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 -mprefer-avx128 @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-mstack-bias -mno-stack-bias @gol 907-munaligned-doubles -mno-unaligned-doubles @gol 908-mv8plus -mno-v8plus -mvis -mno-vis @gol 909-mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol 910-mfmaf -mno-fmaf -mpopc -mno-popc @gol 911-mfix-at697f} 912 913@emph{SPU Options} 914@gccoptlist{-mwarn-reloc -merror-reloc @gol 915-msafe-dma -munsafe-dma @gol 916-mbranch-hints @gol 917-msmall-mem -mlarge-mem -mstdmain @gol 918-mfixed-range=@var{register-range} @gol 919-mea32 -mea64 @gol 920-maddress-space-conversion -mno-address-space-conversion @gol 921-mcache-size=@var{cache-size} @gol 922-matomic-updates -mno-atomic-updates} 923 924@emph{System V Options} 925@gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}} 926 927@emph{TILE-Gx Options} 928@gccoptlist{-mcpu=@var{cpu} -m32 -m64} 929 930@emph{TILEPro Options} 931@gccoptlist{-mcpu=@var{cpu} -m32} 932 933@emph{V850 Options} 934@gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol 935-mprolog-function -mno-prolog-function -mspace @gol 936-mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol 937-mapp-regs -mno-app-regs @gol 938-mdisable-callt -mno-disable-callt @gol 939-mv850e2v3 @gol 940-mv850e2 @gol 941-mv850e1 -mv850es @gol 942-mv850e @gol 943-mv850 -mbig-switch} 944 945@emph{VAX Options} 946@gccoptlist{-mg -mgnu -munix} 947 948@emph{VxWorks Options} 949@gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol 950-Xbind-lazy -Xbind-now} 951 952@emph{x86-64 Options} 953See i386 and x86-64 Options. 954 955@emph{Xstormy16 Options} 956@gccoptlist{-msim} 957 958@emph{Xtensa Options} 959@gccoptlist{-mconst16 -mno-const16 @gol 960-mfused-madd -mno-fused-madd @gol 961-mforce-no-pic @gol 962-mserialize-volatile -mno-serialize-volatile @gol 963-mtext-section-literals -mno-text-section-literals @gol 964-mtarget-align -mno-target-align @gol 965-mlongcalls -mno-longcalls} 966 967@emph{zSeries Options} 968See S/390 and zSeries Options. 969 970@item Code Generation Options 971@xref{Code Gen Options,,Options for Code Generation Conventions}. 972@gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol 973-ffixed-@var{reg} -fexceptions @gol 974-fnon-call-exceptions -funwind-tables @gol 975-fasynchronous-unwind-tables @gol 976-finhibit-size-directive -finstrument-functions @gol 977-finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol 978-finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol 979-fno-common -fno-ident @gol 980-fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol 981-fno-jump-tables @gol 982-frecord-gcc-switches @gol 983-freg-struct-return -fshort-enums @gol 984-fshort-double -fshort-wchar @gol 985-fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol 986-fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol 987-fno-stack-limit -fsplit-stack @gol 988-fleading-underscore -ftls-model=@var{model} @gol 989-ftrapv -fwrapv -fbounds-check @gol 990-fvisibility -fstrict-volatile-bitfields} 991@end table 992 993@menu 994* Overall Options:: Controlling the kind of output: 995 an executable, object files, assembler files, 996 or preprocessed source. 997* C Dialect Options:: Controlling the variant of C language compiled. 998* C++ Dialect Options:: Variations on C++. 999* Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C 1000 and Objective-C++. 1001* Language Independent Options:: Controlling how diagnostics should be 1002 formatted. 1003* Warning Options:: How picky should the compiler be? 1004* Debugging Options:: Symbol tables, measurements, and debugging dumps. 1005* Optimize Options:: How much optimization? 1006* Preprocessor Options:: Controlling header files and macro definitions. 1007 Also, getting dependency information for Make. 1008* Assembler Options:: Passing options to the assembler. 1009* Link Options:: Specifying libraries and so on. 1010* Directory Options:: Where to find header files and libraries. 1011 Where to find the compiler executable files. 1012* Spec Files:: How to pass switches to sub-processes. 1013* Target Options:: Running a cross-compiler, or an old version of GCC. 1014@end menu 1015 1016@node Overall Options 1017@section Options Controlling the Kind of Output 1018 1019Compilation can involve up to four stages: preprocessing, compilation 1020proper, assembly and linking, always in that order. GCC is capable of 1021preprocessing and compiling several files either into several 1022assembler input files, or into one assembler input file; then each 1023assembler input file produces an object file, and linking combines all 1024the object files (those newly compiled, and those specified as input) 1025into an executable file. 1026 1027@cindex file name suffix 1028For any given input file, the file name suffix determines what kind of 1029compilation is done: 1030 1031@table @gcctabopt 1032@item @var{file}.c 1033C source code that must be preprocessed. 1034 1035@item @var{file}.i 1036C source code that should not be preprocessed. 1037 1038@item @var{file}.ii 1039C++ source code that should not be preprocessed. 1040 1041@item @var{file}.m 1042Objective-C source code. Note that you must link with the @file{libobjc} 1043library to make an Objective-C program work. 1044 1045@item @var{file}.mi 1046Objective-C source code that should not be preprocessed. 1047 1048@item @var{file}.mm 1049@itemx @var{file}.M 1050Objective-C++ source code. Note that you must link with the @file{libobjc} 1051library to make an Objective-C++ program work. Note that @samp{.M} refers 1052to a literal capital M@. 1053 1054@item @var{file}.mii 1055Objective-C++ source code that should not be preprocessed. 1056 1057@item @var{file}.h 1058C, C++, Objective-C or Objective-C++ header file to be turned into a 1059precompiled header (default), or C, C++ header file to be turned into an 1060Ada spec (via the @option{-fdump-ada-spec} switch). 1061 1062@item @var{file}.cc 1063@itemx @var{file}.cp 1064@itemx @var{file}.cxx 1065@itemx @var{file}.cpp 1066@itemx @var{file}.CPP 1067@itemx @var{file}.c++ 1068@itemx @var{file}.C 1069C++ source code that must be preprocessed. Note that in @samp{.cxx}, 1070the last two letters must both be literally @samp{x}. Likewise, 1071@samp{.C} refers to a literal capital C@. 1072 1073@item @var{file}.mm 1074@itemx @var{file}.M 1075Objective-C++ source code that must be preprocessed. 1076 1077@item @var{file}.mii 1078Objective-C++ source code that should not be preprocessed. 1079 1080@item @var{file}.hh 1081@itemx @var{file}.H 1082@itemx @var{file}.hp 1083@itemx @var{file}.hxx 1084@itemx @var{file}.hpp 1085@itemx @var{file}.HPP 1086@itemx @var{file}.h++ 1087@itemx @var{file}.tcc 1088C++ header file to be turned into a precompiled header or Ada spec. 1089 1090@item @var{file}.f 1091@itemx @var{file}.for 1092@itemx @var{file}.ftn 1093Fixed form Fortran source code that should not be preprocessed. 1094 1095@item @var{file}.F 1096@itemx @var{file}.FOR 1097@itemx @var{file}.fpp 1098@itemx @var{file}.FPP 1099@itemx @var{file}.FTN 1100Fixed form Fortran source code that must be preprocessed (with the traditional 1101preprocessor). 1102 1103@item @var{file}.f90 1104@itemx @var{file}.f95 1105@itemx @var{file}.f03 1106@itemx @var{file}.f08 1107Free form Fortran source code that should not be preprocessed. 1108 1109@item @var{file}.F90 1110@itemx @var{file}.F95 1111@itemx @var{file}.F03 1112@itemx @var{file}.F08 1113Free form Fortran source code that must be preprocessed (with the 1114traditional preprocessor). 1115 1116@item @var{file}.go 1117Go source code. 1118 1119@c FIXME: Descriptions of Java file types. 1120@c @var{file}.java 1121@c @var{file}.class 1122@c @var{file}.zip 1123@c @var{file}.jar 1124 1125@item @var{file}.ads 1126Ada source code file that contains a library unit declaration (a 1127declaration of a package, subprogram, or generic, or a generic 1128instantiation), or a library unit renaming declaration (a package, 1129generic, or subprogram renaming declaration). Such files are also 1130called @dfn{specs}. 1131 1132@item @var{file}.adb 1133Ada source code file containing a library unit body (a subprogram or 1134package body). Such files are also called @dfn{bodies}. 1135 1136@c GCC also knows about some suffixes for languages not yet included: 1137@c Pascal: 1138@c @var{file}.p 1139@c @var{file}.pas 1140@c Ratfor: 1141@c @var{file}.r 1142 1143@item @var{file}.s 1144Assembler code. 1145 1146@item @var{file}.S 1147@itemx @var{file}.sx 1148Assembler code that must be preprocessed. 1149 1150@item @var{other} 1151An object file to be fed straight into linking. 1152Any file name with no recognized suffix is treated this way. 1153@end table 1154 1155@opindex x 1156You can specify the input language explicitly with the @option{-x} option: 1157 1158@table @gcctabopt 1159@item -x @var{language} 1160Specify explicitly the @var{language} for the following input files 1161(rather than letting the compiler choose a default based on the file 1162name suffix). This option applies to all following input files until 1163the next @option{-x} option. Possible values for @var{language} are: 1164@smallexample 1165c c-header cpp-output 1166c++ c++-header c++-cpp-output 1167objective-c objective-c-header objective-c-cpp-output 1168objective-c++ objective-c++-header objective-c++-cpp-output 1169assembler assembler-with-cpp 1170ada 1171f77 f77-cpp-input f95 f95-cpp-input 1172go 1173java 1174@end smallexample 1175 1176@item -x none 1177Turn off any specification of a language, so that subsequent files are 1178handled according to their file name suffixes (as they are if @option{-x} 1179has not been used at all). 1180 1181@item -pass-exit-codes 1182@opindex pass-exit-codes 1183Normally the @command{gcc} program will exit with the code of 1 if any 1184phase of the compiler returns a non-success return code. If you specify 1185@option{-pass-exit-codes}, the @command{gcc} program will instead return with 1186numerically highest error produced by any phase that returned an error 1187indication. The C, C++, and Fortran frontends return 4, if an internal 1188compiler error is encountered. 1189@end table 1190 1191If you only want some of the stages of compilation, you can use 1192@option{-x} (or filename suffixes) to tell @command{gcc} where to start, and 1193one of the options @option{-c}, @option{-S}, or @option{-E} to say where 1194@command{gcc} is to stop. Note that some combinations (for example, 1195@samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all. 1196 1197@table @gcctabopt 1198@item -c 1199@opindex c 1200Compile or assemble the source files, but do not link. The linking 1201stage simply is not done. The ultimate output is in the form of an 1202object file for each source file. 1203 1204By default, the object file name for a source file is made by replacing 1205the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}. 1206 1207Unrecognized input files, not requiring compilation or assembly, are 1208ignored. 1209 1210@item -S 1211@opindex S 1212Stop after the stage of compilation proper; do not assemble. The output 1213is in the form of an assembler code file for each non-assembler input 1214file specified. 1215 1216By default, the assembler file name for a source file is made by 1217replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}. 1218 1219Input files that don't require compilation are ignored. 1220 1221@item -E 1222@opindex E 1223Stop after the preprocessing stage; do not run the compiler proper. The 1224output is in the form of preprocessed source code, which is sent to the 1225standard output. 1226 1227Input files that don't require preprocessing are ignored. 1228 1229@cindex output file option 1230@item -o @var{file} 1231@opindex o 1232Place output in file @var{file}. This applies regardless to whatever 1233sort of output is being produced, whether it be an executable file, 1234an object file, an assembler file or preprocessed C code. 1235 1236If @option{-o} is not specified, the default is to put an executable 1237file in @file{a.out}, the object file for 1238@file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its 1239assembler file in @file{@var{source}.s}, a precompiled header file in 1240@file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on 1241standard output. 1242 1243@item -v 1244@opindex v 1245Print (on standard error output) the commands executed to run the stages 1246of compilation. Also print the version number of the compiler driver 1247program and of the preprocessor and the compiler proper. 1248 1249@item -### 1250@opindex ### 1251Like @option{-v} except the commands are not executed and arguments 1252are quoted unless they contain only alphanumeric characters or @code{./-_}. 1253This is useful for shell scripts to capture the driver-generated command lines. 1254 1255@item -pipe 1256@opindex pipe 1257Use pipes rather than temporary files for communication between the 1258various stages of compilation. This fails to work on some systems where 1259the assembler is unable to read from a pipe; but the GNU assembler has 1260no trouble. 1261 1262@item --help 1263@opindex help 1264Print (on the standard output) a description of the command-line options 1265understood by @command{gcc}. If the @option{-v} option is also specified 1266then @option{--help} will also be passed on to the various processes 1267invoked by @command{gcc}, so that they can display the command-line options 1268they accept. If the @option{-Wextra} option has also been specified 1269(prior to the @option{--help} option), then command-line options that 1270have no documentation associated with them will also be displayed. 1271 1272@item --target-help 1273@opindex target-help 1274Print (on the standard output) a description of target-specific command-line 1275options for each tool. For some targets extra target-specific 1276information may also be printed. 1277 1278@item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]} 1279Print (on the standard output) a description of the command-line 1280options understood by the compiler that fit into all specified classes 1281and qualifiers. These are the supported classes: 1282 1283@table @asis 1284@item @samp{optimizers} 1285This will display all of the optimization options supported by the 1286compiler. 1287 1288@item @samp{warnings} 1289This will display all of the options controlling warning messages 1290produced by the compiler. 1291 1292@item @samp{target} 1293This will display target-specific options. Unlike the 1294@option{--target-help} option however, target-specific options of the 1295linker and assembler will not be displayed. This is because those 1296tools do not currently support the extended @option{--help=} syntax. 1297 1298@item @samp{params} 1299This will display the values recognized by the @option{--param} 1300option. 1301 1302@item @var{language} 1303This will display the options supported for @var{language}, where 1304@var{language} is the name of one of the languages supported in this 1305version of GCC. 1306 1307@item @samp{common} 1308This will display the options that are common to all languages. 1309@end table 1310 1311These are the supported qualifiers: 1312 1313@table @asis 1314@item @samp{undocumented} 1315Display only those options that are undocumented. 1316 1317@item @samp{joined} 1318Display options taking an argument that appears after an equal 1319sign in the same continuous piece of text, such as: 1320@samp{--help=target}. 1321 1322@item @samp{separate} 1323Display options taking an argument that appears as a separate word 1324following the original option, such as: @samp{-o output-file}. 1325@end table 1326 1327Thus for example to display all the undocumented target-specific 1328switches supported by the compiler the following can be used: 1329 1330@smallexample 1331--help=target,undocumented 1332@end smallexample 1333 1334The sense of a qualifier can be inverted by prefixing it with the 1335@samp{^} character, so for example to display all binary warning 1336options (i.e., ones that are either on or off and that do not take an 1337argument) that have a description, use: 1338 1339@smallexample 1340--help=warnings,^joined,^undocumented 1341@end smallexample 1342 1343The argument to @option{--help=} should not consist solely of inverted 1344qualifiers. 1345 1346Combining several classes is possible, although this usually 1347restricts the output by so much that there is nothing to display. One 1348case where it does work however is when one of the classes is 1349@var{target}. So for example to display all the target-specific 1350optimization options the following can be used: 1351 1352@smallexample 1353--help=target,optimizers 1354@end smallexample 1355 1356The @option{--help=} option can be repeated on the command line. Each 1357successive use will display its requested class of options, skipping 1358those that have already been displayed. 1359 1360If the @option{-Q} option appears on the command line before the 1361@option{--help=} option, then the descriptive text displayed by 1362@option{--help=} is changed. Instead of describing the displayed 1363options, an indication is given as to whether the option is enabled, 1364disabled or set to a specific value (assuming that the compiler 1365knows this at the point where the @option{--help=} option is used). 1366 1367Here is a truncated example from the ARM port of @command{gcc}: 1368 1369@smallexample 1370 % gcc -Q -mabi=2 --help=target -c 1371 The following options are target specific: 1372 -mabi= 2 1373 -mabort-on-noreturn [disabled] 1374 -mapcs [disabled] 1375@end smallexample 1376 1377The output is sensitive to the effects of previous command-line 1378options, so for example it is possible to find out which optimizations 1379are enabled at @option{-O2} by using: 1380 1381@smallexample 1382-Q -O2 --help=optimizers 1383@end smallexample 1384 1385Alternatively you can discover which binary optimizations are enabled 1386by @option{-O3} by using: 1387 1388@smallexample 1389gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 1390gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 1391diff /tmp/O2-opts /tmp/O3-opts | grep enabled 1392@end smallexample 1393 1394@item -no-canonical-prefixes 1395@opindex no-canonical-prefixes 1396Do not expand any symbolic links, resolve references to @samp{/../} 1397or @samp{/./}, or make the path absolute when generating a relative 1398prefix. 1399 1400@item --version 1401@opindex version 1402Display the version number and copyrights of the invoked GCC@. 1403 1404@item -wrapper 1405@opindex wrapper 1406Invoke all subcommands under a wrapper program. The name of the 1407wrapper program and its parameters are passed as a comma separated 1408list. 1409 1410@smallexample 1411gcc -c t.c -wrapper gdb,--args 1412@end smallexample 1413 1414This will invoke all subprograms of @command{gcc} under 1415@samp{gdb --args}, thus the invocation of @command{cc1} will be 1416@samp{gdb --args cc1 @dots{}}. 1417 1418@item -fplugin=@var{name}.so 1419Load the plugin code in file @var{name}.so, assumed to be a 1420shared object to be dlopen'd by the compiler. The base name of 1421the shared object file is used to identify the plugin for the 1422purposes of argument parsing (See 1423@option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below). 1424Each plugin should define the callback functions specified in the 1425Plugins API. 1426 1427@item -fplugin-arg-@var{name}-@var{key}=@var{value} 1428Define an argument called @var{key} with a value of @var{value} 1429for the plugin called @var{name}. 1430 1431@item -fdump-ada-spec@r{[}-slim@r{]} 1432For C and C++ source and include files, generate corresponding Ada 1433specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn, 1434GNAT User's Guide}, which provides detailed documentation on this feature. 1435 1436@item -fdump-go-spec=@var{file} 1437For input files in any language, generate corresponding Go 1438declarations in @var{file}. This generates Go @code{const}, 1439@code{type}, @code{var}, and @code{func} declarations which may be a 1440useful way to start writing a Go interface to code written in some 1441other language. 1442 1443@include @value{srcdir}/../libiberty/at-file.texi 1444@end table 1445 1446@node Invoking G++ 1447@section Compiling C++ Programs 1448 1449@cindex suffixes for C++ source 1450@cindex C++ source file suffixes 1451C++ source files conventionally use one of the suffixes @samp{.C}, 1452@samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or 1453@samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp}, 1454@samp{.H}, or (for shared template code) @samp{.tcc}; and 1455preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes 1456files with these names and compiles them as C++ programs even if you 1457call the compiler the same way as for compiling C programs (usually 1458with the name @command{gcc}). 1459 1460@findex g++ 1461@findex c++ 1462However, the use of @command{gcc} does not add the C++ library. 1463@command{g++} is a program that calls GCC and treats @samp{.c}, 1464@samp{.h} and @samp{.i} files as C++ source files instead of C source 1465files unless @option{-x} is used, and automatically specifies linking 1466against the C++ library. This program is also useful when 1467precompiling a C header file with a @samp{.h} extension for use in C++ 1468compilations. On many systems, @command{g++} is also installed with 1469the name @command{c++}. 1470 1471@cindex invoking @command{g++} 1472When you compile C++ programs, you may specify many of the same 1473command-line options that you use for compiling programs in any 1474language; or command-line options meaningful for C and related 1475languages; or options that are meaningful only for C++ programs. 1476@xref{C Dialect Options,,Options Controlling C Dialect}, for 1477explanations of options for languages related to C@. 1478@xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for 1479explanations of options that are meaningful only for C++ programs. 1480 1481@node C Dialect Options 1482@section Options Controlling C Dialect 1483@cindex dialect options 1484@cindex language dialect options 1485@cindex options, dialect 1486 1487The following options control the dialect of C (or languages derived 1488from C, such as C++, Objective-C and Objective-C++) that the compiler 1489accepts: 1490 1491@table @gcctabopt 1492@cindex ANSI support 1493@cindex ISO support 1494@item -ansi 1495@opindex ansi 1496In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is 1497equivalent to @samp{-std=c++98}. 1498 1499This turns off certain features of GCC that are incompatible with ISO 1500C90 (when compiling C code), or of standard C++ (when compiling C++ code), 1501such as the @code{asm} and @code{typeof} keywords, and 1502predefined macros such as @code{unix} and @code{vax} that identify the 1503type of system you are using. It also enables the undesirable and 1504rarely used ISO trigraph feature. For the C compiler, 1505it disables recognition of C++ style @samp{//} comments as well as 1506the @code{inline} keyword. 1507 1508The alternate keywords @code{__asm__}, @code{__extension__}, 1509@code{__inline__} and @code{__typeof__} continue to work despite 1510@option{-ansi}. You would not want to use them in an ISO C program, of 1511course, but it is useful to put them in header files that might be included 1512in compilations done with @option{-ansi}. Alternate predefined macros 1513such as @code{__unix__} and @code{__vax__} are also available, with or 1514without @option{-ansi}. 1515 1516The @option{-ansi} option does not cause non-ISO programs to be 1517rejected gratuitously. For that, @option{-pedantic} is required in 1518addition to @option{-ansi}. @xref{Warning Options}. 1519 1520The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi} 1521option is used. Some header files may notice this macro and refrain 1522from declaring certain functions or defining certain macros that the 1523ISO standard doesn't call for; this is to avoid interfering with any 1524programs that might use these names for other things. 1525 1526Functions that would normally be built in but do not have semantics 1527defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in 1528functions when @option{-ansi} is used. @xref{Other Builtins,,Other 1529built-in functions provided by GCC}, for details of the functions 1530affected. 1531 1532@item -std= 1533@opindex std 1534Determine the language standard. @xref{Standards,,Language Standards 1535Supported by GCC}, for details of these standard versions. This option 1536is currently only supported when compiling C or C++. 1537 1538The compiler can accept several base standards, such as @samp{c90} or 1539@samp{c++98}, and GNU dialects of those standards, such as 1540@samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the 1541compiler will accept all programs following that standard and those 1542using GNU extensions that do not contradict it. For example, 1543@samp{-std=c90} turns off certain features of GCC that are 1544incompatible with ISO C90, such as the @code{asm} and @code{typeof} 1545keywords, but not other GNU extensions that do not have a meaning in 1546ISO C90, such as omitting the middle term of a @code{?:} 1547expression. On the other hand, by specifying a GNU dialect of a 1548standard, all features the compiler support are enabled, even when 1549those features change the meaning of the base standard and some 1550strict-conforming programs may be rejected. The particular standard 1551is used by @option{-pedantic} to identify which features are GNU 1552extensions given that version of the standard. For example 1553@samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//} 1554comments, while @samp{-std=gnu99 -pedantic} would not. 1555 1556A value for this option must be provided; possible values are 1557 1558@table @samp 1559@item c90 1560@itemx c89 1561@itemx iso9899:1990 1562Support all ISO C90 programs (certain GNU extensions that conflict 1563with ISO C90 are disabled). Same as @option{-ansi} for C code. 1564 1565@item iso9899:199409 1566ISO C90 as modified in amendment 1. 1567 1568@item c99 1569@itemx c9x 1570@itemx iso9899:1999 1571@itemx iso9899:199x 1572ISO C99. Note that this standard is not yet fully supported; see 1573@w{@uref{http://gcc.gnu.org/gcc-4.7/c99status.html}} for more information. The 1574names @samp{c9x} and @samp{iso9899:199x} are deprecated. 1575 1576@item c11 1577@itemx c1x 1578@itemx iso9899:2011 1579ISO C11, the 2011 revision of the ISO C standard. 1580Support is incomplete and experimental. The name @samp{c1x} is 1581deprecated. 1582 1583@item gnu90 1584@itemx gnu89 1585GNU dialect of ISO C90 (including some C99 features). This 1586is the default for C code. 1587 1588@item gnu99 1589@itemx gnu9x 1590GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC, 1591this will become the default. The name @samp{gnu9x} is deprecated. 1592 1593@item gnu11 1594@item gnu1x 1595GNU dialect of ISO C11. Support is incomplete and experimental. The 1596name @samp{gnu1x} is deprecated. 1597 1598@item c++98 1599The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for 1600C++ code. 1601 1602@item gnu++98 1603GNU dialect of @option{-std=c++98}. This is the default for 1604C++ code. 1605 1606@item c++11 1607The 2011 ISO C++ standard plus amendments. Support for C++11 is still 1608experimental, and may change in incompatible ways in future releases. 1609 1610@item gnu++11 1611GNU dialect of @option{-std=c++11}. Support for C++11 is still 1612experimental, and may change in incompatible ways in future releases. 1613@end table 1614 1615@item -fgnu89-inline 1616@opindex fgnu89-inline 1617The option @option{-fgnu89-inline} tells GCC to use the traditional 1618GNU semantics for @code{inline} functions when in C99 mode. 1619@xref{Inline,,An Inline Function is As Fast As a Macro}. This option 1620is accepted and ignored by GCC versions 4.1.3 up to but not including 16214.3. In GCC versions 4.3 and later it changes the behavior of GCC in 1622C99 mode. Using this option is roughly equivalent to adding the 1623@code{gnu_inline} function attribute to all inline functions 1624(@pxref{Function Attributes}). 1625 1626The option @option{-fno-gnu89-inline} explicitly tells GCC to use the 1627C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it 1628specifies the default behavior). This option was first supported in 1629GCC 4.3. This option is not supported in @option{-std=c90} or 1630@option{-std=gnu90} mode. 1631 1632The preprocessor macros @code{__GNUC_GNU_INLINE__} and 1633@code{__GNUC_STDC_INLINE__} may be used to check which semantics are 1634in effect for @code{inline} functions. @xref{Common Predefined 1635Macros,,,cpp,The C Preprocessor}. 1636 1637@item -aux-info @var{filename} 1638@opindex aux-info 1639Output to the given filename prototyped declarations for all functions 1640declared and/or defined in a translation unit, including those in header 1641files. This option is silently ignored in any language other than C@. 1642 1643Besides declarations, the file indicates, in comments, the origin of 1644each declaration (source file and line), whether the declaration was 1645implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or 1646@samp{O} for old, respectively, in the first character after the line 1647number and the colon), and whether it came from a declaration or a 1648definition (@samp{C} or @samp{F}, respectively, in the following 1649character). In the case of function definitions, a K&R-style list of 1650arguments followed by their declarations is also provided, inside 1651comments, after the declaration. 1652 1653@item -fallow-parameterless-variadic-functions 1654Accept variadic functions without named parameters. 1655 1656Although it is possible to define such a function, this is not very 1657useful as it is not possible to read the arguments. This is only 1658supported for C as this construct is allowed by C++. 1659 1660@item -fno-asm 1661@opindex fno-asm 1662Do not recognize @code{asm}, @code{inline} or @code{typeof} as a 1663keyword, so that code can use these words as identifiers. You can use 1664the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__} 1665instead. @option{-ansi} implies @option{-fno-asm}. 1666 1667In C++, this switch only affects the @code{typeof} keyword, since 1668@code{asm} and @code{inline} are standard keywords. You may want to 1669use the @option{-fno-gnu-keywords} flag instead, which has the same 1670effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this 1671switch only affects the @code{asm} and @code{typeof} keywords, since 1672@code{inline} is a standard keyword in ISO C99. 1673 1674@item -fno-builtin 1675@itemx -fno-builtin-@var{function} 1676@opindex fno-builtin 1677@cindex built-in functions 1678Don't recognize built-in functions that do not begin with 1679@samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in 1680functions provided by GCC}, for details of the functions affected, 1681including those which are not built-in functions when @option{-ansi} or 1682@option{-std} options for strict ISO C conformance are used because they 1683do not have an ISO standard meaning. 1684 1685GCC normally generates special code to handle certain built-in functions 1686more efficiently; for instance, calls to @code{alloca} may become single 1687instructions which adjust the stack directly, and calls to @code{memcpy} 1688may become inline copy loops. The resulting code is often both smaller 1689and faster, but since the function calls no longer appear as such, you 1690cannot set a breakpoint on those calls, nor can you change the behavior 1691of the functions by linking with a different library. In addition, 1692when a function is recognized as a built-in function, GCC may use 1693information about that function to warn about problems with calls to 1694that function, or to generate more efficient code, even if the 1695resulting code still contains calls to that function. For example, 1696warnings are given with @option{-Wformat} for bad calls to 1697@code{printf}, when @code{printf} is built in, and @code{strlen} is 1698known not to modify global memory. 1699 1700With the @option{-fno-builtin-@var{function}} option 1701only the built-in function @var{function} is 1702disabled. @var{function} must not begin with @samp{__builtin_}. If a 1703function is named that is not built-in in this version of GCC, this 1704option is ignored. There is no corresponding 1705@option{-fbuiltin-@var{function}} option; if you wish to enable 1706built-in functions selectively when using @option{-fno-builtin} or 1707@option{-ffreestanding}, you may define macros such as: 1708 1709@smallexample 1710#define abs(n) __builtin_abs ((n)) 1711#define strcpy(d, s) __builtin_strcpy ((d), (s)) 1712@end smallexample 1713 1714@item -fhosted 1715@opindex fhosted 1716@cindex hosted environment 1717 1718Assert that compilation takes place in a hosted environment. This implies 1719@option{-fbuiltin}. A hosted environment is one in which the 1720entire standard library is available, and in which @code{main} has a return 1721type of @code{int}. Examples are nearly everything except a kernel. 1722This is equivalent to @option{-fno-freestanding}. 1723 1724@item -ffreestanding 1725@opindex ffreestanding 1726@cindex hosted environment 1727 1728Assert that compilation takes place in a freestanding environment. This 1729implies @option{-fno-builtin}. A freestanding environment 1730is one in which the standard library may not exist, and program startup may 1731not necessarily be at @code{main}. The most obvious example is an OS kernel. 1732This is equivalent to @option{-fno-hosted}. 1733 1734@xref{Standards,,Language Standards Supported by GCC}, for details of 1735freestanding and hosted environments. 1736 1737@item -fopenmp 1738@opindex fopenmp 1739@cindex OpenMP parallel 1740Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and 1741@code{!$omp} in Fortran. When @option{-fopenmp} is specified, the 1742compiler generates parallel code according to the OpenMP Application 1743Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option 1744implies @option{-pthread}, and thus is only supported on targets that 1745have support for @option{-pthread}. 1746 1747@item -fgnu-tm 1748@opindex fgnu-tm 1749When the option @option{-fgnu-tm} is specified, the compiler will 1750generate code for the Linux variant of Intel's current Transactional 1751Memory ABI specification document (Revision 1.1, May 6 2009). This is 1752an experimental feature whose interface may change in future versions 1753of GCC, as the official specification changes. Please note that not 1754all architectures are supported for this feature. 1755 1756For more information on GCC's support for transactional memory, 1757@xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU 1758Transactional Memory Library}. 1759 1760Note that the transactional memory feature is not supported with 1761non-call exceptions (@option{-fnon-call-exceptions}). 1762 1763@item -fms-extensions 1764@opindex fms-extensions 1765Accept some non-standard constructs used in Microsoft header files. 1766 1767In C++ code, this allows member names in structures to be similar 1768to previous types declarations. 1769 1770@smallexample 1771typedef int UOW; 1772struct ABC @{ 1773 UOW UOW; 1774@}; 1775@end smallexample 1776 1777Some cases of unnamed fields in structures and unions are only 1778accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union 1779fields within structs/unions}, for details. 1780 1781@item -fplan9-extensions 1782Accept some non-standard constructs used in Plan 9 code. 1783 1784This enables @option{-fms-extensions}, permits passing pointers to 1785structures with anonymous fields to functions that expect pointers to 1786elements of the type of the field, and permits referring to anonymous 1787fields declared using a typedef. @xref{Unnamed Fields,,Unnamed 1788struct/union fields within structs/unions}, for details. This is only 1789supported for C, not C++. 1790 1791@item -trigraphs 1792@opindex trigraphs 1793Support ISO C trigraphs. The @option{-ansi} option (and @option{-std} 1794options for strict ISO C conformance) implies @option{-trigraphs}. 1795 1796@item -no-integrated-cpp 1797@opindex no-integrated-cpp 1798Performs a compilation in two passes: preprocessing and compiling. This 1799option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the 1800@option{-B} option. The user supplied compilation step can then add in 1801an additional preprocessing step after normal preprocessing but before 1802compiling. The default is to use the integrated cpp (internal cpp) 1803 1804The semantics of this option will change if "cc1", "cc1plus", and 1805"cc1obj" are merged. 1806 1807@cindex traditional C language 1808@cindex C language, traditional 1809@item -traditional 1810@itemx -traditional-cpp 1811@opindex traditional-cpp 1812@opindex traditional 1813Formerly, these options caused GCC to attempt to emulate a pre-standard 1814C compiler. They are now only supported with the @option{-E} switch. 1815The preprocessor continues to support a pre-standard mode. See the GNU 1816CPP manual for details. 1817 1818@item -fcond-mismatch 1819@opindex fcond-mismatch 1820Allow conditional expressions with mismatched types in the second and 1821third arguments. The value of such an expression is void. This option 1822is not supported for C++. 1823 1824@item -flax-vector-conversions 1825@opindex flax-vector-conversions 1826Allow implicit conversions between vectors with differing numbers of 1827elements and/or incompatible element types. This option should not be 1828used for new code. 1829 1830@item -funsigned-char 1831@opindex funsigned-char 1832Let the type @code{char} be unsigned, like @code{unsigned char}. 1833 1834Each kind of machine has a default for what @code{char} should 1835be. It is either like @code{unsigned char} by default or like 1836@code{signed char} by default. 1837 1838Ideally, a portable program should always use @code{signed char} or 1839@code{unsigned char} when it depends on the signedness of an object. 1840But many programs have been written to use plain @code{char} and 1841expect it to be signed, or expect it to be unsigned, depending on the 1842machines they were written for. This option, and its inverse, let you 1843make such a program work with the opposite default. 1844 1845The type @code{char} is always a distinct type from each of 1846@code{signed char} or @code{unsigned char}, even though its behavior 1847is always just like one of those two. 1848 1849@item -fsigned-char 1850@opindex fsigned-char 1851Let the type @code{char} be signed, like @code{signed char}. 1852 1853Note that this is equivalent to @option{-fno-unsigned-char}, which is 1854the negative form of @option{-funsigned-char}. Likewise, the option 1855@option{-fno-signed-char} is equivalent to @option{-funsigned-char}. 1856 1857@item -fsigned-bitfields 1858@itemx -funsigned-bitfields 1859@itemx -fno-signed-bitfields 1860@itemx -fno-unsigned-bitfields 1861@opindex fsigned-bitfields 1862@opindex funsigned-bitfields 1863@opindex fno-signed-bitfields 1864@opindex fno-unsigned-bitfields 1865These options control whether a bit-field is signed or unsigned, when the 1866declaration does not use either @code{signed} or @code{unsigned}. By 1867default, such a bit-field is signed, because this is consistent: the 1868basic integer types such as @code{int} are signed types. 1869@end table 1870 1871@node C++ Dialect Options 1872@section Options Controlling C++ Dialect 1873 1874@cindex compiler options, C++ 1875@cindex C++ options, command-line 1876@cindex options, C++ 1877This section describes the command-line options that are only meaningful 1878for C++ programs; but you can also use most of the GNU compiler options 1879regardless of what language your program is in. For example, you 1880might compile a file @code{firstClass.C} like this: 1881 1882@smallexample 1883g++ -g -frepo -O -c firstClass.C 1884@end smallexample 1885 1886@noindent 1887In this example, only @option{-frepo} is an option meant 1888only for C++ programs; you can use the other options with any 1889language supported by GCC@. 1890 1891Here is a list of options that are @emph{only} for compiling C++ programs: 1892 1893@table @gcctabopt 1894 1895@item -fabi-version=@var{n} 1896@opindex fabi-version 1897Use version @var{n} of the C++ ABI@. Version 2 is the version of the 1898C++ ABI that first appeared in G++ 3.4. Version 1 is the version of 1899the C++ ABI that first appeared in G++ 3.2. Version 0 will always be 1900the version that conforms most closely to the C++ ABI specification. 1901Therefore, the ABI obtained using version 0 will change as ABI bugs 1902are fixed. 1903 1904The default is version 2. 1905 1906Version 3 corrects an error in mangling a constant address as a 1907template argument. 1908 1909Version 4, which first appeared in G++ 4.5, implements a standard 1910mangling for vector types. 1911 1912Version 5, which first appeared in G++ 4.6, corrects the mangling of 1913attribute const/volatile on function pointer types, decltype of a 1914plain decl, and use of a function parameter in the declaration of 1915another parameter. 1916 1917Version 6, which first appeared in G++ 4.7, corrects the promotion 1918behavior of C++11 scoped enums and the mangling of template argument 1919packs, const/static_cast, prefix ++ and --, and a class scope function 1920used as a template argument. 1921 1922See also @option{-Wabi}. 1923 1924@item -fno-access-control 1925@opindex fno-access-control 1926Turn off all access checking. This switch is mainly useful for working 1927around bugs in the access control code. 1928 1929@item -fcheck-new 1930@opindex fcheck-new 1931Check that the pointer returned by @code{operator new} is non-null 1932before attempting to modify the storage allocated. This check is 1933normally unnecessary because the C++ standard specifies that 1934@code{operator new} will only return @code{0} if it is declared 1935@samp{throw()}, in which case the compiler will always check the 1936return value even without this option. In all other cases, when 1937@code{operator new} has a non-empty exception specification, memory 1938exhaustion is signalled by throwing @code{std::bad_alloc}. See also 1939@samp{new (nothrow)}. 1940 1941@item -fconserve-space 1942@opindex fconserve-space 1943Put uninitialized or run-time-initialized global variables into the 1944common segment, as C does. This saves space in the executable at the 1945cost of not diagnosing duplicate definitions. If you compile with this 1946flag and your program mysteriously crashes after @code{main()} has 1947completed, you may have an object that is being destroyed twice because 1948two definitions were merged. 1949 1950This option is no longer useful on most targets, now that support has 1951been added for putting variables into BSS without making them common. 1952 1953@item -fconstexpr-depth=@var{n} 1954@opindex fconstexpr-depth 1955Set the maximum nested evaluation depth for C++11 constexpr functions 1956to @var{n}. A limit is needed to detect endless recursion during 1957constant expression evaluation. The minimum specified by the standard 1958is 512. 1959 1960@item -fdeduce-init-list 1961@opindex fdeduce-init-list 1962Enable deduction of a template type parameter as 1963std::initializer_list from a brace-enclosed initializer list, i.e. 1964 1965@smallexample 1966template <class T> auto forward(T t) -> decltype (realfn (t)) 1967@{ 1968 return realfn (t); 1969@} 1970 1971void f() 1972@{ 1973 forward(@{1,2@}); // call forward<std::initializer_list<int>> 1974@} 1975@end smallexample 1976 1977This deduction was implemented as a possible extension to the 1978originally proposed semantics for the C++11 standard, but was not part 1979of the final standard, so it is disabled by default. This option is 1980deprecated, and may be removed in a future version of G++. 1981 1982@item -ffriend-injection 1983@opindex ffriend-injection 1984Inject friend functions into the enclosing namespace, so that they are 1985visible outside the scope of the class in which they are declared. 1986Friend functions were documented to work this way in the old Annotated 1987C++ Reference Manual, and versions of G++ before 4.1 always worked 1988that way. However, in ISO C++ a friend function that is not declared 1989in an enclosing scope can only be found using argument dependent 1990lookup. This option causes friends to be injected as they were in 1991earlier releases. 1992 1993This option is for compatibility, and may be removed in a future 1994release of G++. 1995 1996@item -fno-elide-constructors 1997@opindex fno-elide-constructors 1998The C++ standard allows an implementation to omit creating a temporary 1999that is only used to initialize another object of the same type. 2000Specifying this option disables that optimization, and forces G++ to 2001call the copy constructor in all cases. 2002 2003@item -fno-enforce-eh-specs 2004@opindex fno-enforce-eh-specs 2005Don't generate code to check for violation of exception specifications 2006at run time. This option violates the C++ standard, but may be useful 2007for reducing code size in production builds, much like defining 2008@samp{NDEBUG}. This does not give user code permission to throw 2009exceptions in violation of the exception specifications; the compiler 2010will still optimize based on the specifications, so throwing an 2011unexpected exception will result in undefined behavior. 2012 2013@item -ffor-scope 2014@itemx -fno-for-scope 2015@opindex ffor-scope 2016@opindex fno-for-scope 2017If @option{-ffor-scope} is specified, the scope of variables declared in 2018a @i{for-init-statement} is limited to the @samp{for} loop itself, 2019as specified by the C++ standard. 2020If @option{-fno-for-scope} is specified, the scope of variables declared in 2021a @i{for-init-statement} extends to the end of the enclosing scope, 2022as was the case in old versions of G++, and other (traditional) 2023implementations of C++. 2024 2025The default if neither flag is given to follow the standard, 2026but to allow and give a warning for old-style code that would 2027otherwise be invalid, or have different behavior. 2028 2029@item -fno-gnu-keywords 2030@opindex fno-gnu-keywords 2031Do not recognize @code{typeof} as a keyword, so that code can use this 2032word as an identifier. You can use the keyword @code{__typeof__} instead. 2033@option{-ansi} implies @option{-fno-gnu-keywords}. 2034 2035@item -fno-implicit-templates 2036@opindex fno-implicit-templates 2037Never emit code for non-inline templates that are instantiated 2038implicitly (i.e.@: by use); only emit code for explicit instantiations. 2039@xref{Template Instantiation}, for more information. 2040 2041@item -fno-implicit-inline-templates 2042@opindex fno-implicit-inline-templates 2043Don't emit code for implicit instantiations of inline templates, either. 2044The default is to handle inlines differently so that compiles with and 2045without optimization will need the same set of explicit instantiations. 2046 2047@item -fno-implement-inlines 2048@opindex fno-implement-inlines 2049To save space, do not emit out-of-line copies of inline functions 2050controlled by @samp{#pragma implementation}. This will cause linker 2051errors if these functions are not inlined everywhere they are called. 2052 2053@item -fms-extensions 2054@opindex fms-extensions 2055Disable pedantic warnings about constructs used in MFC, such as implicit 2056int and getting a pointer to member function via non-standard syntax. 2057 2058@item -fno-nonansi-builtins 2059@opindex fno-nonansi-builtins 2060Disable built-in declarations of functions that are not mandated by 2061ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit}, 2062@code{index}, @code{bzero}, @code{conjf}, and other related functions. 2063 2064@item -fnothrow-opt 2065@opindex fnothrow-opt 2066Treat a @code{throw()} exception specification as though it were a 2067@code{noexcept} specification to reduce or eliminate the text size 2068overhead relative to a function with no exception specification. If 2069the function has local variables of types with non-trivial 2070destructors, the exception specification will actually make the 2071function smaller because the EH cleanups for those variables can be 2072optimized away. The semantic effect is that an exception thrown out of 2073a function with such an exception specification will result in a call 2074to @code{terminate} rather than @code{unexpected}. 2075 2076@item -fno-operator-names 2077@opindex fno-operator-names 2078Do not treat the operator name keywords @code{and}, @code{bitand}, 2079@code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as 2080synonyms as keywords. 2081 2082@item -fno-optional-diags 2083@opindex fno-optional-diags 2084Disable diagnostics that the standard says a compiler does not need to 2085issue. Currently, the only such diagnostic issued by G++ is the one for 2086a name having multiple meanings within a class. 2087 2088@item -fpermissive 2089@opindex fpermissive 2090Downgrade some diagnostics about nonconformant code from errors to 2091warnings. Thus, using @option{-fpermissive} will allow some 2092nonconforming code to compile. 2093 2094@item -fno-pretty-templates 2095@opindex fno-pretty-templates 2096When an error message refers to a specialization of a function 2097template, the compiler will normally print the signature of the 2098template followed by the template arguments and any typedefs or 2099typenames in the signature (e.g. @code{void f(T) [with T = int]} 2100rather than @code{void f(int)}) so that it's clear which template is 2101involved. When an error message refers to a specialization of a class 2102template, the compiler will omit any template arguments that match 2103the default template arguments for that template. If either of these 2104behaviors make it harder to understand the error message rather than 2105easier, using @option{-fno-pretty-templates} will disable them. 2106 2107@item -frepo 2108@opindex frepo 2109Enable automatic template instantiation at link time. This option also 2110implies @option{-fno-implicit-templates}. @xref{Template 2111Instantiation}, for more information. 2112 2113@item -fno-rtti 2114@opindex fno-rtti 2115Disable generation of information about every class with virtual 2116functions for use by the C++ run-time type identification features 2117(@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts 2118of the language, you can save some space by using this flag. Note that 2119exception handling uses the same information, but it will generate it as 2120needed. The @samp{dynamic_cast} operator can still be used for casts that 2121do not require run-time type information, i.e.@: casts to @code{void *} or to 2122unambiguous base classes. 2123 2124@item -fstats 2125@opindex fstats 2126Emit statistics about front-end processing at the end of the compilation. 2127This information is generally only useful to the G++ development team. 2128 2129@item -fstrict-enums 2130@opindex fstrict-enums 2131Allow the compiler to optimize using the assumption that a value of 2132enumerated type can only be one of the values of the enumeration (as 2133defined in the C++ standard; basically, a value that can be 2134represented in the minimum number of bits needed to represent all the 2135enumerators). This assumption may not be valid if the program uses a 2136cast to convert an arbitrary integer value to the enumerated type. 2137 2138@item -ftemplate-depth=@var{n} 2139@opindex ftemplate-depth 2140Set the maximum instantiation depth for template classes to @var{n}. 2141A limit on the template instantiation depth is needed to detect 2142endless recursions during template class instantiation. ANSI/ISO C++ 2143conforming programs must not rely on a maximum depth greater than 17 2144(changed to 1024 in C++11). The default value is 900, as the compiler 2145can run out of stack space before hitting 1024 in some situations. 2146 2147@item -fno-threadsafe-statics 2148@opindex fno-threadsafe-statics 2149Do not emit the extra code to use the routines specified in the C++ 2150ABI for thread-safe initialization of local statics. You can use this 2151option to reduce code size slightly in code that doesn't need to be 2152thread-safe. 2153 2154@item -fuse-cxa-atexit 2155@opindex fuse-cxa-atexit 2156Register destructors for objects with static storage duration with the 2157@code{__cxa_atexit} function rather than the @code{atexit} function. 2158This option is required for fully standards-compliant handling of static 2159destructors, but will only work if your C library supports 2160@code{__cxa_atexit}. 2161 2162@item -fno-use-cxa-get-exception-ptr 2163@opindex fno-use-cxa-get-exception-ptr 2164Don't use the @code{__cxa_get_exception_ptr} runtime routine. This 2165will cause @code{std::uncaught_exception} to be incorrect, but is necessary 2166if the runtime routine is not available. 2167 2168@item -fvisibility-inlines-hidden 2169@opindex fvisibility-inlines-hidden 2170This switch declares that the user does not attempt to compare 2171pointers to inline functions or methods where the addresses of the two functions 2172were taken in different shared objects. 2173 2174The effect of this is that GCC may, effectively, mark inline methods with 2175@code{__attribute__ ((visibility ("hidden")))} so that they do not 2176appear in the export table of a DSO and do not require a PLT indirection 2177when used within the DSO@. Enabling this option can have a dramatic effect 2178on load and link times of a DSO as it massively reduces the size of the 2179dynamic export table when the library makes heavy use of templates. 2180 2181The behavior of this switch is not quite the same as marking the 2182methods as hidden directly, because it does not affect static variables 2183local to the function or cause the compiler to deduce that 2184the function is defined in only one shared object. 2185 2186You may mark a method as having a visibility explicitly to negate the 2187effect of the switch for that method. For example, if you do want to 2188compare pointers to a particular inline method, you might mark it as 2189having default visibility. Marking the enclosing class with explicit 2190visibility will have no effect. 2191 2192Explicitly instantiated inline methods are unaffected by this option 2193as their linkage might otherwise cross a shared library boundary. 2194@xref{Template Instantiation}. 2195 2196@item -fvisibility-ms-compat 2197@opindex fvisibility-ms-compat 2198This flag attempts to use visibility settings to make GCC's C++ 2199linkage model compatible with that of Microsoft Visual Studio. 2200 2201The flag makes these changes to GCC's linkage model: 2202 2203@enumerate 2204@item 2205It sets the default visibility to @code{hidden}, like 2206@option{-fvisibility=hidden}. 2207 2208@item 2209Types, but not their members, are not hidden by default. 2210 2211@item 2212The One Definition Rule is relaxed for types without explicit 2213visibility specifications that are defined in more than one different 2214shared object: those declarations are permitted if they would have 2215been permitted when this option was not used. 2216@end enumerate 2217 2218In new code it is better to use @option{-fvisibility=hidden} and 2219export those classes that are intended to be externally visible. 2220Unfortunately it is possible for code to rely, perhaps accidentally, 2221on the Visual Studio behavior. 2222 2223Among the consequences of these changes are that static data members 2224of the same type with the same name but defined in different shared 2225objects will be different, so changing one will not change the other; 2226and that pointers to function members defined in different shared 2227objects may not compare equal. When this flag is given, it is a 2228violation of the ODR to define types with the same name differently. 2229 2230@item -fno-weak 2231@opindex fno-weak 2232Do not use weak symbol support, even if it is provided by the linker. 2233By default, G++ will use weak symbols if they are available. This 2234option exists only for testing, and should not be used by end-users; 2235it will result in inferior code and has no benefits. This option may 2236be removed in a future release of G++. 2237 2238@item -nostdinc++ 2239@opindex nostdinc++ 2240Do not search for header files in the standard directories specific to 2241C++, but do still search the other standard directories. (This option 2242is used when building the C++ library.) 2243@end table 2244 2245In addition, these optimization, warning, and code generation options 2246have meanings only for C++ programs: 2247 2248@table @gcctabopt 2249@item -fno-default-inline 2250@opindex fno-default-inline 2251Do not assume @samp{inline} for functions defined inside a class scope. 2252@xref{Optimize Options,,Options That Control Optimization}. Note that these 2253functions will have linkage like inline functions; they just won't be 2254inlined by default. 2255 2256@item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)} 2257@opindex Wabi 2258@opindex Wno-abi 2259Warn when G++ generates code that is probably not compatible with the 2260vendor-neutral C++ ABI@. Although an effort has been made to warn about 2261all such cases, there are probably some cases that are not warned about, 2262even though G++ is generating incompatible code. There may also be 2263cases where warnings are emitted even though the code that is generated 2264will be compatible. 2265 2266You should rewrite your code to avoid these warnings if you are 2267concerned about the fact that code generated by G++ may not be binary 2268compatible with code generated by other compilers. 2269 2270The known incompatibilities in @option{-fabi-version=2} (the default) include: 2271 2272@itemize @bullet 2273 2274@item 2275A template with a non-type template parameter of reference type is 2276mangled incorrectly: 2277@smallexample 2278extern int N; 2279template <int &> struct S @{@}; 2280void n (S<N>) @{2@} 2281@end smallexample 2282 2283This is fixed in @option{-fabi-version=3}. 2284 2285@item 2286SIMD vector types declared using @code{__attribute ((vector_size))} are 2287mangled in a non-standard way that does not allow for overloading of 2288functions taking vectors of different sizes. 2289 2290The mangling is changed in @option{-fabi-version=4}. 2291@end itemize 2292 2293The known incompatibilities in @option{-fabi-version=1} include: 2294 2295@itemize @bullet 2296 2297@item 2298Incorrect handling of tail-padding for bit-fields. G++ may attempt to 2299pack data into the same byte as a base class. For example: 2300 2301@smallexample 2302struct A @{ virtual void f(); int f1 : 1; @}; 2303struct B : public A @{ int f2 : 1; @}; 2304@end smallexample 2305 2306@noindent 2307In this case, G++ will place @code{B::f2} into the same byte 2308as@code{A::f1}; other compilers will not. You can avoid this problem 2309by explicitly padding @code{A} so that its size is a multiple of the 2310byte size on your platform; that will cause G++ and other compilers to 2311layout @code{B} identically. 2312 2313@item 2314Incorrect handling of tail-padding for virtual bases. G++ does not use 2315tail padding when laying out virtual bases. For example: 2316 2317@smallexample 2318struct A @{ virtual void f(); char c1; @}; 2319struct B @{ B(); char c2; @}; 2320struct C : public A, public virtual B @{@}; 2321@end smallexample 2322 2323@noindent 2324In this case, G++ will not place @code{B} into the tail-padding for 2325@code{A}; other compilers will. You can avoid this problem by 2326explicitly padding @code{A} so that its size is a multiple of its 2327alignment (ignoring virtual base classes); that will cause G++ and other 2328compilers to layout @code{C} identically. 2329 2330@item 2331Incorrect handling of bit-fields with declared widths greater than that 2332of their underlying types, when the bit-fields appear in a union. For 2333example: 2334 2335@smallexample 2336union U @{ int i : 4096; @}; 2337@end smallexample 2338 2339@noindent 2340Assuming that an @code{int} does not have 4096 bits, G++ will make the 2341union too small by the number of bits in an @code{int}. 2342 2343@item 2344Empty classes can be placed at incorrect offsets. For example: 2345 2346@smallexample 2347struct A @{@}; 2348 2349struct B @{ 2350 A a; 2351 virtual void f (); 2352@}; 2353 2354struct C : public B, public A @{@}; 2355@end smallexample 2356 2357@noindent 2358G++ will place the @code{A} base class of @code{C} at a nonzero offset; 2359it should be placed at offset zero. G++ mistakenly believes that the 2360@code{A} data member of @code{B} is already at offset zero. 2361 2362@item 2363Names of template functions whose types involve @code{typename} or 2364template template parameters can be mangled incorrectly. 2365 2366@smallexample 2367template <typename Q> 2368void f(typename Q::X) @{@} 2369 2370template <template <typename> class Q> 2371void f(typename Q<int>::X) @{@} 2372@end smallexample 2373 2374@noindent 2375Instantiations of these templates may be mangled incorrectly. 2376 2377@end itemize 2378 2379It also warns psABI related changes. The known psABI changes at this 2380point include: 2381 2382@itemize @bullet 2383 2384@item 2385For SYSV/x86-64, when passing union with long double, it is changed to 2386pass in memory as specified in psABI. For example: 2387 2388@smallexample 2389union U @{ 2390 long double ld; 2391 int i; 2392@}; 2393@end smallexample 2394 2395@noindent 2396@code{union U} will always be passed in memory. 2397 2398@end itemize 2399 2400@item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)} 2401@opindex Wctor-dtor-privacy 2402@opindex Wno-ctor-dtor-privacy 2403Warn when a class seems unusable because all the constructors or 2404destructors in that class are private, and it has neither friends nor 2405public static member functions. 2406 2407@item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)} 2408@opindex Wdelete-non-virtual-dtor 2409@opindex Wno-delete-non-virtual-dtor 2410Warn when @samp{delete} is used to destroy an instance of a class that 2411has virtual functions and non-virtual destructor. It is unsafe to delete 2412an instance of a derived class through a pointer to a base class if the 2413base class does not have a virtual destructor. This warning is enabled 2414by @option{-Wall}. 2415 2416@item -Wnarrowing @r{(C++ and Objective-C++ only)} 2417@opindex Wnarrowing 2418@opindex Wno-narrowing 2419Warn when a narrowing conversion prohibited by C++11 occurs within 2420@samp{@{ @}}, e.g. 2421 2422@smallexample 2423int i = @{ 2.2 @}; // error: narrowing from double to int 2424@end smallexample 2425 2426This flag is included in @option{-Wall} and @option{-Wc++11-compat}. 2427 2428With -std=c++11, @option{-Wno-narrowing} suppresses the diagnostic 2429required by the standard. Note that this does not affect the meaning 2430of well-formed code; narrowing conversions are still considered 2431ill-formed in SFINAE context. 2432 2433@item -Wnoexcept @r{(C++ and Objective-C++ only)} 2434@opindex Wnoexcept 2435@opindex Wno-noexcept 2436Warn when a noexcept-expression evaluates to false because of a call 2437to a function that does not have a non-throwing exception 2438specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by 2439the compiler to never throw an exception. 2440 2441@item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)} 2442@opindex Wnon-virtual-dtor 2443@opindex Wno-non-virtual-dtor 2444Warn when a class has virtual functions and accessible non-virtual 2445destructor, in which case it would be possible but unsafe to delete 2446an instance of a derived class through a pointer to the base class. 2447This warning is also enabled if @option{-Weffc++} is specified. 2448 2449@item -Wreorder @r{(C++ and Objective-C++ only)} 2450@opindex Wreorder 2451@opindex Wno-reorder 2452@cindex reordering, warning 2453@cindex warning for reordering of member initializers 2454Warn when the order of member initializers given in the code does not 2455match the order in which they must be executed. For instance: 2456 2457@smallexample 2458struct A @{ 2459 int i; 2460 int j; 2461 A(): j (0), i (1) @{ @} 2462@}; 2463@end smallexample 2464 2465The compiler will rearrange the member initializers for @samp{i} 2466and @samp{j} to match the declaration order of the members, emitting 2467a warning to that effect. This warning is enabled by @option{-Wall}. 2468@end table 2469 2470The following @option{-W@dots{}} options are not affected by @option{-Wall}. 2471 2472@table @gcctabopt 2473@item -Weffc++ @r{(C++ and Objective-C++ only)} 2474@opindex Weffc++ 2475@opindex Wno-effc++ 2476Warn about violations of the following style guidelines from Scott Meyers' 2477@cite{Effective C++, Second Edition} book: 2478 2479@itemize @bullet 2480@item 2481Item 11: Define a copy constructor and an assignment operator for classes 2482with dynamically allocated memory. 2483 2484@item 2485Item 12: Prefer initialization to assignment in constructors. 2486 2487@item 2488Item 14: Make destructors virtual in base classes. 2489 2490@item 2491Item 15: Have @code{operator=} return a reference to @code{*this}. 2492 2493@item 2494Item 23: Don't try to return a reference when you must return an object. 2495 2496@end itemize 2497 2498Also warn about violations of the following style guidelines from 2499Scott Meyers' @cite{More Effective C++} book: 2500 2501@itemize @bullet 2502@item 2503Item 6: Distinguish between prefix and postfix forms of increment and 2504decrement operators. 2505 2506@item 2507Item 7: Never overload @code{&&}, @code{||}, or @code{,}. 2508 2509@end itemize 2510 2511When selecting this option, be aware that the standard library 2512headers do not obey all of these guidelines; use @samp{grep -v} 2513to filter out those warnings. 2514 2515@item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)} 2516@opindex Wstrict-null-sentinel 2517@opindex Wno-strict-null-sentinel 2518Warn also about the use of an uncasted @code{NULL} as sentinel. When 2519compiling only with GCC this is a valid sentinel, as @code{NULL} is defined 2520to @code{__null}. Although it is a null pointer constant not a null pointer, 2521it is guaranteed to be of the same size as a pointer. But this use is 2522not portable across different compilers. 2523 2524@item -Wno-non-template-friend @r{(C++ and Objective-C++ only)} 2525@opindex Wno-non-template-friend 2526@opindex Wnon-template-friend 2527Disable warnings when non-templatized friend functions are declared 2528within a template. Since the advent of explicit template specification 2529support in G++, if the name of the friend is an unqualified-id (i.e., 2530@samp{friend foo(int)}), the C++ language specification demands that the 2531friend declare or define an ordinary, nontemplate function. (Section 253214.5.3). Before G++ implemented explicit specification, unqualified-ids 2533could be interpreted as a particular specialization of a templatized 2534function. Because this non-conforming behavior is no longer the default 2535behavior for G++, @option{-Wnon-template-friend} allows the compiler to 2536check existing code for potential trouble spots and is on by default. 2537This new compiler behavior can be turned off with 2538@option{-Wno-non-template-friend}, which keeps the conformant compiler code 2539but disables the helpful warning. 2540 2541@item -Wold-style-cast @r{(C++ and Objective-C++ only)} 2542@opindex Wold-style-cast 2543@opindex Wno-old-style-cast 2544Warn if an old-style (C-style) cast to a non-void type is used within 2545a C++ program. The new-style casts (@samp{dynamic_cast}, 2546@samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are 2547less vulnerable to unintended effects and much easier to search for. 2548 2549@item -Woverloaded-virtual @r{(C++ and Objective-C++ only)} 2550@opindex Woverloaded-virtual 2551@opindex Wno-overloaded-virtual 2552@cindex overloaded virtual function, warning 2553@cindex warning for overloaded virtual function 2554Warn when a function declaration hides virtual functions from a 2555base class. For example, in: 2556 2557@smallexample 2558struct A @{ 2559 virtual void f(); 2560@}; 2561 2562struct B: public A @{ 2563 void f(int); 2564@}; 2565@end smallexample 2566 2567the @code{A} class version of @code{f} is hidden in @code{B}, and code 2568like: 2569 2570@smallexample 2571B* b; 2572b->f(); 2573@end smallexample 2574 2575will fail to compile. 2576 2577@item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)} 2578@opindex Wno-pmf-conversions 2579@opindex Wpmf-conversions 2580Disable the diagnostic for converting a bound pointer to member function 2581to a plain pointer. 2582 2583@item -Wsign-promo @r{(C++ and Objective-C++ only)} 2584@opindex Wsign-promo 2585@opindex Wno-sign-promo 2586Warn when overload resolution chooses a promotion from unsigned or 2587enumerated type to a signed type, over a conversion to an unsigned type of 2588the same size. Previous versions of G++ would try to preserve 2589unsignedness, but the standard mandates the current behavior. 2590 2591@smallexample 2592struct A @{ 2593 operator int (); 2594 A& operator = (int); 2595@}; 2596 2597main () 2598@{ 2599 A a,b; 2600 a = b; 2601@} 2602@end smallexample 2603 2604In this example, G++ will synthesize a default @samp{A& operator = 2605(const A&);}, while cfront will use the user-defined @samp{operator =}. 2606@end table 2607 2608@node Objective-C and Objective-C++ Dialect Options 2609@section Options Controlling Objective-C and Objective-C++ Dialects 2610 2611@cindex compiler options, Objective-C and Objective-C++ 2612@cindex Objective-C and Objective-C++ options, command-line 2613@cindex options, Objective-C and Objective-C++ 2614(NOTE: This manual does not describe the Objective-C and Objective-C++ 2615languages themselves. @xref{Standards,,Language Standards 2616Supported by GCC}, for references.) 2617 2618This section describes the command-line options that are only meaningful 2619for Objective-C and Objective-C++ programs, but you can also use most of 2620the language-independent GNU compiler options. 2621For example, you might compile a file @code{some_class.m} like this: 2622 2623@smallexample 2624gcc -g -fgnu-runtime -O -c some_class.m 2625@end smallexample 2626 2627@noindent 2628In this example, @option{-fgnu-runtime} is an option meant only for 2629Objective-C and Objective-C++ programs; you can use the other options with 2630any language supported by GCC@. 2631 2632Note that since Objective-C is an extension of the C language, Objective-C 2633compilations may also use options specific to the C front-end (e.g., 2634@option{-Wtraditional}). Similarly, Objective-C++ compilations may use 2635C++-specific options (e.g., @option{-Wabi}). 2636 2637Here is a list of options that are @emph{only} for compiling Objective-C 2638and Objective-C++ programs: 2639 2640@table @gcctabopt 2641@item -fconstant-string-class=@var{class-name} 2642@opindex fconstant-string-class 2643Use @var{class-name} as the name of the class to instantiate for each 2644literal string specified with the syntax @code{@@"@dots{}"}. The default 2645class name is @code{NXConstantString} if the GNU runtime is being used, and 2646@code{NSConstantString} if the NeXT runtime is being used (see below). The 2647@option{-fconstant-cfstrings} option, if also present, will override the 2648@option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals 2649to be laid out as constant CoreFoundation strings. 2650 2651@item -fgnu-runtime 2652@opindex fgnu-runtime 2653Generate object code compatible with the standard GNU Objective-C 2654runtime. This is the default for most types of systems. 2655 2656@item -fnext-runtime 2657@opindex fnext-runtime 2658Generate output compatible with the NeXT runtime. This is the default 2659for NeXT-based systems, including Darwin and Mac OS X@. The macro 2660@code{__NEXT_RUNTIME__} is predefined if (and only if) this option is 2661used. 2662 2663@item -fno-nil-receivers 2664@opindex fno-nil-receivers 2665Assume that all Objective-C message dispatches (@code{[receiver 2666message:arg]}) in this translation unit ensure that the receiver is 2667not @code{nil}. This allows for more efficient entry points in the 2668runtime to be used. This option is only available in conjunction with 2669the NeXT runtime and ABI version 0 or 1. 2670 2671@item -fobjc-abi-version=@var{n} 2672@opindex fobjc-abi-version 2673Use version @var{n} of the Objective-C ABI for the selected runtime. 2674This option is currently supported only for the NeXT runtime. In that 2675case, Version 0 is the traditional (32-bit) ABI without support for 2676properties and other Objective-C 2.0 additions. Version 1 is the 2677traditional (32-bit) ABI with support for properties and other 2678Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If 2679nothing is specified, the default is Version 0 on 32-bit target 2680machines, and Version 2 on 64-bit target machines. 2681 2682@item -fobjc-call-cxx-cdtors 2683@opindex fobjc-call-cxx-cdtors 2684For each Objective-C class, check if any of its instance variables is a 2685C++ object with a non-trivial default constructor. If so, synthesize a 2686special @code{- (id) .cxx_construct} instance method which will run 2687non-trivial default constructors on any such instance variables, in order, 2688and then return @code{self}. Similarly, check if any instance variable 2689is a C++ object with a non-trivial destructor, and if so, synthesize a 2690special @code{- (void) .cxx_destruct} method which will run 2691all such default destructors, in reverse order. 2692 2693The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct} 2694methods thusly generated will only operate on instance variables 2695declared in the current Objective-C class, and not those inherited 2696from superclasses. It is the responsibility of the Objective-C 2697runtime to invoke all such methods in an object's inheritance 2698hierarchy. The @code{- (id) .cxx_construct} methods will be invoked 2699by the runtime immediately after a new object instance is allocated; 2700the @code{- (void) .cxx_destruct} methods will be invoked immediately 2701before the runtime deallocates an object instance. 2702 2703As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has 2704support for invoking the @code{- (id) .cxx_construct} and 2705@code{- (void) .cxx_destruct} methods. 2706 2707@item -fobjc-direct-dispatch 2708@opindex fobjc-direct-dispatch 2709Allow fast jumps to the message dispatcher. On Darwin this is 2710accomplished via the comm page. 2711 2712@item -fobjc-exceptions 2713@opindex fobjc-exceptions 2714Enable syntactic support for structured exception handling in 2715Objective-C, similar to what is offered by C++ and Java. This option 2716is required to use the Objective-C keywords @code{@@try}, 2717@code{@@throw}, @code{@@catch}, @code{@@finally} and 2718@code{@@synchronized}. This option is available with both the GNU 2719runtime and the NeXT runtime (but not available in conjunction with 2720the NeXT runtime on Mac OS X 10.2 and earlier). 2721 2722@item -fobjc-gc 2723@opindex fobjc-gc 2724Enable garbage collection (GC) in Objective-C and Objective-C++ 2725programs. This option is only available with the NeXT runtime; the 2726GNU runtime has a different garbage collection implementation that 2727does not require special compiler flags. 2728 2729@item -fobjc-nilcheck 2730@opindex fobjc-nilcheck 2731For the NeXT runtime with version 2 of the ABI, check for a nil 2732receiver in method invocations before doing the actual method call. 2733This is the default and can be disabled using 2734@option{-fno-objc-nilcheck}. Class methods and super calls are never 2735checked for nil in this way no matter what this flag is set to. 2736Currently this flag does nothing when the GNU runtime, or an older 2737version of the NeXT runtime ABI, is used. 2738 2739@item -fobjc-std=objc1 2740@opindex fobjc-std 2741Conform to the language syntax of Objective-C 1.0, the language 2742recognized by GCC 4.0. This only affects the Objective-C additions to 2743the C/C++ language; it does not affect conformance to C/C++ standards, 2744which is controlled by the separate C/C++ dialect option flags. When 2745this option is used with the Objective-C or Objective-C++ compiler, 2746any Objective-C syntax that is not recognized by GCC 4.0 is rejected. 2747This is useful if you need to make sure that your Objective-C code can 2748be compiled with older versions of GCC. 2749 2750@item -freplace-objc-classes 2751@opindex freplace-objc-classes 2752Emit a special marker instructing @command{ld(1)} not to statically link in 2753the resulting object file, and allow @command{dyld(1)} to load it in at 2754run time instead. This is used in conjunction with the Fix-and-Continue 2755debugging mode, where the object file in question may be recompiled and 2756dynamically reloaded in the course of program execution, without the need 2757to restart the program itself. Currently, Fix-and-Continue functionality 2758is only available in conjunction with the NeXT runtime on Mac OS X 10.3 2759and later. 2760 2761@item -fzero-link 2762@opindex fzero-link 2763When compiling for the NeXT runtime, the compiler ordinarily replaces calls 2764to @code{objc_getClass("@dots{}")} (when the name of the class is known at 2765compile time) with static class references that get initialized at load time, 2766which improves run-time performance. Specifying the @option{-fzero-link} flag 2767suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")} 2768to be retained. This is useful in Zero-Link debugging mode, since it allows 2769for individual class implementations to be modified during program execution. 2770The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")} 2771regardless of command-line options. 2772 2773@item -gen-decls 2774@opindex gen-decls 2775Dump interface declarations for all classes seen in the source file to a 2776file named @file{@var{sourcename}.decl}. 2777 2778@item -Wassign-intercept @r{(Objective-C and Objective-C++ only)} 2779@opindex Wassign-intercept 2780@opindex Wno-assign-intercept 2781Warn whenever an Objective-C assignment is being intercepted by the 2782garbage collector. 2783 2784@item -Wno-protocol @r{(Objective-C and Objective-C++ only)} 2785@opindex Wno-protocol 2786@opindex Wprotocol 2787If a class is declared to implement a protocol, a warning is issued for 2788every method in the protocol that is not implemented by the class. The 2789default behavior is to issue a warning for every method not explicitly 2790implemented in the class, even if a method implementation is inherited 2791from the superclass. If you use the @option{-Wno-protocol} option, then 2792methods inherited from the superclass are considered to be implemented, 2793and no warning is issued for them. 2794 2795@item -Wselector @r{(Objective-C and Objective-C++ only)} 2796@opindex Wselector 2797@opindex Wno-selector 2798Warn if multiple methods of different types for the same selector are 2799found during compilation. The check is performed on the list of methods 2800in the final stage of compilation. Additionally, a check is performed 2801for each selector appearing in a @code{@@selector(@dots{})} 2802expression, and a corresponding method for that selector has been found 2803during compilation. Because these checks scan the method table only at 2804the end of compilation, these warnings are not produced if the final 2805stage of compilation is not reached, for example because an error is 2806found during compilation, or because the @option{-fsyntax-only} option is 2807being used. 2808 2809@item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)} 2810@opindex Wstrict-selector-match 2811@opindex Wno-strict-selector-match 2812Warn if multiple methods with differing argument and/or return types are 2813found for a given selector when attempting to send a message using this 2814selector to a receiver of type @code{id} or @code{Class}. When this flag 2815is off (which is the default behavior), the compiler will omit such warnings 2816if any differences found are confined to types that share the same size 2817and alignment. 2818 2819@item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)} 2820@opindex Wundeclared-selector 2821@opindex Wno-undeclared-selector 2822Warn if a @code{@@selector(@dots{})} expression referring to an 2823undeclared selector is found. A selector is considered undeclared if no 2824method with that name has been declared before the 2825@code{@@selector(@dots{})} expression, either explicitly in an 2826@code{@@interface} or @code{@@protocol} declaration, or implicitly in 2827an @code{@@implementation} section. This option always performs its 2828checks as soon as a @code{@@selector(@dots{})} expression is found, 2829while @option{-Wselector} only performs its checks in the final stage of 2830compilation. This also enforces the coding style convention 2831that methods and selectors must be declared before being used. 2832 2833@item -print-objc-runtime-info 2834@opindex print-objc-runtime-info 2835Generate C header describing the largest structure that is passed by 2836value, if any. 2837 2838@end table 2839 2840@node Language Independent Options 2841@section Options to Control Diagnostic Messages Formatting 2842@cindex options to control diagnostics formatting 2843@cindex diagnostic messages 2844@cindex message formatting 2845 2846Traditionally, diagnostic messages have been formatted irrespective of 2847the output device's aspect (e.g.@: its width, @dots{}). The options described 2848below can be used to control the diagnostic messages formatting 2849algorithm, e.g.@: how many characters per line, how often source location 2850information should be reported. Right now, only the C++ front end can 2851honor these options. However it is expected, in the near future, that 2852the remaining front ends would be able to digest them correctly. 2853 2854@table @gcctabopt 2855@item -fmessage-length=@var{n} 2856@opindex fmessage-length 2857Try to format error messages so that they fit on lines of about @var{n} 2858characters. The default is 72 characters for @command{g++} and 0 for the rest of 2859the front ends supported by GCC@. If @var{n} is zero, then no 2860line-wrapping will be done; each error message will appear on a single 2861line. 2862 2863@opindex fdiagnostics-show-location 2864@item -fdiagnostics-show-location=once 2865Only meaningful in line-wrapping mode. Instructs the diagnostic messages 2866reporter to emit @emph{once} source location information; that is, in 2867case the message is too long to fit on a single physical line and has to 2868be wrapped, the source location won't be emitted (as prefix) again, 2869over and over, in subsequent continuation lines. This is the default 2870behavior. 2871 2872@item -fdiagnostics-show-location=every-line 2873Only meaningful in line-wrapping mode. Instructs the diagnostic 2874messages reporter to emit the same source location information (as 2875prefix) for physical lines that result from the process of breaking 2876a message which is too long to fit on a single line. 2877 2878@item -fno-diagnostics-show-option 2879@opindex fno-diagnostics-show-option 2880@opindex fdiagnostics-show-option 2881By default, each diagnostic emitted includes text indicating the 2882command-line option that directly controls the diagnostic (if such an 2883option is known to the diagnostic machinery). Specifying the 2884@option{-fno-diagnostics-show-option} flag suppresses that behavior. 2885 2886@end table 2887 2888@node Warning Options 2889@section Options to Request or Suppress Warnings 2890@cindex options to control warnings 2891@cindex warning messages 2892@cindex messages, warning 2893@cindex suppressing warnings 2894 2895Warnings are diagnostic messages that report constructions that 2896are not inherently erroneous but that are risky or suggest there 2897may have been an error. 2898 2899The following language-independent options do not enable specific 2900warnings but control the kinds of diagnostics produced by GCC. 2901 2902@table @gcctabopt 2903@cindex syntax checking 2904@item -fsyntax-only 2905@opindex fsyntax-only 2906Check the code for syntax errors, but don't do anything beyond that. 2907 2908@item -fmax-errors=@var{n} 2909@opindex fmax-errors 2910Limits the maximum number of error messages to @var{n}, at which point 2911GCC bails out rather than attempting to continue processing the source 2912code. If @var{n} is 0 (the default), there is no limit on the number 2913of error messages produced. If @option{-Wfatal-errors} is also 2914specified, then @option{-Wfatal-errors} takes precedence over this 2915option. 2916 2917@item -w 2918@opindex w 2919Inhibit all warning messages. 2920 2921@item -Werror 2922@opindex Werror 2923@opindex Wno-error 2924Make all warnings into errors. 2925 2926@item -Werror= 2927@opindex Werror= 2928@opindex Wno-error= 2929Make the specified warning into an error. The specifier for a warning 2930is appended, for example @option{-Werror=switch} turns the warnings 2931controlled by @option{-Wswitch} into errors. This switch takes a 2932negative form, to be used to negate @option{-Werror} for specific 2933warnings, for example @option{-Wno-error=switch} makes 2934@option{-Wswitch} warnings not be errors, even when @option{-Werror} 2935is in effect. 2936 2937The warning message for each controllable warning includes the 2938option that controls the warning. That option can then be used with 2939@option{-Werror=} and @option{-Wno-error=} as described above. 2940(Printing of the option in the warning message can be disabled using the 2941@option{-fno-diagnostics-show-option} flag.) 2942 2943Note that specifying @option{-Werror=}@var{foo} automatically implies 2944@option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not 2945imply anything. 2946 2947@item -Wfatal-errors 2948@opindex Wfatal-errors 2949@opindex Wno-fatal-errors 2950This option causes the compiler to abort compilation on the first error 2951occurred rather than trying to keep going and printing further error 2952messages. 2953 2954@end table 2955 2956You can request many specific warnings with options beginning 2957@samp{-W}, for example @option{-Wimplicit} to request warnings on 2958implicit declarations. Each of these specific warning options also 2959has a negative form beginning @samp{-Wno-} to turn off warnings; for 2960example, @option{-Wno-implicit}. This manual lists only one of the 2961two forms, whichever is not the default. For further, 2962language-specific options also refer to @ref{C++ Dialect Options} and 2963@ref{Objective-C and Objective-C++ Dialect Options}. 2964 2965When an unrecognized warning option is requested (e.g., 2966@option{-Wunknown-warning}), GCC will emit a diagnostic stating 2967that the option is not recognized. However, if the @option{-Wno-} form 2968is used, the behavior is slightly different: No diagnostic will be 2969produced for @option{-Wno-unknown-warning} unless other diagnostics 2970are being produced. This allows the use of new @option{-Wno-} options 2971with old compilers, but if something goes wrong, the compiler will 2972warn that an unrecognized option was used. 2973 2974@table @gcctabopt 2975@item -pedantic 2976@opindex pedantic 2977Issue all the warnings demanded by strict ISO C and ISO C++; 2978reject all programs that use forbidden extensions, and some other 2979programs that do not follow ISO C and ISO C++. For ISO C, follows the 2980version of the ISO C standard specified by any @option{-std} option used. 2981 2982Valid ISO C and ISO C++ programs should compile properly with or without 2983this option (though a rare few will require @option{-ansi} or a 2984@option{-std} option specifying the required version of ISO C)@. However, 2985without this option, certain GNU extensions and traditional C and C++ 2986features are supported as well. With this option, they are rejected. 2987 2988@option{-pedantic} does not cause warning messages for use of the 2989alternate keywords whose names begin and end with @samp{__}. Pedantic 2990warnings are also disabled in the expression that follows 2991@code{__extension__}. However, only system header files should use 2992these escape routes; application programs should avoid them. 2993@xref{Alternate Keywords}. 2994 2995Some users try to use @option{-pedantic} to check programs for strict ISO 2996C conformance. They soon find that it does not do quite what they want: 2997it finds some non-ISO practices, but not all---only those for which 2998ISO C @emph{requires} a diagnostic, and some others for which 2999diagnostics have been added. 3000 3001A feature to report any failure to conform to ISO C might be useful in 3002some instances, but would require considerable additional work and would 3003be quite different from @option{-pedantic}. We don't have plans to 3004support such a feature in the near future. 3005 3006Where the standard specified with @option{-std} represents a GNU 3007extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a 3008corresponding @dfn{base standard}, the version of ISO C on which the GNU 3009extended dialect is based. Warnings from @option{-pedantic} are given 3010where they are required by the base standard. (It would not make sense 3011for such warnings to be given only for features not in the specified GNU 3012C dialect, since by definition the GNU dialects of C include all 3013features the compiler supports with the given option, and there would be 3014nothing to warn about.) 3015 3016@item -pedantic-errors 3017@opindex pedantic-errors 3018Like @option{-pedantic}, except that errors are produced rather than 3019warnings. 3020 3021@item -Wall 3022@opindex Wall 3023@opindex Wno-all 3024This enables all the warnings about constructions that some users 3025consider questionable, and that are easy to avoid (or modify to 3026prevent the warning), even in conjunction with macros. This also 3027enables some language-specific warnings described in @ref{C++ Dialect 3028Options} and @ref{Objective-C and Objective-C++ Dialect Options}. 3029 3030@option{-Wall} turns on the following warning flags: 3031 3032@gccoptlist{-Waddress @gol 3033-Warray-bounds @r{(only with} @option{-O2}@r{)} @gol 3034-Wc++11-compat @gol 3035-Wchar-subscripts @gol 3036-Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol 3037-Wimplicit-int @r{(C and Objective-C only)} @gol 3038-Wimplicit-function-declaration @r{(C and Objective-C only)} @gol 3039-Wcomment @gol 3040-Wformat @gol 3041-Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol 3042-Wmaybe-uninitialized @gol 3043-Wmissing-braces @gol 3044-Wnonnull @gol 3045-Wparentheses @gol 3046-Wpointer-sign @gol 3047-Wreorder @gol 3048-Wreturn-type @gol 3049-Wsequence-point @gol 3050-Wsign-compare @r{(only in C++)} @gol 3051-Wstrict-aliasing @gol 3052-Wstrict-overflow=1 @gol 3053-Wswitch @gol 3054-Wtrigraphs @gol 3055-Wuninitialized @gol 3056-Wunknown-pragmas @gol 3057-Wunused-function @gol 3058-Wunused-label @gol 3059-Wunused-value @gol 3060-Wunused-variable @gol 3061-Wvolatile-register-var @gol 3062} 3063 3064Note that some warning flags are not implied by @option{-Wall}. Some of 3065them warn about constructions that users generally do not consider 3066questionable, but which occasionally you might wish to check for; 3067others warn about constructions that are necessary or hard to avoid in 3068some cases, and there is no simple way to modify the code to suppress 3069the warning. Some of them are enabled by @option{-Wextra} but many of 3070them must be enabled individually. 3071 3072@item -Wextra 3073@opindex W 3074@opindex Wextra 3075@opindex Wno-extra 3076This enables some extra warning flags that are not enabled by 3077@option{-Wall}. (This option used to be called @option{-W}. The older 3078name is still supported, but the newer name is more descriptive.) 3079 3080@gccoptlist{-Wclobbered @gol 3081-Wempty-body @gol 3082-Wignored-qualifiers @gol 3083-Wmissing-field-initializers @gol 3084-Wmissing-parameter-type @r{(C only)} @gol 3085-Wold-style-declaration @r{(C only)} @gol 3086-Woverride-init @gol 3087-Wsign-compare @gol 3088-Wtype-limits @gol 3089-Wuninitialized @gol 3090-Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol 3091-Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol 3092} 3093 3094The option @option{-Wextra} also prints warning messages for the 3095following cases: 3096 3097@itemize @bullet 3098 3099@item 3100A pointer is compared against integer zero with @samp{<}, @samp{<=}, 3101@samp{>}, or @samp{>=}. 3102 3103@item 3104(C++ only) An enumerator and a non-enumerator both appear in a 3105conditional expression. 3106 3107@item 3108(C++ only) Ambiguous virtual bases. 3109 3110@item 3111(C++ only) Subscripting an array that has been declared @samp{register}. 3112 3113@item 3114(C++ only) Taking the address of a variable that has been declared 3115@samp{register}. 3116 3117@item 3118(C++ only) A base class is not initialized in a derived class' copy 3119constructor. 3120 3121@end itemize 3122 3123@item -Wchar-subscripts 3124@opindex Wchar-subscripts 3125@opindex Wno-char-subscripts 3126Warn if an array subscript has type @code{char}. This is a common cause 3127of error, as programmers often forget that this type is signed on some 3128machines. 3129This warning is enabled by @option{-Wall}. 3130 3131@item -Wcomment 3132@opindex Wcomment 3133@opindex Wno-comment 3134Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*} 3135comment, or whenever a Backslash-Newline appears in a @samp{//} comment. 3136This warning is enabled by @option{-Wall}. 3137 3138@item -Wno-coverage-mismatch 3139@opindex Wno-coverage-mismatch 3140Warn if feedback profiles do not match when using the 3141@option{-fprofile-use} option. 3142If a source file was changed between @option{-fprofile-gen} and 3143@option{-fprofile-use}, the files with the profile feedback can fail 3144to match the source file and GCC cannot use the profile feedback 3145information. By default, this warning is enabled and is treated as an 3146error. @option{-Wno-coverage-mismatch} can be used to disable the 3147warning or @option{-Wno-error=coverage-mismatch} can be used to 3148disable the error. Disabling the error for this warning can result in 3149poorly optimized code and is useful only in the 3150case of very minor changes such as bug fixes to an existing code-base. 3151Completely disabling the warning is not recommended. 3152 3153@item -Wno-cpp 3154@r{(C, Objective-C, C++, Objective-C++ and Fortran only)} 3155 3156Suppress warning messages emitted by @code{#warning} directives. 3157 3158@item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)} 3159@opindex Wdouble-promotion 3160@opindex Wno-double-promotion 3161Give a warning when a value of type @code{float} is implicitly 3162promoted to @code{double}. CPUs with a 32-bit ``single-precision'' 3163floating-point unit implement @code{float} in hardware, but emulate 3164@code{double} in software. On such a machine, doing computations 3165using @code{double} values is much more expensive because of the 3166overhead required for software emulation. 3167 3168It is easy to accidentally do computations with @code{double} because 3169floating-point literals are implicitly of type @code{double}. For 3170example, in: 3171@smallexample 3172@group 3173float area(float radius) 3174@{ 3175 return 3.14159 * radius * radius; 3176@} 3177@end group 3178@end smallexample 3179the compiler will perform the entire computation with @code{double} 3180because the floating-point literal is a @code{double}. 3181 3182@item -Wformat 3183@opindex Wformat 3184@opindex Wno-format 3185@opindex ffreestanding 3186@opindex fno-builtin 3187Check calls to @code{printf} and @code{scanf}, etc., to make sure that 3188the arguments supplied have types appropriate to the format string 3189specified, and that the conversions specified in the format string make 3190sense. This includes standard functions, and others specified by format 3191attributes (@pxref{Function Attributes}), in the @code{printf}, 3192@code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension, 3193not in the C standard) families (or other target-specific families). 3194Which functions are checked without format attributes having been 3195specified depends on the standard version selected, and such checks of 3196functions without the attribute specified are disabled by 3197@option{-ffreestanding} or @option{-fno-builtin}. 3198 3199The formats are checked against the format features supported by GNU 3200libc version 2.2. These include all ISO C90 and C99 features, as well 3201as features from the Single Unix Specification and some BSD and GNU 3202extensions. Other library implementations may not support all these 3203features; GCC does not support warning about features that go beyond a 3204particular library's limitations. However, if @option{-pedantic} is used 3205with @option{-Wformat}, warnings will be given about format features not 3206in the selected standard version (but not for @code{strfmon} formats, 3207since those are not in any version of the C standard). @xref{C Dialect 3208Options,,Options Controlling C Dialect}. 3209 3210Since @option{-Wformat} also checks for null format arguments for 3211several functions, @option{-Wformat} also implies @option{-Wnonnull}. 3212 3213@option{-Wformat} is included in @option{-Wall}. For more control over some 3214aspects of format checking, the options @option{-Wformat-y2k}, 3215@option{-Wno-format-extra-args}, @option{-Wno-format-zero-length}, 3216@option{-Wformat-nonliteral}, @option{-Wformat-security}, and 3217@option{-Wformat=2} are available, but are not included in @option{-Wall}. 3218 3219@item -Wformat-y2k 3220@opindex Wformat-y2k 3221@opindex Wno-format-y2k 3222If @option{-Wformat} is specified, also warn about @code{strftime} 3223formats that may yield only a two-digit year. 3224 3225@item -Wno-format-contains-nul 3226@opindex Wno-format-contains-nul 3227@opindex Wformat-contains-nul 3228If @option{-Wformat} is specified, do not warn about format strings that 3229contain NUL bytes. 3230 3231@item -Wno-format-extra-args 3232@opindex Wno-format-extra-args 3233@opindex Wformat-extra-args 3234If @option{-Wformat} is specified, do not warn about excess arguments to a 3235@code{printf} or @code{scanf} format function. The C standard specifies 3236that such arguments are ignored. 3237 3238Where the unused arguments lie between used arguments that are 3239specified with @samp{$} operand number specifications, normally 3240warnings are still given, since the implementation could not know what 3241type to pass to @code{va_arg} to skip the unused arguments. However, 3242in the case of @code{scanf} formats, this option will suppress the 3243warning if the unused arguments are all pointers, since the Single 3244Unix Specification says that such unused arguments are allowed. 3245 3246@item -Wno-format-zero-length 3247@opindex Wno-format-zero-length 3248@opindex Wformat-zero-length 3249If @option{-Wformat} is specified, do not warn about zero-length formats. 3250The C standard specifies that zero-length formats are allowed. 3251 3252@item -Wformat-nonliteral 3253@opindex Wformat-nonliteral 3254@opindex Wno-format-nonliteral 3255If @option{-Wformat} is specified, also warn if the format string is not a 3256string literal and so cannot be checked, unless the format function 3257takes its format arguments as a @code{va_list}. 3258 3259@item -Wformat-security 3260@opindex Wformat-security 3261@opindex Wno-format-security 3262If @option{-Wformat} is specified, also warn about uses of format 3263functions that represent possible security problems. At present, this 3264warns about calls to @code{printf} and @code{scanf} functions where the 3265format string is not a string literal and there are no format arguments, 3266as in @code{printf (foo);}. This may be a security hole if the format 3267string came from untrusted input and contains @samp{%n}. (This is 3268currently a subset of what @option{-Wformat-nonliteral} warns about, but 3269in future warnings may be added to @option{-Wformat-security} that are not 3270included in @option{-Wformat-nonliteral}.) 3271 3272@item -Wformat=2 3273@opindex Wformat=2 3274@opindex Wno-format=2 3275Enable @option{-Wformat} plus format checks not included in 3276@option{-Wformat}. Currently equivalent to @samp{-Wformat 3277-Wformat-nonliteral -Wformat-security -Wformat-y2k}. 3278 3279@item -Wnonnull 3280@opindex Wnonnull 3281@opindex Wno-nonnull 3282Warn about passing a null pointer for arguments marked as 3283requiring a non-null value by the @code{nonnull} function attribute. 3284 3285@option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It 3286can be disabled with the @option{-Wno-nonnull} option. 3287 3288@item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)} 3289@opindex Winit-self 3290@opindex Wno-init-self 3291Warn about uninitialized variables that are initialized with themselves. 3292Note this option can only be used with the @option{-Wuninitialized} option. 3293 3294For example, GCC will warn about @code{i} being uninitialized in the 3295following snippet only when @option{-Winit-self} has been specified: 3296@smallexample 3297@group 3298int f() 3299@{ 3300 int i = i; 3301 return i; 3302@} 3303@end group 3304@end smallexample 3305 3306@item -Wimplicit-int @r{(C and Objective-C only)} 3307@opindex Wimplicit-int 3308@opindex Wno-implicit-int 3309Warn when a declaration does not specify a type. 3310This warning is enabled by @option{-Wall}. 3311 3312@item -Wimplicit-function-declaration @r{(C and Objective-C only)} 3313@opindex Wimplicit-function-declaration 3314@opindex Wno-implicit-function-declaration 3315Give a warning whenever a function is used before being declared. In 3316C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is 3317enabled by default and it is made into an error by 3318@option{-pedantic-errors}. This warning is also enabled by 3319@option{-Wall}. 3320 3321@item -Wimplicit @r{(C and Objective-C only)} 3322@opindex Wimplicit 3323@opindex Wno-implicit 3324Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}. 