1@c Copyright (C) 1988-2021 Free Software Foundation, Inc. 2@c This is part of the GCC manual. 3@c For copying conditions, see the file gcc.texi. 4 5@ignore 6@c man begin INCLUDE 7@include gcc-vers.texi 8@c man end 9 10@c man begin COPYRIGHT 11Copyright @copyright{} 1988-2021 Free Software Foundation, Inc. 12 13Permission is granted to copy, distribute and/or modify this document 14under the terms of the GNU Free Documentation License, Version 1.3 or 15any later version published by the Free Software Foundation; with the 16Invariant Sections being ``GNU General Public License'' and ``Funding 17Free Software'', the Front-Cover texts being (a) (see below), and with 18the Back-Cover Texts being (b) (see below). A copy of the license is 19included in the gfdl(7) man page. 20 21(a) The FSF's Front-Cover Text is: 22 23 A GNU Manual 24 25(b) The FSF's Back-Cover Text is: 26 27 You have freedom to copy and modify this GNU Manual, like GNU 28 software. Copies published by the Free Software Foundation raise 29 funds for GNU development. 30@c man end 31@c Set file name and title for the man page. 32@setfilename gcc 33@settitle GNU project C and C++ compiler 34@c man begin SYNOPSIS 35gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}] 36 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}] 37 [@option{-W}@var{warn}@dots{}] [@option{-Wpedantic}] 38 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}] 39 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}] 40 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}] 41 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{} 42 43Only the most useful options are listed here; see below for the 44remainder. @command{g++} accepts mostly the same options as @command{gcc}. 45@c man end 46@c man begin SEEALSO 47gpl(7), gfdl(7), fsf-funding(7), 48cpp(1), gcov(1), as(1), ld(1), gdb(1), dbx(1) 49and the Info entries for @file{gcc}, @file{cpp}, @file{as}, 50@file{ld}, @file{binutils} and @file{gdb}. 51@c man end 52@c man begin BUGS 53For instructions on reporting bugs, see 54@w{@value{BUGURL}}. 55@c man end 56@c man begin AUTHOR 57See the Info entry for @command{gcc}, or 58@w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}}, 59for contributors to GCC@. 60@c man end 61@end ignore 62 63@node Invoking GCC 64@chapter GCC Command Options 65@cindex GCC command options 66@cindex command options 67@cindex options, GCC command 68 69@c man begin DESCRIPTION 70When you invoke GCC, it normally does preprocessing, compilation, 71assembly and linking. The ``overall options'' allow you to stop this 72process at an intermediate stage. For example, the @option{-c} option 73says not to run the linker. Then the output consists of object files 74output by the assembler. 75@xref{Overall Options,,Options Controlling the Kind of Output}. 76 77Other options are passed on to one or more stages of processing. Some options 78control the preprocessor and others the compiler itself. Yet other 79options control the assembler and linker; most of these are not 80documented here, since you rarely need to use any of them. 81 82@cindex C compilation options 83Most of the command-line options that you can use with GCC are useful 84for C programs; when an option is only useful with another language 85(usually C++), the explanation says so explicitly. If the description 86for a particular option does not mention a source language, you can use 87that option with all supported languages. 88 89@cindex cross compiling 90@cindex specifying machine version 91@cindex specifying compiler version and target machine 92@cindex compiler version, specifying 93@cindex target machine, specifying 94The usual way to run GCC is to run the executable called @command{gcc}, or 95@command{@var{machine}-gcc} when cross-compiling, or 96@command{@var{machine}-gcc-@var{version}} to run a specific version of GCC. 97When you compile C++ programs, you should invoke GCC as @command{g++} 98instead. @xref{Invoking G++,,Compiling C++ Programs}, 99for information about the differences in behavior between @command{gcc} 100and @command{g++} when compiling C++ programs. 101 102@cindex grouping options 103@cindex options, grouping 104The @command{gcc} program accepts options and file names as operands. Many 105options have multi-letter names; therefore multiple single-letter options 106may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d 107-v}}. 108 109@cindex order of options 110@cindex options, order 111You can mix options and other arguments. For the most part, the order 112you use doesn't matter. Order does matter when you use several 113options of the same kind; for example, if you specify @option{-L} more 114than once, the directories are searched in the order specified. Also, 115the placement of the @option{-l} option is significant. 116 117Many options have long names starting with @samp{-f} or with 118@samp{-W}---for example, 119@option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of 120these have both positive and negative forms; the negative form of 121@option{-ffoo} is @option{-fno-foo}. This manual documents 122only one of these two forms, whichever one is not the default. 123 124Some options take one or more arguments typically separated either 125by a space or by the equals sign (@samp{=}) from the option name. 126Unless documented otherwise, an argument can be either numeric or 127a string. Numeric arguments must typically be small unsigned decimal 128or hexadecimal integers. Hexadecimal arguments must begin with 129the @samp{0x} prefix. Arguments to options that specify a size 130threshold of some sort may be arbitrarily large decimal or hexadecimal 131integers followed by a byte size suffix designating a multiple of bytes 132such as @code{kB} and @code{KiB} for kilobyte and kibibyte, respectively, 133@code{MB} and @code{MiB} for megabyte and mebibyte, @code{GB} and 134@code{GiB} for gigabyte and gigibyte, and so on. Such arguments are 135designated by @var{byte-size} in the following text. Refer to the NIST, 136IEC, and other relevant national and international standards for the full 137listing and explanation of the binary and decimal byte size prefixes. 138 139@c man end 140 141@xref{Option Index}, for an index to GCC's options. 142 143@menu 144* Option Summary:: Brief list of all options, without explanations. 145* Overall Options:: Controlling the kind of output: 146 an executable, object files, assembler files, 147 or preprocessed source. 148* Invoking G++:: Compiling C++ programs. 149* C Dialect Options:: Controlling the variant of C language compiled. 150* C++ Dialect Options:: Variations on C++. 151* Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C 152 and Objective-C++. 153* Diagnostic Message Formatting Options:: Controlling how diagnostics should 154 be formatted. 155* Warning Options:: How picky should the compiler be? 156* Static Analyzer Options:: More expensive warnings. 157* Debugging Options:: Producing debuggable code. 158* Optimize Options:: How much optimization? 159* Instrumentation Options:: Enabling profiling and extra run-time error checking. 160* Preprocessor Options:: Controlling header files and macro definitions. 161 Also, getting dependency information for Make. 162* Assembler Options:: Passing options to the assembler. 163* Link Options:: Specifying libraries and so on. 164* Directory Options:: Where to find header files and libraries. 165 Where to find the compiler executable files. 166* Code Gen Options:: Specifying conventions for function calls, data layout 167 and register usage. 168* Developer Options:: Printing GCC configuration info, statistics, and 169 debugging dumps. 170* Submodel Options:: Target-specific options, such as compiling for a 171 specific processor variant. 172* Spec Files:: How to pass switches to sub-processes. 173* Environment Variables:: Env vars that affect GCC. 174* Precompiled Headers:: Compiling a header once, and using it many times. 175* C++ Modules:: Experimental C++20 module system. 176@end menu 177 178@c man begin OPTIONS 179 180@node Option Summary 181@section Option Summary 182 183Here is a summary of all the options, grouped by type. Explanations are 184in the following sections. 185 186@table @emph 187@item Overall Options 188@xref{Overall Options,,Options Controlling the Kind of Output}. 189@gccoptlist{-c -S -E -o @var{file} @gol 190-dumpbase @var{dumpbase} -dumpbase-ext @var{auxdropsuf} @gol 191-dumpdir @var{dumppfx} -x @var{language} @gol 192-v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help --version @gol 193-pass-exit-codes -pipe -specs=@var{file} -wrapper @gol 194@@@var{file} -ffile-prefix-map=@var{old}=@var{new} @gol 195-fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol 196-fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}} 197 198@item C Language Options 199@xref{C Dialect Options,,Options Controlling C Dialect}. 200@gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol 201-fpermitted-flt-eval-methods=@var{standard} @gol 202-aux-info @var{filename} -fallow-parameterless-variadic-functions @gol 203-fno-asm -fno-builtin -fno-builtin-@var{function} -fgimple@gol 204-fhosted -ffreestanding @gol 205-fopenacc -fopenacc-dim=@var{geom} @gol 206-fopenmp -fopenmp-simd @gol 207-fms-extensions -fplan9-extensions -fsso-struct=@var{endianness} @gol 208-fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol 209-fsigned-bitfields -fsigned-char @gol 210-funsigned-bitfields -funsigned-char} 211 212@item C++ Language Options 213@xref{C++ Dialect Options,,Options Controlling C++ Dialect}. 214@gccoptlist{-fabi-version=@var{n} -fno-access-control @gol 215-faligned-new=@var{n} -fargs-in-order=@var{n} -fchar8_t -fcheck-new @gol 216-fconstexpr-depth=@var{n} -fconstexpr-cache-depth=@var{n} @gol 217-fconstexpr-loop-limit=@var{n} -fconstexpr-ops-limit=@var{n} @gol 218-fno-elide-constructors @gol 219-fno-enforce-eh-specs @gol 220-fno-gnu-keywords @gol 221-fno-implicit-templates @gol 222-fno-implicit-inline-templates @gol 223-fno-implement-inlines @gol 224-fmodule-header@r{[}=@var{kind}@r{]} -fmodule-only -fmodules-ts @gol 225-fmodule-implicit-inline @gol 226-fno-module-lazy @gol 227-fmodule-mapper=@var{specification} @gol 228-fmodule-version-ignore @gol 229-fms-extensions @gol 230-fnew-inheriting-ctors @gol 231-fnew-ttp-matching @gol 232-fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol 233-fno-optional-diags -fpermissive @gol 234-fno-pretty-templates @gol 235-fno-rtti -fsized-deallocation @gol 236-ftemplate-backtrace-limit=@var{n} @gol 237-ftemplate-depth=@var{n} @gol 238-fno-threadsafe-statics -fuse-cxa-atexit @gol 239-fno-weak -nostdinc++ @gol 240-fvisibility-inlines-hidden @gol 241-fvisibility-ms-compat @gol 242-fext-numeric-literals @gol 243-flang-info-include-translate@r{[}=@var{header}@r{]} @gol 244-flang-info-include-translate-not @gol 245-flang-info-module-cmi@r{[}=@var{module}@r{]} @gol 246-stdlib=@var{libstdc++,libc++} @gol 247-Wabi-tag -Wcatch-value -Wcatch-value=@var{n} @gol 248-Wno-class-conversion -Wclass-memaccess @gol 249-Wcomma-subscript -Wconditionally-supported @gol 250-Wno-conversion-null -Wctad-maybe-unsupported @gol 251-Wctor-dtor-privacy -Wno-delete-incomplete @gol 252-Wdelete-non-virtual-dtor -Wdeprecated-copy -Wdeprecated-copy-dtor @gol 253-Wno-deprecated-enum-enum-conversion -Wno-deprecated-enum-float-conversion @gol 254-Weffc++ -Wno-exceptions -Wextra-semi -Wno-inaccessible-base @gol 255-Wno-inherited-variadic-ctor -Wno-init-list-lifetime @gol 256-Winvalid-imported-macros @gol 257-Wno-invalid-offsetof -Wno-literal-suffix @gol 258-Wno-mismatched-new-delete -Wmismatched-tags @gol 259-Wmultiple-inheritance -Wnamespaces -Wnarrowing @gol 260-Wnoexcept -Wnoexcept-type -Wnon-virtual-dtor @gol 261-Wpessimizing-move -Wno-placement-new -Wplacement-new=@var{n} @gol 262-Wrange-loop-construct -Wredundant-move -Wredundant-tags @gol 263-Wreorder -Wregister @gol 264-Wstrict-null-sentinel -Wno-subobject-linkage -Wtemplates @gol 265-Wno-non-template-friend -Wold-style-cast @gol 266-Woverloaded-virtual -Wno-pmf-conversions -Wsign-promo @gol 267-Wsized-deallocation -Wsuggest-final-methods @gol 268-Wsuggest-final-types -Wsuggest-override @gol 269-Wno-terminate -Wuseless-cast -Wno-vexing-parse @gol 270-Wvirtual-inheritance @gol 271-Wno-virtual-move-assign -Wvolatile -Wzero-as-null-pointer-constant} 272 273@item Objective-C and Objective-C++ Language Options 274@xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling 275Objective-C and Objective-C++ Dialects}. 276@gccoptlist{-fconstant-string-class=@var{class-name} @gol 277-fgnu-runtime -fnext-runtime @gol 278-fno-nil-receivers @gol 279-fobjc-abi-version=@var{n} @gol 280-fobjc-call-cxx-cdtors @gol 281-fobjc-direct-dispatch @gol 282-fobjc-exceptions @gol 283-fobjc-gc @gol 284-fobjc-nilcheck @gol 285-fobjc-std=objc1 @gol 286-fno-local-ivars @gol 287-fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol 288-freplace-objc-classes @gol 289-fzero-link @gol 290-gen-decls @gol 291-Wassign-intercept -Wno-property-assign-default @gol 292-Wno-protocol -Wobjc-root-class -Wselector @gol 293-Wstrict-selector-match @gol 294-Wundeclared-selector} 295 296@item Diagnostic Message Formatting Options 297@xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}. 298@gccoptlist{-fmessage-length=@var{n} @gol 299-fdiagnostics-plain-output @gol 300-fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol 301-fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol 302-fdiagnostics-urls=@r{[}auto@r{|}never@r{|}always@r{]} @gol 303-fdiagnostics-format=@r{[}text@r{|}json@r{]} @gol 304-fno-diagnostics-show-option -fno-diagnostics-show-caret @gol 305-fno-diagnostics-show-labels -fno-diagnostics-show-line-numbers @gol 306-fno-diagnostics-show-cwe @gol 307-fdiagnostics-minimum-margin-width=@var{width} @gol 308-fdiagnostics-parseable-fixits -fdiagnostics-generate-patch @gol 309-fdiagnostics-show-template-tree -fno-elide-type @gol 310-fdiagnostics-path-format=@r{[}none@r{|}separate-events@r{|}inline-events@r{]} @gol 311-fdiagnostics-show-path-depths @gol 312-fno-show-column @gol 313-fdiagnostics-column-unit=@r{[}display@r{|}byte@r{]} @gol 314-fdiagnostics-column-origin=@var{origin}} 315 316@item Warning Options 317@xref{Warning Options,,Options to Request or Suppress Warnings}. 318@gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol 319-pedantic-errors @gol 320-w -Wextra -Wall -Wabi=@var{n} @gol 321-Waddress -Wno-address-of-packed-member -Waggregate-return @gol 322-Walloc-size-larger-than=@var{byte-size} -Walloc-zero @gol 323-Walloca -Walloca-larger-than=@var{byte-size} @gol 324-Wno-aggressive-loop-optimizations @gol 325-Warith-conversion @gol 326-Warray-bounds -Warray-bounds=@var{n} @gol 327-Wno-attributes -Wattribute-alias=@var{n} -Wno-attribute-alias @gol 328-Wno-attribute-warning -Wbool-compare -Wbool-operation @gol 329-Wno-builtin-declaration-mismatch @gol 330-Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol 331-Wc11-c2x-compat @gol 332-Wc++-compat -Wc++11-compat -Wc++14-compat -Wc++17-compat @gol 333-Wc++20-compat @gol 334-Wcast-align -Wcast-align=strict -Wcast-function-type -Wcast-qual @gol 335-Wchar-subscripts @gol 336-Wclobbered -Wcomment @gol 337-Wconversion -Wno-coverage-mismatch -Wno-cpp @gol 338-Wdangling-else -Wdate-time @gol 339-Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol 340-Wdisabled-optimization @gol 341-Wno-discarded-array-qualifiers -Wno-discarded-qualifiers @gol 342-Wno-div-by-zero -Wdouble-promotion @gol 343-Wduplicated-branches -Wduplicated-cond @gol 344-Wempty-body -Wno-endif-labels -Wenum-compare -Wenum-conversion @gol 345-Werror -Werror=* -Wexpansion-to-defined -Wfatal-errors @gol 346-Wfloat-conversion -Wfloat-equal -Wformat -Wformat=2 @gol 347-Wno-format-contains-nul -Wno-format-extra-args @gol 348-Wformat-nonliteral -Wformat-overflow=@var{n} @gol 349-Wformat-security -Wformat-signedness -Wformat-truncation=@var{n} @gol 350-Wformat-y2k -Wframe-address @gol 351-Wframe-larger-than=@var{byte-size} -Wno-free-nonheap-object @gol 352-Wno-if-not-aligned -Wno-ignored-attributes @gol 353-Wignored-qualifiers -Wno-incompatible-pointer-types @gol 354-Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol 355-Wno-implicit-function-declaration -Wno-implicit-int @gol 356-Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol 357-Wno-int-to-pointer-cast -Wno-invalid-memory-model @gol 358-Winvalid-pch -Wjump-misses-init -Wlarger-than=@var{byte-size} @gol 359-Wlogical-not-parentheses -Wlogical-op -Wlong-long @gol 360-Wno-lto-type-mismatch -Wmain -Wmaybe-uninitialized @gol 361-Wmemset-elt-size -Wmemset-transposed-args @gol 362-Wmisleading-indentation -Wmissing-attributes -Wmissing-braces @gol 363-Wmissing-field-initializers -Wmissing-format-attribute @gol 364-Wmissing-include-dirs -Wmissing-noreturn -Wno-missing-profile @gol 365-Wno-multichar -Wmultistatement-macros -Wnonnull -Wnonnull-compare @gol 366-Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol 367-Wnull-dereference -Wno-odr -Wopenmp-simd @gol 368-Wno-overflow -Woverlength-strings -Wno-override-init-side-effects @gol 369-Wpacked -Wno-packed-bitfield-compat -Wpacked-not-aligned -Wpadded @gol 370-Wparentheses -Wno-pedantic-ms-format @gol 371-Wpointer-arith -Wno-pointer-compare -Wno-pointer-to-int-cast @gol 372-Wno-pragmas -Wno-prio-ctor-dtor -Wredundant-decls @gol 373-Wrestrict -Wno-return-local-addr -Wreturn-type @gol 374-Wno-scalar-storage-order -Wsequence-point @gol 375-Wshadow -Wshadow=global -Wshadow=local -Wshadow=compatible-local @gol 376-Wno-shadow-ivar @gol 377-Wno-shift-count-negative -Wno-shift-count-overflow -Wshift-negative-value @gol 378-Wno-shift-overflow -Wshift-overflow=@var{n} @gol 379-Wsign-compare -Wsign-conversion @gol 380-Wno-sizeof-array-argument @gol 381-Wsizeof-array-div @gol 382-Wsizeof-pointer-div -Wsizeof-pointer-memaccess @gol 383-Wstack-protector -Wstack-usage=@var{byte-size} -Wstrict-aliasing @gol 384-Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol 385-Wstring-compare @gol 386-Wno-stringop-overflow -Wno-stringop-overread @gol 387-Wno-stringop-truncation @gol 388-Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}malloc@r{]} @gol 389-Wswitch -Wno-switch-bool -Wswitch-default -Wswitch-enum @gol 390-Wno-switch-outside-range -Wno-switch-unreachable -Wsync-nand @gol 391-Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol 392-Wtsan -Wtype-limits -Wundef @gol 393-Wuninitialized -Wunknown-pragmas @gol 394-Wunsuffixed-float-constants -Wunused @gol 395-Wunused-but-set-parameter -Wunused-but-set-variable @gol 396-Wunused-const-variable -Wunused-const-variable=@var{n} @gol 397-Wunused-function -Wunused-label -Wunused-local-typedefs @gol 398-Wunused-macros @gol 399-Wunused-parameter -Wno-unused-result @gol 400-Wunused-value -Wunused-variable @gol 401-Wno-varargs -Wvariadic-macros @gol 402-Wvector-operation-performance @gol 403-Wvla -Wvla-larger-than=@var{byte-size} -Wno-vla-larger-than @gol 404-Wvolatile-register-var -Wwrite-strings @gol 405-Wzero-length-bounds} 406 407@item Static Analyzer Options 408@gccoptlist{ 409-fanalyzer @gol 410-fanalyzer-call-summaries @gol 411-fanalyzer-checker=@var{name} @gol 412-fno-analyzer-feasibility @gol 413-fanalyzer-fine-grained @gol 414-fanalyzer-state-merge @gol 415-fanalyzer-state-purge @gol 416-fanalyzer-transitivity @gol 417-fanalyzer-verbose-edges @gol 418-fanalyzer-verbose-state-changes @gol 419-fanalyzer-verbosity=@var{level} @gol 420-fdump-analyzer @gol 421-fdump-analyzer-stderr @gol 422-fdump-analyzer-callgraph @gol 423-fdump-analyzer-exploded-graph @gol 424-fdump-analyzer-exploded-nodes @gol 425-fdump-analyzer-exploded-nodes-2 @gol 426-fdump-analyzer-exploded-nodes-3 @gol 427-fdump-analyzer-feasibility @gol 428-fdump-analyzer-json @gol 429-fdump-analyzer-state-purge @gol 430-fdump-analyzer-supergraph @gol 431-Wno-analyzer-double-fclose @gol 432-Wno-analyzer-double-free @gol 433-Wno-analyzer-exposure-through-output-file @gol 434-Wno-analyzer-file-leak @gol 435-Wno-analyzer-free-of-non-heap @gol 436-Wno-analyzer-malloc-leak @gol 437-Wno-analyzer-mismatching-deallocation @gol 438-Wno-analyzer-null-argument @gol 439-Wno-analyzer-null-dereference @gol 440-Wno-analyzer-possible-null-argument @gol 441-Wno-analyzer-possible-null-dereference @gol 442-Wno-analyzer-shift-count-negative @gol 443-Wno-analyzer-shift-count-overflow @gol 444-Wno-analyzer-stale-setjmp-buffer @gol 445-Wno-analyzer-tainted-array-index @gol 446-Wanalyzer-too-complex @gol 447-Wno-analyzer-unsafe-call-within-signal-handler @gol 448-Wno-analyzer-use-after-free @gol 449-Wno-analyzer-use-of-pointer-in-stale-stack-frame @gol 450-Wno-analyzer-use-of-uninitialized-value @gol 451-Wno-analyzer-write-to-const @gol 452-Wno-analyzer-write-to-string-literal @gol 453} 454 455@item C and Objective-C-only Warning Options 456@gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol 457-Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol 458-Wold-style-declaration -Wold-style-definition @gol 459-Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol 460-Wdeclaration-after-statement -Wpointer-sign} 461 462@item Debugging Options 463@xref{Debugging Options,,Options for Debugging Your Program}. 464@gccoptlist{-g -g@var{level} -gdwarf -gdwarf-@var{version} @gol 465-ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol 466-gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol 467-gas-loc-support -gno-as-loc-support @gol 468-gas-locview-support -gno-as-locview-support @gol 469-gcolumn-info -gno-column-info -gdwarf32 -gdwarf64 @gol 470-gstatement-frontiers -gno-statement-frontiers @gol 471-gvariable-location-views -gno-variable-location-views @gol 472-ginternal-reset-location-views -gno-internal-reset-location-views @gol 473-ginline-points -gno-inline-points @gol 474-gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol 475-gsplit-dwarf -gdescribe-dies -gno-describe-dies @gol 476-fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol 477-fno-eliminate-unused-debug-types @gol 478-femit-struct-debug-baseonly -femit-struct-debug-reduced @gol 479-femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol 480-fno-eliminate-unused-debug-symbols -femit-class-debug-always @gol 481-fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol 482-fvar-tracking -fvar-tracking-assignments} 483 484@item Optimization Options 485@xref{Optimize Options,,Options that Control Optimization}. 486@gccoptlist{-faggressive-loop-optimizations @gol 487-falign-functions[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol 488-falign-jumps[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol 489-falign-labels[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol 490-falign-loops[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol 491-fno-allocation-dce -fallow-store-data-races @gol 492-fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol 493-fauto-inc-dec -fbranch-probabilities @gol 494-fcaller-saves @gol 495-fcombine-stack-adjustments -fconserve-stack @gol 496-fcompare-elim -fcprop-registers -fcrossjumping @gol 497-fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol 498-fcx-limited-range @gol 499-fdata-sections -fdce -fdelayed-branch @gol 500-fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol 501-fdevirtualize-at-ltrans -fdse @gol 502-fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol 503-ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol 504-ffinite-loops @gol 505-fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol 506-fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol 507-fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol 508-fif-conversion2 -findirect-inlining @gol 509-finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol 510-finline-small-functions -fipa-modref -fipa-cp -fipa-cp-clone @gol 511-fipa-bit-cp -fipa-vrp -fipa-pta -fipa-profile -fipa-pure-const @gol 512-fipa-reference -fipa-reference-addressable @gol 513-fipa-stack-alignment -fipa-icf -fira-algorithm=@var{algorithm} @gol 514-flive-patching=@var{level} @gol 515-fira-region=@var{region} -fira-hoist-pressure @gol 516-fira-loop-pressure -fno-ira-share-save-slots @gol 517-fno-ira-share-spill-slots @gol 518-fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol 519-fivopts -fkeep-inline-functions -fkeep-static-functions @gol 520-fkeep-static-consts -flimit-function-alignment -flive-range-shrinkage @gol 521-floop-block -floop-interchange -floop-strip-mine @gol 522-floop-unroll-and-jam -floop-nest-optimize @gol 523-floop-parallelize-all -flra-remat -flto -flto-compression-level @gol 524-flto-partition=@var{alg} -fmerge-all-constants @gol 525-fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol 526-fmove-loop-invariants -fno-branch-count-reg @gol 527-fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol 528-fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol 529-fno-peephole2 -fno-printf-return-value -fno-sched-interblock @gol 530-fno-sched-spec -fno-signed-zeros @gol 531-fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol 532-fomit-frame-pointer -foptimize-sibling-calls @gol 533-fpartial-inlining -fpeel-loops -fpredictive-commoning @gol 534-fprefetch-loop-arrays @gol 535-fprofile-correction @gol 536-fprofile-use -fprofile-use=@var{path} -fprofile-partial-training @gol 537-fprofile-values -fprofile-reorder-functions @gol 538-freciprocal-math -free -frename-registers -freorder-blocks @gol 539-freorder-blocks-algorithm=@var{algorithm} @gol 540-freorder-blocks-and-partition -freorder-functions @gol 541-frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol 542-frounding-math -fsave-optimization-record @gol 543-fsched2-use-superblocks -fsched-pressure @gol 544-fsched-spec-load -fsched-spec-load-dangerous @gol 545-fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol 546-fsched-group-heuristic -fsched-critical-path-heuristic @gol 547-fsched-spec-insn-heuristic -fsched-rank-heuristic @gol 548-fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol 549-fschedule-fusion @gol 550-fschedule-insns -fschedule-insns2 -fsection-anchors @gol 551-fselective-scheduling -fselective-scheduling2 @gol 552-fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol 553-fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol 554-fsignaling-nans @gol 555-fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol 556-fsplit-paths @gol 557-fsplit-wide-types -fsplit-wide-types-early -fssa-backprop -fssa-phiopt @gol 558-fstdarg-opt -fstore-merging -fstrict-aliasing @gol 559-fthread-jumps -ftracer -ftree-bit-ccp @gol 560-ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol 561-ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol 562-ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting @gol 563-ftree-loop-if-convert -ftree-loop-im @gol 564-ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol 565-ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol 566-ftree-loop-vectorize @gol 567-ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol 568-ftree-reassoc -ftree-scev-cprop -ftree-sink -ftree-slsr -ftree-sra @gol 569-ftree-switch-conversion -ftree-tail-merge @gol 570-ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol 571-funit-at-a-time -funroll-all-loops -funroll-loops @gol 572-funsafe-math-optimizations -funswitch-loops @gol 573-fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol 574-fweb -fwhole-program -fwpa -fuse-linker-plugin -fzero-call-used-regs @gol 575--param @var{name}=@var{value} 576-O -O0 -O1 -O2 -O3 -Os -Ofast -Og} 577 578@item Program Instrumentation Options 579@xref{Instrumentation Options,,Program Instrumentation Options}. 580@gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol 581-fprofile-abs-path @gol 582-fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol 583-fprofile-info-section -fprofile-info-section=@var{name} @gol 584-fprofile-note=@var{path} -fprofile-prefix-path=@var{path} @gol 585-fprofile-update=@var{method} -fprofile-filter-files=@var{regex} @gol 586-fprofile-exclude-files=@var{regex} @gol 587-fprofile-reproducible=@r{[}multithreaded@r{|}parallel-runs@r{|}serial@r{]} @gol 588-fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol 589-fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol 590-fsanitize-undefined-trap-on-error -fbounds-check @gol 591-fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{|}check@r{]} @gol 592-fstack-protector -fstack-protector-all -fstack-protector-strong @gol 593-fstack-protector-explicit -fstack-check @gol 594-fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol 595-fno-stack-limit -fsplit-stack @gol 596-fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol 597-fvtv-counts -fvtv-debug @gol 598-finstrument-functions @gol 599-finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol 600-finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}} 601 602@item Preprocessor Options 603@xref{Preprocessor Options,,Options Controlling the Preprocessor}. 604@gccoptlist{-A@var{question}=@var{answer} @gol 605-A-@var{question}@r{[}=@var{answer}@r{]} @gol 606-C -CC -D@var{macro}@r{[}=@var{defn}@r{]} @gol 607-dD -dI -dM -dN -dU @gol 608-fdebug-cpp -fdirectives-only -fdollars-in-identifiers @gol 609-fexec-charset=@var{charset} -fextended-identifiers @gol 610-finput-charset=@var{charset} -flarge-source-files @gol 611-fmacro-prefix-map=@var{old}=@var{new} -fmax-include-depth=@var{depth} @gol 612-fno-canonical-system-headers -fpch-deps -fpch-preprocess @gol 613-fpreprocessed -ftabstop=@var{width} -ftrack-macro-expansion @gol 614-fwide-exec-charset=@var{charset} -fworking-directory @gol 615-H -imacros @var{file} -include @var{file} @gol 616-M -MD -MF -MG -MM -MMD -MP -MQ -MT -Mno-modules @gol 617-no-integrated-cpp -P -pthread -remap @gol 618-traditional -traditional-cpp -trigraphs @gol 619-U@var{macro} -undef @gol 620-Wp,@var{option} -Xpreprocessor @var{option}} 621 622@item Assembler Options 623@xref{Assembler Options,,Passing Options to the Assembler}. 624@gccoptlist{-Wa,@var{option} -Xassembler @var{option}} 625 626@item Linker Options 627@xref{Link Options,,Options for Linking}. 628@gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol 629-nostartfiles -nodefaultlibs -nolibc -nostdlib @gol 630-e @var{entry} --entry=@var{entry} @gol 631-pie -pthread -r -rdynamic @gol 632-s -static -static-pie -static-libgcc -static-libstdc++ @gol 633-static-libasan -static-libtsan -static-liblsan -static-libubsan @gol 634-shared -shared-libgcc -symbolic @gol 635-T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol 636-u @var{symbol} -z @var{keyword}} 637 638@item Directory Options 639@xref{Directory Options,,Options for Directory Search}. 640@gccoptlist{-B@var{prefix} -I@var{dir} -I- @gol 641-idirafter @var{dir} @gol 642-imacros @var{file} -imultilib @var{dir} @gol 643-iplugindir=@var{dir} -iprefix @var{file} @gol 644-iquote @var{dir} -isysroot @var{dir} -isystem @var{dir} @gol 645-iwithprefix @var{dir} -iwithprefixbefore @var{dir} @gol 646-L@var{dir} -no-canonical-prefixes --no-sysroot-suffix @gol 647-nostdinc -nostdinc++ --sysroot=@var{dir}} 648 649@item Code Generation Options 650@xref{Code Gen Options,,Options for Code Generation Conventions}. 651@gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol 652-ffixed-@var{reg} -fexceptions @gol 653-fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol 654-fasynchronous-unwind-tables @gol 655-fno-gnu-unique @gol 656-finhibit-size-directive -fcommon -fno-ident @gol 657-fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol 658-fno-jump-tables -fno-bit-tests @gol 659-frecord-gcc-switches @gol 660-freg-struct-return -fshort-enums -fshort-wchar @gol 661-fverbose-asm -fpack-struct[=@var{n}] @gol 662-fleading-underscore -ftls-model=@var{model} @gol 663-fstack-reuse=@var{reuse_level} @gol 664-ftrampolines -ftrapv -fwrapv @gol 665-fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol 666-fstrict-volatile-bitfields -fsync-libcalls} 667 668@item Developer Options 669@xref{Developer Options,,GCC Developer Options}. 670@gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol 671-dumpfullversion -fcallgraph-info@r{[}=su,da@r{]} 672-fchecking -fchecking=@var{n} 673-fdbg-cnt-list @gol -fdbg-cnt=@var{counter-value-list} @gol 674-fdisable-ipa-@var{pass_name} @gol 675-fdisable-rtl-@var{pass_name} @gol 676-fdisable-rtl-@var{pass-name}=@var{range-list} @gol 677-fdisable-tree-@var{pass_name} @gol 678-fdisable-tree-@var{pass-name}=@var{range-list} @gol 679-fdump-debug -fdump-earlydebug @gol 680-fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol 681-fdump-final-insns@r{[}=@var{file}@r{]} @gol 682-fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol 683-fdump-lang-all @gol 684-fdump-lang-@var{switch} @gol 685-fdump-lang-@var{switch}-@var{options} @gol 686-fdump-lang-@var{switch}-@var{options}=@var{filename} @gol 687-fdump-passes @gol 688-fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol 689-fdump-statistics @gol 690-fdump-tree-all @gol 691-fdump-tree-@var{switch} @gol 692-fdump-tree-@var{switch}-@var{options} @gol 693-fdump-tree-@var{switch}-@var{options}=@var{filename} @gol 694-fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol 695-fenable-@var{kind}-@var{pass} @gol 696-fenable-@var{kind}-@var{pass}=@var{range-list} @gol 697-fira-verbose=@var{n} @gol 698-flto-report -flto-report-wpa -fmem-report-wpa @gol 699-fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol 700-fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol 701-fprofile-report @gol 702-frandom-seed=@var{string} -fsched-verbose=@var{n} @gol 703-fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol 704-fstats -fstack-usage -ftime-report -ftime-report-details @gol 705-fvar-tracking-assignments-toggle -gtoggle @gol 706-print-file-name=@var{library} -print-libgcc-file-name @gol 707-print-multi-directory -print-multi-lib -print-multi-os-directory @gol 708-print-prog-name=@var{program} -print-search-dirs -Q @gol 709-print-sysroot -print-sysroot-headers-suffix @gol 710-save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}} 711 712@item Machine-Dependent Options 713@xref{Submodel Options,,Machine-Dependent Options}. 