-mzarch} is specified, generate code using the 17700instructions available on z/Architecture. 17701When @option{-mesa} is specified, generate code using the 17702instructions available on ESA/390. Note that @option{-mesa} is 17703not possible with @option{-m64}. 17704When generating code compliant to the GNU/Linux for S/390 ABI, 17705the default is @option{-mesa}. When generating code compliant 17706to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}. 17707 17708@item -mmvcle 17709@itemx -mno-mvcle 17710@opindex mmvcle 17711@opindex mno-mvcle 17712Generate (or do not generate) code using the @code{mvcle} instruction 17713to perform block moves. When @option{-mno-mvcle} is specified, 17714use a @code{mvc} loop instead. This is the default unless optimizing for 17715size. 17716 17717@item -mdebug 17718@itemx -mno-debug 17719@opindex mdebug 17720@opindex mno-debug 17721Print (or do not print) additional debug information when compiling. 17722The default is to not print debug information. 17723 17724@item -march=@var{cpu-type} 17725@opindex march 17726Generate code that will run on @var{cpu-type}, which is the name of a system 17727representing a certain processor type. Possible values for 17728@var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990}, 17729@samp{z9-109}, @samp{z9-ec} and @samp{z10}. 17730When generating code using the instructions available on z/Architecture, 17731the default is @option{-march=z900}. Otherwise, the default is 17732@option{-march=g5}. 17733 17734@item -mtune=@var{cpu-type} 17735@opindex mtune 17736Tune to @var{cpu-type} everything applicable about the generated code, 17737except for the ABI and the set of available instructions. 17738The list of @var{cpu-type} values is the same as for @option{-march}. 17739The default is the value used for @option{-march}. 17740 17741@item -mtpf-trace 17742@itemx -mno-tpf-trace 17743@opindex mtpf-trace 17744@opindex mno-tpf-trace 17745Generate code that adds (does not add) in TPF OS specific branches to trace 17746routines in the operating system. This option is off by default, even 17747when compiling for the TPF OS@. 17748 17749@item -mfused-madd 17750@itemx -mno-fused-madd 17751@opindex mfused-madd 17752@opindex mno-fused-madd 17753Generate code that uses (does not use) the floating-point multiply and 17754accumulate instructions. These instructions are generated by default if 17755hardware floating point is used. 17756 17757@item -mwarn-framesize=@var{framesize} 17758@opindex mwarn-framesize 17759Emit a warning if the current function exceeds the given frame size. Because 17760this is a compile-time check it doesn't need to be a real problem when the program 17761runs. It is intended to identify functions that most probably cause 17762a stack overflow. It is useful to be used in an environment with limited stack 17763size e.g.@: the linux kernel. 17764 17765@item -mwarn-dynamicstack 17766@opindex mwarn-dynamicstack 17767Emit a warning if the function calls alloca or uses dynamically 17768sized arrays. This is generally a bad idea with a limited stack size. 17769 17770@item -mstack-guard=@var{stack-guard} 17771@itemx -mstack-size=@var{stack-size} 17772@opindex mstack-guard 17773@opindex mstack-size 17774If these options are provided the s390 back end emits additional instructions in 17775the function prologue which trigger a trap if the stack size is @var{stack-guard} 17776bytes above the @var{stack-size} (remember that the stack on s390 grows downward). 17777If the @var{stack-guard} option is omitted the smallest power of 2 larger than 17778the frame size of the compiled function is chosen. 17779These options are intended to be used to help debugging stack overflow problems. 17780The additionally emitted code causes only little overhead and hence can also be 17781used in production like systems without greater performance degradation. The given 17782values have to be exact powers of 2 and @var{stack-size} has to be greater than 17783@var{stack-guard} without exceeding 64k. 17784In order to be efficient the extra code makes the assumption that the stack starts 17785at an address aligned to the value given by @var{stack-size}. 17786The @var{stack-guard} option can only be used in conjunction with @var{stack-size}. 17787@end table 17788 17789@node Score Options 17790@subsection Score Options 17791@cindex Score Options 17792 17793These options are defined for Score implementations: 17794 17795@table @gcctabopt 17796@item -meb 17797@opindex meb 17798Compile code for big-endian mode. This is the default. 17799 17800@item -mel 17801@opindex mel 17802Compile code for little-endian mode. 17803 17804@item -mnhwloop 17805@opindex mnhwloop 17806Disable generate bcnz instruction. 17807 17808@item -muls 17809@opindex muls 17810Enable generate unaligned load and store instruction. 17811 17812@item -mmac 17813@opindex mmac 17814Enable the use of multiply-accumulate instructions. Disabled by default. 17815 17816@item -mscore5 17817@opindex mscore5 17818Specify the SCORE5 as the target architecture. 17819 17820@item -mscore5u 17821@opindex mscore5u 17822Specify the SCORE5U of the target architecture. 17823 17824@item -mscore7 17825@opindex mscore7 17826Specify the SCORE7 as the target architecture. This is the default. 17827 17828@item -mscore7d 17829@opindex mscore7d 17830Specify the SCORE7D as the target architecture. 17831@end table 17832 17833@node SH Options 17834@subsection SH Options 17835 17836These @samp{-m} options are defined for the SH implementations: 17837 17838@table @gcctabopt 17839@item -m1 17840@opindex m1 17841Generate code for the SH1. 17842 17843@item -m2 17844@opindex m2 17845Generate code for the SH2. 17846 17847@item -m2e 17848Generate code for the SH2e. 17849 17850@item -m2a-nofpu 17851@opindex m2a-nofpu 17852Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way 17853that the floating-point unit is not used. 17854 17855@item -m2a-single-only 17856@opindex m2a-single-only 17857Generate code for the SH2a-FPU, in such a way that no double-precision 17858floating-point operations are used. 17859 17860@item -m2a-single 17861@opindex m2a-single 17862Generate code for the SH2a-FPU assuming the floating-point unit is in 17863single-precision mode by default. 17864 17865@item -m2a 17866@opindex m2a 17867Generate code for the SH2a-FPU assuming the floating-point unit is in 17868double-precision mode by default. 17869 17870@item -m3 17871@opindex m3 17872Generate code for the SH3. 17873 17874@item -m3e 17875@opindex m3e 17876Generate code for the SH3e. 17877 17878@item -m4-nofpu 17879@opindex m4-nofpu 17880Generate code for the SH4 without a floating-point unit. 17881 17882@item -m4-single-only 17883@opindex m4-single-only 17884Generate code for the SH4 with a floating-point unit that only 17885supports single-precision arithmetic. 17886 17887@item -m4-single 17888@opindex m4-single 17889Generate code for the SH4 assuming the floating-point unit is in 17890single-precision mode by default. 17891 17892@item -m4 17893@opindex m4 17894Generate code for the SH4. 17895 17896@item -m4a-nofpu 17897@opindex m4a-nofpu 17898Generate code for the SH4al-dsp, or for a SH4a in such a way that the 17899floating-point unit is not used. 17900 17901@item -m4a-single-only 17902@opindex m4a-single-only 17903Generate code for the SH4a, in such a way that no double-precision 17904floating-point operations are used. 17905 17906@item -m4a-single 17907@opindex m4a-single 17908Generate code for the SH4a assuming the floating-point unit is in 17909single-precision mode by default. 17910 17911@item -m4a 17912@opindex m4a 17913Generate code for the SH4a. 17914 17915@item -m4al 17916@opindex m4al 17917Same as @option{-m4a-nofpu}, except that it implicitly passes 17918@option{-dsp} to the assembler. GCC doesn't generate any DSP 17919instructions at the moment. 17920 17921@item -mb 17922@opindex mb 17923Compile code for the processor in big-endian mode. 17924 17925@item -ml 17926@opindex ml 17927Compile code for the processor in little-endian mode. 17928 17929@item -mdalign 17930@opindex mdalign 17931Align doubles at 64-bit boundaries. Note that this changes the calling 17932conventions, and thus some functions from the standard C library will 17933not work unless you recompile it first with @option{-mdalign}. 17934 17935@item -mrelax 17936@opindex mrelax 17937Shorten some address references at link time, when possible; uses the 17938linker option @option{-relax}. 17939 17940@item -mbigtable 17941@opindex mbigtable 17942Use 32-bit offsets in @code{switch} tables. The default is to use 1794316-bit offsets. 17944 17945@item -mbitops 17946@opindex mbitops 17947Enable the use of bit manipulation instructions on SH2A. 17948 17949@item -mfmovd 17950@opindex mfmovd 17951Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for 17952alignment constraints. 17953 17954@item -mhitachi 17955@opindex mhitachi 17956Comply with the calling conventions defined by Renesas. 17957 17958@item -mrenesas 17959@opindex mhitachi 17960Comply with the calling conventions defined by Renesas. 17961 17962@item -mno-renesas 17963@opindex mhitachi 17964Comply with the calling conventions defined for GCC before the Renesas 17965conventions were available. This option is the default for all 17966targets of the SH toolchain. 17967 17968@item -mnomacsave 17969@opindex mnomacsave 17970Mark the @code{MAC} register as call-clobbered, even if 17971@option{-mhitachi} is given. 17972 17973@item -mieee 17974@item -mno-ieee 17975@opindex mieee 17976@opindex mnoieee 17977Control the IEEE compliance of floating-point comparisons, which affects the 17978handling of cases where the result of a comparison is unordered. By default 17979@option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is 17980enabled @option{-mno-ieee} is implicitly set, which results in faster 17981floating-point greater-equal and less-equal comparisons. The implcit settings 17982can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}. 17983 17984@item -minline-ic_invalidate 17985@opindex minline-ic_invalidate 17986Inline code to invalidate instruction cache entries after setting up 17987nested function trampolines. 17988This option has no effect if -musermode is in effect and the selected 17989code generation option (e.g. -m4) does not allow the use of the icbi 17990instruction. 17991If the selected code generation option does not allow the use of the icbi 17992instruction, and -musermode is not in effect, the inlined code will 17993manipulate the instruction cache address array directly with an associative 17994write. This not only requires privileged mode, but it will also 17995fail if the cache line had been mapped via the TLB and has become unmapped. 17996 17997@item -misize 17998@opindex misize 17999Dump instruction size and location in the assembly code. 18000 18001@item -mpadstruct 18002@opindex mpadstruct 18003This option is deprecated. It pads structures to multiple of 4 bytes, 18004which is incompatible with the SH ABI@. 18005 18006@item -msoft-atomic 18007@opindex msoft-atomic 18008Generate GNU/Linux compatible gUSA software atomic sequences for the atomic 18009built-in functions. The generated atomic sequences require support from the 18010interrupt / exception handling code of the system and are only suitable for 18011single-core systems. They will not perform correctly on multi-core systems. 18012This option is enabled by default when the target is @code{sh-*-linux*}. 18013For details on the atomic built-in functions see @ref{__atomic Builtins}. 18014 18015@item -mspace 18016@opindex mspace 18017Optimize for space instead of speed. Implied by @option{-Os}. 18018 18019@item -mprefergot 18020@opindex mprefergot 18021When generating position-independent code, emit function calls using 18022the Global Offset Table instead of the Procedure Linkage Table. 18023 18024@item -musermode 18025@opindex musermode 18026Don't generate privileged mode only code; implies -mno-inline-ic_invalidate 18027if the inlined code would not work in user mode. 18028This is the default when the target is @code{sh-*-linux*}. 18029 18030@item -multcost=@var{number} 18031@opindex multcost=@var{number} 18032Set the cost to assume for a multiply insn. 18033 18034@item -mdiv=@var{strategy} 18035@opindex mdiv=@var{strategy} 18036Set the division strategy to use for SHmedia code. @var{strategy} must be 18037one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call, 18038inv:call2, inv:fp . 