3325This warning is enabled by @option{-Wall}. 3326 3327@item -Wignored-qualifiers @r{(C and C++ only)} 3328@opindex Wignored-qualifiers 3329@opindex Wno-ignored-qualifiers 3330Warn if the return type of a function has a type qualifier 3331such as @code{const}. For ISO C such a type qualifier has no effect, 3332since the value returned by a function is not an lvalue. 3333For C++, the warning is only emitted for scalar types or @code{void}. 3334ISO C prohibits qualified @code{void} return types on function 3335definitions, so such return types always receive a warning 3336even without this option. 3337 3338This warning is also enabled by @option{-Wextra}. 3339 3340@item -Wmain 3341@opindex Wmain 3342@opindex Wno-main 3343Warn if the type of @samp{main} is suspicious. @samp{main} should be 3344a function with external linkage, returning int, taking either zero 3345arguments, two, or three arguments of appropriate types. This warning 3346is enabled by default in C++ and is enabled by either @option{-Wall} 3347or @option{-pedantic}. 3348 3349@item -Wmissing-braces 3350@opindex Wmissing-braces 3351@opindex Wno-missing-braces 3352Warn if an aggregate or union initializer is not fully bracketed. In 3353the following example, the initializer for @samp{a} is not fully 3354bracketed, but that for @samp{b} is fully bracketed. 3355 3356@smallexample 3357int a[2][2] = @{ 0, 1, 2, 3 @}; 3358int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @}; 3359@end smallexample 3360 3361This warning is enabled by @option{-Wall}. 3362 3363@item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)} 3364@opindex Wmissing-include-dirs 3365@opindex Wno-missing-include-dirs 3366Warn if a user-supplied include directory does not exist. 3367 3368@item -Wparentheses 3369@opindex Wparentheses 3370@opindex Wno-parentheses 3371Warn if parentheses are omitted in certain contexts, such 3372as when there is an assignment in a context where a truth value 3373is expected, or when operators are nested whose precedence people 3374often get confused about. 3375 3376Also warn if a comparison like @samp{x<=y<=z} appears; this is 3377equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different 3378interpretation from that of ordinary mathematical notation. 3379 3380Also warn about constructions where there may be confusion to which 3381@code{if} statement an @code{else} branch belongs. Here is an example of 3382such a case: 3383 3384@smallexample 3385@group 3386@{ 3387 if (a) 3388 if (b) 3389 foo (); 3390 else 3391 bar (); 3392@} 3393@end group 3394@end smallexample 3395 3396In C/C++, every @code{else} branch belongs to the innermost possible 3397@code{if} statement, which in this example is @code{if (b)}. This is 3398often not what the programmer expected, as illustrated in the above 3399example by indentation the programmer chose. When there is the 3400potential for this confusion, GCC will issue a warning when this flag 3401is specified. To eliminate the warning, add explicit braces around 3402the innermost @code{if} statement so there is no way the @code{else} 3403could belong to the enclosing @code{if}. The resulting code would 3404look like this: 3405 3406@smallexample 3407@group 3408@{ 3409 if (a) 3410 @{ 3411 if (b) 3412 foo (); 3413 else 3414 bar (); 3415 @} 3416@} 3417@end group 3418@end smallexample 3419 3420Also warn for dangerous uses of the 3421?: with omitted middle operand GNU extension. When the condition 3422in the ?: operator is a boolean expression the omitted value will 3423be always 1. Often the user expects it to be a value computed 3424inside the conditional expression instead. 3425 3426This warning is enabled by @option{-Wall}. 3427 3428@item -Wsequence-point 3429@opindex Wsequence-point 3430@opindex Wno-sequence-point 3431Warn about code that may have undefined semantics because of violations 3432of sequence point rules in the C and C++ standards. 3433 3434The C and C++ standards defines the order in which expressions in a C/C++ 3435program are evaluated in terms of @dfn{sequence points}, which represent 3436a partial ordering between the execution of parts of the program: those 3437executed before the sequence point, and those executed after it. These 3438occur after the evaluation of a full expression (one which is not part 3439of a larger expression), after the evaluation of the first operand of a 3440@code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a 3441function is called (but after the evaluation of its arguments and the 3442expression denoting the called function), and in certain other places. 3443Other than as expressed by the sequence point rules, the order of 3444evaluation of subexpressions of an expression is not specified. All 3445these rules describe only a partial order rather than a total order, 3446since, for example, if two functions are called within one expression 3447with no sequence point between them, the order in which the functions 3448are called is not specified. However, the standards committee have 3449ruled that function calls do not overlap. 3450 3451It is not specified when between sequence points modifications to the 3452values of objects take effect. Programs whose behavior depends on this 3453have undefined behavior; the C and C++ standards specify that ``Between 3454the previous and next sequence point an object shall have its stored 3455value modified at most once by the evaluation of an expression. 3456Furthermore, the prior value shall be read only to determine the value 3457to be stored.''. If a program breaks these rules, the results on any 3458particular implementation are entirely unpredictable. 3459 3460Examples of code with undefined behavior are @code{a = a++;}, @code{a[n] 3461= b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not 3462diagnosed by this option, and it may give an occasional false positive 3463result, but in general it has been found fairly effective at detecting 3464this sort of problem in programs. 3465 3466The standard is worded confusingly, therefore there is some debate 3467over the precise meaning of the sequence point rules in subtle cases. 3468Links to discussions of the problem, including proposed formal 3469definitions, may be found on the GCC readings page, at 3470@uref{http://gcc.gnu.org/@/readings.html}. 3471 3472This warning is enabled by @option{-Wall} for C and C++. 3473 3474@item -Wreturn-type 3475@opindex Wreturn-type 3476@opindex Wno-return-type 3477Warn whenever a function is defined with a return-type that defaults 3478to @code{int}. Also warn about any @code{return} statement with no 3479return-value in a function whose return-type is not @code{void} 3480(falling off the end of the function body is considered returning 3481without a value), and about a @code{return} statement with an 3482expression in a function whose return-type is @code{void}. 3483 3484For C++, a function without return type always produces a diagnostic 3485message, even when @option{-Wno-return-type} is specified. The only 3486exceptions are @samp{main} and functions defined in system headers. 3487 3488This warning is enabled by @option{-Wall}. 3489 3490@item -Wswitch 3491@opindex Wswitch 3492@opindex Wno-switch 3493Warn whenever a @code{switch} statement has an index of enumerated type 3494and lacks a @code{case} for one or more of the named codes of that 3495enumeration. (The presence of a @code{default} label prevents this 3496warning.) @code{case} labels outside the enumeration range also 3497provoke warnings when this option is used (even if there is a 3498@code{default} label). 3499This warning is enabled by @option{-Wall}. 3500 3501@item -Wswitch-default 3502@opindex Wswitch-default 3503@opindex Wno-switch-default 3504Warn whenever a @code{switch} statement does not have a @code{default} 3505case. 3506 3507@item -Wswitch-enum 3508@opindex Wswitch-enum 3509@opindex Wno-switch-enum 3510Warn whenever a @code{switch} statement has an index of enumerated type 3511and lacks a @code{case} for one or more of the named codes of that 3512enumeration. @code{case} labels outside the enumeration range also 3513provoke warnings when this option is used. The only difference 3514between @option{-Wswitch} and this option is that this option gives a 3515warning about an omitted enumeration code even if there is a 3516@code{default} label. 3517 3518@item -Wsync-nand @r{(C and C++ only)} 3519@opindex Wsync-nand 3520@opindex Wno-sync-nand 3521Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch} 3522built-in functions are used. These functions changed semantics in GCC 4.4. 3523 3524@item -Wtrigraphs 3525@opindex Wtrigraphs 3526@opindex Wno-trigraphs 3527Warn if any trigraphs are encountered that might change the meaning of 3528the program (trigraphs within comments are not warned about). 3529This warning is enabled by @option{-Wall}. 3530 3531@item -Wunused-but-set-parameter 3532@opindex Wunused-but-set-parameter 3533@opindex Wno-unused-but-set-parameter 3534Warn whenever a function parameter is assigned to, but otherwise unused 3535(aside from its declaration). 3536 3537To suppress this warning use the @samp{unused} attribute 3538(@pxref{Variable Attributes}). 3539 3540This warning is also enabled by @option{-Wunused} together with 3541@option{-Wextra}. 3542 3543@item -Wunused-but-set-variable 3544@opindex Wunused-but-set-variable 3545@opindex Wno-unused-but-set-variable 3546Warn whenever a local variable is assigned to, but otherwise unused 3547(aside from its declaration). 3548This warning is enabled by @option{-Wall}. 3549 3550To suppress this warning use the @samp{unused} attribute 3551(@pxref{Variable Attributes}). 3552 3553This warning is also enabled by @option{-Wunused}, which is enabled 3554by @option{-Wall}. 3555 3556@item -Wunused-function 3557@opindex Wunused-function 3558@opindex Wno-unused-function 3559Warn whenever a static function is declared but not defined or a 3560non-inline static function is unused. 3561This warning is enabled by @option{-Wall}. 3562 3563@item -Wunused-label 3564@opindex Wunused-label 3565@opindex Wno-unused-label 3566Warn whenever a label is declared but not used. 3567This warning is enabled by @option{-Wall}. 3568 3569To suppress this warning use the @samp{unused} attribute 3570(@pxref{Variable Attributes}). 3571 3572@item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)} 3573@opindex Wunused-local-typedefs 3574Warn when a typedef locally defined in a function is not used. 3575 3576@item -Wunused-parameter 3577@opindex Wunused-parameter 3578@opindex Wno-unused-parameter 3579Warn whenever a function parameter is unused aside from its declaration. 3580 3581To suppress this warning use the @samp{unused} attribute 3582(@pxref{Variable Attributes}). 3583 3584@item -Wno-unused-result 3585@opindex Wunused-result 3586@opindex Wno-unused-result 3587Do not warn if a caller of a function marked with attribute 3588@code{warn_unused_result} (@pxref{Function Attributes}) does not use 3589its return value. The default is @option{-Wunused-result}. 3590 3591@item -Wunused-variable 3592@opindex Wunused-variable 3593@opindex Wno-unused-variable 3594Warn whenever a local variable or non-constant static variable is unused 3595aside from its declaration. 3596This warning is enabled by @option{-Wall}. 3597 3598To suppress this warning use the @samp{unused} attribute 3599(@pxref{Variable Attributes}). 3600 3601@item -Wunused-value 3602@opindex Wunused-value 3603@opindex Wno-unused-value 3604Warn whenever a statement computes a result that is explicitly not 3605used. To suppress this warning cast the unused expression to 3606@samp{void}. This includes an expression-statement or the left-hand 3607side of a comma expression that contains no side effects. For example, 3608an expression such as @samp{x[i,j]} will cause a warning, while 3609@samp{x[(void)i,j]} will not. 3610 3611This warning is enabled by @option{-Wall}. 3612 3613@item -Wunused 3614@opindex Wunused 3615@opindex Wno-unused 3616All the above @option{-Wunused} options combined. 3617 3618In order to get a warning about an unused function parameter, you must 3619either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies 3620@samp{-Wunused}), or separately specify @option{-Wunused-parameter}. 3621 3622@item -Wuninitialized 3623@opindex Wuninitialized 3624@opindex Wno-uninitialized 3625Warn if an automatic variable is used without first being initialized 3626or if a variable may be clobbered by a @code{setjmp} call. In C++, 3627warn if a non-static reference or non-static @samp{const} member 3628appears in a class without constructors. 3629 3630If you want to warn about code that uses the uninitialized value of the 3631variable in its own initializer, use the @option{-Winit-self} option. 3632 3633These warnings occur for individual uninitialized or clobbered 3634elements of structure, union or array variables as well as for 3635variables that are uninitialized or clobbered as a whole. They do 3636not occur for variables or elements declared @code{volatile}. Because 3637these warnings depend on optimization, the exact variables or elements 3638for which there are warnings will depend on the precise optimization 3639options and version of GCC used. 3640 3641Note that there may be no warning about a variable that is used only 3642to compute a value that itself is never used, because such 3643computations may be deleted by data flow analysis before the warnings 3644are printed. 3645 3646@item -Wmaybe-uninitialized 3647@opindex Wmaybe-uninitialized 3648@opindex Wno-maybe-uninitialized 3649For an automatic variable, if there exists a path from the function 3650entry to a use of the variable that is initialized, but there exist 3651some other paths the variable is not initialized, the compiler will 3652emit a warning if it can not prove the uninitialized paths do not 3653happen at run time. These warnings are made optional because GCC is 3654not smart enough to see all the reasons why the code might be correct 3655despite appearing to have an error. Here is one example of how 3656this can happen: 3657 3658@smallexample 3659@group 3660@{ 3661 int x; 3662 switch (y) 3663 @{ 3664 case 1: x = 1; 3665 break; 3666 case 2: x = 4; 3667 break; 3668 case 3: x = 5; 3669 @} 3670 foo (x); 3671@} 3672@end group 3673@end smallexample 3674 3675@noindent 3676If the value of @code{y} is always 1, 2 or 3, then @code{x} is 3677always initialized, but GCC doesn't know this. To suppress the 3678warning, the user needs to provide a default case with assert(0) or 3679similar code. 3680 3681@cindex @code{longjmp} warnings 3682This option also warns when a non-volatile automatic variable might be 3683changed by a call to @code{longjmp}. These warnings as well are possible 3684only in optimizing compilation. 3685 3686The compiler sees only the calls to @code{setjmp}. It cannot know 3687where @code{longjmp} will be called; in fact, a signal handler could 3688call it at any point in the code. As a result, you may get a warning 3689even when there is in fact no problem because @code{longjmp} cannot 3690in fact be called at the place that would cause a problem. 3691 3692Some spurious warnings can be avoided if you declare all the functions 3693you use that never return as @code{noreturn}. @xref{Function 3694Attributes}. 3695 3696This warning is enabled by @option{-Wall} or @option{-Wextra}. 3697 3698@item -Wunknown-pragmas 3699@opindex Wunknown-pragmas 3700@opindex Wno-unknown-pragmas 3701@cindex warning for unknown pragmas 3702@cindex unknown pragmas, warning 3703@cindex pragmas, warning of unknown 3704Warn when a @code{#pragma} directive is encountered that is not understood by 3705GCC@. If this command-line option is used, warnings will even be issued 3706for unknown pragmas in system header files. This is not the case if 3707the warnings were only enabled by the @option{-Wall} command-line option. 3708 3709@item -Wno-pragmas 3710@opindex Wno-pragmas 3711@opindex Wpragmas 3712Do not warn about misuses of pragmas, such as incorrect parameters, 3713invalid syntax, or conflicts between pragmas. See also 3714@samp{-Wunknown-pragmas}. 3715 3716@item -Wstrict-aliasing 3717@opindex Wstrict-aliasing 3718@opindex Wno-strict-aliasing 3719This option is only active when @option{-fstrict-aliasing} is active. 3720It warns about code that might break the strict aliasing rules that the 3721compiler is using for optimization. The warning does not catch all 3722cases, but does attempt to catch the more common pitfalls. It is 3723included in @option{-Wall}. 3724It is equivalent to @option{-Wstrict-aliasing=3} 3725 3726@item -Wstrict-aliasing=n 3727@opindex Wstrict-aliasing=n 3728@opindex Wno-strict-aliasing=n 3729This option is only active when @option{-fstrict-aliasing} is active. 3730It warns about code that might break the strict aliasing rules that the 3731compiler is using for optimization. 3732Higher levels correspond to higher accuracy (fewer false positives). 3733Higher levels also correspond to more effort, similar to the way -O works. 3734@option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n}, 3735with n=3. 3736 3737Level 1: Most aggressive, quick, least accurate. 3738Possibly useful when higher levels 3739do not warn but -fstrict-aliasing still breaks the code, as it has very few 3740false negatives. However, it has many false positives. 3741Warns for all pointer conversions between possibly incompatible types, 3742even if never dereferenced. Runs in the front end only. 3743 3744Level 2: Aggressive, quick, not too precise. 3745May still have many false positives (not as many as level 1 though), 3746and few false negatives (but possibly more than level 1). 3747Unlike level 1, it only warns when an address is taken. Warns about 3748incomplete types. Runs in the front end only. 3749 3750Level 3 (default for @option{-Wstrict-aliasing}): 3751Should have very few false positives and few false 3752negatives. Slightly slower than levels 1 or 2 when optimization is enabled. 3753Takes care of the common pun+dereference pattern in the front end: 3754@code{*(int*)&some_float}. 3755If optimization is enabled, it also runs in the back end, where it deals 3756with multiple statement cases using flow-sensitive points-to information. 3757Only warns when the converted pointer is dereferenced. 3758Does not warn about incomplete types. 3759 3760@item -Wstrict-overflow 3761@itemx -Wstrict-overflow=@var{n} 3762@opindex Wstrict-overflow 3763@opindex Wno-strict-overflow 3764This option is only active when @option{-fstrict-overflow} is active. 3765It warns about cases where the compiler optimizes based on the 3766assumption that signed overflow does not occur. Note that it does not 3767warn about all cases where the code might overflow: it only warns 3768about cases where the compiler implements some optimization. Thus 3769this warning depends on the optimization level. 3770 3771An optimization that assumes that signed overflow does not occur is 3772perfectly safe if the values of the variables involved are such that 3773overflow never does, in fact, occur. Therefore this warning can 3774easily give a false positive: a warning about code that is not 3775actually a problem. To help focus on important issues, several 3776warning levels are defined. No warnings are issued for the use of 3777undefined signed overflow when estimating how many iterations a loop 3778will require, in particular when determining whether a loop will be 3779executed at all. 3780 3781@table @gcctabopt 3782@item -Wstrict-overflow=1 3783Warn about cases that are both questionable and easy to avoid. For 3784example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the 3785compiler will simplify this to @code{1}. This level of 3786@option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels 3787are not, and must be explicitly requested. 3788 3789@item -Wstrict-overflow=2 3790Also warn about other cases where a comparison is simplified to a 3791constant. For example: @code{abs (x) >= 0}. This can only be 3792simplified when @option{-fstrict-overflow} is in effect, because 3793@code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than 3794zero. @option{-Wstrict-overflow} (with no level) is the same as 3795@option{-Wstrict-overflow=2}. 3796 3797@item -Wstrict-overflow=3 3798Also warn about other cases where a comparison is simplified. For 3799example: @code{x + 1 > 1} will be simplified to @code{x > 0}. 3800 3801@item -Wstrict-overflow=4 3802Also warn about other simplifications not covered by the above cases. 3803For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}. 3804 3805@item -Wstrict-overflow=5 3806Also warn about cases where the compiler reduces the magnitude of a 3807constant involved in a comparison. For example: @code{x + 2 > y} will 3808be simplified to @code{x + 1 >= y}. This is reported only at the 3809highest warning level because this simplification applies to many 3810comparisons, so this warning level will give a very large number of 3811false positives. 3812@end table 3813 3814@item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} 3815@opindex Wsuggest-attribute= 3816@opindex Wno-suggest-attribute= 3817Warn for cases where adding an attribute may be beneficial. The 3818attributes currently supported are listed below. 3819 3820@table @gcctabopt 3821@item -Wsuggest-attribute=pure 3822@itemx -Wsuggest-attribute=const 3823@itemx -Wsuggest-attribute=noreturn 3824@opindex Wsuggest-attribute=pure 3825@opindex Wno-suggest-attribute=pure 3826@opindex Wsuggest-attribute=const 3827@opindex Wno-suggest-attribute=const 3828@opindex Wsuggest-attribute=noreturn 3829@opindex Wno-suggest-attribute=noreturn 3830 3831Warn about functions that might be candidates for attributes 3832@code{pure}, @code{const} or @code{noreturn}. The compiler only warns for 3833functions visible in other compilation units or (in the case of @code{pure} and 3834@code{const}) if it cannot prove that the function returns normally. A function 3835returns normally if it doesn't contain an infinite loop nor returns abnormally 3836by throwing, calling @code{abort()} or trapping. This analysis requires option 3837@option{-fipa-pure-const}, which is enabled by default at @option{-O} and 3838higher. Higher optimization levels improve the accuracy of the analysis. 3839@end table 3840 3841@item -Warray-bounds 3842@opindex Wno-array-bounds 3843@opindex Warray-bounds 3844This option is only active when @option{-ftree-vrp} is active 3845(default for @option{-O2} and above). It warns about subscripts to arrays 3846that are always out of bounds. This warning is enabled by @option{-Wall}. 3847 3848@item -Wno-div-by-zero 3849@opindex Wno-div-by-zero 3850@opindex Wdiv-by-zero 3851Do not warn about compile-time integer division by zero. Floating-point 3852division by zero is not warned about, as it can be a legitimate way of 3853obtaining infinities and NaNs. 3854 3855@item -Wsystem-headers 3856@opindex Wsystem-headers 3857@opindex Wno-system-headers 3858@cindex warnings from system headers 3859@cindex system headers, warnings from 3860Print warning messages for constructs found in system header files. 3861Warnings from system headers are normally suppressed, on the assumption 3862that they usually do not indicate real problems and would only make the 3863compiler output harder to read. Using this command-line option tells 3864GCC to emit warnings from system headers as if they occurred in user 3865code. However, note that using @option{-Wall} in conjunction with this 3866option will @emph{not} warn about unknown pragmas in system 3867headers---for that, @option{-Wunknown-pragmas} must also be used. 3868 3869@item -Wtrampolines 3870@opindex Wtrampolines 3871@opindex Wno-trampolines 3872 Warn about trampolines generated for pointers to nested functions. 3873 3874 A trampoline is a small piece of data or code that is created at run 3875 time on the stack when the address of a nested function is taken, and 3876 is used to call the nested function indirectly. For some targets, it 3877 is made up of data only and thus requires no special treatment. But, 3878 for most targets, it is made up of code and thus requires the stack 3879 to be made executable in order for the program to work properly. 3880 3881@item -Wfloat-equal 3882@opindex Wfloat-equal 3883@opindex Wno-float-equal 3884Warn if floating-point values are used in equality comparisons. 3885 3886The idea behind this is that sometimes it is convenient (for the 3887programmer) to consider floating-point values as approximations to 3888infinitely precise real numbers. If you are doing this, then you need 3889to compute (by analyzing the code, or in some other way) the maximum or 3890likely maximum error that the computation introduces, and allow for it 3891when performing comparisons (and when producing output, but that's a 3892different problem). In particular, instead of testing for equality, you 3893would check to see whether the two values have ranges that overlap; and 3894this is done with the relational operators, so equality comparisons are 3895probably mistaken. 3896 3897@item -Wtraditional @r{(C and Objective-C only)} 3898@opindex Wtraditional 3899@opindex Wno-traditional 3900Warn about certain constructs that behave differently in traditional and 3901ISO C@. Also warn about ISO C constructs that have no traditional C 3902equivalent, and/or problematic constructs that should be avoided. 3903 3904@itemize @bullet 3905@item 3906Macro parameters that appear within string literals in the macro body. 3907In traditional C macro replacement takes place within string literals, 3908but does not in ISO C@. 3909 3910@item 3911In traditional C, some preprocessor directives did not exist. 3912Traditional preprocessors would only consider a line to be a directive 3913if the @samp{#} appeared in column 1 on the line. Therefore 3914@option{-Wtraditional} warns about directives that traditional C 3915understands but would ignore because the @samp{#} does not appear as the 3916first character on the line. It also suggests you hide directives like 3917@samp{#pragma} not understood by traditional C by indenting them. Some 3918traditional implementations would not recognize @samp{#elif}, so it 3919suggests avoiding it altogether. 3920 3921@item 3922A function-like macro that appears without arguments. 3923 3924@item 3925The unary plus operator. 3926 3927@item 3928The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point 3929constant suffixes. (Traditional C does support the @samp{L} suffix on integer 3930constants.) Note, these suffixes appear in macros defined in the system 3931headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}. 3932Use of these macros in user code might normally lead to spurious 3933warnings, however GCC's integrated preprocessor has enough context to 3934avoid warning in these cases. 3935 3936@item 3937A function declared external in one block and then used after the end of 3938the block. 3939 3940@item 3941A @code{switch} statement has an operand of type @code{long}. 3942 3943@item 3944A non-@code{static} function declaration follows a @code{static} one. 3945This construct is not accepted by some traditional C compilers. 3946 3947@item 3948The ISO type of an integer constant has a different width or 3949signedness from its traditional type. This warning is only issued if 3950the base of the constant is ten. I.e.@: hexadecimal or octal values, which 3951typically represent bit patterns, are not warned about. 3952 3953@item 3954Usage of ISO string concatenation is detected. 3955 3956@item 3957Initialization of automatic aggregates. 3958 3959@item 3960Identifier conflicts with labels. Traditional C lacks a separate 3961namespace for labels. 3962 3963@item 3964Initialization of unions. If the initializer is zero, the warning is 3965omitted. This is done under the assumption that the zero initializer in 3966user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing 3967initializer warnings and relies on default initialization to zero in the 3968traditional C case. 3969 3970@item 3971Conversions by prototypes between fixed/floating-point values and vice 3972versa. The absence of these prototypes when compiling with traditional 3973C would cause serious problems. This is a subset of the possible 3974conversion warnings, for the full set use @option{-Wtraditional-conversion}. 3975 3976@item 3977Use of ISO C style function definitions. This warning intentionally is 3978@emph{not} issued for prototype declarations or variadic functions 3979because these ISO C features will appear in your code when using 3980libiberty's traditional C compatibility macros, @code{PARAMS} and 3981@code{VPARAMS}. This warning is also bypassed for nested functions 3982because that feature is already a GCC extension and thus not relevant to 3983traditional C compatibility. 3984@end itemize 3985 3986@item -Wtraditional-conversion @r{(C and Objective-C only)} 3987@opindex Wtraditional-conversion 3988@opindex Wno-traditional-conversion 3989Warn if a prototype causes a type conversion that is different from what 3990would happen to the same argument in the absence of a prototype. This 3991includes conversions of fixed point to floating and vice versa, and 3992conversions changing the width or signedness of a fixed-point argument 3993except when the same as the default promotion. 3994 3995@item -Wdeclaration-after-statement @r{(C and Objective-C only)} 3996@opindex Wdeclaration-after-statement 3997@opindex Wno-declaration-after-statement 3998Warn when a declaration is found after a statement in a block. This 3999construct, known from C++, was introduced with ISO C99 and is by default 4000allowed in GCC@. It is not supported by ISO C90 and was not supported by 4001GCC versions before GCC 3.0. @xref{Mixed Declarations}. 4002 4003@item -Wundef 4004@opindex Wundef 4005@opindex Wno-undef 4006Warn if an undefined identifier is evaluated in an @samp{#if} directive. 4007 4008@item -Wno-endif-labels 4009@opindex Wno-endif-labels 4010@opindex Wendif-labels 4011Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text. 4012 4013@item -Wshadow 4014@opindex Wshadow 4015@opindex Wno-shadow 4016Warn whenever a local variable or type declaration shadows another variable, 4017parameter, type, or class member (in C++), or whenever a built-in function 4018is shadowed. Note that in C++, the compiler will not warn if a local variable 4019shadows a struct/class/enum, but will warn if it shadows an explicit typedef. 4020 4021@item -Wlarger-than=@var{len} 4022@opindex Wlarger-than=@var{len} 4023@opindex Wlarger-than-@var{len} 4024Warn whenever an object of larger than @var{len} bytes is defined. 4025 4026@item -Wframe-larger-than=@var{len} 4027@opindex Wframe-larger-than 4028Warn if the size of a function frame is larger than @var{len} bytes. 4029The computation done to determine the stack frame size is approximate 4030and not conservative. 4031The actual requirements may be somewhat greater than @var{len} 4032even if you do not get a warning. In addition, any space allocated 4033via @code{alloca}, variable-length arrays, or related constructs 4034is not included by the compiler when determining 4035whether or not to issue a warning. 4036 4037@item -Wno-free-nonheap-object 4038@opindex Wno-free-nonheap-object 4039@opindex Wfree-nonheap-object 4040Do not warn when attempting to free an object that was not allocated 4041on the heap. 4042 4043@item -Wstack-usage=@var{len} 4044@opindex Wstack-usage 4045Warn if the stack usage of a function might be larger than @var{len} bytes. 4046The computation done to determine the stack usage is conservative. 4047Any space allocated via @code{alloca}, variable-length arrays, or related 4048constructs is included by the compiler when determining whether or not to 4049issue a warning. 4050 4051The message is in keeping with the output of @option{-fstack-usage}. 4052 4053@itemize 4054@item 4055If the stack usage is fully static but exceeds the specified amount, it's: 4056 4057@smallexample 4058 warning: stack usage is 1120 bytes 4059@end smallexample 4060@item 4061If the stack usage is (partly) dynamic but bounded, it's: 4062 4063@smallexample 4064 warning: stack usage might be 1648 bytes 4065@end smallexample 4066@item 4067If the stack usage is (partly) dynamic and not bounded, it's: 4068 4069@smallexample 4070 warning: stack usage might be unbounded 4071@end smallexample 4072@end itemize 4073 4074@item -Wunsafe-loop-optimizations 4075@opindex Wunsafe-loop-optimizations 4076@opindex Wno-unsafe-loop-optimizations 4077Warn if the loop cannot be optimized because the compiler could not 4078assume anything on the bounds of the loop indices. With 4079@option{-funsafe-loop-optimizations} warn if the compiler made 4080such assumptions. 4081 4082@item -Wno-pedantic-ms-format @r{(MinGW targets only)} 4083@opindex Wno-pedantic-ms-format 4084@opindex Wpedantic-ms-format 4085Disables the warnings about non-ISO @code{printf} / @code{scanf} format 4086width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets 4087depending on the MS runtime, when you are using the options @option{-Wformat} 4088and @option{-pedantic} without gnu-extensions. 4089 4090@item -Wpointer-arith 4091@opindex Wpointer-arith 4092@opindex Wno-pointer-arith 4093Warn about anything that depends on the ``size of'' a function type or 4094of @code{void}. GNU C assigns these types a size of 1, for 4095convenience in calculations with @code{void *} pointers and pointers 4096to functions. In C++, warn also when an arithmetic operation involves 4097@code{NULL}. This warning is also enabled by @option{-pedantic}. 4098 4099@item -Wtype-limits 4100@opindex Wtype-limits 4101@opindex Wno-type-limits 4102Warn if a comparison is always true or always false due to the limited 4103range of the data type, but do not warn for constant expressions. For 4104example, warn if an unsigned variable is compared against zero with 4105@samp{<} or @samp{>=}. This warning is also enabled by 4106@option{-Wextra}. 4107 4108@item -Wbad-function-cast @r{(C and Objective-C only)} 4109@opindex Wbad-function-cast 4110@opindex Wno-bad-function-cast 4111Warn whenever a function call is cast to a non-matching type. 4112For example, warn if @code{int malloc()} is cast to @code{anything *}. 4113 4114@item -Wc++-compat @r{(C and Objective-C only)} 4115Warn about ISO C constructs that are outside of the common subset of 4116ISO C and ISO C++, e.g.@: request for implicit conversion from 4117@code{void *} to a pointer to non-@code{void} type. 4118 4119@item -Wc++11-compat @r{(C++ and Objective-C++ only)} 4120Warn about C++ constructs whose meaning differs between ISO C++ 1998 4121and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords 4122in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is 4123enabled by @option{-Wall}. 4124 4125@item -Wcast-qual 4126@opindex Wcast-qual 4127@opindex Wno-cast-qual 4128Warn whenever a pointer is cast so as to remove a type qualifier from 4129the target type. For example, warn if a @code{const char *} is cast 4130to an ordinary @code{char *}. 4131 4132Also warn when making a cast that introduces a type qualifier in an 4133unsafe way. For example, casting @code{char **} to @code{const char **} 4134is unsafe, as in this example: 4135 4136@smallexample 4137 /* p is char ** value. */ 4138 const char **q = (const char **) p; 4139 /* Assignment of readonly string to const char * is OK. */ 4140 *q = "string"; 4141 /* Now char** pointer points to read-only memory. */ 4142 **p = 'b'; 4143@end smallexample 4144 4145@item -Wcast-align 4146@opindex Wcast-align 4147@opindex Wno-cast-align 4148Warn whenever a pointer is cast such that the required alignment of the 4149target is increased. For example, warn if a @code{char *} is cast to 4150an @code{int *} on machines where integers can only be accessed at 4151two- or four-byte boundaries. 4152 4153@item -Wwrite-strings 4154@opindex Wwrite-strings 4155@opindex Wno-write-strings 4156When compiling C, give string constants the type @code{const 4157char[@var{length}]} so that copying the address of one into a 4158non-@code{const} @code{char *} pointer will get a warning. These 4159warnings will help you find at compile time code that can try to write 4160into a string constant, but only if you have been very careful about 4161using @code{const} in declarations and prototypes. Otherwise, it will 4162just be a nuisance. This is why we did not make @option{-Wall} request 4163these warnings. 4164 4165When compiling C++, warn about the deprecated conversion from string 4166literals to @code{char *}. This warning is enabled by default for C++ 4167programs. 4168 4169@item -Wclobbered 4170@opindex Wclobbered 4171@opindex Wno-clobbered 4172Warn for variables that might be changed by @samp{longjmp} or 4173@samp{vfork}. This warning is also enabled by @option{-Wextra}. 4174 4175@item -Wconversion 4176@opindex Wconversion 4177@opindex Wno-conversion 4178Warn for implicit conversions that may alter a value. This includes 4179conversions between real and integer, like @code{abs (x)} when 4180@code{x} is @code{double}; conversions between signed and unsigned, 4181like @code{unsigned ui = -1}; and conversions to smaller types, like 4182@code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs 4183((int) x)} and @code{ui = (unsigned) -1}, or if the value is not 4184changed by the conversion like in @code{abs (2.0)}. Warnings about 4185conversions between signed and unsigned integers can be disabled by 4186using @option{-Wno-sign-conversion}. 4187 4188For C++, also warn for confusing overload resolution for user-defined 4189conversions; and conversions that will never use a type conversion 4190operator: conversions to @code{void}, the same type, a base class or a 4191reference to them. Warnings about conversions between signed and 4192unsigned integers are disabled by default in C++ unless 4193@option{-Wsign-conversion} is explicitly enabled. 4194 4195@item -Wno-conversion-null @r{(C++ and Objective-C++ only)} 4196@opindex Wconversion-null 4197@opindex Wno-conversion-null 4198Do not warn for conversions between @code{NULL} and non-pointer 4199types. @option{-Wconversion-null} is enabled by default. 4200 4201@item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)} 4202@opindex Wzero-as-null-pointer-constant 4203@opindex Wno-zero-as-null-pointer-constant 4204Warn when a literal '0' is used as null pointer constant. This can 4205be useful to facilitate the conversion to @code{nullptr} in C++11. 4206 4207@item -Wempty-body 4208@opindex Wempty-body 4209@opindex Wno-empty-body 4210Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do 4211while} statement. This warning is also enabled by @option{-Wextra}. 4212 4213@item -Wenum-compare 4214@opindex Wenum-compare 4215@opindex Wno-enum-compare 4216Warn about a comparison between values of different enumerated types. 4217In C++ enumeral mismatches in conditional expressions are also 4218diagnosed and the warning is enabled by default. In C this warning is 4219enabled by @option{-Wall}. 4220 4221@item -Wjump-misses-init @r{(C, Objective-C only)} 4222@opindex Wjump-misses-init 4223@opindex Wno-jump-misses-init 4224Warn if a @code{goto} statement or a @code{switch} statement jumps 4225forward across the initialization of a variable, or jumps backward to a 4226label after the variable has been initialized. This only warns about 4227variables that are initialized when they are declared. This warning is 4228only supported for C and Objective-C; in C++ this sort of branch is an 4229error in any case. 4230 4231@option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It 4232can be disabled with the @option{-Wno-jump-misses-init} option. 4233 4234@item -Wsign-compare 4235@opindex Wsign-compare 4236@opindex Wno-sign-compare 4237@cindex warning for comparison of signed and unsigned values 4238@cindex comparison of signed and unsigned values, warning 4239@cindex signed and unsigned values, comparison warning 4240Warn when a comparison between signed and unsigned values could produce 4241an incorrect result when the signed value is converted to unsigned. 4242This warning is also enabled by @option{-Wextra}; to get the other warnings 4243of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}. 4244 4245@item -Wsign-conversion 4246@opindex Wsign-conversion 4247@opindex Wno-sign-conversion 4248Warn for implicit conversions that may change the sign of an integer 4249value, like assigning a signed integer expression to an unsigned 4250integer variable. An explicit cast silences the warning. In C, this 4251option is enabled also by @option{-Wconversion}. 4252 4253@item -Waddress 4254@opindex Waddress 4255@opindex Wno-address 4256Warn about suspicious uses of memory addresses. These include using 4257the address of a function in a conditional expression, such as 4258@code{void func(void); if (func)}, and comparisons against the memory 4259address of a string literal, such as @code{if (x == "abc")}. Such 4260uses typically indicate a programmer error: the address of a function 4261always evaluates to true, so their use in a conditional usually 4262indicate that the programmer forgot the parentheses in a function 4263call; and comparisons against string literals result in unspecified 4264behavior and are not portable in C, so they usually indicate that the 4265programmer intended to use @code{strcmp}. This warning is enabled by 4266@option{-Wall}. 4267 4268@item -Wlogical-op 4269@opindex Wlogical-op 4270@opindex Wno-logical-op 4271Warn about suspicious uses of logical operators in expressions. 4272This includes using logical operators in contexts where a 4273bit-wise operator is likely to be expected. 4274 4275@item -Waggregate-return 4276@opindex Waggregate-return 4277@opindex Wno-aggregate-return 4278Warn if any functions that return structures or unions are defined or 4279called. (In languages where you can return an array, this also elicits 4280a warning.) 4281 4282@item -Wno-attributes 4283@opindex Wno-attributes 4284@opindex Wattributes 4285Do not warn if an unexpected @code{__attribute__} is used, such as 4286unrecognized attributes, function attributes applied to variables, 4287etc. This will not stop errors for incorrect use of supported 4288attributes. 4289 4290@item -Wno-builtin-macro-redefined 4291@opindex Wno-builtin-macro-redefined 4292@opindex Wbuiltin-macro-redefined 4293Do not warn if certain built-in macros are redefined. This suppresses 4294warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__}, 4295@code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}. 4296 4297@item -Wstrict-prototypes @r{(C and Objective-C only)} 4298@opindex Wstrict-prototypes 4299@opindex Wno-strict-prototypes 4300Warn if a function is declared or defined without specifying the 4301argument types. (An old-style function definition is permitted without 4302a warning if preceded by a declaration that specifies the argument 4303types.) 4304 4305@item -Wold-style-declaration @r{(C and Objective-C only)} 4306@opindex Wold-style-declaration 4307@opindex Wno-old-style-declaration 4308Warn for obsolescent usages, according to the C Standard, in a 4309declaration. For example, warn if storage-class specifiers like 4310@code{static} are not the first things in a declaration. This warning 4311is also enabled by @option{-Wextra}. 4312 4313@item -Wold-style-definition @r{(C and Objective-C only)} 4314@opindex Wold-style-definition 4315@opindex Wno-old-style-definition 4316Warn if an old-style function definition is used. A warning is given 4317even if there is a previous prototype. 4318 4319@item -Wmissing-parameter-type @r{(C and Objective-C only)} 4320@opindex Wmissing-parameter-type 4321@opindex Wno-missing-parameter-type 4322A function parameter is declared without a type specifier in K&R-style 4323functions: 4324 4325@smallexample 4326void foo(bar) @{ @} 4327@end smallexample 4328 4329This warning is also enabled by @option{-Wextra}. 4330 4331@item -Wmissing-prototypes @r{(C and Objective-C only)} 4332@opindex Wmissing-prototypes 4333@opindex Wno-missing-prototypes 4334Warn if a global function is defined without a previous prototype 4335declaration. This warning is issued even if the definition itself 4336provides a prototype. The aim is to detect global functions that 4337are not declared in header files. 4338 4339@item -Wmissing-declarations 4340@opindex Wmissing-declarations 4341@opindex Wno-missing-declarations 4342Warn if a global function is defined without a previous declaration. 4343Do so even if the definition itself provides a prototype. 4344Use this option to detect global functions that are not declared in 4345header files. In C++, no warnings are issued for function templates, 4346or for inline functions, or for functions in anonymous namespaces. 4347 4348@item -Wmissing-field-initializers 4349@opindex Wmissing-field-initializers 4350@opindex Wno-missing-field-initializers 4351@opindex W 4352@opindex Wextra 4353@opindex Wno-extra 4354Warn if a structure's initializer has some fields missing. For 4355example, the following code would cause such a warning, because 4356@code{x.h} is implicitly zero: 4357 4358@smallexample 4359struct s @{ int f, g, h; @}; 4360struct s x = @{ 3, 4 @}; 4361@end smallexample 4362 4363This option does not warn about designated initializers, so the following 4364modification would not trigger a warning: 4365 4366@smallexample 4367struct s @{ int f, g, h; @}; 4368struct s x = @{ .f = 3, .g = 4 @}; 4369@end smallexample 4370 4371This warning is included in @option{-Wextra}. To get other @option{-Wextra} 4372warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}. 4373 4374@item -Wmissing-format-attribute 4375@opindex Wmissing-format-attribute 4376@opindex Wno-missing-format-attribute 4377@opindex Wformat 4378@opindex Wno-format 4379Warn about function pointers that might be candidates for @code{format} 4380attributes. Note these are only possible candidates, not absolute ones. 4381GCC will guess that function pointers with @code{format} attributes that 4382are used in assignment, initialization, parameter passing or return 4383statements should have a corresponding @code{format} attribute in the 4384resulting type. I.e.@: the left-hand side of the assignment or 4385initialization, the type of the parameter variable, or the return type 4386of the containing function respectively should also have a @code{format} 4387attribute to avoid the warning. 4388 4389GCC will also warn about function definitions that might be 4390candidates for @code{format} attributes. Again, these are only 4391possible candidates. GCC will guess that @code{format} attributes 4392might be appropriate for any function that calls a function like 4393@code{vprintf} or @code{vscanf}, but this might not always be the 4394case, and some functions for which @code{format} attributes are 4395appropriate may not be detected. 4396 4397@item -Wno-multichar 4398@opindex Wno-multichar 4399@opindex Wmultichar 4400Do not warn if a multicharacter constant (@samp{'FOOF'}) is used. 4401Usually they indicate a typo in the user's code, as they have 4402implementation-defined values, and should not be used in portable code. 4403 4404@item -Wnormalized=<none|id|nfc|nfkc> 4405@opindex Wnormalized= 4406@cindex NFC 4407@cindex NFKC 4408@cindex character set, input normalization 4409In ISO C and ISO C++, two identifiers are different if they are 4410different sequences of characters. However, sometimes when characters 4411outside the basic ASCII character set are used, you can have two 4412different character sequences that look the same. To avoid confusion, 4413the ISO 10646 standard sets out some @dfn{normalization rules} which 4414when applied ensure that two sequences that look the same are turned into 4415the same sequence. GCC can warn you if you are using identifiers that 4416have not been normalized; this option controls that warning. 4417 4418There are four levels of warning supported by GCC. The default is 4419@option{-Wnormalized=nfc}, which warns about any identifier that is 4420not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the 4421recommended form for most uses. 4422 4423Unfortunately, there are some characters allowed in identifiers by 4424ISO C and ISO C++ that, when turned into NFC, are not allowed in 4425identifiers. That is, there's no way to use these symbols in portable 4426ISO C or C++ and have all your identifiers in NFC@. 4427@option{-Wnormalized=id} suppresses the warning for these characters. 4428It is hoped that future versions of the standards involved will correct 4429this, which is why this option is not the default. 4430 4431You can switch the warning off for all characters by writing 4432@option{-Wnormalized=none}. You would only want to do this if you 4433were using some other normalization scheme (like ``D''), because 4434otherwise you can easily create bugs that are literally impossible to see. 4435 4436Some characters in ISO 10646 have distinct meanings but look identical 4437in some fonts or display methodologies, especially once formatting has 4438been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL 4439LETTER N'', will display just like a regular @code{n} that has been 4440placed in a superscript. ISO 10646 defines the @dfn{NFKC} 4441normalization scheme to convert all these into a standard form as 4442well, and GCC will warn if your code is not in NFKC if you use 4443@option{-Wnormalized=nfkc}. This warning is comparable to warning 4444about every identifier that contains the letter O because it might be 4445confused with the digit 0, and so is not the default, but may be 4446useful as a local coding convention if the programming environment is 4447unable to be fixed to display these characters distinctly. 4448 4449@item -Wno-deprecated 4450@opindex Wno-deprecated 4451@opindex Wdeprecated 4452Do not warn about usage of deprecated features. @xref{Deprecated Features}. 4453 4454@item -Wno-deprecated-declarations 4455@opindex Wno-deprecated-declarations 4456@opindex Wdeprecated-declarations 4457Do not warn about uses of functions (@pxref{Function Attributes}), 4458variables (@pxref{Variable Attributes}), and types (@pxref{Type 4459Attributes}) marked as deprecated by using the @code{deprecated} 4460attribute. 4461 4462@item -Wno-overflow 4463@opindex Wno-overflow 4464@opindex Woverflow 4465Do not warn about compile-time overflow in constant expressions. 4466 4467@item -Woverride-init @r{(C and Objective-C only)} 4468@opindex Woverride-init 4469@opindex Wno-override-init 4470@opindex W 4471@opindex Wextra 4472@opindex Wno-extra 4473Warn if an initialized field without side effects is overridden when 4474using designated initializers (@pxref{Designated Inits, , Designated 4475Initializers}). 4476 4477This warning is included in @option{-Wextra}. To get other 4478@option{-Wextra} warnings without this one, use @samp{-Wextra 4479-Wno-override-init}. 4480 4481@item -Wpacked 4482@opindex Wpacked 4483@opindex Wno-packed 4484Warn if a structure is given the packed attribute, but the packed 4485attribute has no effect on the layout or size of the structure. 4486Such structures may be mis-aligned for little benefit. For 4487instance, in this code, the variable @code{f.x} in @code{struct bar} 4488will be misaligned even though @code{struct bar} does not itself 4489have the packed attribute: 4490 4491@smallexample 4492@group 4493struct foo @{ 4494 int x; 4495 char a, b, c, d; 4496@} __attribute__((packed)); 4497struct bar @{ 4498 char z; 4499 struct foo f; 4500@}; 4501@end group 4502@end smallexample 4503 4504@item -Wpacked-bitfield-compat 4505@opindex Wpacked-bitfield-compat 4506@opindex Wno-packed-bitfield-compat 4507The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute 4508on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but 4509the change can lead to differences in the structure layout. GCC 4510informs you when the offset of such a field has changed in GCC 4.4. 4511For example there is no longer a 4-bit padding between field @code{a} 4512and @code{b} in this structure: 4513 4514@smallexample 4515struct foo 4516@{ 4517 char a:4; 4518 char b:8; 4519@} __attribute__ ((packed)); 4520@end smallexample 4521 4522This warning is enabled by default. Use 4523@option{-Wno-packed-bitfield-compat} to disable this warning. 4524 4525@item -Wpadded 4526@opindex Wpadded 4527@opindex Wno-padded 4528Warn if padding is included in a structure, either to align an element 4529of the structure or to align the whole structure. Sometimes when this 4530happens it is possible to rearrange the fields of the structure to 4531reduce the padding and so make the structure smaller. 4532 4533@item -Wredundant-decls 4534@opindex Wredundant-decls 4535@opindex Wno-redundant-decls 4536Warn if anything is declared more than once in the same scope, even in 4537cases where multiple declaration is valid and changes nothing. 4538 4539@item -Wnested-externs @r{(C and Objective-C only)} 4540@opindex Wnested-externs 4541@opindex Wno-nested-externs 4542Warn if an @code{extern} declaration is encountered within a function. 4543 4544@item -Winline 4545@opindex Winline 4546@opindex Wno-inline 4547Warn if a function can not be inlined and it was declared as inline. 4548Even with this option, the compiler will not warn about failures to 4549inline functions declared in system headers. 4550 4551The compiler uses a variety of heuristics to determine whether or not 4552to inline a function. For example, the compiler takes into account 4553the size of the function being inlined and the amount of inlining 4554that has already been done in the current function. Therefore, 4555seemingly insignificant changes in the source program can cause the 4556warnings produced by @option{-Winline} to appear or disappear. 4557 4558@item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)} 4559@opindex Wno-invalid-offsetof 4560@opindex Winvalid-offsetof 4561Suppress warnings from applying the @samp{offsetof} macro to a non-POD 4562type. According to the 1998 ISO C++ standard, applying @samp{offsetof} 4563to a non-POD type is undefined. In existing C++ implementations, 4564however, @samp{offsetof} typically gives meaningful results even when 4565applied to certain kinds of non-POD types. (Such as a simple 4566@samp{struct} that fails to be a POD type only by virtue of having a 4567constructor.) This flag is for users who are aware that they are 4568writing nonportable code and who have deliberately chosen to ignore the 4569warning about it. 4570 4571The restrictions on @samp{offsetof} may be relaxed in a future version 4572of the C++ standard. 4573 4574@item -Wno-int-to-pointer-cast 4575@opindex Wno-int-to-pointer-cast 4576@opindex Wint-to-pointer-cast 4577Suppress warnings from casts to pointer type of an integer of a 4578different size. In C++, casting to a pointer type of smaller size is 4579an error. @option{Wint-to-pointer-cast} is enabled by default. 4580 4581 4582@item -Wno-pointer-to-int-cast @r{(C and Objective-C only)} 4583@opindex Wno-pointer-to-int-cast 4584@opindex Wpointer-to-int-cast 4585Suppress warnings from casts from a pointer to an integer type of a 4586different size. 4587 4588@item -Winvalid-pch 4589@opindex Winvalid-pch 4590@opindex Wno-invalid-pch 4591Warn if a precompiled header (@pxref{Precompiled Headers}) is found in 4592the search path but can't be used. 4593 4594@item -Wlong-long 4595@opindex Wlong-long 4596@opindex Wno-long-long 4597Warn if @samp{long long} type is used. This is enabled by either 4598@option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98 4599modes. To inhibit the warning messages, use @option{-Wno-long-long}. 4600 4601@item -Wvariadic-macros 4602@opindex Wvariadic-macros 4603@opindex Wno-variadic-macros 4604Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU 4605alternate syntax when in pedantic ISO C99 mode. This is default. 4606To inhibit the warning messages, use @option{-Wno-variadic-macros}. 4607 4608@item -Wvector-operation-performance 4609@opindex Wvector-operation-performance 4610@opindex Wno-vector-operation-performance 4611Warn if vector operation is not implemented via SIMD capabilities of the 4612architecture. Mainly useful for the performance tuning. 4613Vector operation can be implemented @code{piecewise}, which means that the 4614scalar operation is performed on every vector element; 4615@code{in parallel}, which means that the vector operation is implemented 4616using scalars of wider type, which normally is more performance efficient; 4617and @code{as a single scalar}, which means that vector fits into a 4618scalar type. 4619 4620@item -Wvla 4621@opindex Wvla 4622@opindex Wno-vla 4623Warn if variable length array is used in the code. 4624@option{-Wno-vla} will prevent the @option{-pedantic} warning of 4625the variable length array. 4626 4627@item -Wvolatile-register-var 4628@opindex Wvolatile-register-var 4629@opindex Wno-volatile-register-var 4630Warn if a register variable is declared volatile. The volatile 4631modifier does not inhibit all optimizations that may eliminate reads 4632and/or writes to register variables. This warning is enabled by 4633@option{-Wall}. 4634 4635@item -Wdisabled-optimization 4636@opindex Wdisabled-optimization 4637@opindex Wno-disabled-optimization 4638Warn if a requested optimization pass is disabled. This warning does 4639not generally indicate that there is anything wrong with your code; it 4640merely indicates that GCC's optimizers were unable to handle the code 4641effectively. Often, the problem is that your code is too big or too 4642complex; GCC will refuse to optimize programs when the optimization 4643itself is likely to take inordinate amounts of time. 4644 4645@item -Wpointer-sign @r{(C and Objective-C only)} 4646@opindex Wpointer-sign 4647@opindex Wno-pointer-sign 4648Warn for pointer argument passing or assignment with different signedness. 4649This option is only supported for C and Objective-C@. It is implied by 4650@option{-Wall} and by @option{-pedantic}, which can be disabled with 4651@option{-Wno-pointer-sign}. 4652 4653@item -Wstack-protector 4654@opindex Wstack-protector 4655@opindex Wno-stack-protector 4656This option is only active when @option{-fstack-protector} is active. It 4657warns about functions that will not be protected against stack smashing. 4658 4659@item -Wno-mudflap 4660@opindex Wno-mudflap 4661Suppress warnings about constructs that cannot be instrumented by 4662@option{-fmudflap}. 4663 4664@item -Woverlength-strings 4665@opindex Woverlength-strings 4666@opindex Wno-overlength-strings 4667Warn about string constants that are longer than the ``minimum 4668maximum'' length specified in the C standard. Modern compilers 4669generally allow string constants that are much longer than the 4670standard's minimum limit, but very portable programs should avoid 4671using longer strings. 4672 4673The limit applies @emph{after} string constant concatenation, and does 4674not count the trailing NUL@. In C90, the limit was 509 characters; in 4675C99, it was raised to 4095. C++98 does not specify a normative 4676minimum maximum, so we do not diagnose overlength strings in C++@. 4677 4678This option is implied by @option{-pedantic}, and can be disabled with 4679@option{-Wno-overlength-strings}. 4680 4681@item -Wunsuffixed-float-constants @r{(C and Objective-C only)} 4682@opindex Wunsuffixed-float-constants 4683 4684GCC will issue a warning for any floating constant that does not have 4685a suffix. When used together with @option{-Wsystem-headers} it will 4686warn about such constants in system header files. This can be useful 4687when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma 4688from the decimal floating-point extension to C99. 4689@end table 4690 4691@node Debugging Options 4692@section Options for Debugging Your Program or GCC 4693@cindex options, debugging 4694@cindex debugging information options 4695 4696GCC has various special options that are used for debugging 4697either your program or GCC: 4698 4699@table @gcctabopt 4700@item -g 4701@opindex g 4702Produce debugging information in the operating system's native format 4703(stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging 4704information. 4705 4706On most systems that use stabs format, @option{-g} enables use of extra 4707debugging information that only GDB can use; this extra information 4708makes debugging work better in GDB but will probably make other debuggers 4709crash or 4710refuse to read the program. If you want to control for certain whether 4711to generate the extra information, use @option{-gstabs+}, @option{-gstabs}, 4712@option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below). 4713 4714GCC allows you to use @option{-g} with 4715@option{-O}. The shortcuts taken by optimized code may occasionally 4716produce surprising results: some variables you declared may not exist 4717at all; flow of control may briefly move where you did not expect it; 4718some statements may not be executed because they compute constant 4719results or their values were already at hand; some statements may 4720execute in different places because they were moved out of loops. 4721 4722Nevertheless it proves possible to debug optimized output. This makes 4723it reasonable to use the optimizer for programs that might have bugs. 4724 4725The following options are useful when GCC is generated with the 4726capability for more than one debugging format. 4727 4728@item -ggdb 4729@opindex ggdb 4730Produce debugging information for use by GDB@. This means to use the 4731most expressive format available (DWARF 2, stabs, or the native format 4732if neither of those are supported), including GDB extensions if at all 4733possible. 4734 4735@item -gstabs 4736@opindex gstabs 4737Produce debugging information in stabs format (if that is supported), 4738without GDB extensions. This is the format used by DBX on most BSD 4739systems. On MIPS, Alpha and System V Release 4 systems this option 4740produces stabs debugging output that is not understood by DBX or SDB@. 4741On System V Release 4 systems this option requires the GNU assembler. 4742 4743@item -feliminate-unused-debug-symbols 4744@opindex feliminate-unused-debug-symbols 4745Produce debugging information in stabs format (if that is supported), 4746for only symbols that are actually used. 4747 4748@item -femit-class-debug-always 4749Instead of emitting debugging information for a C++ class in only one 4750object file, emit it in all object files using the class. This option 4751should be used only with debuggers that are unable to handle the way GCC 4752normally emits debugging information for classes because using this 4753option will increase the size of debugging information by as much as a 4754factor of two. 4755 4756@item -fno-debug-types-section 4757@opindex fno-debug-types-section 4758@opindex fdebug-types-section 4759By default when using DWARF v4 or higher type DIEs will be put into 4760their own .debug_types section instead of making them part of the 4761.debug_info section. It is more efficient to put them in a separate 4762comdat sections since the linker will then be able to remove duplicates. 4763But not all DWARF consumers support .debug_types sections yet. 4764 4765@item -gstabs+ 4766@opindex gstabs+ 4767Produce debugging information in stabs format (if that is supported), 4768using GNU extensions understood only by the GNU debugger (GDB)@. The 4769use of these extensions is likely to make other debuggers crash or 4770refuse to read the program. 4771 4772@item -gcoff 4773@opindex gcoff 4774Produce debugging information in COFF format (if that is supported). 4775This is the format used by SDB on most System V systems prior to 4776System V Release 4. 4777 4778@item -gxcoff 4779@opindex gxcoff 4780Produce debugging information in XCOFF format (if that is supported). 4781This is the format used by the DBX debugger on IBM RS/6000 systems. 4782 4783@item -gxcoff+ 4784@opindex gxcoff+ 4785Produce debugging information in XCOFF format (if that is supported), 4786using GNU extensions understood only by the GNU debugger (GDB)@. The 4787use of these extensions is likely to make other debuggers crash or 4788refuse to read the program, and may cause assemblers other than the GNU 4789assembler (GAS) to fail with an error. 4790 4791@item -gdwarf-@var{version} 4792@opindex gdwarf-@var{version} 4793Produce debugging information in DWARF format (if that is 4794supported). This is the format used by DBX on IRIX 6. The value 4795of @var{version} may be either 2, 3 or 4; the default version is 2. 4796 4797Note that with DWARF version 2 some ports require, and will always 4798use, some non-conflicting DWARF 3 extensions in the unwind tables. 4799 4800Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments} 4801for maximum benefit. 4802 4803@item -grecord-gcc-switches 4804@opindex grecord-gcc-switches 4805This switch causes the command-line options used to invoke the 4806compiler that may affect code generation to be appended to the 4807DW_AT_producer attribute in DWARF debugging information. The options 4808are concatenated with spaces separating them from each other and from 4809the compiler version. See also @option{-frecord-gcc-switches} for another 4810way of storing compiler options into the object file. 4811 4812@item -gno-record-gcc-switches 4813@opindex gno-record-gcc-switches 4814Disallow appending command-line options to the DW_AT_producer attribute 4815in DWARF debugging information. This is the default. 4816 4817@item -gstrict-dwarf 4818@opindex gstrict-dwarf 4819Disallow using extensions of later DWARF standard version than selected 4820with @option{-gdwarf-@var{version}}. On most targets using non-conflicting 4821DWARF extensions from later standard versions is allowed. 4822 4823@item -gno-strict-dwarf 4824@opindex gno-strict-dwarf 4825Allow using extensions of later DWARF standard version than selected with 4826@option{-gdwarf-@var{version}}. 4827 4828@item -gvms 4829@opindex gvms 4830Produce debugging information in VMS debug format (if that is 4831supported). This is the format used by DEBUG on VMS systems. 4832 4833@item -g@var{level} 4834@itemx -ggdb@var{level} 4835@itemx -gstabs@var{level} 4836@itemx -gcoff@var{level} 4837@itemx -gxcoff@var{level} 4838@itemx -gvms@var{level} 4839Request debugging information and also use @var{level} to specify how 4840much information. The default level is 2. 4841 4842Level 0 produces no debug information at all. Thus, @option{-g0} negates 4843@option{-g}. 4844 4845Level 1 produces minimal information, enough for making backtraces in 4846parts of the program that you don't plan to debug. This includes 4847descriptions of functions and external variables, but no information 4848about local variables and no line numbers. 4849 4850Level 3 includes extra information, such as all the macro definitions 4851present in the program. Some debuggers support macro expansion when 4852you use @option{-g3}. 4853 4854@option{-gdwarf-2} does not accept a concatenated debug level, because 4855GCC used to support an option @option{-gdwarf} that meant to generate 4856debug information in version 1 of the DWARF format (which is very 4857different from version 2), and it would have been too confusing. That 4858debug format is long obsolete, but the option cannot be changed now. 4859Instead use an additional @option{-g@var{level}} option to change the 4860debug level for DWARF. 4861 4862@item -gtoggle 4863@opindex gtoggle 4864Turn off generation of debug info, if leaving out this option would have 4865generated it, or turn it on at level 2 otherwise. The position of this 4866argument in the command line does not matter, it takes effect after all 4867other options are processed, and it does so only once, no matter how 4868many times it is given. This is mainly intended to be used with 4869@option{-fcompare-debug}. 4870 4871@item -fdump-final-insns@r{[}=@var{file}@r{]} 4872@opindex fdump-final-insns 4873Dump the final internal representation (RTL) to @var{file}. If the 4874optional argument is omitted (or if @var{file} is @code{.}), the name 4875of the dump file will be determined by appending @code{.gkd} to the 4876compilation output file name. 4877 4878@item -fcompare-debug@r{[}=@var{opts}@r{]} 4879@opindex fcompare-debug 4880@opindex fno-compare-debug 4881If no error occurs during compilation, run the compiler a second time, 4882adding @var{opts} and @option{-fcompare-debug-second} to the arguments 4883passed to the second compilation. Dump the final internal 4884representation in both compilations, and print an error if they differ. 4885 4886If the equal sign is omitted, the default @option{-gtoggle} is used. 4887 4888The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty 4889and nonzero, implicitly enables @option{-fcompare-debug}. If 4890@env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash, 4891then it is used for @var{opts}, otherwise the default @option{-gtoggle} 4892is used. 4893 4894@option{-fcompare-debug=}, with the equal sign but without @var{opts}, 4895is equivalent to @option{-fno-compare-debug}, which disables the dumping 4896of the final representation and the second compilation, preventing even 4897@env{GCC_COMPARE_DEBUG} from taking effect. 4898 4899To verify full coverage during @option{-fcompare-debug} testing, set 4900@env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden}, 4901which GCC will reject as an invalid option in any actual compilation 4902(rather than preprocessing, assembly or linking). To get just a 4903warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug 4904not overridden} will do. 4905 4906@item -fcompare-debug-second 4907@opindex fcompare-debug-second 4908This option is implicitly passed to the compiler for the second 4909compilation requested by @option{-fcompare-debug}, along with options to 4910silence warnings, and omitting other options that would cause 4911side-effect compiler outputs to files or to the standard output. Dump 4912files and preserved temporary files are renamed so as to contain the 4913@code{.gk} additional extension during the second compilation, to avoid 4914overwriting those generated by the first. 4915 4916When this option is passed to the compiler driver, it causes the 4917@emph{first} compilation to be skipped, which makes it useful for little 4918other than debugging the compiler proper. 4919 4920@item -feliminate-dwarf2-dups 4921@opindex feliminate-dwarf2-dups 4922Compress DWARF2 debugging information by eliminating duplicated 4923information about each symbol. This option only makes sense when 4924generating DWARF2 debugging information with @option{-gdwarf-2}. 4925 4926@item -femit-struct-debug-baseonly 4927Emit debug information for struct-like types 4928only when the base name of the compilation source file 4929matches the base name of file in which the struct was defined. 4930 4931This option substantially reduces the size of debugging information, 4932but at significant potential loss in type information to the debugger. 4933See @option{-femit-struct-debug-reduced} for a less aggressive option. 4934See @option{-femit-struct-debug-detailed} for more detailed control. 4935 4936This option works only with DWARF 2. 4937 4938@item -femit-struct-debug-reduced 4939Emit debug information for struct-like types 4940only when the base name of the compilation source file 4941matches the base name of file in which the type was defined, 4942unless the struct is a template or defined in a system header. 4943 4944This option significantly reduces the size of debugging information, 4945with some potential loss in type information to the debugger. 4946See @option{-femit-struct-debug-baseonly} for a more aggressive option. 4947See @option{-femit-struct-debug-detailed} for more detailed control. 4948 4949This option works only with DWARF 2. 4950 4951@item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} 4952Specify the struct-like types 4953for which the compiler will generate debug information. 4954The intent is to reduce duplicate struct debug information 4955between different object files within the same program. 4956 4957This option is a detailed version of 4958@option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly}, 4959which will serve for most needs. 4960 4961A specification has the syntax@* 4962[@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none}) 4963 4964The optional first word limits the specification to 4965structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}). 4966A struct type is used directly when it is the type of a variable, member. 4967Indirect uses arise through pointers to structs. 4968That is, when use of an incomplete struct would be legal, the use is indirect. 4969An example is 4970@samp{struct one direct; struct two * indirect;}. 4971 4972The optional second word limits the specification to 4973ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}). 4974Generic structs are a bit complicated to explain. 4975For C++, these are non-explicit specializations of template classes, 4976or non-template classes within the above. 4977Other programming languages have generics, 4978but @samp{-femit-struct-debug-detailed} does not yet implement them. 4979 4980The third word specifies the source files for those 4981structs for which the compiler will emit debug information. 4982The values @samp{none} and @samp{any} have the normal meaning. 4983The value @samp{base} means that 4984the base of name of the file in which the type declaration appears 4985must match the base of the name of the main compilation file. 4986In practice, this means that 4987types declared in @file{foo.c} and @file{foo.h} will have debug information, 4988but types declared in other header will not. 4989The value @samp{sys} means those types satisfying @samp{base} 4990or declared in system or compiler headers. 4991 4992You may need to experiment to determine the best settings for your application. 4993 4994The default is @samp{-femit-struct-debug-detailed=all}. 4995 4996This option works only with DWARF 2. 4997 4998@item -fno-merge-debug-strings 4999@opindex fmerge-debug-strings 5000@opindex fno-merge-debug-strings 5001Direct the linker to not merge together strings in the debugging 5002information that are identical in different object files. Merging is 5003not supported by all assemblers or linkers. Merging decreases the size 5004of the debug information in the output file at the cost of increasing 5005link processing time. Merging is enabled by default. 5006 5007@item -fdebug-prefix-map=@var{old}=@var{new} 5008@opindex fdebug-prefix-map 5009When compiling files in directory @file{@var{old}}, record debugging 5010information describing them as in @file{@var{new}} instead. 5011 5012@item -fno-dwarf2-cfi-asm 5013@opindex fdwarf2-cfi-asm 5014@opindex fno-dwarf2-cfi-asm 5015Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section 5016instead of using GAS @code{.cfi_*} directives. 5017 5018@cindex @command{prof} 5019@item -p 5020@opindex p 5021Generate extra code to write profile information suitable for the 5022analysis program @command{prof}. You must use this option when compiling 5023the source files you want data about, and you must also use it when 5024linking. 5025 5026@cindex @command{gprof} 5027@item -pg 5028@opindex pg 5029Generate extra code to write profile information suitable for the 5030analysis program @command{gprof}. You must use this option when compiling 5031the source files you want data about, and you must also use it when 5032linking. 5033 5034@item -Q 5035@opindex Q 5036Makes the compiler print out each function name as it is compiled, and 5037print some statistics about each pass when it finishes. 5038 5039@item -ftime-report 5040@opindex ftime-report 5041Makes the compiler print some statistics about the time consumed by each 5042pass when it finishes. 5043 5044@item -fmem-report 5045@opindex fmem-report 5046Makes the compiler print some statistics about permanent memory 5047allocation when it finishes. 5048 5049@item -fpre-ipa-mem-report 5050@opindex fpre-ipa-mem-report 5051@item -fpost-ipa-mem-report 5052@opindex fpost-ipa-mem-report 5053Makes the compiler print some statistics about permanent memory 5054allocation before or after interprocedural optimization. 5055 5056@item -fstack-usage 5057@opindex fstack-usage 5058Makes the compiler output stack usage information for the program, on a 5059per-function basis. The filename for the dump is made by appending 5060@file{.su} to the @var{auxname}. @var{auxname} is generated from the name of 5061the output file, if explicitly specified and it is not an executable, 5062otherwise it is the basename of the source file. An entry is made up 5063of three fields: 5064 5065@itemize 5066@item 5067The name of the function. 5068@item 5069A number of bytes. 5070@item 5071One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}. 5072@end itemize 5073 5074The qualifier @code{static} means that the function manipulates the stack 5075statically: a fixed number of bytes are allocated for the frame on function 5076entry and released on function exit; no stack adjustments are otherwise made 5077in the function. The second field is this fixed number of bytes. 5078 5079The qualifier @code{dynamic} means that the function manipulates the stack 5080dynamically: in addition to the static allocation described above, stack 5081adjustments are made in the body of the function, for example to push/pop 5082arguments around function calls. If the qualifier @code{bounded} is also 5083present, the amount of these adjustments is bounded at compile time and 5084the second field is an upper bound of the total amount of stack used by 5085the function. If it is not present, the amount of these adjustments is 5086not bounded at compile time and the second field only represents the 5087bounded part. 5088 5089@item -fprofile-arcs 5090@opindex fprofile-arcs 5091Add code so that program flow @dfn{arcs} are instrumented. During 5092execution the program records how many times each branch and call is 5093executed and how many times it is taken or returns. When the compiled 5094program exits it saves this data to a file called 5095@file{@var{auxname}.gcda} for each source file. The data may be used for 5096profile-directed optimizations (@option{-fbranch-probabilities}), or for 5097test coverage analysis (@option{-ftest-coverage}). Each object file's 5098@var{auxname} is generated from the name of the output file, if 5099explicitly specified and it is not the final executable, otherwise it is 5100the basename of the source file. In both cases any suffix is removed 5101(e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or 5102@file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}). 5103@xref{Cross-profiling}. 5104 5105@cindex @command{gcov} 5106@item --coverage 5107@opindex coverage 5108 5109This option is used to compile and link code instrumented for coverage 5110analysis. The option is a synonym for @option{-fprofile-arcs} 5111@option{-ftest-coverage} (when compiling) and @option{-lgcov} (when 5112linking). See the documentation for those options for more details. 5113 5114@itemize 5115 5116@item 5117Compile the source files with @option{-fprofile-arcs} plus optimization 5118and code generation options. For test coverage analysis, use the 5119additional @option{-ftest-coverage} option. You do not need to profile 5120every source file in a program. 5121 5122@item 5123Link your object files with @option{-lgcov} or @option{-fprofile-arcs} 5124(the latter implies the former). 5125 5126@item 5127Run the program on a representative workload to generate the arc profile 5128information. This may be repeated any number of times. You can run 5129concurrent instances of your program, and provided that the file system 5130supports locking, the data files will be correctly updated. Also 5131@code{fork} calls are detected and correctly handled (double counting 5132will not happen). 5133 5134@item 5135For profile-directed optimizations, compile the source files again with 5136the same optimization and code generation options plus 5137@option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that 5138Control Optimization}). 5139 5140@item 5141For test coverage analysis, use @command{gcov} to produce human readable 5142information from the @file{.gcno} and @file{.gcda} files. Refer to the 5143@command{gcov} documentation for further information. 5144 5145@end itemize 5146 5147With @option{-fprofile-arcs}, for each function of your program GCC 5148creates a program flow graph, then finds a spanning tree for the graph. 5149Only arcs that are not on the spanning tree have to be instrumented: the 5150compiler adds code to count the number of times that these arcs are 5151executed. When an arc is the only exit or only entrance to a block, the 5152instrumentation code can be added to the block; otherwise, a new basic 5153block must be created to hold the instrumentation code. 5154 5155@need 2000 5156@item -ftest-coverage 5157@opindex ftest-coverage 5158Produce a notes file that the @command{gcov} code-coverage utility 5159(@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to 5160show program coverage. Each source file's note file is called 5161@file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option 5162above for a description of @var{auxname} and instructions on how to 5163generate test coverage data. Coverage data will match the source files 5164more closely, if you do not optimize. 5165 5166@item -fdbg-cnt-list 5167@opindex fdbg-cnt-list 5168Print the name and the counter upper bound for all debug counters. 5169 5170 5171@item -fdbg-cnt=@var{counter-value-list} 5172@opindex fdbg-cnt 5173Set the internal debug counter upper bound. @var{counter-value-list} 5174is a comma-separated list of @var{name}:@var{value} pairs 5175which sets the upper bound of each debug counter @var{name} to @var{value}. 5176All debug counters have the initial upper bound of @var{UINT_MAX}, 5177thus dbg_cnt() returns true always unless the upper bound is set by this option. 5178e.g. With -fdbg-cnt=dce:10,tail_call:0 5179dbg_cnt(dce) will return true only for first 10 invocations 5180 5181@item -fenable-@var{kind}-@var{pass} 5182@itemx -fdisable-@var{kind}-@var{pass}=@var{range-list} 5183@opindex fdisable- 5184@opindex fenable- 5185 5186This is a set of debugging options that are used to explicitly disable/enable 5187optimization passes. For compiler users, regular options for enabling/disabling 5188passes should be used instead. 5189 5190@itemize 5191 5192@item -fdisable-ipa-@var{pass} 5193Disable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is 5194statically invoked in the compiler multiple times, the pass name should be 5195appended with a sequential number starting from 1. 5196 5197@item -fdisable-rtl-@var{pass} 5198@item -fdisable-rtl-@var{pass}=@var{range-list} 5199Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is 5200statically invoked in the compiler multiple times, the pass name should be 5201appended with a sequential number starting from 1. @var{range-list} is a comma 5202seperated list of function ranges or assembler names. Each range is a number 5203pair seperated by a colon. The range is inclusive in both ends. If the range 5204is trivial, the number pair can be simplified as a single number. If the 5205function's cgraph node's @var{uid} is falling within one of the specified ranges, 5206the @var{pass} is disabled for that function. The @var{uid} is shown in the 5207function header of a dump file, and the pass names can be dumped by using 5208option @option{-fdump-passes}. 5209 5210@item -fdisable-tree-@var{pass} 5211@item -fdisable-tree-@var{pass}=@var{range-list} 5212Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of 5213option arguments. 5214 5215@item -fenable-ipa-@var{pass} 5216Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is 5217statically invoked in the compiler multiple times, the pass name should be 5218appended with a sequential number starting from 1. 5219 5220@item -fenable-rtl-@var{pass} 5221@item -fenable-rtl-@var{pass}=@var{range-list} 5222Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument 5223description and examples. 5224 5225@item -fenable-tree-@var{pass} 5226@item -fenable-tree-@var{pass}=@var{range-list} 5227Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description 5228of option arguments. 5229 5230@smallexample 5231 5232# disable ccp1 for all functions 5233 -fdisable-tree-ccp1 5234# disable complete unroll for function whose cgraph node uid is 1 5235 -fenable-tree-cunroll=1 5236# disable gcse2 for functions at the following ranges [1,1], 5237# [300,400], and [400,1000] 5238# disable gcse2 for functions foo and foo2 5239 -fdisable-rtl-gcse2=foo,foo2 5240# disable early inlining 5241 -fdisable-tree-einline 5242# disable ipa inlining 5243 -fdisable-ipa-inline 5244# enable tree full unroll 5245 -fenable-tree-unroll 5246 5247@end smallexample 5248 5249@end itemize 5250 5251@item -d@var{letters} 5252@itemx -fdump-rtl-@var{pass} 5253@opindex d 5254Says to make debugging dumps during compilation at times specified by 5255@var{letters}. This is used for debugging the RTL-based passes of the 5256compiler. The file names for most of the dumps are made by appending 5257a pass number and a word to the @var{dumpname}, and the files are 5258created in the directory of the output file. Note that the pass 5259number is computed statically as passes get registered into the pass 5260manager. Thus the numbering is not related to the dynamic order of 5261execution of passes. In particular, a pass installed by a plugin 5262could have a number over 200 even if it executed quite early. 5263@var{dumpname} is generated from the name of the output file, if 5264explicitly specified and it is not an executable, otherwise it is the 5265basename of the source file. These switches may have different effects 5266when @option{-E} is used for preprocessing. 5267 5268Debug dumps can be enabled with a @option{-fdump-rtl} switch or some 5269@option{-d} option @var{letters}. Here are the possible 5270letters for use in @var{pass} and @var{letters}, and their meanings: 5271 5272@table @gcctabopt 5273 5274@item -fdump-rtl-alignments 5275@opindex fdump-rtl-alignments 5276Dump after branch alignments have been computed. 5277 5278@item -fdump-rtl-asmcons 5279@opindex fdump-rtl-asmcons 5280Dump after fixing rtl statements that have unsatisfied in/out constraints. 5281 5282@item -fdump-rtl-auto_inc_dec 5283@opindex fdump-rtl-auto_inc_dec 5284Dump after auto-inc-dec discovery. This pass is only run on 5285architectures that have auto inc or auto dec instructions. 5286 5287@item -fdump-rtl-barriers 5288@opindex fdump-rtl-barriers 5289Dump after cleaning up the barrier instructions. 5290 5291@item -fdump-rtl-bbpart 5292@opindex fdump-rtl-bbpart 5293Dump after partitioning hot and cold basic blocks. 5294 5295@item -fdump-rtl-bbro 5296@opindex fdump-rtl-bbro 5297Dump after block reordering. 5298 5299@item -fdump-rtl-btl1 5300@itemx -fdump-rtl-btl2 5301@opindex fdump-rtl-btl2 5302@opindex fdump-rtl-btl2 5303@option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping 5304after the two branch 5305target load optimization passes. 5306 5307@item -fdump-rtl-bypass 5308@opindex fdump-rtl-bypass 5309Dump after jump bypassing and control flow optimizations. 5310 5311@item -fdump-rtl-combine 5312@opindex fdump-rtl-combine 5313Dump after the RTL instruction combination pass. 5314 5315@item -fdump-rtl-compgotos 5316@opindex fdump-rtl-compgotos 5317Dump after duplicating the computed gotos. 5318 5319@item -fdump-rtl-ce1 5320@itemx -fdump-rtl-ce2 5321@itemx -fdump-rtl-ce3 5322@opindex fdump-rtl-ce1 5323@opindex fdump-rtl-ce2 5324@opindex fdump-rtl-ce3 5325@option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and 5326@option{-fdump-rtl-ce3} enable dumping after the three 5327if conversion passes. 5328 5329@item -fdump-rtl-cprop_hardreg 5330@opindex fdump-rtl-cprop_hardreg 5331Dump after hard register copy propagation. 5332 5333@item -fdump-rtl-csa 5334@opindex fdump-rtl-csa 5335Dump after combining stack adjustments. 5336 5337@item -fdump-rtl-cse1 5338@itemx -fdump-rtl-cse2 5339@opindex fdump-rtl-cse1 5340@opindex fdump-rtl-cse2 5341@option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after 5342the two common sub-expression elimination passes. 5343 5344@item -fdump-rtl-dce 5345@opindex fdump-rtl-dce 5346Dump after the standalone dead code elimination passes. 5347 5348@item -fdump-rtl-dbr 5349@opindex fdump-rtl-dbr 5350Dump after delayed branch scheduling. 5351 5352@item -fdump-rtl-dce1 5353@itemx -fdump-rtl-dce2 5354@opindex fdump-rtl-dce1 5355@opindex fdump-rtl-dce2 5356@option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after 5357the two dead store elimination passes. 5358 5359@item -fdump-rtl-eh 5360@opindex fdump-rtl-eh 5361Dump after finalization of EH handling code. 5362 5363@item -fdump-rtl-eh_ranges 5364@opindex fdump-rtl-eh_ranges 5365Dump after conversion of EH handling range regions. 5366 5367@item -fdump-rtl-expand 5368@opindex fdump-rtl-expand 5369Dump after RTL generation. 5370 5371@item -fdump-rtl-fwprop1 5372@itemx -fdump-rtl-fwprop2 5373@opindex fdump-rtl-fwprop1 5374@opindex fdump-rtl-fwprop2 5375@option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable 5376dumping after the two forward propagation passes. 5377 5378@item -fdump-rtl-gcse1 5379@itemx -fdump-rtl-gcse2 5380@opindex fdump-rtl-gcse1 5381@opindex fdump-rtl-gcse2 5382@option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping 5383after global common subexpression elimination. 5384 5385@item -fdump-rtl-init-regs 5386@opindex fdump-rtl-init-regs 5387Dump after the initialization of the registers. 5388 5389@item -fdump-rtl-initvals 5390@opindex fdump-rtl-initvals 5391Dump after the computation of the initial value sets. 5392 5393@item -fdump-rtl-into_cfglayout 5394@opindex fdump-rtl-into_cfglayout 5395Dump after converting to cfglayout mode. 5396 5397@item -fdump-rtl-ira 5398@opindex fdump-rtl-ira 5399Dump after iterated register allocation. 5400 5401@item -fdump-rtl-jump 5402@opindex fdump-rtl-jump 5403Dump after the second jump optimization. 5404 5405@item -fdump-rtl-loop2 5406@opindex fdump-rtl-loop2 5407@option{-fdump-rtl-loop2} enables dumping after the rtl 5408loop optimization passes. 5409 5410@item -fdump-rtl-mach 5411@opindex fdump-rtl-mach 5412Dump after performing the machine dependent reorganization pass, if that 5413pass exists. 5414 5415@item -fdump-rtl-mode_sw 5416@opindex fdump-rtl-mode_sw 5417Dump after removing redundant mode switches. 5418 5419@item -fdump-rtl-rnreg 5420@opindex fdump-rtl-rnreg 5421Dump after register renumbering. 5422 5423@item -fdump-rtl-outof_cfglayout 5424@opindex fdump-rtl-outof_cfglayout 5425Dump after converting from cfglayout mode. 5426 5427@item -fdump-rtl-peephole2 5428@opindex fdump-rtl-peephole2 5429Dump after the peephole pass. 5430 5431@item -fdump-rtl-postreload 5432@opindex fdump-rtl-postreload 5433Dump after post-reload optimizations. 5434 5435@item -fdump-rtl-pro_and_epilogue 5436@opindex fdump-rtl-pro_and_epilogue 5437Dump after generating the function prologues and epilogues. 5438 5439@item -fdump-rtl-regmove 5440@opindex fdump-rtl-regmove 5441Dump after the register move pass. 5442 5443@item -fdump-rtl-sched1 5444@itemx -fdump-rtl-sched2 5445@opindex fdump-rtl-sched1 5446@opindex fdump-rtl-sched2 5447@option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping 5448after the basic block scheduling passes. 5449 5450@item -fdump-rtl-see 5451@opindex fdump-rtl-see 5452Dump after sign extension elimination. 5453 5454@item -fdump-rtl-seqabstr 5455@opindex fdump-rtl-seqabstr 5456Dump after common sequence discovery. 5457 5458@item -fdump-rtl-shorten 5459@opindex fdump-rtl-shorten 5460Dump after shortening branches. 5461 5462@item -fdump-rtl-sibling 5463@opindex fdump-rtl-sibling 5464Dump after sibling call optimizations. 5465 5466@item -fdump-rtl-split1 5467@itemx -fdump-rtl-split2 5468@itemx -fdump-rtl-split3 5469@itemx -fdump-rtl-split4 5470@itemx -fdump-rtl-split5 5471@opindex fdump-rtl-split1 5472@opindex fdump-rtl-split2 5473@opindex fdump-rtl-split3 5474@opindex fdump-rtl-split4 5475@opindex fdump-rtl-split5 5476@option{-fdump-rtl-split1}, @option{-fdump-rtl-split2}, 5477@option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and 5478@option{-fdump-rtl-split5} enable dumping after five rounds of 5479instruction splitting. 5480 5481@item -fdump-rtl-sms 5482@opindex fdump-rtl-sms 5483Dump after modulo scheduling. This pass is only run on some 5484architectures. 5485 5486@item -fdump-rtl-stack 5487@opindex fdump-rtl-stack 5488Dump after conversion from GCC's "flat register file" registers to the 5489x87's stack-like registers. This pass is only run on x86 variants. 5490 5491@item -fdump-rtl-subreg1 5492@itemx -fdump-rtl-subreg2 5493@opindex fdump-rtl-subreg1 5494@opindex fdump-rtl-subreg2 5495@option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after 5496the two subreg expansion passes. 5497 5498@item -fdump-rtl-unshare 5499@opindex fdump-rtl-unshare 5500Dump after all rtl has been unshared. 5501 5502@item -fdump-rtl-vartrack 5503@opindex fdump-rtl-vartrack 5504Dump after variable tracking. 5505 5506@item -fdump-rtl-vregs 5507@opindex fdump-rtl-vregs 5508Dump after converting virtual registers to hard registers. 5509 5510@item -fdump-rtl-web 5511@opindex fdump-rtl-web 5512Dump after live range splitting. 5513 5514@item -fdump-rtl-regclass 5515@itemx -fdump-rtl-subregs_of_mode_init 5516@itemx -fdump-rtl-subregs_of_mode_finish 5517@itemx -fdump-rtl-dfinit 5518@itemx -fdump-rtl-dfinish 5519@opindex fdump-rtl-regclass 5520@opindex fdump-rtl-subregs_of_mode_init 5521@opindex fdump-rtl-subregs_of_mode_finish 5522@opindex fdump-rtl-dfinit 5523@opindex fdump-rtl-dfinish 5524These dumps are defined but always produce empty files. 5525 5526@item -da 5527@itemx -fdump-rtl-all 5528@opindex da 5529@opindex fdump-rtl-all 5530Produce all the dumps listed above. 5531 5532@item -dA 5533@opindex dA 5534Annotate the assembler output with miscellaneous debugging information. 5535 5536@item -dD 5537@opindex dD 5538Dump all macro definitions, at the end of preprocessing, in addition to 5539normal output. 5540 5541@item -dH 5542@opindex dH 5543Produce a core dump whenever an error occurs. 5544 5545@item -dp 5546@opindex dp 5547Annotate the assembler output with a comment indicating which 5548pattern and alternative was used. The length of each instruction is 5549also printed. 5550 5551@item -dP 5552@opindex dP 5553Dump the RTL in the assembler output as a comment before each instruction. 5554Also turns on @option{-dp} annotation. 5555 5556@item -dv 5557@opindex dv 5558For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}), 5559dump a representation of the control flow graph suitable for viewing with VCG 5560to @file{@var{file}.@var{pass}.vcg}. 5561 5562@item -dx 5563@opindex dx 5564Just generate RTL for a function instead of compiling it. Usually used 5565with @option{-fdump-rtl-expand}. 5566@end table 5567 5568@item -fdump-noaddr 5569@opindex fdump-noaddr 5570When doing debugging dumps, suppress address output. This makes it more 5571feasible to use diff on debugging dumps for compiler invocations with 5572different compiler binaries and/or different 5573text / bss / data / heap / stack / dso start locations. 5574 5575@item -fdump-unnumbered 5576@opindex fdump-unnumbered 5577When doing debugging dumps, suppress instruction numbers and address output. 5578This makes it more feasible to use diff on debugging dumps for compiler 5579invocations with different options, in particular with and without 5580@option{-g}. 5581 5582@item -fdump-unnumbered-links 5583@opindex fdump-unnumbered-links 5584When doing debugging dumps (see @option{-d} option above), suppress 5585instruction numbers for the links to the previous and next instructions 5586in a sequence. 5587 5588@item -fdump-translation-unit @r{(C++ only)} 5589@itemx -fdump-translation-unit-@var{options} @r{(C++ only)} 5590@opindex fdump-translation-unit 5591Dump a representation of the tree structure for the entire translation 5592unit to a file. The file name is made by appending @file{.tu} to the 5593source file name, and the file is created in the same directory as the 5594output file. If the @samp{-@var{options}} form is used, @var{options} 5595controls the details of the dump as described for the 5596@option{-fdump-tree} options. 5597 5598@item -fdump-class-hierarchy @r{(C++ only)} 5599@itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)} 5600@opindex fdump-class-hierarchy 5601Dump a representation of each class's hierarchy and virtual function 5602table layout to a file. The file name is made by appending 5603@file{.class} to the source file name, and the file is created in the 5604same directory as the output file. If the @samp{-@var{options}} form 5605is used, @var{options} controls the details of the dump as described 5606for the @option{-fdump-tree} options. 5607 5608@item -fdump-ipa-@var{switch} 5609@opindex fdump-ipa 5610Control the dumping at various stages of inter-procedural analysis 5611language tree to a file. The file name is generated by appending a 5612switch specific suffix to the source file name, and the file is created 5613in the same directory as the output file. The following dumps are 5614possible: 5615 5616@table @samp 5617@item all 5618Enables all inter-procedural analysis dumps. 5619 5620@item cgraph 5621Dumps information about call-graph optimization, unused function removal, 5622and inlining decisions. 5623 5624@item inline 5625Dump after function inlining. 5626 5627@end table 5628 5629@item -fdump-passes 5630@opindex fdump-passes 5631Dump the list of optimization passes that are turned on and off by 5632the current command-line options. 5633 5634@item -fdump-statistics-@var{option} 5635@opindex fdump-statistics 5636Enable and control dumping of pass statistics in a separate file. The 5637file name is generated by appending a suffix ending in 5638@samp{.statistics} to the source file name, and the file is created in 5639the same directory as the output file. If the @samp{-@var{option}} 5640form is used, @samp{-stats} will cause counters to be summed over the 5641whole compilation unit while @samp{-details} will dump every event as 5642the passes generate them. The default with no option is to sum 5643counters for each function compiled. 5644 5645@item -fdump-tree-@var{switch} 5646@itemx -fdump-tree-@var{switch}-@var{options} 5647@opindex fdump-tree 5648Control the dumping at various stages of processing the intermediate 5649language tree to a file. The file name is generated by appending a 5650switch specific suffix to the source file name, and the file is 5651created in the same directory as the output file. If the 5652@samp{-@var{options}} form is used, @var{options} is a list of 5653@samp{-} separated options which control the details of the dump. Not 5654all options are applicable to all dumps; those that are not 5655meaningful will be ignored. The following options are available 5656 5657@table @samp 5658@item address 5659Print the address of each node. Usually this is not meaningful as it 5660changes according to the environment and source file. Its primary use 5661is for tying up a dump file with a debug environment. 5662@item asmname 5663If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that 5664in the dump instead of @code{DECL_NAME}. Its primary use is ease of 5665use working backward from mangled names in the assembly file. 5666@item slim 5667Inhibit dumping of members of a scope or body of a function merely 5668because that scope has been reached. Only dump such items when they 5669are directly reachable by some other path. When dumping pretty-printed 5670trees, this option inhibits dumping the bodies of control structures. 5671@item raw 5672Print a raw representation of the tree. By default, trees are 5673pretty-printed into a C-like representation. 5674@item details 5675Enable more detailed dumps (not honored by every dump option). 5676@item stats 5677Enable dumping various statistics about the pass (not honored by every dump 5678option). 5679@item blocks 5680Enable showing basic block boundaries (disabled in raw dumps). 5681@item vops 5682Enable showing virtual operands for every statement. 5683@item lineno 5684Enable showing line numbers for statements. 5685@item uid 5686Enable showing the unique ID (@code{DECL_UID}) for each variable. 5687@item verbose 5688Enable showing the tree dump for each statement. 5689@item eh 5690Enable showing the EH region number holding each statement. 5691@item scev 5692Enable showing scalar evolution analysis details. 5693@item all 5694Turn on all options, except @option{raw}, @option{slim}, @option{verbose} 5695and @option{lineno}. 5696@end table 5697 5698The following tree dumps are possible: 5699@table @samp 5700 5701@item original 5702@opindex fdump-tree-original 5703Dump before any tree based optimization, to @file{@var{file}.original}. 5704 5705@item optimized 5706@opindex fdump-tree-optimized 5707Dump after all tree based optimization, to @file{@var{file}.optimized}. 5708 5709@item gimple 5710@opindex fdump-tree-gimple 5711Dump each function before and after the gimplification pass to a file. The 5712file name is made by appending @file{.gimple} to the source file name. 5713 5714@item cfg 5715@opindex fdump-tree-cfg 5716Dump the control flow graph of each function to a file. The file name is 5717made by appending @file{.cfg} to the source file name. 5718 5719@item vcg 5720@opindex fdump-tree-vcg 5721Dump the control flow graph of each function to a file in VCG format. The 5722file name is made by appending @file{.vcg} to the source file name. Note 5723that if the file contains more than one function, the generated file cannot 5724be used directly by VCG@. You will need to cut and paste each function's 5725graph into its own separate file first. 