714@c This list is ordered alphanumerically by subsection name. 715@c Try and put the significant identifier (CPU or system) first, 716@c so users have a clue at guessing where the ones they want will be. 717 718@emph{AArch64 Options} 719@gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol 720-mgeneral-regs-only @gol 721-mcmodel=tiny -mcmodel=small -mcmodel=large @gol 722-mstrict-align -mno-strict-align @gol 723-momit-leaf-frame-pointer @gol 724-mtls-dialect=desc -mtls-dialect=traditional @gol 725-mtls-size=@var{size} @gol 726-mfix-cortex-a53-835769 -mfix-cortex-a53-843419 @gol 727-mlow-precision-recip-sqrt -mlow-precision-sqrt -mlow-precision-div @gol 728-mpc-relative-literal-loads @gol 729-msign-return-address=@var{scope} @gol 730-mbranch-protection=@var{none}|@var{standard}|@var{pac-ret}[+@var{leaf} 731+@var{b-key}]|@var{bti} @gol 732-mharden-sls=@var{opts} @gol 733-march=@var{name} -mcpu=@var{name} -mtune=@var{name} @gol 734-moverride=@var{string} -mverbose-cost-dump @gol 735-mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{sysreg} @gol 736-mstack-protector-guard-offset=@var{offset} -mtrack-speculation @gol 737-moutline-atomics } 738 739@emph{Adapteva Epiphany Options} 740@gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol 741-mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol 742-msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol 743-mround-nearest -mlong-calls -mshort-calls -msmall16 @gol 744-mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol 745-msplit-vecmove-early -m1reg-@var{reg}} 746 747@emph{AMD GCN Options} 748@gccoptlist{-march=@var{gpu} -mtune=@var{gpu} -mstack-size=@var{bytes}} 749 750@emph{ARC Options} 751@gccoptlist{-mbarrel-shifter -mjli-always @gol 752-mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol 753-mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol 754-mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol 755-mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol 756-mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol 757-mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol 758-mlong-calls -mmedium-calls -msdata -mirq-ctrl-saved @gol 759-mrgf-banked-regs -mlpc-width=@var{width} -G @var{num} @gol 760-mvolatile-cache -mtp-regno=@var{regno} @gol 761-malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol 762-mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol 763-mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol 764-mlra-priority-compact mlra-priority-noncompact -mmillicode @gol 765-mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol 766-mtune=@var{cpu} -mmultcost=@var{num} -mcode-density-frame @gol 767-munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol 768-mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu} -mrf16 -mbranch-index} 769 770@emph{ARM Options} 771@gccoptlist{-mapcs-frame -mno-apcs-frame @gol 772-mabi=@var{name} @gol 773-mapcs-stack-check -mno-apcs-stack-check @gol 774-mapcs-reentrant -mno-apcs-reentrant @gol 775-mgeneral-regs-only @gol 776-msched-prolog -mno-sched-prolog @gol 777-mlittle-endian -mbig-endian @gol 778-mbe8 -mbe32 @gol 779-mfloat-abi=@var{name} @gol 780-mfp16-format=@var{name} 781-mthumb-interwork -mno-thumb-interwork @gol 782-mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol 783-mtune=@var{name} -mprint-tune-info @gol 784-mstructure-size-boundary=@var{n} @gol 785-mabort-on-noreturn @gol 786-mlong-calls -mno-long-calls @gol 787-msingle-pic-base -mno-single-pic-base @gol 788-mpic-register=@var{reg} @gol 789-mnop-fun-dllimport @gol 790-mpoke-function-name @gol 791-mthumb -marm -mflip-thumb @gol 792-mtpcs-frame -mtpcs-leaf-frame @gol 793-mcaller-super-interworking -mcallee-super-interworking @gol 794-mtp=@var{name} -mtls-dialect=@var{dialect} @gol 795-mword-relocations @gol 796-mfix-cortex-m3-ldrd @gol 797-munaligned-access @gol 798-mneon-for-64bits @gol 799-mslow-flash-data @gol 800-masm-syntax-unified @gol 801-mrestrict-it @gol 802-mverbose-cost-dump @gol 803-mpure-code @gol 804-mcmse @gol 805-mfix-cmse-cve-2021-35465 @gol 806-mfdpic} 807 808@emph{AVR Options} 809@gccoptlist{-mmcu=@var{mcu} -mabsdata -maccumulate-args @gol 810-mbranch-cost=@var{cost} @gol 811-mcall-prologues -mgas-isr-prologues -mint8 @gol 812-mdouble=@var{bits} -mlong-double=@var{bits} @gol 813-mn_flash=@var{size} -mno-interrupts @gol 814-mmain-is-OS_task -mrelax -mrmw -mstrict-X -mtiny-stack @gol 815-mfract-convert-truncate @gol 816-mshort-calls -nodevicelib -nodevicespecs @gol 817-Waddr-space-convert -Wmisspelled-isr} 818 819@emph{Blackfin Options} 820@gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol 821-msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol 822-mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol 823-mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol 824-mno-id-shared-library -mshared-library-id=@var{n} @gol 825-mleaf-id-shared-library -mno-leaf-id-shared-library @gol 826-msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol 827-mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol 828-micplb} 829 830@emph{C6X Options} 831@gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol 832-msim -msdata=@var{sdata-type}} 833 834@emph{CRIS Options} 835@gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol 836-mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol 837-metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol 838-mstack-align -mdata-align -mconst-align @gol 839-m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol 840-melf -maout -melinux -mlinux -sim -sim2 @gol 841-mmul-bug-workaround -mno-mul-bug-workaround} 842 843@emph{CR16 Options} 844@gccoptlist{-mmac @gol 845-mcr16cplus -mcr16c @gol 846-msim -mint32 -mbit-ops 847-mdata-model=@var{model}} 848 849@emph{C-SKY Options} 850@gccoptlist{-march=@var{arch} -mcpu=@var{cpu} @gol 851-mbig-endian -EB -mlittle-endian -EL @gol 852-mhard-float -msoft-float -mfpu=@var{fpu} -mdouble-float -mfdivdu @gol 853-mfloat-abi=@var{name} @gol 854-melrw -mistack -mmp -mcp -mcache -msecurity -mtrust @gol 855-mdsp -medsp -mvdsp @gol 856-mdiv -msmart -mhigh-registers -manchor @gol 857-mpushpop -mmultiple-stld -mconstpool -mstack-size -mccrt @gol 858-mbranch-cost=@var{n} -mcse-cc -msched-prolog -msim} 859 860@emph{Darwin Options} 861@gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol 862-arch_only -bind_at_load -bundle -bundle_loader @gol 863-client_name -compatibility_version -current_version @gol 864-dead_strip @gol 865-dependency-file -dylib_file -dylinker_install_name @gol 866-dynamic -dynamiclib -exported_symbols_list @gol 867-filelist -flat_namespace -force_cpusubtype_ALL @gol 868-force_flat_namespace -headerpad_max_install_names @gol 869-iframework @gol 870-image_base -init -install_name -keep_private_externs @gol 871-multi_module -multiply_defined -multiply_defined_unused @gol 872-noall_load -no_dead_strip_inits_and_terms @gol 873-nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol 874-pagezero_size -prebind -prebind_all_twolevel_modules @gol 875-private_bundle -read_only_relocs -sectalign @gol 876-sectobjectsymbols -whyload -seg1addr @gol 877-sectcreate -sectobjectsymbols -sectorder @gol 878-segaddr -segs_read_only_addr -segs_read_write_addr @gol 879-seg_addr_table -seg_addr_table_filename -seglinkedit @gol 880-segprot -segs_read_only_addr -segs_read_write_addr @gol 881-single_module -static -sub_library -sub_umbrella @gol 882-twolevel_namespace -umbrella -undefined @gol 883-unexported_symbols_list -weak_reference_mismatches @gol 884-whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol 885-mkernel -mone-byte-bool} 886 887@emph{DEC Alpha Options} 888@gccoptlist{-mno-fp-regs -msoft-float @gol 889-mieee -mieee-with-inexact -mieee-conformant @gol 890-mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol 891-mtrap-precision=@var{mode} -mbuild-constants @gol 892-mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol 893-mbwx -mmax -mfix -mcix @gol 894-mfloat-vax -mfloat-ieee @gol 895-mexplicit-relocs -msmall-data -mlarge-data @gol 896-msmall-text -mlarge-text @gol 897-mmemory-latency=@var{time}} 898 899@emph{eBPF Options} 900@gccoptlist{-mbig-endian -mlittle-endian -mkernel=@var{version} 901-mframe-limit=@var{bytes} -mxbpf} 902 903@emph{FR30 Options} 904@gccoptlist{-msmall-model -mno-lsim} 905 906@emph{FT32 Options} 907@gccoptlist{-msim -mlra -mnodiv -mft32b -mcompress -mnopm} 908 909@emph{FRV Options} 910@gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol 911-mhard-float -msoft-float @gol 912-malloc-cc -mfixed-cc -mdword -mno-dword @gol 913-mdouble -mno-double @gol 914-mmedia -mno-media -mmuladd -mno-muladd @gol 915-mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol 916-mlinked-fp -mlong-calls -malign-labels @gol 917-mlibrary-pic -macc-4 -macc-8 @gol 918-mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol 919-moptimize-membar -mno-optimize-membar @gol 920-mscc -mno-scc -mcond-exec -mno-cond-exec @gol 921-mvliw-branch -mno-vliw-branch @gol 922-mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol 923-mno-nested-cond-exec -mtomcat-stats @gol 924-mTLS -mtls @gol 925-mcpu=@var{cpu}} 926 927@emph{GNU/Linux Options} 928@gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol 929-tno-android-cc -tno-android-ld} 930 931@emph{H8/300 Options} 932@gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300} 933 934@emph{HPPA Options} 935@gccoptlist{-march=@var{architecture-type} @gol 936-mcaller-copies -mdisable-fpregs -mdisable-indexing @gol 937-mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol 938-mfixed-range=@var{register-range} @gol 939-mjump-in-delay -mlinker-opt -mlong-calls @gol 940-mlong-load-store -mno-disable-fpregs @gol 941-mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol 942-mno-jump-in-delay -mno-long-load-store @gol 943-mno-portable-runtime -mno-soft-float @gol 944-mno-space-regs -msoft-float -mpa-risc-1-0 @gol 945-mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol 946-mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol 947-munix=@var{unix-std} -nolibdld -static -threads} 948 949@emph{IA-64 Options} 950@gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol 951-mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol 952-mconstant-gp -mauto-pic -mfused-madd @gol 953-minline-float-divide-min-latency @gol 954-minline-float-divide-max-throughput @gol 955-mno-inline-float-divide @gol 956-minline-int-divide-min-latency @gol 957-minline-int-divide-max-throughput @gol 958-mno-inline-int-divide @gol 959-minline-sqrt-min-latency -minline-sqrt-max-throughput @gol 960-mno-inline-sqrt @gol 961-mdwarf2-asm -mearly-stop-bits @gol 962-mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol 963-mtune=@var{cpu-type} -milp32 -mlp64 @gol 964-msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol 965-msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol 966-msched-spec-ldc -msched-spec-control-ldc @gol 967-msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol 968-msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol 969-msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol 970-msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}} 971 972@emph{LM32 Options} 973@gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol 974-msign-extend-enabled -muser-enabled} 975 976@emph{M32R/D Options} 977@gccoptlist{-m32r2 -m32rx -m32r @gol 978-mdebug @gol 979-malign-loops -mno-align-loops @gol 980-missue-rate=@var{number} @gol 981-mbranch-cost=@var{number} @gol 982-mmodel=@var{code-size-model-type} @gol 983-msdata=@var{sdata-type} @gol 984-mno-flush-func -mflush-func=@var{name} @gol 985-mno-flush-trap -mflush-trap=@var{number} @gol 986-G @var{num}} 987 988@emph{M32C Options} 989@gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}} 990 991@emph{M680x0 Options} 992@gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol 993-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol 994-m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol 995-mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol 996-mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol 997-mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol 998-malign-int -mstrict-align -msep-data -mno-sep-data @gol 999-mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol 1000-mxgot -mno-xgot -mlong-jump-table-offsets} 1001 1002@emph{MCore Options} 1003@gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol 1004-mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol 1005-m4byte-functions -mno-4byte-functions -mcallgraph-data @gol 1006-mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol 1007-mlittle-endian -mbig-endian -m210 -m340 -mstack-increment} 1008 1009@emph{MeP Options} 1010@gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol 1011-mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol 1012-mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol 1013-mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol 1014-mtiny=@var{n}} 1015 1016@emph{MicroBlaze Options} 1017@gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol 1018-mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol 1019-mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol 1020-mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol 1021-mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model} @gol 1022-mpic-data-is-text-relative} 1023 1024@emph{MIPS Options} 1025@gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol 1026-mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol 1027-mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol 1028-mips16 -mno-mips16 -mflip-mips16 @gol 1029-minterlink-compressed -mno-interlink-compressed @gol 1030-minterlink-mips16 -mno-interlink-mips16 @gol 1031-mabi=@var{abi} -mabicalls -mno-abicalls @gol 1032-mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol 1033-mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol 1034-mno-float -msingle-float -mdouble-float @gol 1035-modd-spreg -mno-odd-spreg @gol 1036-mabs=@var{mode} -mnan=@var{encoding} @gol 1037-mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol 1038-mmcu -mmno-mcu @gol 1039-meva -mno-eva @gol 1040-mvirt -mno-virt @gol 1041-mxpa -mno-xpa @gol 1042-mcrc -mno-crc @gol 1043-mginv -mno-ginv @gol 1044-mmicromips -mno-micromips @gol 1045-mmsa -mno-msa @gol 1046-mloongson-mmi -mno-loongson-mmi @gol 1047-mloongson-ext -mno-loongson-ext @gol 1048-mloongson-ext2 -mno-loongson-ext2 @gol 1049-mfpu=@var{fpu-type} @gol 1050-msmartmips -mno-smartmips @gol 1051-mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol 1052-mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol 1053-mlong64 -mlong32 -msym32 -mno-sym32 @gol 1054-G@var{num} -mlocal-sdata -mno-local-sdata @gol 1055-mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol 1056-membedded-data -mno-embedded-data @gol 1057-muninit-const-in-rodata -mno-uninit-const-in-rodata @gol 1058-mcode-readable=@var{setting} @gol 1059-msplit-addresses -mno-split-addresses @gol 1060-mexplicit-relocs -mno-explicit-relocs @gol 1061-mcheck-zero-division -mno-check-zero-division @gol 1062-mdivide-traps -mdivide-breaks @gol 1063-mload-store-pairs -mno-load-store-pairs @gol 1064-mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol 1065-mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol 1066-mfix-24k -mno-fix-24k @gol 1067-mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol 1068-mfix-r5900 -mno-fix-r5900 @gol 1069-mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol 1070-mfix-vr4120 -mno-fix-vr4120 @gol 1071-mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol 1072-mflush-func=@var{func} -mno-flush-func @gol 1073-mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol 1074-mcompact-branches=@var{policy} @gol 1075-mfp-exceptions -mno-fp-exceptions @gol 1076-mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol 1077-mlxc1-sxc1 -mno-lxc1-sxc1 -mmadd4 -mno-madd4 @gol 1078-mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol 1079-mframe-header-opt -mno-frame-header-opt} 1080 1081@emph{MMIX Options} 1082@gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol 1083-mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol 1084-melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol 1085-mno-base-addresses -msingle-exit -mno-single-exit} 1086 1087@emph{MN10300 Options} 1088@gccoptlist{-mmult-bug -mno-mult-bug @gol 1089-mno-am33 -mam33 -mam33-2 -mam34 @gol 1090-mtune=@var{cpu-type} @gol 1091-mreturn-pointer-on-d0 @gol 1092-mno-crt0 -mrelax -mliw -msetlb} 1093 1094@emph{Moxie Options} 1095@gccoptlist{-meb -mel -mmul.x -mno-crt0} 1096 1097@emph{MSP430 Options} 1098@gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol 1099-mwarn-mcu @gol 1100-mcode-region= -mdata-region= @gol 1101-msilicon-errata= -msilicon-errata-warn= @gol 1102-mhwmult= -minrt -mtiny-printf -mmax-inline-shift=} 1103 1104@emph{NDS32 Options} 1105@gccoptlist{-mbig-endian -mlittle-endian @gol 1106-mreduced-regs -mfull-regs @gol 1107-mcmov -mno-cmov @gol 1108-mext-perf -mno-ext-perf @gol 1109-mext-perf2 -mno-ext-perf2 @gol 1110-mext-string -mno-ext-string @gol 1111-mv3push -mno-v3push @gol 1112-m16bit -mno-16bit @gol 1113-misr-vector-size=@var{num} @gol 1114-mcache-block-size=@var{num} @gol 1115-march=@var{arch} @gol 1116-mcmodel=@var{code-model} @gol 1117-mctor-dtor -mrelax} 1118 1119@emph{Nios II Options} 1120@gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol 1121-mgprel-sec=@var{regexp} -mr0rel-sec=@var{regexp} @gol 1122-mel -meb @gol 1123-mno-bypass-cache -mbypass-cache @gol 1124-mno-cache-volatile -mcache-volatile @gol 1125-mno-fast-sw-div -mfast-sw-div @gol 1126-mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol 1127-mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol 1128-mcustom-fpu-cfg=@var{name} @gol 1129-mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol 1130-march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx} 1131 1132@emph{Nvidia PTX Options} 1133@gccoptlist{-m64 -mmainkernel -moptimize} 1134 1135@emph{OpenRISC Options} 1136@gccoptlist{-mboard=@var{name} -mnewlib -mhard-mul -mhard-div @gol 1137-msoft-mul -msoft-div @gol 1138-msoft-float -mhard-float -mdouble-float -munordered-float @gol 1139-mcmov -mror -mrori -msext -msfimm -mshftimm} 1140 1141@emph{PDP-11 Options} 1142@gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol 1143-mint32 -mno-int16 -mint16 -mno-int32 @gol 1144-msplit -munix-asm -mdec-asm -mgnu-asm -mlra} 1145 1146@emph{picoChip Options} 1147@gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol 1148-msymbol-as-address -mno-inefficient-warnings} 1149 1150@emph{PowerPC Options} 1151See RS/6000 and PowerPC Options. 1152 1153@emph{PRU Options} 1154@gccoptlist{-mmcu=@var{mcu} -minrt -mno-relax -mloop @gol 1155-mabi=@var{variant} @gol} 1156 1157@emph{RISC-V Options} 1158@gccoptlist{-mbranch-cost=@var{N-instruction} @gol 1159-mplt -mno-plt @gol 1160-mabi=@var{ABI-string} @gol 1161-mfdiv -mno-fdiv @gol 1162-mdiv -mno-div @gol 1163-march=@var{ISA-string} @gol 1164-mtune=@var{processor-string} @gol 1165-mpreferred-stack-boundary=@var{num} @gol 1166-msmall-data-limit=@var{N-bytes} @gol 1167-msave-restore -mno-save-restore @gol 1168-mshorten-memrefs -mno-shorten-memrefs @gol 1169-mstrict-align -mno-strict-align @gol 1170-mcmodel=medlow -mcmodel=medany @gol 1171-mexplicit-relocs -mno-explicit-relocs @gol 1172-mrelax -mno-relax @gol 1173-mriscv-attribute -mmo-riscv-attribute @gol 1174-malign-data=@var{type} @gol 1175-mbig-endian -mlittle-endian @gol 1176+-mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol 1177+-mstack-protector-guard-offset=@var{offset}} 1178 1179@emph{RL78 Options} 1180@gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol 1181-mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol 1182-m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts} 1183 1184@emph{RS/6000 and PowerPC Options} 1185@gccoptlist{-mcpu=@var{cpu-type} @gol 1186-mtune=@var{cpu-type} @gol 1187-mcmodel=@var{code-model} @gol 1188-mpowerpc64 @gol 1189-maltivec -mno-altivec @gol 1190-mpowerpc-gpopt -mno-powerpc-gpopt @gol 1191-mpowerpc-gfxopt -mno-powerpc-gfxopt @gol 1192-mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol 1193-mfprnd -mno-fprnd @gol 1194-mcmpb -mno-cmpb -mhard-dfp -mno-hard-dfp @gol 1195-mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol 1196-m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol 1197-malign-power -malign-natural @gol 1198-msoft-float -mhard-float -mmultiple -mno-multiple @gol 1199-mupdate -mno-update @gol 1200-mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol 1201-mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol 1202-mstrict-align -mno-strict-align -mrelocatable @gol 1203-mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol 1204-mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol 1205-mdynamic-no-pic -mswdiv -msingle-pic-base @gol 1206-mprioritize-restricted-insns=@var{priority} @gol 1207-msched-costly-dep=@var{dependence_type} @gol 1208-minsert-sched-nops=@var{scheme} @gol 1209-mcall-aixdesc -mcall-eabi -mcall-freebsd @gol 1210-mcall-linux -mcall-netbsd -mcall-openbsd @gol 1211-mcall-sysv -mcall-sysv-eabi -mcall-sysv-noeabi @gol 1212-mtraceback=@var{traceback_type} @gol 1213-maix-struct-return -msvr4-struct-return @gol 1214-mabi=@var{abi-type} -msecure-plt -mbss-plt @gol 1215-mlongcall -mno-longcall -mpltseq -mno-pltseq @gol 1216-mblock-move-inline-limit=@var{num} @gol 1217-mblock-compare-inline-limit=@var{num} @gol 1218-mblock-compare-inline-loop-limit=@var{num} @gol 1219-mno-block-ops-unaligned-vsx @gol 1220-mstring-compare-inline-limit=@var{num} @gol 1221-misel -mno-isel @gol 1222-mvrsave -mno-vrsave @gol 1223-mmulhw -mno-mulhw @gol 1224-mdlmzb -mno-dlmzb @gol 1225-mprototype -mno-prototype @gol 1226-msim -mmvme -mads -myellowknife -memb -msdata @gol 1227-msdata=@var{opt} -mreadonly-in-sdata -mvxworks -G @var{num} @gol 1228-mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol 1229-mno-recip-precision @gol 1230-mveclibabi=@var{type} -mfriz -mno-friz @gol 1231-mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol 1232-msave-toc-indirect -mno-save-toc-indirect @gol 1233-mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol 1234-mcrypto -mno-crypto -mhtm -mno-htm @gol 1235-mquad-memory -mno-quad-memory @gol 1236-mquad-memory-atomic -mno-quad-memory-atomic @gol 1237-mcompat-align-parm -mno-compat-align-parm @gol 1238-mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol 1239-mgnu-attribute -mno-gnu-attribute @gol 1240-mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol 1241-mstack-protector-guard-offset=@var{offset} -mprefixed -mno-prefixed @gol 1242-mpcrel -mno-pcrel -mmma -mno-mmma -mrop-protect -mno-rop-protect @gol 1243-mprivileged -mno-privileged} 1244 1245@emph{RX Options} 1246@gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol 1247-mcpu=@gol 1248-mbig-endian-data -mlittle-endian-data @gol 1249-msmall-data @gol 1250-msim -mno-sim@gol 1251-mas100-syntax -mno-as100-syntax@gol 1252-mrelax@gol 1253-mmax-constant-size=@gol 1254-mint-register=@gol 1255-mpid@gol 1256-mallow-string-insns -mno-allow-string-insns@gol 1257-mjsr@gol 1258-mno-warn-multiple-fast-interrupts@gol 1259-msave-acc-in-interrupts} 1260 1261@emph{S/390 and zSeries Options} 1262@gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol 1263-mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol 1264-mlong-double-64 -mlong-double-128 @gol 1265-mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol 1266-msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol 1267-m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol 1268-mhtm -mvx -mzvector @gol 1269-mtpf-trace -mno-tpf-trace -mtpf-trace-skip -mno-tpf-trace-skip @gol 1270-mfused-madd -mno-fused-madd @gol 1271-mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol 1272-mhotpatch=@var{halfwords},@var{halfwords}} 1273 1274@emph{Score Options} 1275@gccoptlist{-meb -mel @gol 1276-mnhwloop @gol 1277-muls @gol 1278-mmac @gol 1279-mscore5 -mscore5u -mscore7 -mscore7d} 1280 1281@emph{SH Options} 1282@gccoptlist{-m1 -m2 -m2e @gol 1283-m2a-nofpu -m2a-single-only -m2a-single -m2a @gol 1284-m3 -m3e @gol 1285-m4-nofpu -m4-single-only -m4-single -m4 @gol 1286-m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol 1287-mb -ml -mdalign -mrelax @gol 1288-mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol 1289-mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol 1290-mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol 1291-mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol 1292-maccumulate-outgoing-args @gol 1293-matomic-model=@var{atomic-model} @gol 1294-mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol 1295-mcbranch-force-delay-slot @gol 1296-mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol 1297-mpretend-cmove -mtas} 1298 1299@emph{Solaris 2 Options} 1300@gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol 1301-pthreads} 1302 1303@emph{SPARC Options} 1304@gccoptlist{-mcpu=@var{cpu-type} @gol 1305-mtune=@var{cpu-type} @gol 1306-mcmodel=@var{code-model} @gol 1307-mmemory-model=@var{mem-model} @gol 1308-m32 -m64 -mapp-regs -mno-app-regs @gol 1309-mfaster-structs -mno-faster-structs -mflat -mno-flat @gol 1310-mfpu -mno-fpu -mhard-float -msoft-float @gol 1311-mhard-quad-float -msoft-quad-float @gol 1312-mstack-bias -mno-stack-bias @gol 1313-mstd-struct-return -mno-std-struct-return @gol 1314-munaligned-doubles -mno-unaligned-doubles @gol 1315-muser-mode -mno-user-mode @gol 1316-mv8plus -mno-v8plus -mvis -mno-vis @gol 1317-mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol 1318-mvis4 -mno-vis4 -mvis4b -mno-vis4b @gol 1319-mcbcond -mno-cbcond -mfmaf -mno-fmaf -mfsmuld -mno-fsmuld @gol 1320-mpopc -mno-popc -msubxc -mno-subxc @gol 1321-mfix-at697f -mfix-ut699 -mfix-ut700 -mfix-gr712rc @gol 1322-mlra -mno-lra} 1323 1324@emph{System V Options} 1325@gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}} 1326 1327@emph{TILE-Gx Options} 1328@gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol 1329-mcmodel=@var{code-model}} 1330 1331@emph{TILEPro Options} 1332@gccoptlist{-mcpu=@var{cpu} -m32} 1333 1334@emph{V850 Options} 1335@gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol 1336-mprolog-function -mno-prolog-function -mspace @gol 1337-mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol 1338-mapp-regs -mno-app-regs @gol 1339-mdisable-callt -mno-disable-callt @gol 1340-mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol 1341-mv850e -mv850 -mv850e3v5 @gol 1342-mloop @gol 1343-mrelax @gol 1344-mlong-jumps @gol 1345-msoft-float @gol 1346-mhard-float @gol 1347-mgcc-abi @gol 1348-mrh850-abi @gol 1349-mbig-switch} 1350 1351@emph{VAX Options} 1352@gccoptlist{-mg -mgnu -munix} 1353 1354@emph{Visium Options} 1355@gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol 1356-mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode} 1357 1358@emph{VMS Options} 1359@gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol 1360-mpointer-size=@var{size}} 1361 1362@emph{VxWorks Options} 1363@gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol 1364-Xbind-lazy -Xbind-now} 1365 1366@emph{x86 Options} 1367@gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol 1368-mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol 1369-mfpmath=@var{unit} @gol 1370-masm=@var{dialect} -mno-fancy-math-387 @gol 1371-mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol 1372-mno-wide-multiply -mrtd -malign-double @gol 1373-mpreferred-stack-boundary=@var{num} @gol 1374-mincoming-stack-boundary=@var{num} @gol 1375-mcld -mcx16 -msahf -mmovbe -mcrc32 -mmwait @gol 1376-mrecip -mrecip=@var{opt} @gol 1377-mvzeroupper -mprefer-avx128 -mprefer-vector-width=@var{opt} @gol 1378-mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol 1379-mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol 1380-mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol 1381-mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mpconfig -mwbnoinvd @gol 1382-mptwrite -mprefetchwt1 -mclflushopt -mclwb -mxsavec -mxsaves @gol 1383-msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol 1384-madx -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mhle -mlwp @gol 1385-mmwaitx -mclzero -mpku -mthreads -mgfni -mvaes -mwaitpkg @gol 1386-mshstk -mmanual-endbr -mforce-indirect-call -mavx512vbmi2 -mavx512bf16 -menqcmd @gol 1387-mvpclmulqdq -mavx512bitalg -mmovdiri -mmovdir64b -mavx512vpopcntdq @gol 1388-mavx5124fmaps -mavx512vnni -mavx5124vnniw -mprfchw -mrdpid @gol 1389-mrdseed -msgx -mavx512vp2intersect -mserialize -mtsxldtrk@gol 1390-mamx-tile -mamx-int8 -mamx-bf16 -muintr -mhreset -mavxvnni@gol 1391-mcldemote -mms-bitfields -mno-align-stringops -minline-all-stringops @gol 1392-minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol 1393-mkl -mwidekl @gol 1394-mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol 1395-mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol 1396-m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol 1397-mregparm=@var{num} -msseregparm @gol 1398-mveclibabi=@var{type} -mvect8-ret-in-mem @gol 1399-mpc32 -mpc64 -mpc80 -mstackrealign @gol 1400-momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol 1401-mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol 1402-m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol 1403-msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol 1404-minstrument-return=@var{type} -mfentry-name=@var{name} -mfentry-section=@var{name} @gol 1405-mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol 1406-malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol 1407-mstack-protector-guard-reg=@var{reg} @gol 1408-mstack-protector-guard-offset=@var{offset} @gol 1409-mstack-protector-guard-symbol=@var{symbol} @gol 1410-mgeneral-regs-only -mcall-ms2sysv-xlogues @gol 1411-mindirect-branch=@var{choice} -mfunction-return=@var{choice} @gol 1412-mindirect-branch-register -mneeded} 1413 1414@emph{x86 Windows Options} 1415@gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol 1416-mnop-fun-dllimport -mthread @gol 1417-municode -mwin32 -mwindows -fno-set-stack-executable} 1418 1419@emph{Xstormy16 Options} 1420@gccoptlist{-msim} 1421 1422@emph{Xtensa Options} 1423@gccoptlist{-mconst16 -mno-const16 @gol 1424-mfused-madd -mno-fused-madd @gol 1425-mforce-no-pic @gol 1426-mserialize-volatile -mno-serialize-volatile @gol 1427-mtext-section-literals -mno-text-section-literals @gol 1428-mauto-litpools -mno-auto-litpools @gol 1429-mtarget-align -mno-target-align @gol 1430-mlongcalls -mno-longcalls @gol 1431-mabi=@var{abi-type}} 1432 1433@emph{zSeries Options} 1434See S/390 and zSeries Options. 1435@end table 1436 1437 1438@node Overall Options 1439@section Options Controlling the Kind of Output 1440 1441Compilation can involve up to four stages: preprocessing, compilation 1442proper, assembly and linking, always in that order. GCC is capable of 1443preprocessing and compiling several files either into several 1444assembler input files, or into one assembler input file; then each 1445assembler input file produces an object file, and linking combines all 1446the object files (those newly compiled, and those specified as input) 1447into an executable file. 1448 1449@cindex file name suffix 1450For any given input file, the file name suffix determines what kind of 1451compilation is done: 1452 1453@table @gcctabopt 1454@item @var{file}.c 1455C source code that must be preprocessed. 1456 1457@item @var{file}.i 1458C source code that should not be preprocessed. 1459 1460@item @var{file}.ii 1461C++ source code that should not be preprocessed. 1462 1463@item @var{file}.m 1464Objective-C source code. Note that you must link with the @file{libobjc} 1465library to make an Objective-C program work. 1466 1467@item @var{file}.mi 1468Objective-C source code that should not be preprocessed. 1469 1470@item @var{file}.mm 1471@itemx @var{file}.M 1472Objective-C++ source code. Note that you must link with the @file{libobjc} 1473library to make an Objective-C++ program work. Note that @samp{.M} refers 1474to a literal capital M@. 1475 1476@item @var{file}.mii 1477Objective-C++ source code that should not be preprocessed. 1478 1479@item @var{file}.h 1480C, C++, Objective-C or Objective-C++ header file to be turned into a 1481precompiled header (default), or C, C++ header file to be turned into an 1482Ada spec (via the @option{-fdump-ada-spec} switch). 