18039"fp" performs the operation in floating point. This has a very high latency, 18040but needs only a few instructions, so it might be a good choice if 18041your code has enough easily-exploitable ILP to allow the compiler to 18042schedule the floating-point instructions together with other instructions. 18043Division by zero causes a floating-point exception. 18044"inv" uses integer operations to calculate the inverse of the divisor, 18045and then multiplies the dividend with the inverse. This strategy allows 18046cse and hoisting of the inverse calculation. Division by zero calculates 18047an unspecified result, but does not trap. 18048"inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities 18049have been found, or if the entire operation has been hoisted to the same 18050place, the last stages of the inverse calculation are intertwined with the 18051final multiply to reduce the overall latency, at the expense of using a few 18052more instructions, and thus offering fewer scheduling opportunities with 18053other code. 18054"call" calls a library function that usually implements the inv:minlat 18055strategy. 18056This gives high code density for m5-*media-nofpu compilations. 18057"call2" uses a different entry point of the same library function, where it 18058assumes that a pointer to a lookup table has already been set up, which 18059exposes the pointer load to cse / code hoisting optimizations. 18060"inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial 18061code generation, but if the code stays unoptimized, revert to the "call", 18062"call2", or "fp" strategies, respectively. Note that the 18063potentially-trapping side effect of division by zero is carried by a 18064separate instruction, so it is possible that all the integer instructions 18065are hoisted out, but the marker for the side effect stays where it is. 18066A recombination to fp operations or a call is not possible in that case. 18067"inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case 18068that the inverse calculation was nor separated from the multiply, they speed 18069up division where the dividend fits into 20 bits (plus sign where applicable), 18070by inserting a test to skip a number of operations in this case; this test 18071slows down the case of larger dividends. inv20u assumes the case of a such 18072a small dividend to be unlikely, and inv20l assumes it to be likely. 18073 18074@item -maccumulate-outgoing-args 18075@opindex maccumulate-outgoing-args 18076Reserve space once for outgoing arguments in the function prologue rather 18077than around each call. Generally beneficial for performance and size. Also 18078needed for unwinding to avoid changing the stack frame around conditional code. 18079 18080@item -mdivsi3_libfunc=@var{name} 18081@opindex mdivsi3_libfunc=@var{name} 18082Set the name of the library function used for 32-bit signed division to 18083@var{name}. This only affect the name used in the call and inv:call 18084division strategies, and the compiler will still expect the same 18085sets of input/output/clobbered registers as if this option was not present. 18086 18087@item -mfixed-range=@var{register-range} 18088@opindex mfixed-range 18089Generate code treating the given register range as fixed registers. 18090A fixed register is one that the register allocator can not use. This is 18091useful when compiling kernel code. A register range is specified as 18092two registers separated by a dash. Multiple register ranges can be 18093specified separated by a comma. 18094 18095@item -madjust-unroll 18096@opindex madjust-unroll 18097Throttle unrolling to avoid thrashing target registers. 18098This option only has an effect if the gcc code base supports the 18099TARGET_ADJUST_UNROLL_MAX target hook. 18100 18101@item -mindexed-addressing 18102@opindex mindexed-addressing 18103Enable the use of the indexed addressing mode for SHmedia32/SHcompact. 18104This is only safe if the hardware and/or OS implement 32-bit wrap-around 18105semantics for the indexed addressing mode. The architecture allows the 18106implementation of processors with 64-bit MMU, which the OS could use to 18107get 32-bit addressing, but since no current hardware implementation supports 18108this or any other way to make the indexed addressing mode safe to use in 18109the 32-bit ABI, the default is @option{-mno-indexed-addressing}. 18110 18111@item -mgettrcost=@var{number} 18112@opindex mgettrcost=@var{number} 18113Set the cost assumed for the gettr instruction to @var{number}. 18114The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise. 18115 18116@item -mpt-fixed 18117@opindex mpt-fixed 18118Assume pt* instructions won't trap. This will generally generate better 18119scheduled code, but is unsafe on current hardware. The current architecture 18120definition says that ptabs and ptrel trap when the target anded with 3 is 3. 18121This has the unintentional effect of making it unsafe to schedule ptabs / 18122ptrel before a branch, or hoist it out of a loop. For example, 18123__do_global_ctors, a part of libgcc that runs constructors at program 18124startup, calls functions in a list which is delimited by @minus{}1. With the 18125-mpt-fixed option, the ptabs will be done before testing against @minus{}1. 18126That means that all the constructors will be run a bit quicker, but when 18127the loop comes to the end of the list, the program crashes because ptabs 18128loads @minus{}1 into a target register. Since this option is unsafe for any 18129hardware implementing the current architecture specification, the default 18130is -mno-pt-fixed. Unless the user specifies a specific cost with 18131@option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100}; 18132this deters register allocation using target registers for storing 18133ordinary integers. 18134 18135@item -minvalid-symbols 18136@opindex minvalid-symbols 18137Assume symbols might be invalid. Ordinary function symbols generated by 18138the compiler will always be valid to load with movi/shori/ptabs or 18139movi/shori/ptrel, but with assembler and/or linker tricks it is possible 18140to generate symbols that will cause ptabs / ptrel to trap. 18141This option is only meaningful when @option{-mno-pt-fixed} is in effect. 18142It will then prevent cross-basic-block cse, hoisting and most scheduling 18143of symbol loads. The default is @option{-mno-invalid-symbols}. 18144 18145@item -mbranch-cost=@var{num} 18146@opindex mbranch-cost=@var{num} 18147Assume @var{num} to be the cost for a branch instruction. Higher numbers 18148will make the compiler try to generate more branch-free code if possible. 18149If not specified the value is selected depending on the processor type that 18150is being compiled for. 18151 18152@item -mcbranchdi 18153@opindex mcbranchdi 18154Enable the @code{cbranchdi4} instruction pattern. 18155 18156@item -mcmpeqdi 18157@opindex mcmpeqdi 18158Emit the @code{cmpeqdi_t} instruction pattern even when @option{-mcbranchdi} 18159is in effect. 18160 18161@item -mfused-madd 18162@opindex mfused-madd 18163Allow the usage of the @code{fmac} instruction (floating-point 18164multiply-accumulate) if the processor type supports it. Enabling this 18165option might generate code that produces different numeric floating-point 18166results compared to strict IEEE 754 arithmetic. 18167 18168@item -mpretend-cmove 18169@opindex mpretend-cmove 18170Prefer zero-displacement conditional branches for conditional move instruction 18171patterns. This can result in faster code on the SH4 processor. 18172 18173@end table 18174 18175@node Solaris 2 Options 18176@subsection Solaris 2 Options 18177@cindex Solaris 2 options 18178 18179These @samp{-m} options are supported on Solaris 2: 18180 18181@table @gcctabopt 18182@item -mimpure-text 18183@opindex mimpure-text 18184@option{-mimpure-text}, used in addition to @option{-shared}, tells 18185the compiler to not pass @option{-z text} to the linker when linking a 18186shared object. Using this option, you can link position-dependent 18187code into a shared object. 18188 18189@option{-mimpure-text} suppresses the ``relocations remain against 18190allocatable but non-writable sections'' linker error message. 18191However, the necessary relocations will trigger copy-on-write, and the 18192shared object is not actually shared across processes. Instead of 18193using @option{-mimpure-text}, you should compile all source code with 18194@option{-fpic} or @option{-fPIC}. 18195 18196@end table 18197 18198These switches are supported in addition to the above on Solaris 2: 18199 18200@table @gcctabopt 18201@item -pthreads 18202@opindex pthreads 18203Add support for multithreading using the POSIX threads library. This 18204option sets flags for both the preprocessor and linker. This option does 18205not affect the thread safety of object code produced by the compiler or 18206that of libraries supplied with it. 18207 18208@item -pthread 18209@opindex pthread 18210This is a synonym for @option{-pthreads}. 18211@end table 18212 18213@node SPARC Options 18214@subsection SPARC Options 18215@cindex SPARC options 18216 18217These @samp{-m} options are supported on the SPARC: 18218 18219@table @gcctabopt 18220@item -mno-app-regs 18221@itemx -mapp-regs 18222@opindex mno-app-regs 18223@opindex mapp-regs 18224Specify @option{-mapp-regs} to generate output using the global registers 182252 through 4, which the SPARC SVR4 ABI reserves for applications. This 18226is the default. 18227 18228To be fully SVR4 ABI compliant at the cost of some performance loss, 18229specify @option{-mno-app-regs}. You should compile libraries and system 18230software with this option. 18231 18232@item -mflat 18233@itemx -mno-flat 18234@opindex mflat 18235@opindex mno-flat 18236With @option{-mflat}, the compiler does not generate save/restore instructions 18237and uses a ``flat'' or single register window model. This model is compatible 18238with the regular register window model. The local registers and the input 18239registers (0--5) are still treated as ``call-saved'' registers and will be 18240saved on the stack as needed. 18241 18242With @option{-mno-flat} (the default), the compiler generates save/restore 18243instructions (except for leaf functions). This is the normal operating mode. 18244 18245@item -mfpu 18246@itemx -mhard-float 18247@opindex mfpu 18248@opindex mhard-float 18249Generate output containing floating-point instructions. This is the 18250default. 18251 18252@item -mno-fpu 18253@itemx -msoft-float 18254@opindex mno-fpu 18255@opindex msoft-float 18256Generate output containing library calls for floating point. 18257@strong{Warning:} the requisite libraries are not available for all SPARC 18258targets. Normally the facilities of the machine's usual C compiler are 18259used, but this cannot be done directly in cross-compilation. You must make 18260your own arrangements to provide suitable library functions for 18261cross-compilation. The embedded targets @samp{sparc-*-aout} and 18262@samp{sparclite-*-*} do provide software floating-point support. 18263 18264@option{-msoft-float} changes the calling convention in the output file; 18265therefore, it is only useful if you compile @emph{all} of a program with 18266this option. In particular, you need to compile @file{libgcc.a}, the 18267library that comes with GCC, with @option{-msoft-float} in order for 18268this to work. 18269 18270@item -mhard-quad-float 18271@opindex mhard-quad-float 18272Generate output containing quad-word (long double) floating-point 18273instructions. 18274 18275@item -msoft-quad-float 18276@opindex msoft-quad-float 18277Generate output containing library calls for quad-word (long double) 18278floating-point instructions. The functions called are those specified 18279in the SPARC ABI@. This is the default. 18280 18281As of this writing, there are no SPARC implementations that have hardware 18282support for the quad-word floating-point instructions. They all invoke 18283a trap handler for one of these instructions, and then the trap handler 18284emulates the effect of the instruction. Because of the trap handler overhead, 18285this is much slower than calling the ABI library routines. Thus the 18286@option{-msoft-quad-float} option is the default. 