5726 5727@item ch 5728@opindex fdump-tree-ch 5729Dump each function after copying loop headers. The file name is made by 5730appending @file{.ch} to the source file name. 5731 5732@item ssa 5733@opindex fdump-tree-ssa 5734Dump SSA related information to a file. The file name is made by appending 5735@file{.ssa} to the source file name. 5736 5737@item alias 5738@opindex fdump-tree-alias 5739Dump aliasing information for each function. The file name is made by 5740appending @file{.alias} to the source file name. 5741 5742@item ccp 5743@opindex fdump-tree-ccp 5744Dump each function after CCP@. The file name is made by appending 5745@file{.ccp} to the source file name. 5746 5747@item storeccp 5748@opindex fdump-tree-storeccp 5749Dump each function after STORE-CCP@. The file name is made by appending 5750@file{.storeccp} to the source file name. 5751 5752@item pre 5753@opindex fdump-tree-pre 5754Dump trees after partial redundancy elimination. The file name is made 5755by appending @file{.pre} to the source file name. 5756 5757@item fre 5758@opindex fdump-tree-fre 5759Dump trees after full redundancy elimination. The file name is made 5760by appending @file{.fre} to the source file name. 5761 5762@item copyprop 5763@opindex fdump-tree-copyprop 5764Dump trees after copy propagation. The file name is made 5765by appending @file{.copyprop} to the source file name. 5766 5767@item store_copyprop 5768@opindex fdump-tree-store_copyprop 5769Dump trees after store copy-propagation. The file name is made 5770by appending @file{.store_copyprop} to the source file name. 5771 5772@item dce 5773@opindex fdump-tree-dce 5774Dump each function after dead code elimination. The file name is made by 5775appending @file{.dce} to the source file name. 5776 5777@item mudflap 5778@opindex fdump-tree-mudflap 5779Dump each function after adding mudflap instrumentation. The file name is 5780made by appending @file{.mudflap} to the source file name. 5781 5782@item sra 5783@opindex fdump-tree-sra 5784Dump each function after performing scalar replacement of aggregates. The 5785file name is made by appending @file{.sra} to the source file name. 5786 5787@item sink 5788@opindex fdump-tree-sink 5789Dump each function after performing code sinking. The file name is made 5790by appending @file{.sink} to the source file name. 5791 5792@item dom 5793@opindex fdump-tree-dom 5794Dump each function after applying dominator tree optimizations. The file 5795name is made by appending @file{.dom} to the source file name. 5796 5797@item dse 5798@opindex fdump-tree-dse 5799Dump each function after applying dead store elimination. The file 5800name is made by appending @file{.dse} to the source file name. 5801 5802@item phiopt 5803@opindex fdump-tree-phiopt 5804Dump each function after optimizing PHI nodes into straightline code. The file 5805name is made by appending @file{.phiopt} to the source file name. 5806 5807@item forwprop 5808@opindex fdump-tree-forwprop 5809Dump each function after forward propagating single use variables. The file 5810name is made by appending @file{.forwprop} to the source file name. 5811 5812@item copyrename 5813@opindex fdump-tree-copyrename 5814Dump each function after applying the copy rename optimization. The file 5815name is made by appending @file{.copyrename} to the source file name. 5816 5817@item nrv 5818@opindex fdump-tree-nrv 5819Dump each function after applying the named return value optimization on 5820generic trees. The file name is made by appending @file{.nrv} to the source 5821file name. 5822 5823@item vect 5824@opindex fdump-tree-vect 5825Dump each function after applying vectorization of loops. The file name is 5826made by appending @file{.vect} to the source file name. 5827 5828@item slp 5829@opindex fdump-tree-slp 5830Dump each function after applying vectorization of basic blocks. The file name 5831is made by appending @file{.slp} to the source file name. 5832 5833@item vrp 5834@opindex fdump-tree-vrp 5835Dump each function after Value Range Propagation (VRP). The file name 5836is made by appending @file{.vrp} to the source file name. 5837 5838@item all 5839@opindex fdump-tree-all 5840Enable all the available tree dumps with the flags provided in this option. 5841@end table 5842 5843@item -ftree-vectorizer-verbose=@var{n} 5844@opindex ftree-vectorizer-verbose 5845This option controls the amount of debugging output the vectorizer prints. 5846This information is written to standard error, unless 5847@option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified, 5848in which case it is output to the usual dump listing file, @file{.vect}. 5849For @var{n}=0 no diagnostic information is reported. 5850If @var{n}=1 the vectorizer reports each loop that got vectorized, 5851and the total number of loops that got vectorized. 5852If @var{n}=2 the vectorizer also reports non-vectorized loops that passed 5853the first analysis phase (vect_analyze_loop_form) - i.e.@: countable, 5854inner-most, single-bb, single-entry/exit loops. This is the same verbosity 5855level that @option{-fdump-tree-vect-stats} uses. 5856Higher verbosity levels mean either more information dumped for each 5857reported loop, or same amount of information reported for more loops: 5858if @var{n}=3, vectorizer cost model information is reported. 5859If @var{n}=4, alignment related information is added to the reports. 5860If @var{n}=5, data-references related information (e.g.@: memory dependences, 5861memory access-patterns) is added to the reports. 5862If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops 5863that did not pass the first analysis phase (i.e., may not be countable, or 5864may have complicated control-flow). 5865If @var{n}=7, the vectorizer reports also non-vectorized nested loops. 5866If @var{n}=8, SLP related information is added to the reports. 5867For @var{n}=9, all the information the vectorizer generates during its 5868analysis and transformation is reported. This is the same verbosity level 5869that @option{-fdump-tree-vect-details} uses. 5870 5871@item -frandom-seed=@var{string} 5872@opindex frandom-seed 5873This option provides a seed that GCC uses when it would otherwise use 5874random numbers. It is used to generate certain symbol names 5875that have to be different in every compiled file. It is also used to 5876place unique stamps in coverage data files and the object files that 5877produce them. You can use the @option{-frandom-seed} option to produce 5878reproducibly identical object files. 5879 5880The @var{string} should be different for every file you compile. 5881 5882@item -fsched-verbose=@var{n} 5883@opindex fsched-verbose 5884On targets that use instruction scheduling, this option controls the 5885amount of debugging output the scheduler prints. This information is 5886written to standard error, unless @option{-fdump-rtl-sched1} or 5887@option{-fdump-rtl-sched2} is specified, in which case it is output 5888to the usual dump listing file, @file{.sched1} or @file{.sched2} 5889respectively. However for @var{n} greater than nine, the output is 5890always printed to standard error. 5891 5892For @var{n} greater than zero, @option{-fsched-verbose} outputs the 5893same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}. 5894For @var{n} greater than one, it also output basic block probabilities, 5895detailed ready list information and unit/insn info. For @var{n} greater 5896than two, it includes RTL at abort point, control-flow and regions info. 5897And for @var{n} over four, @option{-fsched-verbose} also includes 5898dependence info. 5899 5900@item -save-temps 5901@itemx -save-temps=cwd 5902@opindex save-temps 5903Store the usual ``temporary'' intermediate files permanently; place them 5904in the current directory and name them based on the source file. Thus, 5905compiling @file{foo.c} with @samp{-c -save-temps} would produce files 5906@file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a 5907preprocessed @file{foo.i} output file even though the compiler now 5908normally uses an integrated preprocessor. 5909 5910When used in combination with the @option{-x} command-line option, 5911@option{-save-temps} is sensible enough to avoid over writing an 5912input source file with the same extension as an intermediate file. 5913The corresponding intermediate file may be obtained by renaming the 5914source file before using @option{-save-temps}. 5915 5916If you invoke GCC in parallel, compiling several different source 5917files that share a common base name in different subdirectories or the 5918same source file compiled for multiple output destinations, it is 5919likely that the different parallel compilers will interfere with each 5920other, and overwrite the temporary files. For instance: 5921 5922@smallexample 5923gcc -save-temps -o outdir1/foo.o indir1/foo.c& 5924gcc -save-temps -o outdir2/foo.o indir2/foo.c& 5925@end smallexample 5926 5927may result in @file{foo.i} and @file{foo.o} being written to 5928simultaneously by both compilers. 5929 5930@item -save-temps=obj 5931@opindex save-temps=obj 5932Store the usual ``temporary'' intermediate files permanently. If the 5933@option{-o} option is used, the temporary files are based on the 5934object file. If the @option{-o} option is not used, the 5935@option{-save-temps=obj} switch behaves like @option{-save-temps}. 5936 5937For example: 5938 5939@smallexample 5940gcc -save-temps=obj -c foo.c 5941gcc -save-temps=obj -c bar.c -o dir/xbar.o 5942gcc -save-temps=obj foobar.c -o dir2/yfoobar 5943@end smallexample 5944 5945would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i}, 5946@file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and 5947@file{dir2/yfoobar.o}. 5948 5949@item -time@r{[}=@var{file}@r{]} 5950@opindex time 5951Report the CPU time taken by each subprocess in the compilation 5952sequence. For C source files, this is the compiler proper and assembler 5953(plus the linker if linking is done). 5954 5955Without the specification of an output file, the output looks like this: 5956 5957@smallexample 5958# cc1 0.12 0.01 5959# as 0.00 0.01 5960@end smallexample 5961 5962The first number on each line is the ``user time'', that is time spent 5963executing the program itself. The second number is ``system time'', 5964time spent executing operating system routines on behalf of the program. 5965Both numbers are in seconds. 5966 5967With the specification of an output file, the output is appended to the 5968named file, and it looks like this: 5969 5970@smallexample 59710.12 0.01 cc1 @var{options} 59720.00 0.01 as @var{options} 5973@end smallexample 5974 5975The ``user time'' and the ``system time'' are moved before the program 5976name, and the options passed to the program are displayed, so that one 5977can later tell what file was being compiled, and with which options. 5978 5979@item -fvar-tracking 5980@opindex fvar-tracking 5981Run variable tracking pass. It computes where variables are stored at each 5982position in code. Better debugging information is then generated 5983(if the debugging information format supports this information). 5984 5985It is enabled by default when compiling with optimization (@option{-Os}, 5986@option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and 5987the debug info format supports it. 5988 5989@item -fvar-tracking-assignments 5990@opindex fvar-tracking-assignments 5991@opindex fno-var-tracking-assignments 5992Annotate assignments to user variables early in the compilation and 5993attempt to carry the annotations over throughout the compilation all the 5994way to the end, in an attempt to improve debug information while 5995optimizing. Use of @option{-gdwarf-4} is recommended along with it. 5996 5997It can be enabled even if var-tracking is disabled, in which case 5998annotations will be created and maintained, but discarded at the end. 5999 6000@item -fvar-tracking-assignments-toggle 6001@opindex fvar-tracking-assignments-toggle 6002@opindex fno-var-tracking-assignments-toggle 6003Toggle @option{-fvar-tracking-assignments}, in the same way that 6004@option{-gtoggle} toggles @option{-g}. 6005 6006@item -print-file-name=@var{library} 6007@opindex print-file-name 6008Print the full absolute name of the library file @var{library} that 6009would be used when linking---and don't do anything else. With this 6010option, GCC does not compile or link anything; it just prints the 6011file name. 6012 6013@item -print-multi-directory 6014@opindex print-multi-directory 6015Print the directory name corresponding to the multilib selected by any 6016other switches present in the command line. This directory is supposed 6017to exist in @env{GCC_EXEC_PREFIX}. 6018 6019@item -print-multi-lib 6020@opindex print-multi-lib 6021Print the mapping from multilib directory names to compiler switches 6022that enable them. The directory name is separated from the switches by 6023@samp{;}, and each switch starts with an @samp{@@} instead of the 6024@samp{-}, without spaces between multiple switches. This is supposed to 6025ease shell-processing. 6026 6027@item -print-multi-os-directory 6028@opindex print-multi-os-directory 6029Print the path to OS libraries for the selected 6030multilib, relative to some @file{lib} subdirectory. If OS libraries are 6031present in the @file{lib} subdirectory and no multilibs are used, this is 6032usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}} 6033sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or 6034@file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}} 6035subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}. 6036 6037@item -print-multiarch 6038@opindex print-multiarch 6039Print the path to OS libraries for the selected multiarch, 6040relative to some @file{lib} subdirectory. 6041 6042@item -print-prog-name=@var{program} 6043@opindex print-prog-name 6044Like @option{-print-file-name}, but searches for a program such as @samp{cpp}. 6045 6046@item -print-libgcc-file-name 6047@opindex print-libgcc-file-name 6048Same as @option{-print-file-name=libgcc.a}. 6049 6050This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs} 6051but you do want to link with @file{libgcc.a}. You can do 6052 6053@smallexample 6054gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name` 6055@end smallexample 6056 6057@item -print-search-dirs 6058@opindex print-search-dirs 6059Print the name of the configured installation directory and a list of 6060program and library directories @command{gcc} will search---and don't do anything else. 6061 6062This is useful when @command{gcc} prints the error message 6063@samp{installation problem, cannot exec cpp0: No such file or directory}. 6064To resolve this you either need to put @file{cpp0} and the other compiler 6065components where @command{gcc} expects to find them, or you can set the environment 6066variable @env{GCC_EXEC_PREFIX} to the directory where you installed them. 6067Don't forget the trailing @samp{/}. 6068@xref{Environment Variables}. 6069 6070@item -print-sysroot 6071@opindex print-sysroot 6072Print the target sysroot directory that will be used during 6073compilation. This is the target sysroot specified either at configure 6074time or using the @option{--sysroot} option, possibly with an extra 6075suffix that depends on compilation options. If no target sysroot is 6076specified, the option prints nothing. 6077 6078@item -print-sysroot-headers-suffix 6079@opindex print-sysroot-headers-suffix 6080Print the suffix added to the target sysroot when searching for 6081headers, or give an error if the compiler is not configured with such 6082a suffix---and don't do anything else. 6083 6084@item -dumpmachine 6085@opindex dumpmachine 6086Print the compiler's target machine (for example, 6087@samp{i686-pc-linux-gnu})---and don't do anything else. 6088 6089@item -dumpversion 6090@opindex dumpversion 6091Print the compiler version (for example, @samp{3.0})---and don't do 6092anything else. 6093 6094@item -dumpspecs 6095@opindex dumpspecs 6096Print the compiler's built-in specs---and don't do anything else. (This 6097is used when GCC itself is being built.) @xref{Spec Files}. 6098 6099@item -feliminate-unused-debug-types 6100@opindex feliminate-unused-debug-types 6101Normally, when producing DWARF2 output, GCC will emit debugging 6102information for all types declared in a compilation 6103unit, regardless of whether or not they are actually used 6104in that compilation unit. Sometimes this is useful, such as 6105if, in the debugger, you want to cast a value to a type that is 6106not actually used in your program (but is declared). More often, 6107however, this results in a significant amount of wasted space. 6108With this option, GCC will avoid producing debug symbol output 6109for types that are nowhere used in the source file being compiled. 6110@end table 6111 6112@node Optimize Options 6113@section Options That Control Optimization 6114@cindex optimize options 6115@cindex options, optimization 6116 6117These options control various sorts of optimizations. 6118 6119Without any optimization option, the compiler's goal is to reduce the 6120cost of compilation and to make debugging produce the expected 6121results. Statements are independent: if you stop the program with a 6122breakpoint between statements, you can then assign a new value to any 6123variable or change the program counter to any other statement in the 6124function and get exactly the results you would expect from the source 6125code. 6126 6127Turning on optimization flags makes the compiler attempt to improve 6128the performance and/or code size at the expense of compilation time 6129and possibly the ability to debug the program. 6130 6131The compiler performs optimization based on the knowledge it has of the 6132program. Compiling multiple files at once to a single output file mode allows 6133the compiler to use information gained from all of the files when compiling 6134each of them. 6135 6136Not all optimizations are controlled directly by a flag. Only 6137optimizations that have a flag are listed in this section. 6138 6139Most optimizations are only enabled if an @option{-O} level is set on 6140the command line. Otherwise they are disabled, even if individual 6141optimization flags are specified. 6142 6143Depending on the target and how GCC was configured, a slightly different 6144set of optimizations may be enabled at each @option{-O} level than 6145those listed here. You can invoke GCC with @samp{-Q --help=optimizers} 6146to find out the exact set of optimizations that are enabled at each level. 6147@xref{Overall Options}, for examples. 6148 6149@table @gcctabopt 6150@item -O 6151@itemx -O1 6152@opindex O 6153@opindex O1 6154Optimize. Optimizing compilation takes somewhat more time, and a lot 6155more memory for a large function. 6156 6157With @option{-O}, the compiler tries to reduce code size and execution 6158time, without performing any optimizations that take a great deal of 6159compilation time. 6160 6161@option{-O} turns on the following optimization flags: 6162@gccoptlist{ 6163-fauto-inc-dec @gol 6164-fcompare-elim @gol 6165-fcprop-registers @gol 6166-fdce @gol 6167-fdefer-pop @gol 6168-fdelayed-branch @gol 6169-fdse @gol 6170-fguess-branch-probability @gol 6171-fif-conversion2 @gol 6172-fif-conversion @gol 6173-fipa-pure-const @gol 6174-fipa-profile @gol 6175-fipa-reference @gol 6176-fmerge-constants 6177-fsplit-wide-types @gol 6178-ftree-bit-ccp @gol 6179-ftree-builtin-call-dce @gol 6180-ftree-ccp @gol 6181-ftree-ch @gol 6182-ftree-copyrename @gol 6183-ftree-dce @gol 6184-ftree-dominator-opts @gol 6185-ftree-dse @gol 6186-ftree-forwprop @gol 6187-ftree-fre @gol 6188-ftree-phiprop @gol 6189-ftree-sra @gol 6190-ftree-pta @gol 6191-ftree-ter @gol 6192-funit-at-a-time} 6193 6194@option{-O} also turns on @option{-fomit-frame-pointer} on machines 6195where doing so does not interfere with debugging. 6196 6197@item -O2 6198@opindex O2 6199Optimize even more. GCC performs nearly all supported optimizations 6200that do not involve a space-speed tradeoff. 6201As compared to @option{-O}, this option increases both compilation time 6202and the performance of the generated code. 6203 6204@option{-O2} turns on all optimization flags specified by @option{-O}. It 6205also turns on the following optimization flags: 6206@gccoptlist{-fthread-jumps @gol 6207-falign-functions -falign-jumps @gol 6208-falign-loops -falign-labels @gol 6209-fcaller-saves @gol 6210-fcrossjumping @gol 6211-fcse-follow-jumps -fcse-skip-blocks @gol 6212-fdelete-null-pointer-checks @gol 6213-fdevirtualize @gol 6214-fexpensive-optimizations @gol 6215-fgcse -fgcse-lm @gol 6216-finline-small-functions @gol 6217-findirect-inlining @gol 6218-fipa-sra @gol 6219-foptimize-sibling-calls @gol 6220-fpartial-inlining @gol 6221-fpeephole2 @gol 6222-fregmove @gol 6223-freorder-blocks -freorder-functions @gol 6224-frerun-cse-after-loop @gol 6225-fsched-interblock -fsched-spec @gol 6226-fschedule-insns -fschedule-insns2 @gol 6227-fstrict-aliasing -fstrict-overflow @gol 6228-ftree-switch-conversion -ftree-tail-merge @gol 6229-ftree-pre @gol 6230-ftree-vrp} 6231 6232Please note the warning under @option{-fgcse} about 6233invoking @option{-O2} on programs that use computed gotos. 6234 6235@item -O3 6236@opindex O3 6237Optimize yet more. @option{-O3} turns on all optimizations specified 6238by @option{-O2} and also turns on the @option{-finline-functions}, 6239@option{-funswitch-loops}, @option{-fpredictive-commoning}, 6240@option{-fgcse-after-reload}, @option{-ftree-vectorize} and 6241@option{-fipa-cp-clone} options. 6242 6243@item -O0 6244@opindex O0 6245Reduce compilation time and make debugging produce the expected 6246results. This is the default. 6247 6248@item -Os 6249@opindex Os 6250Optimize for size. @option{-Os} enables all @option{-O2} optimizations that 6251do not typically increase code size. It also performs further 6252optimizations designed to reduce code size. 6253 6254@option{-Os} disables the following optimization flags: 6255@gccoptlist{-falign-functions -falign-jumps -falign-loops @gol 6256-falign-labels -freorder-blocks -freorder-blocks-and-partition @gol 6257-fprefetch-loop-arrays -ftree-vect-loop-version} 6258 6259@item -Ofast 6260@opindex Ofast 6261Disregard strict standards compliance. @option{-Ofast} enables all 6262@option{-O3} optimizations. It also enables optimizations that are not 6263valid for all standard compliant programs. 6264It turns on @option{-ffast-math} and the Fortran-specific 6265@option{-fno-protect-parens} and @option{-fstack-arrays}. 6266 6267If you use multiple @option{-O} options, with or without level numbers, 6268the last such option is the one that is effective. 6269@end table 6270 6271Options of the form @option{-f@var{flag}} specify machine-independent 6272flags. Most flags have both positive and negative forms; the negative 6273form of @option{-ffoo} would be @option{-fno-foo}. In the table 6274below, only one of the forms is listed---the one you typically will 6275use. You can figure out the other form by either removing @samp{no-} 6276or adding it. 6277 6278The following options control specific optimizations. They are either 6279activated by @option{-O} options or are related to ones that are. You 6280can use the following flags in the rare cases when ``fine-tuning'' of 6281optimizations to be performed is desired. 6282 6283@table @gcctabopt 6284@item -fno-default-inline 6285@opindex fno-default-inline 6286Do not make member functions inline by default merely because they are 6287defined inside the class scope (C++ only). Otherwise, when you specify 6288@w{@option{-O}}, member functions defined inside class scope are compiled 6289inline by default; i.e., you don't need to add @samp{inline} in front of 6290the member function name. 6291 6292@item -fno-defer-pop 6293@opindex fno-defer-pop 6294Always pop the arguments to each function call as soon as that function 6295returns. For machines that must pop arguments after a function call, 6296the compiler normally lets arguments accumulate on the stack for several 6297function calls and pops them all at once. 6298 6299Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 6300 6301@item -fforward-propagate 6302@opindex fforward-propagate 6303Perform a forward propagation pass on RTL@. The pass tries to combine two 6304instructions and checks if the result can be simplified. If loop unrolling 6305is active, two passes are performed and the second is scheduled after 6306loop unrolling. 6307 6308This option is enabled by default at optimization levels @option{-O}, 6309@option{-O2}, @option{-O3}, @option{-Os}. 6310 6311@item -ffp-contract=@var{style} 6312@opindex ffp-contract 6313@option{-ffp-contract=off} disables floating-point expression contraction. 6314@option{-ffp-contract=fast} enables floating-point expression contraction 6315such as forming of fused multiply-add operations if the target has 6316native support for them. 6317@option{-ffp-contract=on} enables floating-point expression contraction 6318if allowed by the language standard. This is currently not implemented 6319and treated equal to @option{-ffp-contract=off}. 6320 6321The default is @option{-ffp-contract=fast}. 6322 6323@item -fomit-frame-pointer 6324@opindex fomit-frame-pointer 6325Don't keep the frame pointer in a register for functions that 6326don't need one. This avoids the instructions to save, set up and 6327restore frame pointers; it also makes an extra register available 6328in many functions. @strong{It also makes debugging impossible on 6329some machines.} 6330 6331On some machines, such as the VAX, this flag has no effect, because 6332the standard calling sequence automatically handles the frame pointer 6333and nothing is saved by pretending it doesn't exist. The 6334machine-description macro @code{FRAME_POINTER_REQUIRED} controls 6335whether a target machine supports this flag. @xref{Registers,,Register 6336Usage, gccint, GNU Compiler Collection (GCC) Internals}. 6337 6338Starting with GCC version 4.6, the default setting (when not optimizing for 6339size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to 6340@option{-fomit-frame-pointer}. The default can be reverted to 6341@option{-fno-omit-frame-pointer} by configuring GCC with the 6342@option{--enable-frame-pointer} configure option. 6343 6344Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 6345 6346@item -foptimize-sibling-calls 6347@opindex foptimize-sibling-calls 6348Optimize sibling and tail recursive calls. 6349 6350Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6351 6352@item -fno-inline 6353@opindex fno-inline 6354Do not expand any functions inline apart from those marked with 6355the @code{always_inline} attribute. This is the default when not 6356optimizing. 6357 6358Single functions can be exempted from inlining by marking them 6359with the @code{noinline} attribute. 6360 6361@item -finline-small-functions 6362@opindex finline-small-functions 6363Integrate functions into their callers when their body is smaller than expected 6364function call code (so overall size of program gets smaller). The compiler 6365heuristically decides which functions are simple enough to be worth integrating 6366in this way. This inlining applies to all functions, even those not declared 6367inline. 6368 6369Enabled at level @option{-O2}. 6370 6371@item -findirect-inlining 6372@opindex findirect-inlining 6373Inline also indirect calls that are discovered to be known at compile 6374time thanks to previous inlining. This option has any effect only 6375when inlining itself is turned on by the @option{-finline-functions} 6376or @option{-finline-small-functions} options. 6377 6378Enabled at level @option{-O2}. 6379 6380@item -finline-functions 6381@opindex finline-functions 6382Consider all functions for inlining, even if they are not declared inline. 6383The compiler heuristically decides which functions are worth integrating 6384in this way. 6385 6386If all calls to a given function are integrated, and the function is 6387declared @code{static}, then the function is normally not output as 6388assembler code in its own right. 6389 6390Enabled at level @option{-O3}. 6391 6392@item -finline-functions-called-once 6393@opindex finline-functions-called-once 6394Consider all @code{static} functions called once for inlining into their 6395caller even if they are not marked @code{inline}. If a call to a given 6396function is integrated, then the function is not output as assembler code 6397in its own right. 6398 6399Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}. 6400 6401@item -fearly-inlining 6402@opindex fearly-inlining 6403Inline functions marked by @code{always_inline} and functions whose body seems 6404smaller than the function call overhead early before doing 6405@option{-fprofile-generate} instrumentation and real inlining pass. Doing so 6406makes profiling significantly cheaper and usually inlining faster on programs 6407having large chains of nested wrapper functions. 6408 6409Enabled by default. 6410 6411@item -fipa-sra 6412@opindex fipa-sra 6413Perform interprocedural scalar replacement of aggregates, removal of 6414unused parameters and replacement of parameters passed by reference 6415by parameters passed by value. 6416 6417Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}. 6418 6419@item -finline-limit=@var{n} 6420@opindex finline-limit 6421By default, GCC limits the size of functions that can be inlined. This flag 6422allows coarse control of this limit. @var{n} is the size of functions that 6423can be inlined in number of pseudo instructions. 6424 6425Inlining is actually controlled by a number of parameters, which may be 6426specified individually by using @option{--param @var{name}=@var{value}}. 6427The @option{-finline-limit=@var{n}} option sets some of these parameters 6428as follows: 6429 6430@table @gcctabopt 6431@item max-inline-insns-single 6432is set to @var{n}/2. 6433@item max-inline-insns-auto 6434is set to @var{n}/2. 6435@end table 6436 6437See below for a documentation of the individual 6438parameters controlling inlining and for the defaults of these parameters. 6439 6440@emph{Note:} there may be no value to @option{-finline-limit} that results 6441in default behavior. 6442 6443@emph{Note:} pseudo instruction represents, in this particular context, an 6444abstract measurement of function's size. In no way does it represent a count 6445of assembly instructions and as such its exact meaning might change from one 6446release to an another. 6447 6448@item -fno-keep-inline-dllexport 6449@opindex -fno-keep-inline-dllexport 6450This is a more fine-grained version of @option{-fkeep-inline-functions}, 6451which applies only to functions that are declared using the @code{dllexport} 6452attribute or declspec (@xref{Function Attributes,,Declaring Attributes of 6453Functions}.) 6454 6455@item -fkeep-inline-functions 6456@opindex fkeep-inline-functions 6457In C, emit @code{static} functions that are declared @code{inline} 6458into the object file, even if the function has been inlined into all 6459of its callers. This switch does not affect functions using the 6460@code{extern inline} extension in GNU C90@. In C++, emit any and all 6461inline functions into the object file. 6462 6463@item -fkeep-static-consts 6464@opindex fkeep-static-consts 6465Emit variables declared @code{static const} when optimization isn't turned 6466on, even if the variables aren't referenced. 6467 6468GCC enables this option by default. If you want to force the compiler to 6469check if the variable was referenced, regardless of whether or not 6470optimization is turned on, use the @option{-fno-keep-static-consts} option. 6471 6472@item -fmerge-constants 6473@opindex fmerge-constants 6474Attempt to merge identical constants (string constants and floating-point 6475constants) across compilation units. 6476 6477This option is the default for optimized compilation if the assembler and 6478linker support it. Use @option{-fno-merge-constants} to inhibit this 6479behavior. 6480 6481Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 6482 6483@item -fmerge-all-constants 6484@opindex fmerge-all-constants 6485Attempt to merge identical constants and identical variables. 6486 6487This option implies @option{-fmerge-constants}. In addition to 6488@option{-fmerge-constants} this considers e.g.@: even constant initialized 6489arrays or initialized constant variables with integral or floating-point 6490types. Languages like C or C++ require each variable, including multiple 6491instances of the same variable in recursive calls, to have distinct locations, 6492so using this option will result in non-conforming 6493behavior. 6494 6495@item -fmodulo-sched 6496@opindex fmodulo-sched 6497Perform swing modulo scheduling immediately before the first scheduling 6498pass. This pass looks at innermost loops and reorders their 6499instructions by overlapping different iterations. 6500 6501@item -fmodulo-sched-allow-regmoves 6502@opindex fmodulo-sched-allow-regmoves 6503Perform more aggressive SMS based modulo scheduling with register moves 6504allowed. By setting this flag certain anti-dependences edges will be 6505deleted which will trigger the generation of reg-moves based on the 6506life-range analysis. This option is effective only with 6507@option{-fmodulo-sched} enabled. 6508 6509@item -fno-branch-count-reg 6510@opindex fno-branch-count-reg 6511Do not use ``decrement and branch'' instructions on a count register, 6512but instead generate a sequence of instructions that decrement a 6513register, compare it against zero, then branch based upon the result. 6514This option is only meaningful on architectures that support such 6515instructions, which include x86, PowerPC, IA-64 and S/390. 6516 6517The default is @option{-fbranch-count-reg}. 6518 6519@item -fno-function-cse 6520@opindex fno-function-cse 6521Do not put function addresses in registers; make each instruction that 6522calls a constant function contain the function's address explicitly. 6523 6524This option results in less efficient code, but some strange hacks 6525that alter the assembler output may be confused by the optimizations 6526performed when this option is not used. 6527 6528The default is @option{-ffunction-cse} 6529 6530@item -fno-zero-initialized-in-bss 6531@opindex fno-zero-initialized-in-bss 6532If the target supports a BSS section, GCC by default puts variables that 6533are initialized to zero into BSS@. This can save space in the resulting 6534code. 6535 6536This option turns off this behavior because some programs explicitly 6537rely on variables going to the data section. E.g., so that the 6538resulting executable can find the beginning of that section and/or make 6539assumptions based on that. 6540 6541The default is @option{-fzero-initialized-in-bss}. 6542 6543@item -fmudflap -fmudflapth -fmudflapir 6544@opindex fmudflap 6545@opindex fmudflapth 6546@opindex fmudflapir 6547@cindex bounds checking 6548@cindex mudflap 6549For front-ends that support it (C and C++), instrument all risky 6550pointer/array dereferencing operations, some standard library 6551string/heap functions, and some other associated constructs with 6552range/validity tests. Modules so instrumented should be immune to 6553buffer overflows, invalid heap use, and some other classes of C/C++ 6554programming errors. The instrumentation relies on a separate runtime 6555library (@file{libmudflap}), which will be linked into a program if 6556@option{-fmudflap} is given at link time. Run-time behavior of the 6557instrumented program is controlled by the @env{MUDFLAP_OPTIONS} 6558environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out} 6559for its options. 6560 6561Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to 6562link if your program is multi-threaded. Use @option{-fmudflapir}, in 6563addition to @option{-fmudflap} or @option{-fmudflapth}, if 6564instrumentation should ignore pointer reads. This produces less 6565instrumentation (and therefore faster execution) and still provides 6566some protection against outright memory corrupting writes, but allows 6567erroneously read data to propagate within a program. 6568 6569@item -fthread-jumps 6570@opindex fthread-jumps 6571Perform optimizations where we check to see if a jump branches to a 6572location where another comparison subsumed by the first is found. If 6573so, the first branch is redirected to either the destination of the 6574second branch or a point immediately following it, depending on whether 6575the condition is known to be true or false. 6576 6577Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6578 6579@item -fsplit-wide-types 6580@opindex fsplit-wide-types 6581When using a type that occupies multiple registers, such as @code{long 6582long} on a 32-bit system, split the registers apart and allocate them 6583independently. This normally generates better code for those types, 6584but may make debugging more difficult. 6585 6586Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, 6587@option{-Os}. 6588 6589@item -fcse-follow-jumps 6590@opindex fcse-follow-jumps 6591In common subexpression elimination (CSE), scan through jump instructions 6592when the target of the jump is not reached by any other path. For 6593example, when CSE encounters an @code{if} statement with an 6594@code{else} clause, CSE will follow the jump when the condition 6595tested is false. 6596 6597Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6598 6599@item -fcse-skip-blocks 6600@opindex fcse-skip-blocks 6601This is similar to @option{-fcse-follow-jumps}, but causes CSE to 6602follow jumps that conditionally skip over blocks. When CSE 6603encounters a simple @code{if} statement with no else clause, 6604@option{-fcse-skip-blocks} causes CSE to follow the jump around the 6605body of the @code{if}. 6606 6607Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6608 6609@item -frerun-cse-after-loop 6610@opindex frerun-cse-after-loop 6611Re-run common subexpression elimination after loop optimizations has been 6612performed. 6613 6614Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6615 6616@item -fgcse 6617@opindex fgcse 6618Perform a global common subexpression elimination pass. 6619This pass also performs global constant and copy propagation. 6620 6621@emph{Note:} When compiling a program using computed gotos, a GCC 6622extension, you may get better run-time performance if you disable 6623the global common subexpression elimination pass by adding 6624@option{-fno-gcse} to the command line. 6625 6626Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6627 6628@item -fgcse-lm 6629@opindex fgcse-lm 6630When @option{-fgcse-lm} is enabled, global common subexpression elimination will 6631attempt to move loads that are only killed by stores into themselves. This 6632allows a loop containing a load/store sequence to be changed to a load outside 6633the loop, and a copy/store within the loop. 6634 6635Enabled by default when gcse is enabled. 6636 6637@item -fgcse-sm 6638@opindex fgcse-sm 6639When @option{-fgcse-sm} is enabled, a store motion pass is run after 6640global common subexpression elimination. This pass will attempt to move 6641stores out of loops. When used in conjunction with @option{-fgcse-lm}, 6642loops containing a load/store sequence can be changed to a load before 6643the loop and a store after the loop. 6644 6645Not enabled at any optimization level. 6646 6647@item -fgcse-las 6648@opindex fgcse-las 6649When @option{-fgcse-las} is enabled, the global common subexpression 6650elimination pass eliminates redundant loads that come after stores to the 6651same memory location (both partial and full redundancies). 6652 6653Not enabled at any optimization level. 6654 6655@item -fgcse-after-reload 6656@opindex fgcse-after-reload 6657When @option{-fgcse-after-reload} is enabled, a redundant load elimination 6658pass is performed after reload. The purpose of this pass is to cleanup 6659redundant spilling. 6660 6661@item -funsafe-loop-optimizations 6662@opindex funsafe-loop-optimizations 6663If given, the loop optimizer will assume that loop indices do not 6664overflow, and that the loops with nontrivial exit condition are not 6665infinite. This enables a wider range of loop optimizations even if 6666the loop optimizer itself cannot prove that these assumptions are valid. 6667Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you 6668if it finds this kind of loop. 6669 6670@item -fcrossjumping 6671@opindex fcrossjumping 6672Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The 6673resulting code may or may not perform better than without cross-jumping. 6674 6675Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6676 6677@item -fauto-inc-dec 6678@opindex fauto-inc-dec 6679Combine increments or decrements of addresses with memory accesses. 6680This pass is always skipped on architectures that do not have 6681instructions to support this. Enabled by default at @option{-O} and 6682higher on architectures that support this. 6683 6684@item -fdce 6685@opindex fdce 6686Perform dead code elimination (DCE) on RTL@. 6687Enabled by default at @option{-O} and higher. 6688 6689@item -fdse 6690@opindex fdse 6691Perform dead store elimination (DSE) on RTL@. 6692Enabled by default at @option{-O} and higher. 6693 6694@item -fif-conversion 6695@opindex fif-conversion 6696Attempt to transform conditional jumps into branch-less equivalents. This 6697include use of conditional moves, min, max, set flags and abs instructions, and 6698some tricks doable by standard arithmetics. The use of conditional execution 6699on chips where it is available is controlled by @code{if-conversion2}. 6700 6701Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 6702 6703@item -fif-conversion2 6704@opindex fif-conversion2 6705Use conditional execution (where available) to transform conditional jumps into 6706branch-less equivalents. 6707 6708Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 6709 6710@item -fdelete-null-pointer-checks 6711@opindex fdelete-null-pointer-checks 6712Assume that programs cannot safely dereference null pointers, and that 6713no code or data element resides there. This enables simple constant 6714folding optimizations at all optimization levels. In addition, other 6715optimization passes in GCC use this flag to control global dataflow 6716analyses that eliminate useless checks for null pointers; these assume 6717that if a pointer is checked after it has already been dereferenced, 6718it cannot be null. 6719 6720Note however that in some environments this assumption is not true. 6721Use @option{-fno-delete-null-pointer-checks} to disable this optimization 6722for programs that depend on that behavior. 6723 6724Some targets, especially embedded ones, disable this option at all levels. 6725Otherwise it is enabled at all levels: @option{-O0}, @option{-O1}, 6726@option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information 6727are enabled independently at different optimization levels. 6728 6729@item -fdevirtualize 6730@opindex fdevirtualize 6731Attempt to convert calls to virtual functions to direct calls. This 6732is done both within a procedure and interprocedurally as part of 6733indirect inlining (@code{-findirect-inlining}) and interprocedural constant 6734propagation (@option{-fipa-cp}). 6735Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6736 6737@item -fexpensive-optimizations 6738@opindex fexpensive-optimizations 6739Perform a number of minor optimizations that are relatively expensive. 6740 6741Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6742 6743@item -free 6744@opindex free 6745Attempt to remove redundant extension instructions. This is especially 6746helpful for the x86-64 architecture which implicitly zero-extends in 64-bit 6747registers after writing to their lower 32-bit half. 6748 6749Enabled for x86 at levels @option{-O2}, @option{-O3}. 6750 6751@item -foptimize-register-move 6752@itemx -fregmove 6753@opindex foptimize-register-move 6754@opindex fregmove 6755Attempt to reassign register numbers in move instructions and as 6756operands of other simple instructions in order to maximize the amount of 6757register tying. This is especially helpful on machines with two-operand 6758instructions. 6759 6760Note @option{-fregmove} and @option{-foptimize-register-move} are the same 6761optimization. 6762 6763Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6764 6765@item -fira-algorithm=@var{algorithm} 6766Use the specified coloring algorithm for the integrated register 6767allocator. The @var{algorithm} argument can be @samp{priority}, which 6768specifies Chow's priority coloring, or @samp{CB}, which specifies 6769Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented 6770for all architectures, but for those targets that do support it, it is 6771the default because it generates better code. 6772 6773@item -fira-region=@var{region} 6774Use specified regions for the integrated register allocator. The 6775@var{region} argument should be one of the following: 6776 6777@table @samp 6778 6779@item all 6780Use all loops as register allocation regions. 6781This can give the best results for machines with a small and/or 6782irregular register set. 6783 6784@item mixed 6785Use all loops except for loops with small register pressure 6786as the regions. This value usually gives 6787the best results in most cases and for most architectures, 6788and is enabled by default when compiling with optimization for speed 6789(@option{-O}, @option{-O2}, @dots{}). 6790 6791@item one 6792Use all functions as a single region. 6793This typically results in the smallest code size, and is enabled by default for 6794@option{-Os} or @option{-O0}. 6795 6796@end table 6797 6798@item -fira-loop-pressure 6799@opindex fira-loop-pressure 6800Use IRA to evaluate register pressure in loops for decisions to move 6801loop invariants. This option usually results in generation 6802of faster and smaller code on machines with large register files (>= 32 6803registers), but it can slow the compiler down. 6804 6805This option is enabled at level @option{-O3} for some targets. 6806 6807@item -fno-ira-share-save-slots 6808@opindex fno-ira-share-save-slots 6809Disable sharing of stack slots used for saving call-used hard 6810registers living through a call. Each hard register gets a 6811separate stack slot, and as a result function stack frames are 6812larger. 6813 6814@item -fno-ira-share-spill-slots 6815@opindex fno-ira-share-spill-slots 6816Disable sharing of stack slots allocated for pseudo-registers. Each 6817pseudo-register that does not get a hard register gets a separate 6818stack slot, and as a result function stack frames are larger. 6819 6820@item -fira-verbose=@var{n} 6821@opindex fira-verbose 6822Control the verbosity of the dump file for the integrated register allocator. 6823The default value is 5. If the value @var{n} is greater or equal to 10, 6824the dump output is sent to stderr using the same format as @var{n} minus 10. 6825 6826@item -fdelayed-branch 6827@opindex fdelayed-branch 6828If supported for the target machine, attempt to reorder instructions 6829to exploit instruction slots available after delayed branch 6830instructions. 6831 6832Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 6833 6834@item -fschedule-insns 6835@opindex fschedule-insns 6836If supported for the target machine, attempt to reorder instructions to 6837eliminate execution stalls due to required data being unavailable. This 6838helps machines that have slow floating point or memory load instructions 6839by allowing other instructions to be issued until the result of the load 6840or floating-point instruction is required. 6841 6842Enabled at levels @option{-O2}, @option{-O3}. 6843 6844@item -fschedule-insns2 6845@opindex fschedule-insns2 6846Similar to @option{-fschedule-insns}, but requests an additional pass of 6847instruction scheduling after register allocation has been done. This is 6848especially useful on machines with a relatively small number of 6849registers and where memory load instructions take more than one cycle. 6850 6851Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 6852 6853@item -fno-sched-interblock 6854@opindex fno-sched-interblock 6855Don't schedule instructions across basic blocks. This is normally 6856enabled by default when scheduling before register allocation, i.e.@: 6857with @option{-fschedule-insns} or at @option{-O2} or higher. 6858 6859@item -fno-sched-spec 6860@opindex fno-sched-spec 6861Don't allow speculative motion of non-load instructions. This is normally 6862enabled by default when scheduling before register allocation, i.e.@: 6863with @option{-fschedule-insns} or at @option{-O2} or higher. 6864 6865@item -fsched-pressure 6866@opindex fsched-pressure 6867Enable register pressure sensitive insn scheduling before the register 6868allocation. This only makes sense when scheduling before register 6869allocation is enabled, i.e.@: with @option{-fschedule-insns} or at 6870@option{-O2} or higher. Usage of this option can improve the 6871generated code and decrease its size by preventing register pressure 6872increase above the number of available hard registers and as a 6873consequence register spills in the register allocation. 6874 6875@item -fsched-spec-load 6876@opindex fsched-spec-load 6877Allow speculative motion of some load instructions. This only makes 6878sense when scheduling before register allocation, i.e.@: with 6879@option{-fschedule-insns} or at @option{-O2} or higher. 6880 6881@item -fsched-spec-load-dangerous 6882@opindex fsched-spec-load-dangerous 6883Allow speculative motion of more load instructions. This only makes 6884sense when scheduling before register allocation, i.e.@: with 6885@option{-fschedule-insns} or at @option{-O2} or higher. 6886 6887@item -fsched-stalled-insns 6888@itemx -fsched-stalled-insns=@var{n} 6889@opindex fsched-stalled-insns 6890Define how many insns (if any) can be moved prematurely from the queue 6891of stalled insns into the ready list, during the second scheduling pass. 6892@option{-fno-sched-stalled-insns} means that no insns will be moved 6893prematurely, @option{-fsched-stalled-insns=0} means there is no limit 6894on how many queued insns can be moved prematurely. 6895@option{-fsched-stalled-insns} without a value is equivalent to 6896@option{-fsched-stalled-insns=1}. 6897 6898@item -fsched-stalled-insns-dep 6899@itemx -fsched-stalled-insns-dep=@var{n} 6900@opindex fsched-stalled-insns-dep 6901Define how many insn groups (cycles) will be examined for a dependency 6902on a stalled insn that is candidate for premature removal from the queue 6903of stalled insns. This has an effect only during the second scheduling pass, 6904and only if @option{-fsched-stalled-insns} is used. 6905@option{-fno-sched-stalled-insns-dep} is equivalent to 6906@option{-fsched-stalled-insns-dep=0}. 6907@option{-fsched-stalled-insns-dep} without a value is equivalent to 6908@option{-fsched-stalled-insns-dep=1}. 6909 6910@item -fsched2-use-superblocks 6911@opindex fsched2-use-superblocks 6912When scheduling after register allocation, do use superblock scheduling 6913algorithm. Superblock scheduling allows motion across basic block boundaries 6914resulting on faster schedules. This option is experimental, as not all machine 6915descriptions used by GCC model the CPU closely enough to avoid unreliable 6916results from the algorithm. 6917 6918This only makes sense when scheduling after register allocation, i.e.@: with 6919@option{-fschedule-insns2} or at @option{-O2} or higher. 6920 6921@item -fsched-group-heuristic 6922@opindex fsched-group-heuristic 6923Enable the group heuristic in the scheduler. This heuristic favors 6924the instruction that belongs to a schedule group. This is enabled 6925by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns} 6926or @option{-fschedule-insns2} or at @option{-O2} or higher. 6927 6928@item -fsched-critical-path-heuristic 6929@opindex fsched-critical-path-heuristic 6930Enable the critical-path heuristic in the scheduler. This heuristic favors 6931instructions on the critical path. This is enabled by default when 6932scheduling is enabled, i.e.@: with @option{-fschedule-insns} 6933or @option{-fschedule-insns2} or at @option{-O2} or higher. 6934 6935@item -fsched-spec-insn-heuristic 6936@opindex fsched-spec-insn-heuristic 6937Enable the speculative instruction heuristic in the scheduler. This 6938heuristic favors speculative instructions with greater dependency weakness. 6939This is enabled by default when scheduling is enabled, i.e.@: 6940with @option{-fschedule-insns} or @option{-fschedule-insns2} 6941or at @option{-O2} or higher. 6942 6943@item -fsched-rank-heuristic 6944@opindex fsched-rank-heuristic 6945Enable the rank heuristic in the scheduler. This heuristic favors 6946the instruction belonging to a basic block with greater size or frequency. 6947This is enabled by default when scheduling is enabled, i.e.@: 6948with @option{-fschedule-insns} or @option{-fschedule-insns2} or 6949at @option{-O2} or higher. 6950 6951@item -fsched-last-insn-heuristic 6952@opindex fsched-last-insn-heuristic 6953Enable the last-instruction heuristic in the scheduler. This heuristic 6954favors the instruction that is less dependent on the last instruction 6955scheduled. This is enabled by default when scheduling is enabled, 6956i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or 6957at @option{-O2} or higher. 6958 6959@item -fsched-dep-count-heuristic 6960@opindex fsched-dep-count-heuristic 6961Enable the dependent-count heuristic in the scheduler. This heuristic 6962favors the instruction that has more instructions depending on it. 6963This is enabled by default when scheduling is enabled, i.e.@: 6964with @option{-fschedule-insns} or @option{-fschedule-insns2} or 6965at @option{-O2} or higher. 6966 6967@item -freschedule-modulo-scheduled-loops 6968@opindex freschedule-modulo-scheduled-loops 6969The modulo scheduling comes before the traditional scheduling, if a loop 6970was modulo scheduled we may want to prevent the later scheduling passes 6971from changing its schedule, we use this option to control that. 6972 6973@item -fselective-scheduling 6974@opindex fselective-scheduling 6975Schedule instructions using selective scheduling algorithm. Selective 6976scheduling runs instead of the first scheduler pass. 6977 6978@item -fselective-scheduling2 6979@opindex fselective-scheduling2 6980Schedule instructions using selective scheduling algorithm. Selective 6981scheduling runs instead of the second scheduler pass. 6982 6983@item -fsel-sched-pipelining 6984@opindex fsel-sched-pipelining 6985Enable software pipelining of innermost loops during selective scheduling. 6986This option has no effect until one of @option{-fselective-scheduling} or 6987@option{-fselective-scheduling2} is turned on. 6988 6989@item -fsel-sched-pipelining-outer-loops 6990@opindex fsel-sched-pipelining-outer-loops 6991When pipelining loops during selective scheduling, also pipeline outer loops. 6992This option has no effect until @option{-fsel-sched-pipelining} is turned on. 6993 6994@item -fshrink-wrap 6995@opindex fshrink-wrap 6996Emit function prologues only before parts of the function that need it, 6997rather than at the top of the function. This flag is enabled by default at 6998@option{-O} and higher. 6999 7000@item -fcaller-saves 7001@opindex fcaller-saves 7002Enable values to be allocated in registers that will be clobbered by 7003function calls, by emitting extra instructions to save and restore the 7004registers around such calls. Such allocation is done only when it 7005seems to result in better code than would otherwise be produced. 7006 7007This option is always enabled by default on certain machines, usually 7008those which have no call-preserved registers to use instead. 7009 7010Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 7011 7012@item -fcombine-stack-adjustments 7013@opindex fcombine-stack-adjustments 7014Tracks stack adjustments (pushes and pops) and stack memory references 7015and then tries to find ways to combine them. 7016 7017Enabled by default at @option{-O1} and higher. 7018 7019@item -fconserve-stack 7020@opindex fconserve-stack 7021Attempt to minimize stack usage. The compiler will attempt to use less 7022stack space, even if that makes the program slower. This option 7023implies setting the @option{large-stack-frame} parameter to 100 7024and the @option{large-stack-frame-growth} parameter to 400. 7025 7026@item -ftree-reassoc 7027@opindex ftree-reassoc 7028Perform reassociation on trees. This flag is enabled by default 7029at @option{-O} and higher. 7030 7031@item -ftree-pre 7032@opindex ftree-pre 7033Perform partial redundancy elimination (PRE) on trees. This flag is 7034enabled by default at @option{-O2} and @option{-O3}. 7035 7036@item -ftree-forwprop 7037@opindex ftree-forwprop 7038Perform forward propagation on trees. This flag is enabled by default 7039at @option{-O} and higher. 7040 7041@item -ftree-fre 7042@opindex ftree-fre 7043Perform full redundancy elimination (FRE) on trees. The difference 7044between FRE and PRE is that FRE only considers expressions 7045that are computed on all paths leading to the redundant computation. 7046This analysis is faster than PRE, though it exposes fewer redundancies. 7047This flag is enabled by default at @option{-O} and higher. 7048 7049@item -ftree-phiprop 7050@opindex ftree-phiprop 7051Perform hoisting of loads from conditional pointers on trees. This 7052pass is enabled by default at @option{-O} and higher. 7053 7054@item -ftree-copy-prop 7055@opindex ftree-copy-prop 7056Perform copy propagation on trees. This pass eliminates unnecessary 7057copy operations. This flag is enabled by default at @option{-O} and 7058higher. 7059 7060@item -fipa-pure-const 7061@opindex fipa-pure-const 7062Discover which functions are pure or constant. 7063Enabled by default at @option{-O} and higher. 7064 7065@item -fipa-reference 7066@opindex fipa-reference 7067Discover which static variables do not escape cannot escape the 7068compilation unit. 7069Enabled by default at @option{-O} and higher. 7070 7071@item -fipa-pta 7072@opindex fipa-pta 7073Perform interprocedural pointer analysis and interprocedural modification 7074and reference analysis. This option can cause excessive memory and 7075compile-time usage on large compilation units. It is not enabled by 7076default at any optimization level. 7077 7078@item -fipa-profile 7079@opindex fipa-profile 7080Perform interprocedural profile propagation. The functions called only from 7081cold functions are marked as cold. Also functions executed once (such as 7082@code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold 7083functions and loop less parts of functions executed once are then optimized for 7084size. 7085Enabled by default at @option{-O} and higher. 7086 7087@item -fipa-cp 7088@opindex fipa-cp 7089Perform interprocedural constant propagation. 7090This optimization analyzes the program to determine when values passed 7091to functions are constants and then optimizes accordingly. 7092This optimization can substantially increase performance 7093if the application has constants passed to functions. 7094This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}. 7095 7096@item -fipa-cp-clone 7097@opindex fipa-cp-clone 7098Perform function cloning to make interprocedural constant propagation stronger. 7099When enabled, interprocedural constant propagation will perform function cloning 7100when externally visible function can be called with constant arguments. 7101Because this optimization can create multiple copies of functions, 7102it may significantly increase code size 7103(see @option{--param ipcp-unit-growth=@var{value}}). 7104This flag is enabled by default at @option{-O3}. 7105 7106@item -fipa-matrix-reorg 7107@opindex fipa-matrix-reorg 7108Perform matrix flattening and transposing. 7109Matrix flattening tries to replace an @math{m}-dimensional matrix 7110with its equivalent @math{n}-dimensional matrix, where @math{n < m}. 7111This reduces the level of indirection needed for accessing the elements 7112of the matrix. The second optimization is matrix transposing, which 7113attempts to change the order of the matrix's dimensions in order to 7114improve cache locality. 7115Both optimizations need the @option{-fwhole-program} flag. 7116Transposing is enabled only if profiling information is available. 7117 7118@item -ftree-sink 7119@opindex ftree-sink 7120Perform forward store motion on trees. This flag is 7121enabled by default at @option{-O} and higher. 7122 7123@item -ftree-bit-ccp 7124@opindex ftree-bit-ccp 7125Perform sparse conditional bit constant propagation on trees and propagate 7126pointer alignment information. 7127This pass only operates on local scalar variables and is enabled by default 7128at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled. 7129 7130@item -ftree-ccp 7131@opindex ftree-ccp 7132Perform sparse conditional constant propagation (CCP) on trees. This 7133pass only operates on local scalar variables and is enabled by default 7134at @option{-O} and higher. 7135 7136@item -ftree-switch-conversion 7137Perform conversion of simple initializations in a switch to 7138initializations from a scalar array. This flag is enabled by default 7139at @option{-O2} and higher. 7140 7141@item -ftree-tail-merge 7142Look for identical code sequences. When found, replace one with a jump to the 7143other. This optimization is known as tail merging or cross jumping. This flag 7144is enabled by default at @option{-O2} and higher. The compilation time 7145in this pass can 7146be limited using @option{max-tail-merge-comparisons} parameter and 7147@option{max-tail-merge-iterations} parameter. 7148 7149@item -ftree-dce 7150@opindex ftree-dce 7151Perform dead code elimination (DCE) on trees. This flag is enabled by 7152default at @option{-O} and higher. 7153 7154@item -ftree-builtin-call-dce 7155@opindex ftree-builtin-call-dce 7156Perform conditional dead code elimination (DCE) for calls to builtin functions 7157that may set @code{errno} but are otherwise side-effect free. This flag is 7158enabled by default at @option{-O2} and higher if @option{-Os} is not also 7159specified. 7160 7161@item -ftree-dominator-opts 7162@opindex ftree-dominator-opts 7163Perform a variety of simple scalar cleanups (constant/copy 7164propagation, redundancy elimination, range propagation and expression 7165simplification) based on a dominator tree traversal. This also 7166performs jump threading (to reduce jumps to jumps). This flag is 7167enabled by default at @option{-O} and higher. 7168 7169@item -ftree-dse 7170@opindex ftree-dse 7171Perform dead store elimination (DSE) on trees. A dead store is a store into 7172a memory location that is later overwritten by another store without 7173any intervening loads. In this case the earlier store can be deleted. This 7174flag is enabled by default at @option{-O} and higher. 7175 7176@item -ftree-ch 7177@opindex ftree-ch 7178Perform loop header copying on trees. This is beneficial since it increases 7179effectiveness of code motion optimizations. It also saves one jump. This flag 7180is enabled by default at @option{-O} and higher. It is not enabled 7181for @option{-Os}, since it usually increases code size. 7182 7183@item -ftree-loop-optimize 7184@opindex ftree-loop-optimize 7185Perform loop optimizations on trees. This flag is enabled by default 7186at @option{-O} and higher. 7187 7188@item -ftree-loop-linear 7189@opindex ftree-loop-linear 7190Perform loop interchange transformations on tree. Same as 7191@option{-floop-interchange}. To use this code transformation, GCC has 7192to be configured with @option{--with-ppl} and @option{--with-cloog} to 7193enable the Graphite loop transformation infrastructure. 7194 7195@item -floop-interchange 7196@opindex floop-interchange 7197Perform loop interchange transformations on loops. Interchanging two 7198nested loops switches the inner and outer loops. For example, given a 7199loop like: 7200@smallexample 7201DO J = 1, M 7202 DO I = 1, N 7203 A(J, I) = A(J, I) * C 7204 ENDDO 7205ENDDO 7206@end smallexample 7207loop interchange will transform the loop as if the user had written: 7208@smallexample 7209DO I = 1, N 7210 DO J = 1, M 7211 A(J, I) = A(J, I) * C 7212 ENDDO 7213ENDDO 7214@end smallexample 7215which can be beneficial when @code{N} is larger than the caches, 7216because in Fortran, the elements of an array are stored in memory 7217contiguously by column, and the original loop iterates over rows, 7218potentially creating at each access a cache miss. This optimization 7219applies to all the languages supported by GCC and is not limited to 7220Fortran. To use this code transformation, GCC has to be configured 7221with @option{--with-ppl} and @option{--with-cloog} to enable the 7222Graphite loop transformation infrastructure. 7223 7224@item -floop-strip-mine 7225@opindex floop-strip-mine 7226Perform loop strip mining transformations on loops. Strip mining 7227splits a loop into two nested loops. The outer loop has strides 7228equal to the strip size and the inner loop has strides of the 7229original loop within a strip. The strip length can be changed 7230using the @option{loop-block-tile-size} parameter. For example, 7231given a loop like: 7232@smallexample 7233DO I = 1, N 7234 A(I) = A(I) + C 7235ENDDO 7236@end smallexample 7237loop strip mining will transform the loop as if the user had written: 7238@smallexample 7239DO II = 1, N, 51 7240 DO I = II, min (II + 50, N) 7241 A(I) = A(I) + C 7242 ENDDO 7243ENDDO 7244@end smallexample 7245This optimization applies to all the languages supported by GCC and is 7246not limited to Fortran. To use this code transformation, GCC has to 7247be configured with @option{--with-ppl} and @option{--with-cloog} to 7248enable the Graphite loop transformation infrastructure. 7249 7250@item -floop-block 7251@opindex floop-block 7252Perform loop blocking transformations on loops. Blocking strip mines 7253each loop in the loop nest such that the memory accesses of the 7254element loops fit inside caches. The strip length can be changed 7255using the @option{loop-block-tile-size} parameter. For example, given 7256a loop like: 7257@smallexample 7258DO I = 1, N 7259 DO J = 1, M 7260 A(J, I) = B(I) + C(J) 7261 ENDDO 7262ENDDO 7263@end smallexample 7264loop blocking will transform the loop as if the user had written: 7265@smallexample 7266DO II = 1, N, 51 7267 DO JJ = 1, M, 51 7268 DO I = II, min (II + 50, N) 7269 DO J = JJ, min (JJ + 50, M) 7270 A(J, I) = B(I) + C(J) 7271 ENDDO 7272 ENDDO 7273 ENDDO 7274ENDDO 7275@end smallexample 7276which can be beneficial when @code{M} is larger than the caches, 7277because the innermost loop will iterate over a smaller amount of data 7278which can be kept in the caches. This optimization applies to all the 7279languages supported by GCC and is not limited to Fortran. To use this 7280code transformation, GCC has to be configured with @option{--with-ppl} 7281and @option{--with-cloog} to enable the Graphite loop transformation 7282infrastructure. 7283 7284@item -fgraphite-identity 7285@opindex fgraphite-identity 7286Enable the identity transformation for graphite. For every SCoP we generate 7287the polyhedral representation and transform it back to gimple. Using 7288@option{-fgraphite-identity} we can check the costs or benefits of the 7289GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations 7290are also performed by the code generator CLooG, like index splitting and 7291dead code elimination in loops. 7292 7293@item -floop-flatten 7294@opindex floop-flatten 7295Removes the loop nesting structure: transforms the loop nest into a 7296single loop. This transformation can be useful as an enablement 7297transform for vectorization and parallelization. This feature 7298is experimental. 7299To use this code transformation, GCC has to be configured 7300with @option{--with-ppl} and @option{--with-cloog} to enable the 7301Graphite loop transformation infrastructure. 7302 7303@item -floop-parallelize-all 7304@opindex floop-parallelize-all 7305Use the Graphite data dependence analysis to identify loops that can 7306be parallelized. Parallelize all the loops that can be analyzed to 7307not contain loop carried dependences without checking that it is 7308profitable to parallelize the loops. 7309 7310@item -fcheck-data-deps 7311@opindex fcheck-data-deps 7312Compare the results of several data dependence analyzers. This option 7313is used for debugging the data dependence analyzers. 7314 7315@item -ftree-loop-if-convert 7316Attempt to transform conditional jumps in the innermost loops to 7317branch-less equivalents. The intent is to remove control-flow from 7318the innermost loops in order to improve the ability of the 7319vectorization pass to handle these loops. This is enabled by default 7320if vectorization is enabled. 7321 7322@item -ftree-loop-if-convert-stores 7323Attempt to also if-convert conditional jumps containing memory writes. 7324This transformation can be unsafe for multi-threaded programs as it 7325transforms conditional memory writes into unconditional memory writes. 7326For example, 7327@smallexample 7328for (i = 0; i < N; i++) 7329 if (cond) 7330 A[i] = expr; 7331@end smallexample 7332would be transformed to 7333@smallexample 7334for (i = 0; i < N; i++) 7335 A[i] = cond ? expr : A[i]; 7336@end smallexample 7337potentially producing data races. 7338 7339@item -ftree-loop-distribution 7340Perform loop distribution. This flag can improve cache performance on 7341big loop bodies and allow further loop optimizations, like 7342parallelization or vectorization, to take place. For example, the loop 7343@smallexample 7344DO I = 1, N 7345 A(I) = B(I) + C 7346 D(I) = E(I) * F 7347ENDDO 7348@end smallexample 7349is transformed to 7350@smallexample 7351DO I = 1, N 7352 A(I) = B(I) + C 7353ENDDO 7354DO I = 1, N 7355 D(I) = E(I) * F 7356ENDDO 7357@end smallexample 7358 7359@item -ftree-loop-distribute-patterns 7360Perform loop distribution of patterns that can be code generated with 7361calls to a library. This flag is enabled by default at @option{-O3}. 7362 7363This pass distributes the initialization loops and generates a call to 7364memset zero. For example, the loop 7365@smallexample 7366DO I = 1, N 7367 A(I) = 0 7368 B(I) = A(I) + I 7369ENDDO 7370@end smallexample 7371is transformed to 7372@smallexample 7373DO I = 1, N 7374 A(I) = 0 7375ENDDO 7376DO I = 1, N 7377 B(I) = A(I) + I 7378ENDDO 7379@end smallexample 7380and the initialization loop is transformed into a call to memset zero. 7381 7382@item -ftree-loop-im 7383@opindex ftree-loop-im 7384Perform loop invariant motion on trees. This pass moves only invariants that 7385would be hard to handle at RTL level (function calls, operations that expand to 7386nontrivial sequences of insns). With @option{-funswitch-loops} it also moves 7387operands of conditions that are invariant out of the loop, so that we can use 7388just trivial invariantness analysis in loop unswitching. The pass also includes 7389store motion. 7390 7391@item -ftree-loop-ivcanon 7392@opindex ftree-loop-ivcanon 7393Create a canonical counter for number of iterations in loops for which 7394determining number of iterations requires complicated analysis. Later 7395optimizations then may determine the number easily. Useful especially 7396in connection with unrolling. 7397 7398@item -fivopts 7399@opindex fivopts 7400Perform induction variable optimizations (strength reduction, induction 7401variable merging and induction variable elimination) on trees. 7402 7403@item -ftree-parallelize-loops=n 7404@opindex ftree-parallelize-loops 7405Parallelize loops, i.e., split their iteration space to run in n threads. 7406This is only possible for loops whose iterations are independent 7407and can be arbitrarily reordered. The optimization is only 7408profitable on multiprocessor machines, for loops that are CPU-intensive, 7409rather than constrained e.g.@: by memory bandwidth. This option 7410implies @option{-pthread}, and thus is only supported on targets 7411that have support for @option{-pthread}. 7412 7413@item -ftree-pta 7414@opindex ftree-pta 7415Perform function-local points-to analysis on trees. This flag is 7416enabled by default at @option{-O} and higher. 7417 7418@item -ftree-sra 7419@opindex ftree-sra 7420Perform scalar replacement of aggregates. This pass replaces structure 7421references with scalars to prevent committing structures to memory too 7422early. This flag is enabled by default at @option{-O} and higher. 7423 7424@item -ftree-copyrename 7425@opindex ftree-copyrename 7426Perform copy renaming on trees. This pass attempts to rename compiler 7427temporaries to other variables at copy locations, usually resulting in 7428variable names which more closely resemble the original variables. This flag 7429is enabled by default at @option{-O} and higher. 7430 7431@item -ftree-ter 7432@opindex ftree-ter 7433Perform temporary expression replacement during the SSA->normal phase. Single 7434use/single def temporaries are replaced at their use location with their 7435defining expression. This results in non-GIMPLE code, but gives the expanders 7436much more complex trees to work on resulting in better RTL generation. This is 7437enabled by default at @option{-O} and higher. 7438 7439@item -ftree-vectorize 7440@opindex ftree-vectorize 7441Perform loop vectorization on trees. This flag is enabled by default at 7442@option{-O3}. 7443 7444@item -ftree-slp-vectorize 7445@opindex ftree-slp-vectorize 7446Perform basic block vectorization on trees. This flag is enabled by default at 7447@option{-O3} and when @option{-ftree-vectorize} is enabled. 7448 7449@item -ftree-vect-loop-version 7450@opindex ftree-vect-loop-version 7451Perform loop versioning when doing loop vectorization on trees. When a loop 7452appears to be vectorizable except that data alignment or data dependence cannot 7453be determined at compile time, then vectorized and non-vectorized versions of 7454the loop are generated along with run-time checks for alignment or dependence 7455to control which version is executed. This option is enabled by default 7456except at level @option{-Os} where it is disabled. 7457 7458@item -fvect-cost-model 7459@opindex fvect-cost-model 7460Enable cost model for vectorization. 7461 7462@item -ftree-vrp 7463@opindex ftree-vrp 7464Perform Value Range Propagation on trees. This is similar to the 7465constant propagation pass, but instead of values, ranges of values are 7466propagated. This allows the optimizers to remove unnecessary range 7467checks like array bound checks and null pointer checks. This is 7468enabled by default at @option{-O2} and higher. Null pointer check 7469elimination is only done if @option{-fdelete-null-pointer-checks} is 7470enabled. 7471 7472@item -ftracer 7473@opindex ftracer 7474Perform tail duplication to enlarge superblock size. This transformation 7475simplifies the control flow of the function allowing other optimizations to do 7476better job. 7477 7478@item -funroll-loops 7479@opindex funroll-loops 7480Unroll loops whose number of iterations can be determined at compile 7481time or upon entry to the loop. @option{-funroll-loops} implies 7482@option{-frerun-cse-after-loop}. This option makes code larger, 7483and may or may not make it run faster. 7484 7485@item -funroll-all-loops 7486@opindex funroll-all-loops 7487Unroll all loops, even if their number of iterations is uncertain when 7488the loop is entered. This usually makes programs run more slowly. 7489@option{-funroll-all-loops} implies the same options as 7490@option{-funroll-loops}, 7491 7492@item -fsplit-ivs-in-unroller 7493@opindex fsplit-ivs-in-unroller 7494Enables expressing of values of induction variables in later iterations 7495of the unrolled loop using the value in the first iteration. This breaks 7496long dependency chains, thus improving efficiency of the scheduling passes. 7497 7498Combination of @option{-fweb} and CSE is often sufficient to obtain the 7499same effect. However in cases the loop body is more complicated than 7500a single basic block, this is not reliable. It also does not work at all 7501on some of the architectures due to restrictions in the CSE pass. 7502 7503This optimization is enabled by default. 7504 7505@item -fvariable-expansion-in-unroller 7506@opindex fvariable-expansion-in-unroller 7507With this option, the compiler will create multiple copies of some 7508local variables when unrolling a loop which can result in superior code. 7509 7510@item -fpartial-inlining 7511@opindex fpartial-inlining 7512Inline parts of functions. This option has any effect only 7513when inlining itself is turned on by the @option{-finline-functions} 7514or @option{-finline-small-functions} options. 7515 7516Enabled at level @option{-O2}. 7517 7518@item -fpredictive-commoning 7519@opindex fpredictive-commoning 7520Perform predictive commoning optimization, i.e., reusing computations 7521(especially memory loads and stores) performed in previous 7522iterations of loops. 7523 7524This option is enabled at level @option{-O3}. 7525 7526@item -fprefetch-loop-arrays 7527@opindex fprefetch-loop-arrays 7528If supported by the target machine, generate instructions to prefetch 7529memory to improve the performance of loops that access large arrays. 7530 7531This option may generate better or worse code; results are highly 7532dependent on the structure of loops within the source code. 7533 7534Disabled at level @option{-Os}. 7535 7536@item -fno-peephole 7537@itemx -fno-peephole2 7538@opindex fno-peephole 7539@opindex fno-peephole2 7540Disable any machine-specific peephole optimizations. The difference 7541between @option{-fno-peephole} and @option{-fno-peephole2} is in how they 7542are implemented in the compiler; some targets use one, some use the 7543other, a few use both. 7544 7545@option{-fpeephole} is enabled by default. 7546@option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 7547 7548@item -fno-guess-branch-probability 7549@opindex fno-guess-branch-probability 7550Do not guess branch probabilities using heuristics. 7551 7552GCC will use heuristics to guess branch probabilities if they are 7553not provided by profiling feedback (@option{-fprofile-arcs}). These 7554heuristics are based on the control flow graph. If some branch probabilities 7555are specified by @samp{__builtin_expect}, then the heuristics will be 7556used to guess branch probabilities for the rest of the control flow graph, 7557taking the @samp{__builtin_expect} info into account. The interactions 7558between the heuristics and @samp{__builtin_expect} can be complex, and in 7559some cases, it may be useful to disable the heuristics so that the effects 7560of @samp{__builtin_expect} are easier to understand. 7561 7562The default is @option{-fguess-branch-probability} at levels 7563@option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 7564 7565@item -freorder-blocks 7566@opindex freorder-blocks 7567Reorder basic blocks in the compiled function in order to reduce number of 7568taken branches and improve code locality. 7569 7570Enabled at levels @option{-O2}, @option{-O3}. 7571 7572@item -freorder-blocks-and-partition 7573@opindex freorder-blocks-and-partition 7574In addition to reordering basic blocks in the compiled function, in order 7575to reduce number of taken branches, partitions hot and cold basic blocks 7576into separate sections of the assembly and .o files, to improve 7577paging and cache locality performance. 7578 7579This optimization is automatically turned off in the presence of 7580exception handling, for linkonce sections, for functions with a user-defined 7581section attribute and on any architecture that does not support named 7582sections. 7583 7584@item -freorder-functions 7585@opindex freorder-functions 7586Reorder functions in the object file in order to 7587improve code locality. This is implemented by using special 7588subsections @code{.text.hot} for most frequently executed functions and 7589@code{.text.unlikely} for unlikely executed functions. Reordering is done by 7590the linker so object file format must support named sections and linker must 7591place them in a reasonable way. 7592 7593Also profile feedback must be available in to make this option effective. See 7594@option{-fprofile-arcs} for details. 7595 7596Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 7597 7598@item -fstrict-aliasing 7599@opindex fstrict-aliasing 7600Allow the compiler to assume the strictest aliasing rules applicable to 7601the language being compiled. For C (and C++), this activates 7602optimizations based on the type of expressions. In particular, an 7603object of one type is assumed never to reside at the same address as an 7604object of a different type, unless the types are almost the same. For 7605example, an @code{unsigned int} can alias an @code{int}, but not a 7606@code{void*} or a @code{double}. A character type may alias any other 7607type. 7608 7609@anchor{Type-punning}Pay special attention to code like this: 7610@smallexample 7611union a_union @{ 7612 int i; 7613 double d; 7614@}; 7615 7616int f() @{ 7617 union a_union t; 7618 t.d = 3.0; 7619 return t.i; 7620@} 7621@end smallexample 7622The practice of reading from a different union member than the one most 7623recently written to (called ``type-punning'') is common. Even with 7624@option{-fstrict-aliasing}, type-punning is allowed, provided the memory 7625is accessed through the union type. So, the code above will work as 7626expected. @xref{Structures unions enumerations and bit-fields 7627implementation}. However, this code might not: 7628@smallexample 7629int f() @{ 7630 union a_union t; 7631 int* ip; 7632 t.d = 3.0; 7633 ip = &t.i; 7634 return *ip; 7635@} 7636@end smallexample 7637 7638Similarly, access by taking the address, casting the resulting pointer 7639and dereferencing the result has undefined behavior, even if the cast 7640uses a union type, e.g.: 7641@smallexample 7642int f() @{ 7643 double d = 3.0; 7644 return ((union a_union *) &d)->i; 7645@} 7646@end smallexample 7647 7648The @option{-fstrict-aliasing} option is enabled at levels 7649@option{-O2}, @option{-O3}, @option{-Os}. 7650 7651@item -fstrict-overflow 7652@opindex fstrict-overflow 7653Allow the compiler to assume strict signed overflow rules, depending 7654on the language being compiled. For C (and C++) this means that 7655overflow when doing arithmetic with signed numbers is undefined, which 7656means that the compiler may assume that it will not happen. This 7657permits various optimizations. For example, the compiler will assume 7658that an expression like @code{i + 10 > i} will always be true for 7659signed @code{i}. This assumption is only valid if signed overflow is 7660undefined, as the expression is false if @code{i + 10} overflows when 7661using twos complement arithmetic. When this option is in effect any 7662attempt to determine whether an operation on signed numbers will 7663overflow must be written carefully to not actually involve overflow. 7664 7665This option also allows the compiler to assume strict pointer 7666semantics: given a pointer to an object, if adding an offset to that 7667pointer does not produce a pointer to the same object, the addition is 7668undefined. This permits the compiler to conclude that @code{p + u > 7669p} is always true for a pointer @code{p} and unsigned integer 7670@code{u}. This assumption is only valid because pointer wraparound is 7671undefined, as the expression is false if @code{p + u} overflows using 7672twos complement arithmetic. 7673 7674See also the @option{-fwrapv} option. Using @option{-fwrapv} means 7675that integer signed overflow is fully defined: it wraps. When 7676@option{-fwrapv} is used, there is no difference between 7677@option{-fstrict-overflow} and @option{-fno-strict-overflow} for 7678integers. With @option{-fwrapv} certain types of overflow are 7679permitted. For example, if the compiler gets an overflow when doing 7680arithmetic on constants, the overflowed value can still be used with 7681@option{-fwrapv}, but not otherwise. 7682 7683The @option{-fstrict-overflow} option is enabled at levels 7684@option{-O2}, @option{-O3}, @option{-Os}. 7685 7686@item -falign-functions 7687@itemx -falign-functions=@var{n} 7688@opindex falign-functions 7689Align the start of functions to the next power-of-two greater than 7690@var{n}, skipping up to @var{n} bytes. For instance, 7691@option{-falign-functions=32} aligns functions to the next 32-byte 7692boundary, but @option{-falign-functions=24} would align to the next 769332-byte boundary only if this can be done by skipping 23 bytes or less. 7694 7695@option{-fno-align-functions} and @option{-falign-functions=1} are 7696equivalent and mean that functions will not be aligned. 7697 7698Some assemblers only support this flag when @var{n} is a power of two; 7699in that case, it is rounded up. 7700 7701If @var{n} is not specified or is zero, use a machine-dependent default. 7702 7703Enabled at levels @option{-O2}, @option{-O3}. 7704 7705@item -falign-labels 7706@itemx -falign-labels=@var{n} 7707@opindex falign-labels 7708Align all branch targets to a power-of-two boundary, skipping up to 7709@var{n} bytes like @option{-falign-functions}. This option can easily 7710make code slower, because it must insert dummy operations for when the 7711branch target is reached in the usual flow of the code. 7712 7713@option{-fno-align-labels} and @option{-falign-labels=1} are 7714equivalent and mean that labels will not be aligned. 7715 7716If @option{-falign-loops} or @option{-falign-jumps} are applicable and 7717are greater than this value, then their values are used instead. 7718 7719If @var{n} is not specified or is zero, use a machine-dependent default 7720which is very likely to be @samp{1}, meaning no alignment. 7721 7722Enabled at levels @option{-O2}, @option{-O3}. 7723 7724@item -falign-loops 7725@itemx -falign-loops=@var{n} 7726@opindex falign-loops 7727Align loops to a power-of-two boundary, skipping up to @var{n} bytes 7728like @option{-falign-functions}. The hope is that the loop will be 7729executed many times, which will make up for any execution of the dummy 7730operations. 7731 7732@option{-fno-align-loops} and @option{-falign-loops=1} are 7733equivalent and mean that loops will not be aligned. 7734 7735If @var{n} is not specified or is zero, use a machine-dependent default. 7736 7737Enabled at levels @option{-O2}, @option{-O3}. 7738 7739@item -falign-jumps 7740@itemx -falign-jumps=@var{n} 7741@opindex falign-jumps 7742Align branch targets to a power-of-two boundary, for branch targets 7743where the targets can only be reached by jumping, skipping up to @var{n} 7744bytes like @option{-falign-functions}. In this case, no dummy operations 7745need be executed. 7746 7747@option{-fno-align-jumps} and @option{-falign-jumps=1} are 7748equivalent and mean that loops will not be aligned. 7749 7750If @var{n} is not specified or is zero, use a machine-dependent default. 7751 7752Enabled at levels @option{-O2}, @option{-O3}. 7753 7754@item -funit-at-a-time 7755@opindex funit-at-a-time 7756This option is left for compatibility reasons. @option{-funit-at-a-time} 7757has no effect, while @option{-fno-unit-at-a-time} implies 7758@option{-fno-toplevel-reorder} and @option{-fno-section-anchors}. 7759 7760Enabled by default. 7761 7762@item -fno-toplevel-reorder 7763@opindex fno-toplevel-reorder 7764Do not reorder top-level functions, variables, and @code{asm} 7765statements. Output them in the same order that they appear in the 7766input file. When this option is used, unreferenced static variables 7767will not be removed. This option is intended to support existing code 7768that relies on a particular ordering. For new code, it is better to 7769use attributes. 7770 7771Enabled at level @option{-O0}. When disabled explicitly, it also implies 7772@option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some 7773targets. 7774 7775@item -fweb 7776@opindex fweb 7777Constructs webs as commonly used for register allocation purposes and assign 7778each web individual pseudo register. This allows the register allocation pass 7779to operate on pseudos directly, but also strengthens several other optimization 7780passes, such as CSE, loop optimizer and trivial dead code remover. It can, 7781however, make debugging impossible, since variables will no longer stay in a 7782``home register''. 7783 7784Enabled by default with @option{-funroll-loops}. 7785 7786@item -fwhole-program 7787@opindex fwhole-program 7788Assume that the current compilation unit represents the whole program being 7789compiled. All public functions and variables with the exception of @code{main} 7790and those merged by attribute @code{externally_visible} become static functions 7791and 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. 7792While this option is equivalent to proper use of the @code{static} keyword for 7793programs consisting of a single file, in combination with option 7794@option{-flto} this flag can be used to 7795compile many smaller scale programs since the functions and variables become 7796local for the whole combined compilation unit, not for the single source file 7797itself. 7798 7799This option implies @option{-fwhole-file} for Fortran programs. 7800 7801@item -flto[=@var{n}] 7802@opindex flto 7803This option runs the standard link-time optimizer. When invoked 7804with source code, it generates GIMPLE (one of GCC's internal 7805representations) and writes it to special ELF sections in the object 7806file. When the object files are linked together, all the function 7807bodies are read from these ELF sections and instantiated as if they 7808had been part of the same translation unit. 7809 7810To use the link-time optimizer, @option{-flto} needs to be specified at 7811compile time and during the final link. For example: 7812 7813@smallexample 7814gcc -c -O2 -flto foo.c 7815gcc -c -O2 -flto bar.c 7816gcc -o myprog -flto -O2 foo.o bar.o 7817@end smallexample 7818 7819The first two invocations to GCC save a bytecode representation 7820of GIMPLE into special ELF sections inside @file{foo.o} and 7821@file{bar.o}. The final invocation reads the GIMPLE bytecode from 7822@file{foo.o} and @file{bar.o}, merges the two files into a single 7823internal image, and compiles the result as usual. Since both 7824@file{foo.o} and @file{bar.o} are merged into a single image, this 7825causes all the interprocedural analyses and optimizations in GCC to 7826work across the two files as if they were a single one. This means, 7827for example, that the inliner is able to inline functions in 7828@file{bar.o} into functions in @file{foo.o} and vice-versa. 7829 7830Another (simpler) way to enable link-time optimization is: 7831 7832@smallexample 7833gcc -o myprog -flto -O2 foo.c bar.c 7834@end smallexample 7835 7836The above generates bytecode for @file{foo.c} and @file{bar.c}, 7837merges them together into a single GIMPLE representation and optimizes 7838them as usual to produce @file{myprog}. 7839 7840The only important thing to keep in mind is that to enable link-time 7841optimizations the @option{-flto} flag needs to be passed to both the 7842compile and the link commands. 7843 7844To make whole program optimization effective, it is necessary to make 7845certain whole program assumptions. The compiler needs to know 7846what functions and variables can be accessed by libraries and runtime 7847outside of the link-time optimized unit. When supported by the linker, 7848the linker plugin (see @option{-fuse-linker-plugin}) passes information 7849to the compiler about used and externally visible symbols. When 7850the linker plugin is not available, @option{-fwhole-program} should be 7851used to allow the compiler to make these assumptions, which leads 7852to more aggressive optimization decisions. 7853 7854Note that when a file is compiled with @option{-flto}, the generated 7855object file is larger than a regular object file because it 7856contains GIMPLE bytecodes and the usual final code. This means that 7857object files with LTO information can be linked as normal object 7858files; if @option{-flto} is not passed to the linker, no 7859interprocedural optimizations are applied. 7860 7861Additionally, the optimization flags used to compile individual files 7862are not necessarily related to those used at link time. For instance, 7863 7864@smallexample 7865gcc -c -O0 -flto foo.c 7866gcc -c -O0 -flto bar.c 7867gcc -o myprog -flto -O3 foo.o bar.o 7868@end smallexample 7869 7870This produces individual object files with unoptimized assembler 7871code, but the resulting binary @file{myprog} is optimized at 7872@option{-O3}. If, instead, the final binary is generated without 7873@option{-flto}, then @file{myprog} is not optimized. 7874 7875When producing the final binary with @option{-flto}, GCC only 7876applies link-time optimizations to those files that contain bytecode. 7877Therefore, you can mix and match object files and libraries with 7878GIMPLE bytecodes and final object code. GCC automatically selects 7879which files to optimize in LTO mode and which files to link without 7880further processing. 7881 7882There are some code generation flags preserved by GCC when 7883generating bytecodes, as they need to be used during the final link 7884stage. Currently, the following options are saved into the GIMPLE 7885bytecode files: @option{-fPIC}, @option{-fcommon} and all the 7886@option{-m} target flags. 7887 7888At link time, these options are read in and reapplied. Note that the 7889current implementation makes no attempt to recognize conflicting 7890values for these options. If different files have conflicting option 7891values (e.g., one file is compiled with @option{-fPIC} and another 7892isn't), the compiler simply uses the last value read from the 7893bytecode files. It is recommended, then, that you compile all the files 7894participating in the same link with the same options. 7895 7896If LTO encounters objects with C linkage declared with incompatible 7897types in separate translation units to be linked together (undefined 7898behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be 7899issued. The behavior is still undefined at run time. 7900 7901Another feature of LTO is that it is possible to apply interprocedural 7902optimizations on files written in different languages. This requires 7903support in the language front end. Currently, the C, C++ and 7904Fortran front ends are capable of emitting GIMPLE bytecodes, so 7905something like this should work: 7906 7907@smallexample 7908gcc -c -flto foo.c 7909g++ -c -flto bar.cc 7910gfortran -c -flto baz.f90 7911g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran 7912@end smallexample 7913 7914Notice that the final link is done with @command{g++} to get the C++ 7915runtime libraries and @option{-lgfortran} is added to get the Fortran 7916runtime libraries. In general, when mixing languages in LTO mode, you 7917should use the same link command options as when mixing languages in a 7918regular (non-LTO) compilation; all you need to add is @option{-flto} to 7919all the compile and link commands. 7920 7921If object files containing GIMPLE bytecode are stored in a library archive, say 7922@file{libfoo.a}, it is possible to extract and use them in an LTO link if you 7923are using a linker with plugin support. To enable this feature, use 7924the flag @option{-fuse-linker-plugin} at link time: 7925 7926@smallexample 7927gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo 7928@end smallexample 7929 7930With the linker plugin enabled, the linker extracts the needed 7931GIMPLE files from @file{libfoo.a} and passes them on to the running GCC 7932to make them part of the aggregated GIMPLE image to be optimized. 7933 7934If you are not using a linker with plugin support and/or do not 7935enable the linker plugin, then the objects inside @file{libfoo.a} 7936are extracted and linked as usual, but they do not participate 7937in the LTO optimization process. 7938 7939Link-time optimizations do not require the presence of the whole program to 7940operate. If the program does not require any symbols to be exported, it is 7941possible to combine @option{-flto} and @option{-fwhole-program} to allow 7942the interprocedural optimizers to use more aggressive assumptions which may 7943lead to improved optimization opportunities. 7944Use of @option{-fwhole-program} is not needed when linker plugin is 7945active (see @option{-fuse-linker-plugin}). 7946 7947The current implementation of LTO makes no 7948attempt to generate bytecode that is portable between different 7949types of hosts. The bytecode files are versioned and there is a 7950strict version check, so bytecode files generated in one version of 7951GCC will not work with an older/newer version of GCC. 7952 7953Link-time optimization does not work well with generation of debugging 7954information. Combining @option{-flto} with 7955@option{-g} is currently experimental and expected to produce wrong 7956results. 7957 7958If you specify the optional @var{n}, the optimization and code 7959generation done at link time is executed in parallel using @var{n} 7960parallel jobs by utilizing an installed @command{make} program. The 7961environment variable @env{MAKE} may be used to override the program 7962used. The default value for @var{n} is 1. 7963 7964You can also specify @option{-flto=jobserver} to use GNU make's 7965job server mode to determine the number of parallel jobs. This 7966is useful when the Makefile calling GCC is already executing in parallel. 7967You must prepend a @samp{+} to the command recipe in the parent Makefile 7968for this to work. This option likely only works if @env{MAKE} is 7969GNU make. 7970 7971This option is disabled by default 7972 7973@item -flto-partition=@var{alg} 7974@opindex flto-partition 7975Specify the partitioning algorithm used by the link-time optimizer. 7976The value is either @code{1to1} to specify a partitioning mirroring 7977the original source files or @code{balanced} to specify partitioning 7978into equally sized chunks (whenever possible). Specifying @code{none} 7979as an algorithm disables partitioning and streaming completely. The 7980default value is @code{balanced}. 7981 7982@item -flto-compression-level=@var{n} 7983This option specifies the level of compression used for intermediate 7984language written to LTO object files, and is only meaningful in 7985conjunction with LTO mode (@option{-flto}). Valid 7986values are 0 (no compression) to 9 (maximum compression). Values 7987outside this range are clamped to either 0 or 9. If the option is not 7988given, a default balanced compression setting is used. 7989 7990@item -flto-report 7991Prints a report with internal details on the workings of the link-time 7992optimizer. The contents of this report vary from version to version. 7993It is meant to be useful to GCC developers when processing object 7994files in LTO mode (via @option{-flto}). 7995 7996Disabled by default. 7997 7998@item -fuse-linker-plugin 7999Enables the use of a linker plugin during link-time optimization. This 8000option relies on plugin support in the linker, which is available in gold 8001or in GNU ld 2.21 or newer. 8002 8003This option enables the extraction of object files with GIMPLE bytecode out 8004of library archives. This improves the quality of optimization by exposing 8005more code to the link-time optimizer. This information specifies what 8006symbols can be accessed externally (by non-LTO object or during dynamic 8007linking). Resulting code quality improvements on binaries (and shared 8008libraries that use hidden visibility) are similar to @code{-fwhole-program}. 8009See @option{-flto} for a description of the effect of this flag and how to 8010use it. 8011 8012This option is enabled by default when LTO support in GCC is enabled 8013and GCC was configured for use with 8014a linker supporting plugins (GNU ld 2.21 or newer or gold). 8015 8016@item -ffat-lto-objects 8017@opindex ffat-lto-objects 8018Fat LTO objects are object files that contain both the intermediate language 8019and the object code. This makes them usable for both LTO linking and normal 8020linking. This option is effective only when compiling with @option{-flto} 8021and is ignored at link time. 8022 8023@option{-fno-fat-lto-objects} improves compilation time over plain LTO, but 8024requires the complete toolchain to be aware of LTO. It requires a linker with 8025linker plugin support for basic functionality. Additionally, nm, ar and ranlib 8026need to support linker plugins to allow a full-featured build environment 8027(capable of building static libraries etc). 8028 8029The default is @option{-ffat-lto-objects} but this default is intended to 8030change in future releases when linker plugin enabled environments become more 8031common. 8032 8033@item -fcompare-elim 8034@opindex fcompare-elim 8035After register allocation and post-register allocation instruction splitting, 8036identify arithmetic instructions that compute processor flags similar to a 8037comparison operation based on that arithmetic. If possible, eliminate the 8038explicit comparison operation. 8039 8040This pass only applies to certain targets that cannot explicitly represent 8041the comparison operation before register allocation is complete. 8042 8043Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 8044 8045@item -fcprop-registers 8046@opindex fcprop-registers 8047After register allocation and post-register allocation instruction splitting, 8048we perform a copy-propagation pass to try to reduce scheduling dependencies 8049and occasionally eliminate the copy. 8050 8051Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 8052 8053@item -fprofile-correction 8054@opindex fprofile-correction 8055Profiles collected using an instrumented binary for multi-threaded programs may 8056be inconsistent due to missed counter updates. When this option is specified, 8057GCC will use heuristics to correct or smooth out such inconsistencies. By 8058default, GCC will emit an error message when an inconsistent profile is detected. 8059 8060@item -fprofile-dir=@var{path} 8061@opindex fprofile-dir 8062 8063Set the directory to search for the profile data files in to @var{path}. 