1483 1484@item @var{file}.cc 1485@itemx @var{file}.cp 1486@itemx @var{file}.cxx 1487@itemx @var{file}.cpp 1488@itemx @var{file}.CPP 1489@itemx @var{file}.c++ 1490@itemx @var{file}.C 1491C++ source code that must be preprocessed. Note that in @samp{.cxx}, 1492the last two letters must both be literally @samp{x}. Likewise, 1493@samp{.C} refers to a literal capital C@. 1494 1495@item @var{file}.mm 1496@itemx @var{file}.M 1497Objective-C++ source code that must be preprocessed. 1498 1499@item @var{file}.mii 1500Objective-C++ source code that should not be preprocessed. 1501 1502@item @var{file}.hh 1503@itemx @var{file}.H 1504@itemx @var{file}.hp 1505@itemx @var{file}.hxx 1506@itemx @var{file}.hpp 1507@itemx @var{file}.HPP 1508@itemx @var{file}.h++ 1509@itemx @var{file}.tcc 1510C++ header file to be turned into a precompiled header or Ada spec. 1511 1512@item @var{file}.f 1513@itemx @var{file}.for 1514@itemx @var{file}.ftn 1515Fixed form Fortran source code that should not be preprocessed. 1516 1517@item @var{file}.F 1518@itemx @var{file}.FOR 1519@itemx @var{file}.fpp 1520@itemx @var{file}.FPP 1521@itemx @var{file}.FTN 1522Fixed form Fortran source code that must be preprocessed (with the traditional 1523preprocessor). 1524 1525@item @var{file}.f90 1526@itemx @var{file}.f95 1527@itemx @var{file}.f03 1528@itemx @var{file}.f08 1529Free form Fortran source code that should not be preprocessed. 1530 1531@item @var{file}.F90 1532@itemx @var{file}.F95 1533@itemx @var{file}.F03 1534@itemx @var{file}.F08 1535Free form Fortran source code that must be preprocessed (with the 1536traditional preprocessor). 1537 1538@item @var{file}.go 1539Go source code. 1540 1541@item @var{file}.brig 1542BRIG files (binary representation of HSAIL). 1543 1544@item @var{file}.d 1545D source code. 1546 1547@item @var{file}.di 1548D interface file. 1549 1550@item @var{file}.dd 1551D documentation code (Ddoc). 1552 1553@item @var{file}.ads 1554Ada source code file that contains a library unit declaration (a 1555declaration of a package, subprogram, or generic, or a generic 1556instantiation), or a library unit renaming declaration (a package, 1557generic, or subprogram renaming declaration). Such files are also 1558called @dfn{specs}. 1559 1560@item @var{file}.adb 1561Ada source code file containing a library unit body (a subprogram or 1562package body). Such files are also called @dfn{bodies}. 1563 1564@c GCC also knows about some suffixes for languages not yet included: 1565@c Ratfor: 1566@c @var{file}.r 1567 1568@item @var{file}.s 1569Assembler code. 1570 1571@item @var{file}.S 1572@itemx @var{file}.sx 1573Assembler code that must be preprocessed. 1574 1575@item @var{other} 1576An object file to be fed straight into linking. 1577Any file name with no recognized suffix is treated this way. 1578@end table 1579 1580@opindex x 1581You can specify the input language explicitly with the @option{-x} option: 1582 1583@table @gcctabopt 1584@item -x @var{language} 1585Specify explicitly the @var{language} for the following input files 1586(rather than letting the compiler choose a default based on the file 1587name suffix). This option applies to all following input files until 1588the next @option{-x} option. Possible values for @var{language} are: 1589@smallexample 1590c c-header cpp-output 1591c++ c++-header c++-system-header c++-user-header c++-cpp-output 1592objective-c objective-c-header objective-c-cpp-output 1593objective-c++ objective-c++-header objective-c++-cpp-output 1594assembler assembler-with-cpp 1595ada 1596d 1597f77 f77-cpp-input f95 f95-cpp-input 1598go 1599brig 1600@end smallexample 1601 1602@item -x none 1603Turn off any specification of a language, so that subsequent files are 1604handled according to their file name suffixes (as they are if @option{-x} 1605has not been used at all). 1606@end table 1607 1608If you only want some of the stages of compilation, you can use 1609@option{-x} (or filename suffixes) to tell @command{gcc} where to start, and 1610one of the options @option{-c}, @option{-S}, or @option{-E} to say where 1611@command{gcc} is to stop. Note that some combinations (for example, 1612@samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all. 1613 1614@table @gcctabopt 1615@item -c 1616@opindex c 1617Compile or assemble the source files, but do not link. The linking 1618stage simply is not done. The ultimate output is in the form of an 1619object file for each source file. 1620 1621By default, the object file name for a source file is made by replacing 1622the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}. 1623 1624Unrecognized input files, not requiring compilation or assembly, are 1625ignored. 1626 1627@item -S 1628@opindex S 1629Stop after the stage of compilation proper; do not assemble. The output 1630is in the form of an assembler code file for each non-assembler input 1631file specified. 1632 1633By default, the assembler file name for a source file is made by 1634replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}. 1635 1636Input files that don't require compilation are ignored. 1637 1638@item -E 1639@opindex E 1640Stop after the preprocessing stage; do not run the compiler proper. The 1641output is in the form of preprocessed source code, which is sent to the 1642standard output. 1643 1644Input files that don't require preprocessing are ignored. 1645 1646@cindex output file option 1647@item -o @var{file} 1648@opindex o 1649Place the primary output in file @var{file}. This applies to whatever 1650sort of output is being produced, whether it be an executable file, an 1651object file, an assembler file or preprocessed C code. 1652 1653If @option{-o} is not specified, the default is to put an executable 1654file in @file{a.out}, the object file for 1655@file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its 1656assembler file in @file{@var{source}.s}, a precompiled header file in 1657@file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on 1658standard output. 1659 1660Though @option{-o} names only the primary output, it also affects the 1661naming of auxiliary and dump outputs. See the examples below. Unless 1662overridden, both auxiliary outputs and dump outputs are placed in the 1663same directory as the primary output. In auxiliary outputs, the suffix 1664of the input file is replaced with that of the auxiliary output file 1665type; in dump outputs, the suffix of the dump file is appended to the 1666input file suffix. In compilation commands, the base name of both 1667auxiliary and dump outputs is that of the primary output; in compile and 1668link commands, the primary output name, minus the executable suffix, is 1669combined with the input file name. If both share the same base name, 1670disregarding the suffix, the result of the combination is that base 1671name, otherwise, they are concatenated, separated by a dash. 1672 1673@smallexample 1674gcc -c foo.c ... 1675@end smallexample 1676 1677will use @file{foo.o} as the primary output, and place aux outputs and 1678dumps next to it, e.g., aux file @file{foo.dwo} for 1679@option{-gsplit-dwarf}, and dump file @file{foo.c.???r.final} for 1680@option{-fdump-rtl-final}. 1681 1682If a non-linker output file is explicitly specified, aux and dump files 1683by default take the same base name: 1684 1685@smallexample 1686gcc -c foo.c -o dir/foobar.o ... 1687@end smallexample 1688 1689will name aux outputs @file{dir/foobar.*} and dump outputs 1690@file{dir/foobar.c.*}. 1691 1692A linker output will instead prefix aux and dump outputs: 1693 1694@smallexample 1695gcc foo.c bar.c -o dir/foobar ... 1696@end smallexample 1697 1698will generally name aux outputs @file{dir/foobar-foo.*} and 1699@file{dir/foobar-bar.*}, and dump outputs @file{dir/foobar-foo.c.*} and 1700@file{dir/foobar-bar.c.*}. 1701 1702The one exception to the above is when the executable shares the base 1703name with the single input: 1704 1705@smallexample 1706gcc foo.c -o dir/foo ... 1707@end smallexample 1708 1709in which case aux outputs are named @file{dir/foo.*} and dump outputs 1710named @file{dir/foo.c.*}. 1711 1712The location and the names of auxiliary and dump outputs can be adjusted 1713by the options @option{-dumpbase}, @option{-dumpbase-ext}, 1714@option{-dumpdir}, @option{-save-temps=cwd}, and 1715@option{-save-temps=obj}. 1716 1717 1718@item -dumpbase @var{dumpbase} 1719@opindex dumpbase 1720This option sets the base name for auxiliary and dump output files. It 1721does not affect the name of the primary output file. Intermediate 1722outputs, when preserved, are not regarded as primary outputs, but as 1723auxiliary outputs: 1724 1725@smallexample 1726gcc -save-temps -S foo.c 1727@end smallexample 1728 1729saves the (no longer) temporary preprocessed file in @file{foo.i}, and 1730then compiles to the (implied) output file @file{foo.s}, whereas: 1731 1732@smallexample 1733gcc -save-temps -dumpbase save-foo -c foo.c 1734@end smallexample 1735 1736preprocesses to in @file{save-foo.i}, compiles to @file{save-foo.s} (now 1737an intermediate, thus auxiliary output), and then assembles to the 1738(implied) output file @file{foo.o}. 1739 1740Absent this option, dump and aux files take their names from the input 1741file, or from the (non-linker) output file, if one is explicitly 1742specified: dump output files (e.g. those requested by @option{-fdump-*} 1743options) with the input name suffix, and aux output files (those 1744requested by other non-dump options, e.g. @code{-save-temps}, 1745@code{-gsplit-dwarf}, @code{-fcallgraph-info}) without it. 1746 1747Similar suffix differentiation of dump and aux outputs can be attained 1748for explicitly-given @option{-dumpbase basename.suf} by also specifying 1749@option{-dumpbase-ext .suf}. 1750 1751If @var{dumpbase} is explicitly specified with any directory component, 1752any @var{dumppfx} specification (e.g. @option{-dumpdir} or 1753@option{-save-temps=*}) is ignored, and instead of appending to it, 1754@var{dumpbase} fully overrides it: 1755 1756@smallexample 1757gcc foo.c -c -o dir/foo.o -dumpbase alt/foo \ 1758 -dumpdir pfx- -save-temps=cwd ... 1759@end smallexample 1760 1761creates auxiliary and dump outputs named @file{alt/foo.*}, disregarding 1762@file{dir/} in @option{-o}, the @file{./} prefix implied by 1763@option{-save-temps=cwd}, and @file{pfx-} in @option{-dumpdir}. 1764 1765When @option{-dumpbase} is specified in a command that compiles multiple 1766inputs, or that compiles and then links, it may be combined with 1767@var{dumppfx}, as specified under @option{-dumpdir}. Then, each input 1768file is compiled using the combined @var{dumppfx}, and default values 1769for @var{dumpbase} and @var{auxdropsuf} are computed for each input 1770file: 1771 1772@smallexample 1773gcc foo.c bar.c -c -dumpbase main ... 1774@end smallexample 1775 1776creates @file{foo.o} and @file{bar.o} as primary outputs, and avoids 1777overwriting the auxiliary and dump outputs by using the @var{dumpbase} 1778as a prefix, creating auxiliary and dump outputs named @file{main-foo.*} 1779and @file{main-bar.*}. 1780 1781An empty string specified as @var{dumpbase} avoids the influence of the 1782output basename in the naming of auxiliary and dump outputs during 1783compilation, computing default values : 1784 1785@smallexample 1786gcc -c foo.c -o dir/foobar.o -dumpbase '' ... 1787@end smallexample 1788 1789will name aux outputs @file{dir/foo.*} and dump outputs 1790@file{dir/foo.c.*}. Note how their basenames are taken from the input 1791name, but the directory still defaults to that of the output. 1792 1793The empty-string dumpbase does not prevent the use of the output 1794basename for outputs during linking: 1795 1796@smallexample 1797gcc foo.c bar.c -o dir/foobar -dumpbase '' -flto ... 1798@end smallexample 1799 1800The compilation of the source files will name auxiliary outputs 1801@file{dir/foo.*} and @file{dir/bar.*}, and dump outputs 1802@file{dir/foo.c.*} and @file{dir/bar.c.*}. LTO recompilation during 1803linking will use @file{dir/foobar.} as the prefix for dumps and 1804auxiliary files. 1805 1806 1807@item -dumpbase-ext @var{auxdropsuf} 1808@opindex dumpbase-ext 1809When forming the name of an auxiliary (but not a dump) output file, drop 1810trailing @var{auxdropsuf} from @var{dumpbase} before appending any 1811suffixes. If not specified, this option defaults to the suffix of a 1812default @var{dumpbase}, i.e., the suffix of the input file when 1813@option{-dumpbase} is not present in the command line, or @var{dumpbase} 1814is combined with @var{dumppfx}. 1815 1816@smallexample 1817gcc foo.c -c -o dir/foo.o -dumpbase x-foo.c -dumpbase-ext .c ... 1818@end smallexample 1819 1820creates @file{dir/foo.o} as the main output, and generates auxiliary 1821outputs in @file{dir/x-foo.*}, taking the location of the primary 1822output, and dropping the @file{.c} suffix from the @var{dumpbase}. Dump 1823outputs retain the suffix: @file{dir/x-foo.c.*}. 1824 1825This option is disregarded if it does not match the suffix of a 1826specified @var{dumpbase}, except as an alternative to the executable 1827suffix when appending the linker output base name to @var{dumppfx}, as 1828specified below: 1829 1830@smallexample 1831gcc foo.c bar.c -o main.out -dumpbase-ext .out ... 1832@end smallexample 1833 1834creates @file{main.out} as the primary output, and avoids overwriting 1835the auxiliary and dump outputs by using the executable name minus 1836@var{auxdropsuf} as a prefix, creating auxiliary outputs named 1837@file{main-foo.*} and @file{main-bar.*} and dump outputs named 1838@file{main-foo.c.*} and @file{main-bar.c.*}. 1839 1840 1841@item -dumpdir @var{dumppfx} 1842@opindex dumpdir 1843When forming the name of an auxiliary or dump output file, use 1844@var{dumppfx} as a prefix: 1845 1846@smallexample 1847gcc -dumpdir pfx- -c foo.c ... 1848@end smallexample 1849 1850creates @file{foo.o} as the primary output, and auxiliary outputs named 1851@file{pfx-foo.*}, combining the given @var{dumppfx} with the default 1852@var{dumpbase} derived from the default primary output, derived in turn 1853from the input name. Dump outputs also take the input name suffix: 1854@file{pfx-foo.c.*}. 1855 1856If @var{dumppfx} is to be used as a directory name, it must end with a 1857directory separator: 1858 1859@smallexample 1860gcc -dumpdir dir/ -c foo.c -o obj/bar.o ... 1861@end smallexample 1862 1863creates @file{obj/bar.o} as the primary output, and auxiliary outputs 1864named @file{dir/bar.*}, combining the given @var{dumppfx} with the 1865default @var{dumpbase} derived from the primary output name. Dump 1866outputs also take the input name suffix: @file{dir/bar.c.*}. 1867 1868It defaults to the location of the output file; options 1869@option{-save-temps=cwd} and @option{-save-temps=obj} override this 1870default, just like an explicit @option{-dumpdir} option. In case 1871multiple such options are given, the last one prevails: 1872 1873@smallexample 1874gcc -dumpdir pfx- -c foo.c -save-temps=obj ... 1875@end smallexample 1876 1877outputs @file{foo.o}, with auxiliary outputs named @file{foo.*} because 1878@option{-save-temps=*} overrides the @var{dumppfx} given by the earlier 1879@option{-dumpdir} option. It does not matter that @option{=obj} is the 1880default for @option{-save-temps}, nor that the output directory is 1881implicitly the current directory. Dump outputs are named 1882@file{foo.c.*}. 1883 1884When compiling from multiple input files, if @option{-dumpbase} is 1885specified, @var{dumpbase}, minus a @var{auxdropsuf} suffix, and a dash 1886are appended to (or override, if containing any directory components) an 1887explicit or defaulted @var{dumppfx}, so that each of the multiple 1888compilations gets differently-named aux and dump outputs. 1889 1890@smallexample 1891gcc foo.c bar.c -c -dumpdir dir/pfx- -dumpbase main ... 1892@end smallexample 1893 1894outputs auxiliary dumps to @file{dir/pfx-main-foo.*} and 1895@file{dir/pfx-main-bar.*}, appending @var{dumpbase}- to @var{dumppfx}. 1896Dump outputs retain the input file suffix: @file{dir/pfx-main-foo.c.*} 1897and @file{dir/pfx-main-bar.c.*}, respectively. Contrast with the 1898single-input compilation: 1899 1900@smallexample 1901gcc foo.c -c -dumpdir dir/pfx- -dumpbase main ... 1902@end smallexample 1903 1904that, applying @option{-dumpbase} to a single source, does not compute 1905and append a separate @var{dumpbase} per input file. Its auxiliary and 1906dump outputs go in @file{dir/pfx-main.*}. 1907 1908When compiling and then linking from multiple input files, a defaulted 1909or explicitly specified @var{dumppfx} also undergoes the @var{dumpbase}- 1910transformation above (e.g. the compilation of @file{foo.c} and 1911@file{bar.c} above, but without @option{-c}). If neither 1912@option{-dumpdir} nor @option{-dumpbase} are given, the linker output 1913base name, minus @var{auxdropsuf}, if specified, or the executable 1914suffix otherwise, plus a dash is appended to the default @var{dumppfx} 1915instead. Note, however, that unlike earlier cases of linking: 1916 1917@smallexample 1918gcc foo.c bar.c -dumpdir dir/pfx- -o main ... 1919@end smallexample 1920 1921does not append the output name @file{main} to @var{dumppfx}, because 1922@option{-dumpdir} is explicitly specified. The goal is that the 1923explicitly-specified @var{dumppfx} may contain the specified output name 1924as part of the prefix, if desired; only an explicitly-specified 1925@option{-dumpbase} would be combined with it, in order to avoid simply 1926discarding a meaningful option. 1927 1928When compiling and then linking from a single input file, the linker 1929output base name will only be appended to the default @var{dumppfx} as 1930above if it does not share the base name with the single input file 1931name. This has been covered in single-input linking cases above, but 1932not with an explicit @option{-dumpdir} that inhibits the combination, 1933even if overridden by @option{-save-temps=*}: 1934 1935@smallexample 1936gcc foo.c -dumpdir alt/pfx- -o dir/main.exe -save-temps=cwd ... 1937@end smallexample 1938 1939Auxiliary outputs are named @file{foo.*}, and dump outputs 1940@file{foo.c.*}, in the current working directory as ultimately requested 1941by @option{-save-temps=cwd}. 1942 1943Summing it all up for an intuitive though slightly imprecise data flow: 1944the primary output name is broken into a directory part and a basename 1945part; @var{dumppfx} is set to the former, unless overridden by 1946@option{-dumpdir} or @option{-save-temps=*}, and @var{dumpbase} is set 1947to the latter, unless overriden by @option{-dumpbase}. If there are 1948multiple inputs or linking, this @var{dumpbase} may be combined with 1949@var{dumppfx} and taken from each input file. Auxiliary output names 1950for each input are formed by combining @var{dumppfx}, @var{dumpbase} 1951minus suffix, and the auxiliary output suffix; dump output names are 1952only different in that the suffix from @var{dumpbase} is retained. 1953 1954When it comes to auxiliary and dump outputs created during LTO 1955recompilation, a combination of @var{dumppfx} and @var{dumpbase}, as 1956given or as derived from the linker output name but not from inputs, 1957even in cases in which this combination would not otherwise be used as 1958such, is passed down with a trailing period replacing the compiler-added 1959dash, if any, as a @option{-dumpdir} option to @command{lto-wrapper}; 1960being involved in linking, this program does not normally get any 1961@option{-dumpbase} and @option{-dumpbase-ext}, and it ignores them. 1962 1963When running sub-compilers, @command{lto-wrapper} appends LTO stage 1964names to the received @var{dumppfx}, ensures it contains a directory 1965component so that it overrides any @option{-dumpdir}, and passes that as 1966@option{-dumpbase} to sub-compilers. 1967 1968@item -v 1969@opindex v 1970Print (on standard error output) the commands executed to run the stages 1971of compilation. Also print the version number of the compiler driver 1972program and of the preprocessor and the compiler proper. 1973 1974@item -### 1975@opindex ### 1976Like @option{-v} except the commands are not executed and arguments 1977are quoted unless they contain only alphanumeric characters or @code{./-_}. 1978This is useful for shell scripts to capture the driver-generated command lines. 1979 1980@item --help 1981@opindex help 1982Print (on the standard output) a description of the command-line options 1983understood by @command{gcc}. If the @option{-v} option is also specified 1984then @option{--help} is also passed on to the various processes 1985invoked by @command{gcc}, so that they can display the command-line options 1986they accept. If the @option{-Wextra} option has also been specified 1987(prior to the @option{--help} option), then command-line options that 1988have no documentation associated with them are also displayed. 1989 1990@item --target-help 1991@opindex target-help 1992Print (on the standard output) a description of target-specific command-line 1993options for each tool. For some targets extra target-specific 1994information may also be printed. 1995 1996@item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]} 1997Print (on the standard output) a description of the command-line 1998options understood by the compiler that fit into all specified classes 1999and qualifiers. These are the supported classes: 2000 2001@table @asis 2002@item @samp{optimizers} 2003Display all of the optimization options supported by the 2004compiler. 2005 2006@item @samp{warnings} 2007Display all of the options controlling warning messages 2008produced by the compiler. 2009 2010@item @samp{target} 2011Display target-specific options. Unlike the 2012@option{--target-help} option however, target-specific options of the 2013linker and assembler are not displayed. This is because those 2014tools do not currently support the extended @option{--help=} syntax. 2015 2016@item @samp{params} 2017Display the values recognized by the @option{--param} 2018option. 2019 2020@item @var{language} 2021Display the options supported for @var{language}, where 2022@var{language} is the name of one of the languages supported in this 2023version of GCC@. If an option is supported by all languages, one needs 2024to select @samp{common} class. 2025 2026@item @samp{common} 2027Display the options that are common to all languages. 2028@end table 2029 2030These are the supported qualifiers: 2031 2032@table @asis 2033@item @samp{undocumented} 2034Display only those options that are undocumented. 2035 2036@item @samp{joined} 2037Display options taking an argument that appears after an equal 2038sign in the same continuous piece of text, such as: 2039@samp{--help=target}. 2040 2041@item @samp{separate} 2042Display options taking an argument that appears as a separate word 2043following the original option, such as: @samp{-o output-file}. 2044@end table 2045 2046Thus for example to display all the undocumented target-specific 2047switches supported by the compiler, use: 2048 2049@smallexample 2050--help=target,undocumented 2051@end smallexample 2052 2053The sense of a qualifier can be inverted by prefixing it with the 2054@samp{^} character, so for example to display all binary warning 2055options (i.e., ones that are either on or off and that do not take an 2056argument) that have a description, use: 2057 2058@smallexample 2059--help=warnings,^joined,^undocumented 2060@end smallexample 2061 2062The argument to @option{--help=} should not consist solely of inverted 2063qualifiers. 2064 2065Combining several classes is possible, although this usually 2066restricts the output so much that there is nothing to display. One 2067case where it does work, however, is when one of the classes is 2068@var{target}. For example, to display all the target-specific 2069optimization options, use: 2070 2071@smallexample 2072--help=target,optimizers 2073@end smallexample 2074 2075The @option{--help=} option can be repeated on the command line. Each 2076successive use displays its requested class of options, skipping 2077those that have already been displayed. If @option{--help} is also 2078specified anywhere on the command line then this takes precedence 2079over any @option{--help=} option. 2080 2081If the @option{-Q} option appears on the command line before the 2082@option{--help=} option, then the descriptive text displayed by 2083@option{--help=} is changed. Instead of describing the displayed 2084options, an indication is given as to whether the option is enabled, 2085disabled or set to a specific value (assuming that the compiler 2086knows this at the point where the @option{--help=} option is used). 2087 2088Here is a truncated example from the ARM port of @command{gcc}: 2089 2090@smallexample 2091 % gcc -Q -mabi=2 --help=target -c 2092 The following options are target specific: 2093 -mabi= 2 2094 -mabort-on-noreturn [disabled] 2095 -mapcs [disabled] 2096@end smallexample 2097 2098The output is sensitive to the effects of previous command-line 2099options, so for example it is possible to find out which optimizations 2100are enabled at @option{-O2} by using: 2101 2102@smallexample 2103-Q -O2 --help=optimizers 2104@end smallexample 2105 2106Alternatively you can discover which binary optimizations are enabled 2107by @option{-O3} by using: 2108 2109@smallexample 2110gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 2111gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 2112diff /tmp/O2-opts /tmp/O3-opts | grep enabled 2113@end smallexample 2114 2115@item --version 2116@opindex version 2117Display the version number and copyrights of the invoked GCC@. 2118 2119@item -pass-exit-codes 2120@opindex pass-exit-codes 2121Normally the @command{gcc} program exits with the code of 1 if any 2122phase of the compiler returns a non-success return code. If you specify 2123@option{-pass-exit-codes}, the @command{gcc} program instead returns with 2124the numerically highest error produced by any phase returning an error 2125indication. The C, C++, and Fortran front ends return 4 if an internal 2126compiler error is encountered. 2127 2128@item -pipe 2129@opindex pipe 2130Use pipes rather than temporary files for communication between the 2131various stages of compilation. This fails to work on some systems where 2132the assembler is unable to read from a pipe; but the GNU assembler has 2133no trouble. 2134 2135@item -specs=@var{file} 2136@opindex specs 2137Process @var{file} after the compiler reads in the standard @file{specs} 2138file, in order to override the defaults which the @command{gcc} driver 2139program uses when determining what switches to pass to @command{cc1}, 2140@command{cc1plus}, @command{as}, @command{ld}, etc. More than one 2141@option{-specs=@var{file}} can be specified on the command line, and they 2142are processed in order, from left to right. @xref{Spec Files}, for 2143information about the format of the @var{file}. 2144 2145@item -wrapper 2146@opindex wrapper 2147Invoke all subcommands under a wrapper program. The name of the 2148wrapper program and its parameters are passed as a comma separated 2149list. 2150 2151@smallexample 2152gcc -c t.c -wrapper gdb,--args 2153@end smallexample 2154 2155@noindent 2156This invokes all subprograms of @command{gcc} under 2157@samp{gdb --args}, thus the invocation of @command{cc1} is 2158@samp{gdb --args cc1 @dots{}}. 2159 2160@item -ffile-prefix-map=@var{old}=@var{new} 2161@opindex ffile-prefix-map 2162When compiling files residing in directory @file{@var{old}}, record 2163any references to them in the result of the compilation as if the 2164files resided in directory @file{@var{new}} instead. Specifying this 2165option is equivalent to specifying all the individual 2166@option{-f*-prefix-map} options. This can be used to make reproducible 2167builds that are location independent. See also 2168@option{-fmacro-prefix-map} and @option{-fdebug-prefix-map}. 2169 2170@item -fplugin=@var{name}.so 2171@opindex fplugin 2172Load the plugin code in file @var{name}.so, assumed to be a 2173shared object to be dlopen'd by the compiler. The base name of 2174the shared object file is used to identify the plugin for the 2175purposes of argument parsing (See 2176@option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below). 2177Each plugin should define the callback functions specified in the 2178Plugins API. 2179 2180@item -fplugin-arg-@var{name}-@var{key}=@var{value} 2181@opindex fplugin-arg 2182Define an argument called @var{key} with a value of @var{value} 2183for the plugin called @var{name}. 2184 2185@item -fdump-ada-spec@r{[}-slim@r{]} 2186@opindex fdump-ada-spec 2187For C and C++ source and include files, generate corresponding Ada specs. 2188@xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn, 2189GNAT User's Guide}, which provides detailed documentation on this feature. 2190 2191@item -fada-spec-parent=@var{unit} 2192@opindex fada-spec-parent 2193In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate 2194Ada specs as child units of parent @var{unit}. 2195 2196@item -fdump-go-spec=@var{file} 2197@opindex fdump-go-spec 2198For input files in any language, generate corresponding Go 2199declarations in @var{file}. This generates Go @code{const}, 2200@code{type}, @code{var}, and @code{func} declarations which may be a 2201useful way to start writing a Go interface to code written in some 2202other language. 2203 2204@include @value{srcdir}/../libiberty/at-file.texi 2205@end table 2206 2207@node Invoking G++ 2208@section Compiling C++ Programs 2209 2210@cindex suffixes for C++ source 2211@cindex C++ source file suffixes 2212C++ source files conventionally use one of the suffixes @samp{.C}, 2213@samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or 2214@samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp}, 2215@samp{.H}, or (for shared template code) @samp{.tcc}; and 2216preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes 2217files with these names and compiles them as C++ programs even if you 2218call the compiler the same way as for compiling C programs (usually 2219with the name @command{gcc}). 2220 2221@findex g++ 2222@findex c++ 2223However, the use of @command{gcc} does not add the C++ library. 2224@command{g++} is a program that calls GCC and automatically specifies linking 2225against the C++ library. It treats @samp{.c}, 2226@samp{.h} and @samp{.i} files as C++ source files instead of C source 2227files unless @option{-x} is used. This program is also useful when 2228precompiling a C header file with a @samp{.h} extension for use in C++ 2229compilations. On many systems, @command{g++} is also installed with 2230the name @command{c++}. 2231 2232@cindex invoking @command{g++} 2233When you compile C++ programs, you may specify many of the same 2234command-line options that you use for compiling programs in any 2235language; or command-line options meaningful for C and related 2236languages; or options that are meaningful only for C++ programs. 2237@xref{C Dialect Options,,Options Controlling C Dialect}, for 2238explanations of options for languages related to C@. 2239@xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for 2240explanations of options that are meaningful only for C++ programs. 2241 2242@node C Dialect Options 2243@section Options Controlling C Dialect 2244@cindex dialect options 2245@cindex language dialect options 2246@cindex options, dialect 2247 2248The following options control the dialect of C (or languages derived 2249from C, such as C++, Objective-C and Objective-C++) that the compiler 2250accepts: 2251 2252@table @gcctabopt 2253@cindex ANSI support 2254@cindex ISO support 2255@item -ansi 2256@opindex ansi 2257In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is 2258equivalent to @option{-std=c++98}. 2259 2260This turns off certain features of GCC that are incompatible with ISO 2261C90 (when compiling C code), or of standard C++ (when compiling C++ code), 2262such as the @code{asm} and @code{typeof} keywords, and 2263predefined macros such as @code{unix} and @code{vax} that identify the 2264type of system you are using. It also enables the undesirable and 2265rarely used ISO trigraph feature. For the C compiler, 2266it disables recognition of C++ style @samp{//} comments as well as 2267the @code{inline} keyword. 2268 2269The alternate keywords @code{__asm__}, @code{__extension__}, 2270@code{__inline__} and @code{__typeof__} continue to work despite 2271@option{-ansi}. You would not want to use them in an ISO C program, of 2272course, but it is useful to put them in header files that might be included 2273in compilations done with @option{-ansi}. Alternate predefined macros 2274such as @code{__unix__} and @code{__vax__} are also available, with or 2275without @option{-ansi}. 2276 2277The @option{-ansi} option does not cause non-ISO programs to be 2278rejected gratuitously. For that, @option{-Wpedantic} is required in 2279addition to @option{-ansi}. @xref{Warning Options}. 