18287 18288@item -mno-unaligned-doubles 18289@itemx -munaligned-doubles 18290@opindex mno-unaligned-doubles 18291@opindex munaligned-doubles 18292Assume that doubles have 8-byte alignment. This is the default. 18293 18294With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte 18295alignment only if they are contained in another type, or if they have an 18296absolute address. Otherwise, it assumes they have 4-byte alignment. 18297Specifying this option avoids some rare compatibility problems with code 18298generated by other compilers. It is not the default because it results 18299in a performance loss, especially for floating-point code. 18300 18301@item -mno-faster-structs 18302@itemx -mfaster-structs 18303@opindex mno-faster-structs 18304@opindex mfaster-structs 18305With @option{-mfaster-structs}, the compiler assumes that structures 18306should have 8-byte alignment. This enables the use of pairs of 18307@code{ldd} and @code{std} instructions for copies in structure 18308assignment, in place of twice as many @code{ld} and @code{st} pairs. 18309However, the use of this changed alignment directly violates the SPARC 18310ABI@. Thus, it's intended only for use on targets where the developer 18311acknowledges that their resulting code will not be directly in line with 18312the rules of the ABI@. 18313 18314@item -mcpu=@var{cpu_type} 18315@opindex mcpu 18316Set the instruction set, register set, and instruction scheduling parameters 18317for machine type @var{cpu_type}. Supported values for @var{cpu_type} are 18318@samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc}, 18319@samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, 18320@samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, 18321@samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3}, 18322and @samp{niagara4}. 18323 18324Native Solaris and GNU/Linux toolchains also support the value @samp{native}, 18325which selects the best architecture option for the host processor. 18326@option{-mcpu=native} has no effect if GCC does not recognize 18327the processor. 18328 18329Default instruction scheduling parameters are used for values that select 18330an architecture and not an implementation. These are @samp{v7}, @samp{v8}, 18331@samp{sparclite}, @samp{sparclet}, @samp{v9}. 18332 18333Here is a list of each supported architecture and their supported 18334implementations. 18335 18336@table @asis 18337@item v7 18338cypress 18339 18340@item v8 18341supersparc, hypersparc, leon 18342 18343@item sparclite 18344f930, f934, sparclite86x 18345 18346@item sparclet 18347tsc701 18348 18349@item v9 18350ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4 18351@end table 18352 18353By default (unless configured otherwise), GCC generates code for the V7 18354variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler 18355additionally optimizes it for the Cypress CY7C602 chip, as used in the 18356SPARCStation/SPARCServer 3xx series. This is also appropriate for the older 18357SPARCStation 1, 2, IPX etc. 18358 18359With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC 18360architecture. The only difference from V7 code is that the compiler emits 18361the integer multiply and integer divide instructions which exist in SPARC-V8 18362but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally 18363optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and 183642000 series. 18365 18366With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of 18367the SPARC architecture. This adds the integer multiply, integer divide step 18368and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7. 18369With @option{-mcpu=f930}, the compiler additionally optimizes it for the 18370Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With 18371@option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu 18372MB86934 chip, which is the more recent SPARClite with FPU@. 18373 18374With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of 18375the SPARC architecture. This adds the integer multiply, multiply/accumulate, 18376integer divide step and scan (@code{ffs}) instructions which exist in SPARClet 18377but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally 18378optimizes it for the TEMIC SPARClet chip. 18379 18380With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC 18381architecture. This adds 64-bit integer and floating-point move instructions, 183823 additional floating-point condition code registers and conditional move 18383instructions. With @option{-mcpu=ultrasparc}, the compiler additionally 18384optimizes it for the Sun UltraSPARC I/II/IIi chips. With 18385@option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the 18386Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With 18387@option{-mcpu=niagara}, the compiler additionally optimizes it for 18388Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler 18389additionally optimizes it for Sun UltraSPARC T2 chips. With 18390@option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun 18391UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler 18392additionally optimizes it for Sun UltraSPARC T4 chips. 18393 18394@item -mtune=@var{cpu_type} 18395@opindex mtune 18396Set the instruction scheduling parameters for machine type 18397@var{cpu_type}, but do not set the instruction set or register set that the 18398option @option{-mcpu=@var{cpu_type}} would. 18399 18400The same values for @option{-mcpu=@var{cpu_type}} can be used for 18401@option{-mtune=@var{cpu_type}}, but the only useful values are those 18402that select a particular CPU implementation. Those are @samp{cypress}, 18403@samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934}, 18404@samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3}, 18405@samp{niagara}, @samp{niagara2}, @samp{niagara3} and @samp{niagara4}. With 18406native Solaris and GNU/Linux toolchains, @samp{native} can also be used. 18407 18408@item -mv8plus 18409@itemx -mno-v8plus 18410@opindex mv8plus 18411@opindex mno-v8plus 18412With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The 18413difference from the V8 ABI is that the global and out registers are 18414considered 64 bits wide. This is enabled by default on Solaris in 32-bit 18415mode for all SPARC-V9 processors. 18416 18417@item -mvis 18418@itemx -mno-vis 18419@opindex mvis 18420@opindex mno-vis 18421With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC 18422Visual Instruction Set extensions. The default is @option{-mno-vis}. 18423 18424@item -mvis2 18425@itemx -mno-vis2 18426@opindex mvis2 18427@opindex mno-vis2 18428With @option{-mvis2}, GCC generates code that takes advantage of 18429version 2.0 of the UltraSPARC Visual Instruction Set extensions. The 18430default is @option{-mvis2} when targetting a cpu that supports such 18431instructions, such as UltraSPARC-III and later. Setting @option{-mvis2} 18432also sets @option{-mvis}. 18433 18434@item -mvis3 18435@itemx -mno-vis3 18436@opindex mvis3 18437@opindex mno-vis3 18438With @option{-mvis3}, GCC generates code that takes advantage of 18439version 3.0 of the UltraSPARC Visual Instruction Set extensions. The 18440default is @option{-mvis3} when targetting a cpu that supports such 18441instructions, such as niagara-3 and later. Setting @option{-mvis3} 18442also sets @option{-mvis2} and @option{-mvis}. 18443 18444@item -mpopc 18445@itemx -mno-popc 18446@opindex mpopc 18447@opindex mno-popc 18448With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC 18449population count instruction. The default is @option{-mpopc} 18450when targetting a cpu that supports such instructions, such as Niagara-2 and 18451later. 18452 18453@item -mfmaf 18454@itemx -mno-fmaf 18455@opindex mfmaf 18456@opindex mno-fmaf 18457With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC 18458Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf} 18459when targetting a cpu that supports such instructions, such as Niagara-3 and 18460later. 18461 18462@item -mfix-at697f 18463@opindex mfix-at697f 18464Enable the documented workaround for the single erratum of the Atmel AT697F 18465processor (which corresponds to erratum #13 of the AT697E processor). 18466@end table 18467 18468These @samp{-m} options are supported in addition to the above 18469on SPARC-V9 processors in 64-bit environments: 18470 18471@table @gcctabopt 18472@item -mlittle-endian 18473@opindex mlittle-endian 18474Generate code for a processor running in little-endian mode. It is only 18475available for a few configurations and most notably not on Solaris and Linux. 18476 18477@item -m32 18478@itemx -m64 18479@opindex m32 18480@opindex m64 18481Generate code for a 32-bit or 64-bit environment. 18482The 32-bit environment sets int, long and pointer to 32 bits. 18483The 64-bit environment sets int to 32 bits and long and pointer 18484to 64 bits. 18485 18486@item -mcmodel=@var{which} 18487@opindex mcmodel 18488Set the code model to one of 18489 18490@table @samp 18491@item medlow 18492The Medium/Low code model: 64-bit addresses, programs 18493must be linked in the low 32 bits of memory. Programs can be statically 18494or dynamically linked. 18495 18496@item medmid 18497The Medium/Middle code model: 64-bit addresses, programs 18498must be linked in the low 44 bits of memory, the text and data segments must 18499be less than 2GB in size and the data segment must be located within 2GB of 18500the text segment. 18501 18502@item medany 18503The Medium/Anywhere code model: 64-bit addresses, programs 18504may be linked anywhere in memory, the text and data segments must be less 18505than 2GB in size and the data segment must be located within 2GB of the 18506text segment. 18507 18508@item embmedany 18509The Medium/Anywhere code model for embedded systems: 1851064-bit addresses, the text and data segments must be less than 2GB in 18511size, both starting anywhere in memory (determined at link time). The 18512global register %g4 points to the base of the data segment. Programs 18513are statically linked and PIC is not supported. 18514@end table 18515 18516@item -mmemory-model=@var{mem-model} 18517@opindex mmemory-model 18518Set the memory model in force on the processor to one of 18519 18520@table @samp 18521@item default 18522The default memory model for the processor and operating system. 18523 18524@item rmo 18525Relaxed Memory Order 18526 18527@item pso 18528Partial Store Order 18529 18530@item tso 18531Total Store Order 18532 18533@item sc 18534Sequential Consistency 18535@end table 18536 18537These memory models are formally defined in Appendix D of the Sparc V9 18538architecture manual, as set in the processor's @code{PSTATE.MM} field. 18539 18540@item -mstack-bias 18541@itemx -mno-stack-bias 18542@opindex mstack-bias 18543@opindex mno-stack-bias 18544With @option{-mstack-bias}, GCC assumes that the stack pointer, and 18545frame pointer if present, are offset by @minus{}2047 which must be added back 18546when making stack frame references. This is the default in 64-bit mode. 18547Otherwise, assume no such offset is present. 18548@end table 18549 18550@node SPU Options 18551@subsection SPU Options 18552@cindex SPU options 18553 18554These @samp{-m} options are supported on the SPU: 18555 18556@table @gcctabopt 18557@item -mwarn-reloc 18558@itemx -merror-reloc 18559@opindex mwarn-reloc 18560@opindex merror-reloc 18561 18562The loader for SPU does not handle dynamic relocations. By default, GCC 18563will give an error when it generates code that requires a dynamic 18564relocation. @option{-mno-error-reloc} disables the error, 18565@option{-mwarn-reloc} will generate a warning instead. 18566 18567@item -msafe-dma 18568@itemx -munsafe-dma 18569@opindex msafe-dma 18570@opindex munsafe-dma 18571 18572Instructions that initiate or test completion of DMA must not be 18573reordered with respect to loads and stores of the memory that is being 18574accessed. Users typically address this problem using the volatile 18575keyword, but that can lead to inefficient code in places where the 18576memory is known to not change. Rather than mark the memory as volatile 18577we treat the DMA instructions as potentially effecting all memory. With 18578@option{-munsafe-dma} users must use the volatile keyword to protect 18579memory accesses. 