8064This option affects only the profile data generated by 8065@option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs} 8066and used by @option{-fprofile-use} and @option{-fbranch-probabilities} 8067and its related options. Both absolute and relative paths can be used. 8068By default, GCC will use the current directory as @var{path}, thus the 8069profile data file will appear in the same directory as the object file. 8070 8071@item -fprofile-generate 8072@itemx -fprofile-generate=@var{path} 8073@opindex fprofile-generate 8074 8075Enable options usually used for instrumenting application to produce 8076profile useful for later recompilation with profile feedback based 8077optimization. You must use @option{-fprofile-generate} both when 8078compiling and when linking your program. 8079 8080The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}. 8081 8082If @var{path} is specified, GCC will look at the @var{path} to find 8083the profile feedback data files. See @option{-fprofile-dir}. 8084 8085@item -fprofile-use 8086@itemx -fprofile-use=@var{path} 8087@opindex fprofile-use 8088Enable profile feedback directed optimizations, and optimizations 8089generally profitable only with profile feedback available. 8090 8091The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt}, 8092@code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer} 8093 8094By default, GCC emits an error message if the feedback profiles do not 8095match the source code. This error can be turned into a warning by using 8096@option{-Wcoverage-mismatch}. Note this may result in poorly optimized 8097code. 8098 8099If @var{path} is specified, GCC will look at the @var{path} to find 8100the profile feedback data files. See @option{-fprofile-dir}. 8101@end table 8102 8103The following options control compiler behavior regarding floating-point 8104arithmetic. These options trade off between speed and 8105correctness. All must be specifically enabled. 8106 8107@table @gcctabopt 8108@item -ffloat-store 8109@opindex ffloat-store 8110Do not store floating-point variables in registers, and inhibit other 8111options that might change whether a floating-point value is taken from a 8112register or memory. 8113 8114@cindex floating-point precision 8115This option prevents undesirable excess precision on machines such as 8116the 68000 where the floating registers (of the 68881) keep more 8117precision than a @code{double} is supposed to have. Similarly for the 8118x86 architecture. For most programs, the excess precision does only 8119good, but a few programs rely on the precise definition of IEEE floating 8120point. Use @option{-ffloat-store} for such programs, after modifying 8121them to store all pertinent intermediate computations into variables. 8122 8123@item -fexcess-precision=@var{style} 8124@opindex fexcess-precision 8125This option allows further control over excess precision on machines 8126where floating-point registers have more precision than the IEEE 8127@code{float} and @code{double} types and the processor does not 8128support operations rounding to those types. By default, 8129@option{-fexcess-precision=fast} is in effect; this means that 8130operations are carried out in the precision of the registers and that 8131it is unpredictable when rounding to the types specified in the source 8132code takes place. When compiling C, if 8133@option{-fexcess-precision=standard} is specified then excess 8134precision will follow the rules specified in ISO C99; in particular, 8135both casts and assignments cause values to be rounded to their 8136semantic types (whereas @option{-ffloat-store} only affects 8137assignments). This option is enabled by default for C if a strict 8138conformance option such as @option{-std=c99} is used. 8139 8140@opindex mfpmath 8141@option{-fexcess-precision=standard} is not implemented for languages 8142other than C, and has no effect if 8143@option{-funsafe-math-optimizations} or @option{-ffast-math} is 8144specified. On the x86, it also has no effect if @option{-mfpmath=sse} 8145or @option{-mfpmath=sse+387} is specified; in the former case, IEEE 8146semantics apply without excess precision, and in the latter, rounding 8147is unpredictable. 8148 8149@item -ffast-math 8150@opindex ffast-math 8151Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, 8152@option{-ffinite-math-only}, @option{-fno-rounding-math}, 8153@option{-fno-signaling-nans} and @option{-fcx-limited-range}. 8154 8155This option causes the preprocessor macro @code{__FAST_MATH__} to be defined. 8156 8157This option is not turned on by any @option{-O} option besides 8158@option{-Ofast} since it can result in incorrect output for programs 8159that depend on an exact implementation of IEEE or ISO rules/specifications 8160for math functions. It may, however, yield faster code for programs 8161that do not require the guarantees of these specifications. 8162 8163@item -fno-math-errno 8164@opindex fno-math-errno 8165Do not set ERRNO after calling math functions that are executed 8166with a single instruction, e.g., sqrt. A program that relies on 8167IEEE exceptions for math error handling may want to use this flag 8168for speed while maintaining IEEE arithmetic compatibility. 8169 8170This option is not turned on by any @option{-O} option since 8171it can result in incorrect output for programs that depend on 8172an exact implementation of IEEE or ISO rules/specifications for 8173math functions. It may, however, yield faster code for programs 8174that do not require the guarantees of these specifications. 8175 8176The default is @option{-fmath-errno}. 8177 8178On Darwin systems, the math library never sets @code{errno}. There is 8179therefore no reason for the compiler to consider the possibility that 8180it might, and @option{-fno-math-errno} is the default. 8181 8182@item -funsafe-math-optimizations 8183@opindex funsafe-math-optimizations 8184 8185Allow optimizations for floating-point arithmetic that (a) assume 8186that arguments and results are valid and (b) may violate IEEE or 8187ANSI standards. When used at link-time, it may include libraries 8188or startup files that change the default FPU control word or other 8189similar optimizations. 8190 8191This option is not turned on by any @option{-O} option since 8192it can result in incorrect output for programs that depend on 8193an exact implementation of IEEE or ISO rules/specifications for 8194math functions. It may, however, yield faster code for programs 8195that do not require the guarantees of these specifications. 8196Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math}, 8197@option{-fassociative-math} and @option{-freciprocal-math}. 8198 8199The default is @option{-fno-unsafe-math-optimizations}. 8200 8201@item -fassociative-math 8202@opindex fassociative-math 8203 8204Allow re-association of operands in series of floating-point operations. 8205This violates the ISO C and C++ language standard by possibly changing 8206computation result. NOTE: re-ordering may change the sign of zero as 8207well as ignore NaNs and inhibit or create underflow or overflow (and 8208thus cannot be used on code that relies on rounding behavior like 8209@code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons 8210and thus may not be used when ordered comparisons are required. 8211This option requires that both @option{-fno-signed-zeros} and 8212@option{-fno-trapping-math} be in effect. Moreover, it doesn't make 8213much sense with @option{-frounding-math}. For Fortran the option 8214is automatically enabled when both @option{-fno-signed-zeros} and 8215@option{-fno-trapping-math} are in effect. 8216 8217The default is @option{-fno-associative-math}. 8218 8219@item -freciprocal-math 8220@opindex freciprocal-math 8221 8222Allow the reciprocal of a value to be used instead of dividing by 8223the value if this enables optimizations. For example @code{x / y} 8224can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)} 8225is subject to common subexpression elimination. Note that this loses 8226precision and increases the number of flops operating on the value. 8227 8228The default is @option{-fno-reciprocal-math}. 8229 8230@item -ffinite-math-only 8231@opindex ffinite-math-only 8232Allow optimizations for floating-point arithmetic that assume 8233that arguments and results are not NaNs or +-Infs. 8234 8235This option is not turned on by any @option{-O} option since 8236it can result in incorrect output for programs that depend on 8237an exact implementation of IEEE or ISO rules/specifications for 8238math functions. It may, however, yield faster code for programs 8239that do not require the guarantees of these specifications. 8240 8241The default is @option{-fno-finite-math-only}. 8242 8243@item -fno-signed-zeros 8244@opindex fno-signed-zeros 8245Allow optimizations for floating-point arithmetic that ignore the 8246signedness of zero. IEEE arithmetic specifies the behavior of 8247distinct +0.0 and @minus{}0.0 values, which then prohibits simplification 8248of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}). 8249This option implies that the sign of a zero result isn't significant. 8250 8251The default is @option{-fsigned-zeros}. 8252 8253@item -fno-trapping-math 8254@opindex fno-trapping-math 8255Compile code assuming that floating-point operations cannot generate 8256user-visible traps. These traps include division by zero, overflow, 8257underflow, inexact result and invalid operation. This option requires 8258that @option{-fno-signaling-nans} be in effect. Setting this option may 8259allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example. 8260 8261This option should never be turned on by any @option{-O} option since 8262it can result in incorrect output for programs that depend on 8263an exact implementation of IEEE or ISO rules/specifications for 8264math functions. 8265 8266The default is @option{-ftrapping-math}. 8267 8268@item -frounding-math 8269@opindex frounding-math 8270Disable transformations and optimizations that assume default floating-point 8271rounding behavior. This is round-to-zero for all floating point 8272to integer conversions, and round-to-nearest for all other arithmetic 8273truncations. This option should be specified for programs that change 8274the FP rounding mode dynamically, or that may be executed with a 8275non-default rounding mode. This option disables constant folding of 8276floating-point expressions at compile time (which may be affected by 8277rounding mode) and arithmetic transformations that are unsafe in the 8278presence of sign-dependent rounding modes. 8279 8280The default is @option{-fno-rounding-math}. 8281 8282This option is experimental and does not currently guarantee to 8283disable all GCC optimizations that are affected by rounding mode. 8284Future versions of GCC may provide finer control of this setting 8285using C99's @code{FENV_ACCESS} pragma. This command-line option 8286will be used to specify the default state for @code{FENV_ACCESS}. 8287 8288@item -fsignaling-nans 8289@opindex fsignaling-nans 8290Compile code assuming that IEEE signaling NaNs may generate user-visible 8291traps during floating-point operations. Setting this option disables 8292optimizations that may change the number of exceptions visible with 8293signaling NaNs. This option implies @option{-ftrapping-math}. 8294 8295This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to 8296be defined. 8297 8298The default is @option{-fno-signaling-nans}. 8299 8300This option is experimental and does not currently guarantee to 8301disable all GCC optimizations that affect signaling NaN behavior. 8302 8303@item -fsingle-precision-constant 8304@opindex fsingle-precision-constant 8305Treat floating-point constants as single precision instead of 8306implicitly converting them to double-precision constants. 8307 8308@item -fcx-limited-range 8309@opindex fcx-limited-range 8310When enabled, this option states that a range reduction step is not 8311needed when performing complex division. Also, there is no checking 8312whether the result of a complex multiplication or division is @code{NaN 8313+ I*NaN}, with an attempt to rescue the situation in that case. The 8314default is @option{-fno-cx-limited-range}, but is enabled by 8315@option{-ffast-math}. 8316 8317This option controls the default setting of the ISO C99 8318@code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to 8319all languages. 8320 8321@item -fcx-fortran-rules 8322@opindex fcx-fortran-rules 8323Complex multiplication and division follow Fortran rules. Range 8324reduction is done as part of complex division, but there is no checking 8325whether the result of a complex multiplication or division is @code{NaN 8326+ I*NaN}, with an attempt to rescue the situation in that case. 8327 8328The default is @option{-fno-cx-fortran-rules}. 8329 8330@end table 8331 8332The following options control optimizations that may improve 8333performance, but are not enabled by any @option{-O} options. This 8334section includes experimental options that may produce broken code. 8335 8336@table @gcctabopt 8337@item -fbranch-probabilities 8338@opindex fbranch-probabilities 8339After running a program compiled with @option{-fprofile-arcs} 8340(@pxref{Debugging Options,, Options for Debugging Your Program or 8341@command{gcc}}), you can compile it a second time using 8342@option{-fbranch-probabilities}, to improve optimizations based on 8343the number of times each branch was taken. When the program 8344compiled with @option{-fprofile-arcs} exits it saves arc execution 8345counts to a file called @file{@var{sourcename}.gcda} for each source 8346file. The information in this data file is very dependent on the 8347structure of the generated code, so you must use the same source code 8348and the same optimization options for both compilations. 8349 8350With @option{-fbranch-probabilities}, GCC puts a 8351@samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}. 8352These can be used to improve optimization. Currently, they are only 8353used in one place: in @file{reorg.c}, instead of guessing which path a 8354branch is most likely to take, the @samp{REG_BR_PROB} values are used to 8355exactly determine which path is taken more often. 8356 8357@item -fprofile-values 8358@opindex fprofile-values 8359If combined with @option{-fprofile-arcs}, it adds code so that some 8360data about values of expressions in the program is gathered. 8361 8362With @option{-fbranch-probabilities}, it reads back the data gathered 8363from profiling values of expressions for usage in optimizations. 8364 8365Enabled with @option{-fprofile-generate} and @option{-fprofile-use}. 8366 8367@item -fvpt 8368@opindex fvpt 8369If combined with @option{-fprofile-arcs}, it instructs the compiler to add 8370a code to gather information about values of expressions. 8371 8372With @option{-fbranch-probabilities}, it reads back the data gathered 8373and actually performs the optimizations based on them. 8374Currently the optimizations include specialization of division operation 8375using the knowledge about the value of the denominator. 8376 8377@item -frename-registers 8378@opindex frename-registers 8379Attempt to avoid false dependencies in scheduled code by making use 8380of registers left over after register allocation. This optimization 8381will most benefit processors with lots of registers. Depending on the 8382debug information format adopted by the target, however, it can 8383make debugging impossible, since variables will no longer stay in 8384a ``home register''. 8385 8386Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}. 8387 8388@item -ftracer 8389@opindex ftracer 8390Perform tail duplication to enlarge superblock size. This transformation 8391simplifies the control flow of the function allowing other optimizations to do 8392better job. 8393 8394Enabled with @option{-fprofile-use}. 8395 8396@item -funroll-loops 8397@opindex funroll-loops 8398Unroll loops whose number of iterations can be determined at compile time or 8399upon entry to the loop. @option{-funroll-loops} implies 8400@option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}. 8401It also turns on complete loop peeling (i.e.@: complete removal of loops with 8402small constant number of iterations). This option makes code larger, and may 8403or may not make it run faster. 8404 8405Enabled with @option{-fprofile-use}. 8406 8407@item -funroll-all-loops 8408@opindex funroll-all-loops 8409Unroll all loops, even if their number of iterations is uncertain when 8410the loop is entered. This usually makes programs run more slowly. 8411@option{-funroll-all-loops} implies the same options as 8412@option{-funroll-loops}. 8413 8414@item -fpeel-loops 8415@opindex fpeel-loops 8416Peels loops for which there is enough information that they do not 8417roll much (from profile feedback). It also turns on complete loop peeling 8418(i.e.@: complete removal of loops with small constant number of iterations). 8419 8420Enabled with @option{-fprofile-use}. 8421 8422@item -fmove-loop-invariants 8423@opindex fmove-loop-invariants 8424Enables the loop invariant motion pass in the RTL loop optimizer. Enabled 8425at level @option{-O1} 8426 8427@item -funswitch-loops 8428@opindex funswitch-loops 8429Move branches with loop invariant conditions out of the loop, with duplicates 8430of the loop on both branches (modified according to result of the condition). 8431 8432@item -ffunction-sections 8433@itemx -fdata-sections 8434@opindex ffunction-sections 8435@opindex fdata-sections 8436Place each function or data item into its own section in the output 8437file if the target supports arbitrary sections. The name of the 8438function or the name of the data item determines the section's name 8439in the output file. 8440 8441Use these options on systems where the linker can perform optimizations 8442to improve locality of reference in the instruction space. Most systems 8443using the ELF object format and SPARC processors running Solaris 2 have 8444linkers with such optimizations. AIX may have these optimizations in 8445the future. 8446 8447Only use these options when there are significant benefits from doing 8448so. When you specify these options, the assembler and linker will 8449create larger object and executable files and will also be slower. 8450You will not be able to use @code{gprof} on all systems if you 8451specify this option and you may have problems with debugging if 8452you specify both this option and @option{-g}. 8453 8454@item -fbranch-target-load-optimize 8455@opindex fbranch-target-load-optimize 8456Perform branch target register load optimization before prologue / epilogue 8457threading. 8458The use of target registers can typically be exposed only during reload, 8459thus hoisting loads out of loops and doing inter-block scheduling needs 8460a separate optimization pass. 8461 8462@item -fbranch-target-load-optimize2 8463@opindex fbranch-target-load-optimize2 8464Perform branch target register load optimization after prologue / epilogue 8465threading. 8466 8467@item -fbtr-bb-exclusive 8468@opindex fbtr-bb-exclusive 8469When performing branch target register load optimization, don't reuse 8470branch target registers in within any basic block. 8471 8472@item -fstack-protector 8473@opindex fstack-protector 8474Emit extra code to check for buffer overflows, such as stack smashing 8475attacks. This is done by adding a guard variable to functions with 8476vulnerable objects. This includes functions that call alloca, and 8477functions with buffers larger than 8 bytes. The guards are initialized 8478when a function is entered and then checked when the function exits. 8479If a guard check fails, an error message is printed and the program exits. 8480 8481@item -fstack-protector-all 8482@opindex fstack-protector-all 8483Like @option{-fstack-protector} except that all functions are protected. 8484 8485@item -fsection-anchors 8486@opindex fsection-anchors 8487Try to reduce the number of symbolic address calculations by using 8488shared ``anchor'' symbols to address nearby objects. This transformation 8489can help to reduce the number of GOT entries and GOT accesses on some 8490targets. 8491 8492For example, the implementation of the following function @code{foo}: 8493 8494@smallexample 8495static int a, b, c; 8496int foo (void) @{ return a + b + c; @} 8497@end smallexample 8498 8499would usually calculate the addresses of all three variables, but if you 8500compile it with @option{-fsection-anchors}, it will access the variables 8501from a common anchor point instead. The effect is similar to the 8502following pseudocode (which isn't valid C): 8503 8504@smallexample 8505int foo (void) 8506@{ 8507 register int *xr = &x; 8508 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x]; 8509@} 8510@end smallexample 8511 8512Not all targets support this option. 8513 8514@item --param @var{name}=@var{value} 8515@opindex param 8516In some places, GCC uses various constants to control the amount of 8517optimization that is done. For example, GCC will not inline functions 8518that contain more than a certain number of instructions. You can 8519control some of these constants on the command line using the 8520@option{--param} option. 8521 8522The names of specific parameters, and the meaning of the values, are 8523tied to the internals of the compiler, and are subject to change 8524without notice in future releases. 8525 8526In each case, the @var{value} is an integer. The allowable choices for 8527@var{name} are given in the following table: 8528 8529@table @gcctabopt 8530@item predictable-branch-outcome 8531When branch is predicted to be taken with probability lower than this threshold 8532(in percent), then it is considered well predictable. The default is 10. 8533 8534@item max-crossjump-edges 8535The maximum number of incoming edges to consider for crossjumping. 8536The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in 8537the number of edges incoming to each block. Increasing values mean 8538more aggressive optimization, making the compilation time increase with 8539probably small improvement in executable size. 8540 8541@item min-crossjump-insns 8542The minimum number of instructions that must be matched at the end 8543of two blocks before crossjumping will be performed on them. This 8544value is ignored in the case where all instructions in the block being 8545crossjumped from are matched. The default value is 5. 8546 8547@item max-grow-copy-bb-insns 8548The maximum code size expansion factor when copying basic blocks 8549instead of jumping. The expansion is relative to a jump instruction. 8550The default value is 8. 8551 8552@item max-goto-duplication-insns 8553The maximum number of instructions to duplicate to a block that jumps 8554to a computed goto. To avoid @math{O(N^2)} behavior in a number of 8555passes, GCC factors computed gotos early in the compilation process, 8556and unfactors them as late as possible. Only computed jumps at the 8557end of a basic blocks with no more than max-goto-duplication-insns are 8558unfactored. The default value is 8. 8559 8560@item max-delay-slot-insn-search 8561The maximum number of instructions to consider when looking for an 8562instruction to fill a delay slot. If more than this arbitrary number of 8563instructions is searched, the time savings from filling the delay slot 8564will be minimal so stop searching. Increasing values mean more 8565aggressive optimization, making the compilation time increase with probably 8566small improvement in execution time. 8567 8568@item max-delay-slot-live-search 8569When trying to fill delay slots, the maximum number of instructions to 8570consider when searching for a block with valid live register 8571information. Increasing this arbitrarily chosen value means more 8572aggressive optimization, increasing the compilation time. This parameter 8573should be removed when the delay slot code is rewritten to maintain the 8574control-flow graph. 8575 8576@item max-gcse-memory 8577The approximate maximum amount of memory that will be allocated in 8578order to perform the global common subexpression elimination 8579optimization. If more memory than specified is required, the 8580optimization will not be done. 8581 8582@item max-gcse-insertion-ratio 8583If the ratio of expression insertions to deletions is larger than this value 8584for any expression, then RTL PRE will insert or remove the expression and thus 8585leave partially redundant computations in the instruction stream. The default value is 20. 8586 8587@item max-pending-list-length 8588The maximum number of pending dependencies scheduling will allow 8589before flushing the current state and starting over. Large functions 8590with few branches or calls can create excessively large lists which 8591needlessly consume memory and resources. 8592 8593@item max-modulo-backtrack-attempts 8594The maximum number of backtrack attempts the scheduler should make 8595when modulo scheduling a loop. Larger values can exponentially increase 8596compilation time. 8597 8598@item max-inline-insns-single 8599Several parameters control the tree inliner used in gcc. 8600This number sets the maximum number of instructions (counted in GCC's 8601internal representation) in a single function that the tree inliner 8602will consider for inlining. This only affects functions declared 8603inline and methods implemented in a class declaration (C++). 8604The default value is 400. 8605 8606@item max-inline-insns-auto 8607When you use @option{-finline-functions} (included in @option{-O3}), 8608a lot of functions that would otherwise not be considered for inlining 8609by the compiler will be investigated. To those functions, a different 8610(more restrictive) limit compared to functions declared inline can 8611be applied. 8612The default value is 40. 8613 8614@item large-function-insns 8615The limit specifying really large functions. For functions larger than this 8616limit after inlining, inlining is constrained by 8617@option{--param large-function-growth}. This parameter is useful primarily 8618to avoid extreme compilation time caused by non-linear algorithms used by the 8619back end. 8620The default value is 2700. 8621 8622@item large-function-growth 8623Specifies maximal growth of large function caused by inlining in percents. 8624The default value is 100 which limits large function growth to 2.0 times 8625the original size. 8626 8627@item large-unit-insns 8628The limit specifying large translation unit. Growth caused by inlining of 8629units larger than this limit is limited by @option{--param inline-unit-growth}. 8630For small units this might be too tight (consider unit consisting of function A 8631that is inline and B that just calls A three time. If B is small relative to 8632A, the growth of unit is 300\% and yet such inlining is very sane. For very 8633large units consisting of small inlineable functions however the overall unit 8634growth limit is needed to avoid exponential explosion of code size. Thus for 8635smaller units, the size is increased to @option{--param large-unit-insns} 8636before applying @option{--param inline-unit-growth}. The default is 10000 8637 8638@item inline-unit-growth 8639Specifies maximal overall growth of the compilation unit caused by inlining. 8640The default value is 30 which limits unit growth to 1.3 times the original 8641size. 8642 8643@item ipcp-unit-growth 8644Specifies maximal overall growth of the compilation unit caused by 8645interprocedural constant propagation. The default value is 10 which limits 8646unit growth to 1.1 times the original size. 8647 8648@item large-stack-frame 8649The limit specifying large stack frames. While inlining the algorithm is trying 8650to not grow past this limit too much. Default value is 256 bytes. 8651 8652@item large-stack-frame-growth 8653Specifies maximal growth of large stack frames caused by inlining in percents. 8654The default value is 1000 which limits large stack frame growth to 11 times 8655the original size. 8656 8657@item max-inline-insns-recursive 8658@itemx max-inline-insns-recursive-auto 8659Specifies maximum number of instructions out-of-line copy of self recursive inline 8660function can grow into by performing recursive inlining. 8661 8662For functions declared inline @option{--param max-inline-insns-recursive} is 8663taken into account. For function not declared inline, recursive inlining 8664happens only when @option{-finline-functions} (included in @option{-O3}) is 8665enabled and @option{--param max-inline-insns-recursive-auto} is used. The 8666default value is 450. 8667 8668@item max-inline-recursive-depth 8669@itemx max-inline-recursive-depth-auto 8670Specifies maximum recursion depth used by the recursive inlining. 8671 8672For functions declared inline @option{--param max-inline-recursive-depth} is 8673taken into account. For function not declared inline, recursive inlining 8674happens only when @option{-finline-functions} (included in @option{-O3}) is 8675enabled and @option{--param max-inline-recursive-depth-auto} is used. The 8676default value is 8. 8677 8678@item min-inline-recursive-probability 8679Recursive inlining is profitable only for function having deep recursion 8680in average and can hurt for function having little recursion depth by 8681increasing the prologue size or complexity of function body to other 8682optimizers. 8683 8684When profile feedback is available (see @option{-fprofile-generate}) the actual 8685recursion depth can be guessed from probability that function will recurse via 8686given call expression. This parameter limits inlining only to call expression 8687whose probability exceeds given threshold (in percents). The default value is 868810. 8689 8690@item early-inlining-insns 8691Specify growth that early inliner can make. In effect it increases amount of 8692inlining for code having large abstraction penalty. The default value is 10. 8693 8694@item max-early-inliner-iterations 8695@itemx max-early-inliner-iterations 8696Limit of iterations of early inliner. This basically bounds number of nested 8697indirect calls early inliner can resolve. Deeper chains are still handled by 8698late inlining. 8699 8700@item comdat-sharing-probability 8701@itemx comdat-sharing-probability 8702Probability (in percent) that C++ inline function with comdat visibility 8703will be shared across multiple compilation units. The default value is 20. 8704 8705@item min-vect-loop-bound 8706The minimum number of iterations under which a loop will not get vectorized 8707when @option{-ftree-vectorize} is used. The number of iterations after 8708vectorization needs to be greater than the value specified by this option 8709to allow vectorization. The default value is 0. 8710 8711@item gcse-cost-distance-ratio 8712Scaling factor in calculation of maximum distance an expression 8713can be moved by GCSE optimizations. This is currently supported only in the 8714code hoisting pass. The bigger the ratio, the more aggressive code hoisting 8715will be with simple expressions, i.e., the expressions that have cost 8716less than @option{gcse-unrestricted-cost}. Specifying 0 will disable 8717hoisting of simple expressions. The default value is 10. 8718 8719@item gcse-unrestricted-cost 8720Cost, roughly measured as the cost of a single typical machine 8721instruction, at which GCSE optimizations will not constrain 8722the distance an expression can travel. This is currently 8723supported only in the code hoisting pass. The lesser the cost, 8724the more aggressive code hoisting will be. Specifying 0 will 8725allow all expressions to travel unrestricted distances. 8726The default value is 3. 8727 8728@item max-hoist-depth 8729The depth of search in the dominator tree for expressions to hoist. 8730This is used to avoid quadratic behavior in hoisting algorithm. 8731The value of 0 will avoid limiting the search, but may slow down compilation 8732of huge functions. The default value is 30. 8733 8734@item max-tail-merge-comparisons 8735The maximum amount of similar bbs to compare a bb with. This is used to 8736avoid quadratic behavior in tree tail merging. The default value is 10. 8737 8738@item max-tail-merge-iterations 8739The maximum amount of iterations of the pass over the function. This is used to 8740limit compilation time in tree tail merging. The default value is 2. 8741 8742@item max-unrolled-insns 8743The maximum number of instructions that a loop should have if that loop 8744is unrolled, and if the loop is unrolled, it determines how many times 8745the loop code is unrolled. 8746 8747@item max-average-unrolled-insns 8748The maximum number of instructions biased by probabilities of their execution 8749that a loop should have if that loop is unrolled, and if the loop is unrolled, 8750it determines how many times the loop code is unrolled. 8751 8752@item max-unroll-times 8753The maximum number of unrollings of a single loop. 8754 8755@item max-peeled-insns 8756The maximum number of instructions that a loop should have if that loop 8757is peeled, and if the loop is peeled, it determines how many times 8758the loop code is peeled. 8759 8760@item max-peel-times 8761The maximum number of peelings of a single loop. 8762 8763@item max-completely-peeled-insns 8764The maximum number of insns of a completely peeled loop. 8765 8766@item max-completely-peel-times 8767The maximum number of iterations of a loop to be suitable for complete peeling. 8768 8769@item max-completely-peel-loop-nest-depth 8770The maximum depth of a loop nest suitable for complete peeling. 8771 8772@item max-unswitch-insns 8773The maximum number of insns of an unswitched loop. 8774 8775@item max-unswitch-level 8776The maximum number of branches unswitched in a single loop. 8777 8778@item lim-expensive 8779The minimum cost of an expensive expression in the loop invariant motion. 8780 8781@item iv-consider-all-candidates-bound 8782Bound on number of candidates for induction variables below that 8783all candidates are considered for each use in induction variable 8784optimizations. Only the most relevant candidates are considered 8785if there are more candidates, to avoid quadratic time complexity. 8786 8787@item iv-max-considered-uses 8788The induction variable optimizations give up on loops that contain more 8789induction variable uses. 8790 8791@item iv-always-prune-cand-set-bound 8792If number of candidates in the set is smaller than this value, 8793we always try to remove unnecessary ivs from the set during its 8794optimization when a new iv is added to the set. 8795 8796@item scev-max-expr-size 8797Bound on size of expressions used in the scalar evolutions analyzer. 8798Large expressions slow the analyzer. 8799 8800@item scev-max-expr-complexity 8801Bound on the complexity of the expressions in the scalar evolutions analyzer. 8802Complex expressions slow the analyzer. 8803 8804@item omega-max-vars 8805The maximum number of variables in an Omega constraint system. 8806The default value is 128. 8807 8808@item omega-max-geqs 8809The maximum number of inequalities in an Omega constraint system. 8810The default value is 256. 8811 8812@item omega-max-eqs 8813The maximum number of equalities in an Omega constraint system. 8814The default value is 128. 8815 8816@item omega-max-wild-cards 8817The maximum number of wildcard variables that the Omega solver will 8818be able to insert. The default value is 18. 8819 8820@item omega-hash-table-size 8821The size of the hash table in the Omega solver. The default value is 8822550. 8823 8824@item omega-max-keys 8825The maximal number of keys used by the Omega solver. The default 8826value is 500. 8827 8828@item omega-eliminate-redundant-constraints 8829When set to 1, use expensive methods to eliminate all redundant 8830constraints. The default value is 0. 8831 8832@item vect-max-version-for-alignment-checks 8833The maximum number of run-time checks that can be performed when 8834doing loop versioning for alignment in the vectorizer. See option 8835ftree-vect-loop-version for more information. 8836 8837@item vect-max-version-for-alias-checks 8838The maximum number of run-time checks that can be performed when 8839doing loop versioning for alias in the vectorizer. See option 8840ftree-vect-loop-version for more information. 8841 8842@item max-iterations-to-track 8843 8844The maximum number of iterations of a loop the brute force algorithm 8845for analysis of # of iterations of the loop tries to evaluate. 8846 8847@item hot-bb-count-fraction 8848Select fraction of the maximal count of repetitions of basic block in program 8849given basic block needs to have to be considered hot. 8850 8851@item hot-bb-frequency-fraction 8852Select fraction of the entry block frequency of executions of basic block in 8853function given basic block needs to have to be considered hot. 8854 8855@item max-predicted-iterations 8856The maximum number of loop iterations we predict statically. This is useful 8857in cases where function contain single loop with known bound and other loop 8858with unknown. We predict the known number of iterations correctly, while 8859the unknown number of iterations average to roughly 10. This means that the 8860loop without bounds would appear artificially cold relative to the other one. 8861 8862@item align-threshold 8863 8864Select fraction of the maximal frequency of executions of basic block in 8865function given basic block will get aligned. 8866 8867@item align-loop-iterations 8868 8869A loop expected to iterate at lest the selected number of iterations will get 8870aligned. 8871 8872@item tracer-dynamic-coverage 8873@itemx tracer-dynamic-coverage-feedback 8874 8875This value is used to limit superblock formation once the given percentage of 8876executed instructions is covered. This limits unnecessary code size 8877expansion. 8878 8879The @option{tracer-dynamic-coverage-feedback} is used only when profile 8880feedback is available. The real profiles (as opposed to statically estimated 8881ones) are much less balanced allowing the threshold to be larger value. 8882 8883@item tracer-max-code-growth 8884Stop tail duplication once code growth has reached given percentage. This is 8885rather hokey argument, as most of the duplicates will be eliminated later in 8886cross jumping, so it may be set to much higher values than is the desired code 8887growth. 8888 8889@item tracer-min-branch-ratio 8890 8891Stop reverse growth when the reverse probability of best edge is less than this 8892threshold (in percent). 8893 8894@item tracer-min-branch-ratio 8895@itemx tracer-min-branch-ratio-feedback 8896 8897Stop forward growth if the best edge do have probability lower than this 8898threshold. 8899 8900Similarly to @option{tracer-dynamic-coverage} two values are present, one for 8901compilation for profile feedback and one for compilation without. The value 8902for compilation with profile feedback needs to be more conservative (higher) in 8903order to make tracer effective. 8904 8905@item max-cse-path-length 8906 8907Maximum number of basic blocks on path that cse considers. The default is 10. 8908 8909@item max-cse-insns 8910The maximum instructions CSE process before flushing. The default is 1000. 8911 8912@item ggc-min-expand 8913 8914GCC uses a garbage collector to manage its own memory allocation. This 8915parameter specifies the minimum percentage by which the garbage 8916collector's heap should be allowed to expand between collections. 8917Tuning this may improve compilation speed; it has no effect on code 8918generation. 8919 8920The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when 8921RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is 8922the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If 8923GCC is not able to calculate RAM on a particular platform, the lower 8924bound of 30% is used. Setting this parameter and 8925@option{ggc-min-heapsize} to zero causes a full collection to occur at 8926every opportunity. This is extremely slow, but can be useful for 8927debugging. 8928 8929@item ggc-min-heapsize 8930 8931Minimum size of the garbage collector's heap before it begins bothering 8932to collect garbage. The first collection occurs after the heap expands 8933by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again, 8934tuning this may improve compilation speed, and has no effect on code 8935generation. 8936 8937The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that 8938tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but 8939with a lower bound of 4096 (four megabytes) and an upper bound of 8940131072 (128 megabytes). If GCC is not able to calculate RAM on a 8941particular platform, the lower bound is used. Setting this parameter 8942very large effectively disables garbage collection. Setting this 8943parameter and @option{ggc-min-expand} to zero causes a full collection 8944to occur at every opportunity. 8945 8946@item max-reload-search-insns 8947The maximum number of instruction reload should look backward for equivalent 8948register. Increasing values mean more aggressive optimization, making the 8949compilation time increase with probably slightly better performance. 8950The default value is 100. 8951 8952@item max-cselib-memory-locations 8953The maximum number of memory locations cselib should take into account. 8954Increasing values mean more aggressive optimization, making the compilation time 8955increase with probably slightly better performance. The default value is 500. 8956 8957@item reorder-blocks-duplicate 8958@itemx reorder-blocks-duplicate-feedback 8959 8960Used by basic block reordering pass to decide whether to use unconditional 8961branch or duplicate the code on its destination. Code is duplicated when its 8962estimated size is smaller than this value multiplied by the estimated size of 8963unconditional jump in the hot spots of the program. 8964 8965The @option{reorder-block-duplicate-feedback} is used only when profile 8966feedback is available and may be set to higher values than 8967@option{reorder-block-duplicate} since information about the hot spots is more 8968accurate. 8969 8970@item max-sched-ready-insns 8971The maximum number of instructions ready to be issued the scheduler should 8972consider at any given time during the first scheduling pass. Increasing 8973values mean more thorough searches, making the compilation time increase 8974with probably little benefit. The default value is 100. 8975 8976@item max-sched-region-blocks 8977The maximum number of blocks in a region to be considered for 8978interblock scheduling. The default value is 10. 8979 8980@item max-pipeline-region-blocks 8981The maximum number of blocks in a region to be considered for 8982pipelining in the selective scheduler. The default value is 15. 8983 8984@item max-sched-region-insns 8985The maximum number of insns in a region to be considered for 8986interblock scheduling. The default value is 100. 8987 8988@item max-pipeline-region-insns 8989The maximum number of insns in a region to be considered for 8990pipelining in the selective scheduler. The default value is 200. 8991 8992@item min-spec-prob 8993The minimum probability (in percents) of reaching a source block 8994for interblock speculative scheduling. The default value is 40. 8995 8996@item max-sched-extend-regions-iters 8997The maximum number of iterations through CFG to extend regions. 89980 - disable region extension, 8999N - do at most N iterations. 9000The default value is 0. 9001 9002@item max-sched-insn-conflict-delay 9003The maximum conflict delay for an insn to be considered for speculative motion. 9004The default value is 3. 9005 9006@item sched-spec-prob-cutoff 9007The minimal probability of speculation success (in percents), so that 9008speculative insn will be scheduled. 9009The default value is 40. 9010 9011@item sched-mem-true-dep-cost 9012Minimal distance (in CPU cycles) between store and load targeting same 9013memory locations. The default value is 1. 9014 9015@item selsched-max-lookahead 9016The maximum size of the lookahead window of selective scheduling. It is a 9017depth of search for available instructions. 9018The default value is 50. 9019 9020@item selsched-max-sched-times 9021The maximum number of times that an instruction will be scheduled during 9022selective scheduling. This is the limit on the number of iterations 9023through which the instruction may be pipelined. The default value is 2. 9024 9025@item selsched-max-insns-to-rename 9026The maximum number of best instructions in the ready list that are considered 9027for renaming in the selective scheduler. The default value is 2. 9028 9029@item sms-min-sc 9030The minimum value of stage count that swing modulo scheduler will 9031generate. The default value is 2. 9032 9033@item max-last-value-rtl 9034The maximum size measured as number of RTLs that can be recorded in an expression 9035in combiner for a pseudo register as last known value of that register. The default 9036is 10000. 9037 9038@item integer-share-limit 9039Small integer constants can use a shared data structure, reducing the 9040compiler's memory usage and increasing its speed. This sets the maximum 9041value of a shared integer constant. The default value is 256. 9042 9043@item min-virtual-mappings 9044Specifies the minimum number of virtual mappings in the incremental 9045SSA updater that should be registered to trigger the virtual mappings 9046heuristic defined by virtual-mappings-ratio. The default value is 9047100. 9048 9049@item virtual-mappings-ratio 9050If the number of virtual mappings is virtual-mappings-ratio bigger 9051than the number of virtual symbols to be updated, then the incremental 9052SSA updater switches to a full update for those symbols. The default 9053ratio is 3. 9054 9055@item ssp-buffer-size 9056The minimum size of buffers (i.e.@: arrays) that will receive stack smashing 9057protection when @option{-fstack-protection} is used. 9058 9059@item max-jump-thread-duplication-stmts 9060Maximum number of statements allowed in a block that needs to be 9061duplicated when threading jumps. 9062 9063@item max-fields-for-field-sensitive 9064Maximum number of fields in a structure we will treat in 9065a field sensitive manner during pointer analysis. The default is zero 9066for -O0, and -O1 and 100 for -Os, -O2, and -O3. 9067 9068@item prefetch-latency 9069Estimate on average number of instructions that are executed before 9070prefetch finishes. The distance we prefetch ahead is proportional 9071to this constant. Increasing this number may also lead to less 9072streams being prefetched (see @option{simultaneous-prefetches}). 9073 9074@item simultaneous-prefetches 9075Maximum number of prefetches that can run at the same time. 9076 9077@item l1-cache-line-size 9078The size of cache line in L1 cache, in bytes. 9079 9080@item l1-cache-size 9081The size of L1 cache, in kilobytes. 9082 9083@item l2-cache-size 9084The size of L2 cache, in kilobytes. 9085 9086@item min-insn-to-prefetch-ratio 9087The minimum ratio between the number of instructions and the 9088number of prefetches to enable prefetching in a loop. 9089 9090@item prefetch-min-insn-to-mem-ratio 9091The minimum ratio between the number of instructions and the 9092number of memory references to enable prefetching in a loop. 9093 9094@item use-canonical-types 9095Whether the compiler should use the ``canonical'' type system. By 9096default, this should always be 1, which uses a more efficient internal 9097mechanism for comparing types in C++ and Objective-C++. However, if 9098bugs in the canonical type system are causing compilation failures, 9099set this value to 0 to disable canonical types. 9100 9101@item switch-conversion-max-branch-ratio 9102Switch initialization conversion will refuse to create arrays that are 9103bigger than @option{switch-conversion-max-branch-ratio} times the number of 9104branches in the switch. 9105 9106@item max-partial-antic-length 9107Maximum length of the partial antic set computed during the tree 9108partial redundancy elimination optimization (@option{-ftree-pre}) when 9109optimizing at @option{-O3} and above. For some sorts of source code 9110the enhanced partial redundancy elimination optimization can run away, 9111consuming all of the memory available on the host machine. This 9112parameter sets a limit on the length of the sets that are computed, 9113which prevents the runaway behavior. Setting a value of 0 for 9114this parameter will allow an unlimited set length. 9115 9116@item sccvn-max-scc-size 9117Maximum size of a strongly connected component (SCC) during SCCVN 9118processing. If this limit is hit, SCCVN processing for the whole 9119function will not be done and optimizations depending on it will 9120be disabled. The default maximum SCC size is 10000. 9121 9122@item ira-max-loops-num 9123IRA uses regional register allocation by default. If a function 9124contains more loops than the number given by this parameter, only at most 9125the given number of the most frequently-executed loops form regions 9126for regional register allocation. The default value of the 9127parameter is 100. 9128 9129@item ira-max-conflict-table-size 9130Although IRA uses a sophisticated algorithm to compress the conflict 9131table, the table can still require excessive amounts of memory for 9132huge functions. If the conflict table for a function could be more 9133than the size in MB given by this parameter, the register allocator 9134instead uses a faster, simpler, and lower-quality 9135algorithm that does not require building a pseudo-register conflict table. 9136The default value of the parameter is 2000. 9137 9138@item ira-loop-reserved-regs 9139IRA can be used to evaluate more accurate register pressure in loops 9140for decisions to move loop invariants (see @option{-O3}). The number 9141of available registers reserved for some other purposes is given 9142by this parameter. The default value of the parameter is 2, which is 9143the minimal number of registers needed by typical instructions. 9144This value is the best found from numerous experiments. 9145 9146@item loop-invariant-max-bbs-in-loop 9147Loop invariant motion can be very expensive, both in compilation time and 9148in amount of needed compile-time memory, with very large loops. Loops 9149with more basic blocks than this parameter won't have loop invariant 9150motion optimization performed on them. The default value of the 9151parameter is 1000 for -O1 and 10000 for -O2 and above. 9152 9153@item loop-max-datarefs-for-datadeps 9154Building data dapendencies is expensive for very large loops. This 9155parameter limits the number of data references in loops that are 9156considered for data dependence analysis. These large loops will not 9157be handled then by the optimizations using loop data dependencies. 9158The default value is 1000. 9159 9160@item max-vartrack-size 9161Sets a maximum number of hash table slots to use during variable 9162tracking dataflow analysis of any function. If this limit is exceeded 9163with variable tracking at assignments enabled, analysis for that 9164function is retried without it, after removing all debug insns from 9165the function. If the limit is exceeded even without debug insns, var 9166tracking analysis is completely disabled for the function. Setting 9167the parameter to zero makes it unlimited. 9168 9169@item max-vartrack-expr-depth 9170Sets a maximum number of recursion levels when attempting to map 9171variable names or debug temporaries to value expressions. This trades 9172compilation time for more complete debug information. If this is set too 9173low, value expressions that are available and could be represented in 9174debug information may end up not being used; setting this higher may 9175enable the compiler to find more complex debug expressions, but compile 9176time and memory use may grow. The default is 12. 9177 9178@item min-nondebug-insn-uid 9179Use uids starting at this parameter for nondebug insns. The range below 9180the parameter is reserved exclusively for debug insns created by 9181@option{-fvar-tracking-assignments}, but debug insns may get 9182(non-overlapping) uids above it if the reserved range is exhausted. 9183 9184@item ipa-sra-ptr-growth-factor 9185IPA-SRA will replace a pointer to an aggregate with one or more new 9186parameters only when their cumulative size is less or equal to 9187@option{ipa-sra-ptr-growth-factor} times the size of the original 9188pointer parameter. 9189 9190@item tm-max-aggregate-size 9191When making copies of thread-local variables in a transaction, this 9192parameter specifies the size in bytes after which variables will be 9193saved with the logging functions as opposed to save/restore code 9194sequence pairs. This option only applies when using 9195@option{-fgnu-tm}. 9196 9197@item graphite-max-nb-scop-params 9198To avoid exponential effects in the Graphite loop transforms, the 9199number of parameters in a Static Control Part (SCoP) is bounded. The 9200default value is 10 parameters. A variable whose value is unknown at 9201compilation time and defined outside a SCoP is a parameter of the SCoP. 9202 9203@item graphite-max-bbs-per-function 9204To avoid exponential effects in the detection of SCoPs, the size of 9205the functions analyzed by Graphite is bounded. The default value is 9206100 basic blocks. 9207 9208@item loop-block-tile-size 9209Loop blocking or strip mining transforms, enabled with 9210@option{-floop-block} or @option{-floop-strip-mine}, strip mine each 9211loop in the loop nest by a given number of iterations. The strip 9212length can be changed using the @option{loop-block-tile-size} 9213parameter. The default value is 51 iterations. 9214 9215@item ipa-cp-value-list-size 9216IPA-CP attempts to track all possible values and types passed to a function's 9217parameter in order to propagate them and perform devirtualization. 9218@option{ipa-cp-value-list-size} is the maximum number of values and types it 9219stores per one formal parameter of a function. 9220 9221@item lto-partitions 9222Specify desired number of partitions produced during WHOPR compilation. 9223The number of partitions should exceed the number of CPUs used for compilation. 9224The default value is 32. 9225 9226@item lto-minpartition 9227Size of minimal partition for WHOPR (in estimated instructions). 9228This prevents expenses of splitting very small programs into too many 9229partitions. 9230 9231@item cxx-max-namespaces-for-diagnostic-help 9232The maximum number of namespaces to consult for suggestions when C++ 9233name lookup fails for an identifier. The default is 1000. 9234 9235@item sink-frequency-threshold 9236The maximum relative execution frequency (in percents) of the target block 9237relative to a statement's original block to allow statement sinking of a 9238statement. Larger numbers result in more aggressive statement sinking. 9239The default value is 75. A small positive adjustment is applied for 9240statements with memory operands as those are even more profitable so sink. 9241 9242@item max-stores-to-sink 9243The maximum number of conditional stores paires that can be sunk. Set to 0 9244if either vectorization (@option{-ftree-vectorize}) or if-conversion 9245(@option{-ftree-loop-if-convert}) is disabled. The default is 2. 9246 9247@item allow-load-data-races 9248Allow optimizers to introduce new data races on loads. 9249Set to 1 to allow, otherwise to 0. This option is enabled by default 9250unless implicitly set by the @option{-fmemory-model=} option. 9251 9252@item allow-store-data-races 9253Allow optimizers to introduce new data races on stores. 9254Set to 1 to allow, otherwise to 0. This option is enabled by default 9255unless implicitly set by the @option{-fmemory-model=} option. 9256 9257@item allow-packed-load-data-races 9258Allow optimizers to introduce new data races on packed data loads. 9259Set to 1 to allow, otherwise to 0. This option is enabled by default 9260unless implicitly set by the @option{-fmemory-model=} option. 9261 9262@item allow-packed-store-data-races 9263Allow optimizers to introduce new data races on packed data stores. 9264Set to 1 to allow, otherwise to 0. This option is enabled by default 9265unless implicitly set by the @option{-fmemory-model=} option. 9266 9267@item case-values-threshold 9268The smallest number of different values for which it is best to use a 9269jump-table instead of a tree of conditional branches. If the value is 92700, use the default for the machine. The default is 0. 9271 9272@item tree-reassoc-width 9273Set the maximum number of instructions executed in parallel in 9274reassociated tree. This parameter overrides target dependent 9275heuristics used by default if has non zero value. 9276 9277@end table 9278@end table 9279 9280@node Preprocessor Options 9281@section Options Controlling the Preprocessor 9282@cindex preprocessor options 9283@cindex options, preprocessor 9284 9285These options control the C preprocessor, which is run on each C source 9286file before actual compilation. 9287 9288If you use the @option{-E} option, nothing is done except preprocessing. 9289Some of these options make sense only together with @option{-E} because 9290they cause the preprocessor output to be unsuitable for actual 9291compilation. 9292 9293@table @gcctabopt 9294@item -Wp,@var{option} 9295@opindex Wp 9296You can use @option{-Wp,@var{option}} to bypass the compiler driver 9297and pass @var{option} directly through to the preprocessor. If 9298@var{option} contains commas, it is split into multiple options at the 9299commas. However, many options are modified, translated or interpreted 9300by the compiler driver before being passed to the preprocessor, and 9301@option{-Wp} forcibly bypasses this phase. The preprocessor's direct 9302interface is undocumented and subject to change, so whenever possible 9303you should avoid using @option{-Wp} and let the driver handle the 9304options instead. 9305 9306@item -Xpreprocessor @var{option} 9307@opindex Xpreprocessor 9308Pass @var{option} as an option to the preprocessor. You can use this to 9309supply system-specific preprocessor options that GCC does not know how to 9310recognize. 9311 9312If you want to pass an option that takes an argument, you must use 9313@option{-Xpreprocessor} twice, once for the option and once for the argument. 9314@end table 9315 9316@include cppopts.texi 9317 9318@node Assembler Options 9319@section Passing Options to the Assembler 9320 9321@c prevent bad page break with this line 9322You can pass options to the assembler. 9323 9324@table @gcctabopt 9325@item -Wa,@var{option} 9326@opindex Wa 9327Pass @var{option} as an option to the assembler. If @var{option} 9328contains commas, it is split into multiple options at the commas. 9329 9330@item -Xassembler @var{option} 9331@opindex Xassembler 9332Pass @var{option} as an option to the assembler. You can use this to 9333supply system-specific assembler options that GCC does not know how to 9334recognize. 9335 9336If you want to pass an option that takes an argument, you must use 9337@option{-Xassembler} twice, once for the option and once for the argument. 9338 9339@end table 9340 9341@node Link Options 9342@section Options for Linking 9343@cindex link options 9344@cindex options, linking 9345 9346These options come into play when the compiler links object files into 9347an executable output file. They are meaningless if the compiler is 9348not doing a link step. 9349 9350@table @gcctabopt 9351@cindex file names 9352@item @var{object-file-name} 9353A file name that does not end in a special recognized suffix is 9354considered to name an object file or library. (Object files are 9355distinguished from libraries by the linker according to the file 9356contents.) If linking is done, these object files are used as input 9357to the linker. 9358 9359@item -c 9360@itemx -S 9361@itemx -E 9362@opindex c 9363@opindex S 9364@opindex E 9365If any of these options is used, then the linker is not run, and 9366object file names should not be used as arguments. @xref{Overall 9367Options}. 9368 9369@cindex Libraries 9370@item -l@var{library} 9371@itemx -l @var{library} 9372@opindex l 9373Search the library named @var{library} when linking. (The second 9374alternative with the library as a separate argument is only for 9375POSIX compliance and is not recommended.) 9376 9377It makes a difference where in the command you write this option; the 9378linker searches and processes libraries and object files in the order they 9379are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z} 9380after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers 9381to functions in @samp{z}, those functions may not be loaded. 9382 9383The linker searches a standard list of directories for the library, 9384which is actually a file named @file{lib@var{library}.a}. The linker 9385then uses this file as if it had been specified precisely by name. 9386 9387The directories searched include several standard system directories 9388plus any that you specify with @option{-L}. 9389 9390Normally the files found this way are library files---archive files 9391whose members are object files. The linker handles an archive file by 9392scanning through it for members which define symbols that have so far 9393been referenced but not defined. But if the file that is found is an 9394ordinary object file, it is linked in the usual fashion. The only 9395difference between using an @option{-l} option and specifying a file name 9396is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a} 9397and searches several directories. 9398 9399@item -lobjc 9400@opindex lobjc 9401You need this special case of the @option{-l} option in order to 9402link an Objective-C or Objective-C++ program. 9403 9404@item -nostartfiles 9405@opindex nostartfiles 9406Do not use the standard system startup files when linking. 9407The standard system libraries are used normally, unless @option{-nostdlib} 9408or @option{-nodefaultlibs} is used. 9409 9410@item -nodefaultlibs 9411@opindex nodefaultlibs 9412Do not use the standard system libraries when linking. 9413Only the libraries you specify will be passed to the linker, options 9414specifying linkage of the system libraries, such as @code{-static-libgcc} 9415or @code{-shared-libgcc}, will be ignored. 9416The standard startup files are used normally, unless @option{-nostartfiles} 9417is used. The compiler may generate calls to @code{memcmp}, 9418@code{memset}, @code{memcpy} and @code{memmove}. 9419These entries are usually resolved by entries in 9420libc. These entry points should be supplied through some other 9421mechanism when this option is specified. 9422 9423@item -nostdlib 9424@opindex nostdlib 9425Do not use the standard system startup files or libraries when linking. 9426No startup files and only the libraries you specify will be passed to 9427the linker, options specifying linkage of the system libraries, such as 9428@code{-static-libgcc} or @code{-shared-libgcc}, will be ignored. 9429The compiler may generate calls to @code{memcmp}, @code{memset}, 9430@code{memcpy} and @code{memmove}. 9431These entries are usually resolved by entries in 9432libc. These entry points should be supplied through some other 9433mechanism when this option is specified. 9434 9435@cindex @option{-lgcc}, use with @option{-nostdlib} 9436@cindex @option{-nostdlib} and unresolved references 9437@cindex unresolved references and @option{-nostdlib} 9438@cindex @option{-lgcc}, use with @option{-nodefaultlibs} 9439@cindex @option{-nodefaultlibs} and unresolved references 9440@cindex unresolved references and @option{-nodefaultlibs} 9441One of the standard libraries bypassed by @option{-nostdlib} and 9442@option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines 9443which GCC uses to overcome shortcomings of particular machines, or special 9444needs for some languages. 9445(@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler 9446Collection (GCC) Internals}, 9447for more discussion of @file{libgcc.a}.) 9448In most cases, you need @file{libgcc.a} even when you want to avoid 9449other standard libraries. In other words, when you specify @option{-nostdlib} 9450or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well. 9451This ensures that you have no unresolved references to internal GCC 9452library subroutines. (For example, @samp{__main}, used to ensure C++ 9453constructors will be called; @pxref{Collect2,,@code{collect2}, gccint, 9454GNU Compiler Collection (GCC) Internals}.) 9455 9456@item -pie 9457@opindex pie 9458Produce a position independent executable on targets that support it. 9459For predictable results, you must also specify the same set of options 9460that were used to generate code (@option{-fpie}, @option{-fPIE}, 9461or model suboptions) when you specify this option. 9462 9463@item -rdynamic 9464@opindex rdynamic 9465Pass the flag @option{-export-dynamic} to the ELF linker, on targets 9466that support it. This instructs the linker to add all symbols, not 9467only used ones, to the dynamic symbol table. This option is needed 9468for some uses of @code{dlopen} or to allow obtaining backtraces 9469from within a program. 9470 9471@item -s 9472@opindex s 9473Remove all symbol table and relocation information from the executable. 9474 9475@item -static 9476@opindex static 9477On systems that support dynamic linking, this prevents linking with the shared 9478libraries. On other systems, this option has no effect. 9479 9480@item -shared 9481@opindex shared 9482Produce a shared object which can then be linked with other objects to 9483form an executable. Not all systems support this option. For predictable 9484results, you must also specify the same set of options that were used to 9485generate code (@option{-fpic}, @option{-fPIC}, or model suboptions) 9486when you specify this option.@footnote{On some systems, @samp{gcc -shared} 9487needs to build supplementary stub code for constructors to work. On 9488multi-libbed systems, @samp{gcc -shared} must select the correct support 9489libraries to link against. Failing to supply the correct flags may lead 9490to subtle defects. Supplying them in cases where they are not necessary 9491is innocuous.} 9492 9493@item -shared-libgcc 9494@itemx -static-libgcc 9495@opindex shared-libgcc 9496@opindex static-libgcc 9497On systems that provide @file{libgcc} as a shared library, these options 9498force the use of either the shared or static version respectively. 9499If no shared version of @file{libgcc} was built when the compiler was 9500configured, these options have no effect. 9501 9502There are several situations in which an application should use the 9503shared @file{libgcc} instead of the static version. The most common 9504of these is when the application wishes to throw and catch exceptions 9505across different shared libraries. In that case, each of the libraries 9506as well as the application itself should use the shared @file{libgcc}. 9507 9508Therefore, the G++ and GCJ drivers automatically add 9509@option{-shared-libgcc} whenever you build a shared library or a main 9510executable, because C++ and Java programs typically use exceptions, so 9511this is the right thing to do. 9512 9513If, instead, you use the GCC driver to create shared libraries, you may 9514find that they will not always be linked with the shared @file{libgcc}. 9515If GCC finds, at its configuration time, that you have a non-GNU linker 9516or a GNU linker that does not support option @option{--eh-frame-hdr}, 9517it will link the shared version of @file{libgcc} into shared libraries 9518by default. Otherwise, it will take advantage of the linker and optimize 9519away the linking with the shared version of @file{libgcc}, linking with 9520the static version of libgcc by default. This allows exceptions to 9521propagate through such shared libraries, without incurring relocation 9522costs at library load time. 9523 9524However, if a library or main executable is supposed to throw or catch 9525exceptions, you must link it using the G++ or GCJ driver, as appropriate 9526for the languages used in the program, or using the option 9527@option{-shared-libgcc}, such that it is linked with the shared 9528@file{libgcc}. 9529 9530@item -static-libstdc++ 9531When the @command{g++} program is used to link a C++ program, it will 9532normally automatically link against @option{libstdc++}. If 9533@file{libstdc++} is available as a shared library, and the 9534@option{-static} option is not used, then this will link against the 9535shared version of @file{libstdc++}. That is normally fine. However, it 9536is sometimes useful to freeze the version of @file{libstdc++} used by 9537the program without going all the way to a fully static link. The 9538@option{-static-libstdc++} option directs the @command{g++} driver to 9539link @file{libstdc++} statically, without necessarily linking other 9540libraries statically. 9541 9542@item -symbolic 9543@opindex symbolic 9544Bind references to global symbols when building a shared object. Warn 9545about any unresolved references (unless overridden by the link editor 9546option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support 9547this option. 9548 9549@item -T @var{script} 9550@opindex T 9551@cindex linker script 9552Use @var{script} as the linker script. This option is supported by most 9553systems using the GNU linker. On some targets, such as bare-board 9554targets without an operating system, the @option{-T} option may be required 9555when linking to avoid references to undefined symbols. 9556 9557@item -Xlinker @var{option} 9558@opindex Xlinker 9559Pass @var{option} as an option to the linker. You can use this to 9560supply system-specific linker options that GCC does not recognize. 9561 9562If you want to pass an option that takes a separate argument, you must use 9563@option{-Xlinker} twice, once for the option and once for the argument. 9564For example, to pass @option{-assert definitions}, you must write 9565@samp{-Xlinker -assert -Xlinker definitions}. It does not work to write 9566@option{-Xlinker "-assert definitions"}, because this passes the entire 9567string as a single argument, which is not what the linker expects. 9568 9569When using the GNU linker, it is usually more convenient to pass 9570arguments to linker options using the @option{@var{option}=@var{value}} 9571syntax than as separate arguments. For example, you can specify 9572@samp{-Xlinker -Map=output.map} rather than 9573@samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support 9574this syntax for command-line options. 9575 9576@item -Wl,@var{option} 9577@opindex Wl 9578Pass @var{option} as an option to the linker. If @var{option} contains 9579commas, it is split into multiple options at the commas. You can use this 9580syntax to pass an argument to the option. 9581For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the 9582linker. When using the GNU linker, you can also get the same effect with 9583@samp{-Wl,-Map=output.map}. 9584 9585@item -u @var{symbol} 9586@opindex u 9587Pretend the symbol @var{symbol} is undefined, to force linking of 9588library modules to define it. You can use @option{-u} multiple times with 9589different symbols to force loading of additional library modules. 9590@end table 9591 9592@node Directory Options 9593@section Options for Directory Search 9594@cindex directory options 9595@cindex options, directory search 9596@cindex search path 9597 9598These options specify directories to search for header files, for 9599libraries and for parts of the compiler: 9600 9601@table @gcctabopt 9602@item -I@var{dir} 9603@opindex I 9604Add the directory @var{dir} to the head of the list of directories to be 9605searched for header files. This can be used to override a system header 9606file, substituting your own version, since these directories are 9607searched before the system header file directories. However, you should 9608not use this option to add directories that contain vendor-supplied 9609system header files (use @option{-isystem} for that). If you use more than 9610one @option{-I} option, the directories are scanned in left-to-right 9611order; the standard system directories come after. 9612 9613If a standard system include directory, or a directory specified with 9614@option{-isystem}, is also specified with @option{-I}, the @option{-I} 9615option will be ignored. The directory will still be searched but as a 9616system directory at its normal position in the system include chain. 9617This is to ensure that GCC's procedure to fix buggy system headers and 9618the ordering for the include_next directive are not inadvertently changed. 9619If you really need to change the search order for system directories, 9620use the @option{-nostdinc} and/or @option{-isystem} options. 9621 9622@item -iplugindir=@var{dir} 9623Set the directory to search for plugins that are passed 9624by @option{-fplugin=@var{name}} instead of 9625@option{-fplugin=@var{path}/@var{name}.so}. This option is not meant 9626to be used by the user, but only passed by the driver. 9627 9628@item -iquote@var{dir} 9629@opindex iquote 9630Add the directory @var{dir} to the head of the list of directories to 9631be searched for header files only for the case of @samp{#include 9632"@var{file}"}; they are not searched for @samp{#include <@var{file}>}, 9633otherwise just like @option{-I}. 9634 9635@item -L@var{dir} 9636@opindex L 9637Add directory @var{dir} to the list of directories to be searched 9638for @option{-l}. 9639 9640@item -B@var{prefix} 9641@opindex B 9642This option specifies where to find the executables, libraries, 9643include files, and data files of the compiler itself. 9644 9645The compiler driver program runs one or more of the subprograms 9646@file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries 9647@var{prefix} as a prefix for each program it tries to run, both with and 9648without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}). 9649 9650For each subprogram to be run, the compiler driver first tries the 9651@option{-B} prefix, if any. If that name is not found, or if @option{-B} 9652was not specified, the driver tries two standard prefixes, 9653@file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of 9654those results in a file name that is found, the unmodified program 9655name is searched for using the directories specified in your 9656@env{PATH} environment variable. 9657 9658The compiler will check to see if the path provided by the @option{-B} 9659refers to a directory, and if necessary it will add a directory 9660separator character at the end of the path. 9661 9662@option{-B} prefixes that effectively specify directory names also apply 9663to libraries in the linker, because the compiler translates these 9664options into @option{-L} options for the linker. They also apply to 9665includes files in the preprocessor, because the compiler translates these 9666options into @option{-isystem} options for the preprocessor. In this case, 9667the compiler appends @samp{include} to the prefix. 9668 9669The runtime support file @file{libgcc.a} can also be searched for using 9670the @option{-B} prefix, if needed. If it is not found there, the two 9671standard prefixes above are tried, and that is all. The file is left 9672out of the link if it is not found by those means. 9673 9674Another way to specify a prefix much like the @option{-B} prefix is to use 9675the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment 9676Variables}. 9677 9678As a special kludge, if the path provided by @option{-B} is 9679@file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to 96809, then it will be replaced by @file{[dir/]include}. This is to help 9681with boot-strapping the compiler. 9682 9683@item -specs=@var{file} 9684@opindex specs 9685Process @var{file} after the compiler reads in the standard @file{specs} 9686file, in order to override the defaults which the @file{gcc} driver 9687program uses when determining what switches to pass to @file{cc1}, 9688@file{cc1plus}, @file{as}, @file{ld}, etc. More than one 9689@option{-specs=@var{file}} can be specified on the command line, and they 9690are processed in order, from left to right. 9691 9692@item --sysroot=@var{dir} 9693@opindex sysroot 9694Use @var{dir} as the logical root directory for headers and libraries. 9695For example, if the compiler would normally search for headers in 9696@file{/usr/include} and libraries in @file{/usr/lib}, it will instead 9697search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}. 9698 9699If you use both this option and the @option{-isysroot} option, then 9700the @option{--sysroot} option will apply to libraries, but the 9701@option{-isysroot} option will apply to header files. 9702 9703The GNU linker (beginning with version 2.16) has the necessary support 9704for this option. If your linker does not support this option, the 9705header file aspect of @option{--sysroot} will still work, but the 9706library aspect will not. 9707 9708@item -I- 9709@opindex I- 9710This option has been deprecated. Please use @option{-iquote} instead for 9711@option{-I} directories before the @option{-I-} and remove the @option{-I-}. 9712Any directories you specify with @option{-I} options before the @option{-I-} 9713option are searched only for the case of @samp{#include "@var{file}"}; 9714they are not searched for @samp{#include <@var{file}>}. 9715 9716If additional directories are specified with @option{-I} options after 9717the @option{-I-}, these directories are searched for all @samp{#include} 9718directives. (Ordinarily @emph{all} @option{-I} directories are used 9719this way.) 9720 9721In addition, the @option{-I-} option inhibits the use of the current 9722directory (where the current input file came from) as the first search 9723directory for @samp{#include "@var{file}"}. There is no way to 9724override this effect of @option{-I-}. With @option{-I.} you can specify 9725searching the directory that was current when the compiler was 9726invoked. That is not exactly the same as what the preprocessor does 9727by default, but it is often satisfactory. 9728 9729@option{-I-} does not inhibit the use of the standard system directories 9730for header files. Thus, @option{-I-} and @option{-nostdinc} are 9731independent. 9732@end table 9733 9734@c man end 9735 9736@node Spec Files 9737@section Specifying subprocesses and the switches to pass to them 9738@cindex Spec Files 9739 9740@command{gcc} is a driver program. It performs its job by invoking a 9741sequence of other programs to do the work of compiling, assembling and 9742linking. GCC interprets its command-line parameters and uses these to 9743deduce which programs it should invoke, and which command-line options 9744it ought to place on their command lines. This behavior is controlled 9745by @dfn{spec strings}. In most cases there is one spec string for each 9746program that GCC can invoke, but a few programs have multiple spec 9747strings to control their behavior. The spec strings built into GCC can 9748be overridden by using the @option{-specs=} command-line switch to specify 9749a spec file. 9750 9751@dfn{Spec files} are plaintext files that are used to construct spec 9752strings. They consist of a sequence of directives separated by blank 9753lines. The type of directive is determined by the first non-whitespace 9754character on the line, which can be one of the following: 9755 9756@table @code 9757@item %@var{command} 9758Issues a @var{command} to the spec file processor. The commands that can 9759appear here are: 9760 9761@table @code 9762@item %include <@var{file}> 9763@cindex @code{%include} 9764Search for @var{file} and insert its text at the current point in the 9765specs file. 9766 9767@item %include_noerr <@var{file}> 9768@cindex @code{%include_noerr} 9769Just like @samp{%include}, but do not generate an error message if the include 9770file cannot be found. 9771 9772@item %rename @var{old_name} @var{new_name} 9773@cindex @code{%rename} 9774Rename the spec string @var{old_name} to @var{new_name}. 9775 9776@end table 9777 9778@item *[@var{spec_name}]: 9779This tells the compiler to create, override or delete the named spec 9780string. All lines after this directive up to the next directive or 9781blank line are considered to be the text for the spec string. If this 9782results in an empty string then the spec will be deleted. (Or, if the 9783spec did not exist, then nothing will happen.) Otherwise, if the spec 9784does not currently exist a new spec will be created. If the spec does 9785exist then its contents will be overridden by the text of this 9786directive, unless the first character of that text is the @samp{+} 9787character, in which case the text will be appended to the spec. 9788 9789@item [@var{suffix}]: 9790Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive 9791and up to the next directive or blank line are considered to make up the 9792spec string for the indicated suffix. When the compiler encounters an 9793input file with the named suffix, it will processes the spec string in 9794order to work out how to compile that file. For example: 9795 9796@smallexample 9797.ZZ: 9798z-compile -input %i 9799@end smallexample 9800 9801This says that any input file whose name ends in @samp{.ZZ} should be 9802passed to the program @samp{z-compile}, which should be invoked with the 9803command-line switch @option{-input} and with the result of performing the 9804@samp{%i} substitution. (See below.) 9805 9806As an alternative to providing a spec string, the text that follows a 9807suffix directive can be one of the following: 9808 9809@table @code 9810@item @@@var{language} 9811This says that the suffix is an alias for a known @var{language}. This is 9812similar to using the @option{-x} command-line switch to GCC to specify a 9813language explicitly. For example: 9814 9815@smallexample 9816.ZZ: 9817@@c++ 9818@end smallexample 9819 9820Says that .ZZ files are, in fact, C++ source files. 9821 9822@item #@var{name} 9823This causes an error messages saying: 9824 9825@smallexample 9826@var{name} compiler not installed on this system. 9827@end smallexample 9828@end table 9829 9830GCC already has an extensive list of suffixes built into it. 9831This directive will add an entry to the end of the list of suffixes, but 9832since the list is searched from the end backwards, it is effectively 9833possible to override earlier entries using this technique. 9834 9835@end table 9836 9837GCC has the following spec strings built into it. Spec files can 9838override these strings or create their own. Note that individual 9839targets can also add their own spec strings to this list. 9840 9841@smallexample 9842asm Options to pass to the assembler 9843asm_final Options to pass to the assembler post-processor 9844cpp Options to pass to the C preprocessor 9845cc1 Options to pass to the C compiler 9846cc1plus Options to pass to the C++ compiler 9847endfile Object files to include at the end of the link 9848link Options to pass to the linker 9849lib Libraries to include on the command line to the linker 9850libgcc Decides which GCC support library to pass to the linker 9851linker Sets the name of the linker 9852predefines Defines to be passed to the C preprocessor 9853signed_char Defines to pass to CPP to say whether @code{char} is signed 9854 by default 9855startfile Object files to include at the start of the link 9856@end smallexample 9857 9858Here is a small example of a spec file: 9859 9860@smallexample 9861%rename lib old_lib 9862 9863*lib: 9864--start-group -lgcc -lc -leval1 --end-group %(old_lib) 9865@end smallexample 9866 9867This example renames the spec called @samp{lib} to @samp{old_lib} and 9868then overrides the previous definition of @samp{lib} with a new one. 9869The new definition adds in some extra command-line options before 9870including the text of the old definition. 9871 9872@dfn{Spec strings} are a list of command-line options to be passed to their 9873corresponding program. In addition, the spec strings can contain 9874@samp{%}-prefixed sequences to substitute variable text or to 9875conditionally insert text into the command line. Using these constructs 9876it is possible to generate quite complex command lines. 9877 9878Here is a table of all defined @samp{%}-sequences for spec 9879strings. Note that spaces are not generated automatically around the 9880results of expanding these sequences. Therefore you can concatenate them 9881together or combine them with constant text in a single argument. 9882 9883@table @code 9884@item %% 9885Substitute one @samp{%} into the program name or argument. 9886 9887@item %i 9888Substitute the name of the input file being processed. 9889 9890@item %b 9891Substitute the basename of the input file being processed. 9892This is the substring up to (and not including) the last period 9893and not including the directory. 9894 9895@item %B 9896This is the same as @samp{%b}, but include the file suffix (text after 9897the last period). 9898 9899@item %d 9900Marks the argument containing or following the @samp{%d} as a 9901temporary file name, so that that file will be deleted if GCC exits 9902successfully. Unlike @samp{%g}, this contributes no text to the 9903argument. 9904 9905@item %g@var{suffix} 9906Substitute a file name that has suffix @var{suffix} and is chosen 9907once per compilation, and mark the argument in the same way as 9908@samp{%d}. To reduce exposure to denial-of-service attacks, the file 9909name is now chosen in a way that is hard to predict even when previously 9910chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s} 9911might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches 9912the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is 9913treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g} 9914was simply substituted with a file name chosen once per compilation, 9915without regard to any appended suffix (which was therefore treated 9916just like ordinary text), making such attacks more likely to succeed. 9917 9918@item %u@var{suffix} 9919Like @samp{%g}, but generates a new temporary file name even if 9920@samp{%u@var{suffix}} was already seen. 9921 9922@item %U@var{suffix} 9923Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a 9924new one if there is no such last file name. In the absence of any 9925@samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share 9926the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s} 9927would involve the generation of two distinct file names, one 9928for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was 9929simply substituted with a file name chosen for the previous @samp{%u}, 9930without regard to any appended suffix. 9931 9932@item %j@var{suffix} 9933Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is 9934writable, and if save-temps is off; otherwise, substitute the name 9935of a temporary file, just like @samp{%u}. This temporary file is not 9936meant for communication between processes, but rather as a junk 9937disposal mechanism. 9938 9939@item %|@var{suffix} 9940@itemx %m@var{suffix} 9941Like @samp{%g}, except if @option{-pipe} is in effect. In that case 9942@samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at 9943all. These are the two most common ways to instruct a program that it 9944should read from standard input or write to standard output. If you 9945need something more elaborate you can use an @samp{%@{pipe:@code{X}@}} 9946construct: see for example @file{f/lang-specs.h}. 9947 9948@item %.@var{SUFFIX} 9949Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args 9950when it is subsequently output with @samp{%*}. @var{SUFFIX} is 9951terminated by the next space or %. 9952 9953@item %w 9954Marks the argument containing or following the @samp{%w} as the 9955designated output file of this compilation. This puts the argument 9956into the sequence of arguments that @samp{%o} will substitute later. 9957 9958@item %o 9959Substitutes the names of all the output files, with spaces 9960automatically placed around them. You should write spaces 9961around the @samp{%o} as well or the results are undefined. 9962@samp{%o} is for use in the specs for running the linker. 9963Input files whose names have no recognized suffix are not compiled 9964at all, but they are included among the output files, so they will 9965be linked. 9966 9967@item %O 9968Substitutes the suffix for object files. Note that this is 9969handled specially when it immediately follows @samp{%g, %u, or %U}, 9970because of the need for those to form complete file names. The 9971handling is such that @samp{%O} is treated exactly as if it had already 9972been substituted, except that @samp{%g, %u, and %U} do not currently 9973support additional @var{suffix} characters following @samp{%O} as they would 9974following, for example, @samp{.o}. 9975 9976@item %p 9977Substitutes the standard macro predefinitions for the 9978current target machine. Use this when running @code{cpp}. 9979 9980@item %P 9981Like @samp{%p}, but puts @samp{__} before and after the name of each 9982predefined macro, except for macros that start with @samp{__} or with 9983@samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO 9984C@. 9985 9986@item %I 9987Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}), 9988@option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), 9989@option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options) 9990and @option{-imultilib} as necessary. 9991 9992@item %s 9993Current argument is the name of a library or startup file of some sort. 9994Search for that file in a standard list of directories and substitute 9995the full name found. The current working directory is included in the 9996list of directories scanned. 9997 9998@item %T 9999Current argument is the name of a linker script. Search for that file 10000in the current list of directories to scan for libraries. If the file 10001is located insert a @option{--script} option into the command line 10002followed by the full path name found. If the file is not found then 10003generate an error message. Note: the current working directory is not 10004searched. 10005 10006@item %e@var{str} 10007Print @var{str} as an error message. @var{str} is terminated by a newline. 10008Use this when inconsistent options are detected. 10009 10010@item %(@var{name}) 10011Substitute the contents of spec string @var{name} at this point. 10012 10013@item %x@{@var{option}@} 10014Accumulate an option for @samp{%X}. 10015 10016@item %X 10017Output the accumulated linker options specified by @option{-Wl} or a @samp{%x} 10018spec string. 10019 10020@item %Y 10021Output the accumulated assembler options specified by @option{-Wa}. 10022 10023@item %Z 10024Output the accumulated preprocessor options specified by @option{-Wp}. 10025 10026@item %a 10027Process the @code{asm} spec. This is used to compute the 10028switches to be passed to the assembler. 10029 10030@item %A 10031Process the @code{asm_final} spec. This is a spec string for 10032passing switches to an assembler post-processor, if such a program is 10033needed. 10034 10035@item %l 10036Process the @code{link} spec. This is the spec for computing the 10037command line passed to the linker. Typically it will make use of the 10038@samp{%L %G %S %D and %E} sequences. 10039 10040@item %D 10041Dump out a @option{-L} option for each directory that GCC believes might 10042contain startup files. If the target supports multilibs then the 10043current multilib directory will be prepended to each of these paths. 10044 10045@item %L 10046Process the @code{lib} spec. This is a spec string for deciding which 10047libraries should be included on the command line to the linker. 10048 10049@item %G 10050Process the @code{libgcc} spec. This is a spec string for deciding 10051which GCC support library should be included on the command line to the linker. 10052 10053@item %S 10054Process the @code{startfile} spec. This is a spec for deciding which 10055object files should be the first ones passed to the linker. Typically 10056this might be a file named @file{crt0.o}. 10057 10058@item %E 10059Process the @code{endfile} spec. This is a spec string that specifies 10060the last object files that will be passed to the linker. 10061 10062@item %C 10063Process the @code{cpp} spec. This is used to construct the arguments 10064to be passed to the C preprocessor. 10065 10066@item %1 10067Process the @code{cc1} spec. This is used to construct the options to be 10068passed to the actual C compiler (@samp{cc1}). 10069 10070@item %2 10071Process the @code{cc1plus} spec. This is used to construct the options to be 10072passed to the actual C++ compiler (@samp{cc1plus}). 10073 10074@item %* 10075Substitute the variable part of a matched option. See below. 10076Note that each comma in the substituted string is replaced by 10077a single space. 10078 10079@item %<@code{S} 10080Remove all occurrences of @code{-S} from the command line. Note---this 10081command is position dependent. @samp{%} commands in the spec string 10082before this one will see @code{-S}, @samp{%} commands in the spec string 10083after this one will not. 10084 10085@item %:@var{function}(@var{args}) 10086Call the named function @var{function}, passing it @var{args}. 10087@var{args} is first processed as a nested spec string, then split 10088into an argument vector in the usual fashion. The function returns 10089a string which is processed as if it had appeared literally as part 10090of the current spec. 10091 10092The following built-in spec functions are provided: 10093 10094@table @code 10095@item @code{getenv} 10096The @code{getenv} spec function takes two arguments: an environment 10097variable name and a string. If the environment variable is not 10098defined, a fatal error is issued. Otherwise, the return value is the 10099value of the environment variable concatenated with the string. For 10100example, if @env{TOPDIR} is defined as @file{/path/to/top}, then: 10101 10102@smallexample 10103%:getenv(TOPDIR /include) 10104@end smallexample 10105 10106expands to @file{/path/to/top/include}. 10107 10108@item @code{if-exists} 10109The @code{if-exists} spec function takes one argument, an absolute 10110pathname to a file. If the file exists, @code{if-exists} returns the 10111pathname. Here is a small example of its usage: 10112 10113@smallexample 10114*startfile: 10115crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s 10116@end smallexample 10117 10118@item @code{if-exists-else} 10119The @code{if-exists-else} spec function is similar to the @code{if-exists} 10120spec function, except that it takes two arguments. The first argument is 10121an absolute pathname to a file. If the file exists, @code{if-exists-else} 10122returns the pathname. If it does not exist, it returns the second argument. 10123This way, @code{if-exists-else} can be used to select one file or another, 10124based on the existence of the first. Here is a small example of its usage: 10125 10126@smallexample 10127*startfile: 10128crt0%O%s %:if-exists(crti%O%s) \ 10129%:if-exists-else(crtbeginT%O%s crtbegin%O%s) 10130@end smallexample 10131 10132@item @code{replace-outfile} 10133The @code{replace-outfile} spec function takes two arguments. It looks for the 10134first argument in the outfiles array and replaces it with the second argument. Here 10135is a small example of its usage: 10136 10137@smallexample 10138%@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@} 10139@end smallexample 10140 10141@item @code{remove-outfile} 10142The @code{remove-outfile} spec function takes one argument. It looks for the 10143first argument in the outfiles array and removes it. Here is a small example 10144its usage: 10145 10146@smallexample 10147%:remove-outfile(-lm) 10148@end smallexample 10149 10150@item @code{pass-through-libs} 10151The @code{pass-through-libs} spec function takes any number of arguments. It 10152finds any @option{-l} options and any non-options ending in ".a" (which it 10153assumes are the names of linker input library archive files) and returns a 10154result containing all the found arguments each prepended by 10155@option{-plugin-opt=-pass-through=} and joined by spaces. This list is 10156intended to be passed to the LTO linker plugin. 10157 10158@smallexample 10159%:pass-through-libs(%G %L %G) 10160@end smallexample 10161 10162@item @code{print-asm-header} 10163The @code{print-asm-header} function takes no arguments and simply 10164prints a banner like: 10165 10166@smallexample 10167Assembler options 10168================= 10169 10170Use "-Wa,OPTION" to pass "OPTION" to the assembler. 10171@end smallexample 10172 10173It is used to separate compiler options from assembler options 10174in the @option{--target-help} output. 10175@end table 10176 10177@item %@{@code{S}@} 10178Substitutes the @code{-S} switch, if that switch was given to GCC@. 10179If that switch was not specified, this substitutes nothing. Note that 10180the leading dash is omitted when specifying this option, and it is 10181automatically inserted if the substitution is performed. Thus the spec 10182string @samp{%@{foo@}} would match the command-line option @option{-foo} 10183and would output the command-line option @option{-foo}. 10184 10185@item %W@{@code{S}@} 10186Like %@{@code{S}@} but mark last argument supplied within as a file to be 10187deleted on failure. 10188 10189@item %@{@code{S}*@} 10190Substitutes all the switches specified to GCC whose names start 10191with @code{-S}, but which also take an argument. This is used for 10192switches like @option{-o}, @option{-D}, @option{-I}, etc. 10193GCC considers @option{-o foo} as being 10194one switch whose names starts with @samp{o}. %@{o*@} would substitute this 10195text, including the space. Thus two arguments would be generated. 10196 10197@item %@{@code{S}*&@code{T}*@} 10198Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options 10199(the order of @code{S} and @code{T} in the spec is not significant). 10200There can be any number of ampersand-separated variables; for each the 10201wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}. 10202 10203@item %@{@code{S}:@code{X}@} 10204Substitutes @code{X}, if the @samp{-S} switch was given to GCC@. 10205 10206@item %@{!@code{S}:@code{X}@} 10207Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@. 10208 10209@item %@{@code{S}*:@code{X}@} 10210Substitutes @code{X} if one or more switches whose names start with 10211@code{-S} are specified to GCC@. Normally @code{X} is substituted only 10212once, no matter how many such switches appeared. However, if @code{%*} 10213appears somewhere in @code{X}, then @code{X} will be substituted once 10214for each matching switch, with the @code{%*} replaced by the part of 10215that switch that matched the @code{*}. 10216 10217@item %@{.@code{S}:@code{X}@} 10218Substitutes @code{X}, if processing a file with suffix @code{S}. 10219 10220@item %@{!.@code{S}:@code{X}@} 10221Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}. 10222 10223@item %@{,@code{S}:@code{X}@} 10224Substitutes @code{X}, if processing a file for language @code{S}. 10225 10226@item %@{!,@code{S}:@code{X}@} 10227Substitutes @code{X}, if not processing a file for language @code{S}. 10228 10229@item %@{@code{S}|@code{P}:@code{X}@} 10230Substitutes @code{X} if either @code{-S} or @code{-P} was given to 10231GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and 10232@code{*} sequences as well, although they have a stronger binding than 10233the @samp{|}. If @code{%*} appears in @code{X}, all of the 10234alternatives must be starred, and only the first matching alternative 10235is substituted. 10236 10237For example, a spec string like this: 10238 10239@smallexample 10240%@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@} 10241@end smallexample 10242 10243will output the following command-line options from the following input 10244command-line options: 10245 10246@smallexample 10247fred.c -foo -baz 10248jim.d -bar -boggle 10249-d fred.c -foo -baz -boggle 10250-d jim.d -bar -baz -boggle 10251@end smallexample 10252 10253@item %@{S:X; T:Y; :D@} 10254 10255If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was 10256given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can 10257be as many clauses as you need. This may be combined with @code{.}, 10258@code{,}, @code{!}, @code{|}, and @code{*} as needed. 10259 10260 10261@end table 10262 10263The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar 10264construct may contain other nested @samp{%} constructs or spaces, or 10265even newlines. They are processed as usual, as described above. 10266Trailing white space in @code{X} is ignored. White space may also 10267appear anywhere on the left side of the colon in these constructs, 10268except between @code{.} or @code{*} and the corresponding word. 10269 10270The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are 10271handled specifically in these constructs. If another value of 10272@option{-O} or the negated form of a @option{-f}, @option{-m}, or 10273@option{-W} switch is found later in the command line, the earlier 10274switch value is ignored, except with @{@code{S}*@} where @code{S} is 10275just one letter, which passes all matching options. 10276 10277The character @samp{|} at the beginning of the predicate text is used to 10278indicate that a command should be piped to the following command, but 10279only if @option{-pipe} is specified. 10280 10281It is built into GCC which switches take arguments and which do not. 10282(You might think it would be useful to generalize this to allow each 10283compiler's spec to say which switches take arguments. But this cannot 10284be done in a consistent fashion. GCC cannot even decide which input 10285files have been specified without knowing which switches take arguments, 10286and it must know which input files to compile in order to tell which 10287compilers to run). 10288 10289GCC also knows implicitly that arguments starting in @option{-l} are to be 10290treated as compiler output files, and passed to the linker in their 10291proper position among the other output files. 