2280 2281The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi} 2282option is used. Some header files may notice this macro and refrain 2283from declaring certain functions or defining certain macros that the 2284ISO standard doesn't call for; this is to avoid interfering with any 2285programs that might use these names for other things. 2286 2287Functions that are normally built in but do not have semantics 2288defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in 2289functions when @option{-ansi} is used. @xref{Other Builtins,,Other 2290built-in functions provided by GCC}, for details of the functions 2291affected. 2292 2293@item -std= 2294@opindex std 2295Determine the language standard. @xref{Standards,,Language Standards 2296Supported by GCC}, for details of these standard versions. This option 2297is currently only supported when compiling C or C++. 2298 2299The compiler can accept several base standards, such as @samp{c90} or 2300@samp{c++98}, and GNU dialects of those standards, such as 2301@samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the 2302compiler accepts all programs following that standard plus those 2303using GNU extensions that do not contradict it. For example, 2304@option{-std=c90} turns off certain features of GCC that are 2305incompatible with ISO C90, such as the @code{asm} and @code{typeof} 2306keywords, but not other GNU extensions that do not have a meaning in 2307ISO C90, such as omitting the middle term of a @code{?:} 2308expression. On the other hand, when a GNU dialect of a standard is 2309specified, all features supported by the compiler are enabled, even when 2310those features change the meaning of the base standard. As a result, some 2311strict-conforming programs may be rejected. The particular standard 2312is used by @option{-Wpedantic} to identify which features are GNU 2313extensions given that version of the standard. For example 2314@option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//} 2315comments, while @option{-std=gnu99 -Wpedantic} does not. 2316 2317A value for this option must be provided; possible values are 2318 2319@table @samp 2320@item c90 2321@itemx c89 2322@itemx iso9899:1990 2323Support all ISO C90 programs (certain GNU extensions that conflict 2324with ISO C90 are disabled). Same as @option{-ansi} for C code. 2325 2326@item iso9899:199409 2327ISO C90 as modified in amendment 1. 2328 2329@item c99 2330@itemx c9x 2331@itemx iso9899:1999 2332@itemx iso9899:199x 2333ISO C99. This standard is substantially completely supported, modulo 2334bugs and floating-point issues 2335(mainly but not entirely relating to optional C99 features from 2336Annexes F and G). See 2337@w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The 2338names @samp{c9x} and @samp{iso9899:199x} are deprecated. 2339 2340@item c11 2341@itemx c1x 2342@itemx iso9899:2011 2343ISO C11, the 2011 revision of the ISO C standard. This standard is 2344substantially completely supported, modulo bugs, floating-point issues 2345(mainly but not entirely relating to optional C11 features from 2346Annexes F and G) and the optional Annexes K (Bounds-checking 2347interfaces) and L (Analyzability). The name @samp{c1x} is deprecated. 2348 2349@item c17 2350@itemx c18 2351@itemx iso9899:2017 2352@itemx iso9899:2018 2353ISO C17, the 2017 revision of the ISO C standard 2354(published in 2018). This standard is 2355same as C11 except for corrections of defects (all of which are also 2356applied with @option{-std=c11}) and a new value of 2357@code{__STDC_VERSION__}, and so is supported to the same extent as C11. 2358 2359@item c2x 2360The next version of the ISO C standard, still under development. The 2361support for this version is experimental and incomplete. 2362 2363@item gnu90 2364@itemx gnu89 2365GNU dialect of ISO C90 (including some C99 features). 2366 2367@item gnu99 2368@itemx gnu9x 2369GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated. 2370 2371@item gnu11 2372@itemx gnu1x 2373GNU dialect of ISO C11. 2374The name @samp{gnu1x} is deprecated. 2375 2376@item gnu17 2377@itemx gnu18 2378GNU dialect of ISO C17. This is the default for C code. 2379 2380@item gnu2x 2381The next version of the ISO C standard, still under development, plus 2382GNU extensions. The support for this version is experimental and 2383incomplete. 2384 2385@item c++98 2386@itemx c++03 2387The 1998 ISO C++ standard plus the 2003 technical corrigendum and some 2388additional defect reports. Same as @option{-ansi} for C++ code. 2389 2390@item gnu++98 2391@itemx gnu++03 2392GNU dialect of @option{-std=c++98}. 2393 2394@item c++11 2395@itemx c++0x 2396The 2011 ISO C++ standard plus amendments. 2397The name @samp{c++0x} is deprecated. 2398 2399@item gnu++11 2400@itemx gnu++0x 2401GNU dialect of @option{-std=c++11}. 2402The name @samp{gnu++0x} is deprecated. 2403 2404@item c++14 2405@itemx c++1y 2406The 2014 ISO C++ standard plus amendments. 2407The name @samp{c++1y} is deprecated. 2408 2409@item gnu++14 2410@itemx gnu++1y 2411GNU dialect of @option{-std=c++14}. 2412The name @samp{gnu++1y} is deprecated. 2413 2414@item c++17 2415@itemx c++1z 2416The 2017 ISO C++ standard plus amendments. 2417The name @samp{c++1z} is deprecated. 2418 2419@item gnu++17 2420@itemx gnu++1z 2421GNU dialect of @option{-std=c++17}. 2422This is the default for C++ code. 2423The name @samp{gnu++1z} is deprecated. 2424 2425@item c++20 2426@itemx c++2a 2427The 2020 ISO C++ standard plus amendments. 2428Support is experimental, and could change in incompatible ways in 2429future releases. 2430The name @samp{c++2a} is deprecated. 2431 2432@item gnu++20 2433@itemx gnu++2a 2434GNU dialect of @option{-std=c++20}. 2435Support is experimental, and could change in incompatible ways in 2436future releases. 2437The name @samp{gnu++2a} is deprecated. 2438 2439@item c++2b 2440@itemx c++23 2441The next revision of the ISO C++ standard, planned for 24422023. Support is highly experimental, and will almost certainly 2443change in incompatible ways in future releases. 2444 2445@item gnu++2b 2446@itemx gnu++23 2447GNU dialect of @option{-std=c++2b}. Support is highly experimental, 2448and will almost certainly change in incompatible ways in future 2449releases. 2450@end table 2451 2452@item -fgnu89-inline 2453@opindex fgnu89-inline 2454The option @option{-fgnu89-inline} tells GCC to use the traditional 2455GNU semantics for @code{inline} functions when in C99 mode. 2456@xref{Inline,,An Inline Function is As Fast As a Macro}. 2457Using this option is roughly equivalent to adding the 2458@code{gnu_inline} function attribute to all inline functions 2459(@pxref{Function Attributes}). 2460 2461The option @option{-fno-gnu89-inline} explicitly tells GCC to use the 2462C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it 2463specifies the default behavior). 2464This option is not supported in @option{-std=c90} or 2465@option{-std=gnu90} mode. 2466 2467The preprocessor macros @code{__GNUC_GNU_INLINE__} and 2468@code{__GNUC_STDC_INLINE__} may be used to check which semantics are 2469in effect for @code{inline} functions. @xref{Common Predefined 2470Macros,,,cpp,The C Preprocessor}. 2471 2472@item -fpermitted-flt-eval-methods=@var{style} 2473@opindex fpermitted-flt-eval-methods 2474@opindex fpermitted-flt-eval-methods=c11 2475@opindex fpermitted-flt-eval-methods=ts-18661-3 2476ISO/IEC TS 18661-3 defines new permissible values for 2477@code{FLT_EVAL_METHOD} that indicate that operations and constants with 2478a semantic type that is an interchange or extended format should be 2479evaluated to the precision and range of that type. These new values are 2480a superset of those permitted under C99/C11, which does not specify the 2481meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code 2482conforming to C11 may not have been written expecting the possibility of 2483the new values. 2484 2485@option{-fpermitted-flt-eval-methods} specifies whether the compiler 2486should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11, 2487or the extended set of values specified in ISO/IEC TS 18661-3. 2488 2489@var{style} is either @code{c11} or @code{ts-18661-3} as appropriate. 2490 2491The default when in a standards compliant mode (@option{-std=c11} or similar) 2492is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU 2493dialect (@option{-std=gnu11} or similar) is 2494@option{-fpermitted-flt-eval-methods=ts-18661-3}. 2495 2496@item -aux-info @var{filename} 2497@opindex aux-info 2498Output to the given filename prototyped declarations for all functions 2499declared and/or defined in a translation unit, including those in header 2500files. This option is silently ignored in any language other than C@. 2501 2502Besides declarations, the file indicates, in comments, the origin of 2503each declaration (source file and line), whether the declaration was 2504implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or 2505@samp{O} for old, respectively, in the first character after the line 2506number and the colon), and whether it came from a declaration or a 2507definition (@samp{C} or @samp{F}, respectively, in the following 2508character). In the case of function definitions, a K&R-style list of 2509arguments followed by their declarations is also provided, inside 2510comments, after the declaration. 2511 2512@item -fallow-parameterless-variadic-functions 2513@opindex fallow-parameterless-variadic-functions 2514Accept variadic functions without named parameters. 2515 2516Although it is possible to define such a function, this is not very 2517useful as it is not possible to read the arguments. This is only 2518supported for C as this construct is allowed by C++. 2519 2520@item -fno-asm 2521@opindex fno-asm 2522@opindex fasm 2523Do not recognize @code{asm}, @code{inline} or @code{typeof} as a 2524keyword, so that code can use these words as identifiers. You can use 2525the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__} 2526instead. @option{-ansi} implies @option{-fno-asm}. 2527 2528In C++, this switch only affects the @code{typeof} keyword, since 2529@code{asm} and @code{inline} are standard keywords. You may want to 2530use the @option{-fno-gnu-keywords} flag instead, which has the same 2531effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this 2532switch only affects the @code{asm} and @code{typeof} keywords, since 2533@code{inline} is a standard keyword in ISO C99. 2534 2535@item -fno-builtin 2536@itemx -fno-builtin-@var{function} 2537@opindex fno-builtin 2538@opindex fbuiltin 2539@cindex built-in functions 2540Don't recognize built-in functions that do not begin with 2541@samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in 2542functions provided by GCC}, for details of the functions affected, 2543including those which are not built-in functions when @option{-ansi} or 2544@option{-std} options for strict ISO C conformance are used because they 2545do not have an ISO standard meaning. 2546 2547GCC normally generates special code to handle certain built-in functions 2548more efficiently; for instance, calls to @code{alloca} may become single 2549instructions which adjust the stack directly, and calls to @code{memcpy} 2550may become inline copy loops. The resulting code is often both smaller 2551and faster, but since the function calls no longer appear as such, you 2552cannot set a breakpoint on those calls, nor can you change the behavior 2553of the functions by linking with a different library. In addition, 2554when a function is recognized as a built-in function, GCC may use 2555information about that function to warn about problems with calls to 2556that function, or to generate more efficient code, even if the 2557resulting code still contains calls to that function. For example, 2558warnings are given with @option{-Wformat} for bad calls to 2559@code{printf} when @code{printf} is built in and @code{strlen} is 2560known not to modify global memory. 2561 2562With the @option{-fno-builtin-@var{function}} option 2563only the built-in function @var{function} is 2564disabled. @var{function} must not begin with @samp{__builtin_}. If a 2565function is named that is not built-in in this version of GCC, this 2566option is ignored. There is no corresponding 2567@option{-fbuiltin-@var{function}} option; if you wish to enable 2568built-in functions selectively when using @option{-fno-builtin} or 2569@option{-ffreestanding}, you may define macros such as: 2570 2571@smallexample 2572#define abs(n) __builtin_abs ((n)) 2573#define strcpy(d, s) __builtin_strcpy ((d), (s)) 2574@end smallexample 2575 2576@item -fgimple 2577@opindex fgimple 2578 2579Enable parsing of function definitions marked with @code{__GIMPLE}. 2580This is an experimental feature that allows unit testing of GIMPLE 2581passes. 2582 2583@item -fhosted 2584@opindex fhosted 2585@cindex hosted environment 2586 2587Assert that compilation targets a hosted environment. This implies 2588@option{-fbuiltin}. A hosted environment is one in which the 2589entire standard library is available, and in which @code{main} has a return 2590type of @code{int}. Examples are nearly everything except a kernel. 2591This is equivalent to @option{-fno-freestanding}. 2592 2593@item -ffreestanding 2594@opindex ffreestanding 2595@cindex hosted environment 2596 2597Assert that compilation targets a freestanding environment. This 2598implies @option{-fno-builtin}. A freestanding environment 2599is one in which the standard library may not exist, and program startup may 2600not necessarily be at @code{main}. The most obvious example is an OS kernel. 2601This is equivalent to @option{-fno-hosted}. 2602 2603@xref{Standards,,Language Standards Supported by GCC}, for details of 2604freestanding and hosted environments. 2605 2606@item -fopenacc 2607@opindex fopenacc 2608@cindex OpenACC accelerator programming 2609Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and 2610@code{!$acc} in Fortran. When @option{-fopenacc} is specified, the 2611compiler generates accelerated code according to the OpenACC Application 2612Programming Interface v2.6 @w{@uref{https://www.openacc.org}}. This option 2613implies @option{-pthread}, and thus is only supported on targets that 2614have support for @option{-pthread}. 2615 2616@item -fopenacc-dim=@var{geom} 2617@opindex fopenacc-dim 2618@cindex OpenACC accelerator programming 2619Specify default compute dimensions for parallel offload regions that do 2620not explicitly specify. The @var{geom} value is a triple of 2621':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size 2622can be omitted, to use a target-specific default value. 2623 2624@item -fopenmp 2625@opindex fopenmp 2626@cindex OpenMP parallel 2627Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and 2628@code{!$omp} in Fortran. When @option{-fopenmp} is specified, the 2629compiler generates parallel code according to the OpenMP Application 2630Program Interface v4.5 @w{@uref{https://www.openmp.org}}. This option 2631implies @option{-pthread}, and thus is only supported on targets that 2632have support for @option{-pthread}. @option{-fopenmp} implies 2633@option{-fopenmp-simd}. 2634 2635@item -fopenmp-simd 2636@opindex fopenmp-simd 2637@cindex OpenMP SIMD 2638@cindex SIMD 2639Enable handling of OpenMP's SIMD directives with @code{#pragma omp} 2640in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives 2641are ignored. 2642 2643@item -fgnu-tm 2644@opindex fgnu-tm 2645When the option @option{-fgnu-tm} is specified, the compiler 2646generates code for the Linux variant of Intel's current Transactional 2647Memory ABI specification document (Revision 1.1, May 6 2009). This is 2648an experimental feature whose interface may change in future versions 2649of GCC, as the official specification changes. Please note that not 2650all architectures are supported for this feature. 2651 2652For more information on GCC's support for transactional memory, 2653@xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU 2654Transactional Memory Library}. 2655 2656Note that the transactional memory feature is not supported with 2657non-call exceptions (@option{-fnon-call-exceptions}). 2658 2659@item -fms-extensions 2660@opindex fms-extensions 2661Accept some non-standard constructs used in Microsoft header files. 2662 2663In C++ code, this allows member names in structures to be similar 2664to previous types declarations. 2665 2666@smallexample 2667typedef int UOW; 2668struct ABC @{ 2669 UOW UOW; 2670@}; 2671@end smallexample 2672 2673Some cases of unnamed fields in structures and unions are only 2674accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union 2675fields within structs/unions}, for details. 2676 2677Note that this option is off for all targets except for x86 2678targets using ms-abi. 2679 2680@item -fplan9-extensions 2681@opindex fplan9-extensions 2682Accept some non-standard constructs used in Plan 9 code. 2683 2684This enables @option{-fms-extensions}, permits passing pointers to 2685structures with anonymous fields to functions that expect pointers to 2686elements of the type of the field, and permits referring to anonymous 2687fields declared using a typedef. @xref{Unnamed Fields,,Unnamed 2688struct/union fields within structs/unions}, for details. This is only 2689supported for C, not C++. 2690 2691@item -fcond-mismatch 2692@opindex fcond-mismatch 2693Allow conditional expressions with mismatched types in the second and 2694third arguments. The value of such an expression is void. This option 2695is not supported for C++. 2696 2697@item -flax-vector-conversions 2698@opindex flax-vector-conversions 2699Allow implicit conversions between vectors with differing numbers of 2700elements and/or incompatible element types. This option should not be 2701used for new code. 2702 2703@item -funsigned-char 2704@opindex funsigned-char 2705Let the type @code{char} be unsigned, like @code{unsigned char}. 2706 2707Each kind of machine has a default for what @code{char} should 2708be. It is either like @code{unsigned char} by default or like 2709@code{signed char} by default. 2710 2711Ideally, a portable program should always use @code{signed char} or 2712@code{unsigned char} when it depends on the signedness of an object. 2713But many programs have been written to use plain @code{char} and 2714expect it to be signed, or expect it to be unsigned, depending on the 2715machines they were written for. This option, and its inverse, let you 2716make such a program work with the opposite default. 2717 2718The type @code{char} is always a distinct type from each of 2719@code{signed char} or @code{unsigned char}, even though its behavior 2720is always just like one of those two. 2721 2722@item -fsigned-char 2723@opindex fsigned-char 2724Let the type @code{char} be signed, like @code{signed char}. 2725 2726Note that this is equivalent to @option{-fno-unsigned-char}, which is 2727the negative form of @option{-funsigned-char}. Likewise, the option 2728@option{-fno-signed-char} is equivalent to @option{-funsigned-char}. 2729 2730@item -fsigned-bitfields 2731@itemx -funsigned-bitfields 2732@itemx -fno-signed-bitfields 2733@itemx -fno-unsigned-bitfields 2734@opindex fsigned-bitfields 2735@opindex funsigned-bitfields 2736@opindex fno-signed-bitfields 2737@opindex fno-unsigned-bitfields 2738These options control whether a bit-field is signed or unsigned, when the 2739declaration does not use either @code{signed} or @code{unsigned}. By 2740default, such a bit-field is signed, because this is consistent: the 2741basic integer types such as @code{int} are signed types. 2742 2743@item -fsso-struct=@var{endianness} 2744@opindex fsso-struct 2745Set the default scalar storage order of structures and unions to the 2746specified endianness. The accepted values are @samp{big-endian}, 2747@samp{little-endian} and @samp{native} for the native endianness of 2748the target (the default). This option is not supported for C++. 2749 2750@strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate 2751code that is not binary compatible with code generated without it if the 2752specified endianness is not the native endianness of the target. 2753@end table 2754 2755@node C++ Dialect Options 2756@section Options Controlling C++ Dialect 2757 2758@cindex compiler options, C++ 2759@cindex C++ options, command-line 2760@cindex options, C++ 2761This section describes the command-line options that are only meaningful 2762for C++ programs. You can also use most of the GNU compiler options 2763regardless of what language your program is in. For example, you 2764might compile a file @file{firstClass.C} like this: 2765 2766@smallexample 2767g++ -g -fstrict-enums -O -c firstClass.C 2768@end smallexample 2769 2770@noindent 2771In this example, only @option{-fstrict-enums} is an option meant 2772only for C++ programs; you can use the other options with any 2773language supported by GCC@. 2774 2775Some options for compiling C programs, such as @option{-std}, are also 2776relevant for C++ programs. 2777@xref{C Dialect Options,,Options Controlling C Dialect}. 2778 2779Here is a list of options that are @emph{only} for compiling C++ programs: 2780 2781@table @gcctabopt 2782 2783@item -fabi-version=@var{n} 2784@opindex fabi-version 2785Use version @var{n} of the C++ ABI@. The default is version 0. 2786 2787Version 0 refers to the version conforming most closely to 2788the C++ ABI specification. Therefore, the ABI obtained using version 0 2789will change in different versions of G++ as ABI bugs are fixed. 2790 2791Version 1 is the version of the C++ ABI that first appeared in G++ 3.2. 2792 2793Version 2 is the version of the C++ ABI that first appeared in G++ 27943.4, and was the default through G++ 4.9. 2795 2796Version 3 corrects an error in mangling a constant address as a 2797template argument. 2798 2799Version 4, which first appeared in G++ 4.5, implements a standard 2800mangling for vector types. 2801 2802Version 5, which first appeared in G++ 4.6, corrects the mangling of 2803attribute const/volatile on function pointer types, decltype of a 2804plain decl, and use of a function parameter in the declaration of 2805another parameter. 2806 2807Version 6, which first appeared in G++ 4.7, corrects the promotion 2808behavior of C++11 scoped enums and the mangling of template argument 2809packs, const/static_cast, prefix ++ and --, and a class scope function 2810used as a template argument. 2811 2812Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a 2813builtin type and corrects the mangling of lambdas in default argument 2814scope. 2815 2816Version 8, which first appeared in G++ 4.9, corrects the substitution 2817behavior of function types with function-cv-qualifiers. 2818 2819Version 9, which first appeared in G++ 5.2, corrects the alignment of 2820@code{nullptr_t}. 2821 2822Version 10, which first appeared in G++ 6.1, adds mangling of 2823attributes that affect type identity, such as ia32 calling convention 2824attributes (e.g.@: @samp{stdcall}). 2825 2826Version 11, which first appeared in G++ 7, corrects the mangling of 2827sizeof... expressions and operator names. For multiple entities with 2828the same name within a function, that are declared in different scopes, 2829the mangling now changes starting with the twelfth occurrence. It also 2830implies @option{-fnew-inheriting-ctors}. 2831 2832Version 12, which first appeared in G++ 8, corrects the calling 2833conventions for empty classes on the x86_64 target and for classes 2834with only deleted copy/move constructors. It accidentally changes the 2835calling convention for classes with a deleted copy constructor and a 2836trivial move constructor. 2837 2838Version 13, which first appeared in G++ 8.2, fixes the accidental 2839change in version 12. 2840 2841Version 14, which first appeared in G++ 10, corrects the mangling of 2842the nullptr expression. 2843 2844Version 15, which first appeared in G++ 11, changes the mangling of 2845@code{__alignof__} to be distinct from that of @code{alignof}, and 2846dependent operator names. 2847 2848See also @option{-Wabi}. 2849 2850@item -fabi-compat-version=@var{n} 2851@opindex fabi-compat-version 2852On targets that support strong aliases, G++ 2853works around mangling changes by creating an alias with the correct 2854mangled name when defining a symbol with an incorrect mangled name. 2855This switch specifies which ABI version to use for the alias. 2856 2857With @option{-fabi-version=0} (the default), this defaults to 11 (GCC 7 2858compatibility). If another ABI version is explicitly selected, this 2859defaults to 0. For compatibility with GCC versions 3.2 through 4.9, 2860use @option{-fabi-compat-version=2}. 2861 2862If this option is not provided but @option{-Wabi=@var{n}} is, that 2863version is used for compatibility aliases. If this option is provided 2864along with @option{-Wabi} (without the version), the version from this 2865option is used for the warning. 2866 2867@item -fno-access-control 2868@opindex fno-access-control 2869@opindex faccess-control 2870Turn off all access checking. This switch is mainly useful for working 2871around bugs in the access control code. 2872 2873@item -faligned-new 2874@opindex faligned-new 2875Enable support for C++17 @code{new} of types that require more 2876alignment than @code{void* ::operator new(std::size_t)} provides. A 2877numeric argument such as @code{-faligned-new=32} can be used to 2878specify how much alignment (in bytes) is provided by that function, 2879but few users will need to override the default of 2880@code{alignof(std::max_align_t)}. 2881 2882This flag is enabled by default for @option{-std=c++17}. 2883 2884@item -fchar8_t 2885@itemx -fno-char8_t 2886@opindex fchar8_t 2887@opindex fno-char8_t 2888Enable support for @code{char8_t} as adopted for C++20. This includes 2889the addition of a new @code{char8_t} fundamental type, changes to the 2890types of UTF-8 string and character literals, new signatures for 2891user-defined literals, associated standard library updates, and new 2892@code{__cpp_char8_t} and @code{__cpp_lib_char8_t} feature test macros. 2893 2894This option enables functions to be overloaded for ordinary and UTF-8 2895strings: 2896 2897@smallexample 2898int f(const char *); // #1 2899int f(const char8_t *); // #2 2900int v1 = f("text"); // Calls #1 2901int v2 = f(u8"text"); // Calls #2 2902@end smallexample 2903 2904@noindent 2905and introduces new signatures for user-defined literals: 2906 2907@smallexample 2908int operator""_udl1(char8_t); 2909int v3 = u8'x'_udl1; 2910int operator""_udl2(const char8_t*, std::size_t); 2911int v4 = u8"text"_udl2; 2912template<typename T, T...> int operator""_udl3(); 2913int v5 = u8"text"_udl3; 2914@end smallexample 2915 2916@noindent 2917The change to the types of UTF-8 string and character literals introduces 2918incompatibilities with ISO C++11 and later standards. For example, the 2919following code is well-formed under ISO C++11, but is ill-formed when 2920@option{-fchar8_t} is specified. 2921 2922@smallexample 2923char ca[] = u8"xx"; // error: char-array initialized from wide 2924 // string 2925const char *cp = u8"xx";// error: invalid conversion from 2926 // `const char8_t*' to `const char*' 2927int f(const char*); 2928auto v = f(u8"xx"); // error: invalid conversion from 2929 // `const char8_t*' to `const char*' 2930std::string s@{u8"xx"@}; // error: no matching function for call to 2931 // `std::basic_string<char>::basic_string()' 2932using namespace std::literals; 2933s = u8"xx"s; // error: conversion from 2934 // `basic_string<char8_t>' to non-scalar 2935 // type `basic_string<char>' requested 2936@end smallexample 2937 2938@item -fcheck-new 2939@opindex fcheck-new 2940Check that the pointer returned by @code{operator new} is non-null 2941before attempting to modify the storage allocated. This check is 2942normally unnecessary because the C++ standard specifies that 2943@code{operator new} only returns @code{0} if it is declared 2944@code{throw()}, in which case the compiler always checks the 2945return value even without this option. In all other cases, when 2946@code{operator new} has a non-empty exception specification, memory 2947exhaustion is signalled by throwing @code{std::bad_alloc}. See also 2948@samp{new (nothrow)}. 2949 2950@item -fconcepts 2951@itemx -fconcepts-ts 2952@opindex fconcepts 2953@opindex fconcepts-ts 2954Below @option{-std=c++20}, @option{-fconcepts} enables support for the 2955C++ Extensions for Concepts Technical Specification, ISO 19217 (2015). 2956 2957With @option{-std=c++20} and above, Concepts are part of the language 2958standard, so @option{-fconcepts} defaults to on. But the standard 2959specification of Concepts differs significantly from the TS, so some 2960constructs that were allowed in the TS but didn't make it into the 2961standard can still be enabled by @option{-fconcepts-ts}. 2962 2963@item -fconstexpr-depth=@var{n} 2964@opindex fconstexpr-depth 2965Set the maximum nested evaluation depth for C++11 constexpr functions 2966to @var{n}. A limit is needed to detect endless recursion during 2967constant expression evaluation. The minimum specified by the standard 2968is 512. 2969 2970@item -fconstexpr-cache-depth=@var{n} 2971@opindex fconstexpr-cache-depth 2972Set the maximum level of nested evaluation depth for C++11 constexpr 2973functions that will be cached to @var{n}. This is a heuristic that 2974trades off compilation speed (when the cache avoids repeated 2975calculations) against memory consumption (when the cache grows very 2976large from highly recursive evaluations). The default is 8. Very few 2977users are likely to want to adjust it, but if your code does heavy 2978constexpr calculations you might want to experiment to find which 2979value works best for you. 2980 2981@item -fconstexpr-loop-limit=@var{n} 2982@opindex fconstexpr-loop-limit 2983Set the maximum number of iterations for a loop in C++14 constexpr functions 2984to @var{n}. A limit is needed to detect infinite loops during 2985constant expression evaluation. The default is 262144 (1<<18). 2986 2987@item -fconstexpr-ops-limit=@var{n} 2988@opindex fconstexpr-ops-limit 2989Set the maximum number of operations during a single constexpr evaluation. 2990Even when number of iterations of a single loop is limited with the above limit, 2991if there are several nested loops and each of them has many iterations but still 2992smaller than the above limit, or if in a body of some loop or even outside 2993of a loop too many expressions need to be evaluated, the resulting constexpr 2994evaluation might take too long. 2995The default is 33554432 (1<<25). 2996 2997@item -fcoroutines 2998@opindex fcoroutines 2999Enable support for the C++ coroutines extension (experimental). 3000 3001@item -fno-elide-constructors 3002@opindex fno-elide-constructors 3003@opindex felide-constructors 3004The C++ standard allows an implementation to omit creating a temporary 3005that is only used to initialize another object of the same type. 3006Specifying this option disables that optimization, and forces G++ to 3007call the copy constructor in all cases. This option also causes G++ 3008to call trivial member functions which otherwise would be expanded inline. 3009 3010In C++17, the compiler is required to omit these temporaries, but this 3011option still affects trivial member functions. 3012 3013@item -fno-enforce-eh-specs 3014@opindex fno-enforce-eh-specs 3015@opindex fenforce-eh-specs 3016Don't generate code to check for violation of exception specifications 3017at run time. This option violates the C++ standard, but may be useful 3018for reducing code size in production builds, much like defining 3019@code{NDEBUG}. This does not give user code permission to throw 3020exceptions in violation of the exception specifications; the compiler 3021still optimizes based on the specifications, so throwing an 3022unexpected exception results in undefined behavior at run time. 3023 3024@item -fextern-tls-init 3025@itemx -fno-extern-tls-init 3026@opindex fextern-tls-init 3027@opindex fno-extern-tls-init 3028The C++11 and OpenMP standards allow @code{thread_local} and 3029@code{threadprivate} variables to have dynamic (runtime) 3030initialization. To support this, any use of such a variable goes 3031through a wrapper function that performs any necessary initialization. 3032When the use and definition of the variable are in the same 3033translation unit, this overhead can be optimized away, but when the 3034use is in a different translation unit there is significant overhead 3035even if the variable doesn't actually need dynamic initialization. If 3036the programmer can be sure that no use of the variable in a 3037non-defining TU needs to trigger dynamic initialization (either 3038because the variable is statically initialized, or a use of the 3039variable in the defining TU will be executed before any uses in 3040another TU), they can avoid this overhead with the 3041@option{-fno-extern-tls-init} option. 3042 3043On targets that support symbol aliases, the default is 3044@option{-fextern-tls-init}. On targets that do not support symbol 3045aliases, the default is @option{-fno-extern-tls-init}. 