18580 18581@item -mbranch-hints 18582@opindex mbranch-hints 18583 18584By default, GCC will generate a branch hint instruction to avoid 18585pipeline stalls for always taken or probably taken branches. A hint 18586will not be generated closer than 8 instructions away from its branch. 18587There is little reason to disable them, except for debugging purposes, 18588or to make an object a little bit smaller. 18589 18590@item -msmall-mem 18591@itemx -mlarge-mem 18592@opindex msmall-mem 18593@opindex mlarge-mem 18594 18595By default, GCC generates code assuming that addresses are never larger 18596than 18 bits. With @option{-mlarge-mem} code is generated that assumes 18597a full 32-bit address. 18598 18599@item -mstdmain 18600@opindex mstdmain 18601 18602By default, GCC links against startup code that assumes the SPU-style 18603main function interface (which has an unconventional parameter list). 18604With @option{-mstdmain}, GCC will link your program against startup 18605code that assumes a C99-style interface to @code{main}, including a 18606local copy of @code{argv} strings. 18607 18608@item -mfixed-range=@var{register-range} 18609@opindex mfixed-range 18610Generate code treating the given register range as fixed registers. 18611A fixed register is one that the register allocator can not use. This is 18612useful when compiling kernel code. A register range is specified as 18613two registers separated by a dash. Multiple register ranges can be 18614specified separated by a comma. 18615 18616@item -mea32 18617@itemx -mea64 18618@opindex mea32 18619@opindex mea64 18620Compile code assuming that pointers to the PPU address space accessed 18621via the @code{__ea} named address space qualifier are either 32 or 64 18622bits wide. The default is 32 bits. As this is an ABI changing option, 18623all object code in an executable must be compiled with the same setting. 18624 18625@item -maddress-space-conversion 18626@itemx -mno-address-space-conversion 18627@opindex maddress-space-conversion 18628@opindex mno-address-space-conversion 18629Allow/disallow treating the @code{__ea} address space as superset 18630of the generic address space. This enables explicit type casts 18631between @code{__ea} and generic pointer as well as implicit 18632conversions of generic pointers to @code{__ea} pointers. The 18633default is to allow address space pointer conversions. 18634 18635@item -mcache-size=@var{cache-size} 18636@opindex mcache-size 18637This option controls the version of libgcc that the compiler links to an 18638executable and selects a software-managed cache for accessing variables 18639in the @code{__ea} address space with a particular cache size. Possible 18640options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64} 18641and @samp{128}. The default cache size is 64KB. 18642 18643@item -matomic-updates 18644@itemx -mno-atomic-updates 18645@opindex matomic-updates 18646@opindex mno-atomic-updates 18647This option controls the version of libgcc that the compiler links to an 18648executable and selects whether atomic updates to the software-managed 18649cache of PPU-side variables are used. If you use atomic updates, changes 18650to a PPU variable from SPU code using the @code{__ea} named address space 18651qualifier will not interfere with changes to other PPU variables residing 18652in the same cache line from PPU code. If you do not use atomic updates, 18653such interference may occur; however, writing back cache lines will be 18654more efficient. The default behavior is to use atomic updates. 18655 18656@item -mdual-nops 18657@itemx -mdual-nops=@var{n} 18658@opindex mdual-nops 18659By default, GCC will insert nops to increase dual issue when it expects 18660it to increase performance. @var{n} can be a value from 0 to 10. A 18661smaller @var{n} will insert fewer nops. 10 is the default, 0 is the 18662same as @option{-mno-dual-nops}. Disabled with @option{-Os}. 18663 18664@item -mhint-max-nops=@var{n} 18665@opindex mhint-max-nops 18666Maximum number of nops to insert for a branch hint. A branch hint must 18667be at least 8 instructions away from the branch it is effecting. GCC 18668will insert up to @var{n} nops to enforce this, otherwise it will not 18669generate the branch hint. 18670 18671@item -mhint-max-distance=@var{n} 18672@opindex mhint-max-distance 18673The encoding of the branch hint instruction limits the hint to be within 18674256 instructions of the branch it is effecting. By default, GCC makes 18675sure it is within 125. 18676 18677@item -msafe-hints 18678@opindex msafe-hints 18679Work around a hardware bug that causes the SPU to stall indefinitely. 18680By default, GCC will insert the @code{hbrp} instruction to make sure 18681this stall won't happen. 18682 18683@end table 18684 18685@node System V Options 18686@subsection Options for System V 18687 18688These additional options are available on System V Release 4 for 18689compatibility with other compilers on those systems: 18690 18691@table @gcctabopt 18692@item -G 18693@opindex G 18694Create a shared object. 18695It is recommended that @option{-symbolic} or @option{-shared} be used instead. 18696 18697@item -Qy 18698@opindex Qy 18699Identify the versions of each tool used by the compiler, in a 18700@code{.ident} assembler directive in the output. 18701 18702@item -Qn 18703@opindex Qn 18704Refrain from adding @code{.ident} directives to the output file (this is 18705the default). 18706 18707@item -YP,@var{dirs} 18708@opindex YP 18709Search the directories @var{dirs}, and no others, for libraries 18710specified with @option{-l}. 18711 18712@item -Ym,@var{dir} 18713@opindex Ym 18714Look in the directory @var{dir} to find the M4 preprocessor. 18715The assembler uses this option. 18716@c This is supposed to go with a -Yd for predefined M4 macro files, but 18717@c the generic assembler that comes with Solaris takes just -Ym. 18718@end table 18719 18720@node TILE-Gx Options 18721@subsection TILE-Gx Options 18722@cindex TILE-Gx options 18723 18724These @samp{-m} options are supported on the TILE-Gx: 18725 18726@table @gcctabopt 18727@item -mcpu=@var{name} 18728@opindex mcpu 18729Selects the type of CPU to be targeted. Currently the only supported 18730type is @samp{tilegx}. 18731 18732@item -m32 18733@itemx -m64 18734@opindex m32 18735@opindex m64 18736Generate code for a 32-bit or 64-bit environment. The 32-bit 18737environment sets int, long, and pointer to 32 bits. The 64-bit 18738environment sets int to 32 bits and long and pointer to 64 bits. 18739@end table 18740 18741@node TILEPro Options 18742@subsection TILEPro Options 18743@cindex TILEPro options 18744 18745These @samp{-m} options are supported on the TILEPro: 18746 18747@table @gcctabopt 18748@item -mcpu=@var{name} 18749@opindex mcpu 18750Selects the type of CPU to be targeted. Currently the only supported 18751type is @samp{tilepro}. 18752 18753@item -m32 18754@opindex m32 18755Generate code for a 32-bit environment, which sets int, long, and 18756pointer to 32 bits. This is the only supported behavior so the flag 18757is essentially ignored. 18758@end table 18759 18760@node V850 Options 18761@subsection V850 Options 18762@cindex V850 Options 18763 18764These @samp{-m} options are defined for V850 implementations: 18765 18766@table @gcctabopt 18767@item -mlong-calls 18768@itemx -mno-long-calls 18769@opindex mlong-calls 18770@opindex mno-long-calls 18771Treat all calls as being far away (near). If calls are assumed to be 18772far away, the compiler will always load the functions address up into a 18773register, and call indirect through the pointer. 18774 18775@item -mno-ep 18776@itemx -mep 18777@opindex mno-ep 18778@opindex mep 18779Do not optimize (do optimize) basic blocks that use the same index 18780pointer 4 or more times to copy pointer into the @code{ep} register, and 18781use the shorter @code{sld} and @code{sst} instructions. The @option{-mep} 18782option is on by default if you optimize. 18783 18784@item -mno-prolog-function 18785@itemx -mprolog-function 18786@opindex mno-prolog-function 18787@opindex mprolog-function 18788Do not use (do use) external functions to save and restore registers 18789at the prologue and epilogue of a function. The external functions 18790are slower, but use less code space if more than one function saves 18791the same number of registers. The @option{-mprolog-function} option 18792is on by default if you optimize. 18793 18794@item -mspace 18795@opindex mspace 18796Try to make the code as small as possible. At present, this just turns 18797on the @option{-mep} and @option{-mprolog-function} options. 18798 18799@item -mtda=@var{n} 18800@opindex mtda 18801Put static or global variables whose size is @var{n} bytes or less into 18802the tiny data area that register @code{ep} points to. The tiny data 18803area can hold up to 256 bytes in total (128 bytes for byte references). 18804 18805@item -msda=@var{n} 18806@opindex msda 18807Put static or global variables whose size is @var{n} bytes or less into 18808the small data area that register @code{gp} points to. The small data 18809area can hold up to 64 kilobytes. 18810 18811@item -mzda=@var{n} 18812@opindex mzda 18813Put static or global variables whose size is @var{n} bytes or less into 18814the first 32 kilobytes of memory. 18815 18816@item -mv850 18817@opindex mv850 18818Specify that the target processor is the V850. 18819 18820@item -mbig-switch 18821@opindex mbig-switch 18822Generate code suitable for big switch tables. Use this option only if 18823the assembler/linker complain about out of range branches within a switch 18824table. 18825 18826@item -mapp-regs 18827@opindex mapp-regs 18828This option will cause r2 and r5 to be used in the code generated by 18829the compiler. This setting is the default. 18830 18831@item -mno-app-regs 18832@opindex mno-app-regs 18833This option will cause r2 and r5 to be treated as fixed registers. 18834 18835@item -mv850e2v3 18836@opindex mv850e2v3 18837Specify that the target processor is the V850E2V3. The preprocessor 18838constants @samp{__v850e2v3__} will be defined if 18839this option is used. 18840 18841@item -mv850e2 18842@opindex mv850e2 18843Specify that the target processor is the V850E2. The preprocessor 18844constants @samp{__v850e2__} will be defined if this option is used. 18845 18846@item -mv850e1 18847@opindex mv850e1 18848Specify that the target processor is the V850E1. The preprocessor 18849constants @samp{__v850e1__} and @samp{__v850e__} will be defined if 18850this option is used. 18851 18852@item -mv850es 18853@opindex mv850es 18854Specify that the target processor is the V850ES. This is an alias for 18855the @option{-mv850e1} option. 18856 18857@item -mv850e 18858@opindex mv850e 18859Specify that the target processor is the V850E@. The preprocessor 18860constant @samp{__v850e__} will be defined if this option is used. 18861 18862If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1} 18863nor @option{-mv850e2} nor @option{-mv850e2v3} 18864are defined then a default target processor will be chosen and the 18865relevant @samp{__v850*__} preprocessor constant will be defined. 18866 18867The preprocessor constants @samp{__v850} and @samp{__v851__} are always 18868defined, regardless of which processor variant is the target. 18869 18870@item -mdisable-callt 18871@opindex mdisable-callt 18872This option will suppress generation of the CALLT instruction for the 18873v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is 18874@option{-mno-disable-callt} which allows the CALLT instruction to be used. 18875 18876@end table 18877 18878@node VAX Options 18879@subsection VAX Options 18880@cindex VAX options 18881 18882These @samp{-m} options are defined for the VAX: 18883 18884@table @gcctabopt 18885@item -munix 18886@opindex munix 18887Do not output certain jump instructions (@code{aobleq} and so on) 18888that the Unix assembler for the VAX cannot handle across long 18889ranges. 18890 18891@item -mgnu 18892@opindex mgnu 18893Do output those jump instructions, on the assumption that you 18894will assemble with the GNU assembler. 18895 18896@item -mg 18897@opindex mg 18898Output code for G-format floating-point numbers instead of D-format. 18899@end table 18900 18901@node VxWorks Options 18902@subsection VxWorks Options 18903@cindex VxWorks Options 18904 18905The options in this section are defined for all VxWorks targets. 18906Options specific to the target hardware are listed with the other 18907options for that target. 