10292 10293@c man begin OPTIONS 10294 10295@node Target Options 10296@section Specifying Target Machine and Compiler Version 10297@cindex target options 10298@cindex cross compiling 10299@cindex specifying machine version 10300@cindex specifying compiler version and target machine 10301@cindex compiler version, specifying 10302@cindex target machine, specifying 10303 10304The usual way to run GCC is to run the executable called @command{gcc}, or 10305@command{@var{machine}-gcc} when cross-compiling, or 10306@command{@var{machine}-gcc-@var{version}} to run a version other than the 10307one that was installed last. 10308 10309@node Submodel Options 10310@section Hardware Models and Configurations 10311@cindex submodel options 10312@cindex specifying hardware config 10313@cindex hardware models and configurations, specifying 10314@cindex machine dependent options 10315 10316Each target machine types can have its own 10317special options, starting with @samp{-m}, to choose among various 10318hardware models or configurations---for example, 68010 vs 68020, 10319floating coprocessor or none. A single installed version of the 10320compiler can compile for any model or configuration, according to the 10321options specified. 10322 10323Some configurations of the compiler also support additional special 10324options, usually for compatibility with other compilers on the same 10325platform. 10326 10327@c This list is ordered alphanumerically by subsection name. 10328@c It should be the same order and spelling as these options are listed 10329@c in Machine Dependent Options 10330 10331@menu 10332* Adapteva Epiphany Options:: 10333* ARM Options:: 10334* AVR Options:: 10335* Blackfin Options:: 10336* C6X Options:: 10337* CRIS Options:: 10338* CR16 Options:: 10339* Darwin Options:: 10340* DEC Alpha Options:: 10341* DEC Alpha/VMS Options:: 10342* FR30 Options:: 10343* FRV Options:: 10344* GNU/Linux Options:: 10345* H8/300 Options:: 10346* HPPA Options:: 10347* i386 and x86-64 Options:: 10348* i386 and x86-64 Windows Options:: 10349* IA-64 Options:: 10350* IA-64/VMS Options:: 10351* LM32 Options:: 10352* M32C Options:: 10353* M32R/D Options:: 10354* M680x0 Options:: 10355* MCore Options:: 10356* MeP Options:: 10357* MicroBlaze Options:: 10358* MIPS Options:: 10359* MMIX Options:: 10360* MN10300 Options:: 10361* PDP-11 Options:: 10362* picoChip Options:: 10363* PowerPC Options:: 10364* RL78 Options:: 10365* RS/6000 and PowerPC Options:: 10366* RX Options:: 10367* S/390 and zSeries Options:: 10368* Score Options:: 10369* SH Options:: 10370* Solaris 2 Options:: 10371* SPARC Options:: 10372* SPU Options:: 10373* System V Options:: 10374* TILE-Gx Options:: 10375* TILEPro Options:: 10376* V850 Options:: 10377* VAX Options:: 10378* VxWorks Options:: 10379* x86-64 Options:: 10380* Xstormy16 Options:: 10381* Xtensa Options:: 10382* zSeries Options:: 10383@end menu 10384 10385@node Adapteva Epiphany Options 10386@subsection Adapteva Epiphany Options 10387 10388These @samp{-m} options are defined for Adapteva Epiphany: 10389 10390@table @gcctabopt 10391@item -mhalf-reg-file 10392@opindex mhalf-reg-file 10393Don't allocate any register in the range @code{r32}@dots{}@code{r63}. 10394That allows code to run on hardware variants that lack these registers. 10395 10396@item -mprefer-short-insn-regs 10397@opindex mprefer-short-insn-regs 10398Preferrentially allocate registers that allow short instruction generation. 10399This can result in increasesd instruction count, so if this reduces or 10400increases code size might vary from case to case. 10401 10402@item -mbranch-cost=@var{num} 10403@opindex mbranch-cost 10404Set the cost of branches to roughly @var{num} ``simple'' instructions. 10405This cost is only a heuristic and is not guaranteed to produce 10406consistent results across releases. 10407 10408@item -mcmove 10409@opindex mcmove 10410Enable the generation of conditional moves. 10411 10412@item -mnops=@var{num} 10413@opindex mnops 10414Emit @var{num} nops before every other generated instruction. 10415 10416@item -mno-soft-cmpsf 10417@opindex mno-soft-cmpsf 10418For single-precision floating-point comparisons, emit an fsub instruction 10419and test the flags. This is faster than a software comparison, but can 10420get incorrect results in the presence of NaNs, or when two different small 10421numbers are compared such that their difference is calculated as zero. 10422The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant, 10423software comparisons. 10424 10425@item -mstack-offset=@var{num} 10426@opindex mstack-offset 10427Set the offset between the top of the stack and the stack pointer. 10428E.g., a value of 8 means that the eight bytes in the range sp+0@dots{}sp+7 10429can be used by leaf functions without stack allocation. 10430Values other than @samp{8} or @samp{16} are untested and unlikely to work. 10431Note also that this option changes the ABI, compiling a program with a 10432different stack offset than the libraries have been compiled with 10433will generally not work. 10434This option can be useful if you want to evaluate if a different stack 10435offset would give you better code, but to actually use a different stack 10436offset to build working programs, it is recommended to configure the 10437toolchain with the appropriate @samp{--with-stack-offset=@var{num}} option. 10438 10439@item -mno-round-nearest 10440@opindex mno-round-nearest 10441Make the scheduler assume that the rounding mode has been set to 10442truncating. The default is @option{-mround-nearest}. 10443 10444@item -mlong-calls 10445@opindex mlong-calls 10446If not otherwise specified by an attribute, assume all calls might be beyond 10447the offset range of the b / bl instructions, and therefore load the 10448function address into a register before performing a (otherwise direct) call. 10449This is the default. 10450 10451@item -mshort-calls 10452@opindex short-calls 10453If not otherwise specified by an attribute, assume all direct calls are 10454in the range of the b / bl instructions, so use these instructions 10455for direct calls. The default is @option{-mlong-calls}. 10456 10457@item -msmall16 10458@opindex msmall16 10459Assume addresses can be loaded as 16-bit unsigned values. This does not 10460apply to function addresses for which @option{-mlong-calls} semantics 10461are in effect. 10462 10463@item -mfp-mode=@var{mode} 10464@opindex mfp-mode 10465Set the prevailing mode of the floating-point unit. 10466This determines the floating-point mode that is provided and expected 10467at function call and return time. Making this mode match the mode you 10468predominantly need at function start can make your programs smaller and 10469faster by avoiding unnecessary mode switches. 10470 10471@var{mode} can be set to one the following values: 10472 10473@table @samp 10474@item caller 10475Any mode at function entry is valid, and retained or restored when 10476the function returns, and when it calls other functions. 10477This mode is useful for compiling libraries or other compilation units 10478you might want to incorporate into different programs with different 10479prevailing FPU modes, and the convenience of being able to use a single 10480object file outweighs the size and speed overhead for any extra 10481mode switching that might be needed, compared with what would be needed 10482with a more specific choice of prevailing FPU mode. 10483 10484@item truncate 10485This is the mode used for floating-point calculations with 10486truncating (i.e.@: round towards zero) rounding mode. That includes 10487conversion from floating point to integer. 10488 10489@item round-nearest 10490This is the mode used for floating-point calculations with 10491round-to-nearest-or-even rounding mode. 10492 10493@item int 10494This is the mode used to perform integer calculations in the FPU, e.g.@: 10495integer multiply, or integer multiply-and-accumulate. 10496@end table 10497 10498The default is @option{-mfp-mode=caller} 10499 10500@item -mnosplit-lohi 10501@opindex mnosplit-lohi 10502@item -mno-postinc 10503@opindex mno-postinc 10504@item -mno-postmodify 10505@opindex mno-postmodify 10506Code generation tweaks that disable, respectively, splitting of 32-bit 10507loads, generation of post-increment addresses, and generation of 10508post-modify addresses. The defaults are @option{msplit-lohi}, 10509@option{-mpost-inc}, and @option{-mpost-modify}. 10510 10511@item -mnovect-double 10512@opindex mno-vect-double 10513Change the preferred SIMD mode to SImode. The default is 10514@option{-mvect-double}, which uses DImode as preferred SIMD mode. 10515 10516@item -max-vect-align=@var{num} 10517@opindex max-vect-align 10518The maximum alignment for SIMD vector mode types. 10519@var{num} may be 4 or 8. The default is 8. 10520Note that this is an ABI change, even though many library function 10521interfaces will be unaffected, if they don't use SIMD vector modes 10522in places where they affect size and/or alignment of relevant types. 10523 10524@item -msplit-vecmove-early 10525@opindex msplit-vecmove-early 10526Split vector moves into single word moves before reload. In theory this 10527could give better register allocation, but so far the reverse seems to be 10528generally the case. 10529 10530@item -m1reg-@var{reg} 10531@opindex m1reg- 10532Specify a register to hold the constant @minus{}1, which makes loading small negative 10533constants and certain bitmasks faster. 10534Allowable values for reg are r43 and r63, which specify to use that register 10535as a fixed register, and none, which means that no register is used for this 10536purpose. The default is @option{-m1reg-none}. 10537 10538@end table 10539 10540@node ARM Options 10541@subsection ARM Options 10542@cindex ARM options 10543 10544These @samp{-m} options are defined for Advanced RISC Machines (ARM) 10545architectures: 10546 10547@table @gcctabopt 10548@item -mabi=@var{name} 10549@opindex mabi 10550Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu}, 10551@samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}. 10552 10553@item -mapcs-frame 10554@opindex mapcs-frame 10555Generate a stack frame that is compliant with the ARM Procedure Call 10556Standard for all functions, even if this is not strictly necessary for 10557correct execution of the code. Specifying @option{-fomit-frame-pointer} 10558with this option will cause the stack frames not to be generated for 10559leaf functions. The default is @option{-mno-apcs-frame}. 10560 10561@item -mapcs 10562@opindex mapcs 10563This is a synonym for @option{-mapcs-frame}. 10564 10565@ignore 10566@c not currently implemented 10567@item -mapcs-stack-check 10568@opindex mapcs-stack-check 10569Generate code to check the amount of stack space available upon entry to 10570every function (that actually uses some stack space). If there is 10571insufficient space available then either the function 10572@samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be 10573called, depending upon the amount of stack space required. The runtime 10574system is required to provide these functions. The default is 10575@option{-mno-apcs-stack-check}, since this produces smaller code. 10576 10577@c not currently implemented 10578@item -mapcs-float 10579@opindex mapcs-float 10580Pass floating-point arguments using the floating-point registers. This is 10581one of the variants of the APCS@. This option is recommended if the 10582target hardware has a floating-point unit or if a lot of floating-point 10583arithmetic is going to be performed by the code. The default is 10584@option{-mno-apcs-float}, since integer only code is slightly increased in 10585size if @option{-mapcs-float} is used. 10586 10587@c not currently implemented 10588@item -mapcs-reentrant 10589@opindex mapcs-reentrant 10590Generate reentrant, position independent code. The default is 10591@option{-mno-apcs-reentrant}. 10592@end ignore 10593 10594@item -mthumb-interwork 10595@opindex mthumb-interwork 10596Generate code that supports calling between the ARM and Thumb 10597instruction sets. Without this option, on pre-v5 architectures, the 10598two instruction sets cannot be reliably used inside one program. The 10599default is @option{-mno-thumb-interwork}, since slightly larger code 10600is generated when @option{-mthumb-interwork} is specified. In AAPCS 10601configurations this option is meaningless. 10602 10603@item -mno-sched-prolog 10604@opindex mno-sched-prolog 10605Prevent the reordering of instructions in the function prologue, or the 10606merging of those instruction with the instructions in the function's 10607body. This means that all functions will start with a recognizable set 10608of instructions (or in fact one of a choice from a small set of 10609different function prologues), and this information can be used to 10610locate the start if functions inside an executable piece of code. The 10611default is @option{-msched-prolog}. 10612 10613@item -mfloat-abi=@var{name} 10614@opindex mfloat-abi 10615Specifies which floating-point ABI to use. Permissible values 10616are: @samp{soft}, @samp{softfp} and @samp{hard}. 10617 10618Specifying @samp{soft} causes GCC to generate output containing 10619library calls for floating-point operations. 10620@samp{softfp} allows the generation of code using hardware floating-point 10621instructions, but still uses the soft-float calling conventions. 10622@samp{hard} allows generation of floating-point instructions 10623and uses FPU-specific calling conventions. 10624 10625The default depends on the specific target configuration. Note that 10626the hard-float and soft-float ABIs are not link-compatible; you must 10627compile your entire program with the same ABI, and link with a 10628compatible set of libraries. 10629 10630@item -mlittle-endian 10631@opindex mlittle-endian 10632Generate code for a processor running in little-endian mode. This is 10633the default for all standard configurations. 10634 10635@item -mbig-endian 10636@opindex mbig-endian 10637Generate code for a processor running in big-endian mode; the default is 10638to compile code for a little-endian processor. 10639 10640@item -mwords-little-endian 10641@opindex mwords-little-endian 10642This option only applies when generating code for big-endian processors. 10643Generate code for a little-endian word order but a big-endian byte 10644order. That is, a byte order of the form @samp{32107654}. Note: this 10645option should only be used if you require compatibility with code for 10646big-endian ARM processors generated by versions of the compiler prior to 106472.8. This option is now deprecated. 10648 10649@item -mcpu=@var{name} 10650@opindex mcpu 10651This specifies the name of the target ARM processor. GCC uses this name 10652to determine what kind of instructions it can emit when generating 10653assembly code. Permissible names are: @samp{arm2}, @samp{arm250}, 10654@samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610}, 10655@samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm}, 10656@samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700}, 10657@samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100}, 10658@samp{arm720}, 10659@samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s}, 10660@samp{arm710t}, @samp{arm720t}, @samp{arm740t}, 10661@samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100}, 10662@samp{strongarm1110}, 10663@samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920}, 10664@samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s}, 10665@samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi}, 10666@samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s}, 10667@samp{arm10e}, @samp{arm1020e}, @samp{arm1022e}, 10668@samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp}, 10669@samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s}, 10670@samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9}, 10671@samp{cortex-a15}, @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5}, 10672@samp{cortex-m4}, @samp{cortex-m3}, 10673@samp{cortex-m1}, 10674@samp{cortex-m0}, 10675@samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}, 10676@samp{fa526}, @samp{fa626}, 10677@samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}. 10678 10679 10680@option{-mcpu=generic-@var{arch}} is also permissible, and is 10681equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}. 10682See @option{-mtune} for more information. 10683 10684@option{-mcpu=native} causes the compiler to auto-detect the CPU 10685of the build computer. At present, this feature is only supported on 10686Linux, and not all architectures are recognized. If the auto-detect is 10687unsuccessful the option has no effect. 10688 10689@item -mtune=@var{name} 10690@opindex mtune 10691This option is very similar to the @option{-mcpu=} option, except that 10692instead of specifying the actual target processor type, and hence 10693restricting which instructions can be used, it specifies that GCC should 10694tune the performance of the code as if the target were of the type 10695specified in this option, but still choosing the instructions that it 10696will generate based on the CPU specified by a @option{-mcpu=} option. 10697For some ARM implementations better performance can be obtained by using 10698this option. 10699 10700@option{-mtune=generic-@var{arch}} specifies that GCC should tune the 10701performance for a blend of processors within architecture @var{arch}. 10702The aim is to generate code that run well on the current most popular 10703processors, balancing between optimizations that benefit some CPUs in the 10704range, and avoiding performance pitfalls of other CPUs. The effects of 10705this option may change in future GCC versions as CPU models come and go. 10706 10707@option{-mtune=native} causes the compiler to auto-detect the CPU 10708of the build computer. At present, this feature is only supported on 10709Linux, and not all architectures are recognized. If the auto-detect is 10710unsuccessful the option has no effect. 10711 10712@item -march=@var{name} 10713@opindex march 10714This specifies the name of the target ARM architecture. GCC uses this 10715name to determine what kind of instructions it can emit when generating 10716assembly code. This option can be used in conjunction with or instead 10717of the @option{-mcpu=} option. Permissible names are: @samp{armv2}, 10718@samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t}, 10719@samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te}, 10720@samp{armv6}, @samp{armv6j}, 10721@samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m}, 10722@samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, 10723@samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}. 10724 10725@option{-march=native} causes the compiler to auto-detect the architecture 10726of the build computer. At present, this feature is only supported on 10727Linux, and not all architectures are recognized. If the auto-detect is 10728unsuccessful the option has no effect. 10729 10730@item -mfpu=@var{name} 10731@itemx -mfpe=@var{number} 10732@itemx -mfp=@var{number} 10733@opindex mfpu 10734@opindex mfpe 10735@opindex mfp 10736This specifies what floating-point hardware (or hardware emulation) is 10737available on the target. Permissible names are: @samp{fpa}, @samp{fpe2}, 10738@samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16}, 10739@samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16}, 10740@samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16}, 10741@samp{fpv4-sp-d16} and @samp{neon-vfpv4}. 10742@option{-mfp} and @option{-mfpe} are synonyms for 10743@option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions 10744of GCC@. 10745 10746If @option{-msoft-float} is specified this specifies the format of 10747floating-point values. 10748 10749If the selected floating-point hardware includes the NEON extension 10750(e.g. @option{-mfpu}=@samp{neon}), note that floating-point 10751operations will not be used by GCC's auto-vectorization pass unless 10752@option{-funsafe-math-optimizations} is also specified. This is 10753because NEON hardware does not fully implement the IEEE 754 standard for 10754floating-point arithmetic (in particular denormal values are treated as 10755zero), so the use of NEON instructions may lead to a loss of precision. 10756 10757@item -mfp16-format=@var{name} 10758@opindex mfp16-format 10759Specify the format of the @code{__fp16} half-precision floating-point type. 10760Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative}; 10761the default is @samp{none}, in which case the @code{__fp16} type is not 10762defined. @xref{Half-Precision}, for more information. 10763 10764@item -mstructure-size-boundary=@var{n} 10765@opindex mstructure-size-boundary 10766The size of all structures and unions will be rounded up to a multiple 10767of the number of bits set by this option. Permissible values are 8, 32 10768and 64. The default value varies for different toolchains. For the COFF 10769targeted toolchain the default value is 8. A value of 64 is only allowed 10770if the underlying ABI supports it. 10771 10772Specifying the larger number can produce faster, more efficient code, but 10773can also increase the size of the program. Different values are potentially 10774incompatible. Code compiled with one value cannot necessarily expect to 10775work with code or libraries compiled with another value, if they exchange 10776information using structures or unions. 10777 10778@item -mabort-on-noreturn 10779@opindex mabort-on-noreturn 10780Generate a call to the function @code{abort} at the end of a 10781@code{noreturn} function. It will be executed if the function tries to 10782return. 10783 10784@item -mlong-calls 10785@itemx -mno-long-calls 10786@opindex mlong-calls 10787@opindex mno-long-calls 10788Tells the compiler to perform function calls by first loading the 10789address of the function into a register and then performing a subroutine 10790call on this register. This switch is needed if the target function 10791will lie outside of the 64 megabyte addressing range of the offset based 10792version of subroutine call instruction. 10793 10794Even if this switch is enabled, not all function calls will be turned 10795into long calls. The heuristic is that static functions, functions 10796that have the @samp{short-call} attribute, functions that are inside 10797the scope of a @samp{#pragma no_long_calls} directive and functions whose 10798definitions have already been compiled within the current compilation 10799unit, will not be turned into long calls. The exception to this rule is 10800that weak function definitions, functions with the @samp{long-call} 10801attribute or the @samp{section} attribute, and functions that are within 10802the scope of a @samp{#pragma long_calls} directive, will always be 10803turned into long calls. 10804 10805This feature is not enabled by default. Specifying 10806@option{-mno-long-calls} will restore the default behavior, as will 10807placing the function calls within the scope of a @samp{#pragma 10808long_calls_off} directive. Note these switches have no effect on how 10809the compiler generates code to handle function calls via function 10810pointers. 10811 10812@item -msingle-pic-base 10813@opindex msingle-pic-base 10814Treat the register used for PIC addressing as read-only, rather than 10815loading it in the prologue for each function. The runtime system is 10816responsible for initializing this register with an appropriate value 10817before execution begins. 10818 10819@item -mpic-register=@var{reg} 10820@opindex mpic-register 10821Specify the register to be used for PIC addressing. The default is R10 10822unless stack-checking is enabled, when R9 is used. 10823 10824@item -mcirrus-fix-invalid-insns 10825@opindex mcirrus-fix-invalid-insns 10826@opindex mno-cirrus-fix-invalid-insns 10827Insert NOPs into the instruction stream to in order to work around 10828problems with invalid Maverick instruction combinations. This option 10829is only valid if the @option{-mcpu=ep9312} option has been used to 10830enable generation of instructions for the Cirrus Maverick floating-point 10831co-processor. This option is not enabled by default, since the 10832problem is only present in older Maverick implementations. The default 10833can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns} 10834switch. 10835 10836@item -mpoke-function-name 10837@opindex mpoke-function-name 10838Write the name of each function into the text section, directly 10839preceding the function prologue. The generated code is similar to this: 10840 10841@smallexample 10842 t0 10843 .ascii "arm_poke_function_name", 0 10844 .align 10845 t1 10846 .word 0xff000000 + (t1 - t0) 10847 arm_poke_function_name 10848 mov ip, sp 10849 stmfd sp!, @{fp, ip, lr, pc@} 10850 sub fp, ip, #4 10851@end smallexample 10852 10853When performing a stack backtrace, code can inspect the value of 10854@code{pc} stored at @code{fp + 0}. If the trace function then looks at 10855location @code{pc - 12} and the top 8 bits are set, then we know that 10856there is a function name embedded immediately preceding this location 10857and has length @code{((pc[-3]) & 0xff000000)}. 10858 10859@item -mthumb 10860@itemx -marm 10861@opindex marm 10862@opindex mthumb 10863 10864Select between generating code that executes in ARM and Thumb 10865states. The default for most configurations is to generate code 10866that executes in ARM state, but the default can be changed by 10867configuring GCC with the @option{--with-mode=}@var{state} 10868configure option. 10869 10870@item -mtpcs-frame 10871@opindex mtpcs-frame 10872Generate a stack frame that is compliant with the Thumb Procedure Call 10873Standard for all non-leaf functions. (A leaf function is one that does 10874not call any other functions.) The default is @option{-mno-tpcs-frame}. 10875 10876@item -mtpcs-leaf-frame 10877@opindex mtpcs-leaf-frame 10878Generate a stack frame that is compliant with the Thumb Procedure Call 10879Standard for all leaf functions. (A leaf function is one that does 10880not call any other functions.) The default is @option{-mno-apcs-leaf-frame}. 10881 10882@item -mcallee-super-interworking 10883@opindex mcallee-super-interworking 10884Gives all externally visible functions in the file being compiled an ARM 10885instruction set header which switches to Thumb mode before executing the 10886rest of the function. This allows these functions to be called from 10887non-interworking code. This option is not valid in AAPCS configurations 10888because interworking is enabled by default. 10889 10890@item -mcaller-super-interworking 10891@opindex mcaller-super-interworking 10892Allows calls via function pointers (including virtual functions) to 10893execute correctly regardless of whether the target code has been 10894compiled for interworking or not. There is a small overhead in the cost 10895of executing a function pointer if this option is enabled. This option 10896is not valid in AAPCS configurations because interworking is enabled 10897by default. 10898 10899@item -mtp=@var{name} 10900@opindex mtp 10901Specify the access model for the thread local storage pointer. The valid 10902models are @option{soft}, which generates calls to @code{__aeabi_read_tp}, 10903@option{cp15}, which fetches the thread pointer from @code{cp15} directly 10904(supported in the arm6k architecture), and @option{auto}, which uses the 10905best available method for the selected processor. The default setting is 10906@option{auto}. 10907 10908@item -mtls-dialect=@var{dialect} 10909@opindex mtls-dialect 10910Specify the dialect to use for accessing thread local storage. Two 10911dialects are supported --- @option{gnu} and @option{gnu2}. The 10912@option{gnu} dialect selects the original GNU scheme for supporting 10913local and global dynamic TLS models. The @option{gnu2} dialect 10914selects the GNU descriptor scheme, which provides better performance 10915for shared libraries. The GNU descriptor scheme is compatible with 10916the original scheme, but does require new assembler, linker and 10917library support. Initial and local exec TLS models are unaffected by 10918this option and always use the original scheme. 10919 10920@item -mword-relocations 10921@opindex mword-relocations 10922Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32). 10923This is enabled by default on targets (uClinux, SymbianOS) where the runtime 10924loader imposes this restriction, and when @option{-fpic} or @option{-fPIC} 10925is specified. 10926 10927@item -mfix-cortex-m3-ldrd 10928@opindex mfix-cortex-m3-ldrd 10929Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions 10930with overlapping destination and base registers are used. This option avoids 10931generating these instructions. This option is enabled by default when 10932@option{-mcpu=cortex-m3} is specified. 10933 10934@item -munaligned-access 10935@itemx -mno-unaligned-access 10936@opindex munaligned-access 10937@opindex mno-unaligned-access 10938Enables (or disables) reading and writing of 16- and 32- bit values 10939from addresses that are not 16- or 32- bit aligned. By default 10940unaligned access is disabled for all pre-ARMv6 and all ARMv6-M 10941architectures, and enabled for all other architectures. If unaligned 10942access is not enabled then words in packed data structures will be 10943accessed a byte at a time. 10944 10945The ARM attribute @code{Tag_CPU_unaligned_access} will be set in the 10946generated object file to either true or false, depending upon the 10947setting of this option. If unaligned access is enabled then the 10948preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} will also be 10949defined. 10950 10951@end table 10952 10953@node AVR Options 10954@subsection AVR Options 10955@cindex AVR Options 10956 10957@table @gcctabopt 10958@item -mmcu=@var{mcu} 10959@opindex mmcu 10960Specify Atmel AVR instruction set architectures (ISA) or MCU type. 10961 10962The default for this option is@tie{}@code{avr2}. 10963 10964GCC supports the following AVR devices and ISAs: 10965 10966@table @code 10967 10968@item avr2 10969``Classic'' devices with up to 8@tie{}KiB of program memory. 10970@*@var{mcu}@tie{}= @code{attiny22}, @code{attiny26}, @code{at90c8534}, 10971@code{at90s2313}, @code{at90s2323}, @code{at90s2333}, 10972@code{at90s2343}, @code{at90s4414}, @code{at90s4433}, 10973@code{at90s4434}, @code{at90s8515}, @code{at90s8535}. 10974 10975@item avr25 10976``Classic'' devices with up to 8@tie{}KiB of program memory and with 10977the @code{MOVW} instruction. 10978@*@var{mcu}@tie{}= @code{ata6289}, @code{attiny13}, @code{attiny13a}, 10979@code{attiny2313}, @code{attiny2313a}, @code{attiny24}, 10980@code{attiny24a}, @code{attiny25}, @code{attiny261}, 10981@code{attiny261a}, @code{attiny43u}, @code{attiny4313}, 10982@code{attiny44}, @code{attiny44a}, @code{attiny45}, @code{attiny461}, 10983@code{attiny461a}, @code{attiny48}, @code{attiny84}, @code{attiny84a}, 10984@code{attiny85}, @code{attiny861}, @code{attiny861a}, @code{attiny87}, 10985@code{attiny88}, @code{at86rf401}. 10986 10987@item avr3 10988``Classic'' devices with 16@tie{}KiB up to 64@tie{}KiB of program memory. 10989@*@var{mcu}@tie{}= @code{at43usb355}, @code{at76c711}. 10990 10991@item avr31 10992``Classic'' devices with 128@tie{}KiB of program memory. 10993@*@var{mcu}@tie{}= @code{atmega103}, @code{at43usb320}. 10994 10995@item avr35 10996``Classic'' devices with 16@tie{}KiB up to 64@tie{}KiB of program 10997memory and with the @code{MOVW} instruction. 10998@*@var{mcu}@tie{}= @code{atmega16u2}, @code{atmega32u2}, 10999@code{atmega8u2}, @code{attiny167}, @code{at90usb162}, 11000@code{at90usb82}. 11001 11002@item avr4 11003``Enhanced'' devices with up to 8@tie{}KiB of program memory. 11004@*@var{mcu}@tie{}= @code{atmega48}, @code{atmega48a}, 11005@code{atmega48p}, @code{atmega8}, @code{atmega8hva}, 11006@code{atmega8515}, @code{atmega8535}, @code{atmega88}, 11007@code{atmega88a}, @code{atmega88p}, @code{atmega88pa}, 11008@code{at90pwm1}, @code{at90pwm2}, @code{at90pwm2b}, @code{at90pwm3}, 11009@code{at90pwm3b}, @code{at90pwm81}. 11010 11011@item avr5 11012``Enhanced'' devices with 16@tie{}KiB up to 64@tie{}KiB of program memory. 11013@*@var{mcu}@tie{}= @code{atmega16}, @code{atmega16a}, 11014@code{atmega16hva}, @code{atmega16hva2}, @code{atmega16hvb}, 11015@code{atmega16m1}, @code{atmega16u4}, @code{atmega161}, 11016@code{atmega162}, @code{atmega163}, @code{atmega164a}, 11017@code{atmega164p}, @code{atmega165}, @code{atmega165a}, 11018@code{atmega165p}, @code{atmega168}, @code{atmega168a}, 11019@code{atmega168p}, @code{atmega169}, @code{atmega169a}, 11020@code{atmega169p}, @code{atmega169pa}, @code{atmega32}, 11021@code{atmega32c1}, @code{atmega32hvb}, @code{atmega32m1}, 11022@code{atmega32u4}, @code{atmega32u6}, @code{atmega323}, 11023@code{atmega324a}, @code{atmega324p}, @code{atmega324pa}, 11024@code{atmega325}, @code{atmega325a}, @code{atmega325p}, 11025@code{atmega3250}, @code{atmega3250a}, @code{atmega3250p}, 11026@code{atmega328}, @code{atmega328p}, @code{atmega329}, 11027@code{atmega329a}, @code{atmega329p}, @code{atmega329pa}, 11028@code{atmega3290}, @code{atmega3290a}, @code{atmega3290p}, 11029@code{atmega406}, @code{atmega64}, @code{atmega64c1}, 11030@code{atmega64hve}, @code{atmega64m1}, @code{atmega640}, 11031@code{atmega644}, @code{atmega644a}, @code{atmega644p}, 11032@code{atmega644pa}, @code{atmega645}, @code{atmega645a}, 11033@code{atmega645p}, @code{atmega6450}, @code{atmega6450a}, 11034@code{atmega6450p}, @code{atmega649}, @code{atmega649a}, 11035@code{atmega649p}, @code{atmega6490}, @code{at90can32}, 11036@code{at90can64}, @code{at90pwm216}, @code{at90pwm316}, 11037@code{at90scr100}, @code{at90usb646}, @code{at90usb647}, @code{at94k}, 11038@code{m3000}. 11039 11040@item avr51 11041``Enhanced'' devices with 128@tie{}KiB of program memory. 11042@*@var{mcu}@tie{}= @code{atmega128}, @code{atmega128rfa1}, 11043@code{atmega1280}, @code{atmega1281}, @code{atmega1284p}, 11044@code{at90can128}, @code{at90usb1286}, @code{at90usb1287}. 11045 11046@item avr6 11047``Enhanced'' devices with 3-byte PC, i.e.@: with more than 11048128@tie{}KiB of program memory. 11049@*@var{mcu}@tie{}= @code{atmega2560}, @code{atmega2561}. 11050 11051@item avrxmega2 11052``XMEGA'' devices with more than 8@tie{}KiB and up to 64@tie{}KiB of 11053program memory. 11054@*@var{mcu}@tie{}= @code{atxmega16a4}, @code{atxmega16d4}, 11055@code{atxmega16x1}, @code{atxmega32a4}, @code{atxmega32d4}, 11056@code{atxmega32x1}. 11057 11058@item avrxmega4 11059``XMEGA'' devices with more than 64@tie{}KiB and up to 128@tie{}KiB of 11060program memory. 11061@*@var{mcu}@tie{}= @code{atxmega64a3}, @code{atxmega64d3}. 11062 11063@item avrxmega5 11064``XMEGA'' devices with more than 64@tie{}KiB and up to 128@tie{}KiB of 11065program memory and more than 64@tie{}KiB of RAM. 11066@*@var{mcu}@tie{}= @code{atxmega64a1}, @code{atxmega64a1u}. 11067 11068@item avrxmega6 11069``XMEGA'' devices with more than 128@tie{}KiB of program memory. 11070@*@var{mcu}@tie{}= @code{atxmega128a3}, @code{atxmega128d3}, 11071@code{atxmega192a3}, @code{atxmega192d3}, @code{atxmega256a3}, 11072@code{atxmega256a3b}, @code{atxmega256a3bu}, @code{atxmega256d3}. 11073 11074@item avrxmega7 11075``XMEGA'' devices with more than 128@tie{}KiB of program memory and 11076more than 64@tie{}KiB of RAM. 11077@*@var{mcu}@tie{}= @code{atxmega128a1}, @code{atxmega128a1u}. 11078 11079@item avr1 11080This ISA is implemented by the minimal AVR core and supported for 11081assembler only. 11082@*@var{mcu}@tie{}= @code{attiny11}, @code{attiny12}, @code{attiny15}, 11083@code{attiny28}, @code{at90s1200}. 11084 11085@end table 11086 11087@item -maccumulate-args 11088@opindex maccumulate-args 11089Accumulate outgoing function arguments and acquire/release the needed 11090stack space for outgoing function arguments once in function 11091prologue/epilogue. Without this option, outgoing arguments are pushed 11092before calling a function and popped afterwards. 11093 11094Popping the arguments after the function call can be expensive on 11095AVR so that accumulating the stack space might lead to smaller 11096executables because arguments need not to be removed from the 11097stack after such a function call. 11098 11099This option can lead to reduced code size for functions that perform 11100several calls to functions that get their arguments on the stack like 11101calls to printf-like functions. 11102 11103@item -mbranch-cost=@var{cost} 11104@opindex mbranch-cost 11105Set the branch costs for conditional branch instructions to 11106@var{cost}. Reasonable values for @var{cost} are small, non-negative 11107integers. The default branch cost is 0. 11108 11109@item -mcall-prologues 11110@opindex mcall-prologues 11111Functions prologues/epilogues are expanded as calls to appropriate 11112subroutines. Code size is smaller. 11113 11114@item -mint8 11115@opindex mint8 11116Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a 11117@code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes, 11118and @code{long long} is 4 bytes. Please note that this option does not 11119conform to the C standards, but it results in smaller code 11120size. 11121 11122@item -mno-interrupts 11123@opindex mno-interrupts 11124Generated code is not compatible with hardware interrupts. 11125Code size is smaller. 11126 11127@item -mrelax 11128@opindex mrelax 11129Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter 11130@code{RCALL} resp.@: @code{RJMP} instruction if applicable. 11131Setting @code{-mrelax} just adds the @code{--relax} option to the 11132linker command line when the linker is called. 11133 11134Jump relaxing is performed by the linker because jump offsets are not 11135known before code is located. Therefore, the assembler code generated by the 11136compiler is the same, but the instructions in the executable may 11137differ from instructions in the assembler code. 11138 11139Relaxing must be turned on if linker stubs are needed, see the 11140section on @code{EIND} and linker stubs below. 11141 11142@item -mshort-calls 11143@opindex mshort-calls 11144This option has been deprecated and will be removed in GCC 4.8. 11145See @code{-mrelax} for a replacement. 11146 11147Use @code{RCALL}/@code{RJMP} instructions even on devices with 1114816@tie{}KiB or more of program memory, i.e.@: on devices that 11149have the @code{CALL} and @code{JMP} instructions. 11150 11151@item -msp8 11152@opindex msp8 11153Treat the stack pointer register as an 8-bit register, 11154i.e.@: assume the high byte of the stack pointer is zero. 11155In general, you don't need to set this option by hand. 11156 11157This option is used internally by the compiler to select and 11158build multilibs for architectures @code{avr2} and @code{avr25}. 11159These architectures mix devices with and without @code{SPH}. 11160For any setting other than @code{-mmcu=avr2} or @code{-mmcu=avr25} 11161the compiler driver will add or remove this option from the compiler 11162proper's command line, because the compiler then knows if the device 11163or architecture has an 8-bit stack pointer and thus no @code{SPH} 11164register or not. 11165 11166@item -mstrict-X 11167@opindex mstrict-X 11168Use address register @code{X} in a way proposed by the hardware. This means 11169that @code{X} is only used in indirect, post-increment or 11170pre-decrement addressing. 11171 11172Without this option, the @code{X} register may be used in the same way 11173as @code{Y} or @code{Z} which then is emulated by additional 11174instructions. 11175For example, loading a value with @code{X+const} addressing with a 11176small non-negative @code{const < 64} to a register @var{Rn} is 11177performed as 11178 11179@example 11180adiw r26, const ; X += const 11181ld @var{Rn}, X ; @var{Rn} = *X 11182sbiw r26, const ; X -= const 11183@end example 11184 11185@item -mtiny-stack 11186@opindex mtiny-stack 11187Only change the lower 8@tie{}bits of the stack pointer. 11188@end table 11189 11190@subsubsection @code{EIND} and Devices with more than 128 Ki Bytes of Flash 11191@cindex @code{EIND} 11192Pointers in the implementation are 16@tie{}bits wide. 11193The address of a function or label is represented as word address so 11194that indirect jumps and calls can target any code address in the 11195range of 64@tie{}Ki words. 11196 11197In order to facilitate indirect jump on devices with more than 128@tie{}Ki 11198bytes of program memory space, there is a special function register called 11199@code{EIND} that serves as most significant part of the target address 11200when @code{EICALL} or @code{EIJMP} instructions are used. 11201 11202Indirect jumps and calls on these devices are handled as follows by 11203the compiler and are subject to some limitations: 11204 11205@itemize @bullet 11206 11207@item 11208The compiler never sets @code{EIND}. 11209 11210@item 11211The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP} 11212instructions or might read @code{EIND} directly in order to emulate an 11213indirect call/jump by means of a @code{RET} instruction. 11214 11215@item 11216The compiler assumes that @code{EIND} never changes during the startup 11217code or during the application. In particular, @code{EIND} is not 11218saved/restored in function or interrupt service routine 11219prologue/epilogue. 11220 11221@item 11222For indirect calls to functions and computed goto, the linker 11223generates @emph{stubs}. Stubs are jump pads sometimes also called 11224@emph{trampolines}. Thus, the indirect call/jump jumps to such a stub. 11225The stub contains a direct jump to the desired address. 11226 11227@item 11228Linker relaxation must be turned on so that the linker will generate 11229the stubs correctly an all situaltion. See the compiler option 11230@code{-mrelax} and the linler option @code{--relax}. 11231There are corner cases where the linker is supposed to generate stubs 11232but aborts without relaxation and without a helpful error message. 11233 11234@item 11235The default linker script is arranged for code with @code{EIND = 0}. 11236If code is supposed to work for a setup with @code{EIND != 0}, a custom 11237linker script has to be used in order to place the sections whose 11238name start with @code{.trampolines} into the segment where @code{EIND} 11239points to. 11240 11241@item 11242The startup code from libgcc never sets @code{EIND}. 11243Notice that startup code is a blend of code from libgcc and AVR-LibC. 11244For the impact of AVR-LibC on @code{EIND}, see the 11245@w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}. 11246 11247@item 11248It is legitimate for user-specific startup code to set up @code{EIND} 11249early, for example by means of initialization code located in 11250section @code{.init3}. Such code runs prior to general startup code 11251that initializes RAM and calls constructors, but after the bit 11252of startup code from AVR-LibC that sets @code{EIND} to the segment 11253where the vector table is located. 11254@example 11255#include <avr/io.h> 11256 11257static void 11258__attribute__((section(".init3"),naked,used,no_instrument_function)) 11259init3_set_eind (void) 11260@{ 11261 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t" 11262 "out %i0,r24" :: "n" (&EIND) : "r24","memory"); 11263@} 11264@end example 11265 11266@noindent 11267The @code{__trampolines_start} symbol is defined in the linker script. 11268 11269@item 11270Stubs are generated automatically by the linker if 11271the following two conditions are met: 11272@itemize @minus 11273 11274@item The address of a label is taken by means of the @code{gs} modifier 11275(short for @emph{generate stubs}) like so: 11276@example 11277LDI r24, lo8(gs(@var{func})) 11278LDI r25, hi8(gs(@var{func})) 11279@end example 11280@item The final location of that label is in a code segment 11281@emph{outside} the segment where the stubs are located. 11282@end itemize 11283 11284@item 11285The compiler emits such @code{gs} modifiers for code labels in the 11286following situations: 11287@itemize @minus 11288@item Taking address of a function or code label. 11289@item Computed goto. 11290@item If prologue-save function is used, see @option{-mcall-prologues} 11291command-line option. 11292@item Switch/case dispatch tables. If you do not want such dispatch 11293tables you can specify the @option{-fno-jump-tables} command-line option. 11294@item C and C++ constructors/destructors called during startup/shutdown. 11295@item If the tools hit a @code{gs()} modifier explained above. 11296@end itemize 11297 11298@item 11299Jumping to non-symbolic addresses like so is @emph{not} supported: 11300 11301@example 11302int main (void) 11303@{ 11304 /* Call function at word address 0x2 */ 11305 return ((int(*)(void)) 0x2)(); 11306@} 11307@end example 11308 11309Instead, a stub has to be set up, i.e.@: the function has to be called 11310through a symbol (@code{func_4} in the example): 11311 11312@example 11313int main (void) 11314@{ 11315 extern int func_4 (void); 11316 11317 /* Call function at byte address 0x4 */ 11318 return func_4(); 11319@} 11320@end example 11321 11322and the application be linked with @code{-Wl,--defsym,func_4=0x4}. 11323Alternatively, @code{func_4} can be defined in the linker script. 11324@end itemize 11325 11326@subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers 11327@cindex @code{RAMPD} 11328@cindex @code{RAMPX} 11329@cindex @code{RAMPY} 11330@cindex @code{RAMPZ} 11331Some AVR devices support memories larger than the 64@tie{}KiB range 11332that can be accessed with 16-bit pointers. To access memory locations 11333outside this 64@tie{}KiB range, the contentent of a @code{RAMP} 11334register is used as high part of the address: 11335The @code{X}, @code{Y}, @code{Z} address register is concatenated 11336with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function 11337register, respectively, to get a wide address. Similarly, 11338@code{RAMPD} is used together with direct addressing. 11339 11340@itemize 11341@item 11342The startup code initializes the @code{RAMP} special function 11343registers with zero. 11344 11345@item 11346If a @ref{AVR Named Address Spaces,named address space} other than 11347generic or @code{__flash} is used, then @code{RAMPZ} is set 11348as needed before the operation. 11349 11350@item 11351If the device supports RAM larger than 64@tie{KiB} and the compiler 11352needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ} 11353is reset to zero after the operation. 11354 11355@item 11356If the device comes with a specific @code{RAMP} register, the ISR 11357prologue/epilogue saves/restores that SFR and initializes it with 11358zero in case the ISR code might (implicitly) use it. 11359 11360@item 11361RAM larger than 64@tie{KiB} is not supported by GCC for AVR targets. 11362If you use inline assembler to read from locations outside the 1136316-bit address range and change one of the @code{RAMP} registers, 11364you must reset it to zero after the access. 11365 11366@end itemize 11367 11368@subsubsection AVR Built-in Macros 11369 11370GCC defines several built-in macros so that the user code can test 11371for the presence or absence of features. Almost any of the following 11372built-in macros are deduced from device capabilities and thus 11373triggered by the @code{-mmcu=} command-line option. 11374 11375For even more AVR-specific built-in macros see 11376@ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}. 11377 11378@table @code 11379 11380@item __AVR_ARCH__ 11381Build-in macro that resolves to a decimal number that identifies the 11382architecture and depends on the @code{-mmcu=@var{mcu}} option. 11383Possible values are: 11384 11385@code{2}, @code{25}, @code{3}, @code{31}, @code{35}, 11386@code{4}, @code{5}, @code{51}, @code{6}, @code{102}, @code{104}, 11387@code{105}, @code{106}, @code{107} 11388 11389for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, 11390@code{avr31}, @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, 11391@code{avr6}, @code{avrxmega2}, @code{avrxmega4}, @code{avrxmega5}, 11392@code{avrxmega6}, @code{avrxmega7}, respectively. 11393If @var{mcu} specifies a device, this built-in macro is set 11394accordingly. For example, with @code{-mmcu=atmega8} the macro will be 11395defined to @code{4}. 11396 11397@item __AVR_@var{Device}__ 11398Setting @code{-mmcu=@var{device}} defines this built-in macro which reflects 11399the device's name. For example, @code{-mmcu=atmega8} defines the 11400built-in macro @code{__AVR_ATmega8__}, @code{-mmcu=attiny261a} defines 11401@code{__AVR_ATtiny261A__}, etc. 11402 11403The built-in macros' names follow 11404the scheme @code{__AVR_@var{Device}__} where @var{Device} is 11405the device name as from the AVR user manual. The difference between 11406@var{Device} in the built-in macro and @var{device} in 11407@code{-mmcu=@var{device}} is that the latter is always lowercase. 11408 11409If @var{device} is not a device but only a core architecture like 11410@code{avr51}, this macro will not be defined. 11411 11412@item __AVR_XMEGA__ 11413The device/architecture belongs to the XMEGA family of devices. 11414 11415@item __AVR_HAVE_ELPM__ 11416The device has the the @code{ELPM} instruction. 11417 11418@item __AVR_HAVE_ELPMX__ 11419The device has the @code{ELPM R@var{n},Z} and @code{ELPM 11420R@var{n},Z+} instructions. 11421 11422@item __AVR_HAVE_MOVW__ 11423The device has the @code{MOVW} instruction to perform 16-bit 11424register-register moves. 11425 11426@item __AVR_HAVE_LPMX__ 11427The device has the @code{LPM R@var{n},Z} and 11428@code{LPM R@var{n},Z+} instructions. 11429 11430@item __AVR_HAVE_MUL__ 11431The device has a hardware multiplier. 11432 11433@item __AVR_HAVE_JMP_CALL__ 11434The device has the @code{JMP} and @code{CALL} instructions. 11435This is the case for devices with at least 16@tie{}KiB of program 11436memory and if @code{-mshort-calls} is not set. 11437 11438@item __AVR_HAVE_EIJMP_EICALL__ 11439@item __AVR_3_BYTE_PC__ 11440The device has the @code{EIJMP} and @code{EICALL} instructions. 11441This is the case for devices with more than 128@tie{}KiB of program memory. 11442This also means that the program counter 11443(PC) is 3@tie{}bytes wide. 11444 11445@item __AVR_2_BYTE_PC__ 11446The program counter (PC) is 2@tie{}bytes wide. This is the case for devices 11447with up to 128@tie{}KiB of program memory. 11448 11449@item __AVR_HAVE_8BIT_SP__ 11450@item __AVR_HAVE_16BIT_SP__ 11451The stack pointer (SP) register is treated as 8-bit respectively 1145216-bit register by the compiler. 11453The definition of these macros is affected by @code{-mtiny-stack}. 11454 11455@item __AVR_HAVE_SPH__ 11456@item __AVR_SP8__ 11457The device has the SPH (high part of stack pointer) special function 11458register or has an 8-bit stack pointer, respectively. 11459The definition of these macros is affected by @code{-mmcu=} and 11460in the cases of @code{-mmcu=avr2} and @code{-mmcu=avr25} also 11461by @code{-msp8}. 11462 11463@item __AVR_HAVE_RAMPD__ 11464@item __AVR_HAVE_RAMPX__ 11465@item __AVR_HAVE_RAMPY__ 11466@item __AVR_HAVE_RAMPZ__ 11467The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY}, 11468@code{RAMPZ} special function register, respectively. 11469 11470@item __NO_INTERRUPTS__ 11471This macro reflects the @code{-mno-interrupts} command line option. 11472 11473@item __AVR_ERRATA_SKIP__ 11474@item __AVR_ERRATA_SKIP_JMP_CALL__ 11475Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit 11476instructions because of a hardware erratum. Skip instructions are 11477@code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}. 11478The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also 11479set. 11480 11481@item __AVR_SFR_OFFSET__=@var{offset} 11482Instructions that can address I/O special function registers directly 11483like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different 11484address as if addressed by an instruction to access RAM like @code{LD} 11485or @code{STS}. This offset depends on the device architecture and has 11486to be subtracted from the RAM address in order to get the 11487respective I/O@tie{}address. 11488 11489@item __WITH_AVRLIBC__ 11490The compiler is configured to be used together with AVR-Libc. 11491See the @code{--with-avrlibc} configure option. 11492 11493@end table 11494 11495@node Blackfin Options 11496@subsection Blackfin Options 11497@cindex Blackfin Options 11498 11499@table @gcctabopt 11500@item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} 11501@opindex mcpu= 11502Specifies the name of the target Blackfin processor. Currently, @var{cpu} 11503can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518}, 11504@samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526}, 11505@samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533}, 11506@samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539}, 11507@samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549}, 11508@samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m}, 11509@samp{bf561}, @samp{bf592}. 11510The optional @var{sirevision} specifies the silicon revision of the target 11511Blackfin processor. Any workarounds available for the targeted silicon revision 11512will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled. 11513If @var{sirevision} is @samp{any}, all workarounds for the targeted processor 11514will be enabled. The @code{__SILICON_REVISION__} macro is defined to two 11515hexadecimal digits representing the major and minor numbers in the silicon 11516revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__} 11517is not defined. If @var{sirevision} is @samp{any}, the 11518@code{__SILICON_REVISION__} is defined to be @code{0xffff}. 11519If this optional @var{sirevision} is not used, GCC assumes the latest known 11520silicon revision of the targeted Blackfin processor. 11521 11522Support for @samp{bf561} is incomplete. For @samp{bf561}, 11523Only the processor macro is defined. 11524Without this option, @samp{bf532} is used as the processor by default. 11525The corresponding predefined processor macros for @var{cpu} is to 11526be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP 11527provided by libgloss to be linked in if @option{-msim} is not given. 11528 11529@item -msim 11530@opindex msim 11531Specifies that the program will be run on the simulator. This causes 11532the simulator BSP provided by libgloss to be linked in. This option 11533has effect only for @samp{bfin-elf} toolchain. 11534Certain other options, such as @option{-mid-shared-library} and 11535@option{-mfdpic}, imply @option{-msim}. 11536 11537@item -momit-leaf-frame-pointer 11538@opindex momit-leaf-frame-pointer 11539Don't keep the frame pointer in a register for leaf functions. This 11540avoids the instructions to save, set up and restore frame pointers and 11541makes an extra register available in leaf functions. The option 11542@option{-fomit-frame-pointer} removes the frame pointer for all functions, 11543which might make debugging harder. 11544 11545@item -mspecld-anomaly 11546@opindex mspecld-anomaly 11547When enabled, the compiler will ensure that the generated code does not 11548contain speculative loads after jump instructions. If this option is used, 11549@code{__WORKAROUND_SPECULATIVE_LOADS} is defined. 11550 11551@item -mno-specld-anomaly 11552@opindex mno-specld-anomaly 11553Don't generate extra code to prevent speculative loads from occurring. 11554 11555@item -mcsync-anomaly 11556@opindex mcsync-anomaly 11557When enabled, the compiler will ensure that the generated code does not 11558contain CSYNC or SSYNC instructions too soon after conditional branches. 11559If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined. 11560 11561@item -mno-csync-anomaly 11562@opindex mno-csync-anomaly 11563Don't generate extra code to prevent CSYNC or SSYNC instructions from 11564occurring too soon after a conditional branch. 11565 11566@item -mlow-64k 11567@opindex mlow-64k 11568When enabled, the compiler is free to take advantage of the knowledge that 11569the entire program fits into the low 64k of memory. 11570 11571@item -mno-low-64k 11572@opindex mno-low-64k 11573Assume that the program is arbitrarily large. This is the default. 11574 11575@item -mstack-check-l1 11576@opindex mstack-check-l1 11577Do stack checking using information placed into L1 scratchpad memory by the 11578uClinux kernel. 11579 11580@item -mid-shared-library 11581@opindex mid-shared-library 11582Generate code that supports shared libraries via the library ID method. 11583This allows for execute in place and shared libraries in an environment 11584without virtual memory management. This option implies @option{-fPIC}. 11585With a @samp{bfin-elf} target, this option implies @option{-msim}. 11586 11587@item -mno-id-shared-library 11588@opindex mno-id-shared-library 11589Generate code that doesn't assume ID based shared libraries are being used. 11590This is the default. 11591 11592@item -mleaf-id-shared-library 11593@opindex mleaf-id-shared-library 11594Generate code that supports shared libraries via the library ID method, 11595but assumes that this library or executable won't link against any other 11596ID shared libraries. That allows the compiler to use faster code for jumps 11597and calls. 11598 11599@item -mno-leaf-id-shared-library 11600@opindex mno-leaf-id-shared-library 11601Do not assume that the code being compiled won't link against any ID shared 11602libraries. Slower code will be generated for jump and call insns. 11603 11604@item -mshared-library-id=n 11605@opindex mshared-library-id 11606Specified the identification number of the ID based shared library being 11607compiled. Specifying a value of 0 will generate more compact code, specifying 11608other values will force the allocation of that number to the current 11609library but is no more space or time efficient than omitting this option. 11610 11611@item -msep-data 11612@opindex msep-data 11613Generate code that allows the data segment to be located in a different 11614area of memory from the text segment. This allows for execute in place in 11615an environment without virtual memory management by eliminating relocations 11616against the text section. 11617 11618@item -mno-sep-data 11619@opindex mno-sep-data 11620Generate code that assumes that the data segment follows the text segment. 11621This is the default. 11622 11623@item -mlong-calls 11624@itemx -mno-long-calls 11625@opindex mlong-calls 11626@opindex mno-long-calls 11627Tells the compiler to perform function calls by first loading the 11628address of the function into a register and then performing a subroutine 11629call on this register. This switch is needed if the target function 11630lies outside of the 24-bit addressing range of the offset-based 11631version of subroutine call instruction. 11632 11633This feature is not enabled by default. Specifying 11634@option{-mno-long-calls} will restore the default behavior. Note these 11635switches have no effect on how the compiler generates code to handle 11636function calls via function pointers. 11637 11638@item -mfast-fp 11639@opindex mfast-fp 11640Link with the fast floating-point library. This library relaxes some of 11641the IEEE floating-point standard's rules for checking inputs against 11642Not-a-Number (NAN), in the interest of performance. 11643 11644@item -minline-plt 11645@opindex minline-plt 11646Enable inlining of PLT entries in function calls to functions that are 11647not known to bind locally. It has no effect without @option{-mfdpic}. 11648 11649@item -mmulticore 11650@opindex mmulticore 11651Build standalone application for multicore Blackfin processor. Proper 11652start files and link scripts will be used to support multicore. 11653This option defines @code{__BFIN_MULTICORE}. It can only be used with 11654@option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with 11655@option{-mcorea} or @option{-mcoreb}. If it's used without 11656@option{-mcorea} or @option{-mcoreb}, single application/dual core 11657programming model is used. In this model, the main function of Core B 11658should be named as coreb_main. If it's used with @option{-mcorea} or 11659@option{-mcoreb}, one application per core programming model is used. 11660If this option is not used, single core application programming 11661model is used. 11662 11663@item -mcorea 11664@opindex mcorea 11665Build standalone application for Core A of BF561 when using 11666one application per core programming model. Proper start files 11667and link scripts will be used to support Core A. This option 11668defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}. 11669 11670@item -mcoreb 11671@opindex mcoreb 11672Build standalone application for Core B of BF561 when using 11673one application per core programming model. Proper start files 11674and link scripts will be used to support Core B. This option 11675defines @code{__BFIN_COREB}. When this option is used, coreb_main 11676should be used instead of main. It must be used with 11677@option{-mmulticore}. 11678 11679@item -msdram 11680@opindex msdram 11681Build standalone application for SDRAM. Proper start files and 11682link scripts will be used to put the application into SDRAM. 11683Loader should initialize SDRAM before loading the application 11684into SDRAM. This option defines @code{__BFIN_SDRAM}. 11685 11686@item -micplb 11687@opindex micplb 11688Assume that ICPLBs are enabled at run time. This has an effect on certain 11689anomaly workarounds. For Linux targets, the default is to assume ICPLBs 11690are enabled; for standalone applications the default is off. 11691@end table 11692 11693@node C6X Options 11694@subsection C6X Options 11695@cindex C6X Options 11696 11697@table @gcctabopt 11698@item -march=@var{name} 11699@opindex march 11700This specifies the name of the target architecture. GCC uses this 11701name to determine what kind of instructions it can emit when generating 11702assembly code. Permissible names are: @samp{c62x}, 11703@samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}. 11704 11705@item -mbig-endian 11706@opindex mbig-endian 11707Generate code for a big-endian target. 11708 11709@item -mlittle-endian 11710@opindex mlittle-endian 11711Generate code for a little-endian target. This is the default. 11712 11713@item -msim 11714@opindex msim 11715Choose startup files and linker script suitable for the simulator. 11716 11717@item -msdata=default 11718@opindex msdata=default 11719Put small global and static data in the @samp{.neardata} section, 11720which is pointed to by register @code{B14}. Put small uninitialized 11721global and static data in the @samp{.bss} section, which is adjacent 11722to the @samp{.neardata} section. Put small read-only data into the 11723@samp{.rodata} section. The corresponding sections used for large 11724pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}. 11725 11726@item -msdata=all 11727@opindex msdata=all 11728Put all data, not just small objets, into the sections reserved for 11729small data, and use addressing relative to the @code{B14} register to 11730access them. 11731 11732@item -msdata=none 11733@opindex msdata=none 11734Make no use of the sections reserved for small data, and use absolute 11735addresses to access all data. Put all initialized global and static 11736data in the @samp{.fardata} section, and all uninitialized data in the 11737@samp{.far} section. Put all constant data into the @samp{.const} 11738section. 11739@end table 11740 11741@node CRIS Options 11742@subsection CRIS Options 11743@cindex CRIS Options 11744 11745These options are defined specifically for the CRIS ports. 11746 11747@table @gcctabopt 11748@item -march=@var{architecture-type} 11749@itemx -mcpu=@var{architecture-type} 11750@opindex march 11751@opindex mcpu 11752Generate code for the specified architecture. The choices for 11753@var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for 11754respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@. 11755Default is @samp{v0} except for cris-axis-linux-gnu, where the default is 11756@samp{v10}. 11757 11758@item -mtune=@var{architecture-type} 11759@opindex mtune 11760Tune to @var{architecture-type} everything applicable about the generated 11761code, except for the ABI and the set of available instructions. The 11762choices for @var{architecture-type} are the same as for 11763@option{-march=@var{architecture-type}}. 11764 11765@item -mmax-stack-frame=@var{n} 11766@opindex mmax-stack-frame 11767Warn when the stack frame of a function exceeds @var{n} bytes. 11768 11769@item -metrax4 11770@itemx -metrax100 11771@opindex metrax4 11772@opindex metrax100 11773The options @option{-metrax4} and @option{-metrax100} are synonyms for 11774@option{-march=v3} and @option{-march=v8} respectively. 11775 11776@item -mmul-bug-workaround 11777@itemx -mno-mul-bug-workaround 11778@opindex mmul-bug-workaround 11779@opindex mno-mul-bug-workaround 11780Work around a bug in the @code{muls} and @code{mulu} instructions for CPU 11781models where it applies. This option is active by default. 11782 11783@item -mpdebug 11784@opindex mpdebug 11785Enable CRIS-specific verbose debug-related information in the assembly 11786code. This option also has the effect to turn off the @samp{#NO_APP} 11787formatted-code indicator to the assembler at the beginning of the 11788assembly file. 11789 11790@item -mcc-init 11791@opindex mcc-init 11792Do not use condition-code results from previous instruction; always emit 11793compare and test instructions before use of condition codes. 11794 11795@item -mno-side-effects 11796@opindex mno-side-effects 11797Do not emit instructions with side-effects in addressing modes other than 11798post-increment. 11799 11800@item -mstack-align 11801@itemx -mno-stack-align 11802@itemx -mdata-align 11803@itemx -mno-data-align 11804@itemx -mconst-align 11805@itemx -mno-const-align 11806@opindex mstack-align 11807@opindex mno-stack-align 11808@opindex mdata-align 11809@opindex mno-data-align 11810@opindex mconst-align 11811@opindex mno-const-align 11812These options (no-options) arranges (eliminate arrangements) for the 11813stack-frame, individual data and constants to be aligned for the maximum 11814single data access size for the chosen CPU model. The default is to 11815arrange for 32-bit alignment. ABI details such as structure layout are 11816not affected by these options. 11817 11818@item -m32-bit 11819@itemx -m16-bit 11820@itemx -m8-bit 11821@opindex m32-bit 11822@opindex m16-bit 11823@opindex m8-bit 11824Similar to the stack- data- and const-align options above, these options 11825arrange for stack-frame, writable data and constants to all be 32-bit, 1182616-bit or 8-bit aligned. The default is 32-bit alignment. 11827 11828@item -mno-prologue-epilogue 11829@itemx -mprologue-epilogue 11830@opindex mno-prologue-epilogue 11831@opindex mprologue-epilogue 11832With @option{-mno-prologue-epilogue}, the normal function prologue and 11833epilogue which set up the stack frame are omitted and no return 11834instructions or return sequences are generated in the code. Use this 11835option only together with visual inspection of the compiled code: no 11836warnings or errors are generated when call-saved registers must be saved, 11837or storage for local variable needs to be allocated. 11838 11839@item -mno-gotplt 11840@itemx -mgotplt 11841@opindex mno-gotplt 11842@opindex mgotplt 11843With @option{-fpic} and @option{-fPIC}, don't generate (do generate) 11844instruction sequences that load addresses for functions from the PLT part 11845of the GOT rather than (traditional on other architectures) calls to the 11846PLT@. The default is @option{-mgotplt}. 11847 11848@item -melf 11849@opindex melf 11850Legacy no-op option only recognized with the cris-axis-elf and 11851cris-axis-linux-gnu targets. 11852 11853@item -mlinux 11854@opindex mlinux 11855Legacy no-op option only recognized with the cris-axis-linux-gnu target. 11856 11857@item -sim 11858@opindex sim 11859This option, recognized for the cris-axis-elf arranges 11860to link with input-output functions from a simulator library. Code, 11861initialized data and zero-initialized data are allocated consecutively. 11862 11863@item -sim2 11864@opindex sim2 11865Like @option{-sim}, but pass linker options to locate initialized data at 118660x40000000 and zero-initialized data at 0x80000000. 11867@end table 11868 11869@node CR16 Options 11870@subsection CR16 Options 11871@cindex CR16 Options 11872 11873These options are defined specifically for the CR16 ports. 11874 11875@table @gcctabopt 11876 11877@item -mmac 11878@opindex mmac 11879Enable the use of multiply-accumulate instructions. Disabled by default. 11880 11881@item -mcr16cplus 11882@itemx -mcr16c 11883@opindex mcr16cplus 11884@opindex mcr16c 11885Generate code for CR16C or CR16C+ architecture. CR16C+ architecture 11886is default. 11887 11888@item -msim 11889@opindex msim 11890Links the library libsim.a which is in compatible with simulator. Applicable 11891to elf compiler only. 11892 11893@item -mint32 11894@opindex mint32 11895Choose integer type as 32-bit wide. 11896 11897@item -mbit-ops 11898@opindex mbit-ops 11899Generates sbit/cbit instructions for bit manipulations. 11900 11901@item -mdata-model=@var{model} 11902@opindex mdata-model 11903Choose a data model. The choices for @var{model} are @samp{near}, 11904@samp{far} or @samp{medium}. @samp{medium} is default. 11905However, @samp{far} is not valid when -mcr16c option is chosen as 11906CR16C architecture does not support far data model. 11907@end table 11908 11909@node Darwin Options 11910@subsection Darwin Options 11911@cindex Darwin options 11912 11913These options are defined for all architectures running the Darwin operating 11914system. 11915 11916FSF GCC on Darwin does not create ``fat'' object files; it will create 11917an object file for the single architecture that it was built to 11918target. Apple's GCC on Darwin does create ``fat'' files if multiple 11919@option{-arch} options are used; it does so by running the compiler or 11920linker multiple times and joining the results together with 11921@file{lipo}. 11922 11923The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or 11924@samp{i686}) is determined by the flags that specify the ISA 11925that GCC is targetting, like @option{-mcpu} or @option{-march}. The 11926@option{-force_cpusubtype_ALL} option can be used to override this. 11927 11928The Darwin tools vary in their behavior when presented with an ISA 11929mismatch. The assembler, @file{as}, will only permit instructions to 11930be used that are valid for the subtype of the file it is generating, 11931so you cannot put 64-bit instructions in a @samp{ppc750} object file. 11932The linker for shared libraries, @file{/usr/bin/libtool}, will fail 11933and print an error if asked to create a shared library with a less 11934restrictive subtype than its input files (for instance, trying to put 11935a @samp{ppc970} object file in a @samp{ppc7400} library). The linker 11936for executables, @file{ld}, will quietly give the executable the most 11937restrictive subtype of any of its input files. 11938 11939@table @gcctabopt 11940@item -F@var{dir} 11941@opindex F 11942Add the framework directory @var{dir} to the head of the list of 11943directories to be searched for header files. These directories are 11944interleaved with those specified by @option{-I} options and are 11945scanned in a left-to-right order. 11946 11947A framework directory is a directory with frameworks in it. A 11948framework is a directory with a @samp{"Headers"} and/or 11949@samp{"PrivateHeaders"} directory contained directly in it that ends 11950in @samp{".framework"}. The name of a framework is the name of this 11951directory excluding the @samp{".framework"}. Headers associated with 11952the framework are found in one of those two directories, with 11953@samp{"Headers"} being searched first. A subframework is a framework 11954directory that is in a framework's @samp{"Frameworks"} directory. 11955Includes of subframework headers can only appear in a header of a 11956framework that contains the subframework, or in a sibling subframework 11957header. Two subframeworks are siblings if they occur in the same 11958framework. A subframework should not have the same name as a 11959framework, a warning will be issued if this is violated. Currently a 11960subframework cannot have subframeworks, in the future, the mechanism 11961may be extended to support this. The standard frameworks can be found 11962in @samp{"/System/Library/Frameworks"} and 11963@samp{"/Library/Frameworks"}. An example include looks like 11964@code{#include <Framework/header.h>}, where @samp{Framework} denotes 11965the name of the framework and header.h is found in the 11966@samp{"PrivateHeaders"} or @samp{"Headers"} directory. 11967 11968@item -iframework@var{dir} 11969@opindex iframework 11970Like @option{-F} except the directory is a treated as a system 11971directory. The main difference between this @option{-iframework} and 11972@option{-F} is that with @option{-iframework} the compiler does not 11973warn about constructs contained within header files found via 11974@var{dir}. This option is valid only for the C family of languages. 11975 11976@item -gused 11977@opindex gused 11978Emit debugging information for symbols that are used. For STABS 11979debugging format, this enables @option{-feliminate-unused-debug-symbols}. 11980This is by default ON@. 11981 11982@item -gfull 11983@opindex gfull 11984Emit debugging information for all symbols and types. 11985 11986@item -mmacosx-version-min=@var{version} 11987The earliest version of MacOS X that this executable will run on 11988is @var{version}. Typical values of @var{version} include @code{10.1}, 11989@code{10.2}, and @code{10.3.9}. 11990 11991If the compiler was built to use the system's headers by default, 11992then the default for this option is the system version on which the 11993compiler is running, otherwise the default is to make choices that 11994are compatible with as many systems and code bases as possible. 11995 11996@item -mkernel 11997@opindex mkernel 11998Enable kernel development mode. The @option{-mkernel} option sets 11999@option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit}, 12000@option{-fno-exceptions}, @option{-fno-non-call-exceptions}, 12001@option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where 12002applicable. This mode also sets @option{-mno-altivec}, 12003@option{-msoft-float}, @option{-fno-builtin} and 12004@option{-mlong-branch} for PowerPC targets. 12005 12006@item -mone-byte-bool 12007@opindex mone-byte-bool 12008Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}. 12009By default @samp{sizeof(bool)} is @samp{4} when compiling for 12010Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this 12011option has no effect on x86. 12012 12013@strong{Warning:} The @option{-mone-byte-bool} switch causes GCC 12014to generate code that is not binary compatible with code generated 12015without that switch. Using this switch may require recompiling all 12016other modules in a program, including system libraries. Use this 12017switch to conform to a non-default data model. 12018 12019@item -mfix-and-continue 12020@itemx -ffix-and-continue 12021@itemx -findirect-data 12022@opindex mfix-and-continue 12023@opindex ffix-and-continue 12024@opindex findirect-data 12025Generate code suitable for fast turn around development. Needed to 12026enable gdb to dynamically load @code{.o} files into already running 12027programs. @option{-findirect-data} and @option{-ffix-and-continue} 12028are provided for backwards compatibility. 12029 12030@item -all_load 12031@opindex all_load 12032Loads all members of static archive libraries. 12033See man ld(1) for more information. 12034 12035@item -arch_errors_fatal 12036@opindex arch_errors_fatal 12037Cause the errors having to do with files that have the wrong architecture 12038to be fatal. 12039 12040@item -bind_at_load 12041@opindex bind_at_load 12042Causes the output file to be marked such that the dynamic linker will 12043bind all undefined references when the file is loaded or launched. 12044 12045@item -bundle 12046@opindex bundle 12047Produce a Mach-o bundle format file. 12048See man ld(1) for more information. 12049 12050@item -bundle_loader @var{executable} 12051@opindex bundle_loader 12052This option specifies the @var{executable} that will be loading the build 12053output file being linked. See man ld(1) for more information. 12054 12055@item -dynamiclib 12056@opindex dynamiclib 12057When passed this option, GCC will produce a dynamic library instead of 12058an executable when linking, using the Darwin @file{libtool} command. 12059 12060@item -force_cpusubtype_ALL 12061@opindex force_cpusubtype_ALL 12062This causes GCC's output file to have the @var{ALL} subtype, instead of 12063one controlled by the @option{-mcpu} or @option{-march} option. 12064 12065@item -allowable_client @var{client_name} 12066@itemx -client_name 12067@itemx -compatibility_version 12068@itemx -current_version 12069@itemx -dead_strip 12070@itemx -dependency-file 12071@itemx -dylib_file 12072@itemx -dylinker_install_name 12073@itemx -dynamic 12074@itemx -exported_symbols_list 12075@itemx -filelist 12076@need 800 12077@itemx -flat_namespace 12078@itemx -force_flat_namespace 12079@itemx -headerpad_max_install_names 12080@itemx -image_base 12081@itemx -init 12082@itemx -install_name 12083@itemx -keep_private_externs 12084@itemx -multi_module 12085@itemx -multiply_defined 12086@itemx -multiply_defined_unused 12087@need 800 12088@itemx -noall_load 12089@itemx -no_dead_strip_inits_and_terms 12090@itemx -nofixprebinding 12091@itemx -nomultidefs 12092@itemx -noprebind 12093@itemx -noseglinkedit 12094@itemx -pagezero_size 12095@itemx -prebind 12096@itemx -prebind_all_twolevel_modules 12097@itemx -private_bundle 12098@need 800 12099@itemx -read_only_relocs 12100@itemx -sectalign 12101@itemx -sectobjectsymbols 12102@itemx -whyload 12103@itemx -seg1addr 12104@itemx -sectcreate 12105@itemx -sectobjectsymbols 12106@itemx -sectorder 12107@itemx -segaddr 12108@itemx -segs_read_only_addr 12109@need 800 12110@itemx -segs_read_write_addr 12111@itemx -seg_addr_table 12112@itemx -seg_addr_table_filename 12113@itemx -seglinkedit 12114@itemx -segprot 12115@itemx -segs_read_only_addr 12116@itemx -segs_read_write_addr 12117@itemx -single_module 12118@itemx -static 12119@itemx -sub_library 12120@need 800 12121@itemx -sub_umbrella 12122@itemx -twolevel_namespace 12123@itemx -umbrella 12124@itemx -undefined 12125@itemx -unexported_symbols_list 12126@itemx -weak_reference_mismatches 12127@itemx -whatsloaded 12128@opindex allowable_client 12129@opindex client_name 12130@opindex compatibility_version 12131@opindex current_version 12132@opindex dead_strip 12133@opindex dependency-file 12134@opindex dylib_file 12135@opindex dylinker_install_name 12136@opindex dynamic 12137@opindex exported_symbols_list 12138@opindex filelist 12139@opindex flat_namespace 12140@opindex force_flat_namespace 12141@opindex headerpad_max_install_names 12142@opindex image_base 12143@opindex init 12144@opindex install_name 12145@opindex keep_private_externs 12146@opindex multi_module 12147@opindex multiply_defined 12148@opindex multiply_defined_unused 12149@opindex noall_load 12150@opindex no_dead_strip_inits_and_terms 12151@opindex nofixprebinding 12152@opindex nomultidefs 12153@opindex noprebind 12154@opindex noseglinkedit 12155@opindex pagezero_size 12156@opindex prebind 12157@opindex prebind_all_twolevel_modules 12158@opindex private_bundle 12159@opindex read_only_relocs 12160@opindex sectalign 12161@opindex sectobjectsymbols 12162@opindex whyload 12163@opindex seg1addr 12164@opindex sectcreate 12165@opindex sectobjectsymbols 12166@opindex sectorder 12167@opindex segaddr 12168@opindex segs_read_only_addr 12169@opindex segs_read_write_addr 12170@opindex seg_addr_table 12171@opindex seg_addr_table_filename 12172@opindex seglinkedit 12173@opindex segprot 12174@opindex segs_read_only_addr 12175@opindex segs_read_write_addr 12176@opindex single_module 12177@opindex static 12178@opindex sub_library 12179@opindex sub_umbrella 12180@opindex twolevel_namespace 12181@opindex umbrella 12182@opindex undefined 12183@opindex unexported_symbols_list 12184@opindex weak_reference_mismatches 12185@opindex whatsloaded 12186These options are passed to the Darwin linker. The Darwin linker man page 12187describes them in detail. 12188@end table 12189 12190@node DEC Alpha Options 12191@subsection DEC Alpha Options 12192 12193These @samp{-m} options are defined for the DEC Alpha implementations: 12194 12195@table @gcctabopt 12196@item -mno-soft-float 12197@itemx -msoft-float 12198@opindex mno-soft-float 12199@opindex msoft-float 12200Use (do not use) the hardware floating-point instructions for 12201floating-point operations. When @option{-msoft-float} is specified, 12202functions in @file{libgcc.a} will be used to perform floating-point 12203operations. Unless they are replaced by routines that emulate the 12204floating-point operations, or compiled in such a way as to call such 12205emulations routines, these routines will issue floating-point 12206operations. If you are compiling for an Alpha without floating-point 12207operations, you must ensure that the library is built so as not to call 12208them. 12209 12210Note that Alpha implementations without floating-point operations are 12211required to have floating-point registers. 12212 12213@item -mfp-reg 12214@itemx -mno-fp-regs 12215@opindex mfp-reg 12216@opindex mno-fp-regs 12217Generate code that uses (does not use) the floating-point register set. 12218@option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point 12219register set is not used, floating-point operands are passed in integer 12220registers as if they were integers and floating-point results are passed 12221in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence, 12222so any function with a floating-point argument or return value called by code 12223compiled with @option{-mno-fp-regs} must also be compiled with that 12224option. 12225 12226A typical use of this option is building a kernel that does not use, 12227and hence need not save and restore, any floating-point registers. 12228 12229@item -mieee 12230@opindex mieee 12231The Alpha architecture implements floating-point hardware optimized for 12232maximum performance. It is mostly compliant with the IEEE floating-point 12233standard. However, for full compliance, software assistance is 12234required. This option generates code fully IEEE-compliant code 12235@emph{except} that the @var{inexact-flag} is not maintained (see below). 12236If this option is turned on, the preprocessor macro @code{_IEEE_FP} is 12237defined during compilation. The resulting code is less efficient but is 12238able to correctly support denormalized numbers and exceptional IEEE 12239values such as not-a-number and plus/minus infinity. Other Alpha 12240compilers call this option @option{-ieee_with_no_inexact}. 12241 12242@item -mieee-with-inexact 12243@opindex mieee-with-inexact 12244This is like @option{-mieee} except the generated code also maintains 12245the IEEE @var{inexact-flag}. Turning on this option causes the 12246generated code to implement fully-compliant IEEE math. In addition to 12247@code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor 12248macro. On some Alpha implementations the resulting code may execute 12249significantly slower than the code generated by default. Since there is 12250very little code that depends on the @var{inexact-flag}, you should 12251normally not specify this option. Other Alpha compilers call this 12252option @option{-ieee_with_inexact}. 12253 12254@item -mfp-trap-mode=@var{trap-mode} 12255@opindex mfp-trap-mode 12256This option controls what floating-point related traps are enabled. 12257Other Alpha compilers call this option @option{-fptm @var{trap-mode}}. 12258The trap mode can be set to one of four values: 12259 12260@table @samp 12261@item n 12262This is the default (normal) setting. The only traps that are enabled 12263are the ones that cannot be disabled in software (e.g., division by zero 12264trap). 12265 12266@item u 12267In addition to the traps enabled by @samp{n}, underflow traps are enabled 12268as well. 12269 12270@item su 12271Like @samp{u}, but the instructions are marked to be safe for software 12272completion (see Alpha architecture manual for details). 12273 12274@item sui 12275Like @samp{su}, but inexact traps are enabled as well. 12276@end table 12277 12278@item -mfp-rounding-mode=@var{rounding-mode} 12279@opindex mfp-rounding-mode 12280Selects the IEEE rounding mode. Other Alpha compilers call this option 12281@option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one 12282of: 12283 12284@table @samp 12285@item n 12286Normal IEEE rounding mode. Floating-point numbers are rounded towards 12287the nearest machine number or towards the even machine number in case 12288of a tie. 12289 12290@item m 12291Round towards minus infinity. 12292 12293@item c 12294Chopped rounding mode. Floating-point numbers are rounded towards zero. 12295 12296@item d 12297Dynamic rounding mode. A field in the floating-point control register 12298(@var{fpcr}, see Alpha architecture reference manual) controls the 12299rounding mode in effect. The C library initializes this register for 12300rounding towards plus infinity. Thus, unless your program modifies the 12301@var{fpcr}, @samp{d} corresponds to round towards plus infinity. 12302@end table 12303 12304@item -mtrap-precision=@var{trap-precision} 12305@opindex mtrap-precision 12306In the Alpha architecture, floating-point traps are imprecise. This 12307means without software assistance it is impossible to recover from a 12308floating trap and program execution normally needs to be terminated. 12309GCC can generate code that can assist operating system trap handlers 12310in determining the exact location that caused a floating-point trap. 12311Depending on the requirements of an application, different levels of 12312precisions can be selected: 12313 12314@table @samp 12315@item p 12316Program precision. This option is the default and means a trap handler 12317can only identify which program caused a floating-point exception. 12318 12319@item f 12320Function precision. The trap handler can determine the function that 12321caused a floating-point exception. 12322 12323@item i 12324Instruction precision. The trap handler can determine the exact 12325instruction that caused a floating-point exception. 12326@end table 12327 12328Other Alpha compilers provide the equivalent options called 12329@option{-scope_safe} and @option{-resumption_safe}. 12330 12331@item -mieee-conformant 12332@opindex mieee-conformant 12333This option marks the generated code as IEEE conformant. You must not 12334use this option unless you also specify @option{-mtrap-precision=i} and either 12335@option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect 12336is to emit the line @samp{.eflag 48} in the function prologue of the 12337generated assembly file. Under DEC Unix, this has the effect that 12338IEEE-conformant math library routines will be linked in. 12339 12340@item -mbuild-constants 12341@opindex mbuild-constants 12342Normally GCC examines a 32- or 64-bit integer constant to 12343see if it can construct it from smaller constants in two or three 12344instructions. If it cannot, it will output the constant as a literal and 12345generate code to load it from the data segment at run time. 12346 12347Use this option to require GCC to construct @emph{all} integer constants 12348using code, even if it takes more instructions (the maximum is six). 12349 12350You would typically use this option to build a shared library dynamic 12351loader. Itself a shared library, it must relocate itself in memory 12352before it can find the variables and constants in its own data segment. 12353 12354@item -malpha-as 12355@itemx -mgas 12356@opindex malpha-as 12357@opindex mgas 12358Select whether to generate code to be assembled by the vendor-supplied 12359assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}. 12360 12361@item -mbwx 12362@itemx -mno-bwx 12363@itemx -mcix 12364@itemx -mno-cix 12365@itemx -mfix 12366@itemx -mno-fix 12367@itemx -mmax 12368@itemx -mno-max 12369@opindex mbwx 12370@opindex mno-bwx 12371@opindex mcix 12372@opindex mno-cix 12373@opindex mfix 12374@opindex mno-fix 12375@opindex mmax 12376@opindex mno-max 12377Indicate whether GCC should generate code to use the optional BWX, 12378CIX, FIX and MAX instruction sets. The default is to use the instruction 12379sets supported by the CPU type specified via @option{-mcpu=} option or that 12380of the CPU on which GCC was built if none was specified. 12381 12382@item -mfloat-vax 12383@itemx -mfloat-ieee 12384@opindex mfloat-vax 12385@opindex mfloat-ieee 12386Generate code that uses (does not use) VAX F and G floating-point 12387arithmetic instead of IEEE single and double precision. 12388 12389@item -mexplicit-relocs 12390@itemx -mno-explicit-relocs 12391@opindex mexplicit-relocs 12392@opindex mno-explicit-relocs 12393Older Alpha assemblers provided no way to generate symbol relocations 12394except via assembler macros. Use of these macros does not allow 12395optimal instruction scheduling. GNU binutils as of version 2.12 12396supports a new syntax that allows the compiler to explicitly mark 12397which relocations should apply to which instructions. This option 12398is mostly useful for debugging, as GCC detects the capabilities of 12399the assembler when it is built and sets the default accordingly. 12400 12401@item -msmall-data 12402@itemx -mlarge-data 12403@opindex msmall-data 12404@opindex mlarge-data 12405When @option{-mexplicit-relocs} is in effect, static data is 12406accessed via @dfn{gp-relative} relocations. When @option{-msmall-data} 12407is used, objects 8 bytes long or smaller are placed in a @dfn{small data area} 12408(the @code{.sdata} and @code{.sbss} sections) and are accessed via 1240916-bit relocations off of the @code{$gp} register. This limits the 12410size of the small data area to 64KB, but allows the variables to be 12411directly accessed via a single instruction. 12412 12413The default is @option{-mlarge-data}. With this option the data area 12414is limited to just below 2GB@. Programs that require more than 2GB of 12415data must use @code{malloc} or @code{mmap} to allocate the data in the 12416heap instead of in the program's data segment. 12417 12418When generating code for shared libraries, @option{-fpic} implies 12419@option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}. 12420 12421@item -msmall-text 12422@itemx -mlarge-text 12423@opindex msmall-text 12424@opindex mlarge-text 12425When @option{-msmall-text} is used, the compiler assumes that the 12426code of the entire program (or shared library) fits in 4MB, and is 12427thus reachable with a branch instruction. When @option{-msmall-data} 12428is used, the compiler can assume that all local symbols share the 12429same @code{$gp} value, and thus reduce the number of instructions 12430required for a function call from 4 to 1. 12431 12432The default is @option{-mlarge-text}. 12433 12434@item -mcpu=@var{cpu_type} 12435@opindex mcpu 12436Set the instruction set and instruction scheduling parameters for 12437machine type @var{cpu_type}. You can specify either the @samp{EV} 12438style name or the corresponding chip number. GCC supports scheduling 12439parameters for the EV4, EV5 and EV6 family of processors and will 12440choose the default values for the instruction set from the processor 12441you specify. If you do not specify a processor type, GCC will default 12442to the processor on which the compiler was built. 12443 12444Supported values for @var{cpu_type} are 12445 12446@table @samp 12447@item ev4 12448@itemx ev45 12449@itemx 21064 12450Schedules as an EV4 and has no instruction set extensions. 12451 12452@item ev5 12453@itemx 21164 12454Schedules as an EV5 and has no instruction set extensions. 12455 12456@item ev56 12457@itemx 21164a 12458Schedules as an EV5 and supports the BWX extension. 12459 12460@item pca56 12461@itemx 21164pc 12462@itemx 21164PC 12463Schedules as an EV5 and supports the BWX and MAX extensions. 12464 12465@item ev6 12466@itemx 21264 12467Schedules as an EV6 and supports the BWX, FIX, and MAX extensions. 12468 12469@item ev67 12470@itemx 21264a 12471Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions. 12472@end table 12473 12474Native toolchains also support the value @samp{native}, 12475which selects the best architecture option for the host processor. 12476@option{-mcpu=native} has no effect if GCC does not recognize 12477the processor. 12478 12479@item -mtune=@var{cpu_type} 12480@opindex mtune 12481Set only the instruction scheduling parameters for machine type 12482@var{cpu_type}. The instruction set is not changed. 12483 12484Native toolchains also support the value @samp{native}, 12485which selects the best architecture option for the host processor. 12486@option{-mtune=native} has no effect if GCC does not recognize 12487the processor. 12488 12489@item -mmemory-latency=@var{time} 12490@opindex mmemory-latency 12491Sets the latency the scheduler should assume for typical memory 12492references as seen by the application. This number is highly 12493dependent on the memory access patterns used by the application 12494and the size of the external cache on the machine. 12495 12496Valid options for @var{time} are 12497 12498@table @samp 12499@item @var{number} 12500A decimal number representing clock cycles. 12501 12502@item L1 12503@itemx L2 12504@itemx L3 12505@itemx main 12506The compiler contains estimates of the number of clock cycles for 12507``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches 12508(also called Dcache, Scache, and Bcache), as well as to main memory. 12509Note that L3 is only valid for EV5. 12510 12511@end table 12512@end table 12513 12514@node DEC Alpha/VMS Options 12515@subsection DEC Alpha/VMS Options 12516 12517These @samp{-m} options are defined for the DEC Alpha/VMS implementations: 12518 12519@table @gcctabopt 12520@item -mvms-return-codes 12521@opindex mvms-return-codes 12522Return VMS condition codes from main. The default is to return POSIX 12523style condition (e.g.@: error) codes. 12524 12525@item -mdebug-main=@var{prefix} 12526@opindex mdebug-main=@var{prefix} 12527Flag the first routine whose name starts with @var{prefix} as the main 12528routine for the debugger. 12529 12530@item -mmalloc64 12531@opindex mmalloc64 12532Default to 64-bit memory allocation routines. 12533@end table 12534 12535@node FR30 Options 12536@subsection FR30 Options 12537@cindex FR30 Options 12538 12539These options are defined specifically for the FR30 port. 12540 12541@table @gcctabopt 12542 12543@item -msmall-model 12544@opindex msmall-model 12545Use the small address space model. This can produce smaller code, but 12546it does assume that all symbolic values and addresses will fit into a 1254720-bit range. 12548 12549@item -mno-lsim 12550@opindex mno-lsim 12551Assume that runtime support has been provided and so there is no need 12552to include the simulator library (@file{libsim.a}) on the linker 12553command line. 12554 12555@end table 12556 12557@node FRV Options 12558@subsection FRV Options 12559@cindex FRV Options 12560 12561@table @gcctabopt 12562@item -mgpr-32 12563@opindex mgpr-32 12564 12565Only use the first 32 general-purpose registers. 12566 12567@item -mgpr-64 12568@opindex mgpr-64 12569 12570Use all 64 general-purpose registers. 12571 12572@item -mfpr-32 12573@opindex mfpr-32 12574 12575Use only the first 32 floating-point registers. 12576 12577@item -mfpr-64 12578@opindex mfpr-64 12579 12580Use all 64 floating-point registers. 12581 12582@item -mhard-float 12583@opindex mhard-float 12584 12585Use hardware instructions for floating-point operations. 12586 12587@item -msoft-float 12588@opindex msoft-float 12589 12590Use library routines for floating-point operations. 12591 12592@item -malloc-cc 12593@opindex malloc-cc 12594 12595Dynamically allocate condition code registers. 12596 12597@item -mfixed-cc 12598@opindex mfixed-cc 12599 12600Do not try to dynamically allocate condition code registers, only 12601use @code{icc0} and @code{fcc0}. 12602 12603@item -mdword 12604@opindex mdword 12605 12606Change ABI to use double word insns. 12607 12608@item -mno-dword 12609@opindex mno-dword 12610 12611Do not use double word instructions. 12612 12613@item -mdouble 12614@opindex mdouble 12615 12616Use floating-point double instructions. 12617 12618@item -mno-double 12619@opindex mno-double 12620 12621Do not use floating-point double instructions. 12622 12623@item -mmedia 12624@opindex mmedia 12625 12626Use media instructions. 12627 12628@item -mno-media 12629@opindex mno-media 12630 12631Do not use media instructions. 12632 12633@item -mmuladd 12634@opindex mmuladd 12635 12636Use multiply and add/subtract instructions. 12637 12638@item -mno-muladd 12639@opindex mno-muladd 12640 12641Do not use multiply and add/subtract instructions. 12642 12643@item -mfdpic 12644@opindex mfdpic 12645 12646Select the FDPIC ABI, which uses function descriptors to represent 12647pointers to functions. Without any PIC/PIE-related options, it 12648implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it 12649assumes GOT entries and small data are within a 12-bit range from the 12650GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets 12651are computed with 32 bits. 12652With a @samp{bfin-elf} target, this option implies @option{-msim}. 12653 12654@item -minline-plt 12655@opindex minline-plt 12656 12657Enable inlining of PLT entries in function calls to functions that are 12658not known to bind locally. It has no effect without @option{-mfdpic}. 12659It's enabled by default if optimizing for speed and compiling for 12660shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an 12661optimization option such as @option{-O3} or above is present in the 12662command line. 12663 12664@item -mTLS 12665@opindex mTLS 12666 12667Assume a large TLS segment when generating thread-local code. 12668 12669@item -mtls 12670@opindex mtls 12671 12672Do not assume a large TLS segment when generating thread-local code. 12673 12674@item -mgprel-ro 12675@opindex mgprel-ro 12676 12677Enable the use of @code{GPREL} relocations in the FDPIC ABI for data 12678that is known to be in read-only sections. It's enabled by default, 12679except for @option{-fpic} or @option{-fpie}: even though it may help 12680make the global offset table smaller, it trades 1 instruction for 4. 12681With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4, 12682one of which may be shared by multiple symbols, and it avoids the need 12683for a GOT entry for the referenced symbol, so it's more likely to be a 12684win. If it is not, @option{-mno-gprel-ro} can be used to disable it. 12685 12686@item -multilib-library-pic 12687@opindex multilib-library-pic 12688 12689Link with the (library, not FD) pic libraries. It's implied by 12690@option{-mlibrary-pic}, as well as by @option{-fPIC} and 12691@option{-fpic} without @option{-mfdpic}. You should never have to use 12692it explicitly. 12693 12694@item -mlinked-fp 12695@opindex mlinked-fp 12696 12697Follow the EABI requirement of always creating a frame pointer whenever 12698a stack frame is allocated. This option is enabled by default and can 12699be disabled with @option{-mno-linked-fp}. 12700 12701@item -mlong-calls 12702@opindex mlong-calls 12703 12704Use indirect addressing to call functions outside the current 12705compilation unit. This allows the functions to be placed anywhere 12706within the 32-bit address space. 12707 12708@item -malign-labels 12709@opindex malign-labels 12710 12711Try to align labels to an 8-byte boundary by inserting nops into the 12712previous packet. This option only has an effect when VLIW packing 12713is enabled. It doesn't create new packets; it merely adds nops to 12714existing ones. 12715 12716@item -mlibrary-pic 12717@opindex mlibrary-pic 12718 12719Generate position-independent EABI code. 12720 12721@item -macc-4 12722@opindex macc-4 12723 12724Use only the first four media accumulator registers. 12725 12726@item -macc-8 12727@opindex macc-8 12728 12729Use all eight media accumulator registers. 12730 12731@item -mpack 12732@opindex mpack 12733 12734Pack VLIW instructions. 12735 12736@item -mno-pack 12737@opindex mno-pack 12738 12739Do not pack VLIW instructions. 12740 12741@item -mno-eflags 12742@opindex mno-eflags 12743 12744Do not mark ABI switches in e_flags. 12745 12746@item -mcond-move 12747@opindex mcond-move 12748 12749Enable the use of conditional-move instructions (default). 12750 12751This switch is mainly for debugging the compiler and will likely be removed 12752in a future version. 12753 12754@item -mno-cond-move 12755@opindex mno-cond-move 12756 12757Disable the use of conditional-move instructions. 12758 12759This switch is mainly for debugging the compiler and will likely be removed 12760in a future version. 12761 12762@item -mscc 12763@opindex mscc 12764 12765Enable the use of conditional set instructions (default). 12766 12767This switch is mainly for debugging the compiler and will likely be removed 12768in a future version. 12769 12770@item -mno-scc 12771@opindex mno-scc 12772 12773Disable the use of conditional set instructions. 12774 12775This switch is mainly for debugging the compiler and will likely be removed 12776in a future version. 12777 12778@item -mcond-exec 12779@opindex mcond-exec 12780 12781Enable the use of conditional execution (default). 12782 12783This switch is mainly for debugging the compiler and will likely be removed 12784in a future version. 12785 12786@item -mno-cond-exec 12787@opindex mno-cond-exec 12788 12789Disable the use of conditional execution. 12790 12791This switch is mainly for debugging the compiler and will likely be removed 12792in a future version. 12793 12794@item -mvliw-branch 12795@opindex mvliw-branch 12796 12797Run a pass to pack branches into VLIW instructions (default). 