3046 3047@item -fno-gnu-keywords 3048@opindex fno-gnu-keywords 3049@opindex fgnu-keywords 3050Do not recognize @code{typeof} as a keyword, so that code can use this 3051word as an identifier. You can use the keyword @code{__typeof__} instead. 3052This option is implied by the strict ISO C++ dialects: @option{-ansi}, 3053@option{-std=c++98}, @option{-std=c++11}, etc. 3054 3055@item -fno-implicit-templates 3056@opindex fno-implicit-templates 3057@opindex fimplicit-templates 3058Never emit code for non-inline templates that are instantiated 3059implicitly (i.e.@: by use); only emit code for explicit instantiations. 3060If you use this option, you must take care to structure your code to 3061include all the necessary explicit instantiations to avoid getting 3062undefined symbols at link time. 3063@xref{Template Instantiation}, for more information. 3064 3065@item -fno-implicit-inline-templates 3066@opindex fno-implicit-inline-templates 3067@opindex fimplicit-inline-templates 3068Don't emit code for implicit instantiations of inline templates, either. 3069The default is to handle inlines differently so that compiles with and 3070without optimization need the same set of explicit instantiations. 3071 3072@item -fno-implement-inlines 3073@opindex fno-implement-inlines 3074@opindex fimplement-inlines 3075To save space, do not emit out-of-line copies of inline functions 3076controlled by @code{#pragma implementation}. This causes linker 3077errors if these functions are not inlined everywhere they are called. 3078 3079@item -fmodules-ts 3080@itemx -fno-modules-ts 3081@opindex fmodules-ts 3082@opindex fno-modules-ts 3083Enable support for C++20 modules (@xref{C++ Modules}). The 3084@option{-fno-modules-ts} is usually not needed, as that is the 3085default. Even though this is a C++20 feature, it is not currently 3086implicitly enabled by selecting that standard version. 3087 3088@item -fmodule-header 3089@itemx -fmodule-header=user 3090@itemx -fmodule-header=system 3091@opindex fmodule-header 3092Compile a header file to create an importable header unit. 3093 3094@item -fmodule-implicit-inline 3095@opindex fmodule-implicit-inline 3096Member functions defined in their class definitions are not implicitly 3097inline for modular code. This is different to traditional C++ 3098behavior, for good reasons. However, it may result in a difficulty 3099during code porting. This option makes such function definitions 3100implicitly inline. It does however generate an ABI incompatibility, 3101so you must use it everywhere or nowhere. (Such definitions outside 3102of a named module remain implicitly inline, regardless.) 3103 3104@item -fno-module-lazy 3105@opindex fno-module-lazy 3106@opindex fmodule-lazy 3107Disable lazy module importing and module mapper creation. 3108 3109@item -fmodule-mapper=@r{[}@var{hostname}@r{]}:@var{port}@r{[}?@var{ident}@r{]} 3110@itemx -fmodule-mapper=|@var{program}@r{[}?@var{ident}@r{]} @var{args...} 3111@itemx -fmodule-mapper==@var{socket}@r{[}?@var{ident}@r{]} 3112@itemx -fmodule-mapper=<>@r{[}@var{inout}@r{]}@r{[}?@var{ident}@r{]} 3113@itemx -fmodule-mapper=<@var{in}>@var{out}@r{[}?@var{ident}@r{]} 3114@itemx -fmodule-mapper=@var{file}@r{[}?@var{ident}@r{]} 3115@vindex CXX_MODULE_MAPPER @r{environment variable} 3116@opindex fmodule-mapper 3117An oracle to query for module name to filename mappings. If 3118unspecified the @env{CXX_MODULE_MAPPER} environment variable is used, 3119and if that is unset, an in-process default is provided. 3120 3121@item -fmodule-only 3122@opindex fmodule-only 3123Only emit the Compiled Module Interface, inhibiting any object file. 3124 3125@item -fms-extensions 3126@opindex fms-extensions 3127Disable Wpedantic warnings about constructs used in MFC, such as implicit 3128int and getting a pointer to member function via non-standard syntax. 3129 3130@item -fnew-inheriting-ctors 3131@opindex fnew-inheriting-ctors 3132Enable the P0136 adjustment to the semantics of C++11 constructor 3133inheritance. This is part of C++17 but also considered to be a Defect 3134Report against C++11 and C++14. This flag is enabled by default 3135unless @option{-fabi-version=10} or lower is specified. 3136 3137@item -fnew-ttp-matching 3138@opindex fnew-ttp-matching 3139Enable the P0522 resolution to Core issue 150, template template 3140parameters and default arguments: this allows a template with default 3141template arguments as an argument for a template template parameter 3142with fewer template parameters. This flag is enabled by default for 3143@option{-std=c++17}. 3144 3145@item -fno-nonansi-builtins 3146@opindex fno-nonansi-builtins 3147@opindex fnonansi-builtins 3148Disable built-in declarations of functions that are not mandated by 3149ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit}, 3150@code{index}, @code{bzero}, @code{conjf}, and other related functions. 3151 3152@item -fnothrow-opt 3153@opindex fnothrow-opt 3154Treat a @code{throw()} exception specification as if it were a 3155@code{noexcept} specification to reduce or eliminate the text size 3156overhead relative to a function with no exception specification. If 3157the function has local variables of types with non-trivial 3158destructors, the exception specification actually makes the 3159function smaller because the EH cleanups for those variables can be 3160optimized away. The semantic effect is that an exception thrown out of 3161a function with such an exception specification results in a call 3162to @code{terminate} rather than @code{unexpected}. 3163 3164@item -fno-operator-names 3165@opindex fno-operator-names 3166@opindex foperator-names 3167Do not treat the operator name keywords @code{and}, @code{bitand}, 3168@code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as 3169synonyms as keywords. 3170 3171@item -fno-optional-diags 3172@opindex fno-optional-diags 3173@opindex foptional-diags 3174Disable diagnostics that the standard says a compiler does not need to 3175issue. Currently, the only such diagnostic issued by G++ is the one for 3176a name having multiple meanings within a class. 3177 3178@item -fpermissive 3179@opindex fpermissive 3180Downgrade some diagnostics about nonconformant code from errors to 3181warnings. Thus, using @option{-fpermissive} allows some 3182nonconforming code to compile. 3183 3184@item -fno-pretty-templates 3185@opindex fno-pretty-templates 3186@opindex fpretty-templates 3187When an error message refers to a specialization of a function 3188template, the compiler normally prints the signature of the 3189template followed by the template arguments and any typedefs or 3190typenames in the signature (e.g.@: @code{void f(T) [with T = int]} 3191rather than @code{void f(int)}) so that it's clear which template is 3192involved. When an error message refers to a specialization of a class 3193template, the compiler omits any template arguments that match 3194the default template arguments for that template. If either of these 3195behaviors make it harder to understand the error message rather than 3196easier, you can use @option{-fno-pretty-templates} to disable them. 3197 3198@item -fno-rtti 3199@opindex fno-rtti 3200@opindex frtti 3201Disable generation of information about every class with virtual 3202functions for use by the C++ run-time type identification features 3203(@code{dynamic_cast} and @code{typeid}). If you don't use those parts 3204of the language, you can save some space by using this flag. Note that 3205exception handling uses the same information, but G++ generates it as 3206needed. The @code{dynamic_cast} operator can still be used for casts that 3207do not require run-time type information, i.e.@: casts to @code{void *} or to 3208unambiguous base classes. 3209 3210Mixing code compiled with @option{-frtti} with that compiled with 3211@option{-fno-rtti} may not work. For example, programs may 3212fail to link if a class compiled with @option{-fno-rtti} is used as a base 3213for a class compiled with @option{-frtti}. 3214 3215@item -fsized-deallocation 3216@opindex fsized-deallocation 3217Enable the built-in global declarations 3218@smallexample 3219void operator delete (void *, std::size_t) noexcept; 3220void operator delete[] (void *, std::size_t) noexcept; 3221@end smallexample 3222as introduced in C++14. This is useful for user-defined replacement 3223deallocation functions that, for example, use the size of the object 3224to make deallocation faster. Enabled by default under 3225@option{-std=c++14} and above. The flag @option{-Wsized-deallocation} 3226warns about places that might want to add a definition. 3227 3228@item -fstrict-enums 3229@opindex fstrict-enums 3230Allow the compiler to optimize using the assumption that a value of 3231enumerated type can only be one of the values of the enumeration (as 3232defined in the C++ standard; basically, a value that can be 3233represented in the minimum number of bits needed to represent all the 3234enumerators). This assumption may not be valid if the program uses a 3235cast to convert an arbitrary integer value to the enumerated type. 3236 3237@item -fstrong-eval-order 3238@opindex fstrong-eval-order 3239Evaluate member access, array subscripting, and shift expressions in 3240left-to-right order, and evaluate assignment in right-to-left order, 3241as adopted for C++17. Enabled by default with @option{-std=c++17}. 3242@option{-fstrong-eval-order=some} enables just the ordering of member 3243access and shift expressions, and is the default without 3244@option{-std=c++17}. 3245 3246@item -ftemplate-backtrace-limit=@var{n} 3247@opindex ftemplate-backtrace-limit 3248Set the maximum number of template instantiation notes for a single 3249warning or error to @var{n}. The default value is 10. 3250 3251@item -ftemplate-depth=@var{n} 3252@opindex ftemplate-depth 3253Set the maximum instantiation depth for template classes to @var{n}. 3254A limit on the template instantiation depth is needed to detect 3255endless recursions during template class instantiation. ANSI/ISO C++ 3256conforming programs must not rely on a maximum depth greater than 17 3257(changed to 1024 in C++11). The default value is 900, as the compiler 3258can run out of stack space before hitting 1024 in some situations. 3259 3260@item -fno-threadsafe-statics 3261@opindex fno-threadsafe-statics 3262@opindex fthreadsafe-statics 3263Do not emit the extra code to use the routines specified in the C++ 3264ABI for thread-safe initialization of local statics. You can use this 3265option to reduce code size slightly in code that doesn't need to be 3266thread-safe. 3267 3268@item -fuse-cxa-atexit 3269@opindex fuse-cxa-atexit 3270Register destructors for objects with static storage duration with the 3271@code{__cxa_atexit} function rather than the @code{atexit} function. 3272This option is required for fully standards-compliant handling of static 3273destructors, but only works if your C library supports 3274@code{__cxa_atexit}. 3275 3276@item -fno-use-cxa-get-exception-ptr 3277@opindex fno-use-cxa-get-exception-ptr 3278@opindex fuse-cxa-get-exception-ptr 3279Don't use the @code{__cxa_get_exception_ptr} runtime routine. This 3280causes @code{std::uncaught_exception} to be incorrect, but is necessary 3281if the runtime routine is not available. 3282 3283@item -fvisibility-inlines-hidden 3284@opindex fvisibility-inlines-hidden 3285This switch declares that the user does not attempt to compare 3286pointers to inline functions or methods where the addresses of the two functions 3287are taken in different shared objects. 3288 3289The effect of this is that GCC may, effectively, mark inline methods with 3290@code{__attribute__ ((visibility ("hidden")))} so that they do not 3291appear in the export table of a DSO and do not require a PLT indirection 3292when used within the DSO@. Enabling this option can have a dramatic effect 3293on load and link times of a DSO as it massively reduces the size of the 3294dynamic export table when the library makes heavy use of templates. 3295 3296The behavior of this switch is not quite the same as marking the 3297methods as hidden directly, because it does not affect static variables 3298local to the function or cause the compiler to deduce that 3299the function is defined in only one shared object. 3300 3301You may mark a method as having a visibility explicitly to negate the 3302effect of the switch for that method. For example, if you do want to 3303compare pointers to a particular inline method, you might mark it as 3304having default visibility. Marking the enclosing class with explicit 3305visibility has no effect. 3306 3307Explicitly instantiated inline methods are unaffected by this option 3308as their linkage might otherwise cross a shared library boundary. 3309@xref{Template Instantiation}. 3310 3311@item -fvisibility-ms-compat 3312@opindex fvisibility-ms-compat 3313This flag attempts to use visibility settings to make GCC's C++ 3314linkage model compatible with that of Microsoft Visual Studio. 3315 3316The flag makes these changes to GCC's linkage model: 3317 3318@enumerate 3319@item 3320It sets the default visibility to @code{hidden}, like 3321@option{-fvisibility=hidden}. 3322 3323@item 3324Types, but not their members, are not hidden by default. 3325 3326@item 3327The One Definition Rule is relaxed for types without explicit 3328visibility specifications that are defined in more than one 3329shared object: those declarations are permitted if they are 3330permitted when this option is not used. 3331@end enumerate 3332 3333In new code it is better to use @option{-fvisibility=hidden} and 3334export those classes that are intended to be externally visible. 3335Unfortunately it is possible for code to rely, perhaps accidentally, 3336on the Visual Studio behavior. 3337 3338Among the consequences of these changes are that static data members 3339of the same type with the same name but defined in different shared 3340objects are different, so changing one does not change the other; 3341and that pointers to function members defined in different shared 3342objects may not compare equal. When this flag is given, it is a 3343violation of the ODR to define types with the same name differently. 3344 3345@item -fno-weak 3346@opindex fno-weak 3347@opindex fweak 3348Do not use weak symbol support, even if it is provided by the linker. 3349By default, G++ uses weak symbols if they are available. This 3350option exists only for testing, and should not be used by end-users; 3351it results in inferior code and has no benefits. This option may 3352be removed in a future release of G++. 3353 3354@item -fext-numeric-literals @r{(C++ and Objective-C++ only)} 3355@opindex fext-numeric-literals 3356@opindex fno-ext-numeric-literals 3357Accept imaginary, fixed-point, or machine-defined 3358literal number suffixes as GNU extensions. 3359When this option is turned off these suffixes are treated 3360as C++11 user-defined literal numeric suffixes. 3361This is on by default for all pre-C++11 dialects and all GNU dialects: 3362@option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11}, 3363@option{-std=gnu++14}. 3364This option is off by default 3365for ISO C++11 onwards (@option{-std=c++11}, ...). 3366 3367@item -nostdinc++ 3368@opindex nostdinc++ 3369Do not search for header files in the standard directories specific to 3370C++, but do still search the other standard directories. (This option 3371is used when building the C++ library.) 3372 3373@item -flang-info-include-translate 3374@itemx -flang-info-include-translate-not 3375@itemx -flang-info-include-translate=@var{header} 3376@opindex flang-info-include-translate 3377@opindex flang-info-include-translate-not 3378Inform of include translation events. The first will note accepted 3379include translations, the second will note declined include 3380translations. The @var{header} form will inform of include 3381translations relating to that specific header. If @var{header} is of 3382the form @code{"user"} or @code{<system>} it will be resolved to a 3383specific user or system header using the include path. 3384 3385@item -flang-info-module-cmi 3386@itemx -flang-info-module-cmi=@var{module} 3387@opindex flang-info-module-cmi 3388Inform of Compiled Module Interface pathnames. The first will note 3389all read CMI pathnames. The @var{module} form will not reading a 3390specific module's CMI. @var{module} may be a named module or a 3391header-unit (the latter indicated by either being a pathname containing 3392directory separators or enclosed in @code{<>} or @code{""}). 3393 3394@item -stdlib=@var{libstdc++,libc++} 3395@opindex stdlib 3396When G++ is configured to support this option, it allows specification of 3397alternate C++ runtime libraries. Two options are available: @var{libstdc++} 3398(the default, native C++ runtime for G++) and @var{libc++} which is the 3399C++ runtime installed on some operating systems (e.g. Darwin versions from 3400Darwin11 onwards). The option switches G++ to use the headers from the 3401specified library and to emit @code{-lstdc++} or @code{-lc++} respectively, 3402when a C++ runtime is required for linking. 3403@end table 3404 3405In addition, these warning options have meanings only for C++ programs: 3406 3407@table @gcctabopt 3408@item -Wabi-tag @r{(C++ and Objective-C++ only)} 3409@opindex Wabi-tag 3410Warn when a type with an ABI tag is used in a context that does not 3411have that ABI tag. See @ref{C++ Attributes} for more information 3412about ABI tags. 3413 3414@item -Wcomma-subscript @r{(C++ and Objective-C++ only)} 3415@opindex Wcomma-subscript 3416@opindex Wno-comma-subscript 3417Warn about uses of a comma expression within a subscripting expression. 3418This usage was deprecated in C++20. However, a comma expression wrapped 3419in @code{( )} is not deprecated. Example: 3420 3421@smallexample 3422@group 3423void f(int *a, int b, int c) @{ 3424 a[b,c]; // deprecated 3425 a[(b,c)]; // OK 3426@} 3427@end group 3428@end smallexample 3429 3430Enabled by default with @option{-std=c++20}. 3431 3432@item -Wctad-maybe-unsupported @r{(C++ and Objective-C++ only)} 3433@opindex Wctad-maybe-unsupported 3434@opindex Wno-ctad-maybe-unsupported 3435Warn when performing class template argument deduction (CTAD) on a type with 3436no explicitly written deduction guides. This warning will point out cases 3437where CTAD succeeded only because the compiler synthesized the implicit 3438deduction guides, which might not be what the programmer intended. Certain 3439style guides allow CTAD only on types that specifically "opt-in"; i.e., on 3440types that are designed to support CTAD. This warning can be suppressed with 3441the following pattern: 3442 3443@smallexample 3444struct allow_ctad_t; // any name works 3445template <typename T> struct S @{ 3446 S(T) @{ @} 3447@}; 3448S(allow_ctad_t) -> S<void>; // guide with incomplete parameter type will never be considered 3449@end smallexample 3450 3451@item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)} 3452@opindex Wctor-dtor-privacy 3453@opindex Wno-ctor-dtor-privacy 3454Warn when a class seems unusable because all the constructors or 3455destructors in that class are private, and it has neither friends nor 3456public static member functions. Also warn if there are no non-private 3457methods, and there's at least one private member function that isn't 3458a constructor or destructor. 3459 3460@item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)} 3461@opindex Wdelete-non-virtual-dtor 3462@opindex Wno-delete-non-virtual-dtor 3463Warn when @code{delete} is used to destroy an instance of a class that 3464has virtual functions and non-virtual destructor. It is unsafe to delete 3465an instance of a derived class through a pointer to a base class if the 3466base class does not have a virtual destructor. This warning is enabled 3467by @option{-Wall}. 3468 3469@item -Wdeprecated-copy @r{(C++ and Objective-C++ only)} 3470@opindex Wdeprecated-copy 3471@opindex Wno-deprecated-copy 3472Warn that the implicit declaration of a copy constructor or copy 3473assignment operator is deprecated if the class has a user-provided 3474copy constructor or copy assignment operator, in C++11 and up. This 3475warning is enabled by @option{-Wextra}. With 3476@option{-Wdeprecated-copy-dtor}, also deprecate if the class has a 3477user-provided destructor. 3478 3479@item -Wno-deprecated-enum-enum-conversion @r{(C++ and Objective-C++ only)} 3480@opindex Wdeprecated-enum-enum-conversion 3481@opindex Wno-deprecated-enum-enum-conversion 3482Disable the warning about the case when the usual arithmetic conversions 3483are applied on operands where one is of enumeration type and the other is 3484of a different enumeration type. This conversion was deprecated in C++20. 3485For example: 3486 3487@smallexample 3488enum E1 @{ e @}; 3489enum E2 @{ f @}; 3490int k = f - e; 3491@end smallexample 3492 3493@option{-Wdeprecated-enum-enum-conversion} is enabled by default with 3494@option{-std=c++20}. In pre-C++20 dialects, this warning can be enabled 3495by @option{-Wenum-conversion}. 3496 3497@item -Wno-deprecated-enum-float-conversion @r{(C++ and Objective-C++ only)} 3498@opindex Wdeprecated-enum-float-conversion 3499@opindex Wno-deprecated-enum-float-conversion 3500Disable the warning about the case when the usual arithmetic conversions 3501are applied on operands where one is of enumeration type and the other is 3502of a floating-point type. This conversion was deprecated in C++20. For 3503example: 3504 3505@smallexample 3506enum E1 @{ e @}; 3507enum E2 @{ f @}; 3508bool b = e <= 3.7; 3509@end smallexample 3510 3511@option{-Wdeprecated-enum-float-conversion} is enabled by default with 3512@option{-std=c++20}. In pre-C++20 dialects, this warning can be enabled 3513by @option{-Wenum-conversion}. 3514 3515@item -Wno-init-list-lifetime @r{(C++ and Objective-C++ only)} 3516@opindex Winit-list-lifetime 3517@opindex Wno-init-list-lifetime 3518Do not warn about uses of @code{std::initializer_list} that are likely 3519to result in dangling pointers. Since the underlying array for an 3520@code{initializer_list} is handled like a normal C++ temporary object, 3521it is easy to inadvertently keep a pointer to the array past the end 3522of the array's lifetime. For example: 3523 3524@itemize @bullet 3525@item 3526If a function returns a temporary @code{initializer_list}, or a local 3527@code{initializer_list} variable, the array's lifetime ends at the end 3528of the return statement, so the value returned has a dangling pointer. 3529 3530@item 3531If a new-expression creates an @code{initializer_list}, the array only 3532lives until the end of the enclosing full-expression, so the 3533@code{initializer_list} in the heap has a dangling pointer. 3534 3535@item 3536When an @code{initializer_list} variable is assigned from a 3537brace-enclosed initializer list, the temporary array created for the 3538right side of the assignment only lives until the end of the 3539full-expression, so at the next statement the @code{initializer_list} 3540variable has a dangling pointer. 3541 3542@smallexample 3543// li's initial underlying array lives as long as li 3544std::initializer_list<int> li = @{ 1,2,3 @}; 3545// assignment changes li to point to a temporary array 3546li = @{ 4, 5 @}; 3547// now the temporary is gone and li has a dangling pointer 3548int i = li.begin()[0] // undefined behavior 3549@end smallexample 3550 3551@item 3552When a list constructor stores the @code{begin} pointer from the 3553@code{initializer_list} argument, this doesn't extend the lifetime of 3554the array, so if a class variable is constructed from a temporary 3555@code{initializer_list}, the pointer is left dangling by the end of 3556the variable declaration statement. 3557 3558@end itemize 3559 3560@item -Winvalid-imported-macros 3561@opindex Winvalid-imported-macros 3562@opindex Wno-invalid-imported-macros 3563Verify all imported macro definitions are valid at the end of 3564compilation. This is not enabled by default, as it requires 3565additional processing to determine. It may be useful when preparing 3566sets of header-units to ensure consistent macros. 3567 3568@item -Wno-literal-suffix @r{(C++ and Objective-C++ only)} 3569@opindex Wliteral-suffix 3570@opindex Wno-literal-suffix 3571Do not warn when a string or character literal is followed by a 3572ud-suffix which does not begin with an underscore. As a conforming 3573extension, GCC treats such suffixes as separate preprocessing tokens 3574in order to maintain backwards compatibility with code that uses 3575formatting macros from @code{<inttypes.h>}. For example: 3576 3577@smallexample 3578#define __STDC_FORMAT_MACROS 3579#include <inttypes.h> 3580#include <stdio.h> 3581 3582int main() @{ 3583 int64_t i64 = 123; 3584 printf("My int64: %" PRId64"\n", i64); 3585@} 3586@end smallexample 3587 3588In this case, @code{PRId64} is treated as a separate preprocessing token. 3589 3590This option also controls warnings when a user-defined literal 3591operator is declared with a literal suffix identifier that doesn't 3592begin with an underscore. Literal suffix identifiers that don't begin 3593with an underscore are reserved for future standardization. 3594 3595These warnings are enabled by default. 3596 3597@item -Wno-narrowing @r{(C++ and Objective-C++ only)} 3598@opindex Wnarrowing 3599@opindex Wno-narrowing 3600For C++11 and later standards, narrowing conversions are diagnosed by default, 3601as required by the standard. A narrowing conversion from a constant produces 3602an error, and a narrowing conversion from a non-constant produces a warning, 3603but @option{-Wno-narrowing} suppresses the diagnostic. 3604Note that this does not affect the meaning of well-formed code; 3605narrowing conversions are still considered ill-formed in SFINAE contexts. 3606 3607With @option{-Wnarrowing} in C++98, warn when a narrowing 3608conversion prohibited by C++11 occurs within 3609@samp{@{ @}}, e.g. 3610 3611@smallexample 3612int i = @{ 2.2 @}; // error: narrowing from double to int 3613@end smallexample 3614 3615This flag is included in @option{-Wall} and @option{-Wc++11-compat}. 3616 3617@item -Wnoexcept @r{(C++ and Objective-C++ only)} 3618@opindex Wnoexcept 3619@opindex Wno-noexcept 3620Warn when a noexcept-expression evaluates to false because of a call 3621to a function that does not have a non-throwing exception 3622specification (i.e. @code{throw()} or @code{noexcept}) but is known by 3623the compiler to never throw an exception. 3624 3625@item -Wnoexcept-type @r{(C++ and Objective-C++ only)} 3626@opindex Wnoexcept-type 3627@opindex Wno-noexcept-type 3628Warn if the C++17 feature making @code{noexcept} part of a function 3629type changes the mangled name of a symbol relative to C++14. Enabled 3630by @option{-Wabi} and @option{-Wc++17-compat}. 3631 3632As an example: 3633 3634@smallexample 3635template <class T> void f(T t) @{ t(); @}; 3636void g() noexcept; 3637void h() @{ f(g); @} 3638@end smallexample 3639 3640@noindent 3641In C++14, @code{f} calls @code{f<void(*)()>}, but in 3642C++17 it calls @code{f<void(*)()noexcept>}. 3643 3644@item -Wclass-memaccess @r{(C++ and Objective-C++ only)} 3645@opindex Wclass-memaccess 3646@opindex Wno-class-memaccess 3647Warn when the destination of a call to a raw memory function such as 3648@code{memset} or @code{memcpy} is an object of class type, and when writing 3649into such an object might bypass the class non-trivial or deleted constructor 3650or copy assignment, violate const-correctness or encapsulation, or corrupt 3651virtual table pointers. Modifying the representation of such objects may 3652violate invariants maintained by member functions of the class. For example, 3653the call to @code{memset} below is undefined because it modifies a non-trivial 3654class object and is, therefore, diagnosed. The safe way to either initialize 3655or clear the storage of objects of such types is by using the appropriate 3656constructor or assignment operator, if one is available. 3657@smallexample 3658std::string str = "abc"; 3659memset (&str, 0, sizeof str); 3660@end smallexample 3661The @option{-Wclass-memaccess} option is enabled by @option{-Wall}. 3662Explicitly casting the pointer to the class object to @code{void *} or 3663to a type that can be safely accessed by the raw memory function suppresses 3664the warning. 3665 3666@item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)} 3667@opindex Wnon-virtual-dtor 3668@opindex Wno-non-virtual-dtor 3669Warn when a class has virtual functions and an accessible non-virtual 3670destructor itself or in an accessible polymorphic base class, in which 3671case it is possible but unsafe to delete an instance of a derived 3672class through a pointer to the class itself or base class. This 3673warning is automatically enabled if @option{-Weffc++} is specified. 3674 3675@item -Wregister @r{(C++ and Objective-C++ only)} 3676@opindex Wregister 3677@opindex Wno-register 3678Warn on uses of the @code{register} storage class specifier, except 3679when it is part of the GNU @ref{Explicit Register Variables} extension. 3680The use of the @code{register} keyword as storage class specifier has 3681been deprecated in C++11 and removed in C++17. 3682Enabled by default with @option{-std=c++17}. 3683 3684@item -Wreorder @r{(C++ and Objective-C++ only)} 3685@opindex Wreorder 3686@opindex Wno-reorder 3687@cindex reordering, warning 3688@cindex warning for reordering of member initializers 3689Warn when the order of member initializers given in the code does not 3690match the order in which they must be executed. For instance: 3691 3692@smallexample 3693struct A @{ 3694 int i; 3695 int j; 3696 A(): j (0), i (1) @{ @} 3697@}; 3698@end smallexample 3699 3700@noindent 3701The compiler rearranges the member initializers for @code{i} 3702and @code{j} to match the declaration order of the members, emitting 3703a warning to that effect. This warning is enabled by @option{-Wall}. 3704 3705@item -Wno-pessimizing-move @r{(C++ and Objective-C++ only)} 3706@opindex Wpessimizing-move 3707@opindex Wno-pessimizing-move 3708This warning warns when a call to @code{std::move} prevents copy 3709elision. A typical scenario when copy elision can occur is when returning in 3710a function with a class return type, when the expression being returned is the 3711name of a non-volatile automatic object, and is not a function parameter, and 3712has the same type as the function return type. 3713 3714@smallexample 3715struct T @{ 3716@dots{} 3717@}; 3718T fn() 3719@{ 3720 T t; 3721 @dots{} 3722 return std::move (t); 3723@} 3724@end smallexample 3725 3726But in this example, the @code{std::move} call prevents copy elision. 3727 3728This warning is enabled by @option{-Wall}. 3729 3730@item -Wno-redundant-move @r{(C++ and Objective-C++ only)} 3731@opindex Wredundant-move 3732@opindex Wno-redundant-move 3733This warning warns about redundant calls to @code{std::move}; that is, when 3734a move operation would have been performed even without the @code{std::move} 3735call. This happens because the compiler is forced to treat the object as if 3736it were an rvalue in certain situations such as returning a local variable, 3737where copy elision isn't applicable. Consider: 3738 3739@smallexample 3740struct T @{ 3741@dots{} 3742@}; 3743T fn(T t) 3744@{ 3745 @dots{} 3746 return std::move (t); 3747@} 3748@end smallexample 3749 3750Here, the @code{std::move} call is redundant. Because G++ implements Core 3751Issue 1579, another example is: 3752 3753@smallexample 3754struct T @{ // convertible to U 3755@dots{} 3756@}; 3757struct U @{ 3758@dots{} 3759@}; 3760U fn() 3761@{ 3762 T t; 3763 @dots{} 3764 return std::move (t); 3765@} 3766@end smallexample 3767In this example, copy elision isn't applicable because the type of the 3768expression being returned and the function return type differ, yet G++ 3769treats the return value as if it were designated by an rvalue. 3770 3771This warning is enabled by @option{-Wextra}. 3772 3773@item -Wrange-loop-construct @r{(C++ and Objective-C++ only)} 3774@opindex Wrange-loop-construct 3775@opindex Wno-range-loop-construct 3776This warning warns when a C++ range-based for-loop is creating an unnecessary 3777copy. This can happen when the range declaration is not a reference, but 3778probably should be. For example: 3779 3780@smallexample 3781struct S @{ char arr[128]; @}; 3782void fn () @{ 3783 S arr[5]; 3784 for (const auto x : arr) @{ @dots{} @} 3785@} 3786@end smallexample 3787 3788It does not warn when the type being copied is a trivially-copyable type whose 3789size is less than 64 bytes. 3790 3791This warning also warns when a loop variable in a range-based for-loop is 3792initialized with a value of a different type resulting in a copy. For example: 3793 3794@smallexample 3795void fn() @{ 3796 int arr[10]; 3797 for (const double &x : arr) @{ @dots{} @} 3798@} 3799@end smallexample 3800 3801In the example above, in every iteration of the loop a temporary value of 3802type @code{double} is created and destroyed, to which the reference 3803@code{const double &} is bound. 