18908 18909@table @gcctabopt 18910@item -mrtp 18911@opindex mrtp 18912GCC can generate code for both VxWorks kernels and real time processes 18913(RTPs). This option switches from the former to the latter. It also 18914defines the preprocessor macro @code{__RTP__}. 18915 18916@item -non-static 18917@opindex non-static 18918Link an RTP executable against shared libraries rather than static 18919libraries. The options @option{-static} and @option{-shared} can 18920also be used for RTPs (@pxref{Link Options}); @option{-static} 18921is the default. 18922 18923@item -Bstatic 18924@itemx -Bdynamic 18925@opindex Bstatic 18926@opindex Bdynamic 18927These options are passed down to the linker. They are defined for 18928compatibility with Diab. 18929 18930@item -Xbind-lazy 18931@opindex Xbind-lazy 18932Enable lazy binding of function calls. This option is equivalent to 18933@option{-Wl,-z,now} and is defined for compatibility with Diab. 18934 18935@item -Xbind-now 18936@opindex Xbind-now 18937Disable lazy binding of function calls. This option is the default and 18938is defined for compatibility with Diab. 18939@end table 18940 18941@node x86-64 Options 18942@subsection x86-64 Options 18943@cindex x86-64 options 18944 18945These are listed under @xref{i386 and x86-64 Options}. 18946 18947@node Xstormy16 Options 18948@subsection Xstormy16 Options 18949@cindex Xstormy16 Options 18950 18951These options are defined for Xstormy16: 18952 18953@table @gcctabopt 18954@item -msim 18955@opindex msim 18956Choose startup files and linker script suitable for the simulator. 18957@end table 18958 18959@node Xtensa Options 18960@subsection Xtensa Options 18961@cindex Xtensa Options 18962 18963These options are supported for Xtensa targets: 18964 18965@table @gcctabopt 18966@item -mconst16 18967@itemx -mno-const16 18968@opindex mconst16 18969@opindex mno-const16 18970Enable or disable use of @code{CONST16} instructions for loading 18971constant values. The @code{CONST16} instruction is currently not a 18972standard option from Tensilica. When enabled, @code{CONST16} 18973instructions are always used in place of the standard @code{L32R} 18974instructions. The use of @code{CONST16} is enabled by default only if 18975the @code{L32R} instruction is not available. 18976 18977@item -mfused-madd 18978@itemx -mno-fused-madd 18979@opindex mfused-madd 18980@opindex mno-fused-madd 18981Enable or disable use of fused multiply/add and multiply/subtract 18982instructions in the floating-point option. This has no effect if the 18983floating-point option is not also enabled. Disabling fused multiply/add 18984and multiply/subtract instructions forces the compiler to use separate 18985instructions for the multiply and add/subtract operations. This may be 18986desirable in some cases where strict IEEE 754-compliant results are 18987required: the fused multiply add/subtract instructions do not round the 18988intermediate result, thereby producing results with @emph{more} bits of 18989precision than specified by the IEEE standard. Disabling fused multiply 18990add/subtract instructions also ensures that the program output is not 18991sensitive to the compiler's ability to combine multiply and add/subtract 18992operations. 18993 18994@item -mserialize-volatile 18995@itemx -mno-serialize-volatile 18996@opindex mserialize-volatile 18997@opindex mno-serialize-volatile 18998When this option is enabled, GCC inserts @code{MEMW} instructions before 18999@code{volatile} memory references to guarantee sequential consistency. 19000The default is @option{-mserialize-volatile}. Use 19001@option{-mno-serialize-volatile} to omit the @code{MEMW} instructions. 19002 19003@item -mforce-no-pic 19004@opindex mforce-no-pic 19005For targets, like GNU/Linux, where all user-mode Xtensa code must be 19006position-independent code (PIC), this option disables PIC for compiling 19007kernel code. 19008 19009@item -mtext-section-literals 19010@itemx -mno-text-section-literals 19011@opindex mtext-section-literals 19012@opindex mno-text-section-literals 19013Control the treatment of literal pools. The default is 19014@option{-mno-text-section-literals}, which places literals in a separate 19015section in the output file. This allows the literal pool to be placed 19016in a data RAM/ROM, and it also allows the linker to combine literal 19017pools from separate object files to remove redundant literals and 19018improve code size. With @option{-mtext-section-literals}, the literals 19019are interspersed in the text section in order to keep them as close as 19020possible to their references. This may be necessary for large assembly 19021files. 19022 19023@item -mtarget-align 19024@itemx -mno-target-align 19025@opindex mtarget-align 19026@opindex mno-target-align 19027When this option is enabled, GCC instructs the assembler to 19028automatically align instructions to reduce branch penalties at the 19029expense of some code density. The assembler attempts to widen density 19030instructions to align branch targets and the instructions following call 19031instructions. If there are not enough preceding safe density 19032instructions to align a target, no widening will be performed. The 19033default is @option{-mtarget-align}. These options do not affect the 19034treatment of auto-aligned instructions like @code{LOOP}, which the 19035assembler will always align, either by widening density instructions or 19036by inserting no-op instructions. 19037 19038@item -mlongcalls 19039@itemx -mno-longcalls 19040@opindex mlongcalls 19041@opindex mno-longcalls 19042When this option is enabled, GCC instructs the assembler to translate 19043direct calls to indirect calls unless it can determine that the target 19044of a direct call is in the range allowed by the call instruction. This 19045translation typically occurs for calls to functions in other source 19046files. Specifically, the assembler translates a direct @code{CALL} 19047instruction into an @code{L32R} followed by a @code{CALLX} instruction. 19048The default is @option{-mno-longcalls}. This option should be used in 19049programs where the call target can potentially be out of range. This 19050option is implemented in the assembler, not the compiler, so the 19051assembly code generated by GCC will still show direct call 19052instructions---look at the disassembled object code to see the actual 19053instructions. Note that the assembler will use an indirect call for 19054every cross-file call, not just those that really will be out of range. 19055@end table 19056 19057@node zSeries Options 19058@subsection zSeries Options 19059@cindex zSeries options 19060 19061These are listed under @xref{S/390 and zSeries Options}. 19062 19063@node Code Gen Options 19064@section Options for Code Generation Conventions 19065@cindex code generation conventions 19066@cindex options, code generation 19067@cindex run-time options 19068 19069These machine-independent options control the interface conventions 19070used in code generation. 19071 19072Most of them have both positive and negative forms; the negative form 19073of @option{-ffoo} would be @option{-fno-foo}. In the table below, only 19074one of the forms is listed---the one that is not the default. You 19075can figure out the other form by either removing @samp{no-} or adding 19076it. 19077 19078@table @gcctabopt 19079@item -fbounds-check 19080@opindex fbounds-check 19081For front ends that support it, generate additional code to check that 19082indices used to access arrays are within the declared range. This is 19083currently only supported by the Java and Fortran front ends, where 19084this option defaults to true and false respectively. 19085 19086@item -ftrapv 19087@opindex ftrapv 19088This option generates traps for signed overflow on addition, subtraction, 19089multiplication operations. 19090 19091@item -fwrapv 19092@opindex fwrapv 19093This option instructs the compiler to assume that signed arithmetic 19094overflow of addition, subtraction and multiplication wraps around 19095using twos-complement representation. This flag enables some optimizations 19096and disables others. This option is enabled by default for the Java 19097front end, as required by the Java language specification. 19098 19099@item -fexceptions 19100@opindex fexceptions 19101Enable exception handling. Generates extra code needed to propagate 19102exceptions. For some targets, this implies GCC will generate frame 19103unwind information for all functions, which can produce significant data 19104size overhead, although it does not affect execution. If you do not 19105specify this option, GCC will enable it by default for languages like 19106C++ that normally require exception handling, and disable it for 19107languages like C that do not normally require it. However, you may need 19108to enable this option when compiling C code that needs to interoperate 19109properly with exception handlers written in C++. You may also wish to 19110disable this option if you are compiling older C++ programs that don't 19111use exception handling. 19112 19113@item -fnon-call-exceptions 19114@opindex fnon-call-exceptions 19115Generate code that allows trapping instructions to throw exceptions. 19116Note that this requires platform-specific runtime support that does 19117not exist everywhere. Moreover, it only allows @emph{trapping} 19118instructions to throw exceptions, i.e.@: memory references or floating-point 19119instructions. It does not allow exceptions to be thrown from 19120arbitrary signal handlers such as @code{SIGALRM}. 19121 19122@item -funwind-tables 19123@opindex funwind-tables 19124Similar to @option{-fexceptions}, except that it will just generate any needed 19125static data, but will not affect the generated code in any other way. 19126You will normally not enable this option; instead, a language processor 19127that needs this handling would enable it on your behalf. 19128 19129@item -fasynchronous-unwind-tables 19130@opindex fasynchronous-unwind-tables 19131Generate unwind table in dwarf2 format, if supported by target machine. The 19132table is exact at each instruction boundary, so it can be used for stack 19133unwinding from asynchronous events (such as debugger or garbage collector). 19134 19135@item -fpcc-struct-return 19136@opindex fpcc-struct-return 19137Return ``short'' @code{struct} and @code{union} values in memory like 19138longer ones, rather than in registers. This convention is less 19139efficient, but it has the advantage of allowing intercallability between 19140GCC-compiled files and files compiled with other compilers, particularly 19141the Portable C Compiler (pcc). 19142 19143The precise convention for returning structures in memory depends 19144on the target configuration macros. 19145 19146Short structures and unions are those whose size and alignment match 19147that of some integer type. 19148 19149@strong{Warning:} code compiled with the @option{-fpcc-struct-return} 19150switch is not binary compatible with code compiled with the 19151@option{-freg-struct-return} switch. 19152Use it to conform to a non-default application binary interface. 19153 19154@item -freg-struct-return 19155@opindex freg-struct-return 19156Return @code{struct} and @code{union} values in registers when possible. 19157This is more efficient for small structures than 19158@option{-fpcc-struct-return}. 19159 19160If you specify neither @option{-fpcc-struct-return} nor 19161@option{-freg-struct-return}, GCC defaults to whichever convention is 19162standard for the target. If there is no standard convention, GCC 19163defaults to @option{-fpcc-struct-return}, except on targets where GCC is 19164the principal compiler. In those cases, we can choose the standard, and 19165we chose the more efficient register return alternative. 19166 19167@strong{Warning:} code compiled with the @option{-freg-struct-return} 19168switch is not binary compatible with code compiled with the 19169@option{-fpcc-struct-return} switch. 19170Use it to conform to a non-default application binary interface. 19171 19172@item -fshort-enums 19173@opindex fshort-enums 19174Allocate to an @code{enum} type only as many bytes as it needs for the 19175declared range of possible values. Specifically, the @code{enum} type 19176will be equivalent to the smallest integer type that has enough room. 19177 19178@strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate 19179code that is not binary compatible with code generated without that switch. 