12798 12799This switch is mainly for debugging the compiler and will likely be removed 12800in a future version. 12801 12802@item -mno-vliw-branch 12803@opindex mno-vliw-branch 12804 12805Do not run a pass to pack branches into VLIW instructions. 12806 12807This switch is mainly for debugging the compiler and will likely be removed 12808in a future version. 12809 12810@item -mmulti-cond-exec 12811@opindex mmulti-cond-exec 12812 12813Enable optimization of @code{&&} and @code{||} in conditional execution 12814(default). 12815 12816This switch is mainly for debugging the compiler and will likely be removed 12817in a future version. 12818 12819@item -mno-multi-cond-exec 12820@opindex mno-multi-cond-exec 12821 12822Disable optimization of @code{&&} and @code{||} in conditional execution. 12823 12824This switch is mainly for debugging the compiler and will likely be removed 12825in a future version. 12826 12827@item -mnested-cond-exec 12828@opindex mnested-cond-exec 12829 12830Enable nested conditional execution optimizations (default). 12831 12832This switch is mainly for debugging the compiler and will likely be removed 12833in a future version. 12834 12835@item -mno-nested-cond-exec 12836@opindex mno-nested-cond-exec 12837 12838Disable nested conditional execution optimizations. 12839 12840This switch is mainly for debugging the compiler and will likely be removed 12841in a future version. 12842 12843@item -moptimize-membar 12844@opindex moptimize-membar 12845 12846This switch removes redundant @code{membar} instructions from the 12847compiler generated code. It is enabled by default. 12848 12849@item -mno-optimize-membar 12850@opindex mno-optimize-membar 12851 12852This switch disables the automatic removal of redundant @code{membar} 12853instructions from the generated code. 12854 12855@item -mtomcat-stats 12856@opindex mtomcat-stats 12857 12858Cause gas to print out tomcat statistics. 12859 12860@item -mcpu=@var{cpu} 12861@opindex mcpu 12862 12863Select the processor type for which to generate code. Possible values are 12864@samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450}, 12865@samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}. 12866 12867@end table 12868 12869@node GNU/Linux Options 12870@subsection GNU/Linux Options 12871 12872These @samp{-m} options are defined for GNU/Linux targets: 12873 12874@table @gcctabopt 12875@item -mglibc 12876@opindex mglibc 12877Use the GNU C library. This is the default except 12878on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets. 12879 12880@item -muclibc 12881@opindex muclibc 12882Use uClibc C library. This is the default on 12883@samp{*-*-linux-*uclibc*} targets. 12884 12885@item -mbionic 12886@opindex mbionic 12887Use Bionic C library. This is the default on 12888@samp{*-*-linux-*android*} targets. 12889 12890@item -mandroid 12891@opindex mandroid 12892Compile code compatible with Android platform. This is the default on 12893@samp{*-*-linux-*android*} targets. 12894 12895When compiling, this option enables @option{-mbionic}, @option{-fPIC}, 12896@option{-fno-exceptions} and @option{-fno-rtti} by default. When linking, 12897this option makes the GCC driver pass Android-specific options to the linker. 12898Finally, this option causes the preprocessor macro @code{__ANDROID__} 12899to be defined. 12900 12901@item -tno-android-cc 12902@opindex tno-android-cc 12903Disable compilation effects of @option{-mandroid}, i.e., do not enable 12904@option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and 12905@option{-fno-rtti} by default. 12906 12907@item -tno-android-ld 12908@opindex tno-android-ld 12909Disable linking effects of @option{-mandroid}, i.e., pass standard Linux 12910linking options to the linker. 12911 12912@end table 12913 12914@node H8/300 Options 12915@subsection H8/300 Options 12916 12917These @samp{-m} options are defined for the H8/300 implementations: 12918 12919@table @gcctabopt 12920@item -mrelax 12921@opindex mrelax 12922Shorten some address references at link time, when possible; uses the 12923linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300, 12924ld, Using ld}, for a fuller description. 12925 12926@item -mh 12927@opindex mh 12928Generate code for the H8/300H@. 12929 12930@item -ms 12931@opindex ms 12932Generate code for the H8S@. 12933 12934@item -mn 12935@opindex mn 12936Generate code for the H8S and H8/300H in the normal mode. This switch 12937must be used either with @option{-mh} or @option{-ms}. 12938 12939@item -ms2600 12940@opindex ms2600 12941Generate code for the H8S/2600. This switch must be used with @option{-ms}. 12942 12943@item -mint32 12944@opindex mint32 12945Make @code{int} data 32 bits by default. 12946 12947@item -malign-300 12948@opindex malign-300 12949On the H8/300H and H8S, use the same alignment rules as for the H8/300. 12950The default for the H8/300H and H8S is to align longs and floats on 129514-byte boundaries. 12952@option{-malign-300} causes them to be aligned on 2-byte boundaries. 12953This option has no effect on the H8/300. 12954@end table 12955 12956@node HPPA Options 12957@subsection HPPA Options 12958@cindex HPPA Options 12959 12960These @samp{-m} options are defined for the HPPA family of computers: 12961 12962@table @gcctabopt 12963@item -march=@var{architecture-type} 12964@opindex march 12965Generate code for the specified architecture. The choices for 12966@var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA 129671.1, and @samp{2.0} for PA 2.0 processors. Refer to 12968@file{/usr/lib/sched.models} on an HP-UX system to determine the proper 12969architecture option for your machine. Code compiled for lower numbered 12970architectures will run on higher numbered architectures, but not the 12971other way around. 12972 12973@item -mpa-risc-1-0 12974@itemx -mpa-risc-1-1 12975@itemx -mpa-risc-2-0 12976@opindex mpa-risc-1-0 12977@opindex mpa-risc-1-1 12978@opindex mpa-risc-2-0 12979Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively. 12980 12981@item -mbig-switch 12982@opindex mbig-switch 12983Generate code suitable for big switch tables. Use this option only if 12984the assembler/linker complain about out of range branches within a switch 12985table. 12986 12987@item -mjump-in-delay 12988@opindex mjump-in-delay 12989Fill delay slots of function calls with unconditional jump instructions 12990by modifying the return pointer for the function call to be the target 12991of the conditional jump. 12992 12993@item -mdisable-fpregs 12994@opindex mdisable-fpregs 12995Prevent floating-point registers from being used in any manner. This is 12996necessary for compiling kernels that perform lazy context switching of 12997floating-point registers. If you use this option and attempt to perform 12998floating-point operations, the compiler aborts. 12999 13000@item -mdisable-indexing 13001@opindex mdisable-indexing 13002Prevent the compiler from using indexing address modes. This avoids some 13003rather obscure problems when compiling MIG generated code under MACH@. 13004 13005@item -mno-space-regs 13006@opindex mno-space-regs 13007Generate code that assumes the target has no space registers. This allows 13008GCC to generate faster indirect calls and use unscaled index address modes. 13009 13010Such code is suitable for level 0 PA systems and kernels. 13011 13012@item -mfast-indirect-calls 13013@opindex mfast-indirect-calls 13014Generate code that assumes calls never cross space boundaries. This 13015allows GCC to emit code that performs faster indirect calls. 13016 13017This option will not work in the presence of shared libraries or nested 13018functions. 13019 13020@item -mfixed-range=@var{register-range} 13021@opindex mfixed-range 13022Generate code treating the given register range as fixed registers. 13023A fixed register is one that the register allocator can not use. This is 13024useful when compiling kernel code. A register range is specified as 13025two registers separated by a dash. Multiple register ranges can be 13026specified separated by a comma. 13027 13028@item -mlong-load-store 13029@opindex mlong-load-store 13030Generate 3-instruction load and store sequences as sometimes required by 13031the HP-UX 10 linker. This is equivalent to the @samp{+k} option to 13032the HP compilers. 13033 13034@item -mportable-runtime 13035@opindex mportable-runtime 13036Use the portable calling conventions proposed by HP for ELF systems. 13037 13038@item -mgas 13039@opindex mgas 13040Enable the use of assembler directives only GAS understands. 13041 13042@item -mschedule=@var{cpu-type} 13043@opindex mschedule 13044Schedule code according to the constraints for the machine type 13045@var{cpu-type}. The choices for @var{cpu-type} are @samp{700} 13046@samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer 13047to @file{/usr/lib/sched.models} on an HP-UX system to determine the 13048proper scheduling option for your machine. The default scheduling is 13049@samp{8000}. 13050 13051@item -mlinker-opt 13052@opindex mlinker-opt 13053Enable the optimization pass in the HP-UX linker. Note this makes symbolic 13054debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9 13055linkers in which they give bogus error messages when linking some programs. 13056 13057@item -msoft-float 13058@opindex msoft-float 13059Generate output containing library calls for floating point. 13060@strong{Warning:} the requisite libraries are not available for all HPPA 13061targets. Normally the facilities of the machine's usual C compiler are 13062used, but this cannot be done directly in cross-compilation. You must make 13063your own arrangements to provide suitable library functions for 13064cross-compilation. 13065 13066@option{-msoft-float} changes the calling convention in the output file; 13067therefore, it is only useful if you compile @emph{all} of a program with 13068this option. In particular, you need to compile @file{libgcc.a}, the 13069library that comes with GCC, with @option{-msoft-float} in order for 13070this to work. 13071 13072@item -msio 13073@opindex msio 13074Generate the predefine, @code{_SIO}, for server IO@. The default is 13075@option{-mwsio}. This generates the predefines, @code{__hp9000s700}, 13076@code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These 13077options are available under HP-UX and HI-UX@. 13078 13079@item -mgnu-ld 13080@opindex mgnu-ld 13081Use GNU ld specific options. This passes @option{-shared} to ld when 13082building a shared library. It is the default when GCC is configured, 13083explicitly or implicitly, with the GNU linker. This option does not 13084have any affect on which ld is called, it only changes what parameters 13085are passed to that ld. The ld that is called is determined by the 13086@option{--with-ld} configure option, GCC's program search path, and 13087finally by the user's @env{PATH}. The linker used by GCC can be printed 13088using @samp{which `gcc -print-prog-name=ld`}. This option is only available 13089on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}. 13090 13091@item -mhp-ld 13092@opindex mhp-ld 13093Use HP ld specific options. This passes @option{-b} to ld when building 13094a shared library and passes @option{+Accept TypeMismatch} to ld on all 13095links. It is the default when GCC is configured, explicitly or 13096implicitly, with the HP linker. This option does not have any affect on 13097which ld is called, it only changes what parameters are passed to that 13098ld. The ld that is called is determined by the @option{--with-ld} 13099configure option, GCC's program search path, and finally by the user's 13100@env{PATH}. The linker used by GCC can be printed using @samp{which 13101`gcc -print-prog-name=ld`}. This option is only available on the 64-bit 13102HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}. 13103 13104@item -mlong-calls 13105@opindex mno-long-calls 13106Generate code that uses long call sequences. This ensures that a call 13107is always able to reach linker generated stubs. The default is to generate 13108long calls only when the distance from the call site to the beginning 13109of the function or translation unit, as the case may be, exceeds a 13110predefined limit set by the branch type being used. The limits for 13111normal calls are 7,600,000 and 240,000 bytes, respectively for the 13112PA 2.0 and PA 1.X architectures. Sibcalls are always limited at 13113240,000 bytes. 13114 13115Distances are measured from the beginning of functions when using the 13116@option{-ffunction-sections} option, or when using the @option{-mgas} 13117and @option{-mno-portable-runtime} options together under HP-UX with 13118the SOM linker. 13119 13120It is normally not desirable to use this option as it will degrade 13121performance. However, it may be useful in large applications, 13122particularly when partial linking is used to build the application. 13123 13124The types of long calls used depends on the capabilities of the 13125assembler and linker, and the type of code being generated. The 13126impact on systems that support long absolute calls, and long pic 13127symbol-difference or pc-relative calls should be relatively small. 13128However, an indirect call is used on 32-bit ELF systems in pic code 13129and it is quite long. 13130 13131@item -munix=@var{unix-std} 13132@opindex march 13133Generate compiler predefines and select a startfile for the specified 13134UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95} 13135and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95} 13136is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX 1313711.11 and later. The default values are @samp{93} for HP-UX 10.00, 13138@samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11 13139and later. 13140 13141@option{-munix=93} provides the same predefines as GCC 3.3 and 3.4. 13142@option{-munix=95} provides additional predefines for @code{XOPEN_UNIX} 13143and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}. 13144@option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX}, 13145@code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and 13146@code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}. 13147 13148It is @emph{important} to note that this option changes the interfaces 13149for various library routines. It also affects the operational behavior 13150of the C library. Thus, @emph{extreme} care is needed in using this 13151option. 13152 13153Library code that is intended to operate with more than one UNIX 13154standard must test, set and restore the variable @var{__xpg4_extended_mask} 13155as appropriate. Most GNU software doesn't provide this capability. 13156 13157@item -nolibdld 13158@opindex nolibdld 13159Suppress the generation of link options to search libdld.sl when the 13160@option{-static} option is specified on HP-UX 10 and later. 13161 13162@item -static 13163@opindex static 13164The HP-UX implementation of setlocale in libc has a dependency on 13165libdld.sl. There isn't an archive version of libdld.sl. Thus, 13166when the @option{-static} option is specified, special link options 13167are needed to resolve this dependency. 13168 13169On HP-UX 10 and later, the GCC driver adds the necessary options to 13170link with libdld.sl when the @option{-static} option is specified. 13171This causes the resulting binary to be dynamic. On the 64-bit port, 13172the linkers generate dynamic binaries by default in any case. The 13173@option{-nolibdld} option can be used to prevent the GCC driver from 13174adding these link options. 13175 13176@item -threads 13177@opindex threads 13178Add support for multithreading with the @dfn{dce thread} library 13179under HP-UX@. This option sets flags for both the preprocessor and 13180linker. 13181@end table 13182 13183@node i386 and x86-64 Options 13184@subsection Intel 386 and AMD x86-64 Options 13185@cindex i386 Options 13186@cindex x86-64 Options 13187@cindex Intel 386 Options 13188@cindex AMD x86-64 Options 13189 13190These @samp{-m} options are defined for the i386 and x86-64 family of 13191computers: 13192 13193@table @gcctabopt 13194@item -mtune=@var{cpu-type} 13195@opindex mtune 13196Tune to @var{cpu-type} everything applicable about the generated code, except 13197for the ABI and the set of available instructions. The choices for 13198@var{cpu-type} are: 13199@table @emph 13200@item generic 13201Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors. 13202If you know the CPU on which your code will run, then you should use 13203the corresponding @option{-mtune} option instead of 13204@option{-mtune=generic}. But, if you do not know exactly what CPU users 13205of your application will have, then you should use this option. 13206 13207As new processors are deployed in the marketplace, the behavior of this 13208option will change. Therefore, if you upgrade to a newer version of 13209GCC, the code generated option will change to reflect the processors 13210that were most common when that version of GCC was released. 13211 13212There is no @option{-march=generic} option because @option{-march} 13213indicates the instruction set the compiler can use, and there is no 13214generic instruction set applicable to all processors. In contrast, 13215@option{-mtune} indicates the processor (or, in this case, collection of 13216processors) for which the code is optimized. 13217@item native 13218This selects the CPU to tune for at compilation time by determining 13219the processor type of the compiling machine. Using @option{-mtune=native} 13220will produce code optimized for the local machine under the constraints 13221of the selected instruction set. Using @option{-march=native} will 13222enable all instruction subsets supported by the local machine (hence 13223the result might not run on different machines). 13224@item i386 13225Original Intel's i386 CPU@. 13226@item i486 13227Intel's i486 CPU@. (No scheduling is implemented for this chip.) 13228@item i586, pentium 13229Intel Pentium CPU with no MMX support. 13230@item pentium-mmx 13231Intel PentiumMMX CPU based on Pentium core with MMX instruction set support. 13232@item pentiumpro 13233Intel PentiumPro CPU@. 13234@item i686 13235Same as @code{generic}, but when used as @code{march} option, PentiumPro 13236instruction set will be used, so the code will run on all i686 family chips. 13237@item pentium2 13238Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support. 13239@item pentium3, pentium3m 13240Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set 13241support. 13242@item pentium-m 13243Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set 13244support. Used by Centrino notebooks. 13245@item pentium4, pentium4m 13246Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support. 13247@item prescott 13248Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction 13249set support. 13250@item nocona 13251Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE, 13252SSE2 and SSE3 instruction set support. 13253@item core2 13254Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3 13255instruction set support. 13256@item corei7 13257Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1 13258and SSE4.2 instruction set support. 13259@item corei7-avx 13260Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, 13261SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support. 13262@item core-avx-i 13263Intel Core CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, 13264SSE4.1, SSE4.2, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C instruction 13265set support. 13266@item atom 13267Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3 13268instruction set support. 13269@item k6 13270AMD K6 CPU with MMX instruction set support. 13271@item k6-2, k6-3 13272Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support. 13273@item athlon, athlon-tbird 13274AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions 13275support. 13276@item athlon-4, athlon-xp, athlon-mp 13277Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE 13278instruction set support. 13279@item k8, opteron, athlon64, athlon-fx 13280AMD K8 core based CPUs with x86-64 instruction set support. (This supersets 13281MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.) 13282@item k8-sse3, opteron-sse3, athlon64-sse3 13283Improved versions of k8, opteron and athlon64 with SSE3 instruction set support. 13284@item amdfam10, barcelona 13285AMD Family 10h core based CPUs with x86-64 instruction set support. (This 13286supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit 13287instruction set extensions.) 13288@item bdver1 13289AMD Family 15h core based CPUs with x86-64 instruction set support. (This 13290supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, 13291SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.) 13292@item bdver2 13293AMD Family 15h core based CPUs with x86-64 instruction set support. (This 13294supersets BMI, TBM, F16C, FMA, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, 13295SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set 13296extensions.) 13297@item btver1 13298AMD Family 14h core based CPUs with x86-64 instruction set support. (This 13299supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit 13300instruction set extensions.) 13301@item winchip-c6 13302IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction 13303set support. 13304@item winchip2 13305IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@: 13306instruction set support. 13307@item c3 13308Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is 13309implemented for this chip.) 13310@item c3-2 13311Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is 13312implemented for this chip.) 13313@item geode 13314Embedded AMD CPU with MMX and 3DNow!@: instruction set support. 13315@end table 13316 13317While picking a specific @var{cpu-type} will schedule things appropriately 13318for that particular chip, the compiler will not generate any code that 13319does not run on the default machine type without the @option{-march=@var{cpu-type}} 13320option being used. For example, if GCC is configured for i686-pc-linux-gnu 13321then @option{-mtune=pentium4} will generate code that is tuned for Pentium4 13322but will still run on i686 machines. 13323 13324@item -march=@var{cpu-type} 13325@opindex march 13326Generate instructions for the machine type @var{cpu-type}. The choices 13327for @var{cpu-type} are the same as for @option{-mtune}. Moreover, 13328specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}. 13329 13330@item -mcpu=@var{cpu-type} 13331@opindex mcpu 13332A deprecated synonym for @option{-mtune}. 13333 13334@item -mfpmath=@var{unit} 13335@opindex mfpmath 13336Generate floating-point arithmetic for selected unit @var{unit}. The choices 13337for @var{unit} are: 13338 13339@table @samp 13340@item 387 13341Use the standard 387 floating-point coprocessor present on the majority of chips and 13342emulated otherwise. Code compiled with this option runs almost everywhere. 13343The temporary results are computed in 80-bit precision instead of the precision 13344specified by the type, resulting in slightly different results compared to most 13345of other chips. See @option{-ffloat-store} for more detailed description. 13346 13347This is the default choice for i386 compiler. 13348 13349@item sse 13350Use scalar floating-point instructions present in the SSE instruction set. 13351This instruction set is supported by Pentium3 and newer chips, in the AMD line 13352by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE 13353instruction set supports only single-precision arithmetic, thus the double and 13354extended-precision arithmetic are still done using 387. A later version, present 13355only in Pentium4 and the future AMD x86-64 chips, supports double-precision 13356arithmetic too. 13357 13358For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse} 13359or @option{-msse2} switches to enable SSE extensions and make this option 13360effective. For the x86-64 compiler, these extensions are enabled by default. 13361 13362The resulting code should be considerably faster in the majority of cases and avoid 13363the numerical instability problems of 387 code, but may break some existing 13364code that expects temporaries to be 80 bits. 13365 13366This is the default choice for the x86-64 compiler. 13367 13368@item sse,387 13369@itemx sse+387 13370@itemx both 13371Attempt to utilize both instruction sets at once. This effectively double the 13372amount of available registers and on chips with separate execution units for 13373387 and SSE the execution resources too. Use this option with care, as it is 13374still experimental, because the GCC register allocator does not model separate 13375functional units well resulting in instable performance. 13376@end table 13377 13378@item -masm=@var{dialect} 13379@opindex masm=@var{dialect} 13380Output asm instructions using selected @var{dialect}. Supported 13381choices are @samp{intel} or @samp{att} (the default one). Darwin does 13382not support @samp{intel}. 13383 13384@item -mieee-fp 13385@itemx -mno-ieee-fp 13386@opindex mieee-fp 13387@opindex mno-ieee-fp 13388Control whether or not the compiler uses IEEE floating-point 13389comparisons. These handle correctly the case where the result of a 13390comparison is unordered. 13391 13392@item -msoft-float 13393@opindex msoft-float 13394Generate output containing library calls for floating point. 13395@strong{Warning:} the requisite libraries are not part of GCC@. 13396Normally the facilities of the machine's usual C compiler are used, but 13397this can't be done directly in cross-compilation. You must make your 13398own arrangements to provide suitable library functions for 13399cross-compilation. 13400 13401On machines where a function returns floating-point results in the 80387 13402register stack, some floating-point opcodes may be emitted even if 13403@option{-msoft-float} is used. 13404 13405@item -mno-fp-ret-in-387 13406@opindex mno-fp-ret-in-387 13407Do not use the FPU registers for return values of functions. 13408 13409The usual calling convention has functions return values of types 13410@code{float} and @code{double} in an FPU register, even if there 13411is no FPU@. The idea is that the operating system should emulate 13412an FPU@. 13413 13414The option @option{-mno-fp-ret-in-387} causes such values to be returned 13415in ordinary CPU registers instead. 13416 13417@item -mno-fancy-math-387 13418@opindex mno-fancy-math-387 13419Some 387 emulators do not support the @code{sin}, @code{cos} and 13420@code{sqrt} instructions for the 387. Specify this option to avoid 13421generating those instructions. This option is the default on FreeBSD, 13422OpenBSD and NetBSD@. This option is overridden when @option{-march} 13423indicates that the target CPU will always have an FPU and so the 13424instruction will not need emulation. As of revision 2.6.1, these 13425instructions are not generated unless you also use the 13426@option{-funsafe-math-optimizations} switch. 13427 13428@item -malign-double 13429@itemx -mno-align-double 13430@opindex malign-double 13431@opindex mno-align-double 13432Control whether GCC aligns @code{double}, @code{long double}, and 13433@code{long long} variables on a two-word boundary or a one-word 13434boundary. Aligning @code{double} variables on a two-word boundary 13435produces code that runs somewhat faster on a @samp{Pentium} at the 13436expense of more memory. 13437 13438On x86-64, @option{-malign-double} is enabled by default. 13439 13440@strong{Warning:} if you use the @option{-malign-double} switch, 13441structures containing the above types will be aligned differently than 13442the published application binary interface specifications for the 386 13443and will not be binary compatible with structures in code compiled 13444without that switch. 13445 13446@item -m96bit-long-double 13447@itemx -m128bit-long-double 13448@opindex m96bit-long-double 13449@opindex m128bit-long-double 13450These switches control the size of @code{long double} type. The i386 13451application binary interface specifies the size to be 96 bits, 13452so @option{-m96bit-long-double} is the default in 32-bit mode. 13453 13454Modern architectures (Pentium and newer) prefer @code{long double} 13455to be aligned to an 8- or 16-byte boundary. In arrays or structures 13456conforming to the ABI, this is not possible. So specifying 13457@option{-m128bit-long-double} aligns @code{long double} 13458to a 16-byte boundary by padding the @code{long double} with an additional 1345932-bit zero. 13460 13461In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as 13462its ABI specifies that @code{long double} is to be aligned on 16-byte boundary. 13463 13464Notice that neither of these options enable any extra precision over the x87 13465standard of 80 bits for a @code{long double}. 13466 13467@strong{Warning:} if you override the default value for your target ABI, the 13468structures and arrays containing @code{long double} variables will change 13469their size as well as function calling convention for function taking 13470@code{long double} will be modified. Hence they will not be binary 13471compatible with arrays or structures in code compiled without that switch. 13472 13473@item -mlarge-data-threshold=@var{number} 13474@opindex mlarge-data-threshold=@var{number} 13475When @option{-mcmodel=medium} is specified, the data greater than 13476@var{threshold} are placed in large data section. This value must be the 13477same across all object linked into the binary and defaults to 65535. 13478 13479@item -mrtd 13480@opindex mrtd 13481Use a different function-calling convention, in which functions that 13482take a fixed number of arguments return with the @code{ret} @var{num} 13483instruction, which pops their arguments while returning. This saves one 13484instruction in the caller since there is no need to pop the arguments 13485there. 13486 13487You can specify that an individual function is called with this calling 13488sequence with the function attribute @samp{stdcall}. You can also 13489override the @option{-mrtd} option by using the function attribute 13490@samp{cdecl}. @xref{Function Attributes}. 13491 13492@strong{Warning:} this calling convention is incompatible with the one 13493normally used on Unix, so you cannot use it if you need to call 13494libraries compiled with the Unix compiler. 13495 13496Also, you must provide function prototypes for all functions that 13497take variable numbers of arguments (including @code{printf}); 13498otherwise incorrect code will be generated for calls to those 13499functions. 13500 13501In addition, seriously incorrect code will result if you call a 13502function with too many arguments. (Normally, extra arguments are 13503harmlessly ignored.) 13504 13505@item -mregparm=@var{num} 13506@opindex mregparm 13507Control how many registers are used to pass integer arguments. By 13508default, no registers are used to pass arguments, and at most 3 13509registers can be used. You can control this behavior for a specific 13510function by using the function attribute @samp{regparm}. 13511@xref{Function Attributes}. 13512 13513@strong{Warning:} if you use this switch, and 13514@var{num} is nonzero, then you must build all modules with the same 13515value, including any libraries. This includes the system libraries and 13516startup modules. 13517 13518@item -msseregparm 13519@opindex msseregparm 13520Use SSE register passing conventions for float and double arguments 13521and return values. You can control this behavior for a specific 13522function by using the function attribute @samp{sseregparm}. 13523@xref{Function Attributes}. 13524 13525@strong{Warning:} if you use this switch then you must build all 13526modules with the same value, including any libraries. This includes 13527the system libraries and startup modules. 13528 13529@item -mvect8-ret-in-mem 13530@opindex mvect8-ret-in-mem 13531Return 8-byte vectors in memory instead of MMX registers. This is the 13532default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun 13533Studio compilers until version 12. Later compiler versions (starting 13534with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which 13535is the default on Solaris@tie{}10 and later. @emph{Only} use this option if 13536you need to remain compatible with existing code produced by those 13537previous compiler versions or older versions of GCC. 13538 13539@item -mpc32 13540@itemx -mpc64 13541@itemx -mpc80 13542@opindex mpc32 13543@opindex mpc64 13544@opindex mpc80 13545 13546Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32} 13547is specified, the significands of results of floating-point operations are 13548rounded to 24 bits (single precision); @option{-mpc64} rounds the 13549significands of results of floating-point operations to 53 bits (double 13550precision) and @option{-mpc80} rounds the significands of results of 13551floating-point operations to 64 bits (extended double precision), which is 13552the default. When this option is used, floating-point operations in higher 13553precisions are not available to the programmer without setting the FPU 13554control word explicitly. 13555 13556Setting the rounding of floating-point operations to less than the default 1355780 bits can speed some programs by 2% or more. Note that some mathematical 13558libraries assume that extended-precision (80-bit) floating-point operations 13559are enabled by default; routines in such libraries could suffer significant 13560loss of accuracy, typically through so-called "catastrophic cancellation", 13561when this option is used to set the precision to less than extended precision. 13562 13563@item -mstackrealign 13564@opindex mstackrealign 13565Realign the stack at entry. On the Intel x86, the @option{-mstackrealign} 13566option will generate an alternate prologue and epilogue that realigns the 13567run-time stack if necessary. This supports mixing legacy codes that keep 13568a 4-byte aligned stack with modern codes that keep a 16-byte stack for 13569SSE compatibility. See also the attribute @code{force_align_arg_pointer}, 13570applicable to individual functions. 13571 13572@item -mpreferred-stack-boundary=@var{num} 13573@opindex mpreferred-stack-boundary 13574Attempt to keep the stack boundary aligned to a 2 raised to @var{num} 13575byte boundary. If @option{-mpreferred-stack-boundary} is not specified, 13576the default is 4 (16 bytes or 128 bits). 13577 13578@item -mincoming-stack-boundary=@var{num} 13579@opindex mincoming-stack-boundary 13580Assume the incoming stack is aligned to a 2 raised to @var{num} byte 13581boundary. If @option{-mincoming-stack-boundary} is not specified, 13582the one specified by @option{-mpreferred-stack-boundary} will be used. 13583 13584On Pentium and PentiumPro, @code{double} and @code{long double} values 13585should be aligned to an 8-byte boundary (see @option{-malign-double}) or 13586suffer significant run time performance penalties. On Pentium III, the 13587Streaming SIMD Extension (SSE) data type @code{__m128} may not work 13588properly if it is not 16-byte aligned. 13589 13590To ensure proper alignment of this values on the stack, the stack boundary 13591must be as aligned as that required by any value stored on the stack. 13592Further, every function must be generated such that it keeps the stack 13593aligned. Thus calling a function compiled with a higher preferred 13594stack boundary from a function compiled with a lower preferred stack 13595boundary will most likely misalign the stack. It is recommended that 13596libraries that use callbacks always use the default setting. 13597 13598This extra alignment does consume extra stack space, and generally 13599increases code size. Code that is sensitive to stack space usage, such 13600as embedded systems and operating system kernels, may want to reduce the 13601preferred alignment to @option{-mpreferred-stack-boundary=2}. 13602 13603@item -mmmx 13604@itemx -mno-mmx 13605@itemx -msse 13606@itemx -mno-sse 13607@itemx -msse2 13608@itemx -mno-sse2 13609@itemx -msse3 13610@itemx -mno-sse3 13611@itemx -mssse3 13612@itemx -mno-ssse3 13613@itemx -msse4.1 13614@need 800 13615@itemx -mno-sse4.1 13616@itemx -msse4.2 13617@itemx -mno-sse4.2 13618@itemx -msse4 13619@itemx -mno-sse4 13620@itemx -mavx 13621@itemx -mno-avx 13622@itemx -mavx2 13623@itemx -mno-avx2 13624@itemx -maes 13625@itemx -mno-aes 13626@itemx -mpclmul 13627@need 800 13628@itemx -mno-pclmul 13629@itemx -mfsgsbase 13630@itemx -mno-fsgsbase 13631@itemx -mrdrnd 13632@itemx -mno-rdrnd 13633@itemx -mf16c 13634@itemx -mno-f16c 13635@itemx -mfma 13636@itemx -mno-fma 13637@itemx -msse4a 13638@itemx -mno-sse4a 13639@itemx -mfma4 13640@need 800 13641@itemx -mno-fma4 13642@itemx -mxop 13643@itemx -mno-xop 13644@itemx -mlwp 13645@itemx -mno-lwp 13646@itemx -m3dnow 13647@itemx -mno-3dnow 13648@itemx -mpopcnt 13649@itemx -mno-popcnt 13650@itemx -mabm 13651@itemx -mno-abm 13652@itemx -mbmi 13653@itemx -mbmi2 13654@itemx -mno-bmi 13655@itemx -mno-bmi2 13656@itemx -mlzcnt 13657@itemx -mno-lzcnt 13658@itemx -mtbm 13659@itemx -mno-tbm 13660@opindex mmmx 13661@opindex mno-mmx 13662@opindex msse 13663@opindex mno-sse 13664@opindex m3dnow 13665@opindex mno-3dnow 13666These switches enable or disable the use of instructions in the MMX, SSE, 13667SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, F16C, 13668FMA, SSE4A, FMA4, XOP, LWP, ABM, BMI, BMI2, LZCNT or 3DNow! 13669@: extended instruction sets. 13670These extensions are also available as built-in functions: see 13671@ref{X86 Built-in Functions}, for details of the functions enabled and 13672disabled by these switches. 13673 13674To have SSE/SSE2 instructions generated automatically from floating-point 13675code (as opposed to 387 instructions), see @option{-mfpmath=sse}. 13676 13677GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it 13678generates new AVX instructions or AVX equivalence for all SSEx instructions 13679when needed. 13680 13681These options will enable GCC to use these extended instructions in 13682generated code, even without @option{-mfpmath=sse}. Applications that 13683perform run-time CPU detection must compile separate files for each 13684supported architecture, using the appropriate flags. In particular, 13685the file containing the CPU detection code should be compiled without 13686these options. 13687 13688@item -mcld 13689@opindex mcld 13690This option instructs GCC to emit a @code{cld} instruction in the prologue 13691of functions that use string instructions. String instructions depend on 13692the DF flag to select between autoincrement or autodecrement mode. While the 13693ABI specifies the DF flag to be cleared on function entry, some operating 13694systems violate this specification by not clearing the DF flag in their 13695exception dispatchers. The exception handler can be invoked with the DF flag 13696set, which leads to wrong direction mode when string instructions are used. 13697This option can be enabled by default on 32-bit x86 targets by configuring 13698GCC with the @option{--enable-cld} configure option. Generation of @code{cld} 13699instructions can be suppressed with the @option{-mno-cld} compiler option 13700in this case. 13701 13702@item -mvzeroupper 13703@opindex mvzeroupper 13704This option instructs GCC to emit a @code{vzeroupper} instruction 13705before a transfer of control flow out of the function to minimize 13706AVX to SSE transition penalty as well as remove unnecessary zeroupper 13707intrinsics. 13708 13709@item -mprefer-avx128 13710@opindex mprefer-avx128 13711This option instructs GCC to use 128-bit AVX instructions instead of 13712256-bit AVX instructions in the auto-vectorizer. 13713 13714@item -mcx16 13715@opindex mcx16 13716This option will enable GCC to use CMPXCHG16B instruction in generated code. 13717CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword) 13718data types. This is useful for high resolution counters that could be updated 13719by multiple processors (or cores). This instruction is generated as part of 13720atomic built-in functions: see @ref{__sync Builtins} or 13721@ref{__atomic Builtins} for details. 13722 13723@item -msahf 13724@opindex msahf 13725This option will enable GCC to use SAHF instruction in generated 64-bit code. 13726Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported 13727by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and 13728SAHF are load and store instructions, respectively, for certain status flags. 13729In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem} 13730or @code{remainder} built-in functions: see @ref{Other Builtins} for details. 13731 13732@item -mmovbe 13733@opindex mmovbe 13734This option will enable GCC to use movbe instruction to implement 13735@code{__builtin_bswap32} and @code{__builtin_bswap64}. 13736 13737@item -mcrc32 13738@opindex mcrc32 13739This option will enable built-in functions, @code{__builtin_ia32_crc32qi}, 13740@code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and 13741@code{__builtin_ia32_crc32di} to generate the crc32 machine instruction. 13742 13743@item -mrecip 13744@opindex mrecip 13745This option will enable GCC to use RCPSS and RSQRTSS instructions (and their 13746vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step 13747to increase precision instead of DIVSS and SQRTSS (and their vectorized 13748variants) for single-precision floating-point arguments. These instructions 13749are generated only when @option{-funsafe-math-optimizations} is enabled 13750together with @option{-finite-math-only} and @option{-fno-trapping-math}. 13751Note that while the throughput of the sequence is higher than the throughput 13752of the non-reciprocal instruction, the precision of the sequence can be 13753decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994). 13754 13755Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of RSQRTSS 13756(or RSQRTPS) already with @option{-ffast-math} (or the above option 13757combination), and doesn't need @option{-mrecip}. 13758 13759Also note that GCC emits the above sequence with additional Newton-Raphson step 13760for vectorized single-float division and vectorized @code{sqrtf(@var{x})} 13761already with @option{-ffast-math} (or the above option combination), and 13762doesn't need @option{-mrecip}. 13763 13764@item -mrecip=@var{opt} 13765@opindex mrecip=opt 13766This option allows to control which reciprocal estimate instructions 13767may be used. @var{opt} is a comma separated list of options, which may 13768be preceded by a @code{!} to invert the option: 13769@code{all}: enable all estimate instructions, 13770@code{default}: enable the default instructions, equivalent to @option{-mrecip}, 13771@code{none}: disable all estimate instructions, equivalent to @option{-mno-recip}, 13772@code{div}: enable the approximation for scalar division, 13773@code{vec-div}: enable the approximation for vectorized division, 13774@code{sqrt}: enable the approximation for scalar square root, 13775@code{vec-sqrt}: enable the approximation for vectorized square root. 13776 13777So for example, @option{-mrecip=all,!sqrt} would enable 13778all of the reciprocal approximations, except for square root. 13779 13780@item -mveclibabi=@var{type} 13781@opindex mveclibabi 13782Specifies the ABI type to use for vectorizing intrinsics using an 13783external library. Supported types are @code{svml} for the Intel short 13784vector math library and @code{acml} for the AMD math core library style 13785of interfacing. GCC will currently emit calls to @code{vmldExp2}, 13786@code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2}, 13787@code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2}, 13788@code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2}, 13789@code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2}, 13790@code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104}, 13791@code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4}, 13792@code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4}, 13793@code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4}, 13794@code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding 13795function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin}, 13796@code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2}, 13797@code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf}, 13798@code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f}, 13799@code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type 13800when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and 13801@option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI 13802compatible library will have to be specified at link time. 13803 13804@item -mabi=@var{name} 13805@opindex mabi 13806Generate code for the specified calling convention. Permissible values 13807are: @samp{sysv} for the ABI used on GNU/Linux and other systems and 13808@samp{ms} for the Microsoft ABI. The default is to use the Microsoft 13809ABI when targeting Windows. On all other systems, the default is the 13810SYSV ABI. You can control this behavior for a specific function by 13811using the function attribute @samp{ms_abi}/@samp{sysv_abi}. 13812@xref{Function Attributes}. 13813 13814@item -mtls-dialect=@var{type} 13815@opindex mtls-dialect 13816Generate code to access thread-local storage using the @samp{gnu} or 13817@samp{gnu2} conventions. @samp{gnu} is the conservative default; 13818@samp{gnu2} is more efficient, but it may add compile- and run-time 13819requirements that cannot be satisfied on all systems. 13820 13821@item -mpush-args 13822@itemx -mno-push-args 13823@opindex mpush-args 13824@opindex mno-push-args 13825Use PUSH operations to store outgoing parameters. This method is shorter 13826and usually equally fast as method using SUB/MOV operations and is enabled 13827by default. In some cases disabling it may improve performance because of 13828improved scheduling and reduced dependencies. 13829 13830@item -maccumulate-outgoing-args 13831@opindex maccumulate-outgoing-args 13832If enabled, the maximum amount of space required for outgoing arguments will be 13833computed in the function prologue. This is faster on most modern CPUs 13834because of reduced dependencies, improved scheduling and reduced stack usage 13835when preferred stack boundary is not equal to 2. The drawback is a notable 13836increase in code size. This switch implies @option{-mno-push-args}. 13837 13838@item -mthreads 13839@opindex mthreads 13840Support thread-safe exception handling on @samp{Mingw32}. Code that relies 13841on thread-safe exception handling must compile and link all code with the 13842@option{-mthreads} option. When compiling, @option{-mthreads} defines 13843@option{-D_MT}; when linking, it links in a special thread helper library 13844@option{-lmingwthrd} which cleans up per thread exception handling data. 13845 13846@item -mno-align-stringops 13847@opindex mno-align-stringops 13848Do not align destination of inlined string operations. This switch reduces 13849code size and improves performance in case the destination is already aligned, 13850but GCC doesn't know about it. 13851 13852@item -minline-all-stringops 13853@opindex minline-all-stringops 13854By default GCC inlines string operations only when the destination is 13855known to be aligned to least a 4-byte boundary. 13856This enables more inlining, increase code 13857size, but may improve performance of code that depends on fast memcpy, strlen 13858and memset for short lengths. 13859 13860@item -minline-stringops-dynamically 13861@opindex minline-stringops-dynamically 13862For string operations of unknown size, use run-time checks with 13863inline code for small blocks and a library call for large blocks. 13864 13865@item -mstringop-strategy=@var{alg} 13866@opindex mstringop-strategy=@var{alg} 13867Overwrite internal decision heuristic about particular algorithm to inline 13868string operation with. The allowed values are @code{rep_byte}, 13869@code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix 13870of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for 13871expanding inline loop, @code{libcall} for always expanding library call. 13872 13873@item -momit-leaf-frame-pointer 13874@opindex momit-leaf-frame-pointer 13875Don't keep the frame pointer in a register for leaf functions. This 13876avoids the instructions to save, set up and restore frame pointers and 13877makes an extra register available in leaf functions. The option 13878@option{-fomit-frame-pointer} removes the frame pointer for all functions, 13879which might make debugging harder. 13880 13881@item -mtls-direct-seg-refs 13882@itemx -mno-tls-direct-seg-refs 13883@opindex mtls-direct-seg-refs 13884Controls whether TLS variables may be accessed with offsets from the 13885TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit), 13886or whether the thread base pointer must be added. Whether or not this 13887is legal depends on the operating system, and whether it maps the 13888segment to cover the entire TLS area. 13889 13890For systems that use GNU libc, the default is on. 13891 13892@item -msse2avx 13893@itemx -mno-sse2avx 13894@opindex msse2avx 13895Specify that the assembler should encode SSE instructions with VEX 13896prefix. The option @option{-mavx} turns this on by default. 13897 13898@item -mfentry 13899@itemx -mno-fentry 13900@opindex mfentry 13901If profiling is active @option{-pg} put the profiling 13902counter call before prologue. 13903Note: On x86 architectures the attribute @code{ms_hook_prologue} 13904isn't possible at the moment for @option{-mfentry} and @option{-pg}. 13905 13906@item -m8bit-idiv 13907@itemx -mno-8bit-idiv 13908@opindex 8bit-idiv 13909On some processors, like Intel Atom, 8-bit unsigned integer divide is 13910much faster than 32-bit/64-bit integer divide. This option generates a 13911run-time check. If both dividend and divisor are within range of 0 13912to 255, 8-bit unsigned integer divide is used instead of 1391332-bit/64-bit integer divide. 13914 13915@item -mavx256-split-unaligned-load 13916@item -mavx256-split-unaligned-store 13917@opindex avx256-split-unaligned-load 13918@opindex avx256-split-unaligned-store 13919Split 32-byte AVX unaligned load and store. 13920 13921@end table 13922 13923These @samp{-m} switches are supported in addition to the above 13924on AMD x86-64 processors in 64-bit environments. 13925 13926@table @gcctabopt 13927@item -m32 13928@itemx -m64 13929@itemx -mx32 13930@opindex m32 13931@opindex m64 13932@opindex mx32 13933Generate code for a 32-bit or 64-bit environment. 13934The @option{-m32} option sets int, long and pointer to 32 bits and 13935generates code that runs on any i386 system. 13936The @option{-m64} option sets int to 32 bits and long and pointer 13937to 64 bits and generates code for AMD's x86-64 architecture. 13938The @option{-mx32} option sets int, long and pointer to 32 bits and 13939generates code for AMD's x86-64 architecture. 13940For darwin only the @option{-m64} option turns off the @option{-fno-pic} 13941and @option{-mdynamic-no-pic} options. 13942 13943@item -mno-red-zone 13944@opindex mno-red-zone 13945Do not use a so called red zone for x86-64 code. The red zone is mandated 13946by the x86-64 ABI, it is a 128-byte area beyond the location of the 13947stack pointer that will not be modified by signal or interrupt handlers 13948and therefore can be used for temporary data without adjusting the stack 13949pointer. The flag @option{-mno-red-zone} disables this red zone. 13950 13951@item -mcmodel=small 13952@opindex mcmodel=small 13953Generate code for the small code model: the program and its symbols must 13954be linked in the lower 2 GB of the address space. Pointers are 64 bits. 13955Programs can be statically or dynamically linked. This is the default 13956code model. 13957 13958@item -mcmodel=kernel 13959@opindex mcmodel=kernel 13960Generate code for the kernel code model. The kernel runs in the 13961negative 2 GB of the address space. 13962This model has to be used for Linux kernel code. 13963 13964@item -mcmodel=medium 13965@opindex mcmodel=medium 13966Generate code for the medium model: The program is linked in the lower 2 13967GB of the address space. Small symbols are also placed there. Symbols 13968with sizes larger than @option{-mlarge-data-threshold} are put into 13969large data or bss sections and can be located above 2GB. Programs can 13970be statically or dynamically linked. 13971 13972@item -mcmodel=large 13973@opindex mcmodel=large 13974Generate code for the large model: This model makes no assumptions 13975about addresses and sizes of sections. 13976@end table 13977 13978@node i386 and x86-64 Windows Options 13979@subsection i386 and x86-64 Windows Options 13980@cindex i386 and x86-64 Windows Options 13981 13982These additional options are available for Windows targets: 13983 13984@table @gcctabopt 13985@item -mconsole 13986@opindex mconsole 13987This option is available for Cygwin and MinGW targets. It 13988specifies that a console application is to be generated, by 13989instructing the linker to set the PE header subsystem type 13990required for console applications. 13991This is the default behavior for Cygwin and MinGW targets. 13992 13993@item -mdll 13994@opindex mdll 13995This option is available for Cygwin and MinGW targets. It 13996specifies that a DLL - a dynamic link library - is to be 13997generated, enabling the selection of the required runtime 13998startup object and entry point. 13999 14000@item -mnop-fun-dllimport 14001@opindex mnop-fun-dllimport 14002This option is available for Cygwin and MinGW targets. It 14003specifies that the dllimport attribute should be ignored. 14004 14005@item -mthread 14006@opindex mthread 14007This option is available for MinGW targets. It specifies 14008that MinGW-specific thread support is to be used. 14009 14010@item -municode 14011@opindex municode 14012This option is available for mingw-w64 targets. It specifies 14013that the UNICODE macro is getting pre-defined and that the 14014unicode capable runtime startup code is chosen. 14015 14016@item -mwin32 14017@opindex mwin32 14018This option is available for Cygwin and MinGW targets. It 14019specifies that the typical Windows pre-defined macros are to 14020be set in the pre-processor, but does not influence the choice 14021of runtime library/startup code. 14022 14023@item -mwindows 14024@opindex mwindows 14025This option is available for Cygwin and MinGW targets. It 14026specifies that a GUI application is to be generated by 14027instructing the linker to set the PE header subsystem type 14028appropriately. 14029 14030@item -fno-set-stack-executable 14031@opindex fno-set-stack-executable 14032This option is available for MinGW targets. It specifies that 14033the executable flag for stack used by nested functions isn't 14034set. This is necessary for binaries running in kernel mode of 14035Windows, as there the user32 API, which is used to set executable 14036privileges, isn't available. 14037 14038@item -mpe-aligned-commons 14039@opindex mpe-aligned-commons 14040This option is available for Cygwin and MinGW targets. It 14041specifies that the GNU extension to the PE file format that 14042permits the correct alignment of COMMON variables should be 14043used when generating code. It will be enabled by default if 14044GCC detects that the target assembler found during configuration 14045supports the feature. 14046@end table 14047 14048See also under @ref{i386 and x86-64 Options} for standard options. 14049 14050@node IA-64 Options 14051@subsection IA-64 Options 14052@cindex IA-64 Options 14053 14054These are the @samp{-m} options defined for the Intel IA-64 architecture. 14055 14056@table @gcctabopt 14057@item -mbig-endian 14058@opindex mbig-endian 14059Generate code for a big-endian target. This is the default for HP-UX@. 14060 14061@item -mlittle-endian 14062@opindex mlittle-endian 14063Generate code for a little-endian target. This is the default for AIX5 14064and GNU/Linux. 14065 14066@item -mgnu-as 14067@itemx -mno-gnu-as 14068@opindex mgnu-as 14069@opindex mno-gnu-as 14070Generate (or don't) code for the GNU assembler. This is the default. 14071@c Also, this is the default if the configure option @option{--with-gnu-as} 14072@c is used. 14073 14074@item -mgnu-ld 14075@itemx -mno-gnu-ld 14076@opindex mgnu-ld 14077@opindex mno-gnu-ld 14078Generate (or don't) code for the GNU linker. This is the default. 14079@c Also, this is the default if the configure option @option{--with-gnu-ld} 14080@c is used. 14081 14082@item -mno-pic 14083@opindex mno-pic 14084Generate code that does not use a global pointer register. The result 14085is not position independent code, and violates the IA-64 ABI@. 14086 14087@item -mvolatile-asm-stop 14088@itemx -mno-volatile-asm-stop 14089@opindex mvolatile-asm-stop 14090@opindex mno-volatile-asm-stop 14091Generate (or don't) a stop bit immediately before and after volatile asm 14092statements. 14093 14094@item -mregister-names 14095@itemx -mno-register-names 14096@opindex mregister-names 14097@opindex mno-register-names 14098Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for 14099the stacked registers. This may make assembler output more readable. 14100 14101@item -mno-sdata 14102@itemx -msdata 14103@opindex mno-sdata 14104@opindex msdata 14105Disable (or enable) optimizations that use the small data section. This may 14106be useful for working around optimizer bugs. 14107 14108@item -mconstant-gp 14109@opindex mconstant-gp 14110Generate code that uses a single constant global pointer value. This is 14111useful when compiling kernel code. 14112 14113@item -mauto-pic 14114@opindex mauto-pic 14115Generate code that is self-relocatable. This implies @option{-mconstant-gp}. 14116This is useful when compiling firmware code. 14117 14118@item -minline-float-divide-min-latency 14119@opindex minline-float-divide-min-latency 14120Generate code for inline divides of floating-point values 14121using the minimum latency algorithm. 14122 14123@item -minline-float-divide-max-throughput 14124@opindex minline-float-divide-max-throughput 14125Generate code for inline divides of floating-point values 14126using the maximum throughput algorithm. 14127 14128@item -mno-inline-float-divide 14129@opindex mno-inline-float-divide 14130Do not generate inline code for divides of floating-point values. 14131 14132@item -minline-int-divide-min-latency 14133@opindex minline-int-divide-min-latency 14134Generate code for inline divides of integer values 14135using the minimum latency algorithm. 14136 14137@item -minline-int-divide-max-throughput 14138@opindex minline-int-divide-max-throughput 14139Generate code for inline divides of integer values 14140using the maximum throughput algorithm. 14141 14142@item -mno-inline-int-divide 14143@opindex mno-inline-int-divide 14144Do not generate inline code for divides of integer values. 14145 14146@item -minline-sqrt-min-latency 14147@opindex minline-sqrt-min-latency 14148Generate code for inline square roots 14149using the minimum latency algorithm. 14150 14151@item -minline-sqrt-max-throughput 14152@opindex minline-sqrt-max-throughput 14153Generate code for inline square roots 14154using the maximum throughput algorithm. 14155 14156@item -mno-inline-sqrt 14157@opindex mno-inline-sqrt 14158Do not generate inline code for sqrt. 14159 14160@item -mfused-madd 14161@itemx -mno-fused-madd 14162@opindex mfused-madd 14163@opindex mno-fused-madd 14164Do (don't) generate code that uses the fused multiply/add or multiply/subtract 14165instructions. The default is to use these instructions. 14166 14167@item -mno-dwarf2-asm 14168@itemx -mdwarf2-asm 14169@opindex mno-dwarf2-asm 14170@opindex mdwarf2-asm 14171Don't (or do) generate assembler code for the DWARF2 line number debugging 14172info. This may be useful when not using the GNU assembler. 14173 14174@item -mearly-stop-bits 14175@itemx -mno-early-stop-bits 14176@opindex mearly-stop-bits 14177@opindex mno-early-stop-bits 14178Allow stop bits to be placed earlier than immediately preceding the 14179instruction that triggered the stop bit. This can improve instruction 14180scheduling, but does not always do so. 14181 14182@item -mfixed-range=@var{register-range} 14183@opindex mfixed-range 14184Generate code treating the given register range as fixed registers. 14185A fixed register is one that the register allocator can not use. This is 14186useful when compiling kernel code. A register range is specified as 14187two registers separated by a dash. Multiple register ranges can be 14188specified separated by a comma. 14189 14190@item -mtls-size=@var{tls-size} 14191@opindex mtls-size 14192Specify bit size of immediate TLS offsets. Valid values are 14, 22, and 1419364. 14194 14195@item -mtune=@var{cpu-type} 14196@opindex mtune 14197Tune the instruction scheduling for a particular CPU, Valid values are 14198itanium, itanium1, merced, itanium2, and mckinley. 14199 14200@item -milp32 14201@itemx -mlp64 14202@opindex milp32 14203@opindex mlp64 14204Generate code for a 32-bit or 64-bit environment. 14205The 32-bit environment sets int, long and pointer to 32 bits. 14206The 64-bit environment sets int to 32 bits and long and pointer 14207to 64 bits. These are HP-UX specific flags. 14208 14209@item -mno-sched-br-data-spec 14210@itemx -msched-br-data-spec 14211@opindex mno-sched-br-data-spec 14212@opindex msched-br-data-spec 14213(Dis/En)able data speculative scheduling before reload. 14214This will result in generation of the ld.a instructions and 14215the corresponding check instructions (ld.c / chk.a). 14216The default is 'disable'. 14217 14218@item -msched-ar-data-spec 14219@itemx -mno-sched-ar-data-spec 14220@opindex msched-ar-data-spec 14221@opindex mno-sched-ar-data-spec 14222(En/Dis)able data speculative scheduling after reload. 14223This will result in generation of the ld.a instructions and 14224the corresponding check instructions (ld.c / chk.a). 14225The default is 'enable'. 14226 14227@item -mno-sched-control-spec 14228@itemx -msched-control-spec 14229@opindex mno-sched-control-spec 14230@opindex msched-control-spec 14231(Dis/En)able control speculative scheduling. This feature is 14232available only during region scheduling (i.e.@: before reload). 14233This will result in generation of the ld.s instructions and 14234the corresponding check instructions chk.s . 14235The default is 'disable'. 14236 14237@item -msched-br-in-data-spec 14238@itemx -mno-sched-br-in-data-spec 14239@opindex msched-br-in-data-spec 14240@opindex mno-sched-br-in-data-spec 14241(En/Dis)able speculative scheduling of the instructions that 14242are dependent on the data speculative loads before reload. 14243This is effective only with @option{-msched-br-data-spec} enabled. 14244The default is 'enable'. 14245 14246@item -msched-ar-in-data-spec 14247@itemx -mno-sched-ar-in-data-spec 14248@opindex msched-ar-in-data-spec 14249@opindex mno-sched-ar-in-data-spec 14250(En/Dis)able speculative scheduling of the instructions that 14251are dependent on the data speculative loads after reload. 14252This is effective only with @option{-msched-ar-data-spec} enabled. 14253The default is 'enable'. 14254 14255@item -msched-in-control-spec 14256@itemx -mno-sched-in-control-spec 14257@opindex msched-in-control-spec 14258@opindex mno-sched-in-control-spec 14259(En/Dis)able speculative scheduling of the instructions that 14260are dependent on the control speculative loads. 14261This is effective only with @option{-msched-control-spec} enabled. 14262The default is 'enable'. 14263 14264@item -mno-sched-prefer-non-data-spec-insns 14265@itemx -msched-prefer-non-data-spec-insns 14266@opindex mno-sched-prefer-non-data-spec-insns 14267@opindex msched-prefer-non-data-spec-insns 14268If enabled, data speculative instructions will be chosen for schedule 14269only if there are no other choices at the moment. This will make 14270the use of the data speculation much more conservative. 14271The default is 'disable'. 14272 14273@item -mno-sched-prefer-non-control-spec-insns 14274@itemx -msched-prefer-non-control-spec-insns 14275@opindex mno-sched-prefer-non-control-spec-insns 14276@opindex msched-prefer-non-control-spec-insns 14277If enabled, control speculative instructions will be chosen for schedule 14278only if there are no other choices at the moment. This will make 14279the use of the control speculation much more conservative. 14280The default is 'disable'. 14281 14282@item -mno-sched-count-spec-in-critical-path 14283@itemx -msched-count-spec-in-critical-path 14284@opindex mno-sched-count-spec-in-critical-path 14285@opindex msched-count-spec-in-critical-path 14286If enabled, speculative dependencies will be considered during 14287computation of the instructions priorities. This will make the use of the 14288speculation a bit more conservative. 14289The default is 'disable'. 14290 14291@item -msched-spec-ldc 14292@opindex msched-spec-ldc 14293Use a simple data speculation check. This option is on by default. 14294 14295@item -msched-control-spec-ldc 14296@opindex msched-spec-ldc 14297Use a simple check for control speculation. This option is on by default. 14298 14299@item -msched-stop-bits-after-every-cycle 14300@opindex msched-stop-bits-after-every-cycle 14301Place a stop bit after every cycle when scheduling. This option is on 14302by default. 14303 14304@item -msched-fp-mem-deps-zero-cost 14305@opindex msched-fp-mem-deps-zero-cost 14306Assume that floating-point stores and loads are not likely to cause a conflict 14307when placed into the same instruction group. This option is disabled by 14308default. 14309 14310@item -msel-sched-dont-check-control-spec 14311@opindex msel-sched-dont-check-control-spec 14312Generate checks for control speculation in selective scheduling. 14313This flag is disabled by default. 14314 14315@item -msched-max-memory-insns=@var{max-insns} 14316@opindex msched-max-memory-insns 14317Limit on the number of memory insns per instruction group, giving lower 14318priority to subsequent memory insns attempting to schedule in the same 14319instruction group. Frequently useful to prevent cache bank conflicts. 14320The default value is 1. 14321 14322@item -msched-max-memory-insns-hard-limit 14323@opindex msched-max-memory-insns-hard-limit 14324Disallow more than `msched-max-memory-insns' in instruction group. 14325Otherwise, limit is `soft' meaning that we would prefer non-memory operations 14326when limit is reached but may still schedule memory operations. 14327 14328@end table 14329 14330@node IA-64/VMS Options 14331@subsection IA-64/VMS Options 14332 14333These @samp{-m} options are defined for the IA-64/VMS implementations: 14334 14335@table @gcctabopt 14336@item -mvms-return-codes 14337@opindex mvms-return-codes 14338Return VMS condition codes from main. The default is to return POSIX 14339style condition (e.g.@ error) codes. 14340 14341@item -mdebug-main=@var{prefix} 14342@opindex mdebug-main=@var{prefix} 14343Flag the first routine whose name starts with @var{prefix} as the main 14344routine for the debugger. 14345 14346@item -mmalloc64 14347@opindex mmalloc64 14348Default to 64-bit memory allocation routines. 14349@end table 14350 14351@node LM32 Options 14352@subsection LM32 Options 14353@cindex LM32 options 14354 14355These @option{-m} options are defined for the Lattice Mico32 architecture: 14356 14357@table @gcctabopt 14358@item -mbarrel-shift-enabled 14359@opindex mbarrel-shift-enabled 14360Enable barrel-shift instructions. 14361 14362@item -mdivide-enabled 14363@opindex mdivide-enabled 14364Enable divide and modulus instructions. 14365 14366@item -mmultiply-enabled 14367@opindex multiply-enabled 14368Enable multiply instructions. 14369 14370@item -msign-extend-enabled 14371@opindex msign-extend-enabled 14372Enable sign extend instructions. 14373 14374@item -muser-enabled 14375@opindex muser-enabled 14376Enable user-defined instructions. 14377 14378@end table 14379 14380@node M32C Options 14381@subsection M32C Options 14382@cindex M32C options 14383 14384@table @gcctabopt 14385@item -mcpu=@var{name} 14386@opindex mcpu= 14387Select the CPU for which code is generated. @var{name} may be one of 14388@samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to 14389/60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for 14390the M32C/80 series. 14391 14392@item -msim 14393@opindex msim 14394Specifies that the program will be run on the simulator. This causes 14395an alternate runtime library to be linked in which supports, for 14396example, file I/O@. You must not use this option when generating 14397programs that will run on real hardware; you must provide your own 14398runtime library for whatever I/O functions are needed. 14399 14400@item -memregs=@var{number} 14401@opindex memregs= 14402Specifies the number of memory-based pseudo-registers GCC will use 14403during code generation. These pseudo-registers will be used like real 14404registers, so there is a tradeoff between GCC's ability to fit the 14405code into available registers, and the performance penalty of using 14406memory instead of registers. Note that all modules in a program must 14407be compiled with the same value for this option. Because of that, you 14408must not use this option with the default runtime libraries gcc 14409builds. 14410 14411@end table 14412 14413@node M32R/D Options 14414@subsection M32R/D Options 14415@cindex M32R/D options 14416 14417These @option{-m} options are defined for Renesas M32R/D architectures: 14418 14419@table @gcctabopt 14420@item -m32r2 14421@opindex m32r2 14422Generate code for the M32R/2@. 14423 14424@item -m32rx 14425@opindex m32rx 14426Generate code for the M32R/X@. 14427 14428@item -m32r 14429@opindex m32r 14430Generate code for the M32R@. This is the default. 14431 14432@item -mmodel=small 14433@opindex mmodel=small 14434Assume all objects live in the lower 16MB of memory (so that their addresses 14435can be loaded with the @code{ld24} instruction), and assume all subroutines 14436are reachable with the @code{bl} instruction. 14437This is the default. 14438 14439The addressability of a particular object can be set with the 14440@code{model} attribute. 14441 14442@item -mmodel=medium 14443@opindex mmodel=medium 14444Assume objects may be anywhere in the 32-bit address space (the compiler 14445will generate @code{seth/add3} instructions to load their addresses), and 14446assume all subroutines are reachable with the @code{bl} instruction. 14447 14448@item -mmodel=large 14449@opindex mmodel=large 14450Assume objects may be anywhere in the 32-bit address space (the compiler 14451will generate @code{seth/add3} instructions to load their addresses), and 14452assume subroutines may not be reachable with the @code{bl} instruction 14453(the compiler will generate the much slower @code{seth/add3/jl} 14454instruction sequence). 14455 14456@item -msdata=none 14457@opindex msdata=none 14458Disable use of the small data area. Variables will be put into 14459one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the 14460@code{section} attribute has been specified). 14461This is the default. 14462 14463The small data area consists of sections @samp{.sdata} and @samp{.sbss}. 14464Objects may be explicitly put in the small data area with the 14465@code{section} attribute using one of these sections. 14466 14467@item -msdata=sdata 14468@opindex msdata=sdata 14469Put small global and static data in the small data area, but do not 14470generate special code to reference them. 14471 14472@item -msdata=use 14473@opindex msdata=use 14474Put small global and static data in the small data area, and generate 14475special instructions to reference them. 14476 14477@item -G @var{num} 14478@opindex G 14479@cindex smaller data references 14480Put global and static objects less than or equal to @var{num} bytes 14481into the small data or bss sections instead of the normal data or bss 14482sections. The default value of @var{num} is 8. 14483The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use} 14484for this option to have any effect. 14485 14486All modules should be compiled with the same @option{-G @var{num}} value. 14487Compiling with different values of @var{num} may or may not work; if it 14488doesn't the linker will give an error message---incorrect code will not be 14489generated. 14490 14491@item -mdebug 14492@opindex mdebug 14493Makes the M32R specific code in the compiler display some statistics 14494that might help in debugging programs. 14495 14496@item -malign-loops 14497@opindex malign-loops 14498Align all loops to a 32-byte boundary. 14499 14500@item -mno-align-loops 14501@opindex mno-align-loops 14502Do not enforce a 32-byte alignment for loops. This is the default. 14503 14504@item -missue-rate=@var{number} 14505@opindex missue-rate=@var{number} 14506Issue @var{number} instructions per cycle. @var{number} can only be 1 14507or 2. 14508 14509@item -mbranch-cost=@var{number} 14510@opindex mbranch-cost=@var{number} 14511@var{number} can only be 1 or 2. If it is 1 then branches will be 14512preferred over conditional code, if it is 2, then the opposite will 14513apply. 14514 14515@item -mflush-trap=@var{number} 14516@opindex mflush-trap=@var{number} 14517Specifies the trap number to use to flush the cache. The default is 1451812. Valid numbers are between 0 and 15 inclusive. 14519 14520@item -mno-flush-trap 14521@opindex mno-flush-trap 14522Specifies that the cache cannot be flushed by using a trap. 14523 14524@item -mflush-func=@var{name} 14525@opindex mflush-func=@var{name} 14526Specifies the name of the operating system function to call to flush 14527the cache. The default is @emph{_flush_cache}, but a function call 14528will only be used if a trap is not available. 14529 14530@item -mno-flush-func 14531@opindex mno-flush-func 14532Indicates that there is no OS function for flushing the cache. 14533 14534@end table 14535 14536@node M680x0 Options 14537@subsection M680x0 Options 14538@cindex M680x0 options 14539 14540These are the @samp{-m} options defined for M680x0 and ColdFire processors. 14541The default settings depend on which architecture was selected when 14542the compiler was configured; the defaults for the most common choices 14543are given below. 14544 14545@table @gcctabopt 14546@item -march=@var{arch} 14547@opindex march 14548Generate code for a specific M680x0 or ColdFire instruction set 14549architecture. Permissible values of @var{arch} for M680x0 14550architectures are: @samp{68000}, @samp{68010}, @samp{68020}, 14551@samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire 14552architectures are selected according to Freescale's ISA classification 14553and the permissible values are: @samp{isaa}, @samp{isaaplus}, 14554@samp{isab} and @samp{isac}. 14555 14556gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating 14557code for a ColdFire target. The @var{arch} in this macro is one of the 14558@option{-march} arguments given above. 14559 14560When used together, @option{-march} and @option{-mtune} select code 14561that runs on a family of similar processors but that is optimized 14562for a particular microarchitecture. 14563 14564@item -mcpu=@var{cpu} 14565@opindex mcpu 14566Generate code for a specific M680x0 or ColdFire processor. 14567The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020}, 14568@samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332} 14569and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table 14570below, which also classifies the CPUs into families: 14571 14572@multitable @columnfractions 0.20 0.80 14573@item @strong{Family} @tab @strong{@samp{-mcpu} arguments} 14574@item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe} 14575@item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206} 14576@item @samp{5206e} @tab @samp{5206e} 14577@item @samp{5208} @tab @samp{5207} @samp{5208} 14578@item @samp{5211a} @tab @samp{5210a} @samp{5211a} 14579@item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213} 14580@item @samp{5216} @tab @samp{5214} @samp{5216} 14581@item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235} 14582@item @samp{5225} @tab @samp{5224} @samp{5225} 14583@item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259} 14584@item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x} 14585@item @samp{5249} @tab @samp{5249} 14586@item @samp{5250} @tab @samp{5250} 14587@item @samp{5271} @tab @samp{5270} @samp{5271} 14588@item @samp{5272} @tab @samp{5272} 14589@item @samp{5275} @tab @samp{5274} @samp{5275} 14590@item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x} 14591@item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017} 14592@item @samp{5307} @tab @samp{5307} 14593@item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x} 14594@item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x} 14595@item @samp{5407} @tab @samp{5407} 14596@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} 14597@end multitable 14598 14599@option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if 14600@var{arch} is compatible with @var{cpu}. Other combinations of 14601@option{-mcpu} and @option{-march} are rejected. 14602 14603gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target 14604@var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}}, 14605where the value of @var{family} is given by the table above. 14606 14607@item -mtune=@var{tune} 14608@opindex mtune 14609Tune the code for a particular microarchitecture, within the 14610constraints set by @option{-march} and @option{-mcpu}. 14611The M680x0 microarchitectures are: @samp{68000}, @samp{68010}, 14612@samp{68020}, @samp{68030}, @samp{68040}, @samp{68060} 14613and @samp{cpu32}. The ColdFire microarchitectures 14614are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}. 14615 14616You can also use @option{-mtune=68020-40} for code that needs 14617to run relatively well on 68020, 68030 and 68040 targets. 14618@option{-mtune=68020-60} is similar but includes 68060 targets 14619as well. These two options select the same tuning decisions as 14620@option{-m68020-40} and @option{-m68020-60} respectively. 14621 14622gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__} 14623when tuning for 680x0 architecture @var{arch}. It also defines 14624@samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std} 14625option is used. If gcc is tuning for a range of architectures, 14626as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60}, 14627it defines the macros for every architecture in the range. 14628 14629gcc also defines the macro @samp{__m@var{uarch}__} when tuning for 14630ColdFire microarchitecture @var{uarch}, where @var{uarch} is one 14631of the arguments given above. 14632 14633@item -m68000 14634@itemx -mc68000 14635@opindex m68000 14636@opindex mc68000 14637Generate output for a 68000. This is the default 14638when the compiler is configured for 68000-based systems. 14639It is equivalent to @option{-march=68000}. 14640 14641Use this option for microcontrollers with a 68000 or EC000 core, 14642including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356. 14643 14644@item -m68010 14645@opindex m68010 14646Generate output for a 68010. This is the default 14647when the compiler is configured for 68010-based systems. 14648It is equivalent to @option{-march=68010}. 14649 14650@item -m68020 14651@itemx -mc68020 14652@opindex m68020 14653@opindex mc68020 14654Generate output for a 68020. This is the default 14655when the compiler is configured for 68020-based systems. 14656It is equivalent to @option{-march=68020}. 14657 14658@item -m68030 14659@opindex m68030 14660Generate output for a 68030. This is the default when the compiler is 14661configured for 68030-based systems. It is equivalent to 14662@option{-march=68030}. 14663 14664@item -m68040 14665@opindex m68040 14666Generate output for a 68040. This is the default when the compiler is 14667configured for 68040-based systems. It is equivalent to 14668@option{-march=68040}. 14669 14670This option inhibits the use of 68881/68882 instructions that have to be 14671emulated by software on the 68040. Use this option if your 68040 does not 14672have code to emulate those instructions. 14673 14674@item -m68060 14675@opindex m68060 14676Generate output for a 68060. This is the default when the compiler is 14677configured for 68060-based systems. It is equivalent to 14678@option{-march=68060}. 14679 14680This option inhibits the use of 68020 and 68881/68882 instructions that 14681have to be emulated by software on the 68060. Use this option if your 68060 14682does not have code to emulate those instructions. 14683 14684@item -mcpu32 14685@opindex mcpu32 14686Generate output for a CPU32. This is the default 14687when the compiler is configured for CPU32-based systems. 14688It is equivalent to @option{-march=cpu32}. 14689 14690Use this option for microcontrollers with a 14691CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334, 1469268336, 68340, 68341, 68349 and 68360. 14693 14694@item -m5200 14695@opindex m5200 14696Generate output for a 520X ColdFire CPU@. This is the default 14697when the compiler is configured for 520X-based systems. 14698It is equivalent to @option{-mcpu=5206}, and is now deprecated 14699in favor of that option. 14700 14701Use this option for microcontroller with a 5200 core, including 14702the MCF5202, MCF5203, MCF5204 and MCF5206. 14703 14704@item -m5206e 14705@opindex m5206e 14706Generate output for a 5206e ColdFire CPU@. The option is now 14707deprecated in favor of the equivalent @option{-mcpu=5206e}. 14708 14709@item -m528x 14710@opindex m528x 14711Generate output for a member of the ColdFire 528X family. 14712The option is now deprecated in favor of the equivalent 14713@option{-mcpu=528x}. 14714 14715@item -m5307 14716@opindex m5307 14717Generate output for a ColdFire 5307 CPU@. The option is now deprecated 14718in favor of the equivalent @option{-mcpu=5307}. 14719 14720@item -m5407 14721@opindex m5407 14722Generate output for a ColdFire 5407 CPU@. The option is now deprecated 14723in favor of the equivalent @option{-mcpu=5407}. 14724 14725@item -mcfv4e 14726@opindex mcfv4e 14727Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x). 14728This includes use of hardware floating-point instructions. 14729The option is equivalent to @option{-mcpu=547x}, and is now 14730deprecated in favor of that option. 14731 14732@item -m68020-40 14733@opindex m68020-40 14734Generate output for a 68040, without using any of the new instructions. 14735This results in code that can run relatively efficiently on either a 1473668020/68881 or a 68030 or a 68040. The generated code does use the 1473768881 instructions that are emulated on the 68040. 14738 14739The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}. 14740 14741@item -m68020-60 14742@opindex m68020-60 14743Generate output for a 68060, without using any of the new instructions. 14744This results in code that can run relatively efficiently on either a 1474568020/68881 or a 68030 or a 68040. The generated code does use the 1474668881 instructions that are emulated on the 68060. 14747 14748The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}. 14749 14750@item -mhard-float 14751@itemx -m68881 14752@opindex mhard-float 14753@opindex m68881 14754Generate floating-point instructions. This is the default for 68020 14755and above, and for ColdFire devices that have an FPU@. It defines the 14756macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__} 14757on ColdFire targets. 14758 14759@item -msoft-float 14760@opindex msoft-float 14761Do not generate floating-point instructions; use library calls instead. 14762This is the default for 68000, 68010, and 68832 targets. It is also 14763the default for ColdFire devices that have no FPU. 14764 14765@item -mdiv 14766@itemx -mno-div 14767@opindex mdiv 14768@opindex mno-div 14769Generate (do not generate) ColdFire hardware divide and remainder 14770instructions. If @option{-march} is used without @option{-mcpu}, 14771the default is ``on'' for ColdFire architectures and ``off'' for M680x0 14772architectures. Otherwise, the default is taken from the target CPU 14773(either the default CPU, or the one specified by @option{-mcpu}). For 14774example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for 14775@option{-mcpu=5206e}. 14776 14777gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled. 14778 14779@item -mshort 14780@opindex mshort 14781Consider type @code{int} to be 16 bits wide, like @code{short int}. 14782Additionally, parameters passed on the stack are also aligned to a 1478316-bit boundary even on targets whose API mandates promotion to 32-bit. 14784 14785@item -mno-short 14786@opindex mno-short 14787Do not consider type @code{int} to be 16 bits wide. This is the default. 14788 14789@item -mnobitfield 14790@itemx -mno-bitfield 14791@opindex mnobitfield 14792@opindex mno-bitfield 14793Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32} 14794and @option{-m5200} options imply @w{@option{-mnobitfield}}. 14795 14796@item -mbitfield 14797@opindex mbitfield 14798Do use the bit-field instructions. The @option{-m68020} option implies 14799@option{-mbitfield}. This is the default if you use a configuration 14800designed for a 68020. 14801 14802@item -mrtd 14803@opindex mrtd 14804Use a different function-calling convention, in which functions 14805that take a fixed number of arguments return with the @code{rtd} 14806instruction, which pops their arguments while returning. This 14807saves one instruction in the caller since there is no need to pop 14808the arguments there. 14809 14810This calling convention is incompatible with the one normally 14811used on Unix, so you cannot use it if you need to call libraries 14812compiled with the Unix compiler. 14813 14814Also, you must provide function prototypes for all functions that 14815take variable numbers of arguments (including @code{printf}); 14816otherwise incorrect code will be generated for calls to those 14817functions. 14818 14819In addition, seriously incorrect code will result if you call a 14820function with too many arguments. (Normally, extra arguments are 14821harmlessly ignored.) 14822 14823The @code{rtd} instruction is supported by the 68010, 68020, 68030, 1482468040, 68060 and CPU32 processors, but not by the 68000 or 5200. 14825 14826@item -mno-rtd 14827@opindex mno-rtd 14828Do not use the calling conventions selected by @option{-mrtd}. 14829This is the default. 14830 14831@item -malign-int 14832@itemx -mno-align-int 14833@opindex malign-int 14834@opindex mno-align-int 14835Control whether GCC aligns @code{int}, @code{long}, @code{long long}, 14836@code{float}, @code{double}, and @code{long double} variables on a 32-bit 14837boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}). 14838Aligning variables on 32-bit boundaries produces code that runs somewhat 14839faster on processors with 32-bit busses at the expense of more memory. 14840 14841@strong{Warning:} if you use the @option{-malign-int} switch, GCC will 14842align structures containing the above types differently than 14843most published application binary interface specifications for the m68k. 14844 14845@item -mpcrel 14846@opindex mpcrel 14847Use the pc-relative addressing mode of the 68000 directly, instead of 14848using a global offset table. At present, this option implies @option{-fpic}, 14849allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is 14850not presently supported with @option{-mpcrel}, though this could be supported for 1485168020 and higher processors. 14852 14853@item -mno-strict-align 14854@itemx -mstrict-align 14855@opindex mno-strict-align 14856@opindex mstrict-align 14857Do not (do) assume that unaligned memory references will be handled by 14858the system. 14859 14860@item -msep-data 14861Generate code that allows the data segment to be located in a different 14862area of memory from the text segment. This allows for execute in place in 14863an environment without virtual memory management. This option implies 14864@option{-fPIC}. 14865 14866@item -mno-sep-data 14867Generate code that assumes that the data segment follows the text segment. 14868This is the default. 14869 14870@item -mid-shared-library 14871Generate code that supports shared libraries via the library ID method. 14872This allows for execute in place and shared libraries in an environment 14873without virtual memory management. This option implies @option{-fPIC}. 14874 14875@item -mno-id-shared-library 14876Generate code that doesn't assume ID based shared libraries are being used. 14877This is the default. 14878 14879@item -mshared-library-id=n 14880Specified the identification number of the ID based shared library being 14881compiled. Specifying a value of 0 will generate more compact code, specifying 14882other values will force the allocation of that number to the current 14883library but is no more space or time efficient than omitting this option. 14884 14885@item -mxgot 14886@itemx -mno-xgot 14887@opindex mxgot 14888@opindex mno-xgot 14889When generating position-independent code for ColdFire, generate code 14890that works if the GOT has more than 8192 entries. This code is 14891larger and slower than code generated without this option. On M680x0 14892processors, this option is not needed; @option{-fPIC} suffices. 14893 14894GCC normally uses a single instruction to load values from the GOT@. 14895While this is relatively efficient, it only works if the GOT 14896is smaller than about 64k. Anything larger causes the linker 14897to report an error such as: 14898 14899@cindex relocation truncated to fit (ColdFire) 14900@smallexample 14901relocation truncated to fit: R_68K_GOT16O foobar 14902@end smallexample 14903 14904If this happens, you should recompile your code with @option{-mxgot}. 14905It should then work with very large GOTs. However, code generated with 14906@option{-mxgot} is less efficient, since it takes 4 instructions to fetch 14907the value of a global symbol. 14908 14909Note that some linkers, including newer versions of the GNU linker, 14910can create multiple GOTs and sort GOT entries. If you have such a linker, 14911you should only need to use @option{-mxgot} when compiling a single 14912object file that accesses more than 8192 GOT entries. Very few do. 14913 14914These options have no effect unless GCC is generating 14915position-independent code. 14916 14917@end table 14918 14919@node MCore Options 14920@subsection MCore Options 14921@cindex MCore options 14922 14923These are the @samp{-m} options defined for the Motorola M*Core 14924processors. 14925 14926@table @gcctabopt 14927 14928@item -mhardlit 14929@itemx -mno-hardlit 14930@opindex mhardlit 14931@opindex mno-hardlit 14932Inline constants into the code stream if it can be done in two 14933instructions or less. 14934 14935@item -mdiv 14936@itemx -mno-div 14937@opindex mdiv 14938@opindex mno-div 14939Use the divide instruction. (Enabled by default). 14940 14941@item -mrelax-immediate 14942@itemx -mno-relax-immediate 14943@opindex mrelax-immediate 14944@opindex mno-relax-immediate 14945Allow arbitrary sized immediates in bit operations. 14946 14947@item -mwide-bitfields 14948@itemx -mno-wide-bitfields 14949@opindex mwide-bitfields 14950@opindex mno-wide-bitfields 14951Always treat bit-fields as int-sized. 14952 14953@item -m4byte-functions 14954@itemx -mno-4byte-functions 14955@opindex m4byte-functions 14956@opindex mno-4byte-functions 14957Force all functions to be aligned to a 4-byte boundary. 14958 14959@item -mcallgraph-data 14960@itemx -mno-callgraph-data 14961@opindex mcallgraph-data 14962@opindex mno-callgraph-data 14963Emit callgraph information. 14964 14965@item -mslow-bytes 14966@itemx -mno-slow-bytes 14967@opindex mslow-bytes 14968@opindex mno-slow-bytes 14969Prefer word access when reading byte quantities. 14970 14971@item -mlittle-endian 14972@itemx -mbig-endian 14973@opindex mlittle-endian 14974@opindex mbig-endian 14975Generate code for a little-endian target. 14976 14977@item -m210 14978@itemx -m340 14979@opindex m210 14980@opindex m340 14981Generate code for the 210 processor. 14982 14983@item -mno-lsim 14984@opindex mno-lsim 14985Assume that runtime support has been provided and so omit the 14986simulator library (@file{libsim.a)} from the linker command line. 14987 14988@item -mstack-increment=@var{size} 14989@opindex mstack-increment 14990Set the maximum amount for a single stack increment operation. Large 14991values can increase the speed of programs that contain functions 14992that need a large amount of stack space, but they can also trigger a 14993segmentation fault if the stack is extended too much. The default 14994value is 0x1000. 14995 14996@end table 14997 14998@node MeP Options 14999@subsection MeP Options 15000@cindex MeP options 15001 15002@table @gcctabopt 15003 15004@item -mabsdiff 15005@opindex mabsdiff 15006Enables the @code{abs} instruction, which is the absolute difference 15007between two registers. 15008 15009@item -mall-opts 15010@opindex mall-opts 15011Enables all the optional instructions - average, multiply, divide, bit 15012operations, leading zero, absolute difference, min/max, clip, and 15013saturation. 15014 15015 15016@item -maverage 15017@opindex maverage 15018Enables the @code{ave} instruction, which computes the average of two 15019registers. 15020 15021@item -mbased=@var{n} 15022@opindex mbased= 15023Variables of size @var{n} bytes or smaller will be placed in the 15024@code{.based} section by default. Based variables use the @code{$tp} 15025register as a base register, and there is a 128-byte limit to the 15026@code{.based} section. 15027 15028@item -mbitops 15029@opindex mbitops 15030Enables the bit operation instructions - bit test (@code{btstm}), set 15031(@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and 15032test-and-set (@code{tas}). 15033 15034@item -mc=@var{name} 15035@opindex mc= 15036Selects which section constant data will be placed in. @var{name} may 15037be @code{tiny}, @code{near}, or @code{far}. 15038 15039@item -mclip 15040@opindex mclip 15041Enables the @code{clip} instruction. Note that @code{-mclip} is not 15042useful unless you also provide @code{-mminmax}. 15043 15044@item -mconfig=@var{name} 15045@opindex mconfig= 15046Selects one of the build-in core configurations. Each MeP chip has 15047one or more modules in it; each module has a core CPU and a variety of 15048coprocessors, optional instructions, and peripherals. The 15049@code{MeP-Integrator} tool, not part of GCC, provides these 15050configurations through this option; using this option is the same as 15051using all the corresponding command-line options. The default 15052configuration is @code{default}. 15053 15054@item -mcop 15055@opindex mcop 15056Enables the coprocessor instructions. By default, this is a 32-bit 15057coprocessor. Note that the coprocessor is normally enabled via the 15058@code{-mconfig=} option. 15059 15060@item -mcop32 15061@opindex mcop32 15062Enables the 32-bit coprocessor's instructions. 15063 15064@item -mcop64 15065@opindex mcop64 15066Enables the 64-bit coprocessor's instructions. 15067 15068@item -mivc2 15069@opindex mivc2 15070Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor. 15071 15072@item -mdc 15073@opindex mdc 15074Causes constant variables to be placed in the @code{.near} section. 15075 15076@item -mdiv 15077@opindex mdiv 15078Enables the @code{div} and @code{divu} instructions. 15079 15080@item -meb 15081@opindex meb 15082Generate big-endian code. 15083 15084@item -mel 15085@opindex mel 15086Generate little-endian code. 15087 15088@item -mio-volatile 15089@opindex mio-volatile 15090Tells the compiler that any variable marked with the @code{io} 15091attribute is to be considered volatile. 15092 15093@item -ml 15094@opindex ml 15095Causes variables to be assigned to the @code{.far} section by default. 15096 15097@item -mleadz 15098@opindex mleadz 15099Enables the @code{leadz} (leading zero) instruction. 15100 15101@item -mm 15102@opindex mm 15103Causes variables to be assigned to the @code{.near} section by default. 15104 15105@item -mminmax 15106@opindex mminmax 15107Enables the @code{min} and @code{max} instructions. 15108 15109@item -mmult 15110@opindex mmult 15111Enables the multiplication and multiply-accumulate instructions. 15112 15113@item -mno-opts 15114@opindex mno-opts 15115Disables all the optional instructions enabled by @code{-mall-opts}. 15116 15117@item -mrepeat 15118@opindex mrepeat 15119Enables the @code{repeat} and @code{erepeat} instructions, used for 15120low-overhead looping. 15121 15122@item -ms 15123@opindex ms 15124Causes all variables to default to the @code{.tiny} section. Note 15125that there is a 65536-byte limit to this section. Accesses to these 15126variables use the @code{%gp} base register. 15127 15128@item -msatur 15129@opindex msatur 15130Enables the saturation instructions. Note that the compiler does not 15131currently generate these itself, but this option is included for 15132compatibility with other tools, like @code{as}. 15133 15134@item -msdram 15135@opindex msdram 15136Link the SDRAM-based runtime instead of the default ROM-based runtime. 15137 15138@item -msim 15139@opindex msim 15140Link the simulator runtime libraries. 15141 15142@item -msimnovec 15143@opindex msimnovec 15144Link the simulator runtime libraries, excluding built-in support 15145for reset and exception vectors and tables. 15146 15147@item -mtf 15148@opindex mtf 15149Causes all functions to default to the @code{.far} section. Without 15150this option, functions default to the @code{.near} section. 15151 15152@item -mtiny=@var{n} 15153@opindex mtiny= 15154Variables that are @var{n} bytes or smaller will be allocated to the 15155@code{.tiny} section. These variables use the @code{$gp} base 15156register. The default for this option is 4, but note that there's a 1515765536-byte limit to the @code{.tiny} section. 15158 15159@end table 15160 15161@node MicroBlaze Options 15162@subsection MicroBlaze Options 15163@cindex MicroBlaze Options 15164 15165@table @gcctabopt 15166 15167@item -msoft-float 15168@opindex msoft-float 15169Use software emulation for floating point (default). 15170 15171@item -mhard-float 15172@opindex mhard-float 15173Use hardware floating-point instructions. 15174 15175@item -mmemcpy 15176@opindex mmemcpy 15177Do not optimize block moves, use @code{memcpy}. 15178 15179@item -mno-clearbss 15180@opindex mno-clearbss 15181This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead. 15182 15183@item -mcpu=@var{cpu-type} 15184@opindex mcpu= 15185Use features of and schedule code for given CPU. 15186Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}}, 15187where @var{X} is a major version, @var{YY} is the minor version, and 15188@var{Z} is compatibility code. Example values are @samp{v3.00.a}, 15189@samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}. 15190 15191@item -mxl-soft-mul 15192@opindex mxl-soft-mul 15193Use software multiply emulation (default). 15194 15195@item -mxl-soft-div 15196@opindex mxl-soft-div 15197Use software emulation for divides (default). 15198 15199@item -mxl-barrel-shift 15200@opindex mxl-barrel-shift 15201Use the hardware barrel shifter. 15202 15203@item -mxl-pattern-compare 15204@opindex mxl-pattern-compare 15205Use pattern compare instructions. 15206 15207@item -msmall-divides 15208@opindex msmall-divides 15209Use table lookup optimization for small signed integer divisions. 15210 15211@item -mxl-stack-check 15212@opindex mxl-stack-check 15213This option is deprecated. Use -fstack-check instead. 15214 15215@item -mxl-gp-opt 15216@opindex mxl-gp-opt 15217Use GP relative sdata/sbss sections. 15218 15219@item -mxl-multiply-high 15220@opindex mxl-multiply-high 15221Use multiply high instructions for high part of 32x32 multiply. 15222 15223@item -mxl-float-convert 15224@opindex mxl-float-convert 15225Use hardware floating-point conversion instructions. 15226 15227@item -mxl-float-sqrt 15228@opindex mxl-float-sqrt 15229Use hardware floating-point square root instruction. 15230 15231@item -mxl-mode-@var{app-model} 15232Select application model @var{app-model}. Valid models are 15233@table @samp 15234@item executable 15235normal executable (default), uses startup code @file{crt0.o}. 15236 15237@item xmdstub 15238for use with Xilinx Microprocessor Debugger (XMD) based 15239software intrusive debug agent called xmdstub. This uses startup file 15240@file{crt1.o} and sets the start address of the program to be 0x800. 15241 15242@item bootstrap 15243for applications that are loaded using a bootloader. 15244This model uses startup file @file{crt2.o} which does not contain a processor 15245reset vector handler. This is suitable for transferring control on a 15246processor reset to the bootloader rather than the application. 15247 15248@item novectors 15249for applications that do not require any of the 15250MicroBlaze vectors. This option may be useful for applications running 15251within a monitoring application. This model uses @file{crt3.o} as a startup file. 15252@end table 15253 15254Option @option{-xl-mode-@var{app-model}} is a deprecated alias for 15255@option{-mxl-mode-@var{app-model}}. 15256 15257@end table 15258 15259@node MIPS Options 15260@subsection MIPS Options 15261@cindex MIPS options 15262 15263@table @gcctabopt 15264 15265@item -EB 15266@opindex EB 15267Generate big-endian code. 15268 15269@item -EL 15270@opindex EL 15271Generate little-endian code. This is the default for @samp{mips*el-*-*} 15272configurations. 15273 15274@item -march=@var{arch} 15275@opindex march 15276Generate code that will run on @var{arch}, which can be the name of a 15277generic MIPS ISA, or the name of a particular processor. 15278The ISA names are: 15279@samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4}, 15280@samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}. 