3804 3805This warning is enabled by @option{-Wall}. 3806 3807@item -Wredundant-tags @r{(C++ and Objective-C++ only)} 3808@opindex Wredundant-tags 3809@opindex Wno-redundant-tags 3810Warn about redundant class-key and enum-key in references to class types 3811and enumerated types in contexts where the key can be eliminated without 3812causing an ambiguity. For example: 3813 3814@smallexample 3815struct foo; 3816struct foo *p; // warn that keyword struct can be eliminated 3817@end smallexample 3818 3819@noindent 3820On the other hand, in this example there is no warning: 3821 3822@smallexample 3823struct foo; 3824void foo (); // "hides" struct foo 3825void bar (struct foo&); // no warning, keyword struct is necessary 3826@end smallexample 3827 3828@item -Wno-subobject-linkage @r{(C++ and Objective-C++ only)} 3829@opindex Wsubobject-linkage 3830@opindex Wno-subobject-linkage 3831Do not warn 3832if a class type has a base or a field whose type uses the anonymous 3833namespace or depends on a type with no linkage. If a type A depends on 3834a type B with no or internal linkage, defining it in multiple 3835translation units would be an ODR violation because the meaning of B 3836is different in each translation unit. If A only appears in a single 3837translation unit, the best way to silence the warning is to give it 3838internal linkage by putting it in an anonymous namespace as well. The 3839compiler doesn't give this warning for types defined in the main .C 3840file, as those are unlikely to have multiple definitions. 3841@option{-Wsubobject-linkage} is enabled by default. 3842 3843@item -Weffc++ @r{(C++ and Objective-C++ only)} 3844@opindex Weffc++ 3845@opindex Wno-effc++ 3846Warn about violations of the following style guidelines from Scott Meyers' 3847@cite{Effective C++} series of books: 3848 3849@itemize @bullet 3850@item 3851Define a copy constructor and an assignment operator for classes 3852with dynamically-allocated memory. 3853 3854@item 3855Prefer initialization to assignment in constructors. 3856 3857@item 3858Have @code{operator=} return a reference to @code{*this}. 3859 3860@item 3861Don't try to return a reference when you must return an object. 3862 3863@item 3864Distinguish between prefix and postfix forms of increment and 3865decrement operators. 3866 3867@item 3868Never overload @code{&&}, @code{||}, or @code{,}. 3869 3870@end itemize 3871 3872This option also enables @option{-Wnon-virtual-dtor}, which is also 3873one of the effective C++ recommendations. However, the check is 3874extended to warn about the lack of virtual destructor in accessible 3875non-polymorphic bases classes too. 3876 3877When selecting this option, be aware that the standard library 3878headers do not obey all of these guidelines; use @samp{grep -v} 3879to filter out those warnings. 3880 3881@item -Wno-exceptions @r{(C++ and Objective-C++ only)} 3882@opindex Wexceptions 3883@opindex Wno-exceptions 3884Disable the warning about the case when an exception handler is shadowed by 3885another handler, which can point out a wrong ordering of exception handlers. 3886 3887@item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)} 3888@opindex Wstrict-null-sentinel 3889@opindex Wno-strict-null-sentinel 3890Warn about the use of an uncasted @code{NULL} as sentinel. When 3891compiling only with GCC this is a valid sentinel, as @code{NULL} is defined 3892to @code{__null}. Although it is a null pointer constant rather than a 3893null pointer, it is guaranteed to be of the same size as a pointer. 3894But this use is not portable across different compilers. 3895 3896@item -Wno-non-template-friend @r{(C++ and Objective-C++ only)} 3897@opindex Wno-non-template-friend 3898@opindex Wnon-template-friend 3899Disable warnings when non-template friend functions are declared 3900within a template. In very old versions of GCC that predate implementation 3901of the ISO standard, declarations such as 3902@samp{friend int foo(int)}, where the name of the friend is an unqualified-id, 3903could be interpreted as a particular specialization of a template 3904function; the warning exists to diagnose compatibility problems, 3905and is enabled by default. 3906 3907@item -Wold-style-cast @r{(C++ and Objective-C++ only)} 3908@opindex Wold-style-cast 3909@opindex Wno-old-style-cast 3910Warn if an old-style (C-style) cast to a non-void type is used within 3911a C++ program. The new-style casts (@code{dynamic_cast}, 3912@code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are 3913less vulnerable to unintended effects and much easier to search for. 3914 3915@item -Woverloaded-virtual @r{(C++ and Objective-C++ only)} 3916@opindex Woverloaded-virtual 3917@opindex Wno-overloaded-virtual 3918@cindex overloaded virtual function, warning 3919@cindex warning for overloaded virtual function 3920Warn when a function declaration hides virtual functions from a 3921base class. For example, in: 3922 3923@smallexample 3924struct A @{ 3925 virtual void f(); 3926@}; 3927 3928struct B: public A @{ 3929 void f(int); 3930@}; 3931@end smallexample 3932 3933the @code{A} class version of @code{f} is hidden in @code{B}, and code 3934like: 3935 3936@smallexample 3937B* b; 3938b->f(); 3939@end smallexample 3940 3941@noindent 3942fails to compile. 3943 3944@item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)} 3945@opindex Wno-pmf-conversions 3946@opindex Wpmf-conversions 3947Disable the diagnostic for converting a bound pointer to member function 3948to a plain pointer. 3949 3950@item -Wsign-promo @r{(C++ and Objective-C++ only)} 3951@opindex Wsign-promo 3952@opindex Wno-sign-promo 3953Warn when overload resolution chooses a promotion from unsigned or 3954enumerated type to a signed type, over a conversion to an unsigned type of 3955the same size. Previous versions of G++ tried to preserve 3956unsignedness, but the standard mandates the current behavior. 3957 3958@item -Wtemplates @r{(C++ and Objective-C++ only)} 3959@opindex Wtemplates 3960@opindex Wno-templates 3961Warn when a primary template declaration is encountered. Some coding 3962rules disallow templates, and this may be used to enforce that rule. 3963The warning is inactive inside a system header file, such as the STL, so 3964one can still use the STL. One may also instantiate or specialize 3965templates. 3966 3967@item -Wno-mismatched-new-delete @r{(C++ and Objective-C++ only)} 3968@opindex Wmismatched-new-delete 3969@opindex Wno-mismatched-new-delete 3970Warn for mismatches between calls to @code{operator new} or @code{operator 3971delete} and the corresponding call to the allocation or deallocation function. 3972This includes invocations of C++ @code{operator delete} with pointers 3973returned from either mismatched forms of @code{operator new}, or from other 3974functions that allocate objects for which the @code{operator delete} isn't 3975a suitable deallocator, as well as calls to other deallocation functions 3976with pointers returned from @code{operator new} for which the deallocation 3977function isn't suitable. 3978 3979For example, the @code{delete} expression in the function below is diagnosed 3980because it doesn't match the array form of the @code{new} expression 3981the pointer argument was returned from. Similarly, the call to @code{free} 3982is also diagnosed. 3983 3984@smallexample 3985void f () 3986@{ 3987 int *a = new int[n]; 3988 delete a; // warning: mismatch in array forms of expressions 3989 3990 char *p = new char[n]; 3991 free (p); // warning: mismatch between new and free 3992@} 3993@end smallexample 3994 3995The related option @option{-Wmismatched-dealloc} diagnoses mismatches 3996involving allocation and deallocation functions other than @code{operator 3997new} and @code{operator delete}. 3998 3999@option{-Wmismatched-new-delete} is enabled by default. 4000 4001@item -Wmismatched-tags @r{(C++ and Objective-C++ only)} 4002@opindex Wmismatched-tags 4003@opindex Wno-mismatched-tags 4004Warn for declarations of structs, classes, and class templates and their 4005specializations with a class-key that does not match either the definition 4006or the first declaration if no definition is provided. 4007 4008For example, the declaration of @code{struct Object} in the argument list 4009of @code{draw} triggers the warning. To avoid it, either remove the redundant 4010class-key @code{struct} or replace it with @code{class} to match its definition. 4011@smallexample 4012class Object @{ 4013public: 4014 virtual ~Object () = 0; 4015@}; 4016void draw (struct Object*); 4017@end smallexample 4018 4019It is not wrong to declare a class with the class-key @code{struct} as 4020the example above shows. The @option{-Wmismatched-tags} option is intended 4021to help achieve a consistent style of class declarations. In code that is 4022intended to be portable to Windows-based compilers the warning helps prevent 4023unresolved references due to the difference in the mangling of symbols 4024declared with different class-keys. The option can be used either on its 4025own or in conjunction with @option{-Wredundant-tags}. 4026 4027@item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)} 4028@opindex Wmultiple-inheritance 4029@opindex Wno-multiple-inheritance 4030Warn when a class is defined with multiple direct base classes. Some 4031coding rules disallow multiple inheritance, and this may be used to 4032enforce that rule. The warning is inactive inside a system header file, 4033such as the STL, so one can still use the STL. One may also define 4034classes that indirectly use multiple inheritance. 4035 4036@item -Wvirtual-inheritance 4037@opindex Wvirtual-inheritance 4038@opindex Wno-virtual-inheritance 4039Warn when a class is defined with a virtual direct base class. Some 4040coding rules disallow multiple inheritance, and this may be used to 4041enforce that rule. The warning is inactive inside a system header file, 4042such as the STL, so one can still use the STL. One may also define 4043classes that indirectly use virtual inheritance. 4044 4045@item -Wno-virtual-move-assign 4046@opindex Wvirtual-move-assign 4047@opindex Wno-virtual-move-assign 4048Suppress warnings about inheriting from a virtual base with a 4049non-trivial C++11 move assignment operator. This is dangerous because 4050if the virtual base is reachable along more than one path, it is 4051moved multiple times, which can mean both objects end up in the 4052moved-from state. If the move assignment operator is written to avoid 4053moving from a moved-from object, this warning can be disabled. 4054 4055@item -Wnamespaces 4056@opindex Wnamespaces 4057@opindex Wno-namespaces 4058Warn when a namespace definition is opened. Some coding rules disallow 4059namespaces, and this may be used to enforce that rule. The warning is 4060inactive inside a system header file, such as the STL, so one can still 4061use the STL. One may also use using directives and qualified names. 4062 4063@item -Wno-terminate @r{(C++ and Objective-C++ only)} 4064@opindex Wterminate 4065@opindex Wno-terminate 4066Disable the warning about a throw-expression that will immediately 4067result in a call to @code{terminate}. 4068 4069@item -Wno-vexing-parse @r{(C++ and Objective-C++ only)} 4070@opindex Wvexing-parse 4071@opindex Wno-vexing-parse 4072Warn about the most vexing parse syntactic ambiguity. This warns about 4073the cases when a declaration looks like a variable definition, but the 4074C++ language requires it to be interpreted as a function declaration. 4075For instance: 4076 4077@smallexample 4078void f(double a) @{ 4079 int i(); // extern int i (void); 4080 int n(int(a)); // extern int n (int); 4081@} 4082@end smallexample 4083 4084Another example: 4085 4086@smallexample 4087struct S @{ S(int); @}; 4088void f(double a) @{ 4089 S x(int(a)); // extern struct S x (int); 4090 S y(int()); // extern struct S y (int (*) (void)); 4091 S z(); // extern struct S z (void); 4092@} 4093@end smallexample 4094 4095The warning will suggest options how to deal with such an ambiguity; e.g., 4096it can suggest removing the parentheses or using braces instead. 4097 4098This warning is enabled by default. 4099 4100@item -Wno-class-conversion @r{(C++ and Objective-C++ only)} 4101@opindex Wno-class-conversion 4102@opindex Wclass-conversion 4103Do not warn when a conversion function converts an 4104object to the same type, to a base class of that type, or to void; such 4105a conversion function will never be called. 4106 4107@item -Wvolatile @r{(C++ and Objective-C++ only)} 4108@opindex Wvolatile 4109@opindex Wno-volatile 4110Warn about deprecated uses of the @code{volatile} qualifier. This includes 4111postfix and prefix @code{++} and @code{--} expressions of 4112@code{volatile}-qualified types, using simple assignments where the left 4113operand is a @code{volatile}-qualified non-class type for their value, 4114compound assignments where the left operand is a @code{volatile}-qualified 4115non-class type, @code{volatile}-qualified function return type, 4116@code{volatile}-qualified parameter type, and structured bindings of a 4117@code{volatile}-qualified type. This usage was deprecated in C++20. 4118 4119Enabled by default with @option{-std=c++20}. 4120 4121@item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)} 4122@opindex Wzero-as-null-pointer-constant 4123@opindex Wno-zero-as-null-pointer-constant 4124Warn when a literal @samp{0} is used as null pointer constant. This can 4125be useful to facilitate the conversion to @code{nullptr} in C++11. 4126 4127@item -Waligned-new 4128@opindex Waligned-new 4129@opindex Wno-aligned-new 4130Warn about a new-expression of a type that requires greater alignment 4131than the @code{alignof(std::max_align_t)} but uses an allocation 4132function without an explicit alignment parameter. This option is 4133enabled by @option{-Wall}. 4134 4135Normally this only warns about global allocation functions, but 4136@option{-Waligned-new=all} also warns about class member allocation 4137functions. 4138 4139@item -Wno-placement-new 4140@itemx -Wplacement-new=@var{n} 4141@opindex Wplacement-new 4142@opindex Wno-placement-new 4143Warn about placement new expressions with undefined behavior, such as 4144constructing an object in a buffer that is smaller than the type of 4145the object. For example, the placement new expression below is diagnosed 4146because it attempts to construct an array of 64 integers in a buffer only 414764 bytes large. 4148@smallexample 4149char buf [64]; 4150new (buf) int[64]; 4151@end smallexample 4152This warning is enabled by default. 4153 4154@table @gcctabopt 4155@item -Wplacement-new=1 4156This is the default warning level of @option{-Wplacement-new}. At this 4157level the warning is not issued for some strictly undefined constructs that 4158GCC allows as extensions for compatibility with legacy code. For example, 4159the following @code{new} expression is not diagnosed at this level even 4160though it has undefined behavior according to the C++ standard because 4161it writes past the end of the one-element array. 4162@smallexample 4163struct S @{ int n, a[1]; @}; 4164S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]); 4165new (s->a)int [32](); 4166@end smallexample 4167 4168@item -Wplacement-new=2 4169At this level, in addition to diagnosing all the same constructs as at level 41701, a diagnostic is also issued for placement new expressions that construct 4171an object in the last member of structure whose type is an array of a single 4172element and whose size is less than the size of the object being constructed. 4173While the previous example would be diagnosed, the following construct makes 4174use of the flexible member array extension to avoid the warning at level 2. 4175@smallexample 4176struct S @{ int n, a[]; @}; 4177S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]); 4178new (s->a)int [32](); 4179@end smallexample 4180 4181@end table 4182 4183@item -Wcatch-value 4184@itemx -Wcatch-value=@var{n} @r{(C++ and Objective-C++ only)} 4185@opindex Wcatch-value 4186@opindex Wno-catch-value 4187Warn about catch handlers that do not catch via reference. 4188With @option{-Wcatch-value=1} (or @option{-Wcatch-value} for short) 4189warn about polymorphic class types that are caught by value. 4190With @option{-Wcatch-value=2} warn about all class types that are caught 4191by value. With @option{-Wcatch-value=3} warn about all types that are 4192not caught by reference. @option{-Wcatch-value} is enabled by @option{-Wall}. 4193 4194@item -Wconditionally-supported @r{(C++ and Objective-C++ only)} 4195@opindex Wconditionally-supported 4196@opindex Wno-conditionally-supported 4197Warn for conditionally-supported (C++11 [intro.defs]) constructs. 4198 4199@item -Wno-delete-incomplete @r{(C++ and Objective-C++ only)} 4200@opindex Wdelete-incomplete 4201@opindex Wno-delete-incomplete 4202Do not warn when deleting a pointer to incomplete type, which may cause 4203undefined behavior at runtime. This warning is enabled by default. 4204 4205@item -Wextra-semi @r{(C++, Objective-C++ only)} 4206@opindex Wextra-semi 4207@opindex Wno-extra-semi 4208Warn about redundant semicolons after in-class function definitions. 4209 4210@item -Wno-inaccessible-base @r{(C++, Objective-C++ only)} 4211@opindex Winaccessible-base 4212@opindex Wno-inaccessible-base 4213This option controls warnings 4214when a base class is inaccessible in a class derived from it due to 4215ambiguity. The warning is enabled by default. 4216Note that the warning for ambiguous virtual 4217bases is enabled by the @option{-Wextra} option. 4218@smallexample 4219@group 4220struct A @{ int a; @}; 4221 4222struct B : A @{ @}; 4223 4224struct C : B, A @{ @}; 4225@end group 4226@end smallexample 4227 4228@item -Wno-inherited-variadic-ctor 4229@opindex Winherited-variadic-ctor 4230@opindex Wno-inherited-variadic-ctor 4231Suppress warnings about use of C++11 inheriting constructors when the 4232base class inherited from has a C variadic constructor; the warning is 4233on by default because the ellipsis is not inherited. 4234 4235@item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)} 4236@opindex Wno-invalid-offsetof 4237@opindex Winvalid-offsetof 4238Suppress warnings from applying the @code{offsetof} macro to a non-POD 4239type. According to the 2014 ISO C++ standard, applying @code{offsetof} 4240to a non-standard-layout type is undefined. In existing C++ implementations, 4241however, @code{offsetof} typically gives meaningful results. 4242This flag is for users who are aware that they are 4243writing nonportable code and who have deliberately chosen to ignore the 4244warning about it. 4245 4246The restrictions on @code{offsetof} may be relaxed in a future version 4247of the C++ standard. 4248 4249@item -Wsized-deallocation @r{(C++ and Objective-C++ only)} 4250@opindex Wsized-deallocation 4251@opindex Wno-sized-deallocation 4252Warn about a definition of an unsized deallocation function 4253@smallexample 4254void operator delete (void *) noexcept; 4255void operator delete[] (void *) noexcept; 4256@end smallexample 4257without a definition of the corresponding sized deallocation function 4258@smallexample 4259void operator delete (void *, std::size_t) noexcept; 4260void operator delete[] (void *, std::size_t) noexcept; 4261@end smallexample 4262or vice versa. Enabled by @option{-Wextra} along with 4263@option{-fsized-deallocation}. 4264 4265@item -Wsuggest-final-types 4266@opindex Wno-suggest-final-types 4267@opindex Wsuggest-final-types 4268Warn about types with virtual methods where code quality would be improved 4269if the type were declared with the C++11 @code{final} specifier, 4270or, if possible, 4271declared in an anonymous namespace. This allows GCC to more aggressively 4272devirtualize the polymorphic calls. This warning is more effective with 4273link-time optimization, 4274where the information about the class hierarchy graph is 4275more complete. 4276 4277@item -Wsuggest-final-methods 4278@opindex Wno-suggest-final-methods 4279@opindex Wsuggest-final-methods 4280Warn about virtual methods where code quality would be improved if the method 4281were declared with the C++11 @code{final} specifier, 4282or, if possible, its type were 4283declared in an anonymous namespace or with the @code{final} specifier. 4284This warning is 4285more effective with link-time optimization, where the information about the 4286class hierarchy graph is more complete. It is recommended to first consider 4287suggestions of @option{-Wsuggest-final-types} and then rebuild with new 4288annotations. 4289 4290@item -Wsuggest-override 4291@opindex Wsuggest-override 4292@opindex Wno-suggest-override 4293Warn about overriding virtual functions that are not marked with the 4294@code{override} keyword. 4295 4296@item -Wuseless-cast @r{(C++ and Objective-C++ only)} 4297@opindex Wuseless-cast 4298@opindex Wno-useless-cast 4299Warn when an expression is casted to its own type. 4300 4301@item -Wno-conversion-null @r{(C++ and Objective-C++ only)} 4302@opindex Wconversion-null 4303@opindex Wno-conversion-null 4304Do not warn for conversions between @code{NULL} and non-pointer 4305types. @option{-Wconversion-null} is enabled by default. 4306 4307@end table 4308 4309@node Objective-C and Objective-C++ Dialect Options 4310@section Options Controlling Objective-C and Objective-C++ Dialects 4311 4312@cindex compiler options, Objective-C and Objective-C++ 4313@cindex Objective-C and Objective-C++ options, command-line 4314@cindex options, Objective-C and Objective-C++ 4315(NOTE: This manual does not describe the Objective-C and Objective-C++ 4316languages themselves. @xref{Standards,,Language Standards 4317Supported by GCC}, for references.) 4318 4319This section describes the command-line options that are only meaningful 4320for Objective-C and Objective-C++ programs. You can also use most of 4321the language-independent GNU compiler options. 4322For example, you might compile a file @file{some_class.m} like this: 4323 4324@smallexample 4325gcc -g -fgnu-runtime -O -c some_class.m 4326@end smallexample 4327 4328@noindent 4329In this example, @option{-fgnu-runtime} is an option meant only for 4330Objective-C and Objective-C++ programs; you can use the other options with 4331any language supported by GCC@. 4332 4333Note that since Objective-C is an extension of the C language, Objective-C 4334compilations may also use options specific to the C front-end (e.g., 4335@option{-Wtraditional}). Similarly, Objective-C++ compilations may use 4336C++-specific options (e.g., @option{-Wabi}). 4337 4338Here is a list of options that are @emph{only} for compiling Objective-C 4339and Objective-C++ programs: 4340 4341@table @gcctabopt 4342@item -fconstant-string-class=@var{class-name} 4343@opindex fconstant-string-class 4344Use @var{class-name} as the name of the class to instantiate for each 4345literal string specified with the syntax @code{@@"@dots{}"}. The default 4346class name is @code{NXConstantString} if the GNU runtime is being used, and 4347@code{NSConstantString} if the NeXT runtime is being used (see below). The 4348@option{-fconstant-cfstrings} option, if also present, overrides the 4349@option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals 4350to be laid out as constant CoreFoundation strings. 4351 4352@item -fgnu-runtime 4353@opindex fgnu-runtime 4354Generate object code compatible with the standard GNU Objective-C 4355runtime. This is the default for most types of systems. 4356 4357@item -fnext-runtime 4358@opindex fnext-runtime 4359Generate output compatible with the NeXT runtime. This is the default 4360for NeXT-based systems, including Darwin and Mac OS X@. The macro 4361@code{__NEXT_RUNTIME__} is predefined if (and only if) this option is 4362used. 4363 4364@item -fno-nil-receivers 4365@opindex fno-nil-receivers 4366@opindex fnil-receivers 4367Assume that all Objective-C message dispatches (@code{[receiver 4368message:arg]}) in this translation unit ensure that the receiver is 4369not @code{nil}. This allows for more efficient entry points in the 4370runtime to be used. This option is only available in conjunction with 4371the NeXT runtime and ABI version 0 or 1. 4372 4373@item -fobjc-abi-version=@var{n} 4374@opindex fobjc-abi-version 4375Use version @var{n} of the Objective-C ABI for the selected runtime. 4376This option is currently supported only for the NeXT runtime. In that 4377case, Version 0 is the traditional (32-bit) ABI without support for 4378properties and other Objective-C 2.0 additions. Version 1 is the 4379traditional (32-bit) ABI with support for properties and other 4380Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If 4381nothing is specified, the default is Version 0 on 32-bit target 4382machines, and Version 2 on 64-bit target machines. 4383 4384@item -fobjc-call-cxx-cdtors 4385@opindex fobjc-call-cxx-cdtors 4386For each Objective-C class, check if any of its instance variables is a 4387C++ object with a non-trivial default constructor. If so, synthesize a 4388special @code{- (id) .cxx_construct} instance method which runs 4389non-trivial default constructors on any such instance variables, in order, 4390and then return @code{self}. Similarly, check if any instance variable 4391is a C++ object with a non-trivial destructor, and if so, synthesize a 4392special @code{- (void) .cxx_destruct} method which runs 4393all such default destructors, in reverse order. 4394 4395The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct} 4396methods thusly generated only operate on instance variables 4397declared in the current Objective-C class, and not those inherited 4398from superclasses. It is the responsibility of the Objective-C 4399runtime to invoke all such methods in an object's inheritance 4400hierarchy. The @code{- (id) .cxx_construct} methods are invoked 4401by the runtime immediately after a new object instance is allocated; 4402the @code{- (void) .cxx_destruct} methods are invoked immediately 4403before the runtime deallocates an object instance. 4404 4405As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has 4406support for invoking the @code{- (id) .cxx_construct} and 4407@code{- (void) .cxx_destruct} methods. 4408 4409@item -fobjc-direct-dispatch 4410@opindex fobjc-direct-dispatch 4411Allow fast jumps to the message dispatcher. On Darwin this is 4412accomplished via the comm page. 4413 4414@item -fobjc-exceptions 4415@opindex fobjc-exceptions 4416Enable syntactic support for structured exception handling in 4417Objective-C, similar to what is offered by C++. This option 4418is required to use the Objective-C keywords @code{@@try}, 4419@code{@@throw}, @code{@@catch}, @code{@@finally} and 4420@code{@@synchronized}. This option is available with both the GNU 4421runtime and the NeXT runtime (but not available in conjunction with 4422the NeXT runtime on Mac OS X 10.2 and earlier). 4423 4424@item -fobjc-gc 4425@opindex fobjc-gc 4426Enable garbage collection (GC) in Objective-C and Objective-C++ 4427programs. This option is only available with the NeXT runtime; the 4428GNU runtime has a different garbage collection implementation that 4429does not require special compiler flags. 4430 4431@item -fobjc-nilcheck 4432@opindex fobjc-nilcheck 4433For the NeXT runtime with version 2 of the ABI, check for a nil 4434receiver in method invocations before doing the actual method call. 4435This is the default and can be disabled using 4436@option{-fno-objc-nilcheck}. Class methods and super calls are never 4437checked for nil in this way no matter what this flag is set to. 4438Currently this flag does nothing when the GNU runtime, or an older 4439version of the NeXT runtime ABI, is used. 4440 4441@item -fobjc-std=objc1 4442@opindex fobjc-std 4443Conform to the language syntax of Objective-C 1.0, the language 4444recognized by GCC 4.0. This only affects the Objective-C additions to 4445the C/C++ language; it does not affect conformance to C/C++ standards, 4446which is controlled by the separate C/C++ dialect option flags. When 4447this option is used with the Objective-C or Objective-C++ compiler, 4448any Objective-C syntax that is not recognized by GCC 4.0 is rejected. 4449This is useful if you need to make sure that your Objective-C code can 4450be compiled with older versions of GCC@. 4451 4452@item -freplace-objc-classes 4453@opindex freplace-objc-classes 4454Emit a special marker instructing @command{ld(1)} not to statically link in 4455the resulting object file, and allow @command{dyld(1)} to load it in at 4456run time instead. This is used in conjunction with the Fix-and-Continue 4457debugging mode, where the object file in question may be recompiled and 4458dynamically reloaded in the course of program execution, without the need 4459to restart the program itself. Currently, Fix-and-Continue functionality 4460is only available in conjunction with the NeXT runtime on Mac OS X 10.3 4461and later. 4462 4463@item -fzero-link 4464@opindex fzero-link 4465When compiling for the NeXT runtime, the compiler ordinarily replaces calls 4466to @code{objc_getClass("@dots{}")} (when the name of the class is known at 4467compile time) with static class references that get initialized at load time, 4468which improves run-time performance. Specifying the @option{-fzero-link} flag 4469suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")} 4470to be retained. This is useful in Zero-Link debugging mode, since it allows 4471for individual class implementations to be modified during program execution. 4472The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")} 4473regardless of command-line options. 4474 4475@item -fno-local-ivars 4476@opindex fno-local-ivars 4477@opindex flocal-ivars 4478By default instance variables in Objective-C can be accessed as if 4479they were local variables from within the methods of the class they're 4480declared in. This can lead to shadowing between instance variables 4481and other variables declared either locally inside a class method or 4482globally with the same name. Specifying the @option{-fno-local-ivars} 4483flag disables this behavior thus avoiding variable shadowing issues. 4484 4485@item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} 4486@opindex fivar-visibility 4487Set the default instance variable visibility to the specified option 4488so that instance variables declared outside the scope of any access 4489modifier directives default to the specified visibility. 4490 4491@item -gen-decls 4492@opindex gen-decls 4493Dump interface declarations for all classes seen in the source file to a 4494file named @file{@var{sourcename}.decl}. 4495 4496@item -Wassign-intercept @r{(Objective-C and Objective-C++ only)} 4497@opindex Wassign-intercept 4498@opindex Wno-assign-intercept 4499Warn whenever an Objective-C assignment is being intercepted by the 4500garbage collector. 4501 4502@item -Wno-property-assign-default @r{(Objective-C and Objective-C++ only)} 4503@opindex Wproperty-assign-default 4504@opindex Wno-property-assign-default 4505Do not warn if a property for an Objective-C object has no assign 4506semantics specified. 4507 4508@item -Wno-protocol @r{(Objective-C and Objective-C++ only)} 4509@opindex Wno-protocol 4510@opindex Wprotocol 4511If a class is declared to implement a protocol, a warning is issued for 4512every method in the protocol that is not implemented by the class. The 4513default behavior is to issue a warning for every method not explicitly 4514implemented in the class, even if a method implementation is inherited 4515from the superclass. If you use the @option{-Wno-protocol} option, then 4516methods inherited from the superclass are considered to be implemented, 4517and no warning is issued for them. 4518 4519@item -Wobjc-root-class @r{(Objective-C and Objective-C++ only)} 4520@opindex Wobjc-root-class 4521Warn if a class interface lacks a superclass. Most classes will inherit 4522from @code{NSObject} (or @code{Object}) for example. When declaring 4523classes intended to be root classes, the warning can be suppressed by 4524marking their interfaces with @code{__attribute__((objc_root_class))}. 4525 4526@item -Wselector @r{(Objective-C and Objective-C++ only)} 4527@opindex Wselector 4528@opindex Wno-selector 4529Warn if multiple methods of different types for the same selector are 4530found during compilation. The check is performed on the list of methods 4531in the final stage of compilation. Additionally, a check is performed 4532for each selector appearing in a @code{@@selector(@dots{})} 4533expression, and a corresponding method for that selector has been found 4534during compilation. Because these checks scan the method table only at 4535the end of compilation, these warnings are not produced if the final 4536stage of compilation is not reached, for example because an error is 4537found during compilation, or because the @option{-fsyntax-only} option is 4538being used. 4539 4540@item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)} 4541@opindex Wstrict-selector-match 4542@opindex Wno-strict-selector-match 4543Warn if multiple methods with differing argument and/or return types are 4544found for a given selector when attempting to send a message using this 4545selector to a receiver of type @code{id} or @code{Class}. When this flag 4546is off (which is the default behavior), the compiler omits such warnings 4547if any differences found are confined to types that share the same size 4548and alignment. 