19180Use it to conform to a non-default application binary interface. 19181 19182@item -fshort-double 19183@opindex fshort-double 19184Use the same size for @code{double} as for @code{float}. 19185 19186@strong{Warning:} the @option{-fshort-double} switch causes GCC to generate 19187code that is not binary compatible with code generated without that switch. 19188Use it to conform to a non-default application binary interface. 19189 19190@item -fshort-wchar 19191@opindex fshort-wchar 19192Override the underlying type for @samp{wchar_t} to be @samp{short 19193unsigned int} instead of the default for the target. This option is 19194useful for building programs to run under WINE@. 19195 19196@strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate 19197code that is not binary compatible with code generated without that switch. 19198Use it to conform to a non-default application binary interface. 19199 19200@item -fno-common 19201@opindex fno-common 19202In C code, controls the placement of uninitialized global variables. 19203Unix C compilers have traditionally permitted multiple definitions of 19204such variables in different compilation units by placing the variables 19205in a common block. 19206This is the behavior specified by @option{-fcommon}, and is the default 19207for GCC on most targets. 19208On the other hand, this behavior is not required by ISO C, and on some 19209targets may carry a speed or code size penalty on variable references. 19210The @option{-fno-common} option specifies that the compiler should place 19211uninitialized global variables in the data section of the object file, 19212rather than generating them as common blocks. 19213This has the effect that if the same variable is declared 19214(without @code{extern}) in two different compilations, 19215you will get a multiple-definition error when you link them. 19216In this case, you must compile with @option{-fcommon} instead. 19217Compiling with @option{-fno-common} is useful on targets for which 19218it provides better performance, or if you wish to verify that the 19219program will work on other systems that always treat uninitialized 19220variable declarations this way. 19221 19222@item -fno-ident 19223@opindex fno-ident 19224Ignore the @samp{#ident} directive. 19225 19226@item -finhibit-size-directive 19227@opindex finhibit-size-directive 19228Don't output a @code{.size} assembler directive, or anything else that 19229would cause trouble if the function is split in the middle, and the 19230two halves are placed at locations far apart in memory. This option is 19231used when compiling @file{crtstuff.c}; you should not need to use it 19232for anything else. 19233 19234@item -fverbose-asm 19235@opindex fverbose-asm 19236Put extra commentary information in the generated assembly code to 19237make it more readable. This option is generally only of use to those 19238who actually need to read the generated assembly code (perhaps while 19239debugging the compiler itself). 19240 19241@option{-fno-verbose-asm}, the default, causes the 19242extra information to be omitted and is useful when comparing two assembler 19243files. 19244 19245@item -frecord-gcc-switches 19246@opindex frecord-gcc-switches 19247This switch causes the command line that was used to invoke the 19248compiler to be recorded into the object file that is being created. 19249This switch is only implemented on some targets and the exact format 19250of the recording is target and binary file format dependent, but it 19251usually takes the form of a section containing ASCII text. This 19252switch is related to the @option{-fverbose-asm} switch, but that 19253switch only records information in the assembler output file as 19254comments, so it never reaches the object file. 19255See also @option{-grecord-gcc-switches} for another 19256way of storing compiler options into the object file. 19257 19258@item -fpic 19259@opindex fpic 19260@cindex global offset table 19261@cindex PIC 19262Generate position-independent code (PIC) suitable for use in a shared 19263library, if supported for the target machine. Such code accesses all 19264constant addresses through a global offset table (GOT)@. The dynamic 19265loader resolves the GOT entries when the program starts (the dynamic 19266loader is not part of GCC; it is part of the operating system). If 19267the GOT size for the linked executable exceeds a machine-specific 19268maximum size, you get an error message from the linker indicating that 19269@option{-fpic} does not work; in that case, recompile with @option{-fPIC} 19270instead. (These maximums are 8k on the SPARC and 32k 19271on the m68k and RS/6000. The 386 has no such limit.) 19272 19273Position-independent code requires special support, and therefore works 19274only on certain machines. For the 386, GCC supports PIC for System V 19275but not for the Sun 386i. Code generated for the IBM RS/6000 is always 19276position-independent. 19277 19278When this flag is set, the macros @code{__pic__} and @code{__PIC__} 19279are defined to 1. 19280 19281@item -fPIC 19282@opindex fPIC 19283If supported for the target machine, emit position-independent code, 19284suitable for dynamic linking and avoiding any limit on the size of the 19285global offset table. This option makes a difference on the m68k, 19286PowerPC and SPARC@. 19287 19288Position-independent code requires special support, and therefore works 19289only on certain machines. 19290 19291When this flag is set, the macros @code{__pic__} and @code{__PIC__} 19292are defined to 2. 19293 19294@item -fpie 19295@itemx -fPIE 19296@opindex fpie 19297@opindex fPIE 19298These options are similar to @option{-fpic} and @option{-fPIC}, but 19299generated position independent code can be only linked into executables. 19300Usually these options are used when @option{-pie} GCC option will be 19301used during linking. 19302 19303@option{-fpie} and @option{-fPIE} both define the macros 19304@code{__pie__} and @code{__PIE__}. The macros have the value 1 19305for @option{-fpie} and 2 for @option{-fPIE}. 19306 19307@item -fno-jump-tables 19308@opindex fno-jump-tables 19309Do not use jump tables for switch statements even where it would be 19310more efficient than other code generation strategies. This option is 19311of use in conjunction with @option{-fpic} or @option{-fPIC} for 19312building code that forms part of a dynamic linker and cannot 19313reference the address of a jump table. On some targets, jump tables 19314do not require a GOT and this option is not needed. 19315 19316@item -ffixed-@var{reg} 19317@opindex ffixed 19318Treat the register named @var{reg} as a fixed register; generated code 19319should never refer to it (except perhaps as a stack pointer, frame 19320pointer or in some other fixed role). 19321 19322@var{reg} must be the name of a register. The register names accepted 19323are machine-specific and are defined in the @code{REGISTER_NAMES} 19324macro in the machine description macro file. 19325 19326This flag does not have a negative form, because it specifies a 19327three-way choice. 19328 19329@item -fcall-used-@var{reg} 19330@opindex fcall-used 19331Treat the register named @var{reg} as an allocable register that is 19332clobbered by function calls. It may be allocated for temporaries or 19333variables that do not live across a call. Functions compiled this way 19334will not save and restore the register @var{reg}. 19335 19336It is an error to used this flag with the frame pointer or stack pointer. 19337Use of this flag for other registers that have fixed pervasive roles in 19338the machine's execution model will produce disastrous results. 19339 19340This flag does not have a negative form, because it specifies a 19341three-way choice. 19342 19343@item -fcall-saved-@var{reg} 19344@opindex fcall-saved 19345Treat the register named @var{reg} as an allocable register saved by 19346functions. It may be allocated even for temporaries or variables that 19347live across a call. Functions compiled this way will save and restore 19348the register @var{reg} if they use it. 19349 19350It is an error to used this flag with the frame pointer or stack pointer. 19351Use of this flag for other registers that have fixed pervasive roles in 19352the machine's execution model will produce disastrous results. 19353 19354A different sort of disaster will result from the use of this flag for 19355a register in which function values may be returned. 19356 19357This flag does not have a negative form, because it specifies a 19358three-way choice. 19359 19360@item -fpack-struct[=@var{n}] 19361@opindex fpack-struct 19362Without a value specified, pack all structure members together without 19363holes. When a value is specified (which must be a small power of two), pack 19364structure members according to this value, representing the maximum 19365alignment (that is, objects with default alignment requirements larger than 19366this will be output potentially unaligned at the next fitting location. 19367 19368@strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate 19369code that is not binary compatible with code generated without that switch. 19370Additionally, it makes the code suboptimal. 19371Use it to conform to a non-default application binary interface. 19372 19373@item -finstrument-functions 19374@opindex finstrument-functions 19375Generate instrumentation calls for entry and exit to functions. Just 19376after function entry and just before function exit, the following 19377profiling functions will be called with the address of the current 19378function and its call site. (On some platforms, 19379@code{__builtin_return_address} does not work beyond the current 19380function, so the call site information may not be available to the 19381profiling functions otherwise.) 19382 19383@smallexample 19384void __cyg_profile_func_enter (void *this_fn, 19385 void *call_site); 19386void __cyg_profile_func_exit (void *this_fn, 19387 void *call_site); 19388@end smallexample 19389 19390The first argument is the address of the start of the current function, 19391which may be looked up exactly in the symbol table. 19392 19393This instrumentation is also done for functions expanded inline in other 19394functions. The profiling calls will indicate where, conceptually, the 19395inline function is entered and exited. This means that addressable 19396versions of such functions must be available. If all your uses of a 19397function are expanded inline, this may mean an additional expansion of 19398code size. If you use @samp{extern inline} in your C code, an 19399addressable version of such functions must be provided. (This is 19400normally the case anyways, but if you get lucky and the optimizer always 19401expands the functions inline, you might have gotten away without 19402providing static copies.) 19403 19404A function may be given the attribute @code{no_instrument_function}, in 19405which case this instrumentation will not be done. This can be used, for 19406example, for the profiling functions listed above, high-priority 19407interrupt routines, and any functions from which the profiling functions 19408cannot safely be called (perhaps signal handlers, if the profiling 19409routines generate output or allocate memory). 19410 19411@item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} 19412@opindex finstrument-functions-exclude-file-list 19413 19414Set the list of functions that are excluded from instrumentation (see 19415the description of @code{-finstrument-functions}). If the file that 19416contains a function definition matches with one of @var{file}, then 19417that function is not instrumented. The match is done on substrings: 19418if the @var{file} parameter is a substring of the file name, it is 19419considered to be a match. 19420 19421For example: 19422 19423@smallexample 19424-finstrument-functions-exclude-file-list=/bits/stl,include/sys 19425@end smallexample 19426 19427@noindent 19428will exclude any inline function defined in files whose pathnames 19429contain @code{/bits/stl} or @code{include/sys}. 19430 19431If, for some reason, you want to include letter @code{','} in one of 19432@var{sym}, write @code{'\,'}. For example, 19433@code{-finstrument-functions-exclude-file-list='\,\,tmp'} 19434(note the single quote surrounding the option). 