15281The processor names are: 15282@samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc}, 15283@samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd}, 15284@samp{5kc}, @samp{5kf}, 15285@samp{20kc}, 15286@samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1}, 15287@samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1}, 15288@samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, 15289@samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2}, 15290@samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1}, 15291@samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a}, 15292@samp{m4k}, 15293@samp{octeon}, @samp{octeon+}, @samp{octeon2}, 15294@samp{orion}, 15295@samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400}, 15296@samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, 15297@samp{rm7000}, @samp{rm9000}, 15298@samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000}, 15299@samp{sb1}, 15300@samp{sr71000}, 15301@samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300}, 15302@samp{vr5000}, @samp{vr5400}, @samp{vr5500} 15303and @samp{xlr}. 15304The special value @samp{from-abi} selects the 15305most compatible architecture for the selected ABI (that is, 15306@samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@. 15307 15308Native Linux/GNU and IRIX toolchains also support the value @samp{native}, 15309which selects the best architecture option for the host processor. 15310@option{-march=native} has no effect if GCC does not recognize 15311the processor. 15312 15313In processor names, a final @samp{000} can be abbreviated as @samp{k} 15314(for example, @samp{-march=r2k}). Prefixes are optional, and 15315@samp{vr} may be written @samp{r}. 15316 15317Names of the form @samp{@var{n}f2_1} refer to processors with 15318FPUs clocked at half the rate of the core, names of the form 15319@samp{@var{n}f1_1} refer to processors with FPUs clocked at the same 15320rate as the core, and names of the form @samp{@var{n}f3_2} refer to 15321processors with FPUs clocked a ratio of 3:2 with respect to the core. 15322For compatibility reasons, @samp{@var{n}f} is accepted as a synonym 15323for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are 15324accepted as synonyms for @samp{@var{n}f1_1}. 15325 15326GCC defines two macros based on the value of this option. The first 15327is @samp{_MIPS_ARCH}, which gives the name of target architecture, as 15328a string. The second has the form @samp{_MIPS_ARCH_@var{foo}}, 15329where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@. 15330For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH} 15331to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}. 15332 15333Note that the @samp{_MIPS_ARCH} macro uses the processor names given 15334above. In other words, it will have the full prefix and will not 15335abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi}, 15336the macro names the resolved architecture (either @samp{"mips1"} or 15337@samp{"mips3"}). It names the default architecture when no 15338@option{-march} option is given. 15339 15340@item -mtune=@var{arch} 15341@opindex mtune 15342Optimize for @var{arch}. Among other things, this option controls 15343the way instructions are scheduled, and the perceived cost of arithmetic 15344operations. The list of @var{arch} values is the same as for 15345@option{-march}. 15346 15347When this option is not used, GCC will optimize for the processor 15348specified by @option{-march}. By using @option{-march} and 15349@option{-mtune} together, it is possible to generate code that will 15350run on a family of processors, but optimize the code for one 15351particular member of that family. 15352 15353@samp{-mtune} defines the macros @samp{_MIPS_TUNE} and 15354@samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the 15355@samp{-march} ones described above. 15356 15357@item -mips1 15358@opindex mips1 15359Equivalent to @samp{-march=mips1}. 15360 15361@item -mips2 15362@opindex mips2 15363Equivalent to @samp{-march=mips2}. 15364 15365@item -mips3 15366@opindex mips3 15367Equivalent to @samp{-march=mips3}. 15368 15369@item -mips4 15370@opindex mips4 15371Equivalent to @samp{-march=mips4}. 15372 15373@item -mips32 15374@opindex mips32 15375Equivalent to @samp{-march=mips32}. 15376 15377@item -mips32r2 15378@opindex mips32r2 15379Equivalent to @samp{-march=mips32r2}. 15380 15381@item -mips64 15382@opindex mips64 15383Equivalent to @samp{-march=mips64}. 15384 15385@item -mips64r2 15386@opindex mips64r2 15387Equivalent to @samp{-march=mips64r2}. 15388 15389@item -mips16 15390@itemx -mno-mips16 15391@opindex mips16 15392@opindex mno-mips16 15393Generate (do not generate) MIPS16 code. If GCC is targetting a 15394MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@. 15395 15396MIPS16 code generation can also be controlled on a per-function basis 15397by means of @code{mips16} and @code{nomips16} attributes. 15398@xref{Function Attributes}, for more information. 15399 15400@item -mflip-mips16 15401@opindex mflip-mips16 15402Generate MIPS16 code on alternating functions. This option is provided 15403for regression testing of mixed MIPS16/non-MIPS16 code generation, and is 15404not intended for ordinary use in compiling user code. 15405 15406@item -minterlink-mips16 15407@itemx -mno-interlink-mips16 15408@opindex minterlink-mips16 15409@opindex mno-interlink-mips16 15410Require (do not require) that non-MIPS16 code be link-compatible with 15411MIPS16 code. 15412 15413For example, non-MIPS16 code cannot jump directly to MIPS16 code; 15414it must either use a call or an indirect jump. @option{-minterlink-mips16} 15415therefore disables direct jumps unless GCC knows that the target of the 15416jump is not MIPS16. 15417 15418@item -mabi=32 15419@itemx -mabi=o64 15420@itemx -mabi=n32 15421@itemx -mabi=64 15422@itemx -mabi=eabi 15423@opindex mabi=32 15424@opindex mabi=o64 15425@opindex mabi=n32 15426@opindex mabi=64 15427@opindex mabi=eabi 15428Generate code for the given ABI@. 15429 15430Note that the EABI has a 32-bit and a 64-bit variant. GCC normally 15431generates 64-bit code when you select a 64-bit architecture, but you 15432can use @option{-mgp32} to get 32-bit code instead. 15433 15434For information about the O64 ABI, see 15435@uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}. 15436 15437GCC supports a variant of the o32 ABI in which floating-point registers 15438are 64 rather than 32 bits wide. You can select this combination with 15439@option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1} 15440and @samp{mfhc1} instructions and is therefore only supported for 15441MIPS32R2 processors. 15442 15443The register assignments for arguments and return values remain the 15444same, but each scalar value is passed in a single 64-bit register 15445rather than a pair of 32-bit registers. For example, scalar 15446floating-point values are returned in @samp{$f0} only, not a 15447@samp{$f0}/@samp{$f1} pair. The set of call-saved registers also 15448remains the same, but all 64 bits are saved. 15449 15450@item -mabicalls 15451@itemx -mno-abicalls 15452@opindex mabicalls 15453@opindex mno-abicalls 15454Generate (do not generate) code that is suitable for SVR4-style 15455dynamic objects. @option{-mabicalls} is the default for SVR4-based 15456systems. 15457 15458@item -mshared 15459@itemx -mno-shared 15460Generate (do not generate) code that is fully position-independent, 15461and that can therefore be linked into shared libraries. This option 15462only affects @option{-mabicalls}. 15463 15464All @option{-mabicalls} code has traditionally been position-independent, 15465regardless of options like @option{-fPIC} and @option{-fpic}. However, 15466as an extension, the GNU toolchain allows executables to use absolute 15467accesses for locally-binding symbols. It can also use shorter GP 15468initialization sequences and generate direct calls to locally-defined 15469functions. This mode is selected by @option{-mno-shared}. 15470 15471@option{-mno-shared} depends on binutils 2.16 or higher and generates 15472objects that can only be linked by the GNU linker. However, the option 15473does not affect the ABI of the final executable; it only affects the ABI 15474of relocatable objects. Using @option{-mno-shared} will generally make 15475executables both smaller and quicker. 15476 15477@option{-mshared} is the default. 15478 15479@item -mplt 15480@itemx -mno-plt 15481@opindex mplt 15482@opindex mno-plt 15483Assume (do not assume) that the static and dynamic linkers 15484support PLTs and copy relocations. This option only affects 15485@samp{-mno-shared -mabicalls}. For the n64 ABI, this option 15486has no effect without @samp{-msym32}. 15487 15488You can make @option{-mplt} the default by configuring 15489GCC with @option{--with-mips-plt}. The default is 15490@option{-mno-plt} otherwise. 15491 15492@item -mxgot 15493@itemx -mno-xgot 15494@opindex mxgot 15495@opindex mno-xgot 15496Lift (do not lift) the usual restrictions on the size of the global 15497offset table. 15498 15499GCC normally uses a single instruction to load values from the GOT@. 15500While this is relatively efficient, it will only work if the GOT 15501is smaller than about 64k. Anything larger will cause the linker 15502to report an error such as: 15503 15504@cindex relocation truncated to fit (MIPS) 15505@smallexample 15506relocation truncated to fit: R_MIPS_GOT16 foobar 15507@end smallexample 15508 15509If this happens, you should recompile your code with @option{-mxgot}. 15510It should then work with very large GOTs, although it will also be 15511less efficient, since it will take three instructions to fetch the 15512value of a global symbol. 15513 15514Note that some linkers can create multiple GOTs. If you have such a 15515linker, you should only need to use @option{-mxgot} when a single object 15516file accesses more than 64k's worth of GOT entries. Very few do. 15517 15518These options have no effect unless GCC is generating position 15519independent code. 15520 15521@item -mgp32 15522@opindex mgp32 15523Assume that general-purpose registers are 32 bits wide. 15524 15525@item -mgp64 15526@opindex mgp64 15527Assume that general-purpose registers are 64 bits wide. 15528 15529@item -mfp32 15530@opindex mfp32 15531Assume that floating-point registers are 32 bits wide. 15532 15533@item -mfp64 15534@opindex mfp64 15535Assume that floating-point registers are 64 bits wide. 15536 15537@item -mhard-float 15538@opindex mhard-float 15539Use floating-point coprocessor instructions. 15540 15541@item -msoft-float 15542@opindex msoft-float 15543Do not use floating-point coprocessor instructions. Implement 15544floating-point calculations using library calls instead. 15545 15546@item -msingle-float 15547@opindex msingle-float 15548Assume that the floating-point coprocessor only supports single-precision 15549operations. 15550 15551@item -mdouble-float 15552@opindex mdouble-float 15553Assume that the floating-point coprocessor supports double-precision 15554operations. This is the default. 15555 15556@item -mllsc 15557@itemx -mno-llsc 15558@opindex mllsc 15559@opindex mno-llsc 15560Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to 15561implement atomic memory built-in functions. When neither option is 15562specified, GCC will use the instructions if the target architecture 15563supports them. 15564 15565@option{-mllsc} is useful if the runtime environment can emulate the 15566instructions and @option{-mno-llsc} can be useful when compiling for 15567nonstandard ISAs. You can make either option the default by 15568configuring GCC with @option{--with-llsc} and @option{--without-llsc} 15569respectively. @option{--with-llsc} is the default for some 15570configurations; see the installation documentation for details. 15571 15572@item -mdsp 15573@itemx -mno-dsp 15574@opindex mdsp 15575@opindex mno-dsp 15576Use (do not use) revision 1 of the MIPS DSP ASE@. 15577@xref{MIPS DSP Built-in Functions}. This option defines the 15578preprocessor macro @samp{__mips_dsp}. It also defines 15579@samp{__mips_dsp_rev} to 1. 15580 15581@item -mdspr2 15582@itemx -mno-dspr2 15583@opindex mdspr2 15584@opindex mno-dspr2 15585Use (do not use) revision 2 of the MIPS DSP ASE@. 15586@xref{MIPS DSP Built-in Functions}. This option defines the 15587preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}. 15588It also defines @samp{__mips_dsp_rev} to 2. 15589 15590@item -msmartmips 15591@itemx -mno-smartmips 15592@opindex msmartmips 15593@opindex mno-smartmips 15594Use (do not use) the MIPS SmartMIPS ASE. 15595 15596@item -mpaired-single 15597@itemx -mno-paired-single 15598@opindex mpaired-single 15599@opindex mno-paired-single 15600Use (do not use) paired-single floating-point instructions. 15601@xref{MIPS Paired-Single Support}. This option requires 15602hardware floating-point support to be enabled. 15603 15604@item -mdmx 15605@itemx -mno-mdmx 15606@opindex mdmx 15607@opindex mno-mdmx 15608Use (do not use) MIPS Digital Media Extension instructions. 15609This option can only be used when generating 64-bit code and requires 15610hardware floating-point support to be enabled. 15611 15612@item -mips3d 15613@itemx -mno-mips3d 15614@opindex mips3d 15615@opindex mno-mips3d 15616Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}. 15617The option @option{-mips3d} implies @option{-mpaired-single}. 15618 15619@item -mmt 15620@itemx -mno-mt 15621@opindex mmt 15622@opindex mno-mt 15623Use (do not use) MT Multithreading instructions. 15624 15625@item -mlong64 15626@opindex mlong64 15627Force @code{long} types to be 64 bits wide. See @option{-mlong32} for 15628an explanation of the default and the way that the pointer size is 15629determined. 15630 15631@item -mlong32 15632@opindex mlong32 15633Force @code{long}, @code{int}, and pointer types to be 32 bits wide. 15634 15635The default size of @code{int}s, @code{long}s and pointers depends on 15636the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI 15637uses 64-bit @code{long}s, as does the 64-bit EABI; the others use 1563832-bit @code{long}s. Pointers are the same size as @code{long}s, 15639or the same size as integer registers, whichever is smaller. 15640 15641@item -msym32 15642@itemx -mno-sym32 15643@opindex msym32 15644@opindex mno-sym32 15645Assume (do not assume) that all symbols have 32-bit values, regardless 15646of the selected ABI@. This option is useful in combination with 15647@option{-mabi=64} and @option{-mno-abicalls} because it allows GCC 15648to generate shorter and faster references to symbolic addresses. 15649 15650@item -G @var{num} 15651@opindex G 15652Put definitions of externally-visible data in a small data section 15653if that data is no bigger than @var{num} bytes. GCC can then access 15654the data more efficiently; see @option{-mgpopt} for details. 15655 15656The default @option{-G} option depends on the configuration. 15657 15658@item -mlocal-sdata 15659@itemx -mno-local-sdata 15660@opindex mlocal-sdata 15661@opindex mno-local-sdata 15662Extend (do not extend) the @option{-G} behavior to local data too, 15663such as to static variables in C@. @option{-mlocal-sdata} is the 15664default for all configurations. 15665 15666If the linker complains that an application is using too much small data, 15667you might want to try rebuilding the less performance-critical parts with 15668@option{-mno-local-sdata}. You might also want to build large 15669libraries with @option{-mno-local-sdata}, so that the libraries leave 15670more room for the main program. 15671 15672@item -mextern-sdata 15673@itemx -mno-extern-sdata 15674@opindex mextern-sdata 15675@opindex mno-extern-sdata 15676Assume (do not assume) that externally-defined data will be in 15677a small data section if that data is within the @option{-G} limit. 15678@option{-mextern-sdata} is the default for all configurations. 15679 15680If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G 15681@var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var} 15682that is no bigger than @var{num} bytes, you must make sure that @var{Var} 15683is placed in a small data section. If @var{Var} is defined by another 15684module, you must either compile that module with a high-enough 15685@option{-G} setting or attach a @code{section} attribute to @var{Var}'s 15686definition. If @var{Var} is common, you must link the application 15687with a high-enough @option{-G} setting. 15688 15689The easiest way of satisfying these restrictions is to compile 15690and link every module with the same @option{-G} option. However, 15691you may wish to build a library that supports several different 15692small data limits. You can do this by compiling the library with 15693the highest supported @option{-G} setting and additionally using 15694@option{-mno-extern-sdata} to stop the library from making assumptions 15695about externally-defined data. 15696 15697@item -mgpopt 15698@itemx -mno-gpopt 15699@opindex mgpopt 15700@opindex mno-gpopt 15701Use (do not use) GP-relative accesses for symbols that are known to be 15702in a small data section; see @option{-G}, @option{-mlocal-sdata} and 15703@option{-mextern-sdata}. @option{-mgpopt} is the default for all 15704configurations. 15705 15706@option{-mno-gpopt} is useful for cases where the @code{$gp} register 15707might not hold the value of @code{_gp}. For example, if the code is 15708part of a library that might be used in a boot monitor, programs that 15709call boot monitor routines will pass an unknown value in @code{$gp}. 15710(In such situations, the boot monitor itself would usually be compiled 15711with @option{-G0}.) 15712 15713@option{-mno-gpopt} implies @option{-mno-local-sdata} and 15714@option{-mno-extern-sdata}. 15715 15716@item -membedded-data 15717@itemx -mno-embedded-data 15718@opindex membedded-data 15719@opindex mno-embedded-data 15720Allocate variables to the read-only data section first if possible, then 15721next in the small data section if possible, otherwise in data. This gives 15722slightly slower code than the default, but reduces the amount of RAM required 15723when executing, and thus may be preferred for some embedded systems. 15724 15725@item -muninit-const-in-rodata 15726@itemx -mno-uninit-const-in-rodata 15727@opindex muninit-const-in-rodata 15728@opindex mno-uninit-const-in-rodata 15729Put uninitialized @code{const} variables in the read-only data section. 15730This option is only meaningful in conjunction with @option{-membedded-data}. 15731 15732@item -mcode-readable=@var{setting} 15733@opindex mcode-readable 15734Specify whether GCC may generate code that reads from executable sections. 15735There are three possible settings: 15736 15737@table @gcctabopt 15738@item -mcode-readable=yes 15739Instructions may freely access executable sections. This is the 15740default setting. 15741 15742@item -mcode-readable=pcrel 15743MIPS16 PC-relative load instructions can access executable sections, 15744but other instructions must not do so. This option is useful on 4KSc 15745and 4KSd processors when the code TLBs have the Read Inhibit bit set. 15746It is also useful on processors that can be configured to have a dual 15747instruction/data SRAM interface and that, like the M4K, automatically 15748redirect PC-relative loads to the instruction RAM. 15749 15750@item -mcode-readable=no 15751Instructions must not access executable sections. This option can be 15752useful on targets that are configured to have a dual instruction/data 15753SRAM interface but that (unlike the M4K) do not automatically redirect 15754PC-relative loads to the instruction RAM. 15755@end table 15756 15757@item -msplit-addresses 15758@itemx -mno-split-addresses 15759@opindex msplit-addresses 15760@opindex mno-split-addresses 15761Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler 15762relocation operators. This option has been superseded by 15763@option{-mexplicit-relocs} but is retained for backwards compatibility. 15764 15765@item -mexplicit-relocs 15766@itemx -mno-explicit-relocs 15767@opindex mexplicit-relocs 15768@opindex mno-explicit-relocs 15769Use (do not use) assembler relocation operators when dealing with symbolic 15770addresses. The alternative, selected by @option{-mno-explicit-relocs}, 15771is to use assembler macros instead. 15772 15773@option{-mexplicit-relocs} is the default if GCC was configured 15774to use an assembler that supports relocation operators. 15775 15776@item -mcheck-zero-division 15777@itemx -mno-check-zero-division 15778@opindex mcheck-zero-division 15779@opindex mno-check-zero-division 15780Trap (do not trap) on integer division by zero. 15781 15782The default is @option{-mcheck-zero-division}. 15783 15784@item -mdivide-traps 15785@itemx -mdivide-breaks 15786@opindex mdivide-traps 15787@opindex mdivide-breaks 15788MIPS systems check for division by zero by generating either a 15789conditional trap or a break instruction. Using traps results in 15790smaller code, but is only supported on MIPS II and later. Also, some 15791versions of the Linux kernel have a bug that prevents trap from 15792generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to 15793allow conditional traps on architectures that support them and 15794@option{-mdivide-breaks} to force the use of breaks. 15795 15796The default is usually @option{-mdivide-traps}, but this can be 15797overridden at configure time using @option{--with-divide=breaks}. 15798Divide-by-zero checks can be completely disabled using 15799@option{-mno-check-zero-division}. 15800 15801@item -mmemcpy 15802@itemx -mno-memcpy 15803@opindex mmemcpy 15804@opindex mno-memcpy 15805Force (do not force) the use of @code{memcpy()} for non-trivial block 15806moves. The default is @option{-mno-memcpy}, which allows GCC to inline 15807most constant-sized copies. 15808 15809@item -mlong-calls 15810@itemx -mno-long-calls 15811@opindex mlong-calls 15812@opindex mno-long-calls 15813Disable (do not disable) use of the @code{jal} instruction. Calling 15814functions using @code{jal} is more efficient but requires the caller 15815and callee to be in the same 256 megabyte segment. 15816 15817This option has no effect on abicalls code. The default is 15818@option{-mno-long-calls}. 15819 15820@item -mmad 15821@itemx -mno-mad 15822@opindex mmad 15823@opindex mno-mad 15824Enable (disable) use of the @code{mad}, @code{madu} and @code{mul} 15825instructions, as provided by the R4650 ISA@. 15826 15827@item -mfused-madd 15828@itemx -mno-fused-madd 15829@opindex mfused-madd 15830@opindex mno-fused-madd 15831Enable (disable) use of the floating-point multiply-accumulate 15832instructions, when they are available. The default is 15833@option{-mfused-madd}. 15834 15835When multiply-accumulate instructions are used, the intermediate 15836product is calculated to infinite precision and is not subject to 15837the FCSR Flush to Zero bit. This may be undesirable in some 15838circumstances. 15839 15840@item -nocpp 15841@opindex nocpp 15842Tell the MIPS assembler to not run its preprocessor over user 15843assembler files (with a @samp{.s} suffix) when assembling them. 15844 15845@item -mfix-24k 15846@item -mno-fix-24k 15847@opindex mfix-24k 15848@opindex mno-fix-24k 15849Work around the 24K E48 (lost data on stores during refill) errata. 15850The workarounds are implemented by the assembler rather than by GCC. 15851 15852@item -mfix-r4000 15853@itemx -mno-fix-r4000 15854@opindex mfix-r4000 15855@opindex mno-fix-r4000 15856Work around certain R4000 CPU errata: 15857@itemize @minus 15858@item 15859A double-word or a variable shift may give an incorrect result if executed 15860immediately after starting an integer division. 15861@item 15862A double-word or a variable shift may give an incorrect result if executed 15863while an integer multiplication is in progress. 15864@item 15865An integer division may give an incorrect result if started in a delay slot 15866of a taken branch or a jump. 15867@end itemize 15868 15869@item -mfix-r4400 15870@itemx -mno-fix-r4400 15871@opindex mfix-r4400 15872@opindex mno-fix-r4400 15873Work around certain R4400 CPU errata: 15874@itemize @minus 15875@item 15876A double-word or a variable shift may give an incorrect result if executed 15877immediately after starting an integer division. 15878@end itemize 15879 15880@item -mfix-r10000 15881@itemx -mno-fix-r10000 15882@opindex mfix-r10000 15883@opindex mno-fix-r10000 15884Work around certain R10000 errata: 15885@itemize @minus 15886@item 15887@code{ll}/@code{sc} sequences may not behave atomically on revisions 15888prior to 3.0. They may deadlock on revisions 2.6 and earlier. 15889@end itemize 15890 15891This option can only be used if the target architecture supports 15892branch-likely instructions. @option{-mfix-r10000} is the default when 15893@option{-march=r10000} is used; @option{-mno-fix-r10000} is the default 15894otherwise. 15895 15896@item -mfix-vr4120 15897@itemx -mno-fix-vr4120 15898@opindex mfix-vr4120 15899Work around certain VR4120 errata: 15900@itemize @minus 15901@item 15902@code{dmultu} does not always produce the correct result. 15903@item 15904@code{div} and @code{ddiv} do not always produce the correct result if one 15905of the operands is negative. 15906@end itemize 15907The workarounds for the division errata rely on special functions in 15908@file{libgcc.a}. At present, these functions are only provided by 15909the @code{mips64vr*-elf} configurations. 15910 15911Other VR4120 errata require a nop to be inserted between certain pairs of 15912instructions. These errata are handled by the assembler, not by GCC itself. 15913 15914@item -mfix-vr4130 15915@opindex mfix-vr4130 15916Work around the VR4130 @code{mflo}/@code{mfhi} errata. The 15917workarounds are implemented by the assembler rather than by GCC, 15918although GCC will avoid using @code{mflo} and @code{mfhi} if the 15919VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi} 15920instructions are available instead. 15921 15922@item -mfix-sb1 15923@itemx -mno-fix-sb1 15924@opindex mfix-sb1 15925Work around certain SB-1 CPU core errata. 15926(This flag currently works around the SB-1 revision 2 15927``F1'' and ``F2'' floating-point errata.) 15928 15929@item -mr10k-cache-barrier=@var{setting} 15930@opindex mr10k-cache-barrier 15931Specify whether GCC should insert cache barriers to avoid the 15932side-effects of speculation on R10K processors. 15933 15934In common with many processors, the R10K tries to predict the outcome 15935of a conditional branch and speculatively executes instructions from 15936the ``taken'' branch. It later aborts these instructions if the 15937predicted outcome was wrong. However, on the R10K, even aborted 15938instructions can have side effects. 15939 15940This problem only affects kernel stores and, depending on the system, 15941kernel loads. As an example, a speculatively-executed store may load 15942the target memory into cache and mark the cache line as dirty, even if 15943the store itself is later aborted. If a DMA operation writes to the 15944same area of memory before the ``dirty'' line is flushed, the cached 15945data will overwrite the DMA-ed data. See the R10K processor manual 15946for a full description, including other potential problems. 15947 15948One workaround is to insert cache barrier instructions before every memory 15949access that might be speculatively executed and that might have side 15950effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}} 15951controls GCC's implementation of this workaround. It assumes that 15952aborted accesses to any byte in the following regions will not have 15953side effects: 15954 15955@enumerate 15956@item 15957the memory occupied by the current function's stack frame; 15958 15959@item 15960the memory occupied by an incoming stack argument; 15961 15962@item 15963the memory occupied by an object with a link-time-constant address. 15964@end enumerate 15965 15966It is the kernel's responsibility to ensure that speculative 15967accesses to these regions are indeed safe. 15968 15969If the input program contains a function declaration such as: 15970 15971@smallexample 15972void foo (void); 15973@end smallexample 15974 15975then the implementation of @code{foo} must allow @code{j foo} and 15976@code{jal foo} to be executed speculatively. GCC honors this 15977restriction for functions it compiles itself. It expects non-GCC 15978functions (such as hand-written assembly code) to do the same. 15979 15980The option has three forms: 15981 15982@table @gcctabopt 15983@item -mr10k-cache-barrier=load-store 15984Insert a cache barrier before a load or store that might be 15985speculatively executed and that might have side effects even 15986if aborted. 15987 15988@item -mr10k-cache-barrier=store 15989Insert a cache barrier before a store that might be speculatively 15990executed and that might have side effects even if aborted. 15991 15992@item -mr10k-cache-barrier=none 15993Disable the insertion of cache barriers. This is the default setting. 15994@end table 15995 15996@item -mflush-func=@var{func} 15997@itemx -mno-flush-func 15998@opindex mflush-func 15999Specifies the function to call to flush the I and D caches, or to not 16000call any such function. If called, the function must take the same 16001arguments as the common @code{_flush_func()}, that is, the address of the 16002memory range for which the cache is being flushed, the size of the 16003memory range, and the number 3 (to flush both caches). The default 16004depends on the target GCC was configured for, but commonly is either 16005@samp{_flush_func} or @samp{__cpu_flush}. 16006 16007@item mbranch-cost=@var{num} 16008@opindex mbranch-cost 16009Set the cost of branches to roughly @var{num} ``simple'' instructions. 16010This cost is only a heuristic and is not guaranteed to produce 16011consistent results across releases. A zero cost redundantly selects 16012the default, which is based on the @option{-mtune} setting. 16013 16014@item -mbranch-likely 16015@itemx -mno-branch-likely 16016@opindex mbranch-likely 16017@opindex mno-branch-likely 16018Enable or disable use of Branch Likely instructions, regardless of the 16019default for the selected architecture. By default, Branch Likely 16020instructions may be generated if they are supported by the selected 16021architecture. An exception is for the MIPS32 and MIPS64 architectures 16022and processors that implement those architectures; for those, Branch 16023Likely instructions will not be generated by default because the MIPS32 16024and MIPS64 architectures specifically deprecate their use. 16025 16026@item -mfp-exceptions 16027@itemx -mno-fp-exceptions 16028@opindex mfp-exceptions 16029Specifies whether FP exceptions are enabled. This affects how we schedule 16030FP instructions for some processors. The default is that FP exceptions are 16031enabled. 16032 16033For instance, on the SB-1, if FP exceptions are disabled, and we are emitting 1603464-bit code, then we can use both FP pipes. Otherwise, we can only use one 16035FP pipe. 16036 16037@item -mvr4130-align 16038@itemx -mno-vr4130-align 16039@opindex mvr4130-align 16040The VR4130 pipeline is two-way superscalar, but can only issue two 16041instructions together if the first one is 8-byte aligned. When this 16042option is enabled, GCC will align pairs of instructions that it 16043thinks should execute in parallel. 16044 16045This option only has an effect when optimizing for the VR4130. 16046It normally makes code faster, but at the expense of making it bigger. 16047It is enabled by default at optimization level @option{-O3}. 16048 16049@item -msynci 16050@itemx -mno-synci 16051@opindex msynci 16052Enable (disable) generation of @code{synci} instructions on 16053architectures that support it. The @code{synci} instructions (if 16054enabled) will be generated when @code{__builtin___clear_cache()} is 16055compiled. 16056 16057This option defaults to @code{-mno-synci}, but the default can be 16058overridden by configuring with @code{--with-synci}. 16059 16060When compiling code for single processor systems, it is generally safe 16061to use @code{synci}. However, on many multi-core (SMP) systems, it 16062will not invalidate the instruction caches on all cores and may lead 16063to undefined behavior. 16064 16065@item -mrelax-pic-calls 16066@itemx -mno-relax-pic-calls 16067@opindex mrelax-pic-calls 16068Try to turn PIC calls that are normally dispatched via register 16069@code{$25} into direct calls. This is only possible if the linker can 16070resolve the destination at link-time and if the destination is within 16071range for a direct call. 16072 16073@option{-mrelax-pic-calls} is the default if GCC was configured to use 16074an assembler and a linker that supports the @code{.reloc} assembly 16075directive and @code{-mexplicit-relocs} is in effect. With 16076@code{-mno-explicit-relocs}, this optimization can be performed by the 16077assembler and the linker alone without help from the compiler. 16078 16079@item -mmcount-ra-address 16080@itemx -mno-mcount-ra-address 16081@opindex mmcount-ra-address 16082@opindex mno-mcount-ra-address 16083Emit (do not emit) code that allows @code{_mcount} to modify the 16084calling function's return address. When enabled, this option extends 16085the usual @code{_mcount} interface with a new @var{ra-address} 16086parameter, which has type @code{intptr_t *} and is passed in register 16087@code{$12}. @code{_mcount} can then modify the return address by 16088doing both of the following: 16089@itemize 16090@item 16091Returning the new address in register @code{$31}. 16092@item 16093Storing the new address in @code{*@var{ra-address}}, 16094if @var{ra-address} is nonnull. 16095@end itemize 16096 16097The default is @option{-mno-mcount-ra-address}. 16098 16099@end table 16100 16101@node MMIX Options 16102@subsection MMIX Options 16103@cindex MMIX Options 16104 16105These options are defined for the MMIX: 16106 16107@table @gcctabopt 16108@item -mlibfuncs 16109@itemx -mno-libfuncs 16110@opindex mlibfuncs 16111@opindex mno-libfuncs 16112Specify that intrinsic library functions are being compiled, passing all 16113values in registers, no matter the size. 16114 16115@item -mepsilon 16116@itemx -mno-epsilon 16117@opindex mepsilon 16118@opindex mno-epsilon 16119Generate floating-point comparison instructions that compare with respect 16120to the @code{rE} epsilon register. 16121 16122@item -mabi=mmixware 16123@itemx -mabi=gnu 16124@opindex mabi=mmixware 16125@opindex mabi=gnu 16126Generate code that passes function parameters and return values that (in 16127the called function) are seen as registers @code{$0} and up, as opposed to 16128the GNU ABI which uses global registers @code{$231} and up. 16129 16130@item -mzero-extend 16131@itemx -mno-zero-extend 16132@opindex mzero-extend 16133@opindex mno-zero-extend 16134When reading data from memory in sizes shorter than 64 bits, use (do not 16135use) zero-extending load instructions by default, rather than 16136sign-extending ones. 16137 16138@item -mknuthdiv 16139@itemx -mno-knuthdiv 16140@opindex mknuthdiv 16141@opindex mno-knuthdiv 16142Make the result of a division yielding a remainder have the same sign as 16143the divisor. With the default, @option{-mno-knuthdiv}, the sign of the 16144remainder follows the sign of the dividend. Both methods are 16145arithmetically valid, the latter being almost exclusively used. 16146 16147@item -mtoplevel-symbols 16148@itemx -mno-toplevel-symbols 16149@opindex mtoplevel-symbols 16150@opindex mno-toplevel-symbols 16151Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly 16152code can be used with the @code{PREFIX} assembly directive. 16153 16154@item -melf 16155@opindex melf 16156Generate an executable in the ELF format, rather than the default 16157@samp{mmo} format used by the @command{mmix} simulator. 16158 16159@item -mbranch-predict 16160@itemx -mno-branch-predict 16161@opindex mbranch-predict 16162@opindex mno-branch-predict 16163Use (do not use) the probable-branch instructions, when static branch 16164prediction indicates a probable branch. 16165 16166@item -mbase-addresses 16167@itemx -mno-base-addresses 16168@opindex mbase-addresses 16169@opindex mno-base-addresses 16170Generate (do not generate) code that uses @emph{base addresses}. Using a 16171base address automatically generates a request (handled by the assembler 16172and the linker) for a constant to be set up in a global register. The 16173register is used for one or more base address requests within the range 0 16174to 255 from the value held in the register. The generally leads to short 16175and fast code, but the number of different data items that can be 16176addressed is limited. This means that a program that uses lots of static 16177data may require @option{-mno-base-addresses}. 16178 16179@item -msingle-exit 16180@itemx -mno-single-exit 16181@opindex msingle-exit 16182@opindex mno-single-exit 16183Force (do not force) generated code to have a single exit point in each 16184function. 16185@end table 16186 16187@node MN10300 Options 16188@subsection MN10300 Options 16189@cindex MN10300 options 16190 16191These @option{-m} options are defined for Matsushita MN10300 architectures: 16192 16193@table @gcctabopt 16194@item -mmult-bug 16195@opindex mmult-bug 16196Generate code to avoid bugs in the multiply instructions for the MN10300 16197processors. This is the default. 16198 16199@item -mno-mult-bug 16200@opindex mno-mult-bug 16201Do not generate code to avoid bugs in the multiply instructions for the 16202MN10300 processors. 16203 16204@item -mam33 16205@opindex mam33 16206Generate code using features specific to the AM33 processor. 16207 16208@item -mno-am33 16209@opindex mno-am33 16210Do not generate code using features specific to the AM33 processor. This 16211is the default. 16212 16213@item -mam33-2 16214@opindex mam33-2 16215Generate code using features specific to the AM33/2.0 processor. 16216 16217@item -mam34 16218@opindex mam34 16219Generate code using features specific to the AM34 processor. 16220 16221@item -mtune=@var{cpu-type} 16222@opindex mtune 16223Use the timing characteristics of the indicated CPU type when 16224scheduling instructions. This does not change the targeted processor 16225type. The CPU type must be one of @samp{mn10300}, @samp{am33}, 16226@samp{am33-2} or @samp{am34}. 16227 16228@item -mreturn-pointer-on-d0 16229@opindex mreturn-pointer-on-d0 16230When generating a function that returns a pointer, return the pointer 16231in both @code{a0} and @code{d0}. Otherwise, the pointer is returned 16232only in a0, and attempts to call such functions without a prototype 16233would result in errors. Note that this option is on by default; use 16234@option{-mno-return-pointer-on-d0} to disable it. 16235 16236@item -mno-crt0 16237@opindex mno-crt0 16238Do not link in the C run-time initialization object file. 16239 16240@item -mrelax 16241@opindex mrelax 16242Indicate to the linker that it should perform a relaxation optimization pass 16243to shorten branches, calls and absolute memory addresses. This option only 16244has an effect when used on the command line for the final link step. 16245 16246This option makes symbolic debugging impossible. 16247 16248@item -mliw 16249@opindex mliw 16250Allow the compiler to generate @emph{Long Instruction Word} 16251instructions if the target is the @samp{AM33} or later. This is the 16252default. This option defines the preprocessor macro @samp{__LIW__}. 16253 16254@item -mnoliw 16255@opindex mnoliw 16256Do not allow the compiler to generate @emph{Long Instruction Word} 16257instructions. This option defines the preprocessor macro 16258@samp{__NO_LIW__}. 16259 16260@item -msetlb 16261@opindex msetlb 16262Allow the compiler to generate the @emph{SETLB} and @emph{Lcc} 16263instructions if the target is the @samp{AM33} or later. This is the 16264default. This option defines the preprocessor macro @samp{__SETLB__}. 16265 16266@item -mnosetlb 16267@opindex mnosetlb 16268Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc} 16269instructions. This option defines the preprocessor macro 16270@samp{__NO_SETLB__}. 16271 16272@end table 16273 16274@node PDP-11 Options 16275@subsection PDP-11 Options 16276@cindex PDP-11 Options 16277 16278These options are defined for the PDP-11: 16279 16280@table @gcctabopt 16281@item -mfpu 16282@opindex mfpu 16283Use hardware FPP floating point. This is the default. (FIS floating 16284point on the PDP-11/40 is not supported.) 16285 16286@item -msoft-float 16287@opindex msoft-float 16288Do not use hardware floating point. 16289 16290@item -mac0 16291@opindex mac0 16292Return floating-point results in ac0 (fr0 in Unix assembler syntax). 16293 16294@item -mno-ac0 16295@opindex mno-ac0 16296Return floating-point results in memory. This is the default. 16297 16298@item -m40 16299@opindex m40 16300Generate code for a PDP-11/40. 16301 16302@item -m45 16303@opindex m45 16304Generate code for a PDP-11/45. This is the default. 16305 16306@item -m10 16307@opindex m10 16308Generate code for a PDP-11/10. 16309 16310@item -mbcopy-builtin 16311@opindex mbcopy-builtin 16312Use inline @code{movmemhi} patterns for copying memory. This is the 16313default. 16314 16315@item -mbcopy 16316@opindex mbcopy 16317Do not use inline @code{movmemhi} patterns for copying memory. 16318 16319@item -mint16 16320@itemx -mno-int32 16321@opindex mint16 16322@opindex mno-int32 16323Use 16-bit @code{int}. This is the default. 16324 16325@item -mint32 16326@itemx -mno-int16 16327@opindex mint32 16328@opindex mno-int16 16329Use 32-bit @code{int}. 16330 16331@item -mfloat64 16332@itemx -mno-float32 16333@opindex mfloat64 16334@opindex mno-float32 16335Use 64-bit @code{float}. This is the default. 16336 16337@item -mfloat32 16338@itemx -mno-float64 16339@opindex mfloat32 16340@opindex mno-float64 16341Use 32-bit @code{float}. 16342 16343@item -mabshi 16344@opindex mabshi 16345Use @code{abshi2} pattern. This is the default. 16346 16347@item -mno-abshi 16348@opindex mno-abshi 16349Do not use @code{abshi2} pattern. 16350 16351@item -mbranch-expensive 16352@opindex mbranch-expensive 16353Pretend that branches are expensive. This is for experimenting with 16354code generation only. 16355 16356@item -mbranch-cheap 16357@opindex mbranch-cheap 16358Do not pretend that branches are expensive. This is the default. 16359 16360@item -munix-asm 16361@opindex munix-asm 16362Use Unix assembler syntax. This is the default when configured for 16363@samp{pdp11-*-bsd}. 16364 16365@item -mdec-asm 16366@opindex mdec-asm 16367Use DEC assembler syntax. This is the default when configured for any 16368PDP-11 target other than @samp{pdp11-*-bsd}. 16369@end table 16370 16371@node picoChip Options 16372@subsection picoChip Options 16373@cindex picoChip options 16374 16375These @samp{-m} options are defined for picoChip implementations: 16376 16377@table @gcctabopt 16378 16379@item -mae=@var{ae_type} 16380@opindex mcpu 16381Set the instruction set, register set, and instruction scheduling 16382parameters for array element type @var{ae_type}. Supported values 16383for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}. 16384 16385@option{-mae=ANY} selects a completely generic AE type. Code 16386generated with this option will run on any of the other AE types. The 16387code will not be as efficient as it would be if compiled for a specific 16388AE type, and some types of operation (e.g., multiplication) will not 16389work properly on all types of AE. 16390 16391@option{-mae=MUL} selects a MUL AE type. This is the most useful AE type 16392for compiled code, and is the default. 16393 16394@option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this 16395option may suffer from poor performance of byte (char) manipulation, 16396since the DSP AE does not provide hardware support for byte load/stores. 16397 16398@item -msymbol-as-address 16399Enable the compiler to directly use a symbol name as an address in a 16400load/store instruction, without first loading it into a 16401register. Typically, the use of this option will generate larger 16402programs, which run faster than when the option isn't used. However, the 16403results vary from program to program, so it is left as a user option, 16404rather than being permanently enabled. 16405 16406@item -mno-inefficient-warnings 16407Disables warnings about the generation of inefficient code. These 16408warnings can be generated, for example, when compiling code that 16409performs byte-level memory operations on the MAC AE type. The MAC AE has 16410no hardware support for byte-level memory operations, so all byte 16411load/stores must be synthesized from word load/store operations. This is 16412inefficient and a warning will be generated indicating to the programmer 16413that they should rewrite the code to avoid byte operations, or to target 16414an AE type that has the necessary hardware support. This option enables 16415the warning to be turned off. 16416 16417@end table 16418 16419@node PowerPC Options 16420@subsection PowerPC Options 16421@cindex PowerPC options 16422 16423These are listed under @xref{RS/6000 and PowerPC Options}. 16424 16425@node RL78 Options 16426@subsection RL78 Options 16427@cindex RL78 Options 16428 16429@table @gcctabopt 16430 16431@item -msim 16432@opindex msim 16433Links in additional target libraries to support operation within a 16434simulator. 16435 16436@item -mmul=none 16437@itemx -mmul=g13 16438@itemx -mmul=rl78 16439@opindex mmul 16440Specifies the type of hardware multiplication support to be used. The 16441default is @code{none}, which uses software multiplication functions. 16442The @code{g13} option is for the hardware multiply/divide peripheral 16443only on the RL78/G13 targets. The @code{rl78} option is for the 16444standard hardware multiplication defined in the RL78 software manual. 16445 16446@end table 16447 16448@node RS/6000 and PowerPC Options 16449@subsection IBM RS/6000 and PowerPC Options 16450@cindex RS/6000 and PowerPC Options 16451@cindex IBM RS/6000 and PowerPC Options 16452 16453These @samp{-m} options are defined for the IBM RS/6000 and PowerPC: 16454@table @gcctabopt 16455@item -mpower 16456@itemx -mno-power 16457@itemx -mpower2 16458@itemx -mno-power2 16459@itemx -mpowerpc 16460@itemx -mno-powerpc 16461@itemx -mpowerpc-gpopt 16462@itemx -mno-powerpc-gpopt 16463@itemx -mpowerpc-gfxopt 16464@itemx -mno-powerpc-gfxopt 16465@need 800 16466@itemx -mpowerpc64 16467@itemx -mno-powerpc64 16468@itemx -mmfcrf 16469@itemx -mno-mfcrf 16470@itemx -mpopcntb 16471@itemx -mno-popcntb 16472@itemx -mpopcntd 16473@itemx -mno-popcntd 16474@itemx -mfprnd 16475@itemx -mno-fprnd 16476@need 800 16477@itemx -mcmpb 16478@itemx -mno-cmpb 16479@itemx -mmfpgpr 16480@itemx -mno-mfpgpr 16481@itemx -mhard-dfp 16482@itemx -mno-hard-dfp 16483@opindex mpower 16484@opindex mno-power 16485@opindex mpower2 16486@opindex mno-power2 16487@opindex mpowerpc 16488@opindex mno-powerpc 16489@opindex mpowerpc-gpopt 16490@opindex mno-powerpc-gpopt 16491@opindex mpowerpc-gfxopt 16492@opindex mno-powerpc-gfxopt 16493@opindex mpowerpc64 16494@opindex mno-powerpc64 16495@opindex mmfcrf 16496@opindex mno-mfcrf 16497@opindex mpopcntb 16498@opindex mno-popcntb 16499@opindex mpopcntd 16500@opindex mno-popcntd 16501@opindex mfprnd 16502@opindex mno-fprnd 16503@opindex mcmpb 16504@opindex mno-cmpb 16505@opindex mmfpgpr 16506@opindex mno-mfpgpr 16507@opindex mhard-dfp 16508@opindex mno-hard-dfp 16509GCC supports two related instruction set architectures for the 16510RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those 16511instructions supported by the @samp{rios} chip set used in the original 16512RS/6000 systems and the @dfn{PowerPC} instruction set is the 16513architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and 16514the IBM 4xx, 6xx, and follow-on microprocessors. 16515 16516Neither architecture is a subset of the other. However there is a 16517large common subset of instructions supported by both. An MQ 16518register is included in processors supporting the POWER architecture. 16519 16520You use these options to specify which instructions are available on the 16521processor you are using. The default value of these options is 16522determined when configuring GCC@. Specifying the 16523@option{-mcpu=@var{cpu_type}} overrides the specification of these 16524options. We recommend you use the @option{-mcpu=@var{cpu_type}} option 16525rather than the options listed above. 16526 16527The @option{-mpower} option allows GCC to generate instructions that 16528are found only in the POWER architecture and to use the MQ register. 16529Specifying @option{-mpower2} implies @option{-power} and also allows GCC 16530to generate instructions that are present in the POWER2 architecture but 16531not the original POWER architecture. 16532 16533The @option{-mpowerpc} option allows GCC to generate instructions that 16534are found only in the 32-bit subset of the PowerPC architecture. 16535Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows 16536GCC to use the optional PowerPC architecture instructions in the 16537General Purpose group, including floating-point square root. Specifying 16538@option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to 16539use the optional PowerPC architecture instructions in the Graphics 16540group, including floating-point select. 16541 16542The @option{-mmfcrf} option allows GCC to generate the move from 16543condition register field instruction implemented on the POWER4 16544processor and other processors that support the PowerPC V2.01 16545architecture. 16546The @option{-mpopcntb} option allows GCC to generate the popcount and 16547double-precision FP reciprocal estimate instruction implemented on the 16548POWER5 processor and other processors that support the PowerPC V2.02 16549architecture. 16550The @option{-mpopcntd} option allows GCC to generate the popcount 16551instruction implemented on the POWER7 processor and other processors 16552that support the PowerPC V2.06 architecture. 16553The @option{-mfprnd} option allows GCC to generate the FP round to 16554integer instructions implemented on the POWER5+ processor and other 16555processors that support the PowerPC V2.03 architecture. 16556The @option{-mcmpb} option allows GCC to generate the compare bytes 16557instruction implemented on the POWER6 processor and other processors 16558that support the PowerPC V2.05 architecture. 16559The @option{-mmfpgpr} option allows GCC to generate the FP move to/from 16560general-purpose register instructions implemented on the POWER6X 16561processor and other processors that support the extended PowerPC V2.05 16562architecture. 16563The @option{-mhard-dfp} option allows GCC to generate the decimal 16564floating-point instructions implemented on some POWER processors. 16565 16566The @option{-mpowerpc64} option allows GCC to generate the additional 1656764-bit instructions that are found in the full PowerPC64 architecture 16568and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to 16569@option{-mno-powerpc64}. 16570 16571If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC 16572will use only the instructions in the common subset of both 16573architectures plus some special AIX common-mode calls, and will not use 16574the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc} 16575permits GCC to use any instruction from either architecture and to 16576allow use of the MQ register; specify this for the Motorola MPC601. 16577 16578@item -mnew-mnemonics 16579@itemx -mold-mnemonics 16580@opindex mnew-mnemonics 16581@opindex mold-mnemonics 16582Select which mnemonics to use in the generated assembler code. With 16583@option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for 16584the PowerPC architecture. With @option{-mold-mnemonics} it uses the 16585assembler mnemonics defined for the POWER architecture. Instructions 16586defined in only one architecture have only one mnemonic; GCC uses that 16587mnemonic irrespective of which of these options is specified. 16588 16589GCC defaults to the mnemonics appropriate for the architecture in 16590use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the 16591value of these option. Unless you are building a cross-compiler, you 16592should normally not specify either @option{-mnew-mnemonics} or 16593@option{-mold-mnemonics}, but should instead accept the default. 16594 16595@item -mcpu=@var{cpu_type} 16596@opindex mcpu 16597Set architecture type, register usage, choice of mnemonics, and 16598instruction scheduling parameters for machine type @var{cpu_type}. 16599Supported values for @var{cpu_type} are @samp{401}, @samp{403}, 16600@samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp}, 16601@samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603}, 16602@samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740}, 16603@samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823}, 16604@samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2}, 16605@samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3}, 16606@samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3}, 16607@samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, 16608@samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios}, 16609@samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}. 16610 16611@option{-mcpu=common} selects a completely generic processor. Code 16612generated under this option will run on any POWER or PowerPC processor. 16613GCC will use only the instructions in the common subset of both 16614architectures, and will not use the MQ register. GCC assumes a generic 16615processor model for scheduling purposes. 16616 16617@option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and 16618@option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit 16619PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine 16620types, with an appropriate, generic processor model assumed for 16621scheduling purposes. 16622 16623The other options specify a specific processor. Code generated under 16624those options will run best on that processor, and may not run at all on 16625others. 16626 16627The @option{-mcpu} options automatically enable or disable the 16628following options: 16629 16630@gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol 16631-mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol 16632-mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol 16633-msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx} 16634 16635The particular options set for any particular CPU will vary between 16636compiler versions, depending on what setting seems to produce optimal 16637code for that CPU; it doesn't necessarily reflect the actual hardware's 16638capabilities. If you wish to set an individual option to a particular 16639value, you may specify it after the @option{-mcpu} option, like 16640@samp{-mcpu=970 -mno-altivec}. 16641 16642On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are 16643not enabled or disabled by the @option{-mcpu} option at present because 16644AIX does not have full support for these options. You may still 16645enable or disable them individually if you're sure it'll work in your 16646environment. 16647 16648@item -mtune=@var{cpu_type} 16649@opindex mtune 16650Set the instruction scheduling parameters for machine type 16651@var{cpu_type}, but do not set the architecture type, register usage, or 16652choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same 16653values for @var{cpu_type} are used for @option{-mtune} as for 16654@option{-mcpu}. If both are specified, the code generated will use the 16655architecture, registers, and mnemonics set by @option{-mcpu}, but the 16656scheduling parameters set by @option{-mtune}. 16657 16658@item -mcmodel=small 16659@opindex mcmodel=small 16660Generate PowerPC64 code for the small model: The TOC is limited to 1666164k. 16662 16663@item -mcmodel=medium 16664@opindex mcmodel=medium 16665Generate PowerPC64 code for the medium model: The TOC and other static 16666data may be up to a total of 4G in size. 16667 16668@item -mcmodel=large 16669@opindex mcmodel=large 16670Generate PowerPC64 code for the large model: The TOC may be up to 4G 16671in size. Other data and code is only limited by the 64-bit address 16672space. 16673 16674@item -maltivec 16675@itemx -mno-altivec 16676@opindex maltivec 16677@opindex mno-altivec 16678Generate code that uses (does not use) AltiVec instructions, and also 16679enable the use of built-in functions that allow more direct access to 16680the AltiVec instruction set. You may also need to set 16681@option{-mabi=altivec} to adjust the current ABI with AltiVec ABI 16682enhancements. 16683 16684@item -mvrsave 16685@itemx -mno-vrsave 16686@opindex mvrsave 16687@opindex mno-vrsave 16688Generate VRSAVE instructions when generating AltiVec code. 16689 16690@item -mgen-cell-microcode 16691@opindex mgen-cell-microcode 16692Generate Cell microcode instructions 16693 16694@item -mwarn-cell-microcode 16695@opindex mwarn-cell-microcode 16696Warning when a Cell microcode instruction is going to emitted. An example 16697of a Cell microcode instruction is a variable shift. 16698 16699@item -msecure-plt 16700@opindex msecure-plt 16701Generate code that allows ld and ld.so to build executables and shared 16702libraries with non-exec .plt and .got sections. This is a PowerPC 1670332-bit SYSV ABI option. 16704 16705@item -mbss-plt 16706@opindex mbss-plt 16707Generate code that uses a BSS .plt section that ld.so fills in, and 16708requires .plt and .got sections that are both writable and executable. 16709This is a PowerPC 32-bit SYSV ABI option. 16710 16711@item -misel 16712@itemx -mno-isel 16713@opindex misel 16714@opindex mno-isel 16715This switch enables or disables the generation of ISEL instructions. 16716 16717@item -misel=@var{yes/no} 16718This switch has been deprecated. Use @option{-misel} and 16719@option{-mno-isel} instead. 16720 16721@item -mspe 16722@itemx -mno-spe 16723@opindex mspe 16724@opindex mno-spe 16725This switch enables or disables the generation of SPE simd 16726instructions. 16727 16728@item -mpaired 16729@itemx -mno-paired 16730@opindex mpaired 16731@opindex mno-paired 16732This switch enables or disables the generation of PAIRED simd 16733instructions. 16734 16735@item -mspe=@var{yes/no} 16736This option has been deprecated. Use @option{-mspe} and 16737@option{-mno-spe} instead. 16738 16739@item -mvsx 16740@itemx -mno-vsx 16741@opindex mvsx 16742@opindex mno-vsx 16743Generate code that uses (does not use) vector/scalar (VSX) 16744instructions, and also enable the use of built-in functions that allow 16745more direct access to the VSX instruction set. 16746 16747@item -mfloat-gprs=@var{yes/single/double/no} 16748@itemx -mfloat-gprs 16749@opindex mfloat-gprs 16750This switch enables or disables the generation of floating-point 16751operations on the general-purpose registers for architectures that 16752support it. 16753 16754The argument @var{yes} or @var{single} enables the use of 16755single-precision floating-point operations. 16756 16757The argument @var{double} enables the use of single and 16758double-precision floating-point operations. 16759 16760The argument @var{no} disables floating-point operations on the 16761general-purpose registers. 16762 16763This option is currently only available on the MPC854x. 16764 16765@item -m32 16766@itemx -m64 16767@opindex m32 16768@opindex m64 16769Generate code for 32-bit or 64-bit environments of Darwin and SVR4 16770targets (including GNU/Linux). The 32-bit environment sets int, long 16771and pointer to 32 bits and generates code that runs on any PowerPC 16772variant. The 64-bit environment sets int to 32 bits and long and 16773pointer to 64 bits, and generates code for PowerPC64, as for 16774@option{-mpowerpc64}. 16775 16776@item -mfull-toc 16777@itemx -mno-fp-in-toc 16778@itemx -mno-sum-in-toc 16779@itemx -mminimal-toc 16780@opindex mfull-toc 16781@opindex mno-fp-in-toc 16782@opindex mno-sum-in-toc 16783@opindex mminimal-toc 16784Modify generation of the TOC (Table Of Contents), which is created for 16785every executable file. The @option{-mfull-toc} option is selected by 16786default. In that case, GCC will allocate at least one TOC entry for 16787each unique non-automatic variable reference in your program. GCC 16788will also place floating-point constants in the TOC@. However, only 1678916,384 entries are available in the TOC@. 16790 16791If you receive a linker error message that saying you have overflowed 16792the available TOC space, you can reduce the amount of TOC space used 16793with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options. 16794@option{-mno-fp-in-toc} prevents GCC from putting floating-point 16795constants in the TOC and @option{-mno-sum-in-toc} forces GCC to 16796generate code to calculate the sum of an address and a constant at 16797run time instead of putting that sum into the TOC@. You may specify one 16798or both of these options. Each causes GCC to produce very slightly 16799slower and larger code at the expense of conserving TOC space. 16800 16801If you still run out of space in the TOC even when you specify both of 16802these options, specify @option{-mminimal-toc} instead. This option causes 16803GCC to make only one TOC entry for every file. When you specify this 16804option, GCC will produce code that is slower and larger but which 16805uses extremely little TOC space. You may wish to use this option 16806only on files that contain less frequently executed code. 16807 16808@item -maix64 16809@itemx -maix32 16810@opindex maix64 16811@opindex maix32 16812Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit 16813@code{long} type, and the infrastructure needed to support them. 16814Specifying @option{-maix64} implies @option{-mpowerpc64} and 16815@option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and 16816implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}. 16817 16818@item -mxl-compat 16819@itemx -mno-xl-compat 16820@opindex mxl-compat 16821@opindex mno-xl-compat 16822Produce code that conforms more closely to IBM XL compiler semantics 16823when using AIX-compatible ABI@. Pass floating-point arguments to 16824prototyped functions beyond the register save area (RSA) on the stack 16825in addition to argument FPRs. Do not assume that most significant 16826double in 128-bit long double value is properly rounded when comparing 16827values and converting to double. Use XL symbol names for long double 16828support routines. 16829 16830The AIX calling convention was extended but not initially documented to 16831handle an obscure K&R C case of calling a function that takes the 16832address of its arguments with fewer arguments than declared. IBM XL 16833compilers access floating-point arguments that do not fit in the 16834RSA from the stack when a subroutine is compiled without 16835optimization. Because always storing floating-point arguments on the 16836stack is inefficient and rarely needed, this option is not enabled by 16837default and only is necessary when calling subroutines compiled by IBM 16838XL compilers without optimization. 16839 16840@item -mpe 16841@opindex mpe 16842Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an 16843application written to use message passing with special startup code to 16844enable the application to run. The system must have PE installed in the 16845standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file 16846must be overridden with the @option{-specs=} option to specify the 16847appropriate directory location. The Parallel Environment does not 16848support threads, so the @option{-mpe} option and the @option{-pthread} 16849option are incompatible. 16850 16851@item -malign-natural 16852@itemx -malign-power 16853@opindex malign-natural 16854@opindex malign-power 16855On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option 16856@option{-malign-natural} overrides the ABI-defined alignment of larger 16857types, such as floating-point doubles, on their natural size-based boundary. 16858The option @option{-malign-power} instructs GCC to follow the ABI-specified 16859alignment rules. GCC defaults to the standard alignment defined in the ABI@. 16860 16861On 64-bit Darwin, natural alignment is the default, and @option{-malign-power} 16862is not supported. 16863 16864@item -msoft-float 16865@itemx -mhard-float 16866@opindex msoft-float 16867@opindex mhard-float 16868Generate code that does not use (uses) the floating-point register set. 16869Software floating-point emulation is provided if you use the 16870@option{-msoft-float} option, and pass the option to GCC when linking. 16871 16872@item -msingle-float 16873@itemx -mdouble-float 16874@opindex msingle-float 16875@opindex mdouble-float 16876Generate code for single- or double-precision floating-point operations. 16877@option{-mdouble-float} implies @option{-msingle-float}. 16878 16879@item -msimple-fpu 16880@opindex msimple-fpu 16881Do not generate sqrt and div instructions for hardware floating-point unit. 16882 16883@item -mfpu 16884@opindex mfpu 16885Specify type of floating-point unit. Valid values are @var{sp_lite} 16886(equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent 16887to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float), 16888and @var{dp_full} (equivalent to -mdouble-float). 16889 16890@item -mxilinx-fpu 16891@opindex mxilinx-fpu 16892Perform optimizations for the floating-point unit on Xilinx PPC 405/440. 16893 16894@item -mmultiple 16895@itemx -mno-multiple 16896@opindex mmultiple 16897@opindex mno-multiple 16898Generate code that uses (does not use) the load multiple word 16899instructions and the store multiple word instructions. These 16900instructions are generated by default on POWER systems, and not 16901generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian 16902PowerPC systems, since those instructions do not work when the 16903processor is in little-endian mode. The exceptions are PPC740 and 16904PPC750 which permit these instructions in little-endian mode. 16905 16906@item -mstring 16907@itemx -mno-string 16908@opindex mstring 16909@opindex mno-string 16910Generate code that uses (does not use) the load string instructions 16911and the store string word instructions to save multiple registers and 16912do small block moves. These instructions are generated by default on 16913POWER systems, and not generated on PowerPC systems. Do not use 16914@option{-mstring} on little-endian PowerPC systems, since those 16915instructions do not work when the processor is in little-endian mode. 16916The exceptions are PPC740 and PPC750 which permit these instructions 16917in little-endian mode. 16918 16919@item -mupdate 16920@itemx -mno-update 16921@opindex mupdate 16922@opindex mno-update 16923Generate code that uses (does not use) the load or store instructions 16924that update the base register to the address of the calculated memory 16925location. These instructions are generated by default. If you use 16926@option{-mno-update}, there is a small window between the time that the 16927stack pointer is updated and the address of the previous frame is 16928stored, which means code that walks the stack frame across interrupts or 16929signals may get corrupted data. 16930 16931@item -mavoid-indexed-addresses 16932@itemx -mno-avoid-indexed-addresses 16933@opindex mavoid-indexed-addresses 16934@opindex mno-avoid-indexed-addresses 16935Generate code that tries to avoid (not avoid) the use of indexed load 16936or store instructions. These instructions can incur a performance 16937penalty on Power6 processors in certain situations, such as when 16938stepping through large arrays that cross a 16M boundary. This option 16939is enabled by default when targetting Power6 and disabled otherwise. 16940 16941@item -mfused-madd 16942@itemx -mno-fused-madd 16943@opindex mfused-madd 16944@opindex mno-fused-madd 16945Generate code that uses (does not use) the floating-point multiply and 16946accumulate instructions. These instructions are generated by default 16947if hardware floating point is used. The machine-dependent 16948@option{-mfused-madd} option is now mapped to the machine-independent 16949@option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is 16950mapped to @option{-ffp-contract=off}. 16951 16952@item -mmulhw 16953@itemx -mno-mulhw 16954@opindex mmulhw 16955@opindex mno-mulhw 16956Generate code that uses (does not use) the half-word multiply and 16957multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors. 16958These instructions are generated by default when targetting those 16959processors. 16960 16961@item -mdlmzb 16962@itemx -mno-dlmzb 16963@opindex mdlmzb 16964@opindex mno-dlmzb 16965Generate code that uses (does not use) the string-search @samp{dlmzb} 16966instruction on the IBM 405, 440, 464 and 476 processors. This instruction is 16967generated by default when targetting those processors. 16968 16969@item -mno-bit-align 16970@itemx -mbit-align 16971@opindex mno-bit-align 16972@opindex mbit-align 16973On System V.4 and embedded PowerPC systems do not (do) force structures 16974and unions that contain bit-fields to be aligned to the base type of the 16975bit-field. 16976 16977For example, by default a structure containing nothing but 8 16978@code{unsigned} bit-fields of length 1 is aligned to a 4-byte 16979boundary and has a size of 4 bytes. By using @option{-mno-bit-align}, 16980the structure is aligned to a 1-byte boundary and is 1 byte in 16981size. 16982 16983@item -mno-strict-align 16984@itemx -mstrict-align 16985@opindex mno-strict-align 16986@opindex mstrict-align 16987On System V.4 and embedded PowerPC systems do not (do) assume that 16988unaligned memory references will be handled by the system. 16989 16990@item -mrelocatable 16991@itemx -mno-relocatable 16992@opindex mrelocatable 16993@opindex mno-relocatable 16994Generate code that allows (does not allow) a static executable to be 16995relocated to a different address at run time. A simple embedded 16996PowerPC system loader should relocate the entire contents of 16997@code{.got2} and 4-byte locations listed in the @code{.fixup} section, 16998a table of 32-bit addresses generated by this option. For this to 16999work, all objects linked together must be compiled with 17000@option{-mrelocatable} or @option{-mrelocatable-lib}. 17001@option{-mrelocatable} code aligns the stack to an 8-byte boundary. 17002 17003@item -mrelocatable-lib 17004@itemx -mno-relocatable-lib 17005@opindex mrelocatable-lib 17006@opindex mno-relocatable-lib 17007Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a 17008@code{.fixup} section to allow static executables to be relocated at 17009run time, but @option{-mrelocatable-lib} does not use the smaller stack 17010alignment of @option{-mrelocatable}. Objects compiled with 17011@option{-mrelocatable-lib} may be linked with objects compiled with 17012any combination of the @option{-mrelocatable} options. 17013 17014@item -mno-toc 17015@itemx -mtoc 17016@opindex mno-toc 17017@opindex mtoc 17018On System V.4 and embedded PowerPC systems do not (do) assume that 17019register 2 contains a pointer to a global area pointing to the addresses 17020used in the program. 17021 17022@item -mlittle 17023@itemx -mlittle-endian 17024@opindex mlittle 17025@opindex mlittle-endian 17026On System V.4 and embedded PowerPC systems compile code for the 17027processor in little-endian mode. The @option{-mlittle-endian} option is 17028the same as @option{-mlittle}. 17029 17030@item -mbig 17031@itemx -mbig-endian 17032@opindex mbig 17033@opindex mbig-endian 17034On System V.4 and embedded PowerPC systems compile code for the 17035processor in big-endian mode. The @option{-mbig-endian} option is 17036the same as @option{-mbig}. 17037 17038@item -mdynamic-no-pic 17039@opindex mdynamic-no-pic 17040On Darwin and Mac OS X systems, compile code so that it is not 17041relocatable, but that its external references are relocatable. The 17042resulting code is suitable for applications, but not shared 17043libraries. 17044 17045@item -msingle-pic-base 17046@opindex msingle-pic-base 17047Treat the register used for PIC addressing as read-only, rather than 17048loading it in the prologue for each function. The runtime system is 17049responsible for initializing this register with an appropriate value 17050before execution begins. 17051 17052@item -mprioritize-restricted-insns=@var{priority} 17053@opindex mprioritize-restricted-insns 17054This option controls the priority that is assigned to 17055dispatch-slot restricted instructions during the second scheduling 17056pass. The argument @var{priority} takes the value @var{0/1/2} to assign 17057@var{no/highest/second-highest} priority to dispatch slot restricted 17058instructions. 17059 17060@item -msched-costly-dep=@var{dependence_type} 17061@opindex msched-costly-dep 17062This option controls which dependences are considered costly 17063by the target during instruction scheduling. The argument 17064@var{dependence_type} takes one of the following values: 17065@var{no}: no dependence is costly, 17066@var{all}: all dependences are costly, 17067@var{true_store_to_load}: a true dependence from store to load is costly, 17068@var{store_to_load}: any dependence from store to load is costly, 17069@var{number}: any dependence for which latency >= @var{number} is costly. 17070 17071@item -minsert-sched-nops=@var{scheme} 17072@opindex minsert-sched-nops 17073This option controls which nop insertion scheme will be used during 17074the second scheduling pass. The argument @var{scheme} takes one of the 17075following values: 17076@var{no}: Don't insert nops. 17077@var{pad}: Pad with nops any dispatch group that has vacant issue slots, 17078according to the scheduler's grouping. 17079@var{regroup_exact}: Insert nops to force costly dependent insns into 17080separate groups. Insert exactly as many nops as needed to force an insn 17081to a new group, according to the estimated processor grouping. 17082@var{number}: Insert nops to force costly dependent insns into 17083separate groups. Insert @var{number} nops to force an insn to a new group. 17084 17085@item -mcall-sysv 17086@opindex mcall-sysv 17087On System V.4 and embedded PowerPC systems compile code using calling 17088conventions that adheres to the March 1995 draft of the System V 17089Application Binary Interface, PowerPC processor supplement. This is the 17090default unless you configured GCC using @samp{powerpc-*-eabiaix}. 17091 17092@item -mcall-sysv-eabi 17093@itemx -mcall-eabi 17094@opindex mcall-sysv-eabi 17095@opindex mcall-eabi 17096Specify both @option{-mcall-sysv} and @option{-meabi} options. 17097 17098@item -mcall-sysv-noeabi 17099@opindex mcall-sysv-noeabi 17100Specify both @option{-mcall-sysv} and @option{-mno-eabi} options. 17101 17102@item -mcall-aixdesc 17103@opindex m 17104On System V.4 and embedded PowerPC systems compile code for the AIX 17105operating system. 17106 17107@item -mcall-linux 17108@opindex mcall-linux 17109On System V.4 and embedded PowerPC systems compile code for the 17110Linux-based GNU system. 17111 17112@item -mcall-freebsd 17113@opindex mcall-freebsd 17114On System V.4 and embedded PowerPC systems compile code for the 17115FreeBSD operating system. 17116 17117@item -mcall-netbsd 17118@opindex mcall-netbsd 17119On System V.4 and embedded PowerPC systems compile code for the 17120NetBSD operating system. 17121 17122@item -mcall-openbsd 17123@opindex mcall-netbsd 17124On System V.4 and embedded PowerPC systems compile code for the 17125OpenBSD operating system. 17126 17127@item -maix-struct-return 17128@opindex maix-struct-return 17129Return all structures in memory (as specified by the AIX ABI)@. 17130 17131@item -msvr4-struct-return 17132@opindex msvr4-struct-return 17133Return structures smaller than 8 bytes in registers (as specified by the 17134SVR4 ABI)@. 17135 17136@item -mabi=@var{abi-type} 17137@opindex mabi 17138Extend the current ABI with a particular extension, or remove such extension. 17139Valid values are @var{altivec}, @var{no-altivec}, @var{spe}, 17140@var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@. 17141 17142@item -mabi=spe 17143@opindex mabi=spe 17144Extend the current ABI with SPE ABI extensions. This does not change 17145the default ABI, instead it adds the SPE ABI extensions to the current 17146ABI@. 17147 17148@item -mabi=no-spe 17149@opindex mabi=no-spe 17150Disable Booke SPE ABI extensions for the current ABI@. 17151 17152@item -mabi=ibmlongdouble 17153@opindex mabi=ibmlongdouble 17154Change the current ABI to use IBM extended-precision long double. 17155This is a PowerPC 32-bit SYSV ABI option. 17156 17157@item -mabi=ieeelongdouble 17158@opindex mabi=ieeelongdouble 17159Change the current ABI to use IEEE extended-precision long double. 17160This is a PowerPC 32-bit Linux ABI option. 17161 17162@item -mprototype 17163@itemx -mno-prototype 17164@opindex mprototype 17165@opindex mno-prototype 17166On System V.4 and embedded PowerPC systems assume that all calls to 17167variable argument functions are properly prototyped. Otherwise, the 17168compiler must insert an instruction before every non prototyped call to 17169set or clear bit 6 of the condition code register (@var{CR}) to 17170indicate whether floating-point values were passed in the floating-point 17171registers in case the function takes variable arguments. With 17172@option{-mprototype}, only calls to prototyped variable argument functions 17173will set or clear the bit. 17174 17175@item -msim 17176@opindex msim 17177On embedded PowerPC systems, assume that the startup module is called 17178@file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and 17179@file{libc.a}. This is the default for @samp{powerpc-*-eabisim} 17180configurations. 17181 17182@item -mmvme 17183@opindex mmvme 17184On embedded PowerPC systems, assume that the startup module is called 17185@file{crt0.o} and the standard C libraries are @file{libmvme.a} and 17186@file{libc.a}. 17187 17188@item -mads 17189@opindex mads 17190On embedded PowerPC systems, assume that the startup module is called 17191@file{crt0.o} and the standard C libraries are @file{libads.a} and 17192@file{libc.a}. 17193 17194@item -myellowknife 17195@opindex myellowknife 17196On embedded PowerPC systems, assume that the startup module is called 17197@file{crt0.o} and the standard C libraries are @file{libyk.a} and 17198@file{libc.a}. 17199 17200@item -mvxworks 17201@opindex mvxworks 17202On System V.4 and embedded PowerPC systems, specify that you are 17203compiling for a VxWorks system. 17204 17205@item -memb 17206@opindex memb 17207On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags 17208header to indicate that @samp{eabi} extended relocations are used. 17209 17210@item -meabi 17211@itemx -mno-eabi 17212@opindex meabi 17213@opindex mno-eabi 17214On System V.4 and embedded PowerPC systems do (do not) adhere to the 17215Embedded Applications Binary Interface (eabi) which is a set of 17216modifications to the System V.4 specifications. Selecting @option{-meabi} 17217means that the stack is aligned to an 8-byte boundary, a function 17218@code{__eabi} is called to from @code{main} to set up the eabi 17219environment, and the @option{-msdata} option can use both @code{r2} and 17220@code{r13} to point to two separate small data areas. Selecting 17221@option{-mno-eabi} means that the stack is aligned to a 16-byte boundary, 17222do not call an initialization function from @code{main}, and the 17223@option{-msdata} option will only use @code{r13} to point to a single 17224small data area. The @option{-meabi} option is on by default if you 17225configured GCC using one of the @samp{powerpc*-*-eabi*} options. 17226 17227@item -msdata=eabi 17228@opindex msdata=eabi 17229On System V.4 and embedded PowerPC systems, put small initialized 17230@code{const} global and static data in the @samp{.sdata2} section, which 17231is pointed to by register @code{r2}. Put small initialized 17232non-@code{const} global and static data in the @samp{.sdata} section, 17233which is pointed to by register @code{r13}. Put small uninitialized 17234global and static data in the @samp{.sbss} section, which is adjacent to 17235the @samp{.sdata} section. The @option{-msdata=eabi} option is 17236incompatible with the @option{-mrelocatable} option. The 17237@option{-msdata=eabi} option also sets the @option{-memb} option. 17238 17239@item -msdata=sysv 17240@opindex msdata=sysv 17241On System V.4 and embedded PowerPC systems, put small global and static 17242data in the @samp{.sdata} section, which is pointed to by register 17243@code{r13}. Put small uninitialized global and static data in the 17244@samp{.sbss} section, which is adjacent to the @samp{.sdata} section. 17245The @option{-msdata=sysv} option is incompatible with the 17246@option{-mrelocatable} option. 17247 17248@item -msdata=default 17249@itemx -msdata 17250@opindex msdata=default 17251@opindex msdata 17252On System V.4 and embedded PowerPC systems, if @option{-meabi} is used, 17253compile code the same as @option{-msdata=eabi}, otherwise compile code the 17254same as @option{-msdata=sysv}. 17255 17256@item -msdata=data 17257@opindex msdata=data 17258On System V.4 and embedded PowerPC systems, put small global 17259data in the @samp{.sdata} section. Put small uninitialized global 17260data in the @samp{.sbss} section. Do not use register @code{r13} 17261to address small data however. This is the default behavior unless 17262other @option{-msdata} options are used. 17263 17264@item -msdata=none 17265@itemx -mno-sdata 17266@opindex msdata=none 17267@opindex mno-sdata 17268On embedded PowerPC systems, put all initialized global and static data 17269in the @samp{.data} section, and all uninitialized data in the 17270@samp{.bss} section. 17271 17272@item -mblock-move-inline-limit=@var{num} 17273@opindex mblock-move-inline-limit 17274Inline all block moves (such as calls to @code{memcpy} or structure 17275copies) less than or equal to @var{num} bytes. The minimum value for 17276@var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit 17277targets. The default value is target-specific. 17278 17279@item -G @var{num} 17280@opindex G 17281@cindex smaller data references (PowerPC) 17282@cindex .sdata/.sdata2 references (PowerPC) 17283On embedded PowerPC systems, put global and static items less than or 17284equal to @var{num} bytes into the small data or bss sections instead of 17285the normal data or bss section. By default, @var{num} is 8. The 17286@option{-G @var{num}} switch is also passed to the linker. 17287All modules should be compiled with the same @option{-G @var{num}} value. 17288 17289@item -mregnames 17290@itemx -mno-regnames 17291@opindex mregnames 17292@opindex mno-regnames 17293On System V.4 and embedded PowerPC systems do (do not) emit register 17294names in the assembly language output using symbolic forms. 17295 17296@item -mlongcall 17297@itemx -mno-longcall 17298@opindex mlongcall 17299@opindex mno-longcall 17300By default assume that all calls are far away so that a longer more 17301expensive calling sequence is required. This is required for calls 17302further than 32 megabytes (33,554,432 bytes) from the current location. 17303A short call will be generated if the compiler knows 17304the call cannot be that far away. This setting can be overridden by 17305the @code{shortcall} function attribute, or by @code{#pragma 17306longcall(0)}. 17307 17308Some linkers are capable of detecting out-of-range calls and generating 17309glue code on the fly. On these systems, long calls are unnecessary and 17310generate slower code. As of this writing, the AIX linker can do this, 17311as can the GNU linker for PowerPC/64. It is planned to add this feature 17312to the GNU linker for 32-bit PowerPC systems as well. 17313 17314On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr 17315callee, L42'', plus a ``branch island'' (glue code). The two target 17316addresses represent the callee and the ``branch island''. The 17317Darwin/PPC linker will prefer the first address and generate a ``bl 17318callee'' if the PPC ``bl'' instruction will reach the callee directly; 17319otherwise, the linker will generate ``bl L42'' to call the ``branch 17320island''. The ``branch island'' is appended to the body of the 17321calling function; it computes the full 32-bit address of the callee 17322and jumps to it. 17323 17324On Mach-O (Darwin) systems, this option directs the compiler emit to 17325the glue for every direct call, and the Darwin linker decides whether 17326to use or discard it. 17327 17328In the future, we may cause GCC to ignore all longcall specifications 17329when the linker is known to generate glue. 17330 17331@item -mtls-markers 17332@itemx -mno-tls-markers 17333@opindex mtls-markers 17334@opindex mno-tls-markers 17335Mark (do not mark) calls to @code{__tls_get_addr} with a relocation 17336specifying the function argument. The relocation allows ld to 17337reliably associate function call with argument setup instructions for 17338TLS optimization, which in turn allows gcc to better schedule the 17339sequence. 17340 17341@item -pthread 17342@opindex pthread 17343Adds support for multithreading with the @dfn{pthreads} library. 17344This option sets flags for both the preprocessor and linker. 17345 17346@item -mrecip 17347@itemx -mno-recip 17348@opindex mrecip 17349This option will enable GCC to use the reciprocal estimate and 17350reciprocal square root estimate instructions with additional 17351Newton-Raphson steps to increase precision instead of doing a divide or 17352square root and divide for floating-point arguments. You should use 17353the @option{-ffast-math} option when using @option{-mrecip} (or at 17354least @option{-funsafe-math-optimizations}, 17355@option{-finite-math-only}, @option{-freciprocal-math} and 17356@option{-fno-trapping-math}). Note that while the throughput of the 17357sequence is generally higher than the throughput of the non-reciprocal 17358instruction, the precision of the sequence can be decreased by up to 2 17359ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square 17360roots. 17361 17362@item -mrecip=@var{opt} 17363@opindex mrecip=opt 17364This option allows to control which reciprocal estimate instructions 17365may be used. @var{opt} is a comma separated list of options, which may 17366be preceded by a @code{!} to invert the option: 17367@code{all}: enable all estimate instructions, 17368@code{default}: enable the default instructions, equivalent to @option{-mrecip}, 17369@code{none}: disable all estimate instructions, equivalent to @option{-mno-recip}; 17370@code{div}: enable the reciprocal approximation instructions for both single and double precision; 17371@code{divf}: enable the single-precision reciprocal approximation instructions; 17372@code{divd}: enable the double-precision reciprocal approximation instructions; 17373@code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision; 17374@code{rsqrtf}: enable the single-precision reciprocal square root approximation instructions; 17375@code{rsqrtd}: enable the double-precision reciprocal square root approximation instructions; 17376 17377So for example, @option{-mrecip=all,!rsqrtd} would enable the 17378all of the reciprocal estimate instructions, except for the 17379@code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions 17380which handle the double-precision reciprocal square root calculations. 17381 17382@item -mrecip-precision 17383@itemx -mno-recip-precision 17384@opindex mrecip-precision 17385Assume (do not assume) that the reciprocal estimate instructions 17386provide higher-precision estimates than is mandated by the PowerPC 17387ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7} 17388automatically selects @option{-mrecip-precision}. The double-precision 17389square root estimate instructions are not generated by 17390default on low-precision machines, since they do not provide an 17391estimate that converges after three steps. 17392 17393@item -mveclibabi=@var{type} 17394@opindex mveclibabi 17395Specifies the ABI type to use for vectorizing intrinsics using an 17396external library. The only type supported at present is @code{mass}, 17397which specifies to use IBM's Mathematical Acceleration Subsystem 17398(MASS) libraries for vectorizing intrinsics using external libraries. 17399GCC will currently emit calls to @code{acosd2}, @code{acosf4}, 17400@code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4}, 17401@code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4}, 17402@code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4}, 17403@code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4}, 17404@code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4}, 17405@code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4}, 17406@code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4}, 17407@code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4}, 17408@code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4}, 17409@code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4}, 17410@code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2}, 17411@code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2}, 17412@code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code 17413for power7. Both @option{-ftree-vectorize} and 17414@option{-funsafe-math-optimizations} have to be enabled. The MASS 17415libraries will have to be specified at link time. 17416 17417@item -mfriz 17418@itemx -mno-friz 17419@opindex mfriz 17420Generate (do not generate) the @code{friz} instruction when the 17421@option{-funsafe-math-optimizations} option is used to optimize 17422rounding of floating-point values to 64-bit integer and back to floating 17423point. The @code{friz} instruction does not return the same value if 17424the floating-point number is too large to fit in an integer. 17425 17426@item -mpointers-to-nested-functions 17427@itemx -mno-pointers-to-nested-functions 17428@opindex mpointers-to-nested-functions 17429Generate (do not generate) code to load up the static chain register 17430(@var{r11}) when calling through a pointer on AIX and 64-bit Linux 17431systems where a function pointer points to a 3-word descriptor giving 17432the function address, TOC value to be loaded in register @var{r2}, and 17433static chain value to be loaded in register @var{r11}. The 17434@option{-mpointers-to-nested-functions} is on by default. You will 17435not be able to call through pointers to nested functions or pointers 17436to functions compiled in other languages that use the static chain if 17437you use the @option{-mno-pointers-to-nested-functions}. 17438 17439@item -msave-toc-indirect 17440@itemx -mno-save-toc-indirect 17441@opindex msave-toc-indirect 17442Generate (do not generate) code to save the TOC value in the reserved 17443stack location in the function prologue if the function calls through 17444a pointer on AIX and 64-bit Linux systems. If the TOC value is not 17445saved in the prologue, it is saved just before the call through the 17446pointer. The @option{-mno-save-toc-indirect} option is the default. 17447@end table 17448 17449@node RX Options 17450@subsection RX Options 17451@cindex RX Options 17452 17453These command-line options are defined for RX targets: 17454 17455@table @gcctabopt 17456@item -m64bit-doubles 17457@itemx -m32bit-doubles 17458@opindex m64bit-doubles 17459@opindex m32bit-doubles 17460Make the @code{double} data type be 64 bits (@option{-m64bit-doubles}) 17461or 32 bits (@option{-m32bit-doubles}) in size. The default is 17462@option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only 17463works on 32-bit values, which is why the default is 17464@option{-m32bit-doubles}. 17465 17466@item -fpu 17467@itemx -nofpu 17468@opindex fpu 17469@opindex nofpu 17470Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX 17471floating-point hardware. The default is enabled for the @var{RX600} 17472series and disabled for the @var{RX200} series. 17473 17474Floating-point instructions will only be generated for 32-bit floating-point 17475values however, so if the @option{-m64bit-doubles} option is in 17476use then the FPU hardware will not be used for doubles. 17477 17478@emph{Note} If the @option{-fpu} option is enabled then 17479@option{-funsafe-math-optimizations} is also enabled automatically. 17480This is because the RX FPU instructions are themselves unsafe. 17481 17482@item -mcpu=@var{name} 17483@opindex -mcpu 17484Selects the type of RX CPU to be targeted. Currently three types are 17485supported, the generic @var{RX600} and @var{RX200} series hardware and 17486the specific @var{RX610} CPU. The default is @var{RX600}. 17487 17488The only difference between @var{RX600} and @var{RX610} is that the 17489@var{RX610} does not support the @code{MVTIPL} instruction. 17490 17491The @var{RX200} series does not have a hardware floating-point unit 17492and so @option{-nofpu} is enabled by default when this type is 17493selected. 17494 17495@item -mbig-endian-data 17496@itemx -mlittle-endian-data 17497@opindex mbig-endian-data 17498@opindex mlittle-endian-data 17499Store data (but not code) in the big-endian format. The default is 17500@option{-mlittle-endian-data}, i.e.@: to store data in the little-endian 17501format. 17502 17503@item -msmall-data-limit=@var{N} 17504@opindex msmall-data-limit 17505Specifies the maximum size in bytes of global and static variables 17506which can be placed into the small data area. Using the small data 17507area can lead to smaller and faster code, but the size of area is 17508limited and it is up to the programmer to ensure that the area does 17509not overflow. Also when the small data area is used one of the RX's 17510registers (usually @code{r13}) is reserved for use pointing to this 17511area, so it is no longer available for use by the compiler. This 17512could result in slower and/or larger code if variables which once 17513could have been held in the reserved register are now pushed onto the 17514stack. 17515 17516Note, common variables (variables that have not been initialized) and 17517constants are not placed into the small data area as they are assigned 17518to other sections in the output executable. 17519 17520The default value is zero, which disables this feature. Note, this 17521feature is not enabled by default with higher optimization levels 17522(@option{-O2} etc) because of the potentially detrimental effects of 17523reserving a register. It is up to the programmer to experiment and 17524discover whether this feature is of benefit to their program. See the 17525description of the @option{-mpid} option for a description of how the 17526actual register to hold the small data area pointer is chosen. 17527 17528@item -msim 17529@itemx -mno-sim 17530@opindex msim 17531@opindex mno-sim 17532Use the simulator runtime. The default is to use the libgloss board 17533specific runtime. 17534 17535@item -mas100-syntax 17536@itemx -mno-as100-syntax 17537@opindex mas100-syntax 17538@opindex mno-as100-syntax 17539When generating assembler output use a syntax that is compatible with 17540Renesas's AS100 assembler. This syntax can also be handled by the GAS 17541assembler but it has some restrictions so generating it is not the 17542default option. 17543 17544@item -mmax-constant-size=@var{N} 17545@opindex mmax-constant-size 17546Specifies the maximum size, in bytes, of a constant that can be used as 17547an operand in a RX instruction. Although the RX instruction set does 17548allow constants of up to 4 bytes in length to be used in instructions, 17549a longer value equates to a longer instruction. Thus in some 17550circumstances it can be beneficial to restrict the size of constants 17551that are used in instructions. Constants that are too big are instead 17552placed into a constant pool and referenced via register indirection. 17553 17554The value @var{N} can be between 0 and 4. A value of 0 (the default) 17555or 4 means that constants of any size are allowed. 17556 17557@item -mrelax 17558@opindex mrelax 17559Enable linker relaxation. Linker relaxation is a process whereby the 17560linker will attempt to reduce the size of a program by finding shorter 17561versions of various instructions. Disabled by default. 17562 17563@item -mint-register=@var{N} 17564@opindex mint-register 17565Specify the number of registers to reserve for fast interrupt handler 17566functions. The value @var{N} can be between 0 and 4. A value of 1 17567means that register @code{r13} will be reserved for the exclusive use 17568of fast interrupt handlers. A value of 2 reserves @code{r13} and 17569@code{r12}. A value of 3 reserves @code{r13}, @code{r12} and 17570@code{r11}, and a value of 4 reserves @code{r13} through @code{r10}. 17571A value of 0, the default, does not reserve any registers. 17572 17573@item -msave-acc-in-interrupts 17574@opindex msave-acc-in-interrupts 17575Specifies that interrupt handler functions should preserve the 17576accumulator register. This is only necessary if normal code might use 17577the accumulator register, for example because it performs 64-bit 17578multiplications. The default is to ignore the accumulator as this 17579makes the interrupt handlers faster. 17580 17581@item -mpid 17582@itemx -mno-pid 17583@opindex mpid 17584@opindex mno-pid 17585Enables the generation of position independent data. When enabled any 17586access to constant data will done via an offset from a base address 17587held in a register. This allows the location of constant data to be 17588determined at run time without requiring the executable to be 17589relocated, which is a benefit to embedded applications with tight 17590memory constraints. Data that can be modified is not affected by this 17591option. 17592 17593Note, using this feature reserves a register, usually @code{r13}, for 17594the constant data base address. This can result in slower and/or 17595larger code, especially in complicated functions. 17596 17597The actual register chosen to hold the constant data base address 17598depends upon whether the @option{-msmall-data-limit} and/or the 17599@option{-mint-register} command-line options are enabled. Starting 17600with register @code{r13} and proceeding downwards, registers are 17601allocated first to satisfy the requirements of @option{-mint-register}, 17602then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it 17603is possible for the small data area register to be @code{r8} if both 17604@option{-mint-register=4} and @option{-mpid} are specified on the 17605command line. 17606 17607By default this feature is not enabled. The default can be restored 17608via the @option{-mno-pid} command-line option. 17609 17610@end table 17611 17612@emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}} 17613has special significance to the RX port when used with the 17614@code{interrupt} function attribute. This attribute indicates a 17615function intended to process fast interrupts. GCC will will ensure 17616that it only uses the registers @code{r10}, @code{r11}, @code{r12} 17617and/or @code{r13} and only provided that the normal use of the 17618corresponding registers have been restricted via the 17619@option{-ffixed-@var{reg}} or @option{-mint-register} command-line 17620options. 17621 17622@node S/390 and zSeries Options 17623@subsection S/390 and zSeries Options 17624@cindex S/390 and zSeries Options 17625 17626These are the @samp{-m} options defined for the S/390 and zSeries architecture. 17627 17628@table @gcctabopt 17629@item -mhard-float 17630@itemx -msoft-float 17631@opindex mhard-float 17632@opindex msoft-float 17633Use (do not use) the hardware floating-point instructions and registers 17634for floating-point operations. When @option{-msoft-float} is specified, 17635functions in @file{libgcc.a} will be used to perform floating-point 17636operations. When @option{-mhard-float} is specified, the compiler 17637generates IEEE floating-point instructions. This is the default. 17638 17639@item -mhard-dfp 17640@itemx -mno-hard-dfp 17641@opindex mhard-dfp 17642@opindex mno-hard-dfp 17643Use (do not use) the hardware decimal-floating-point instructions for 17644decimal-floating-point operations. When @option{-mno-hard-dfp} is 17645specified, functions in @file{libgcc.a} will be used to perform 17646decimal-floating-point operations. When @option{-mhard-dfp} is 17647specified, the compiler generates decimal-floating-point hardware 17648instructions. This is the default for @option{-march=z9-ec} or higher. 17649 17650@item -mlong-double-64 17651@itemx -mlong-double-128 17652@opindex mlong-double-64 17653@opindex mlong-double-128 17654These switches control the size of @code{long double} type. A size 17655of 64 bits makes the @code{long double} type equivalent to the @code{double} 17656type. This is the default. 17657 17658@item -mbackchain 17659@itemx -mno-backchain 17660@opindex mbackchain 17661@opindex mno-backchain 17662Store (do not store) the address of the caller's frame as backchain pointer 17663into the callee's stack frame. 17664A backchain may be needed to allow debugging using tools that do not understand 17665DWARF-2 call frame information. 17666When @option{-mno-packed-stack} is in effect, the backchain pointer is stored 17667at the bottom of the stack frame; when @option{-mpacked-stack} is in effect, 17668the backchain is placed into the topmost word of the 96/160 byte register 17669save area. 17670 17671In general, code compiled with @option{-mbackchain} is call-compatible with 17672code compiled with @option{-mmo-backchain}; however, use of the backchain 17673for debugging purposes usually requires that the whole binary is built with 17674@option{-mbackchain}. Note that the combination of @option{-mbackchain}, 17675@option{-mpacked-stack} and @option{-mhard-float} is not supported. In order 17676to build a linux kernel use @option{-msoft-float}. 17677 17678The default is to not maintain the backchain. 17679 17680@item -mpacked-stack 17681@itemx -mno-packed-stack 17682@opindex mpacked-stack 17683@opindex mno-packed-stack 17684Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is 17685specified, the compiler uses the all fields of the 96/160 byte register save 17686area only for their default purpose; unused fields still take up stack space. 17687When @option{-mpacked-stack} is specified, register save slots are densely 17688packed at the top of the register save area; unused space is reused for other 17689purposes, allowing for more efficient use of the available stack space. 