4549 4550@item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)} 4551@opindex Wundeclared-selector 4552@opindex Wno-undeclared-selector 4553Warn if a @code{@@selector(@dots{})} expression referring to an 4554undeclared selector is found. A selector is considered undeclared if no 4555method with that name has been declared before the 4556@code{@@selector(@dots{})} expression, either explicitly in an 4557@code{@@interface} or @code{@@protocol} declaration, or implicitly in 4558an @code{@@implementation} section. This option always performs its 4559checks as soon as a @code{@@selector(@dots{})} expression is found, 4560while @option{-Wselector} only performs its checks in the final stage of 4561compilation. This also enforces the coding style convention 4562that methods and selectors must be declared before being used. 4563 4564@item -print-objc-runtime-info 4565@opindex print-objc-runtime-info 4566Generate C header describing the largest structure that is passed by 4567value, if any. 4568 4569@end table 4570 4571@node Diagnostic Message Formatting Options 4572@section Options to Control Diagnostic Messages Formatting 4573@cindex options to control diagnostics formatting 4574@cindex diagnostic messages 4575@cindex message formatting 4576 4577Traditionally, diagnostic messages have been formatted irrespective of 4578the output device's aspect (e.g.@: its width, @dots{}). You can use the 4579options described below 4580to control the formatting algorithm for diagnostic messages, 4581e.g.@: how many characters per line, how often source location 4582information should be reported. Note that some language front ends may not 4583honor these options. 4584 4585@table @gcctabopt 4586@item -fmessage-length=@var{n} 4587@opindex fmessage-length 4588Try to format error messages so that they fit on lines of about 4589@var{n} characters. If @var{n} is zero, then no line-wrapping is 4590done; each error message appears on a single line. This is the 4591default for all front ends. 4592 4593Note - this option also affects the display of the @samp{#error} and 4594@samp{#warning} pre-processor directives, and the @samp{deprecated} 4595function/type/variable attribute. It does not however affect the 4596@samp{pragma GCC warning} and @samp{pragma GCC error} pragmas. 4597 4598@item -fdiagnostics-plain-output 4599This option requests that diagnostic output look as plain as possible, which 4600may be useful when running @command{dejagnu} or other utilities that need to 4601parse diagnostics output and prefer that it remain more stable over time. 4602@option{-fdiagnostics-plain-output} is currently equivalent to the following 4603options: 4604@gccoptlist{-fno-diagnostics-show-caret @gol 4605-fno-diagnostics-show-line-numbers @gol 4606-fdiagnostics-color=never @gol 4607-fdiagnostics-urls=never @gol 4608-fdiagnostics-path-format=separate-events} 4609In the future, if GCC changes the default appearance of its diagnostics, the 4610corresponding option to disable the new behavior will be added to this list. 4611 4612@item -fdiagnostics-show-location=once 4613@opindex fdiagnostics-show-location 4614Only meaningful in line-wrapping mode. Instructs the diagnostic messages 4615reporter to emit source location information @emph{once}; that is, in 4616case the message is too long to fit on a single physical line and has to 4617be wrapped, the source location won't be emitted (as prefix) again, 4618over and over, in subsequent continuation lines. This is the default 4619behavior. 4620 4621@item -fdiagnostics-show-location=every-line 4622Only meaningful in line-wrapping mode. Instructs the diagnostic 4623messages reporter to emit the same source location information (as 4624prefix) for physical lines that result from the process of breaking 4625a message which is too long to fit on a single line. 4626 4627@item -fdiagnostics-color[=@var{WHEN}] 4628@itemx -fno-diagnostics-color 4629@opindex fdiagnostics-color 4630@cindex highlight, color 4631@vindex GCC_COLORS @r{environment variable} 4632Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always}, 4633or @samp{auto}. The default depends on how the compiler has been configured, 4634it can be any of the above @var{WHEN} options or also @samp{never} 4635if @env{GCC_COLORS} environment variable isn't present in the environment, 4636and @samp{auto} otherwise. 4637@samp{auto} makes GCC use color only when the standard error is a terminal, 4638and when not executing in an emacs shell. 4639The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are 4640aliases for @option{-fdiagnostics-color=always} and 4641@option{-fdiagnostics-color=never}, respectively. 4642 4643The colors are defined by the environment variable @env{GCC_COLORS}. 4644Its value is a colon-separated list of capabilities and Select Graphic 4645Rendition (SGR) substrings. SGR commands are interpreted by the 4646terminal or terminal emulator. (See the section in the documentation 4647of your text terminal for permitted values and their meanings as 4648character attributes.) These substring values are integers in decimal 4649representation and can be concatenated with semicolons. 4650Common values to concatenate include 4651@samp{1} for bold, 4652@samp{4} for underline, 4653@samp{5} for blink, 4654@samp{7} for inverse, 4655@samp{39} for default foreground color, 4656@samp{30} to @samp{37} for foreground colors, 4657@samp{90} to @samp{97} for 16-color mode foreground colors, 4658@samp{38;5;0} to @samp{38;5;255} 4659for 88-color and 256-color modes foreground colors, 4660@samp{49} for default background color, 4661@samp{40} to @samp{47} for background colors, 4662@samp{100} to @samp{107} for 16-color mode background colors, 4663and @samp{48;5;0} to @samp{48;5;255} 4664for 88-color and 256-color modes background colors. 4665 4666The default @env{GCC_COLORS} is 4667@smallexample 4668error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\ 4669quote=01:path=01;36:fixit-insert=32:fixit-delete=31:\ 4670diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\ 4671type-diff=01;32 4672@end smallexample 4673@noindent 4674where @samp{01;31} is bold red, @samp{01;35} is bold magenta, 4675@samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue, 4676@samp{01} is bold, and @samp{31} is red. 4677Setting @env{GCC_COLORS} to the empty string disables colors. 4678Supported capabilities are as follows. 4679 4680@table @code 4681@item error= 4682@vindex error GCC_COLORS @r{capability} 4683SGR substring for error: markers. 4684 4685@item warning= 4686@vindex warning GCC_COLORS @r{capability} 4687SGR substring for warning: markers. 4688 4689@item note= 4690@vindex note GCC_COLORS @r{capability} 4691SGR substring for note: markers. 4692 4693@item path= 4694@vindex path GCC_COLORS @r{capability} 4695SGR substring for colorizing paths of control-flow events as printed 4696via @option{-fdiagnostics-path-format=}, such as the identifiers of 4697individual events and lines indicating interprocedural calls and returns. 4698 4699@item range1= 4700@vindex range1 GCC_COLORS @r{capability} 4701SGR substring for first additional range. 4702 4703@item range2= 4704@vindex range2 GCC_COLORS @r{capability} 4705SGR substring for second additional range. 4706 4707@item locus= 4708@vindex locus GCC_COLORS @r{capability} 4709SGR substring for location information, @samp{file:line} or 4710@samp{file:line:column} etc. 4711 4712@item quote= 4713@vindex quote GCC_COLORS @r{capability} 4714SGR substring for information printed within quotes. 4715 4716@item fixit-insert= 4717@vindex fixit-insert GCC_COLORS @r{capability} 4718SGR substring for fix-it hints suggesting text to 4719be inserted or replaced. 4720 4721@item fixit-delete= 4722@vindex fixit-delete GCC_COLORS @r{capability} 4723SGR substring for fix-it hints suggesting text to 4724be deleted. 4725 4726@item diff-filename= 4727@vindex diff-filename GCC_COLORS @r{capability} 4728SGR substring for filename headers within generated patches. 4729 4730@item diff-hunk= 4731@vindex diff-hunk GCC_COLORS @r{capability} 4732SGR substring for the starts of hunks within generated patches. 4733 4734@item diff-delete= 4735@vindex diff-delete GCC_COLORS @r{capability} 4736SGR substring for deleted lines within generated patches. 4737 4738@item diff-insert= 4739@vindex diff-insert GCC_COLORS @r{capability} 4740SGR substring for inserted lines within generated patches. 4741 4742@item type-diff= 4743@vindex type-diff GCC_COLORS @r{capability} 4744SGR substring for highlighting mismatching types within template 4745arguments in the C++ frontend. 4746@end table 4747 4748@item -fdiagnostics-urls[=@var{WHEN}] 4749@opindex fdiagnostics-urls 4750@cindex urls 4751@vindex GCC_URLS @r{environment variable} 4752@vindex TERM_URLS @r{environment variable} 4753Use escape sequences to embed URLs in diagnostics. For example, when 4754@option{-fdiagnostics-show-option} emits text showing the command-line 4755option controlling a diagnostic, embed a URL for documentation of that 4756option. 4757 4758@var{WHEN} is @samp{never}, @samp{always}, or @samp{auto}. 4759@samp{auto} makes GCC use URL escape sequences only when the standard error 4760is a terminal, and when not executing in an emacs shell or any graphical 4761terminal which is known to be incompatible with this feature, see below. 4762 4763The default depends on how the compiler has been configured. 4764It can be any of the above @var{WHEN} options. 4765 4766GCC can also be configured (via the 4767@option{--with-diagnostics-urls=auto-if-env} configure-time option) 4768so that the default is affected by environment variables. 4769Under such a configuration, GCC defaults to using @samp{auto} 4770if either @env{GCC_URLS} or @env{TERM_URLS} environment variables are 4771present and non-empty in the environment of the compiler, or @samp{never} 4772if neither are. 4773 4774However, even with @option{-fdiagnostics-urls=always} the behavior is 4775dependent on those environment variables: 4776If @env{GCC_URLS} is set to empty or @samp{no}, do not embed URLs in 4777diagnostics. If set to @samp{st}, URLs use ST escape sequences. 4778If set to @samp{bel}, the default, URLs use BEL escape sequences. 4779Any other non-empty value enables the feature. 4780If @env{GCC_URLS} is not set, use @env{TERM_URLS} as a fallback. 4781Note: ST is an ANSI escape sequence, string terminator @samp{ESC \}, 4782BEL is an ASCII character, CTRL-G that usually sounds like a beep. 4783 4784At this time GCC tries to detect also a few terminals that are known to 4785not implement the URL feature, and have bugs or at least had bugs in 4786some versions that are still in use, where the URL escapes are likely 4787to misbehave, i.e. print garbage on the screen. 4788That list is currently xfce4-terminal, certain known to be buggy 4789gnome-terminal versions, the linux console, and mingw. 4790This check can be skipped with the @option{-fdiagnostics-urls=always}. 4791 4792@item -fno-diagnostics-show-option 4793@opindex fno-diagnostics-show-option 4794@opindex fdiagnostics-show-option 4795By default, each diagnostic emitted includes text indicating the 4796command-line option that directly controls the diagnostic (if such an 4797option is known to the diagnostic machinery). Specifying the 4798@option{-fno-diagnostics-show-option} flag suppresses that behavior. 4799 4800@item -fno-diagnostics-show-caret 4801@opindex fno-diagnostics-show-caret 4802@opindex fdiagnostics-show-caret 4803By default, each diagnostic emitted includes the original source line 4804and a caret @samp{^} indicating the column. This option suppresses this 4805information. The source line is truncated to @var{n} characters, if 4806the @option{-fmessage-length=n} option is given. When the output is done 4807to the terminal, the width is limited to the width given by the 4808@env{COLUMNS} environment variable or, if not set, to the terminal width. 4809 4810@item -fno-diagnostics-show-labels 4811@opindex fno-diagnostics-show-labels 4812@opindex fdiagnostics-show-labels 4813By default, when printing source code (via @option{-fdiagnostics-show-caret}), 4814diagnostics can label ranges of source code with pertinent information, such 4815as the types of expressions: 4816 4817@smallexample 4818 printf ("foo %s bar", long_i + long_j); 4819 ~^ ~~~~~~~~~~~~~~~ 4820 | | 4821 char * long int 4822@end smallexample 4823 4824This option suppresses the printing of these labels (in the example above, 4825the vertical bars and the ``char *'' and ``long int'' text). 4826 4827@item -fno-diagnostics-show-cwe 4828@opindex fno-diagnostics-show-cwe 4829@opindex fdiagnostics-show-cwe 4830Diagnostic messages can optionally have an associated 4831@url{https://cwe.mitre.org/index.html, CWE} identifier. 4832GCC itself only provides such metadata for some of the @option{-fanalyzer} 4833diagnostics. GCC plugins may also provide diagnostics with such metadata. 4834By default, if this information is present, it will be printed with 4835the diagnostic. This option suppresses the printing of this metadata. 4836 4837@item -fno-diagnostics-show-line-numbers 4838@opindex fno-diagnostics-show-line-numbers 4839@opindex fdiagnostics-show-line-numbers 4840By default, when printing source code (via @option{-fdiagnostics-show-caret}), 4841a left margin is printed, showing line numbers. This option suppresses this 4842left margin. 4843 4844@item -fdiagnostics-minimum-margin-width=@var{width} 4845@opindex fdiagnostics-minimum-margin-width 4846This option controls the minimum width of the left margin printed by 4847@option{-fdiagnostics-show-line-numbers}. It defaults to 6. 4848 4849@item -fdiagnostics-parseable-fixits 4850@opindex fdiagnostics-parseable-fixits 4851Emit fix-it hints in a machine-parseable format, suitable for consumption 4852by IDEs. For each fix-it, a line will be printed after the relevant 4853diagnostic, starting with the string ``fix-it:''. For example: 4854 4855@smallexample 4856fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all" 4857@end smallexample 4858 4859The location is expressed as a half-open range, expressed as a count of 4860bytes, starting at byte 1 for the initial column. In the above example, 4861bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the 4862given string: 4863 4864@smallexample 486500000000011111111112222222222 486612345678901234567890123456789 4867 gtk_widget_showall (dlg); 4868 ^^^^^^^^^^^^^^^^^^ 4869 gtk_widget_show_all 4870@end smallexample 4871 4872The filename and replacement string escape backslash as ``\\", tab as ``\t'', 4873newline as ``\n'', double quotes as ``\"'', non-printable characters as octal 4874(e.g. vertical tab as ``\013''). 4875 4876An empty replacement string indicates that the given range is to be removed. 4877An empty range (e.g. ``45:3-45:3'') indicates that the string is to 4878be inserted at the given position. 4879 4880@item -fdiagnostics-generate-patch 4881@opindex fdiagnostics-generate-patch 4882Print fix-it hints to stderr in unified diff format, after any diagnostics 4883are printed. For example: 4884 4885@smallexample 4886--- test.c 4887+++ test.c 4888@@ -42,5 +42,5 @@ 4889 4890 void show_cb(GtkDialog *dlg) 4891 @{ 4892- gtk_widget_showall(dlg); 4893+ gtk_widget_show_all(dlg); 4894 @} 4895 4896@end smallexample 4897 4898The diff may or may not be colorized, following the same rules 4899as for diagnostics (see @option{-fdiagnostics-color}). 4900 4901@item -fdiagnostics-show-template-tree 4902@opindex fdiagnostics-show-template-tree 4903 4904In the C++ frontend, when printing diagnostics showing mismatching 4905template types, such as: 4906 4907@smallexample 4908 could not convert 'std::map<int, std::vector<double> >()' 4909 from 'map<[...],vector<double>>' to 'map<[...],vector<float>> 4910@end smallexample 4911 4912the @option{-fdiagnostics-show-template-tree} flag enables printing a 4913tree-like structure showing the common and differing parts of the types, 4914such as: 4915 4916@smallexample 4917 map< 4918 [...], 4919 vector< 4920 [double != float]>> 4921@end smallexample 4922 4923The parts that differ are highlighted with color (``double'' and 4924``float'' in this case). 4925 4926@item -fno-elide-type 4927@opindex fno-elide-type 4928@opindex felide-type 4929By default when the C++ frontend prints diagnostics showing mismatching 4930template types, common parts of the types are printed as ``[...]'' to 4931simplify the error message. For example: 4932 4933@smallexample 4934 could not convert 'std::map<int, std::vector<double> >()' 4935 from 'map<[...],vector<double>>' to 'map<[...],vector<float>> 4936@end smallexample 4937 4938Specifying the @option{-fno-elide-type} flag suppresses that behavior. 4939This flag also affects the output of the 4940@option{-fdiagnostics-show-template-tree} flag. 4941 4942@item -fdiagnostics-path-format=@var{KIND} 4943@opindex fdiagnostics-path-format 4944Specify how to print paths of control-flow events for diagnostics that 4945have such a path associated with them. 4946 4947@var{KIND} is @samp{none}, @samp{separate-events}, or @samp{inline-events}, 4948the default. 4949 4950@samp{none} means to not print diagnostic paths. 4951 4952@samp{separate-events} means to print a separate ``note'' diagnostic for 4953each event within the diagnostic. For example: 4954 4955@smallexample 4956test.c:29:5: error: passing NULL as argument 1 to 'PyList_Append' which requires a non-NULL parameter 4957test.c:25:10: note: (1) when 'PyList_New' fails, returning NULL 4958test.c:27:3: note: (2) when 'i < count' 4959test.c:29:5: note: (3) when calling 'PyList_Append', passing NULL from (1) as argument 1 4960@end smallexample 4961 4962@samp{inline-events} means to print the events ``inline'' within the source 4963code. This view attempts to consolidate the events into runs of 4964sufficiently-close events, printing them as labelled ranges within the source. 4965 4966For example, the same events as above might be printed as: 4967 4968@smallexample 4969 'test': events 1-3 4970 | 4971 | 25 | list = PyList_New(0); 4972 | | ^~~~~~~~~~~~~ 4973 | | | 4974 | | (1) when 'PyList_New' fails, returning NULL 4975 | 26 | 4976 | 27 | for (i = 0; i < count; i++) @{ 4977 | | ~~~ 4978 | | | 4979 | | (2) when 'i < count' 4980 | 28 | item = PyLong_FromLong(random()); 4981 | 29 | PyList_Append(list, item); 4982 | | ~~~~~~~~~~~~~~~~~~~~~~~~~ 4983 | | | 4984 | | (3) when calling 'PyList_Append', passing NULL from (1) as argument 1 4985 | 4986@end smallexample 4987 4988Interprocedural control flow is shown by grouping the events by stack frame, 4989and using indentation to show how stack frames are nested, pushed, and popped. 4990 4991For example: 4992 4993@smallexample 4994 'test': events 1-2 4995 | 4996 | 133 | @{ 4997 | | ^ 4998 | | | 4999 | | (1) entering 'test' 5000 | 134 | boxed_int *obj = make_boxed_int (i); 5001 | | ~~~~~~~~~~~~~~~~~~ 5002 | | | 5003 | | (2) calling 'make_boxed_int' 5004 | 5005 +--> 'make_boxed_int': events 3-4 5006 | 5007 | 120 | @{ 5008 | | ^ 5009 | | | 5010 | | (3) entering 'make_boxed_int' 5011 | 121 | boxed_int *result = (boxed_int *)wrapped_malloc (sizeof (boxed_int)); 5012 | | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 5013 | | | 5014 | | (4) calling 'wrapped_malloc' 5015 | 5016 +--> 'wrapped_malloc': events 5-6 5017 | 5018 | 7 | @{ 5019 | | ^ 5020 | | | 5021 | | (5) entering 'wrapped_malloc' 5022 | 8 | return malloc (size); 5023 | | ~~~~~~~~~~~~~ 5024 | | | 5025 | | (6) calling 'malloc' 5026 | 5027 <-------------+ 5028 | 5029 'test': event 7 5030 | 5031 | 138 | free_boxed_int (obj); 5032 | | ^~~~~~~~~~~~~~~~~~~~ 5033 | | | 5034 | | (7) calling 'free_boxed_int' 5035 | 5036(etc) 5037@end smallexample 5038 5039@item -fdiagnostics-show-path-depths 5040@opindex fdiagnostics-show-path-depths 5041This option provides additional information when printing control-flow paths 5042associated with a diagnostic. 5043 5044If this is option is provided then the stack depth will be printed for 5045each run of events within @option{-fdiagnostics-path-format=separate-events}. 5046 5047This is intended for use by GCC developers and plugin developers when 5048debugging diagnostics that report interprocedural control flow. 5049 5050@item -fno-show-column 5051@opindex fno-show-column 5052@opindex fshow-column 5053Do not print column numbers in diagnostics. This may be necessary if 5054diagnostics are being scanned by a program that does not understand the 5055column numbers, such as @command{dejagnu}. 5056 5057@item -fdiagnostics-column-unit=@var{UNIT} 5058@opindex fdiagnostics-column-unit 5059Select the units for the column number. This affects traditional diagnostics 5060(in the absence of @option{-fno-show-column}), as well as JSON format 5061diagnostics if requested. 5062 5063The default @var{UNIT}, @samp{display}, considers the number of display 5064columns occupied by each character. This may be larger than the number 5065of bytes required to encode the character, in the case of tab 5066characters, or it may be smaller, in the case of multibyte characters. 5067For example, the character ``GREEK SMALL LETTER PI (U+03C0)'' occupies one 5068display column, and its UTF-8 encoding requires two bytes; the character 5069``SLIGHTLY SMILING FACE (U+1F642)'' occupies two display columns, and 5070its UTF-8 encoding requires four bytes. 5071 5072Setting @var{UNIT} to @samp{byte} changes the column number to the raw byte 5073count in all cases, as was traditionally output by GCC prior to version 11.1.0. 5074 5075@item -fdiagnostics-column-origin=@var{ORIGIN} 5076@opindex fdiagnostics-column-origin 5077Select the origin for column numbers, i.e. the column number assigned to the 5078first column. The default value of 1 corresponds to traditional GCC 5079behavior and to the GNU style guide. Some utilities may perform better with an 5080origin of 0; any non-negative value may be specified. 5081 5082@item -fdiagnostics-format=@var{FORMAT} 5083@opindex fdiagnostics-format 5084Select a different format for printing diagnostics. 5085@var{FORMAT} is @samp{text} or @samp{json}. 5086The default is @samp{text}. 5087 5088The @samp{json} format consists of a top-level JSON array containing JSON 5089objects representing the diagnostics. 5090 5091The JSON is emitted as one line, without formatting; the examples below 5092have been formatted for clarity. 5093 5094Diagnostics can have child diagnostics. For example, this error and note: 5095 5096@smallexample 5097misleading-indentation.c:15:3: warning: this 'if' clause does not 5098 guard... [-Wmisleading-indentation] 5099 15 | if (flag) 5100 | ^~ 5101misleading-indentation.c:17:5: note: ...this statement, but the latter 5102 is misleadingly indented as if it were guarded by the 'if' 5103 17 | y = 2; 5104 | ^ 5105@end smallexample 5106 5107@noindent 5108might be printed in JSON form (after formatting) like this: 5109 5110@smallexample 5111[ 5112 @{ 5113 "kind": "warning", 5114 "locations": [ 5115 @{ 5116 "caret": @{ 5117 "display-column": 3, 5118 "byte-column": 3, 5119 "column": 3, 5120 "file": "misleading-indentation.c", 5121 "line": 15 5122 @}, 5123 "finish": @{ 5124 "display-column": 4, 5125 "byte-column": 4, 5126 "column": 4, 5127 "file": "misleading-indentation.c", 5128 "line": 15 5129 @} 5130 @} 5131 ], 5132 "message": "this \u2018if\u2019 clause does not guard...", 5133 "option": "-Wmisleading-indentation", 5134 "option_url": "https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#index-Wmisleading-indentation", 5135 "children": [ 5136 @{ 5137 "kind": "note", 5138 "locations": [ 5139 @{ 5140 "caret": @{ 5141 "display-column": 5, 5142 "byte-column": 5, 5143 "column": 5, 5144 "file": "misleading-indentation.c", 5145 "line": 17 5146 @} 5147 @} 5148 ], 5149 "message": "...this statement, but the latter is @dots{}" 5150 @} 5151 ] 5152 "column-origin": 1, 5153 @}, 5154 @dots{} 5155] 5156@end smallexample 5157 5158@noindent 5159where the @code{note} is a child of the @code{warning}. 5160 5161A diagnostic has a @code{kind}. If this is @code{warning}, then there is 5162an @code{option} key describing the command-line option controlling the 5163warning. 5164 5165A diagnostic can contain zero or more locations. Each location has an 5166optional @code{label} string and up to three positions within it: a 5167@code{caret} position and optional @code{start} and @code{finish} positions. 5168A position is described by a @code{file} name, a @code{line} number, and 5169three numbers indicating a column position: 5170@itemize @bullet 5171 5172@item 5173@code{display-column} counts display columns, accounting for tabs and 5174multibyte characters. 5175 5176@item 5177@code{byte-column} counts raw bytes. 5178 5179@item 5180@code{column} is equal to one of 5181the previous two, as dictated by the @option{-fdiagnostics-column-unit} 5182option. 5183 5184@end itemize 5185All three columns are relative to the origin specified by 5186@option{-fdiagnostics-column-origin}, which is typically equal to 1 but may 5187be set, for instance, to 0 for compatibility with other utilities that 5188number columns from 0. The column origin is recorded in the JSON output in 5189the @code{column-origin} tag. In the remaining examples below, the extra 5190column number outputs have been omitted for brevity. 5191 5192For example, this error: 5193 5194@smallexample 5195bad-binary-ops.c:64:23: error: invalid operands to binary + (have 'S' @{aka 5196 'struct s'@} and 'T' @{aka 'struct t'@}) 5197 64 | return callee_4a () + callee_4b (); 5198 | ~~~~~~~~~~~~ ^ ~~~~~~~~~~~~ 5199 | | | 5200 | | T @{aka struct t@} 5201 | S @{aka struct s@} 5202@end smallexample 5203 5204@noindent 5205has three locations. Its primary location is at the ``+'' token at column 520623. It has two secondary locations, describing the left and right-hand sides 5207of the expression, which have labels. It might be printed in JSON form as: 5208 5209@smallexample 5210 @{ 5211 "children": [], 5212 "kind": "error", 5213 "locations": [ 5214 @{ 5215 "caret": @{ 5216 "column": 23, "file": "bad-binary-ops.c", "line": 64 5217 @} 5218 @}, 5219 @{ 5220 "caret": @{ 5221 "column": 10, "file": "bad-binary-ops.c", "line": 64 5222 @}, 5223 "finish": @{ 5224 "column": 21, "file": "bad-binary-ops.c", "line": 64 5225 @}, 5226 "label": "S @{aka struct s@}" 5227 @}, 5228 @{ 5229 "caret": @{ 5230 "column": 25, "file": "bad-binary-ops.c", "line": 64 5231 @}, 5232 "finish": @{ 5233 "column": 36, "file": "bad-binary-ops.c", "line": 64 5234 @}, 5235 "label": "T @{aka struct t@}" 5236 @} 5237 ], 5238 "message": "invalid operands to binary + @dots{}" 5239 @} 5240@end smallexample 5241 5242If a diagnostic contains fix-it hints, it has a @code{fixits} array, 5243consisting of half-open intervals, similar to the output of 5244@option{-fdiagnostics-parseable-fixits}. For example, this diagnostic 5245with a replacement fix-it hint: 5246 5247@smallexample 5248demo.c:8:15: error: 'struct s' has no member named 'colour'; did you 5249 mean 'color'? 5250 8 | return ptr->colour; 5251 | ^~~~~~ 5252 | color 5253@end smallexample 5254 5255@noindent 5256might be printed in JSON form as: 5257 5258@smallexample 5259 @{ 5260 "children": [], 5261 "fixits": [ 5262 @{ 5263 "next": @{ 5264 "column": 21, 5265 "file": "demo.c", 5266 "line": 8 5267 @}, 5268 "start": @{ 5269 "column": 15, 5270 "file": "demo.c", 5271 "line": 8 5272 @}, 5273 "string": "color" 5274 @} 5275 ], 5276 "kind": "error", 5277 "locations": [ 5278 @{ 5279 "caret": @{ 5280 "column": 15, 5281 "file": "demo.c", 5282 "line": 8 5283 @}, 5284 "finish": @{ 5285 "column": 20, 5286 "file": "demo.c", 5287 "line": 8 5288 @} 5289 @} 5290 ], 5291 "message": "\u2018struct s\u2019 has no member named @dots{}" 5292 @} 5293@end smallexample 5294 5295@noindent 5296where the fix-it hint suggests replacing the text from @code{start} up 5297to but not including @code{next} with @code{string}'s value. Deletions 5298are expressed via an empty value for @code{string}, insertions by 5299having @code{start} equal @code{next}. 5300 5301If the diagnostic has a path of control-flow events associated with it, 5302it has a @code{path} array of objects representing the events. Each 5303event object has a @code{description} string, a @code{location} object, 5304along with a @code{function} string and a @code{depth} number for 5305representing interprocedural paths. The @code{function} represents the 5306current function at that event, and the @code{depth} represents the 5307stack depth relative to some baseline: the higher, the more frames are 5308within the stack. 5309 5310For example, the intraprocedural example shown for 5311@option{-fdiagnostics-path-format=} might have this JSON for its path: 5312 5313@smallexample 5314 "path": [ 5315 @{ 5316 "depth": 0, 5317 "description": "when 'PyList_New' fails, returning NULL", 5318 "function": "test", 5319 "location": @{ 5320 "column": 10, 5321 "file": "test.c", 5322 "line": 25 5323 @} 5324 @}, 5325 @{ 5326 "depth": 0, 5327 "description": "when 'i < count'", 5328 "function": "test", 5329 "location": @{ 5330 "column": 3, 5331 "file": "test.c", 5332 "line": 27 5333 @} 5334 @}, 5335 @{ 5336 "depth": 0, 5337 "description": "when calling 'PyList_Append', passing NULL from (1) as argument 1", 5338 "function": "test", 5339 "location": @{ 5340 "column": 5, 5341 "file": "test.c", 5342 "line": 29 5343 @} 5344 @} 5345 ] 5346@end smallexample 5347 5348@end table 5349 5350@node Warning Options 5351@section Options to Request or Suppress Warnings 5352@cindex options to control warnings 5353@cindex warning messages 5354@cindex messages, warning 5355@cindex suppressing warnings 5356 5357Warnings are diagnostic messages that report constructions that 5358are not inherently erroneous but that are risky or suggest there 5359may have been an error. 5360 5361The following language-independent options do not enable specific 5362warnings but control the kinds of diagnostics produced by GCC@. 5363 5364@table @gcctabopt 5365@cindex syntax checking 5366@item -fsyntax-only 5367@opindex fsyntax-only 5368Check the code for syntax errors, but don't do anything beyond that. 5369 5370@item -fmax-errors=@var{n} 5371@opindex fmax-errors 5372Limits the maximum number of error messages to @var{n}, at which point 5373GCC bails out rather than attempting to continue processing the source 5374code. If @var{n} is 0 (the default), there is no limit on the number 5375of error messages produced. If @option{-Wfatal-errors} is also 5376specified, then @option{-Wfatal-errors} takes precedence over this 5377option. 5378 5379@item -w 5380@opindex w 5381Inhibit all warning messages. 5382 5383@item -Werror 5384@opindex Werror 5385@opindex Wno-error 5386Make all warnings into errors. 5387 5388@item -Werror= 5389@opindex Werror= 5390@opindex Wno-error= 5391Make the specified warning into an error. The specifier for a warning 5392is appended; for example @option{-Werror=switch} turns the warnings 5393controlled by @option{-Wswitch} into errors. This switch takes a 5394negative form, to be used to negate @option{-Werror} for specific 5395warnings; for example @option{-Wno-error=switch} makes 5396@option{-Wswitch} warnings not be errors, even when @option{-Werror} 5397is in effect. 5398 5399The warning message for each controllable warning includes the 5400option that controls the warning. That option can then be used with 5401@option{-Werror=} and @option{-Wno-error=} as described above. 5402(Printing of the option in the warning message can be disabled using the 5403@option{-fno-diagnostics-show-option} flag.) 5404 5405Note that specifying @option{-Werror=}@var{foo} automatically implies 5406@option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not 5407imply anything. 5408 5409@item -Wfatal-errors 5410@opindex Wfatal-errors 5411@opindex Wno-fatal-errors 5412This option causes the compiler to abort compilation on the first error 5413occurred rather than trying to keep going and printing further error 5414messages. 5415 5416@end table 5417 5418You can request many specific warnings with options beginning with 5419@samp{-W}, for example @option{-Wimplicit} to request warnings on 5420implicit declarations. Each of these specific warning options also 5421has a negative form beginning @samp{-Wno-} to turn off warnings; for 5422example, @option{-Wno-implicit}. This manual lists only one of the 5423two forms, whichever is not the default. For further 5424language-specific options also refer to @ref{C++ Dialect Options} and 5425@ref{Objective-C and Objective-C++ Dialect Options}. 5426Additional warnings can be produced by enabling the static analyzer; 5427@xref{Static Analyzer Options}. 