19435 19436@item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} 19437@opindex finstrument-functions-exclude-function-list 19438 19439This is similar to @code{-finstrument-functions-exclude-file-list}, 19440but this option sets the list of function names to be excluded from 19441instrumentation. The function name to be matched is its user-visible 19442name, such as @code{vector<int> blah(const vector<int> &)}, not the 19443internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The 19444match is done on substrings: if the @var{sym} parameter is a substring 19445of the function name, it is considered to be a match. For C99 and C++ 19446extended identifiers, the function name must be given in UTF-8, not 19447using universal character names. 19448 19449@item -fstack-check 19450@opindex fstack-check 19451Generate code to verify that you do not go beyond the boundary of the 19452stack. You should specify this flag if you are running in an 19453environment with multiple threads, but only rarely need to specify it in 19454a single-threaded environment since stack overflow is automatically 19455detected on nearly all systems if there is only one stack. 19456 19457Note that this switch does not actually cause checking to be done; the 19458operating system or the language runtime must do that. The switch causes 19459generation of code to ensure that they see the stack being extended. 19460 19461You can additionally specify a string parameter: @code{no} means no 19462checking, @code{generic} means force the use of old-style checking, 19463@code{specific} means use the best checking method and is equivalent 19464to bare @option{-fstack-check}. 19465 19466Old-style checking is a generic mechanism that requires no specific 19467target support in the compiler but comes with the following drawbacks: 19468 19469@enumerate 19470@item 19471Modified allocation strategy for large objects: they will always be 19472allocated dynamically if their size exceeds a fixed threshold. 19473 19474@item 19475Fixed limit on the size of the static frame of functions: when it is 19476topped by a particular function, stack checking is not reliable and 19477a warning is issued by the compiler. 19478 19479@item 19480Inefficiency: because of both the modified allocation strategy and the 19481generic implementation, the performances of the code are hampered. 19482@end enumerate 19483 19484Note that old-style stack checking is also the fallback method for 19485@code{specific} if no target support has been added in the compiler. 19486 19487@item -fstack-limit-register=@var{reg} 19488@itemx -fstack-limit-symbol=@var{sym} 19489@itemx -fno-stack-limit 19490@opindex fstack-limit-register 19491@opindex fstack-limit-symbol 19492@opindex fno-stack-limit 19493Generate code to ensure that the stack does not grow beyond a certain value, 19494either the value of a register or the address of a symbol. If the stack 19495would grow beyond the value, a signal is raised. For most targets, 19496the signal is raised before the stack overruns the boundary, so 19497it is possible to catch the signal without taking special precautions. 19498 19499For instance, if the stack starts at absolute address @samp{0x80000000} 19500and grows downwards, you can use the flags 19501@option{-fstack-limit-symbol=__stack_limit} and 19502@option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit 19503of 128KB@. Note that this may only work with the GNU linker. 19504 19505@item -fsplit-stack 19506@opindex fsplit-stack 19507Generate code to automatically split the stack before it overflows. 19508The resulting program has a discontiguous stack which can only 19509overflow if the program is unable to allocate any more memory. This 19510is most useful when running threaded programs, as it is no longer 19511necessary to calculate a good stack size to use for each thread. This 19512is currently only implemented for the i386 and x86_64 back ends running 19513GNU/Linux. 19514 19515When code compiled with @option{-fsplit-stack} calls code compiled 19516without @option{-fsplit-stack}, there may not be much stack space 19517available for the latter code to run. If compiling all code, 19518including library code, with @option{-fsplit-stack} is not an option, 19519then the linker can fix up these calls so that the code compiled 19520without @option{-fsplit-stack} always has a large stack. Support for 19521this is implemented in the gold linker in GNU binutils release 2.21 19522and later. 19523 19524@item -fleading-underscore 19525@opindex fleading-underscore 19526This option and its counterpart, @option{-fno-leading-underscore}, forcibly 19527change the way C symbols are represented in the object file. One use 19528is to help link with legacy assembly code. 19529 19530@strong{Warning:} the @option{-fleading-underscore} switch causes GCC to 19531generate code that is not binary compatible with code generated without that 19532switch. Use it to conform to a non-default application binary interface. 19533Not all targets provide complete support for this switch. 19534 19535@item -ftls-model=@var{model} 19536@opindex ftls-model 19537Alter the thread-local storage model to be used (@pxref{Thread-Local}). 19538The @var{model} argument should be one of @code{global-dynamic}, 19539@code{local-dynamic}, @code{initial-exec} or @code{local-exec}. 19540 19541The default without @option{-fpic} is @code{initial-exec}; with 19542@option{-fpic} the default is @code{global-dynamic}. 19543 19544@item -fvisibility=@var{default|internal|hidden|protected} 19545@opindex fvisibility 19546Set the default ELF image symbol visibility to the specified option---all 19547symbols will be marked with this unless overridden within the code. 19548Using this feature can very substantially improve linking and 19549load times of shared object libraries, produce more optimized 19550code, provide near-perfect API export and prevent symbol clashes. 19551It is @strong{strongly} recommended that you use this in any shared objects 19552you distribute. 19553 19554Despite the nomenclature, @code{default} always means public; i.e., 19555available to be linked against from outside the shared object. 19556@code{protected} and @code{internal} are pretty useless in real-world 19557usage so the only other commonly used option will be @code{hidden}. 19558The default if @option{-fvisibility} isn't specified is 19559@code{default}, i.e., make every 19560symbol public---this causes the same behavior as previous versions of 19561GCC@. 19562 19563A good explanation of the benefits offered by ensuring ELF 19564symbols have the correct visibility is given by ``How To Write 19565Shared Libraries'' by Ulrich Drepper (which can be found at 19566@w{@uref{http://people.redhat.com/~drepper/}})---however a superior 19567solution made possible by this option to marking things hidden when 19568the default is public is to make the default hidden and mark things 19569public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden} 19570and @code{__attribute__ ((visibility("default")))} instead of 19571@code{__declspec(dllexport)} you get almost identical semantics with 19572identical syntax. This is a great boon to those working with 19573cross-platform projects. 19574 19575For those adding visibility support to existing code, you may find 19576@samp{#pragma GCC visibility} of use. This works by you enclosing 19577the declarations you wish to set visibility for with (for example) 19578@samp{#pragma GCC visibility push(hidden)} and 19579@samp{#pragma GCC visibility pop}. 19580Bear in mind that symbol visibility should be viewed @strong{as 19581part of the API interface contract} and thus all new code should 19582always specify visibility when it is not the default; i.e., declarations 19583only for use within the local DSO should @strong{always} be marked explicitly 19584as hidden as so to avoid PLT indirection overheads---making this 19585abundantly clear also aids readability and self-documentation of the code. 19586Note that due to ISO C++ specification requirements, operator new and 19587operator delete must always be of default visibility. 19588 19589Be aware that headers from outside your project, in particular system 19590headers and headers from any other library you use, may not be 19591expecting to be compiled with visibility other than the default. You 19592may need to explicitly say @samp{#pragma GCC visibility push(default)} 19593before including any such headers. 19594 19595@samp{extern} declarations are not affected by @samp{-fvisibility}, so 19596a lot of code can be recompiled with @samp{-fvisibility=hidden} with 19597no modifications. However, this means that calls to @samp{extern} 19598functions with no explicit visibility will use the PLT, so it is more 19599effective to use @samp{__attribute ((visibility))} and/or 19600@samp{#pragma GCC visibility} to tell the compiler which @samp{extern} 19601declarations should be treated as hidden. 19602 19603Note that @samp{-fvisibility} does affect C++ vague linkage 19604entities. This means that, for instance, an exception class that will 19605be thrown between DSOs must be explicitly marked with default 19606visibility so that the @samp{type_info} nodes will be unified between 19607the DSOs. 19608 19609An overview of these techniques, their benefits and how to use them 19610is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}. 19611 19612@item -fstrict-volatile-bitfields 19613@opindex fstrict-volatile-bitfields 19614This option should be used if accesses to volatile bit-fields (or other 19615structure fields, although the compiler usually honors those types 19616anyway) should use a single access of the width of the 19617field's type, aligned to a natural alignment if possible. For 19618example, targets with memory-mapped peripheral registers might require 19619all such accesses to be 16 bits wide; with this flag the user could 19620declare all peripheral bit-fields as ``unsigned short'' (assuming short 19621is 16 bits on these targets) to force GCC to use 16-bit accesses 19622instead of, perhaps, a more efficient 32-bit access. 19623 19624If this option is disabled, the compiler will use the most efficient 19625instruction. In the previous example, that might be a 32-bit load 19626instruction, even though that will access bytes that do not contain 19627any portion of the bit-field, or memory-mapped registers unrelated to 19628the one being updated. 19629 19630If the target requires strict alignment, and honoring the field 19631type would require violating this alignment, a warning is issued. 19632If the field has @code{packed} attribute, the access is done without 19633honoring the field type. If the field doesn't have @code{packed} 19634attribute, the access is done honoring the field type. In both cases, 19635GCC assumes that the user knows something about the target hardware 19636that it is unaware of. 19637 19638The default value of this option is determined by the application binary 19639interface for the target processor. 19640 19641@end table 19642 19643@c man end 19644 19645@node Environment Variables 19646@section Environment Variables Affecting GCC 19647@cindex environment variables 19648 19649@c man begin ENVIRONMENT 19650This section describes several environment variables that affect how GCC 19651operates. Some of them work by specifying directories or prefixes to use 19652when searching for various kinds of files. Some are used to specify other 19653aspects of the compilation environment. 19654 19655Note that you can also specify places to search using options such as 19656@option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These 19657take precedence over places specified using environment variables, which 19658in turn take precedence over those specified by the configuration of GCC@. 19659@xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint, 19660GNU Compiler Collection (GCC) Internals}. 