17690However, when @option{-mbackchain} is also in effect, the topmost word of 17691the save area is always used to store the backchain, and the return address 17692register is always saved two words below the backchain. 17693 17694As long as the stack frame backchain is not used, code generated with 17695@option{-mpacked-stack} is call-compatible with code generated with 17696@option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for 17697S/390 or zSeries generated code that uses the stack frame backchain at run 17698time, not just for debugging purposes. Such code is not call-compatible 17699with code compiled with @option{-mpacked-stack}. Also, note that the 17700combination of @option{-mbackchain}, 17701@option{-mpacked-stack} and @option{-mhard-float} is not supported. In order 17702to build a linux kernel use @option{-msoft-float}. 17703 17704The default is to not use the packed stack layout. 17705 17706@item -msmall-exec 17707@itemx -mno-small-exec 17708@opindex msmall-exec 17709@opindex mno-small-exec 17710Generate (or do not generate) code using the @code{bras} instruction 17711to do subroutine calls. 17712This only works reliably if the total executable size does not 17713exceed 64k. The default is to use the @code{basr} instruction instead, 17714which does not have this limitation. 17715 17716@item -m64 17717@itemx -m31 17718@opindex m64 17719@opindex m31 17720When @option{-m31} is specified, generate code compliant to the 17721GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate 17722code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in 17723particular to generate 64-bit instructions. For the @samp{s390} 17724targets, the default is @option{-m31}, while the @samp{s390x} 17725targets default to @option{-m64}. 17726 17727@item -mzarch 17728@itemx -mesa 17729@opindex mzarch 17730@opindex mesa 17731When @option{-mzarch} is specified, generate code using the 17732instructions available on z/Architecture. 17733When @option{-mesa} is specified, generate code using the 17734instructions available on ESA/390. Note that @option{-mesa} is 17735not possible with @option{-m64}. 17736When generating code compliant to the GNU/Linux for S/390 ABI, 17737the default is @option{-mesa}. When generating code compliant 17738to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}. 17739 17740@item -mmvcle 17741@itemx -mno-mvcle 17742@opindex mmvcle 17743@opindex mno-mvcle 17744Generate (or do not generate) code using the @code{mvcle} instruction 17745to perform block moves. When @option{-mno-mvcle} is specified, 17746use a @code{mvc} loop instead. This is the default unless optimizing for 17747size. 17748 17749@item -mdebug 17750@itemx -mno-debug 17751@opindex mdebug 17752@opindex mno-debug 17753Print (or do not print) additional debug information when compiling. 17754The default is to not print debug information. 17755 17756@item -march=@var{cpu-type} 17757@opindex march 17758Generate code that will run on @var{cpu-type}, which is the name of a system 17759representing a certain processor type. Possible values for 17760@var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990}, 17761@samp{z9-109}, @samp{z9-ec} and @samp{z10}. 17762When generating code using the instructions available on z/Architecture, 17763the default is @option{-march=z900}. Otherwise, the default is 17764@option{-march=g5}. 17765 17766@item -mtune=@var{cpu-type} 17767@opindex mtune 17768Tune to @var{cpu-type} everything applicable about the generated code, 17769except for the ABI and the set of available instructions. 17770The list of @var{cpu-type} values is the same as for @option{-march}. 17771The default is the value used for @option{-march}. 17772 17773@item -mtpf-trace 17774@itemx -mno-tpf-trace 17775@opindex mtpf-trace 17776@opindex mno-tpf-trace 17777Generate code that adds (does not add) in TPF OS specific branches to trace 17778routines in the operating system. This option is off by default, even 17779when compiling for the TPF OS@. 17780 17781@item -mfused-madd 17782@itemx -mno-fused-madd 17783@opindex mfused-madd 17784@opindex mno-fused-madd 17785Generate code that uses (does not use) the floating-point multiply and 17786accumulate instructions. These instructions are generated by default if 17787hardware floating point is used. 17788 17789@item -mwarn-framesize=@var{framesize} 17790@opindex mwarn-framesize 17791Emit a warning if the current function exceeds the given frame size. Because 17792this is a compile-time check it doesn't need to be a real problem when the program 17793runs. It is intended to identify functions that most probably cause 17794a stack overflow. It is useful to be used in an environment with limited stack 17795size e.g.@: the linux kernel. 17796 17797@item -mwarn-dynamicstack 17798@opindex mwarn-dynamicstack 17799Emit a warning if the function calls alloca or uses dynamically 17800sized arrays. This is generally a bad idea with a limited stack size. 17801 17802@item -mstack-guard=@var{stack-guard} 17803@itemx -mstack-size=@var{stack-size} 17804@opindex mstack-guard 17805@opindex mstack-size 17806If these options are provided the s390 back end emits additional instructions in 17807the function prologue which trigger a trap if the stack size is @var{stack-guard} 17808bytes above the @var{stack-size} (remember that the stack on s390 grows downward). 17809If the @var{stack-guard} option is omitted the smallest power of 2 larger than 17810the frame size of the compiled function is chosen. 17811These options are intended to be used to help debugging stack overflow problems. 17812The additionally emitted code causes only little overhead and hence can also be 17813used in production like systems without greater performance degradation. The given 17814values have to be exact powers of 2 and @var{stack-size} has to be greater than 17815@var{stack-guard} without exceeding 64k. 17816In order to be efficient the extra code makes the assumption that the stack starts 17817at an address aligned to the value given by @var{stack-size}. 17818The @var{stack-guard} option can only be used in conjunction with @var{stack-size}. 17819@end table 17820 17821@node Score Options 17822@subsection Score Options 17823@cindex Score Options 17824 17825These options are defined for Score implementations: 17826 17827@table @gcctabopt 17828@item -meb 17829@opindex meb 17830Compile code for big-endian mode. This is the default. 17831 17832@item -mel 17833@opindex mel 17834Compile code for little-endian mode. 17835 17836@item -mnhwloop 17837@opindex mnhwloop 17838Disable generate bcnz instruction. 17839 17840@item -muls 17841@opindex muls 17842Enable generate unaligned load and store instruction. 17843 17844@item -mmac 17845@opindex mmac 17846Enable the use of multiply-accumulate instructions. Disabled by default. 17847 17848@item -mscore5 17849@opindex mscore5 17850Specify the SCORE5 as the target architecture. 17851 17852@item -mscore5u 17853@opindex mscore5u 17854Specify the SCORE5U of the target architecture. 17855 17856@item -mscore7 17857@opindex mscore7 17858Specify the SCORE7 as the target architecture. This is the default. 17859 17860@item -mscore7d 17861@opindex mscore7d 17862Specify the SCORE7D as the target architecture. 17863@end table 17864 17865@node SH Options 17866@subsection SH Options 17867 17868These @samp{-m} options are defined for the SH implementations: 17869 17870@table @gcctabopt 17871@item -m1 17872@opindex m1 17873Generate code for the SH1. 17874 17875@item -m2 17876@opindex m2 17877Generate code for the SH2. 17878 17879@item -m2e 17880Generate code for the SH2e. 17881 17882@item -m2a-nofpu 17883@opindex m2a-nofpu 17884Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way 17885that the floating-point unit is not used. 17886 17887@item -m2a-single-only 17888@opindex m2a-single-only 17889Generate code for the SH2a-FPU, in such a way that no double-precision 17890floating-point operations are used. 17891 17892@item -m2a-single 17893@opindex m2a-single 17894Generate code for the SH2a-FPU assuming the floating-point unit is in 17895single-precision mode by default. 17896 17897@item -m2a 17898@opindex m2a 17899Generate code for the SH2a-FPU assuming the floating-point unit is in 17900double-precision mode by default. 17901 17902@item -m3 17903@opindex m3 17904Generate code for the SH3. 17905 17906@item -m3e 17907@opindex m3e 17908Generate code for the SH3e. 17909 17910@item -m4-nofpu 17911@opindex m4-nofpu 17912Generate code for the SH4 without a floating-point unit. 17913 17914@item -m4-single-only 17915@opindex m4-single-only 17916Generate code for the SH4 with a floating-point unit that only 17917supports single-precision arithmetic. 17918 17919@item -m4-single 17920@opindex m4-single 17921Generate code for the SH4 assuming the floating-point unit is in 17922single-precision mode by default. 17923 17924@item -m4 17925@opindex m4 17926Generate code for the SH4. 17927 17928@item -m4a-nofpu 17929@opindex m4a-nofpu 17930Generate code for the SH4al-dsp, or for a SH4a in such a way that the 17931floating-point unit is not used. 17932 17933@item -m4a-single-only 17934@opindex m4a-single-only 17935Generate code for the SH4a, in such a way that no double-precision 17936floating-point operations are used. 17937 17938@item -m4a-single 17939@opindex m4a-single 17940Generate code for the SH4a assuming the floating-point unit is in 17941single-precision mode by default. 17942 17943@item -m4a 17944@opindex m4a 17945Generate code for the SH4a. 17946 17947@item -m4al 17948@opindex m4al 17949Same as @option{-m4a-nofpu}, except that it implicitly passes 17950@option{-dsp} to the assembler. GCC doesn't generate any DSP 17951instructions at the moment. 17952 17953@item -mb 17954@opindex mb 17955Compile code for the processor in big-endian mode. 17956 17957@item -ml 17958@opindex ml 17959Compile code for the processor in little-endian mode. 17960 17961@item -mdalign 17962@opindex mdalign 17963Align doubles at 64-bit boundaries. Note that this changes the calling 17964conventions, and thus some functions from the standard C library will 17965not work unless you recompile it first with @option{-mdalign}. 17966 17967@item -mrelax 17968@opindex mrelax 17969Shorten some address references at link time, when possible; uses the 17970linker option @option{-relax}. 17971 17972@item -mbigtable 17973@opindex mbigtable 17974Use 32-bit offsets in @code{switch} tables. The default is to use 1797516-bit offsets. 17976 17977@item -mbitops 17978@opindex mbitops 17979Enable the use of bit manipulation instructions on SH2A. 17980 17981@item -mfmovd 17982@opindex mfmovd 17983Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for 17984alignment constraints. 17985 17986@item -mhitachi 17987@opindex mhitachi 17988Comply with the calling conventions defined by Renesas. 17989 17990@item -mrenesas 17991@opindex mhitachi 17992Comply with the calling conventions defined by Renesas. 17993 17994@item -mno-renesas 17995@opindex mhitachi 17996Comply with the calling conventions defined for GCC before the Renesas 17997conventions were available. This option is the default for all 17998targets of the SH toolchain. 17999 18000@item -mnomacsave 18001@opindex mnomacsave 18002Mark the @code{MAC} register as call-clobbered, even if 18003@option{-mhitachi} is given. 18004 18005@item -mieee 18006@item -mno-ieee 18007@opindex mieee 18008@opindex mnoieee 18009Control the IEEE compliance of floating-point comparisons, which affects the 18010handling of cases where the result of a comparison is unordered. By default 18011@option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is 18012enabled @option{-mno-ieee} is implicitly set, which results in faster 18013floating-point greater-equal and less-equal comparisons. The implcit settings 18014can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}. 18015 18016@item -minline-ic_invalidate 18017@opindex minline-ic_invalidate 18018Inline code to invalidate instruction cache entries after setting up 18019nested function trampolines. 18020This option has no effect if -musermode is in effect and the selected 18021code generation option (e.g. -m4) does not allow the use of the icbi 18022instruction. 18023If the selected code generation option does not allow the use of the icbi 18024instruction, and -musermode is not in effect, the inlined code will 18025manipulate the instruction cache address array directly with an associative 18026write. This not only requires privileged mode, but it will also 18027fail if the cache line had been mapped via the TLB and has become unmapped. 18028 18029@item -misize 18030@opindex misize 18031Dump instruction size and location in the assembly code. 18032 18033@item -mpadstruct 18034@opindex mpadstruct 18035This option is deprecated. It pads structures to multiple of 4 bytes, 18036which is incompatible with the SH ABI@. 18037 18038@item -msoft-atomic 18039@opindex msoft-atomic 18040Generate GNU/Linux compatible gUSA software atomic sequences for the atomic 18041built-in functions. The generated atomic sequences require support from the 18042interrupt / exception handling code of the system and are only suitable for 18043single-core systems. They will not perform correctly on multi-core systems. 18044This option is enabled by default when the target is @code{sh-*-linux*}. 18045For details on the atomic built-in functions see @ref{__atomic Builtins}. 18046 18047@item -mspace 18048@opindex mspace 18049Optimize for space instead of speed. Implied by @option{-Os}. 18050 18051@item -mprefergot 18052@opindex mprefergot 18053When generating position-independent code, emit function calls using 18054the Global Offset Table instead of the Procedure Linkage Table. 18055 18056@item -musermode 18057@opindex musermode 18058Don't generate privileged mode only code; implies -mno-inline-ic_invalidate 18059if the inlined code would not work in user mode. 18060This is the default when the target is @code{sh-*-linux*}. 18061 18062@item -multcost=@var{number} 18063@opindex multcost=@var{number} 18064Set the cost to assume for a multiply insn. 18065 18066@item -mdiv=@var{strategy} 18067@opindex mdiv=@var{strategy} 18068Set the division strategy to be used for integer division operations. 18069For SHmedia @var{strategy} can be one of: 18070 18071@table @samp 18072 18073@item fp 18074Performs the operation in floating point. This has a very high latency, 18075but needs only a few instructions, so it might be a good choice if 18076your code has enough easily-exploitable ILP to allow the compiler to 18077schedule the floating-point instructions together with other instructions. 18078Division by zero causes a floating-point exception. 18079 18080@item inv 18081Uses integer operations to calculate the inverse of the divisor, 18082and then multiplies the dividend with the inverse. This strategy allows 18083CSE and hoisting of the inverse calculation. Division by zero calculates 18084an unspecified result, but does not trap. 18085 18086@item inv:minlat 18087A variant of @samp{inv} where, if no CSE or hoisting opportunities 18088have been found, or if the entire operation has been hoisted to the same 18089place, the last stages of the inverse calculation are intertwined with the 18090final multiply to reduce the overall latency, at the expense of using a few 18091more instructions, and thus offering fewer scheduling opportunities with 18092other code. 18093 18094@item call 18095Calls a library function that usually implements the @samp{inv:minlat} 18096strategy. 18097This gives high code density for @code{m5-*media-nofpu} compilations. 18098 18099@item call2 18100Uses a different entry point of the same library function, where it 18101assumes that a pointer to a lookup table has already been set up, which 18102exposes the pointer load to CSE and code hoisting optimizations. 18103 18104@item inv:call 18105@itemx inv:call2 18106@itemx inv:fp 18107Use the @samp{inv} algorithm for initial 18108code generation, but if the code stays unoptimized, revert to the @samp{call}, 18109@samp{call2}, or @samp{fp} strategies, respectively. Note that the 18110potentially-trapping side effect of division by zero is carried by a 18111separate instruction, so it is possible that all the integer instructions 18112are hoisted out, but the marker for the side effect stays where it is. 18113A recombination to floating-point operations or a call is not possible 18114in that case. 18115 18116@item inv20u 18117@itemx inv20l 18118Variants of the @samp{inv:minlat} strategy. In the case 18119that the inverse calculation is not separated from the multiply, they speed 18120up division where the dividend fits into 20 bits (plus sign where applicable) 18121by inserting a test to skip a number of operations in this case; this test 18122slows down the case of larger dividends. @samp{inv20u} assumes the case of a such 18123a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely. 18124 18125@end table 18126 18127For targets other than SHmedia @var{strategy} can be one of: 18128 18129@table @samp 18130 18131@item call-div1 18132Calls a library function that uses the single-step division instruction 18133@code{div1} to perform the operation. Division by zero calculates an 18134unspecified result and does not trap. This is the default except for SH4, 18135SH2A and SHcompact. 18136 18137@item call-fp 18138Calls a library function that performs the operation in double precision 18139floating point. Division by zero causes a floating-point exception. This is 18140the default for SHcompact with FPU. Specifying this for targets that do not 18141have a double precision FPU will default to @code{call-div1}. 18142 18143@item call-table 18144Calls a library function that uses a lookup table for small divisors and 18145the @code{div1} instruction with case distinction for larger divisors. Division 18146by zero calculates an unspecified result and does not trap. This is the default 18147for SH4. Specifying this for targets that do not have dynamic shift 18148instructions will default to @code{call-div1}. 18149 18150@end table 18151 18152When a division strategy has not been specified the default strategy will be 18153selected based on the current target. For SH2A the default strategy is to 18154use the @code{divs} and @code{divu} instructions instead of library function 18155calls. 18156 18157@item -maccumulate-outgoing-args 18158@opindex maccumulate-outgoing-args 18159Reserve space once for outgoing arguments in the function prologue rather 18160than around each call. Generally beneficial for performance and size. Also 18161needed for unwinding to avoid changing the stack frame around conditional code. 18162 18163@item -mdivsi3_libfunc=@var{name} 18164@opindex mdivsi3_libfunc=@var{name} 18165Set the name of the library function used for 32-bit signed division to 18166@var{name}. This only affect the name used in the call and inv:call 18167division strategies, and the compiler will still expect the same 18168sets of input/output/clobbered registers as if this option was not present. 18169 18170@item -mfixed-range=@var{register-range} 18171@opindex mfixed-range 18172Generate code treating the given register range as fixed registers. 18173A fixed register is one that the register allocator can not use. This is 18174useful when compiling kernel code. A register range is specified as 18175two registers separated by a dash. Multiple register ranges can be 18176specified separated by a comma. 18177 18178@item -madjust-unroll 18179@opindex madjust-unroll 18180Throttle unrolling to avoid thrashing target registers. 18181This option only has an effect if the gcc code base supports the 18182TARGET_ADJUST_UNROLL_MAX target hook. 18183 18184@item -mindexed-addressing 18185@opindex mindexed-addressing 18186Enable the use of the indexed addressing mode for SHmedia32/SHcompact. 18187This is only safe if the hardware and/or OS implement 32-bit wrap-around 18188semantics for the indexed addressing mode. The architecture allows the 18189implementation of processors with 64-bit MMU, which the OS could use to 18190get 32-bit addressing, but since no current hardware implementation supports 18191this or any other way to make the indexed addressing mode safe to use in 18192the 32-bit ABI, the default is @option{-mno-indexed-addressing}. 18193 18194@item -mgettrcost=@var{number} 18195@opindex mgettrcost=@var{number} 18196Set the cost assumed for the gettr instruction to @var{number}. 18197The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise. 18198 18199@item -mpt-fixed 18200@opindex mpt-fixed 18201Assume pt* instructions won't trap. This will generally generate better 18202scheduled code, but is unsafe on current hardware. The current architecture 18203definition says that ptabs and ptrel trap when the target anded with 3 is 3. 18204This has the unintentional effect of making it unsafe to schedule ptabs / 18205ptrel before a branch, or hoist it out of a loop. For example, 18206__do_global_ctors, a part of libgcc that runs constructors at program 18207startup, calls functions in a list which is delimited by @minus{}1. With the 18208-mpt-fixed option, the ptabs will be done before testing against @minus{}1. 18209That means that all the constructors will be run a bit quicker, but when 18210the loop comes to the end of the list, the program crashes because ptabs 18211loads @minus{}1 into a target register. Since this option is unsafe for any 18212hardware implementing the current architecture specification, the default 18213is -mno-pt-fixed. Unless the user specifies a specific cost with 18214@option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100}; 18215this deters register allocation using target registers for storing 18216ordinary integers. 18217 18218@item -minvalid-symbols 18219@opindex minvalid-symbols 18220Assume symbols might be invalid. Ordinary function symbols generated by 18221the compiler will always be valid to load with movi/shori/ptabs or 18222movi/shori/ptrel, but with assembler and/or linker tricks it is possible 18223to generate symbols that will cause ptabs / ptrel to trap. 18224This option is only meaningful when @option{-mno-pt-fixed} is in effect. 18225It will then prevent cross-basic-block cse, hoisting and most scheduling 18226of symbol loads. The default is @option{-mno-invalid-symbols}. 18227 18228@item -mbranch-cost=@var{num} 18229@opindex mbranch-cost=@var{num} 18230Assume @var{num} to be the cost for a branch instruction. Higher numbers 18231will make the compiler try to generate more branch-free code if possible. 18232If not specified the value is selected depending on the processor type that 18233is being compiled for. 18234 18235@item -mcbranchdi 18236@opindex mcbranchdi 18237Enable the @code{cbranchdi4} instruction pattern. 18238 18239@item -mcmpeqdi 18240@opindex mcmpeqdi 18241Emit the @code{cmpeqdi_t} instruction pattern even when @option{-mcbranchdi} 18242is in effect. 18243 18244@item -mfused-madd 18245@opindex mfused-madd 18246Allow the usage of the @code{fmac} instruction (floating-point 18247multiply-accumulate) if the processor type supports it. Enabling this 18248option might generate code that produces different numeric floating-point 18249results compared to strict IEEE 754 arithmetic. 18250 18251@item -mpretend-cmove 18252@opindex mpretend-cmove 18253Prefer zero-displacement conditional branches for conditional move instruction 18254patterns. This can result in faster code on the SH4 processor. 18255 18256@end table 18257 18258@node Solaris 2 Options 18259@subsection Solaris 2 Options 18260@cindex Solaris 2 options 18261 18262These @samp{-m} options are supported on Solaris 2: 18263 18264@table @gcctabopt 18265@item -mimpure-text 18266@opindex mimpure-text 18267@option{-mimpure-text}, used in addition to @option{-shared}, tells 18268the compiler to not pass @option{-z text} to the linker when linking a 18269shared object. Using this option, you can link position-dependent 18270code into a shared object. 18271 18272@option{-mimpure-text} suppresses the ``relocations remain against 18273allocatable but non-writable sections'' linker error message. 18274However, the necessary relocations will trigger copy-on-write, and the 18275shared object is not actually shared across processes. Instead of 18276using @option{-mimpure-text}, you should compile all source code with 18277@option{-fpic} or @option{-fPIC}. 18278 18279@end table 18280 18281These switches are supported in addition to the above on Solaris 2: 18282 18283@table @gcctabopt 18284@item -pthreads 18285@opindex pthreads 18286Add support for multithreading using the POSIX threads library. This 18287option sets flags for both the preprocessor and linker. This option does 18288not affect the thread safety of object code produced by the compiler or 18289that of libraries supplied with it. 18290 18291@item -pthread 18292@opindex pthread 18293This is a synonym for @option{-pthreads}. 18294@end table 18295 18296@node SPARC Options 18297@subsection SPARC Options 18298@cindex SPARC options 18299 18300These @samp{-m} options are supported on the SPARC: 18301 18302@table @gcctabopt 18303@item -mno-app-regs 18304@itemx -mapp-regs 18305@opindex mno-app-regs 18306@opindex mapp-regs 18307Specify @option{-mapp-regs} to generate output using the global registers 183082 through 4, which the SPARC SVR4 ABI reserves for applications. This 18309is the default. 18310 18311To be fully SVR4 ABI compliant at the cost of some performance loss, 18312specify @option{-mno-app-regs}. You should compile libraries and system 18313software with this option. 18314 18315@item -mflat 18316@itemx -mno-flat 18317@opindex mflat 18318@opindex mno-flat 18319With @option{-mflat}, the compiler does not generate save/restore instructions 18320and uses a ``flat'' or single register window model. This model is compatible 18321with the regular register window model. The local registers and the input 18322registers (0--5) are still treated as ``call-saved'' registers and will be 18323saved on the stack as needed. 18324 18325With @option{-mno-flat} (the default), the compiler generates save/restore 18326instructions (except for leaf functions). This is the normal operating mode. 18327 18328@item -mfpu 18329@itemx -mhard-float 18330@opindex mfpu 18331@opindex mhard-float 18332Generate output containing floating-point instructions. This is the 18333default. 18334 18335@item -mno-fpu 18336@itemx -msoft-float 18337@opindex mno-fpu 18338@opindex msoft-float 18339Generate output containing library calls for floating point. 18340@strong{Warning:} the requisite libraries are not available for all SPARC 18341targets. Normally the facilities of the machine's usual C compiler are 18342used, but this cannot be done directly in cross-compilation. You must make 18343your own arrangements to provide suitable library functions for 18344cross-compilation. The embedded targets @samp{sparc-*-aout} and 18345@samp{sparclite-*-*} do provide software floating-point support. 18346 18347@option{-msoft-float} changes the calling convention in the output file; 18348therefore, it is only useful if you compile @emph{all} of a program with 18349this option. In particular, you need to compile @file{libgcc.a}, the 18350library that comes with GCC, with @option{-msoft-float} in order for 18351this to work. 18352 18353@item -mhard-quad-float 18354@opindex mhard-quad-float 18355Generate output containing quad-word (long double) floating-point 18356instructions. 18357 18358@item -msoft-quad-float 18359@opindex msoft-quad-float 18360Generate output containing library calls for quad-word (long double) 18361floating-point instructions. The functions called are those specified 18362in the SPARC ABI@. This is the default. 18363 18364As of this writing, there are no SPARC implementations that have hardware 18365support for the quad-word floating-point instructions. They all invoke 18366a trap handler for one of these instructions, and then the trap handler 18367emulates the effect of the instruction. Because of the trap handler overhead, 18368this is much slower than calling the ABI library routines. Thus the 18369@option{-msoft-quad-float} option is the default. 18370 18371@item -mno-unaligned-doubles 18372@itemx -munaligned-doubles 18373@opindex mno-unaligned-doubles 18374@opindex munaligned-doubles 18375Assume that doubles have 8-byte alignment. This is the default. 18376 18377With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte 18378alignment only if they are contained in another type, or if they have an 18379absolute address. Otherwise, it assumes they have 4-byte alignment. 18380Specifying this option avoids some rare compatibility problems with code 18381generated by other compilers. It is not the default because it results 18382in a performance loss, especially for floating-point code. 18383 18384@item -mno-faster-structs 18385@itemx -mfaster-structs 18386@opindex mno-faster-structs 18387@opindex mfaster-structs 18388With @option{-mfaster-structs}, the compiler assumes that structures 18389should have 8-byte alignment. This enables the use of pairs of 18390@code{ldd} and @code{std} instructions for copies in structure 18391assignment, in place of twice as many @code{ld} and @code{st} pairs. 18392However, the use of this changed alignment directly violates the SPARC 18393ABI@. Thus, it's intended only for use on targets where the developer 18394acknowledges that their resulting code will not be directly in line with 18395the rules of the ABI@. 18396 18397@item -mcpu=@var{cpu_type} 18398@opindex mcpu 18399Set the instruction set, register set, and instruction scheduling parameters 18400for machine type @var{cpu_type}. Supported values for @var{cpu_type} are 18401@samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc}, 18402@samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, 18403@samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, 18404@samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3}, 18405and @samp{niagara4}. 18406 18407Native Solaris and GNU/Linux toolchains also support the value @samp{native}, 18408which selects the best architecture option for the host processor. 18409@option{-mcpu=native} has no effect if GCC does not recognize 18410the processor. 18411 18412Default instruction scheduling parameters are used for values that select 18413an architecture and not an implementation. These are @samp{v7}, @samp{v8}, 18414@samp{sparclite}, @samp{sparclet}, @samp{v9}. 18415 18416Here is a list of each supported architecture and their supported 18417implementations. 18418 18419@table @asis 18420@item v7 18421cypress 18422 18423@item v8 18424supersparc, hypersparc, leon 18425 18426@item sparclite 18427f930, f934, sparclite86x 18428 18429@item sparclet 18430tsc701 18431 18432@item v9 18433ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4 18434@end table 18435 18436By default (unless configured otherwise), GCC generates code for the V7 18437variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler 18438additionally optimizes it for the Cypress CY7C602 chip, as used in the 18439SPARCStation/SPARCServer 3xx series. This is also appropriate for the older 18440SPARCStation 1, 2, IPX etc. 18441 18442With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC 18443architecture. The only difference from V7 code is that the compiler emits 18444the integer multiply and integer divide instructions which exist in SPARC-V8 18445but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally 18446optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and 184472000 series. 18448 18449With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of 18450the SPARC architecture. This adds the integer multiply, integer divide step 18451and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7. 18452With @option{-mcpu=f930}, the compiler additionally optimizes it for the 18453Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With 18454@option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu 18455MB86934 chip, which is the more recent SPARClite with FPU@. 18456 18457With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of 18458the SPARC architecture. This adds the integer multiply, multiply/accumulate, 18459integer divide step and scan (@code{ffs}) instructions which exist in SPARClet 18460but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally 18461optimizes it for the TEMIC SPARClet chip. 18462 18463With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC 18464architecture. This adds 64-bit integer and floating-point move instructions, 184653 additional floating-point condition code registers and conditional move 18466instructions. With @option{-mcpu=ultrasparc}, the compiler additionally 18467optimizes it for the Sun UltraSPARC I/II/IIi chips. With 18468@option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the 18469Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With 18470@option{-mcpu=niagara}, the compiler additionally optimizes it for 18471Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler 18472additionally optimizes it for Sun UltraSPARC T2 chips. With 18473@option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun 18474UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler 18475additionally optimizes it for Sun UltraSPARC T4 chips. 18476 18477@item -mtune=@var{cpu_type} 18478@opindex mtune 18479Set the instruction scheduling parameters for machine type 18480@var{cpu_type}, but do not set the instruction set or register set that the 18481option @option{-mcpu=@var{cpu_type}} would. 18482 18483The same values for @option{-mcpu=@var{cpu_type}} can be used for 18484@option{-mtune=@var{cpu_type}}, but the only useful values are those 18485that select a particular CPU implementation. Those are @samp{cypress}, 18486@samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934}, 18487@samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3}, 18488@samp{niagara}, @samp{niagara2}, @samp{niagara3} and @samp{niagara4}. With 18489native Solaris and GNU/Linux toolchains, @samp{native} can also be used. 18490 18491@item -mv8plus 18492@itemx -mno-v8plus 18493@opindex mv8plus 18494@opindex mno-v8plus 18495With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The 18496difference from the V8 ABI is that the global and out registers are 18497considered 64 bits wide. This is enabled by default on Solaris in 32-bit 18498mode for all SPARC-V9 processors. 18499 18500@item -mvis 18501@itemx -mno-vis 18502@opindex mvis 18503@opindex mno-vis 18504With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC 18505Visual Instruction Set extensions. The default is @option{-mno-vis}. 18506 18507@item -mvis2 18508@itemx -mno-vis2 18509@opindex mvis2 18510@opindex mno-vis2 18511With @option{-mvis2}, GCC generates code that takes advantage of 18512version 2.0 of the UltraSPARC Visual Instruction Set extensions. The 18513default is @option{-mvis2} when targetting a cpu that supports such 18514instructions, such as UltraSPARC-III and later. Setting @option{-mvis2} 18515also sets @option{-mvis}. 18516 18517@item -mvis3 18518@itemx -mno-vis3 18519@opindex mvis3 18520@opindex mno-vis3 18521With @option{-mvis3}, GCC generates code that takes advantage of 18522version 3.0 of the UltraSPARC Visual Instruction Set extensions. The 18523default is @option{-mvis3} when targetting a cpu that supports such 18524instructions, such as niagara-3 and later. Setting @option{-mvis3} 18525also sets @option{-mvis2} and @option{-mvis}. 18526 18527@item -mpopc 18528@itemx -mno-popc 18529@opindex mpopc 18530@opindex mno-popc 18531With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC 18532population count instruction. The default is @option{-mpopc} 18533when targetting a cpu that supports such instructions, such as Niagara-2 and 18534later. 18535 18536@item -mfmaf 18537@itemx -mno-fmaf 18538@opindex mfmaf 18539@opindex mno-fmaf 18540With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC 18541Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf} 18542when targetting a cpu that supports such instructions, such as Niagara-3 and 18543later. 18544 18545@item -mfix-at697f 18546@opindex mfix-at697f 18547Enable the documented workaround for the single erratum of the Atmel AT697F 18548processor (which corresponds to erratum #13 of the AT697E processor). 18549@end table 18550 18551These @samp{-m} options are supported in addition to the above 18552on SPARC-V9 processors in 64-bit environments: 18553 18554@table @gcctabopt 18555@item -m32 18556@itemx -m64 18557@opindex m32 18558@opindex m64 18559Generate code for a 32-bit or 64-bit environment. 18560The 32-bit environment sets int, long and pointer to 32 bits. 18561The 64-bit environment sets int to 32 bits and long and pointer 18562to 64 bits. 18563 18564@item -mcmodel=@var{which} 18565@opindex mcmodel 18566Set the code model to one of 18567 18568@table @samp 18569@item medlow 18570The Medium/Low code model: 64-bit addresses, programs 18571must be linked in the low 32 bits of memory. Programs can be statically 18572or dynamically linked. 18573 18574@item medmid 18575The Medium/Middle code model: 64-bit addresses, programs 18576must be linked in the low 44 bits of memory, the text and data segments must 18577be less than 2GB in size and the data segment must be located within 2GB of 18578the text segment. 18579 18580@item medany 18581The Medium/Anywhere code model: 64-bit addresses, programs 18582may be linked anywhere in memory, the text and data segments must be less 18583than 2GB in size and the data segment must be located within 2GB of the 18584text segment. 18585 18586@item embmedany 18587The Medium/Anywhere code model for embedded systems: 1858864-bit addresses, the text and data segments must be less than 2GB in 18589size, both starting anywhere in memory (determined at link time). The 18590global register %g4 points to the base of the data segment. Programs 18591are statically linked and PIC is not supported. 18592@end table 18593 18594@item -mmemory-model=@var{mem-model} 18595@opindex mmemory-model 18596Set the memory model in force on the processor to one of 18597 18598@table @samp 18599@item default 18600The default memory model for the processor and operating system. 18601 18602@item rmo 18603Relaxed Memory Order 18604 18605@item pso 18606Partial Store Order 18607 18608@item tso 18609Total Store Order 18610 18611@item sc 18612Sequential Consistency 18613@end table 18614 18615These memory models are formally defined in Appendix D of the Sparc V9 18616architecture manual, as set in the processor's @code{PSTATE.MM} field. 18617 18618@item -mstack-bias 18619@itemx -mno-stack-bias 18620@opindex mstack-bias 18621@opindex mno-stack-bias 18622With @option{-mstack-bias}, GCC assumes that the stack pointer, and 18623frame pointer if present, are offset by @minus{}2047 which must be added back 18624when making stack frame references. This is the default in 64-bit mode. 18625Otherwise, assume no such offset is present. 18626@end table 18627 18628@node SPU Options 18629@subsection SPU Options 18630@cindex SPU options 18631 18632These @samp{-m} options are supported on the SPU: 18633 18634@table @gcctabopt 18635@item -mwarn-reloc 18636@itemx -merror-reloc 18637@opindex mwarn-reloc 18638@opindex merror-reloc 18639 18640The loader for SPU does not handle dynamic relocations. By default, GCC 18641will give an error when it generates code that requires a dynamic 18642relocation. @option{-mno-error-reloc} disables the error, 18643@option{-mwarn-reloc} will generate a warning instead. 18644 18645@item -msafe-dma 18646@itemx -munsafe-dma 18647@opindex msafe-dma 18648@opindex munsafe-dma 18649 18650Instructions that initiate or test completion of DMA must not be 18651reordered with respect to loads and stores of the memory that is being 18652accessed. Users typically address this problem using the volatile 18653keyword, but that can lead to inefficient code in places where the 18654memory is known to not change. Rather than mark the memory as volatile 18655we treat the DMA instructions as potentially effecting all memory. With 18656@option{-munsafe-dma} users must use the volatile keyword to protect 18657memory accesses. 18658 18659@item -mbranch-hints 18660@opindex mbranch-hints 18661 18662By default, GCC will generate a branch hint instruction to avoid 18663pipeline stalls for always taken or probably taken branches. A hint 18664will not be generated closer than 8 instructions away from its branch. 18665There is little reason to disable them, except for debugging purposes, 18666or to make an object a little bit smaller. 18667 18668@item -msmall-mem 18669@itemx -mlarge-mem 18670@opindex msmall-mem 18671@opindex mlarge-mem 18672 18673By default, GCC generates code assuming that addresses are never larger 18674than 18 bits. With @option{-mlarge-mem} code is generated that assumes 18675a full 32-bit address. 18676 18677@item -mstdmain 18678@opindex mstdmain 18679 18680By default, GCC links against startup code that assumes the SPU-style 18681main function interface (which has an unconventional parameter list). 18682With @option{-mstdmain}, GCC will link your program against startup 18683code that assumes a C99-style interface to @code{main}, including a 18684local copy of @code{argv} strings. 18685 18686@item -mfixed-range=@var{register-range} 18687@opindex mfixed-range 18688Generate code treating the given register range as fixed registers. 18689A fixed register is one that the register allocator can not use. This is 18690useful when compiling kernel code. A register range is specified as 18691two registers separated by a dash. Multiple register ranges can be 18692specified separated by a comma. 18693 18694@item -mea32 18695@itemx -mea64 18696@opindex mea32 18697@opindex mea64 18698Compile code assuming that pointers to the PPU address space accessed 18699via the @code{__ea} named address space qualifier are either 32 or 64 18700bits wide. The default is 32 bits. As this is an ABI changing option, 18701all object code in an executable must be compiled with the same setting. 18702 18703@item -maddress-space-conversion 18704@itemx -mno-address-space-conversion 18705@opindex maddress-space-conversion 18706@opindex mno-address-space-conversion 18707Allow/disallow treating the @code{__ea} address space as superset 18708of the generic address space. This enables explicit type casts 18709between @code{__ea} and generic pointer as well as implicit 18710conversions of generic pointers to @code{__ea} pointers. The 18711default is to allow address space pointer conversions. 18712 18713@item -mcache-size=@var{cache-size} 18714@opindex mcache-size 18715This option controls the version of libgcc that the compiler links to an 18716executable and selects a software-managed cache for accessing variables 18717in the @code{__ea} address space with a particular cache size. Possible 18718options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64} 18719and @samp{128}. The default cache size is 64KB. 18720 18721@item -matomic-updates 18722@itemx -mno-atomic-updates 18723@opindex matomic-updates 18724@opindex mno-atomic-updates 18725This option controls the version of libgcc that the compiler links to an 18726executable and selects whether atomic updates to the software-managed 18727cache of PPU-side variables are used. If you use atomic updates, changes 18728to a PPU variable from SPU code using the @code{__ea} named address space 18729qualifier will not interfere with changes to other PPU variables residing 18730in the same cache line from PPU code. If you do not use atomic updates, 18731such interference may occur; however, writing back cache lines will be 18732more efficient. The default behavior is to use atomic updates. 18733 18734@item -mdual-nops 18735@itemx -mdual-nops=@var{n} 18736@opindex mdual-nops 18737By default, GCC will insert nops to increase dual issue when it expects 18738it to increase performance. @var{n} can be a value from 0 to 10. A 18739smaller @var{n} will insert fewer nops. 10 is the default, 0 is the 18740same as @option{-mno-dual-nops}. Disabled with @option{-Os}. 18741 18742@item -mhint-max-nops=@var{n} 18743@opindex mhint-max-nops 18744Maximum number of nops to insert for a branch hint. A branch hint must 18745be at least 8 instructions away from the branch it is effecting. GCC 18746will insert up to @var{n} nops to enforce this, otherwise it will not 18747generate the branch hint. 18748 18749@item -mhint-max-distance=@var{n} 18750@opindex mhint-max-distance 18751The encoding of the branch hint instruction limits the hint to be within 18752256 instructions of the branch it is effecting. By default, GCC makes 18753sure it is within 125. 18754 18755@item -msafe-hints 18756@opindex msafe-hints 18757Work around a hardware bug that causes the SPU to stall indefinitely. 18758By default, GCC will insert the @code{hbrp} instruction to make sure 18759this stall won't happen. 18760 18761@end table 18762 18763@node System V Options 18764@subsection Options for System V 18765 18766These additional options are available on System V Release 4 for 18767compatibility with other compilers on those systems: 18768 18769@table @gcctabopt 18770@item -G 18771@opindex G 18772Create a shared object. 18773It is recommended that @option{-symbolic} or @option{-shared} be used instead. 18774 18775@item -Qy 18776@opindex Qy 18777Identify the versions of each tool used by the compiler, in a 18778@code{.ident} assembler directive in the output. 18779 18780@item -Qn 18781@opindex Qn 18782Refrain from adding @code{.ident} directives to the output file (this is 18783the default). 18784 18785@item -YP,@var{dirs} 18786@opindex YP 18787Search the directories @var{dirs}, and no others, for libraries 18788specified with @option{-l}. 18789 18790@item -Ym,@var{dir} 18791@opindex Ym 18792Look in the directory @var{dir} to find the M4 preprocessor. 18793The assembler uses this option. 18794@c This is supposed to go with a -Yd for predefined M4 macro files, but 18795@c the generic assembler that comes with Solaris takes just -Ym. 18796@end table 18797 18798@node TILE-Gx Options 18799@subsection TILE-Gx Options 18800@cindex TILE-Gx options 18801 18802These @samp{-m} options are supported on the TILE-Gx: 18803 18804@table @gcctabopt 18805@item -mcpu=@var{name} 18806@opindex mcpu 18807Selects the type of CPU to be targeted. Currently the only supported 18808type is @samp{tilegx}. 18809 18810@item -m32 18811@itemx -m64 18812@opindex m32 18813@opindex m64 18814Generate code for a 32-bit or 64-bit environment. The 32-bit 18815environment sets int, long, and pointer to 32 bits. The 64-bit 18816environment sets int to 32 bits and long and pointer to 64 bits. 18817@end table 18818 18819@node TILEPro Options 18820@subsection TILEPro Options 18821@cindex TILEPro options 18822 18823These @samp{-m} options are supported on the TILEPro: 18824 18825@table @gcctabopt 18826@item -mcpu=@var{name} 18827@opindex mcpu 18828Selects the type of CPU to be targeted. Currently the only supported 18829type is @samp{tilepro}. 18830 18831@item -m32 18832@opindex m32 18833Generate code for a 32-bit environment, which sets int, long, and 18834pointer to 32 bits. This is the only supported behavior so the flag 18835is essentially ignored. 18836@end table 18837 18838@node V850 Options 18839@subsection V850 Options 18840@cindex V850 Options 18841 18842These @samp{-m} options are defined for V850 implementations: 18843 18844@table @gcctabopt 18845@item -mlong-calls 18846@itemx -mno-long-calls 18847@opindex mlong-calls 18848@opindex mno-long-calls 18849Treat all calls as being far away (near). If calls are assumed to be 18850far away, the compiler will always load the functions address up into a 18851register, and call indirect through the pointer. 18852 18853@item -mno-ep 18854@itemx -mep 18855@opindex mno-ep 18856@opindex mep 18857Do not optimize (do optimize) basic blocks that use the same index 18858pointer 4 or more times to copy pointer into the @code{ep} register, and 18859use the shorter @code{sld} and @code{sst} instructions. The @option{-mep} 18860option is on by default if you optimize. 18861 18862@item -mno-prolog-function 18863@itemx -mprolog-function 18864@opindex mno-prolog-function 18865@opindex mprolog-function 18866Do not use (do use) external functions to save and restore registers 18867at the prologue and epilogue of a function. The external functions 18868are slower, but use less code space if more than one function saves 18869the same number of registers. The @option{-mprolog-function} option 18870is on by default if you optimize. 18871 18872@item -mspace 18873@opindex mspace 18874Try to make the code as small as possible. At present, this just turns 18875on the @option{-mep} and @option{-mprolog-function} options. 18876 18877@item -mtda=@var{n} 18878@opindex mtda 18879Put static or global variables whose size is @var{n} bytes or less into 18880the tiny data area that register @code{ep} points to. The tiny data 18881area can hold up to 256 bytes in total (128 bytes for byte references). 18882 18883@item -msda=@var{n} 18884@opindex msda 18885Put static or global variables whose size is @var{n} bytes or less into 18886the small data area that register @code{gp} points to. The small data 18887area can hold up to 64 kilobytes. 18888 18889@item -mzda=@var{n} 18890@opindex mzda 18891Put static or global variables whose size is @var{n} bytes or less into 18892the first 32 kilobytes of memory. 18893 18894@item -mv850 18895@opindex mv850 18896Specify that the target processor is the V850. 18897 18898@item -mbig-switch 18899@opindex mbig-switch 18900Generate code suitable for big switch tables. Use this option only if 18901the assembler/linker complain about out of range branches within a switch 18902table. 18903 18904@item -mapp-regs 18905@opindex mapp-regs 18906This option will cause r2 and r5 to be used in the code generated by 18907the compiler. This setting is the default. 18908 18909@item -mno-app-regs 18910@opindex mno-app-regs 18911This option will cause r2 and r5 to be treated as fixed registers. 18912 18913@item -mv850e2v3 18914@opindex mv850e2v3 18915Specify that the target processor is the V850E2V3. The preprocessor 18916constants @samp{__v850e2v3__} will be defined if 18917this option is used. 18918 18919@item -mv850e2 18920@opindex mv850e2 18921Specify that the target processor is the V850E2. The preprocessor 18922constants @samp{__v850e2__} will be defined if this option is used. 18923 18924@item -mv850e1 18925@opindex mv850e1 18926Specify that the target processor is the V850E1. The preprocessor 18927constants @samp{__v850e1__} and @samp{__v850e__} will be defined if 18928this option is used. 18929 18930@item -mv850es 18931@opindex mv850es 18932Specify that the target processor is the V850ES. This is an alias for 18933the @option{-mv850e1} option. 18934 18935@item -mv850e 18936@opindex mv850e 18937Specify that the target processor is the V850E@. The preprocessor 18938constant @samp{__v850e__} will be defined if this option is used. 18939 18940If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1} 18941nor @option{-mv850e2} nor @option{-mv850e2v3} 18942are defined then a default target processor will be chosen and the 18943relevant @samp{__v850*__} preprocessor constant will be defined. 18944 18945The preprocessor constants @samp{__v850} and @samp{__v851__} are always 18946defined, regardless of which processor variant is the target. 18947 18948@item -mdisable-callt 18949@opindex mdisable-callt 18950This option will suppress generation of the CALLT instruction for the 18951v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is 18952@option{-mno-disable-callt} which allows the CALLT instruction to be used. 18953 18954@end table 18955 18956@node VAX Options 18957@subsection VAX Options 18958@cindex VAX options 18959 18960These @samp{-m} options are defined for the VAX: 18961 18962@table @gcctabopt 18963@item -munix 18964@opindex munix 18965Do not output certain jump instructions (@code{aobleq} and so on) 18966that the Unix assembler for the VAX cannot handle across long 18967ranges. 18968 18969@item -mgnu 18970@opindex mgnu 18971Do output those jump instructions, on the assumption that you 18972will assemble with the GNU assembler. 18973 18974@item -mg 18975@opindex mg 18976Output code for G-format floating-point numbers instead of D-format. 18977@end table 18978 18979@node VxWorks Options 18980@subsection VxWorks Options 18981@cindex VxWorks Options 18982 18983The options in this section are defined for all VxWorks targets. 18984Options specific to the target hardware are listed with the other 18985options for that target. 18986 18987@table @gcctabopt 18988@item -mrtp 18989@opindex mrtp 18990GCC can generate code for both VxWorks kernels and real time processes 18991(RTPs). This option switches from the former to the latter. It also 18992defines the preprocessor macro @code{__RTP__}. 18993 18994@item -non-static 18995@opindex non-static 18996Link an RTP executable against shared libraries rather than static 18997libraries. The options @option{-static} and @option{-shared} can 18998also be used for RTPs (@pxref{Link Options}); @option{-static} 18999is the default. 19000 19001@item -Bstatic 19002@itemx -Bdynamic 19003@opindex Bstatic 19004@opindex Bdynamic 19005These options are passed down to the linker. They are defined for 19006compatibility with Diab. 19007 19008@item -Xbind-lazy 19009@opindex Xbind-lazy 19010Enable lazy binding of function calls. This option is equivalent to 19011@option{-Wl,-z,now} and is defined for compatibility with Diab. 19012 19013@item -Xbind-now 19014@opindex Xbind-now 19015Disable lazy binding of function calls. This option is the default and 19016is defined for compatibility with Diab. 19017@end table 19018 19019@node x86-64 Options 19020@subsection x86-64 Options 19021@cindex x86-64 options 19022 19023These are listed under @xref{i386 and x86-64 Options}. 19024 19025@node Xstormy16 Options 19026@subsection Xstormy16 Options 19027@cindex Xstormy16 Options 19028 19029These options are defined for Xstormy16: 19030 19031@table @gcctabopt 19032@item -msim 19033@opindex msim 19034Choose startup files and linker script suitable for the simulator. 19035@end table 19036 19037@node Xtensa Options 19038@subsection Xtensa Options 19039@cindex Xtensa Options 19040 19041These options are supported for Xtensa targets: 19042 19043@table @gcctabopt 19044@item -mconst16 19045@itemx -mno-const16 19046@opindex mconst16 19047@opindex mno-const16 19048Enable or disable use of @code{CONST16} instructions for loading 19049constant values. The @code{CONST16} instruction is currently not a 19050standard option from Tensilica. When enabled, @code{CONST16} 19051instructions are always used in place of the standard @code{L32R} 19052instructions. The use of @code{CONST16} is enabled by default only if 19053the @code{L32R} instruction is not available. 19054 19055@item -mfused-madd 19056@itemx -mno-fused-madd 19057@opindex mfused-madd 19058@opindex mno-fused-madd 19059Enable or disable use of fused multiply/add and multiply/subtract 19060instructions in the floating-point option. This has no effect if the 19061floating-point option is not also enabled. Disabling fused multiply/add 19062and multiply/subtract instructions forces the compiler to use separate 19063instructions for the multiply and add/subtract operations. This may be 19064desirable in some cases where strict IEEE 754-compliant results are 19065required: the fused multiply add/subtract instructions do not round the 19066intermediate result, thereby producing results with @emph{more} bits of 19067precision than specified by the IEEE standard. Disabling fused multiply 19068add/subtract instructions also ensures that the program output is not 19069sensitive to the compiler's ability to combine multiply and add/subtract 19070operations. 19071 19072@item -mserialize-volatile 19073@itemx -mno-serialize-volatile 19074@opindex mserialize-volatile 19075@opindex mno-serialize-volatile 19076When this option is enabled, GCC inserts @code{MEMW} instructions before 19077@code{volatile} memory references to guarantee sequential consistency. 19078The default is @option{-mserialize-volatile}. Use 19079@option{-mno-serialize-volatile} to omit the @code{MEMW} instructions. 19080 19081@item -mforce-no-pic 19082@opindex mforce-no-pic 19083For targets, like GNU/Linux, where all user-mode Xtensa code must be 19084position-independent code (PIC), this option disables PIC for compiling 19085kernel code. 19086 19087@item -mtext-section-literals 19088@itemx -mno-text-section-literals 19089@opindex mtext-section-literals 19090@opindex mno-text-section-literals 19091Control the treatment of literal pools. The default is 19092@option{-mno-text-section-literals}, which places literals in a separate 19093section in the output file. This allows the literal pool to be placed 19094in a data RAM/ROM, and it also allows the linker to combine literal 19095pools from separate object files to remove redundant literals and 19096improve code size. With @option{-mtext-section-literals}, the literals 19097are interspersed in the text section in order to keep them as close as 19098possible to their references. This may be necessary for large assembly 19099files. 19100 19101@item -mtarget-align 19102@itemx -mno-target-align 19103@opindex mtarget-align 19104@opindex mno-target-align 19105When this option is enabled, GCC instructs the assembler to 19106automatically align instructions to reduce branch penalties at the 19107expense of some code density. The assembler attempts to widen density 19108instructions to align branch targets and the instructions following call 19109instructions. If there are not enough preceding safe density 19110instructions to align a target, no widening will be performed. The 19111default is @option{-mtarget-align}. These options do not affect the 19112treatment of auto-aligned instructions like @code{LOOP}, which the 19113assembler will always align, either by widening density instructions or 19114by inserting no-op instructions. 19115 19116@item -mlongcalls 19117@itemx -mno-longcalls 19118@opindex mlongcalls 19119@opindex mno-longcalls 19120When this option is enabled, GCC instructs the assembler to translate 19121direct calls to indirect calls unless it can determine that the target 19122of a direct call is in the range allowed by the call instruction. This 19123translation typically occurs for calls to functions in other source 19124files. Specifically, the assembler translates a direct @code{CALL} 19125instruction into an @code{L32R} followed by a @code{CALLX} instruction. 19126The default is @option{-mno-longcalls}. This option should be used in 19127programs where the call target can potentially be out of range. This 19128option is implemented in the assembler, not the compiler, so the 19129assembly code generated by GCC will still show direct call 19130instructions---look at the disassembled object code to see the actual 19131instructions. Note that the assembler will use an indirect call for 19132every cross-file call, not just those that really will be out of range. 19133@end table 19134 19135@node zSeries Options 19136@subsection zSeries Options 19137@cindex zSeries options 19138 19139These are listed under @xref{S/390 and zSeries Options}. 19140 19141@node Code Gen Options 19142@section Options for Code Generation Conventions 19143@cindex code generation conventions 19144@cindex options, code generation 19145@cindex run-time options 19146 19147These machine-independent options control the interface conventions 19148used in code generation. 19149 19150Most of them have both positive and negative forms; the negative form 19151of @option{-ffoo} would be @option{-fno-foo}. In the table below, only 19152one of the forms is listed---the one that is not the default. You 19153can figure out the other form by either removing @samp{no-} or adding 19154it. 19155 19156@table @gcctabopt 19157@item -fbounds-check 19158@opindex fbounds-check 19159For front ends that support it, generate additional code to check that 19160indices used to access arrays are within the declared range. This is 19161currently only supported by the Java and Fortran front ends, where 19162this option defaults to true and false respectively. 19163 19164@item -ftrapv 19165@opindex ftrapv 19166This option generates traps for signed overflow on addition, subtraction, 19167multiplication operations. 19168 19169@item -fwrapv 19170@opindex fwrapv 19171This option instructs the compiler to assume that signed arithmetic 19172overflow of addition, subtraction and multiplication wraps around 19173using twos-complement representation. This flag enables some optimizations 19174and disables others. This option is enabled by default for the Java 19175front end, as required by the Java language specification. 19176 19177@item -fexceptions 19178@opindex fexceptions 19179Enable exception handling. Generates extra code needed to propagate 19180exceptions. For some targets, this implies GCC will generate frame 19181unwind information for all functions, which can produce significant data 19182size overhead, although it does not affect execution. If you do not 19183specify this option, GCC will enable it by default for languages like 19184C++ that normally require exception handling, and disable it for 19185languages like C that do not normally require it. However, you may need 19186to enable this option when compiling C code that needs to interoperate 19187properly with exception handlers written in C++. You may also wish to 19188disable this option if you are compiling older C++ programs that don't 19189use exception handling. 19190 19191@item -fnon-call-exceptions 19192@opindex fnon-call-exceptions 19193Generate code that allows trapping instructions to throw exceptions. 19194Note that this requires platform-specific runtime support that does 19195not exist everywhere. Moreover, it only allows @emph{trapping} 19196instructions to throw exceptions, i.e.@: memory references or floating-point 19197instructions. It does not allow exceptions to be thrown from 19198arbitrary signal handlers such as @code{SIGALRM}. 19199 19200@item -funwind-tables 19201@opindex funwind-tables 19202Similar to @option{-fexceptions}, except that it will just generate any needed 19203static data, but will not affect the generated code in any other way. 19204You will normally not enable this option; instead, a language processor 19205that needs this handling would enable it on your behalf. 19206 19207@item -fasynchronous-unwind-tables 19208@opindex fasynchronous-unwind-tables 19209Generate unwind table in dwarf2 format, if supported by target machine. The 19210table is exact at each instruction boundary, so it can be used for stack 19211unwinding from asynchronous events (such as debugger or garbage collector). 19212 19213@item -fpcc-struct-return 19214@opindex fpcc-struct-return 19215Return ``short'' @code{struct} and @code{union} values in memory like 19216longer ones, rather than in registers. This convention is less 19217efficient, but it has the advantage of allowing intercallability between 19218GCC-compiled files and files compiled with other compilers, particularly 19219the Portable C Compiler (pcc). 19220 19221The precise convention for returning structures in memory depends 19222on the target configuration macros. 19223 19224Short structures and unions are those whose size and alignment match 19225that of some integer type. 19226 19227@strong{Warning:} code compiled with the @option{-fpcc-struct-return} 19228switch is not binary compatible with code compiled with the 19229@option{-freg-struct-return} switch. 19230Use it to conform to a non-default application binary interface. 19231 19232@item -freg-struct-return 19233@opindex freg-struct-return 19234Return @code{struct} and @code{union} values in registers when possible. 19235This is more efficient for small structures than 19236@option{-fpcc-struct-return}. 19237 19238If you specify neither @option{-fpcc-struct-return} nor 19239@option{-freg-struct-return}, GCC defaults to whichever convention is 19240standard for the target. If there is no standard convention, GCC 19241defaults to @option{-fpcc-struct-return}, except on targets where GCC is 19242the principal compiler. In those cases, we can choose the standard, and 19243we chose the more efficient register return alternative. 19244 19245@strong{Warning:} code compiled with the @option{-freg-struct-return} 19246switch is not binary compatible with code compiled with the 19247@option{-fpcc-struct-return} switch. 19248Use it to conform to a non-default application binary interface. 19249 19250@item -fshort-enums 19251@opindex fshort-enums 19252Allocate to an @code{enum} type only as many bytes as it needs for the 19253declared range of possible values. Specifically, the @code{enum} type 19254will be equivalent to the smallest integer type that has enough room. 19255 19256@strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate 19257code that is not binary compatible with code generated without that switch. 19258Use it to conform to a non-default application binary interface. 19259 19260@item -fshort-double 19261@opindex fshort-double 19262Use the same size for @code{double} as for @code{float}. 19263 19264@strong{Warning:} the @option{-fshort-double} switch causes GCC to generate 19265code that is not binary compatible with code generated without that switch. 19266Use it to conform to a non-default application binary interface. 19267 19268@item -fshort-wchar 19269@opindex fshort-wchar 19270Override the underlying type for @samp{wchar_t} to be @samp{short 19271unsigned int} instead of the default for the target. This option is 19272useful for building programs to run under WINE@. 19273 19274@strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate 19275code that is not binary compatible with code generated without that switch. 19276Use it to conform to a non-default application binary interface. 19277 19278@item -fno-common 19279@opindex fno-common 19280In C code, controls the placement of uninitialized global variables. 19281Unix C compilers have traditionally permitted multiple definitions of 19282such variables in different compilation units by placing the variables 19283in a common block. 19284This is the behavior specified by @option{-fcommon}, and is the default 19285for GCC on most targets. 19286On the other hand, this behavior is not required by ISO C, and on some 19287targets may carry a speed or code size penalty on variable references. 19288The @option{-fno-common} option specifies that the compiler should place 19289uninitialized global variables in the data section of the object file, 19290rather than generating them as common blocks. 19291This has the effect that if the same variable is declared 19292(without @code{extern}) in two different compilations, 19293you will get a multiple-definition error when you link them. 19294In this case, you must compile with @option{-fcommon} instead. 19295Compiling with @option{-fno-common} is useful on targets for which 19296it provides better performance, or if you wish to verify that the 19297program will work on other systems that always treat uninitialized 19298variable declarations this way. 19299 19300@item -fno-ident 19301@opindex fno-ident 19302Ignore the @samp{#ident} directive. 19303 19304@item -finhibit-size-directive 19305@opindex finhibit-size-directive 19306Don't output a @code{.size} assembler directive, or anything else that 19307would cause trouble if the function is split in the middle, and the 19308two halves are placed at locations far apart in memory. This option is 19309used when compiling @file{crtstuff.c}; you should not need to use it 19310for anything else. 19311 19312@item -fverbose-asm 19313@opindex fverbose-asm 19314Put extra commentary information in the generated assembly code to 19315make it more readable. This option is generally only of use to those 19316who actually need to read the generated assembly code (perhaps while 19317debugging the compiler itself). 19318 19319@option{-fno-verbose-asm}, the default, causes the 19320extra information to be omitted and is useful when comparing two assembler 19321files. 19322 19323@item -frecord-gcc-switches 19324@opindex frecord-gcc-switches 19325This switch causes the command line that was used to invoke the 19326compiler to be recorded into the object file that is being created. 19327This switch is only implemented on some targets and the exact format 19328of the recording is target and binary file format dependent, but it 19329usually takes the form of a section containing ASCII text. This 19330switch is related to the @option{-fverbose-asm} switch, but that 19331switch only records information in the assembler output file as 19332comments, so it never reaches the object file. 19333See also @option{-grecord-gcc-switches} for another 19334way of storing compiler options into the object file. 19335 19336@item -fpic 19337@opindex fpic 19338@cindex global offset table 19339@cindex PIC 19340Generate position-independent code (PIC) suitable for use in a shared 19341library, if supported for the target machine. Such code accesses all 19342constant addresses through a global offset table (GOT)@. The dynamic 19343loader resolves the GOT entries when the program starts (the dynamic 19344loader is not part of GCC; it is part of the operating system). If 19345the GOT size for the linked executable exceeds a machine-specific 19346maximum size, you get an error message from the linker indicating that 19347@option{-fpic} does not work; in that case, recompile with @option{-fPIC} 19348instead. (These maximums are 8k on the SPARC and 32k 19349on the m68k and RS/6000. The 386 has no such limit.) 19350 19351Position-independent code requires special support, and therefore works 19352only on certain machines. For the 386, GCC supports PIC for System V 19353but not for the Sun 386i. Code generated for the IBM RS/6000 is always 19354position-independent. 19355 19356When this flag is set, the macros @code{__pic__} and @code{__PIC__} 19357are defined to 1. 19358 19359@item -fPIC 19360@opindex fPIC 19361If supported for the target machine, emit position-independent code, 19362suitable for dynamic linking and avoiding any limit on the size of the 19363global offset table. This option makes a difference on the m68k, 19364PowerPC and SPARC@. 19365 19366Position-independent code requires special support, and therefore works 19367only on certain machines. 19368 19369When this flag is set, the macros @code{__pic__} and @code{__PIC__} 19370are defined to 2. 19371 19372@item -fpie 19373@itemx -fPIE 19374@opindex fpie 19375@opindex fPIE 19376These options are similar to @option{-fpic} and @option{-fPIC}, but 19377generated position independent code can be only linked into executables. 19378Usually these options are used when @option{-pie} GCC option will be 19379used during linking. 19380 19381@option{-fpie} and @option{-fPIE} both define the macros 19382@code{__pie__} and @code{__PIE__}. The macros have the value 1 19383for @option{-fpie} and 2 for @option{-fPIE}. 19384 19385@item -fno-jump-tables 19386@opindex fno-jump-tables 19387Do not use jump tables for switch statements even where it would be 19388more efficient than other code generation strategies. This option is 19389of use in conjunction with @option{-fpic} or @option{-fPIC} for 19390building code that forms part of a dynamic linker and cannot 19391reference the address of a jump table. On some targets, jump tables 19392do not require a GOT and this option is not needed. 19393 19394@item -ffixed-@var{reg} 19395@opindex ffixed 19396Treat the register named @var{reg} as a fixed register; generated code 19397should never refer to it (except perhaps as a stack pointer, frame 19398pointer or in some other fixed role). 19399 19400@var{reg} must be the name of a register. The register names accepted 19401are machine-specific and are defined in the @code{REGISTER_NAMES} 19402macro in the machine description macro file. 19403 19404This flag does not have a negative form, because it specifies a 19405three-way choice. 19406 19407@item -fcall-used-@var{reg} 19408@opindex fcall-used 19409Treat the register named @var{reg} as an allocable register that is 19410clobbered by function calls. It may be allocated for temporaries or 19411variables that do not live across a call. Functions compiled this way 19412will not save and restore the register @var{reg}. 19413 19414It is an error to used this flag with the frame pointer or stack pointer. 19415Use of this flag for other registers that have fixed pervasive roles in 19416the machine's execution model will produce disastrous results. 19417 19418This flag does not have a negative form, because it specifies a 19419three-way choice. 19420 19421@item -fcall-saved-@var{reg} 19422@opindex fcall-saved 19423Treat the register named @var{reg} as an allocable register saved by 19424functions. It may be allocated even for temporaries or variables that 19425live across a call. Functions compiled this way will save and restore 19426the register @var{reg} if they use it. 19427 19428It is an error to used this flag with the frame pointer or stack pointer. 19429Use of this flag for other registers that have fixed pervasive roles in 19430the machine's execution model will produce disastrous results. 19431 19432A different sort of disaster will result from the use of this flag for 19433a register in which function values may be returned. 19434 19435This flag does not have a negative form, because it specifies a 19436three-way choice. 19437 19438@item -fpack-struct[=@var{n}] 19439@opindex fpack-struct 19440Without a value specified, pack all structure members together without 19441holes. When a value is specified (which must be a small power of two), pack 19442structure members according to this value, representing the maximum 19443alignment (that is, objects with default alignment requirements larger than 19444this will be output potentially unaligned at the next fitting location. 19445 19446@strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate 19447code that is not binary compatible with code generated without that switch. 19448Additionally, it makes the code suboptimal. 19449Use it to conform to a non-default application binary interface. 19450 19451@item -finstrument-functions 19452@opindex finstrument-functions 19453Generate instrumentation calls for entry and exit to functions. Just 19454after function entry and just before function exit, the following 19455profiling functions will be called with the address of the current 19456function and its call site. (On some platforms, 19457@code{__builtin_return_address} does not work beyond the current 19458function, so the call site information may not be available to the 19459profiling functions otherwise.) 19460 19461@smallexample 19462void __cyg_profile_func_enter (void *this_fn, 19463 void *call_site); 19464void __cyg_profile_func_exit (void *this_fn, 19465 void *call_site); 19466@end smallexample 19467 19468The first argument is the address of the start of the current function, 19469which may be looked up exactly in the symbol table. 19470 19471This instrumentation is also done for functions expanded inline in other 19472functions. The profiling calls will indicate where, conceptually, the 19473inline function is entered and exited. This means that addressable 19474versions of such functions must be available. If all your uses of a 19475function are expanded inline, this may mean an additional expansion of 19476code size. If you use @samp{extern inline} in your C code, an 19477addressable version of such functions must be provided. (This is 19478normally the case anyways, but if you get lucky and the optimizer always 19479expands the functions inline, you might have gotten away without 19480providing static copies.) 19481 19482A function may be given the attribute @code{no_instrument_function}, in 19483which case this instrumentation will not be done. This can be used, for 19484example, for the profiling functions listed above, high-priority 19485interrupt routines, and any functions from which the profiling functions 19486cannot safely be called (perhaps signal handlers, if the profiling 19487routines generate output or allocate memory). 19488 19489@item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} 19490@opindex finstrument-functions-exclude-file-list 19491 19492Set the list of functions that are excluded from instrumentation (see 19493the description of @code{-finstrument-functions}). If the file that 19494contains a function definition matches with one of @var{file}, then 19495that function is not instrumented. The match is done on substrings: 19496if the @var{file} parameter is a substring of the file name, it is 19497considered to be a match. 19498 19499For example: 19500 19501@smallexample 19502-finstrument-functions-exclude-file-list=/bits/stl,include/sys 19503@end smallexample 19504 19505@noindent 19506will exclude any inline function defined in files whose pathnames 19507contain @code{/bits/stl} or @code{include/sys}. 19508 19509If, for some reason, you want to include letter @code{','} in one of 19510@var{sym}, write @code{'\,'}. For example, 19511@code{-finstrument-functions-exclude-file-list='\,\,tmp'} 19512(note the single quote surrounding the option). 19513 19514@item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} 19515@opindex finstrument-functions-exclude-function-list 19516 19517This is similar to @code{-finstrument-functions-exclude-file-list}, 19518but this option sets the list of function names to be excluded from 19519instrumentation. The function name to be matched is its user-visible 19520name, such as @code{vector<int> blah(const vector<int> &)}, not the 19521internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The 19522match is done on substrings: if the @var{sym} parameter is a substring 19523of the function name, it is considered to be a match. For C99 and C++ 19524extended identifiers, the function name must be given in UTF-8, not 19525using universal character names. 19526 19527@item -fstack-check 19528@opindex fstack-check 19529Generate code to verify that you do not go beyond the boundary of the 19530stack. You should specify this flag if you are running in an 19531environment with multiple threads, but only rarely need to specify it in 19532a single-threaded environment since stack overflow is automatically 19533detected on nearly all systems if there is only one stack. 19534 19535Note that this switch does not actually cause checking to be done; the 19536operating system or the language runtime must do that. The switch causes 19537generation of code to ensure that they see the stack being extended. 19538 19539You can additionally specify a string parameter: @code{no} means no 19540checking, @code{generic} means force the use of old-style checking, 19541@code{specific} means use the best checking method and is equivalent 19542to bare @option{-fstack-check}. 19543 19544Old-style checking is a generic mechanism that requires no specific 19545target support in the compiler but comes with the following drawbacks: 19546 19547@enumerate 19548@item 19549Modified allocation strategy for large objects: they will always be 19550allocated dynamically if their size exceeds a fixed threshold. 19551 19552@item 19553Fixed limit on the size of the static frame of functions: when it is 19554topped by a particular function, stack checking is not reliable and 19555a warning is issued by the compiler. 19556 19557@item 19558Inefficiency: because of both the modified allocation strategy and the 19559generic implementation, the performances of the code are hampered. 19560@end enumerate 19561 19562Note that old-style stack checking is also the fallback method for 19563@code{specific} if no target support has been added in the compiler. 19564 19565@item -fstack-limit-register=@var{reg} 19566@itemx -fstack-limit-symbol=@var{sym} 19567@itemx -fno-stack-limit 19568@opindex fstack-limit-register 19569@opindex fstack-limit-symbol 19570@opindex fno-stack-limit 19571Generate code to ensure that the stack does not grow beyond a certain value, 19572either the value of a register or the address of a symbol. If the stack 19573would grow beyond the value, a signal is raised. For most targets, 19574the signal is raised before the stack overruns the boundary, so 19575it is possible to catch the signal without taking special precautions. 19576 19577For instance, if the stack starts at absolute address @samp{0x80000000} 19578and grows downwards, you can use the flags 19579@option{-fstack-limit-symbol=__stack_limit} and 19580@option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit 19581of 128KB@. Note that this may only work with the GNU linker. 19582 19583@item -fsplit-stack 19584@opindex fsplit-stack 19585Generate code to automatically split the stack before it overflows. 19586The resulting program has a discontiguous stack which can only 19587overflow if the program is unable to allocate any more memory. This 19588is most useful when running threaded programs, as it is no longer 19589necessary to calculate a good stack size to use for each thread. This 19590is currently only implemented for the i386 and x86_64 back ends running 19591GNU/Linux. 19592 19593When code compiled with @option{-fsplit-stack} calls code compiled 19594without @option{-fsplit-stack}, there may not be much stack space 19595available for the latter code to run. If compiling all code, 19596including library code, with @option{-fsplit-stack} is not an option, 19597then the linker can fix up these calls so that the code compiled 19598without @option{-fsplit-stack} always has a large stack. Support for 19599this is implemented in the gold linker in GNU binutils release 2.21 19600and later. 19601 19602@item -fleading-underscore 19603@opindex fleading-underscore 19604This option and its counterpart, @option{-fno-leading-underscore}, forcibly 19605change the way C symbols are represented in the object file. One use 19606is to help link with legacy assembly code. 19607 19608@strong{Warning:} the @option{-fleading-underscore} switch causes GCC to 19609generate code that is not binary compatible with code generated without that 19610switch. Use it to conform to a non-default application binary interface. 19611Not all targets provide complete support for this switch. 19612 19613@item -ftls-model=@var{model} 19614@opindex ftls-model 19615Alter the thread-local storage model to be used (@pxref{Thread-Local}). 19616The @var{model} argument should be one of @code{global-dynamic}, 19617@code{local-dynamic}, @code{initial-exec} or @code{local-exec}. 19618 19619The default without @option{-fpic} is @code{initial-exec}; with 19620@option{-fpic} the default is @code{global-dynamic}. 19621 19622@item -fvisibility=@var{default|internal|hidden|protected} 19623@opindex fvisibility 19624Set the default ELF image symbol visibility to the specified option---all 19625symbols will be marked with this unless overridden within the code. 19626Using this feature can very substantially improve linking and 19627load times of shared object libraries, produce more optimized 19628code, provide near-perfect API export and prevent symbol clashes. 19629It is @strong{strongly} recommended that you use this in any shared objects 19630you distribute. 19631 19632Despite the nomenclature, @code{default} always means public; i.e., 19633available to be linked against from outside the shared object. 19634@code{protected} and @code{internal} are pretty useless in real-world 19635usage so the only other commonly used option will be @code{hidden}. 19636The default if @option{-fvisibility} isn't specified is 19637@code{default}, i.e., make every 19638symbol public---this causes the same behavior as previous versions of 19639GCC@. 19640 19641A good explanation of the benefits offered by ensuring ELF 19642symbols have the correct visibility is given by ``How To Write 19643Shared Libraries'' by Ulrich Drepper (which can be found at 19644@w{@uref{http://people.redhat.com/~drepper/}})---however a superior 19645solution made possible by this option to marking things hidden when 19646the default is public is to make the default hidden and mark things 19647public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden} 19648and @code{__attribute__ ((visibility("default")))} instead of 19649@code{__declspec(dllexport)} you get almost identical semantics with 19650identical syntax. This is a great boon to those working with 19651cross-platform projects. 19652 19653For those adding visibility support to existing code, you may find 19654@samp{#pragma GCC visibility} of use. This works by you enclosing 19655the declarations you wish to set visibility for with (for example) 19656@samp{#pragma GCC visibility push(hidden)} and 19657@samp{#pragma GCC visibility pop}. 19658Bear in mind that symbol visibility should be viewed @strong{as 19659part of the API interface contract} and thus all new code should 19660always specify visibility when it is not the default; i.e., declarations 19661only for use within the local DSO should @strong{always} be marked explicitly 19662as hidden as so to avoid PLT indirection overheads---making this 19663abundantly clear also aids readability and self-documentation of the code. 19664Note that due to ISO C++ specification requirements, operator new and 19665operator delete must always be of default visibility. 19666 19667Be aware that headers from outside your project, in particular system 19668headers and headers from any other library you use, may not be 19669expecting to be compiled with visibility other than the default. You 19670may need to explicitly say @samp{#pragma GCC visibility push(default)} 19671before including any such headers. 19672 19673@samp{extern} declarations are not affected by @samp{-fvisibility}, so 19674a lot of code can be recompiled with @samp{-fvisibility=hidden} with 19675no modifications. However, this means that calls to @samp{extern} 19676functions with no explicit visibility will use the PLT, so it is more 19677effective to use @samp{__attribute ((visibility))} and/or 19678@samp{#pragma GCC visibility} to tell the compiler which @samp{extern} 19679declarations should be treated as hidden. 19680 19681Note that @samp{-fvisibility} does affect C++ vague linkage 19682entities. This means that, for instance, an exception class that will 19683be thrown between DSOs must be explicitly marked with default 19684visibility so that the @samp{type_info} nodes will be unified between 19685the DSOs. 19686 19687An overview of these techniques, their benefits and how to use them 19688is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}. 19689 19690@item -fstrict-volatile-bitfields 19691@opindex fstrict-volatile-bitfields 19692This option should be used if accesses to volatile bit-fields (or other 19693structure fields, although the compiler usually honors those types 19694anyway) should use a single access of the width of the 19695field's type, aligned to a natural alignment if possible. For 19696example, targets with memory-mapped peripheral registers might require 19697all such accesses to be 16 bits wide; with this flag the user could 19698declare all peripheral bit-fields as ``unsigned short'' (assuming short 19699is 16 bits on these targets) to force GCC to use 16-bit accesses 19700instead of, perhaps, a more efficient 32-bit access. 19701 19702If this option is disabled, the compiler will use the most efficient 19703instruction. In the previous example, that might be a 32-bit load 19704instruction, even though that will access bytes that do not contain 19705any portion of the bit-field, or memory-mapped registers unrelated to 19706the one being updated. 19707 19708If the target requires strict alignment, and honoring the field 19709type would require violating this alignment, a warning is issued. 19710If the field has @code{packed} attribute, the access is done without 19711honoring the field type. If the field doesn't have @code{packed} 19712attribute, the access is done honoring the field type. In both cases, 19713GCC assumes that the user knows something about the target hardware 19714that it is unaware of. 19715 19716The default value of this option is determined by the application binary 19717interface for the target processor. 19718 19719@end table 19720 19721@c man end 19722 19723@node Environment Variables 19724@section Environment Variables Affecting GCC 19725@cindex environment variables 19726 19727@c man begin ENVIRONMENT 19728This section describes several environment variables that affect how GCC 19729operates. Some of them work by specifying directories or prefixes to use 19730when searching for various kinds of files. Some are used to specify other 19731aspects of the compilation environment. 19732 19733Note that you can also specify places to search using options such as 19734@option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These 19735take precedence over places specified using environment variables, which 19736in turn take precedence over those specified by the configuration of GCC@. 19737@xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint, 19738GNU Compiler Collection (GCC) Internals}. 19739 19740@table @env 19741@item LANG 19742@itemx LC_CTYPE 19743@c @itemx LC_COLLATE 19744@itemx LC_MESSAGES 19745@c @itemx LC_MONETARY 19746@c @itemx LC_NUMERIC 19747@c @itemx LC_TIME 19748@itemx LC_ALL 19749@findex LANG 19750@findex LC_CTYPE 19751@c @findex LC_COLLATE 19752@findex LC_MESSAGES 19753@c @findex LC_MONETARY 19754@c @findex LC_NUMERIC 19755@c @findex LC_TIME 19756@findex LC_ALL 19757@cindex locale 19758These environment variables control the way that GCC uses 19759localization information which allows GCC to work with different 19760national conventions. GCC inspects the locale categories 19761@env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do 19762so. These locale categories can be set to any value supported by your 19763installation. A typical value is @samp{en_GB.UTF-8} for English in the United 19764Kingdom encoded in UTF-8. 19765 19766The @env{LC_CTYPE} environment variable specifies character 19767classification. GCC uses it to determine the character boundaries in 19768a string; this is needed for some multibyte encodings that contain quote 19769and escape characters that would otherwise be interpreted as a string 19770end or escape. 19771 19772The @env{LC_MESSAGES} environment variable specifies the language to 19773use in diagnostic messages. 19774 19775If the @env{LC_ALL} environment variable is set, it overrides the value 19776of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE} 19777and @env{LC_MESSAGES} default to the value of the @env{LANG} 19778environment variable. If none of these variables are set, GCC 19779defaults to traditional C English behavior. 19780 19781@item TMPDIR 19782@findex TMPDIR 19783If @env{TMPDIR} is set, it specifies the directory to use for temporary 19784files. GCC uses temporary files to hold the output of one stage of 19785compilation which is to be used as input to the next stage: for example, 19786the output of the preprocessor, which is the input to the compiler 19787proper. 19788 19789@item GCC_COMPARE_DEBUG 19790@findex GCC_COMPARE_DEBUG 19791Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing 19792@option{-fcompare-debug} to the compiler driver. See the documentation 19793of this option for more details. 19794 19795@item GCC_EXEC_PREFIX 19796@findex GCC_EXEC_PREFIX 19797If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the 19798names of the subprograms executed by the compiler. No slash is added 19799when this prefix is combined with the name of a subprogram, but you can 19800specify a prefix that ends with a slash if you wish. 19801 19802If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out 19803an appropriate prefix to use based on the pathname it was invoked with. 19804 19805If GCC cannot find the subprogram using the specified prefix, it 19806tries looking in the usual places for the subprogram. 19807 19808The default value of @env{GCC_EXEC_PREFIX} is 19809@file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to 19810the installed compiler. In many cases @var{prefix} is the value 19811of @code{prefix} when you ran the @file{configure} script. 19812 19813Other prefixes specified with @option{-B} take precedence over this prefix. 19814 19815This prefix is also used for finding files such as @file{crt0.o} that are 19816used for linking. 19817 19818In addition, the prefix is used in an unusual way in finding the 19819directories to search for header files. For each of the standard 19820directories whose name normally begins with @samp{/usr/local/lib/gcc} 19821(more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries 19822replacing that beginning with the specified prefix to produce an 19823alternate directory name. Thus, with @option{-Bfoo/}, GCC will search 19824@file{foo/bar} where it would normally search @file{/usr/local/lib/bar}. 19825These alternate directories are searched first; the standard directories 19826come next. If a standard directory begins with the configured 19827@var{prefix} then the value of @var{prefix} is replaced by 19828@env{GCC_EXEC_PREFIX} when looking for header files. 19829 19830@item COMPILER_PATH 19831@findex COMPILER_PATH 19832The value of @env{COMPILER_PATH} is a colon-separated list of 19833directories, much like @env{PATH}. GCC tries the directories thus 19834specified when searching for subprograms, if it can't find the 19835subprograms using @env{GCC_EXEC_PREFIX}. 19836 19837@item LIBRARY_PATH 19838@findex LIBRARY_PATH 19839The value of @env{LIBRARY_PATH} is a colon-separated list of 19840directories, much like @env{PATH}. When configured as a native compiler, 19841GCC tries the directories thus specified when searching for special 19842linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking 19843using GCC also uses these directories when searching for ordinary 19844libraries for the @option{-l} option (but directories specified with 19845@option{-L} come first). 19846 19847@item LANG 19848@findex LANG 19849@cindex locale definition 19850This variable is used to pass locale information to the compiler. One way in 19851which this information is used is to determine the character set to be used 19852when character literals, string literals and comments are parsed in C and C++. 19853When the compiler is configured to allow multibyte characters, 19854the following values for @env{LANG} are recognized: 19855 19856@table @samp 19857@item C-JIS 19858Recognize JIS characters. 19859@item C-SJIS 19860Recognize SJIS characters. 19861@item C-EUCJP 19862Recognize EUCJP characters. 19863@end table 19864 19865If @env{LANG} is not defined, or if it has some other value, then the 19866compiler will use mblen and mbtowc as defined by the default locale to 19867recognize and translate multibyte characters. 19868@end table 19869 19870@noindent 19871Some additional environments variables affect the behavior of the 19872preprocessor. 19873 19874@include cppenv.texi 19875 19876@c man end 19877 19878@node Precompiled Headers 19879@section Using Precompiled Headers 19880@cindex precompiled headers 19881@cindex speed of compilation 19882 19883Often large projects have many header files that are included in every 19884source file. The time the compiler takes to process these header files 19885over and over again can account for nearly all of the time required to 19886build the project. To make builds faster, GCC allows users to 19887`precompile' a header file; then, if builds can use the precompiled 19888header file they will be much faster. 19889 19890To create a precompiled header file, simply compile it as you would any 19891other file, if necessary using the @option{-x} option to make the driver 19892treat it as a C or C++ header file. You will probably want to use a 19893tool like @command{make} to keep the precompiled header up-to-date when 19894the headers it contains change. 19895 19896A precompiled header file will be searched for when @code{#include} is 19897seen in the compilation. As it searches for the included file 19898(@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the 19899compiler looks for a precompiled header in each directory just before it 19900looks for the include file in that directory. The name searched for is 19901the name specified in the @code{#include} with @samp{.gch} appended. If 19902the precompiled header file can't be used, it is ignored. 19903 19904For instance, if you have @code{#include "all.h"}, and you have 19905@file{all.h.gch} in the same directory as @file{all.h}, then the 19906precompiled header file will be used if possible, and the original 19907header will be used otherwise. 19908 19909Alternatively, you might decide to put the precompiled header file in a 19910directory and use @option{-I} to ensure that directory is searched 19911before (or instead of) the directory containing the original header. 19912Then, if you want to check that the precompiled header file is always 19913used, you can put a file of the same name as the original header in this 19914directory containing an @code{#error} command. 19915 19916This also works with @option{-include}. So yet another way to use 19917precompiled headers, good for projects not designed with precompiled 19918header files in mind, is to simply take most of the header files used by 19919a project, include them from another header file, precompile that header 19920file, and @option{-include} the precompiled header. If the header files 19921have guards against multiple inclusion, they will be skipped because 19922they've already been included (in the precompiled header). 19923 19924If you need to precompile the same header file for different 19925languages, targets, or compiler options, you can instead make a 19926@emph{directory} named like @file{all.h.gch}, and put each precompiled 19927header in the directory, perhaps using @option{-o}. It doesn't matter 19928what you call the files in the directory, every precompiled header in 19929the directory will be considered. The first precompiled header 19930encountered in the directory that is valid for this compilation will 19931be used; they're searched in no particular order. 19932 19933There are many other possibilities, limited only by your imagination, 19934good sense, and the constraints of your build system. 19935 19936A precompiled header file can be used only when these conditions apply: 19937 19938@itemize 19939@item 19940Only one precompiled header can be used in a particular compilation. 19941 19942@item 19943A precompiled header can't be used once the first C token is seen. You 19944can have preprocessor directives before a precompiled header; you can 19945even include a precompiled header from inside another header, so long as 19946there are no C tokens before the @code{#include}. 19947 19948@item 19949The precompiled header file must be produced for the same language as 19950the current compilation. You can't use a C precompiled header for a C++ 19951compilation. 19952 19953@item 19954The precompiled header file must have been produced by the same compiler 19955binary as the current compilation is using. 19956 19957@item 19958Any macros defined before the precompiled header is included must 19959either be defined in the same way as when the precompiled header was 19960generated, or must not affect the precompiled header, which usually 19961means that they don't appear in the precompiled header at all. 19962 19963The @option{-D} option is one way to define a macro before a 19964precompiled header is included; using a @code{#define} can also do it. 19965There are also some options that define macros implicitly, like 19966@option{-O} and @option{-Wdeprecated}; the same rule applies to macros 19967defined this way. 19968 19969@item If debugging information is output when using the precompiled 19970header, using @option{-g} or similar, the same kind of debugging information 19971must have been output when building the precompiled header. However, 19972a precompiled header built using @option{-g} can be used in a compilation 19973when no debugging information is being output. 19974 19975@item The same @option{-m} options must generally be used when building 19976and using the precompiled header. @xref{Submodel Options}, 19977for any cases where this rule is relaxed. 19978 19979@item Each of the following options must be the same when building and using 19980the precompiled header: 19981 19982@gccoptlist{-fexceptions} 19983 19984@item 19985Some other command-line options starting with @option{-f}, 19986@option{-p}, or @option{-O} must be defined in the same way as when 19987the precompiled header was generated. At present, it's not clear 19988which options are safe to change and which are not; the safest choice 19989is to use exactly the same options when generating and using the 19990precompiled header. The following are known to be safe: 19991 19992@gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol 19993-fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol 19994-fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol 19995-pedantic-errors} 19996 19997@end itemize 19998 19999For all of these except the last, the compiler will automatically 20000ignore the precompiled header if the conditions aren't met. If you 20001find an option combination that doesn't work and doesn't cause the 20002precompiled header to be ignored, please consider filing a bug report, 20003see @ref{Bugs}. 20004 20005If you do use differing options when generating and using the 20006precompiled header, the actual behavior will be a mixture of the 20007behavior for the options. For instance, if you use @option{-g} to 20008generate the precompiled header but not when using it, you may or may 20009not get debugging information for routines in the precompiled header. 20010