19661 19662@table @env 19663@item LANG 19664@itemx LC_CTYPE 19665@c @itemx LC_COLLATE 19666@itemx LC_MESSAGES 19667@c @itemx LC_MONETARY 19668@c @itemx LC_NUMERIC 19669@c @itemx LC_TIME 19670@itemx LC_ALL 19671@findex LANG 19672@findex LC_CTYPE 19673@c @findex LC_COLLATE 19674@findex LC_MESSAGES 19675@c @findex LC_MONETARY 19676@c @findex LC_NUMERIC 19677@c @findex LC_TIME 19678@findex LC_ALL 19679@cindex locale 19680These environment variables control the way that GCC uses 19681localization information which allows GCC to work with different 19682national conventions. GCC inspects the locale categories 19683@env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do 19684so. These locale categories can be set to any value supported by your 19685installation. A typical value is @samp{en_GB.UTF-8} for English in the United 19686Kingdom encoded in UTF-8. 19687 19688The @env{LC_CTYPE} environment variable specifies character 19689classification. GCC uses it to determine the character boundaries in 19690a string; this is needed for some multibyte encodings that contain quote 19691and escape characters that would otherwise be interpreted as a string 19692end or escape. 19693 19694The @env{LC_MESSAGES} environment variable specifies the language to 19695use in diagnostic messages. 19696 19697If the @env{LC_ALL} environment variable is set, it overrides the value 19698of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE} 19699and @env{LC_MESSAGES} default to the value of the @env{LANG} 19700environment variable. If none of these variables are set, GCC 19701defaults to traditional C English behavior. 19702 19703@item TMPDIR 19704@findex TMPDIR 19705If @env{TMPDIR} is set, it specifies the directory to use for temporary 19706files. GCC uses temporary files to hold the output of one stage of 19707compilation which is to be used as input to the next stage: for example, 19708the output of the preprocessor, which is the input to the compiler 19709proper. 19710 19711@item GCC_COMPARE_DEBUG 19712@findex GCC_COMPARE_DEBUG 19713Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing 19714@option{-fcompare-debug} to the compiler driver. See the documentation 19715of this option for more details. 19716 19717@item GCC_EXEC_PREFIX 19718@findex GCC_EXEC_PREFIX 19719If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the 19720names of the subprograms executed by the compiler. No slash is added 19721when this prefix is combined with the name of a subprogram, but you can 19722specify a prefix that ends with a slash if you wish. 19723 19724If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out 19725an appropriate prefix to use based on the pathname it was invoked with. 19726 19727If GCC cannot find the subprogram using the specified prefix, it 19728tries looking in the usual places for the subprogram. 19729 19730The default value of @env{GCC_EXEC_PREFIX} is 19731@file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to 19732the installed compiler. In many cases @var{prefix} is the value 19733of @code{prefix} when you ran the @file{configure} script. 19734 19735Other prefixes specified with @option{-B} take precedence over this prefix. 19736 19737This prefix is also used for finding files such as @file{crt0.o} that are 19738used for linking. 19739 19740In addition, the prefix is used in an unusual way in finding the 19741directories to search for header files. For each of the standard 19742directories whose name normally begins with @samp{/usr/local/lib/gcc} 19743(more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries 19744replacing that beginning with the specified prefix to produce an 19745alternate directory name. Thus, with @option{-Bfoo/}, GCC will search 19746@file{foo/bar} where it would normally search @file{/usr/local/lib/bar}. 19747These alternate directories are searched first; the standard directories 19748come next. If a standard directory begins with the configured 19749@var{prefix} then the value of @var{prefix} is replaced by 19750@env{GCC_EXEC_PREFIX} when looking for header files. 19751 19752@item COMPILER_PATH 19753@findex COMPILER_PATH 19754The value of @env{COMPILER_PATH} is a colon-separated list of 19755directories, much like @env{PATH}. GCC tries the directories thus 19756specified when searching for subprograms, if it can't find the 19757subprograms using @env{GCC_EXEC_PREFIX}. 19758 19759@item LIBRARY_PATH 19760@findex LIBRARY_PATH 19761The value of @env{LIBRARY_PATH} is a colon-separated list of 19762directories, much like @env{PATH}. When configured as a native compiler, 19763GCC tries the directories thus specified when searching for special 19764linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking 19765using GCC also uses these directories when searching for ordinary 19766libraries for the @option{-l} option (but directories specified with 19767@option{-L} come first). 19768 19769@item LANG 19770@findex LANG 19771@cindex locale definition 19772This variable is used to pass locale information to the compiler. One way in 19773which this information is used is to determine the character set to be used 19774when character literals, string literals and comments are parsed in C and C++. 19775When the compiler is configured to allow multibyte characters, 19776the following values for @env{LANG} are recognized: 19777 19778@table @samp 19779@item C-JIS 19780Recognize JIS characters. 19781@item C-SJIS 19782Recognize SJIS characters. 19783@item C-EUCJP 19784Recognize EUCJP characters. 19785@end table 19786 19787If @env{LANG} is not defined, or if it has some other value, then the 19788compiler will use mblen and mbtowc as defined by the default locale to 19789recognize and translate multibyte characters. 19790@end table 19791 19792@noindent 19793Some additional environments variables affect the behavior of the 19794preprocessor. 19795 19796@include cppenv.texi 19797 19798@c man end 19799 19800@node Precompiled Headers 19801@section Using Precompiled Headers 19802@cindex precompiled headers 19803@cindex speed of compilation 19804 19805Often large projects have many header files that are included in every 19806source file. The time the compiler takes to process these header files 19807over and over again can account for nearly all of the time required to 19808build the project. To make builds faster, GCC allows users to 19809`precompile' a header file; then, if builds can use the precompiled 19810header file they will be much faster. 19811 19812To create a precompiled header file, simply compile it as you would any 19813other file, if necessary using the @option{-x} option to make the driver 19814treat it as a C or C++ header file. You will probably want to use a 19815tool like @command{make} to keep the precompiled header up-to-date when 19816the headers it contains change. 19817 19818A precompiled header file will be searched for when @code{#include} is 19819seen in the compilation. As it searches for the included file 19820(@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the 19821compiler looks for a precompiled header in each directory just before it 19822looks for the include file in that directory. The name searched for is 19823the name specified in the @code{#include} with @samp{.gch} appended. If 19824the precompiled header file can't be used, it is ignored. 19825 19826For instance, if you have @code{#include "all.h"}, and you have 19827@file{all.h.gch} in the same directory as @file{all.h}, then the 19828precompiled header file will be used if possible, and the original 19829header will be used otherwise. 19830 19831Alternatively, you might decide to put the precompiled header file in a 19832directory and use @option{-I} to ensure that directory is searched 19833before (or instead of) the directory containing the original header. 19834Then, if you want to check that the precompiled header file is always 19835used, you can put a file of the same name as the original header in this 19836directory containing an @code{#error} command. 19837 19838This also works with @option{-include}. So yet another way to use 19839precompiled headers, good for projects not designed with precompiled 19840header files in mind, is to simply take most of the header files used by 19841a project, include them from another header file, precompile that header 19842file, and @option{-include} the precompiled header. If the header files 19843have guards against multiple inclusion, they will be skipped because 19844they've already been included (in the precompiled header). 19845 19846If you need to precompile the same header file for different 19847languages, targets, or compiler options, you can instead make a 19848@emph{directory} named like @file{all.h.gch}, and put each precompiled 19849header in the directory, perhaps using @option{-o}. It doesn't matter 19850what you call the files in the directory, every precompiled header in 19851the directory will be considered. The first precompiled header 19852encountered in the directory that is valid for this compilation will 19853be used; they're searched in no particular order. 19854 19855There are many other possibilities, limited only by your imagination, 19856good sense, and the constraints of your build system. 19857 19858A precompiled header file can be used only when these conditions apply: 19859 19860@itemize 19861@item 19862Only one precompiled header can be used in a particular compilation. 19863 19864@item 19865A precompiled header can't be used once the first C token is seen. You 19866can have preprocessor directives before a precompiled header; you can 19867even include a precompiled header from inside another header, so long as 19868there are no C tokens before the @code{#include}. 19869 19870@item 19871The precompiled header file must be produced for the same language as 19872the current compilation. You can't use a C precompiled header for a C++ 19873compilation. 19874 19875@item 19876The precompiled header file must have been produced by the same compiler 19877binary as the current compilation is using. 19878 19879@item 19880Any macros defined before the precompiled header is included must 19881either be defined in the same way as when the precompiled header was 19882generated, or must not affect the precompiled header, which usually 19883means that they don't appear in the precompiled header at all. 19884 19885The @option{-D} option is one way to define a macro before a 19886precompiled header is included; using a @code{#define} can also do it. 19887There are also some options that define macros implicitly, like 19888@option{-O} and @option{-Wdeprecated}; the same rule applies to macros 19889defined this way. 19890 19891@item If debugging information is output when using the precompiled 19892header, using @option{-g} or similar, the same kind of debugging information 19893must have been output when building the precompiled header. However, 19894a precompiled header built using @option{-g} can be used in a compilation 19895when no debugging information is being output. 19896 19897@item The same @option{-m} options must generally be used when building 19898and using the precompiled header. @xref{Submodel Options}, 19899for any cases where this rule is relaxed. 19900 19901@item Each of the following options must be the same when building and using 19902the precompiled header: 19903 19904@gccoptlist{-fexceptions} 19905 19906@item 19907Some other command-line options starting with @option{-f}, 19908@option{-p}, or @option{-O} must be defined in the same way as when 19909the precompiled header was generated. At present, it's not clear 19910which options are safe to change and which are not; the safest choice 19911is to use exactly the same options when generating and using the 19912precompiled header. The following are known to be safe: 19913 19914@gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol 19915-fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol 19916-fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol 19917-pedantic-errors} 19918 19919@end itemize 19920 19921For all of these except the last, the compiler will automatically 19922ignore the precompiled header if the conditions aren't met. If you 19923find an option combination that doesn't work and doesn't cause the 19924precompiled header to be ignored, please consider filing a bug report, 19925see @ref{Bugs}. 19926 19927If you do use differing options when generating and using the 19928precompiled header, the actual behavior will be a mixture of the 19929behavior for the options. For instance, if you use @option{-g} to 19930generate the precompiled header but not when using it, you may or may 19931not get debugging information for routines in the precompiled header. 19932