5428 5429Some options, such as @option{-Wall} and @option{-Wextra}, turn on other 5430options, such as @option{-Wunused}, which may turn on further options, 5431such as @option{-Wunused-value}. The combined effect of positive and 5432negative forms is that more specific options have priority over less 5433specific ones, independently of their position in the command-line. For 5434options of the same specificity, the last one takes effect. Options 5435enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect 5436as if they appeared at the end of the command-line. 5437 5438When an unrecognized warning option is requested (e.g., 5439@option{-Wunknown-warning}), GCC emits a diagnostic stating 5440that the option is not recognized. However, if the @option{-Wno-} form 5441is used, the behavior is slightly different: no diagnostic is 5442produced for @option{-Wno-unknown-warning} unless other diagnostics 5443are being produced. This allows the use of new @option{-Wno-} options 5444with old compilers, but if something goes wrong, the compiler 5445warns that an unrecognized option is present. 5446 5447The effectiveness of some warnings depends on optimizations also being 5448enabled. For example @option{-Wsuggest-final-types} is more effective 5449with link-time optimization and @option{-Wmaybe-uninitialized} does not 5450warn at all unless optimization is enabled. 5451 5452@table @gcctabopt 5453@item -Wpedantic 5454@itemx -pedantic 5455@opindex pedantic 5456@opindex Wpedantic 5457@opindex Wno-pedantic 5458Issue all the warnings demanded by strict ISO C and ISO C++; 5459reject all programs that use forbidden extensions, and some other 5460programs that do not follow ISO C and ISO C++. For ISO C, follows the 5461version of the ISO C standard specified by any @option{-std} option used. 5462 5463Valid ISO C and ISO C++ programs should compile properly with or without 5464this option (though a rare few require @option{-ansi} or a 5465@option{-std} option specifying the required version of ISO C)@. However, 5466without this option, certain GNU extensions and traditional C and C++ 5467features are supported as well. With this option, they are rejected. 5468 5469@option{-Wpedantic} does not cause warning messages for use of the 5470alternate keywords whose names begin and end with @samp{__}. This alternate 5471format can also be used to disable warnings for non-ISO @samp{__intN} types, 5472i.e. @samp{__intN__}. 5473Pedantic warnings are also disabled in the expression that follows 5474@code{__extension__}. However, only system header files should use 5475these escape routes; application programs should avoid them. 5476@xref{Alternate Keywords}. 5477 5478Some users try to use @option{-Wpedantic} to check programs for strict ISO 5479C conformance. They soon find that it does not do quite what they want: 5480it finds some non-ISO practices, but not all---only those for which 5481ISO C @emph{requires} a diagnostic, and some others for which 5482diagnostics have been added. 5483 5484A feature to report any failure to conform to ISO C might be useful in 5485some instances, but would require considerable additional work and would 5486be quite different from @option{-Wpedantic}. We don't have plans to 5487support such a feature in the near future. 5488 5489Where the standard specified with @option{-std} represents a GNU 5490extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a 5491corresponding @dfn{base standard}, the version of ISO C on which the GNU 5492extended dialect is based. Warnings from @option{-Wpedantic} are given 5493where they are required by the base standard. (It does not make sense 5494for such warnings to be given only for features not in the specified GNU 5495C dialect, since by definition the GNU dialects of C include all 5496features the compiler supports with the given option, and there would be 5497nothing to warn about.) 5498 5499@item -pedantic-errors 5500@opindex pedantic-errors 5501Give an error whenever the @dfn{base standard} (see @option{-Wpedantic}) 5502requires a diagnostic, in some cases where there is undefined behavior 5503at compile-time and in some other cases that do not prevent compilation 5504of programs that are valid according to the standard. This is not 5505equivalent to @option{-Werror=pedantic}, since there are errors enabled 5506by this option and not enabled by the latter and vice versa. 5507 5508@item -Wall 5509@opindex Wall 5510@opindex Wno-all 5511This enables all the warnings about constructions that some users 5512consider questionable, and that are easy to avoid (or modify to 5513prevent the warning), even in conjunction with macros. This also 5514enables some language-specific warnings described in @ref{C++ Dialect 5515Options} and @ref{Objective-C and Objective-C++ Dialect Options}. 5516 5517@option{-Wall} turns on the following warning flags: 5518 5519@gccoptlist{-Waddress @gol 5520-Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol 5521-Warray-parameter=2 @r{(C and Objective-C only)} @gol 5522-Wbool-compare @gol 5523-Wbool-operation @gol 5524-Wc++11-compat -Wc++14-compat @gol 5525-Wcatch-value @r{(C++ and Objective-C++ only)} @gol 5526-Wchar-subscripts @gol 5527-Wcomment @gol 5528-Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol 5529-Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol 5530-Wformat @gol 5531-Wformat-overflow @gol 5532-Wformat-truncation @gol 5533-Wint-in-bool-context @gol 5534-Wimplicit @r{(C and Objective-C only)} @gol 5535-Wimplicit-int @r{(C and Objective-C only)} @gol 5536-Wimplicit-function-declaration @r{(C and Objective-C only)} @gol 5537-Winit-self @r{(only for C++)} @gol 5538-Wlogical-not-parentheses @gol 5539-Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol 5540-Wmaybe-uninitialized @gol 5541-Wmemset-elt-size @gol 5542-Wmemset-transposed-args @gol 5543-Wmisleading-indentation @r{(only for C/C++)} @gol 5544-Wmissing-attributes @gol 5545-Wmissing-braces @r{(only for C/ObjC)} @gol 5546-Wmultistatement-macros @gol 5547-Wnarrowing @r{(only for C++)} @gol 5548-Wnonnull @gol 5549-Wnonnull-compare @gol 5550-Wopenmp-simd @gol 5551-Wparentheses @gol 5552-Wpessimizing-move @r{(only for C++)} @gol 5553-Wpointer-sign @gol 5554-Wrange-loop-construct @r{(only for C++)} @gol 5555-Wreorder @gol 5556-Wrestrict @gol 5557-Wreturn-type @gol 5558-Wsequence-point @gol 5559-Wsign-compare @r{(only in C++)} @gol 5560-Wsizeof-array-div @gol 5561-Wsizeof-pointer-div @gol 5562-Wsizeof-pointer-memaccess @gol 5563-Wstrict-aliasing @gol 5564-Wstrict-overflow=1 @gol 5565-Wswitch @gol 5566-Wtautological-compare @gol 5567-Wtrigraphs @gol 5568-Wuninitialized @gol 5569-Wunknown-pragmas @gol 5570-Wunused-function @gol 5571-Wunused-label @gol 5572-Wunused-value @gol 5573-Wunused-variable @gol 5574-Wvla-parameter @r{(C and Objective-C only)} @gol 5575-Wvolatile-register-var @gol 5576-Wzero-length-bounds} 5577 5578Note that some warning flags are not implied by @option{-Wall}. Some of 5579them warn about constructions that users generally do not consider 5580questionable, but which occasionally you might wish to check for; 5581others warn about constructions that are necessary or hard to avoid in 5582some cases, and there is no simple way to modify the code to suppress 5583the warning. Some of them are enabled by @option{-Wextra} but many of 5584them must be enabled individually. 5585 5586@item -Wextra 5587@opindex W 5588@opindex Wextra 5589@opindex Wno-extra 5590This enables some extra warning flags that are not enabled by 5591@option{-Wall}. (This option used to be called @option{-W}. The older 5592name is still supported, but the newer name is more descriptive.) 5593 5594@gccoptlist{-Wclobbered @gol 5595-Wcast-function-type @gol 5596-Wdeprecated-copy @r{(C++ only)} @gol 5597-Wempty-body @gol 5598-Wenum-conversion @r{(C only)} @gol 5599-Wignored-qualifiers @gol 5600-Wimplicit-fallthrough=3 @gol 5601-Wmissing-field-initializers @gol 5602-Wmissing-parameter-type @r{(C only)} @gol 5603-Wold-style-declaration @r{(C only)} @gol 5604-Woverride-init @gol 5605-Wsign-compare @r{(C only)} @gol 5606-Wstring-compare @gol 5607-Wredundant-move @r{(only for C++)} @gol 5608-Wtype-limits @gol 5609-Wuninitialized @gol 5610-Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol 5611-Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol 5612-Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)}} 5613 5614 5615The option @option{-Wextra} also prints warning messages for the 5616following cases: 5617 5618@itemize @bullet 5619 5620@item 5621A pointer is compared against integer zero with @code{<}, @code{<=}, 5622@code{>}, or @code{>=}. 5623 5624@item 5625(C++ only) An enumerator and a non-enumerator both appear in a 5626conditional expression. 5627 5628@item 5629(C++ only) Ambiguous virtual bases. 5630 5631@item 5632(C++ only) Subscripting an array that has been declared @code{register}. 5633 5634@item 5635(C++ only) Taking the address of a variable that has been declared 5636@code{register}. 5637 5638@item 5639(C++ only) A base class is not initialized in the copy constructor 5640of a derived class. 5641 5642@end itemize 5643 5644@item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)} 5645@opindex Wabi 5646@opindex Wno-abi 5647 5648Warn about code affected by ABI changes. This includes code that may 5649not be compatible with the vendor-neutral C++ ABI as well as the psABI 5650for the particular target. 5651 5652Since G++ now defaults to updating the ABI with each major release, 5653normally @option{-Wabi} warns only about C++ ABI compatibility 5654problems if there is a check added later in a release series for an 5655ABI issue discovered since the initial release. @option{-Wabi} warns 5656about more things if an older ABI version is selected (with 5657@option{-fabi-version=@var{n}}). 5658 5659@option{-Wabi} can also be used with an explicit version number to 5660warn about C++ ABI compatibility with a particular @option{-fabi-version} 5661level, e.g.@: @option{-Wabi=2} to warn about changes relative to 5662@option{-fabi-version=2}. 5663 5664If an explicit version number is provided and 5665@option{-fabi-compat-version} is not specified, the version number 5666from this option is used for compatibility aliases. If no explicit 5667version number is provided with this option, but 5668@option{-fabi-compat-version} is specified, that version number is 5669used for C++ ABI warnings. 5670 5671Although an effort has been made to warn about 5672all such cases, there are probably some cases that are not warned about, 5673even though G++ is generating incompatible code. There may also be 5674cases where warnings are emitted even though the code that is generated 5675is compatible. 5676 5677You should rewrite your code to avoid these warnings if you are 5678concerned about the fact that code generated by G++ may not be binary 5679compatible with code generated by other compilers. 5680 5681Known incompatibilities in @option{-fabi-version=2} (which was the 5682default from GCC 3.4 to 4.9) include: 5683 5684@itemize @bullet 5685 5686@item 5687A template with a non-type template parameter of reference type was 5688mangled incorrectly: 5689@smallexample 5690extern int N; 5691template <int &> struct S @{@}; 5692void n (S<N>) @{2@} 5693@end smallexample 5694 5695This was fixed in @option{-fabi-version=3}. 5696 5697@item 5698SIMD vector types declared using @code{__attribute ((vector_size))} were 5699mangled in a non-standard way that does not allow for overloading of 5700functions taking vectors of different sizes. 5701 5702The mangling was changed in @option{-fabi-version=4}. 5703 5704@item 5705@code{__attribute ((const))} and @code{noreturn} were mangled as type 5706qualifiers, and @code{decltype} of a plain declaration was folded away. 5707 5708These mangling issues were fixed in @option{-fabi-version=5}. 5709 5710@item 5711Scoped enumerators passed as arguments to a variadic function are 5712promoted like unscoped enumerators, causing @code{va_arg} to complain. 5713On most targets this does not actually affect the parameter passing 5714ABI, as there is no way to pass an argument smaller than @code{int}. 5715 5716Also, the ABI changed the mangling of template argument packs, 5717@code{const_cast}, @code{static_cast}, prefix increment/decrement, and 5718a class scope function used as a template argument. 5719 5720These issues were corrected in @option{-fabi-version=6}. 5721 5722@item 5723Lambdas in default argument scope were mangled incorrectly, and the 5724ABI changed the mangling of @code{nullptr_t}. 5725 5726These issues were corrected in @option{-fabi-version=7}. 5727 5728@item 5729When mangling a function type with function-cv-qualifiers, the 5730un-qualified function type was incorrectly treated as a substitution 5731candidate. 5732 5733This was fixed in @option{-fabi-version=8}, the default for GCC 5.1. 5734 5735@item 5736@code{decltype(nullptr)} incorrectly had an alignment of 1, leading to 5737unaligned accesses. Note that this did not affect the ABI of a 5738function with a @code{nullptr_t} parameter, as parameters have a 5739minimum alignment. 5740 5741This was fixed in @option{-fabi-version=9}, the default for GCC 5.2. 5742 5743@item 5744Target-specific attributes that affect the identity of a type, such as 5745ia32 calling conventions on a function type (stdcall, regparm, etc.), 5746did not affect the mangled name, leading to name collisions when 5747function pointers were used as template arguments. 5748 5749This was fixed in @option{-fabi-version=10}, the default for GCC 6.1. 5750 5751@end itemize 5752 5753This option also enables warnings about psABI-related changes. 5754The known psABI changes at this point include: 5755 5756@itemize @bullet 5757 5758@item 5759For SysV/x86-64, unions with @code{long double} members are 5760passed in memory as specified in psABI. Prior to GCC 4.4, this was not 5761the case. For example: 5762 5763@smallexample 5764union U @{ 5765 long double ld; 5766 int i; 5767@}; 5768@end smallexample 5769 5770@noindent 5771@code{union U} is now always passed in memory. 5772 5773@end itemize 5774 5775@item -Wchar-subscripts 5776@opindex Wchar-subscripts 5777@opindex Wno-char-subscripts 5778Warn if an array subscript has type @code{char}. This is a common cause 5779of error, as programmers often forget that this type is signed on some 5780machines. 5781This warning is enabled by @option{-Wall}. 5782 5783@item -Wno-coverage-mismatch 5784@opindex Wno-coverage-mismatch 5785@opindex Wcoverage-mismatch 5786Warn if feedback profiles do not match when using the 5787@option{-fprofile-use} option. 5788If a source file is changed between compiling with @option{-fprofile-generate} 5789and with @option{-fprofile-use}, the files with the profile feedback can fail 5790to match the source file and GCC cannot use the profile feedback 5791information. By default, this warning is enabled and is treated as an 5792error. @option{-Wno-coverage-mismatch} can be used to disable the 5793warning or @option{-Wno-error=coverage-mismatch} can be used to 5794disable the error. Disabling the error for this warning can result in 5795poorly optimized code and is useful only in the 5796case of very minor changes such as bug fixes to an existing code-base. 5797Completely disabling the warning is not recommended. 5798 5799@item -Wno-cpp 5800@r{(C, Objective-C, C++, Objective-C++ and Fortran only)} 5801@opindex Wno-cpp 5802@opindex Wcpp 5803Suppress warning messages emitted by @code{#warning} directives. 5804 5805@item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)} 5806@opindex Wdouble-promotion 5807@opindex Wno-double-promotion 5808Give a warning when a value of type @code{float} is implicitly 5809promoted to @code{double}. CPUs with a 32-bit ``single-precision'' 5810floating-point unit implement @code{float} in hardware, but emulate 5811@code{double} in software. On such a machine, doing computations 5812using @code{double} values is much more expensive because of the 5813overhead required for software emulation. 5814 5815It is easy to accidentally do computations with @code{double} because 5816floating-point literals are implicitly of type @code{double}. For 5817example, in: 5818@smallexample 5819@group 5820float area(float radius) 5821@{ 5822 return 3.14159 * radius * radius; 5823@} 5824@end group 5825@end smallexample 5826the compiler performs the entire computation with @code{double} 5827because the floating-point literal is a @code{double}. 5828 5829@item -Wduplicate-decl-specifier @r{(C and Objective-C only)} 5830@opindex Wduplicate-decl-specifier 5831@opindex Wno-duplicate-decl-specifier 5832Warn if a declaration has duplicate @code{const}, @code{volatile}, 5833@code{restrict} or @code{_Atomic} specifier. This warning is enabled by 5834@option{-Wall}. 5835 5836@item -Wformat 5837@itemx -Wformat=@var{n} 5838@opindex Wformat 5839@opindex Wno-format 5840@opindex ffreestanding 5841@opindex fno-builtin 5842@opindex Wformat= 5843Check calls to @code{printf} and @code{scanf}, etc., to make sure that 5844the arguments supplied have types appropriate to the format string 5845specified, and that the conversions specified in the format string make 5846sense. This includes standard functions, and others specified by format 5847attributes (@pxref{Function Attributes}), in the @code{printf}, 5848@code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension, 5849not in the C standard) families (or other target-specific families). 5850Which functions are checked without format attributes having been 5851specified depends on the standard version selected, and such checks of 5852functions without the attribute specified are disabled by 5853@option{-ffreestanding} or @option{-fno-builtin}. 5854 5855The formats are checked against the format features supported by GNU 5856libc version 2.2. These include all ISO C90 and C99 features, as well 5857as features from the Single Unix Specification and some BSD and GNU 5858extensions. Other library implementations may not support all these 5859features; GCC does not support warning about features that go beyond a 5860particular library's limitations. However, if @option{-Wpedantic} is used 5861with @option{-Wformat}, warnings are given about format features not 5862in the selected standard version (but not for @code{strfmon} formats, 5863since those are not in any version of the C standard). @xref{C Dialect 5864Options,,Options Controlling C Dialect}. 5865 5866@table @gcctabopt 5867@item -Wformat=1 5868@itemx -Wformat 5869@opindex Wformat 5870@opindex Wformat=1 5871Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and 5872@option{-Wno-format} is equivalent to @option{-Wformat=0}. Since 5873@option{-Wformat} also checks for null format arguments for several 5874functions, @option{-Wformat} also implies @option{-Wnonnull}. Some 5875aspects of this level of format checking can be disabled by the 5876options: @option{-Wno-format-contains-nul}, 5877@option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}. 5878@option{-Wformat} is enabled by @option{-Wall}. 5879 5880@item -Wformat=2 5881@opindex Wformat=2 5882Enable @option{-Wformat} plus additional format checks. Currently 5883equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security 5884-Wformat-y2k}. 5885@end table 5886 5887@item -Wno-format-contains-nul 5888@opindex Wno-format-contains-nul 5889@opindex Wformat-contains-nul 5890If @option{-Wformat} is specified, do not warn about format strings that 5891contain NUL bytes. 5892 5893@item -Wno-format-extra-args 5894@opindex Wno-format-extra-args 5895@opindex Wformat-extra-args 5896If @option{-Wformat} is specified, do not warn about excess arguments to a 5897@code{printf} or @code{scanf} format function. The C standard specifies 5898that such arguments are ignored. 5899 5900Where the unused arguments lie between used arguments that are 5901specified with @samp{$} operand number specifications, normally 5902warnings are still given, since the implementation could not know what 5903type to pass to @code{va_arg} to skip the unused arguments. However, 5904in the case of @code{scanf} formats, this option suppresses the 5905warning if the unused arguments are all pointers, since the Single 5906Unix Specification says that such unused arguments are allowed. 5907 5908@item -Wformat-overflow 5909@itemx -Wformat-overflow=@var{level} 5910@opindex Wformat-overflow 5911@opindex Wno-format-overflow 5912Warn about calls to formatted input/output functions such as @code{sprintf} 5913and @code{vsprintf} that might overflow the destination buffer. When the 5914exact number of bytes written by a format directive cannot be determined 5915at compile-time it is estimated based on heuristics that depend on the 5916@var{level} argument and on optimization. While enabling optimization 5917will in most cases improve the accuracy of the warning, it may also 5918result in false positives. 5919 5920@table @gcctabopt 5921@item -Wformat-overflow 5922@itemx -Wformat-overflow=1 5923@opindex Wformat-overflow 5924@opindex Wno-format-overflow 5925Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat} 5926employs a conservative approach that warns only about calls that most 5927likely overflow the buffer. At this level, numeric arguments to format 5928directives with unknown values are assumed to have the value of one, and 5929strings of unknown length to be empty. Numeric arguments that are known 5930to be bounded to a subrange of their type, or string arguments whose output 5931is bounded either by their directive's precision or by a finite set of 5932string literals, are assumed to take on the value within the range that 5933results in the most bytes on output. For example, the call to @code{sprintf} 5934below is diagnosed because even with both @var{a} and @var{b} equal to zero, 5935the terminating NUL character (@code{'\0'}) appended by the function 5936to the destination buffer will be written past its end. Increasing 5937the size of the buffer by a single byte is sufficient to avoid the 5938warning, though it may not be sufficient to avoid the overflow. 5939 5940@smallexample 5941void f (int a, int b) 5942@{ 5943 char buf [13]; 5944 sprintf (buf, "a = %i, b = %i\n", a, b); 5945@} 5946@end smallexample 5947 5948@item -Wformat-overflow=2 5949Level @var{2} warns also about calls that might overflow the destination 5950buffer given an argument of sufficient length or magnitude. At level 5951@var{2}, unknown numeric arguments are assumed to have the minimum 5952representable value for signed types with a precision greater than 1, and 5953the maximum representable value otherwise. Unknown string arguments whose 5954length cannot be assumed to be bounded either by the directive's precision, 5955or by a finite set of string literals they may evaluate to, or the character 5956array they may point to, are assumed to be 1 character long. 5957 5958At level @var{2}, the call in the example above is again diagnosed, but 5959this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first 5960@code{%i} directive will write some of its digits beyond the end of 5961the destination buffer. To make the call safe regardless of the values 5962of the two variables, the size of the destination buffer must be increased 5963to at least 34 bytes. GCC includes the minimum size of the buffer in 5964an informational note following the warning. 5965 5966An alternative to increasing the size of the destination buffer is to 5967constrain the range of formatted values. The maximum length of string 5968arguments can be bounded by specifying the precision in the format 5969directive. When numeric arguments of format directives can be assumed 5970to be bounded by less than the precision of their type, choosing 5971an appropriate length modifier to the format specifier will reduce 5972the required buffer size. For example, if @var{a} and @var{b} in the 5973example above can be assumed to be within the precision of 5974the @code{short int} type then using either the @code{%hi} format 5975directive or casting the argument to @code{short} reduces the maximum 5976required size of the buffer to 24 bytes. 5977 5978@smallexample 5979void f (int a, int b) 5980@{ 5981 char buf [23]; 5982 sprintf (buf, "a = %hi, b = %i\n", a, (short)b); 5983@} 5984@end smallexample 5985@end table 5986 5987@item -Wno-format-zero-length 5988@opindex Wno-format-zero-length 5989@opindex Wformat-zero-length 5990If @option{-Wformat} is specified, do not warn about zero-length formats. 5991The C standard specifies that zero-length formats are allowed. 5992 5993@item -Wformat-nonliteral 5994@opindex Wformat-nonliteral 5995@opindex Wno-format-nonliteral 5996If @option{-Wformat} is specified, also warn if the format string is not a 5997string literal and so cannot be checked, unless the format function 5998takes its format arguments as a @code{va_list}. 5999 6000@item -Wformat-security 6001@opindex Wformat-security 6002@opindex Wno-format-security 6003If @option{-Wformat} is specified, also warn about uses of format 6004functions that represent possible security problems. At present, this 6005warns about calls to @code{printf} and @code{scanf} functions where the 6006format string is not a string literal and there are no format arguments, 6007as in @code{printf (foo);}. This may be a security hole if the format 6008string came from untrusted input and contains @samp{%n}. (This is 6009currently a subset of what @option{-Wformat-nonliteral} warns about, but 6010in future warnings may be added to @option{-Wformat-security} that are not 6011included in @option{-Wformat-nonliteral}.) 6012 6013@item -Wformat-signedness 6014@opindex Wformat-signedness 6015@opindex Wno-format-signedness 6016If @option{-Wformat} is specified, also warn if the format string 6017requires an unsigned argument and the argument is signed and vice versa. 6018 6019@item -Wformat-truncation 6020@itemx -Wformat-truncation=@var{level} 6021@opindex Wformat-truncation 6022@opindex Wno-format-truncation 6023Warn about calls to formatted input/output functions such as @code{snprintf} 6024and @code{vsnprintf} that might result in output truncation. When the exact 6025number of bytes written by a format directive cannot be determined at 6026compile-time it is estimated based on heuristics that depend on 6027the @var{level} argument and on optimization. While enabling optimization 6028will in most cases improve the accuracy of the warning, it may also result 6029in false positives. Except as noted otherwise, the option uses the same 6030logic @option{-Wformat-overflow}. 6031 6032@table @gcctabopt 6033@item -Wformat-truncation 6034@itemx -Wformat-truncation=1 6035@opindex Wformat-truncation 6036@opindex Wno-format-truncation 6037Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat} 6038employs a conservative approach that warns only about calls to bounded 6039functions whose return value is unused and that will most likely result 6040in output truncation. 6041 6042@item -Wformat-truncation=2 6043Level @var{2} warns also about calls to bounded functions whose return 6044value is used and that might result in truncation given an argument of 6045sufficient length or magnitude. 6046@end table 6047 6048@item -Wformat-y2k 6049@opindex Wformat-y2k 6050@opindex Wno-format-y2k 6051If @option{-Wformat} is specified, also warn about @code{strftime} 6052formats that may yield only a two-digit year. 6053 6054@item -Wnonnull 6055@opindex Wnonnull 6056@opindex Wno-nonnull 6057Warn about passing a null pointer for arguments marked as 6058requiring a non-null value by the @code{nonnull} function attribute. 6059 6060@option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It 6061can be disabled with the @option{-Wno-nonnull} option. 6062 6063@item -Wnonnull-compare 6064@opindex Wnonnull-compare 6065@opindex Wno-nonnull-compare 6066Warn when comparing an argument marked with the @code{nonnull} 6067function attribute against null inside the function. 6068 6069@option{-Wnonnull-compare} is included in @option{-Wall}. It 6070can be disabled with the @option{-Wno-nonnull-compare} option. 6071 6072@item -Wnull-dereference 6073@opindex Wnull-dereference 6074@opindex Wno-null-dereference 6075Warn if the compiler detects paths that trigger erroneous or 6076undefined behavior due to dereferencing a null pointer. This option 6077is only active when @option{-fdelete-null-pointer-checks} is active, 6078which is enabled by optimizations in most targets. The precision of 6079the warnings depends on the optimization options used. 6080 6081@item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)} 6082@opindex Winit-self 6083@opindex Wno-init-self 6084Warn about uninitialized variables that are initialized with themselves. 6085Note this option can only be used with the @option{-Wuninitialized} option. 6086 6087For example, GCC warns about @code{i} being uninitialized in the 6088following snippet only when @option{-Winit-self} has been specified: 6089@smallexample 6090@group 6091int f() 6092@{ 6093 int i = i; 6094 return i; 6095@} 6096@end group 6097@end smallexample 6098 6099This warning is enabled by @option{-Wall} in C++. 6100 6101@item -Wno-implicit-int @r{(C and Objective-C only)} 6102@opindex Wimplicit-int 6103@opindex Wno-implicit-int 6104This option controls warnings when a declaration does not specify a type. 6105This warning is enabled by default in C99 and later dialects of C, 6106and also by @option{-Wall}. 6107 6108@item -Wno-implicit-function-declaration @r{(C and Objective-C only)} 6109@opindex Wimplicit-function-declaration 6110@opindex Wno-implicit-function-declaration 6111This option controls warnings when a function is used before being declared. 6112This warning is enabled by default in C99 and later dialects of C, 6113and also by @option{-Wall}. 6114The warning is made into an error by @option{-pedantic-errors}. 6115 6116@item -Wimplicit @r{(C and Objective-C only)} 6117@opindex Wimplicit 6118@opindex Wno-implicit 6119Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}. 6120This warning is enabled by @option{-Wall}. 6121 6122@item -Wimplicit-fallthrough 6123@opindex Wimplicit-fallthrough 6124@opindex Wno-implicit-fallthrough 6125@option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3} 6126and @option{-Wno-implicit-fallthrough} is the same as 6127@option{-Wimplicit-fallthrough=0}. 6128 6129@item -Wimplicit-fallthrough=@var{n} 6130@opindex Wimplicit-fallthrough= 6131Warn when a switch case falls through. For example: 6132 6133@smallexample 6134@group 6135switch (cond) 6136 @{ 6137 case 1: 6138 a = 1; 6139 break; 6140 case 2: 6141 a = 2; 6142 case 3: 6143 a = 3; 6144 break; 6145 @} 6146@end group 6147@end smallexample 6148 6149This warning does not warn when the last statement of a case cannot 6150fall through, e.g. when there is a return statement or a call to function 6151declared with the noreturn attribute. @option{-Wimplicit-fallthrough=} 6152also takes into account control flow statements, such as ifs, and only 6153warns when appropriate. E.g.@: 6154 6155@smallexample 6156@group 6157switch (cond) 6158 @{ 6159 case 1: 6160 if (i > 3) @{ 6161 bar (5); 6162 break; 6163 @} else if (i < 1) @{ 6164 bar (0); 6165 @} else 6166 return; 6167 default: 6168 @dots{} 6169 @} 6170@end group 6171@end smallexample 6172 6173Since there are occasions where a switch case fall through is desirable, 6174GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is 6175to be used along with a null statement to suppress this warning that 6176would normally occur: 6177 6178@smallexample 6179@group 6180switch (cond) 6181 @{ 6182 case 1: 6183 bar (0); 6184 __attribute__ ((fallthrough)); 6185 default: 6186 @dots{} 6187 @} 6188@end group 6189@end smallexample 6190 6191C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough} 6192warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11 6193or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension. 6194Instead of these attributes, it is also possible to add a fallthrough comment 6195to silence the warning. The whole body of the C or C++ style comment should 6196match the given regular expressions listed below. The option argument @var{n} 6197specifies what kind of comments are accepted: 6198 6199@itemize @bullet 6200 6201@item @option{-Wimplicit-fallthrough=0} disables the warning altogether. 6202 6203@item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular 6204expression, any comment is used as fallthrough comment. 6205 6206@item @option{-Wimplicit-fallthrough=2} case insensitively matches 6207@code{.*falls?[ \t-]*thr(ough|u).*} regular expression. 6208 6209@item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the 6210following regular expressions: 6211 6212@itemize @bullet 6213 6214@item @code{-fallthrough} 6215 6216@item @code{@@fallthrough@@} 6217 6218@item @code{lint -fallthrough[ \t]*} 6219 6220@item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?} 6221 6222@item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?} 6223 6224@item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?} 6225 6226@end itemize 6227 6228@item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the 6229following regular expressions: 6230 6231@itemize @bullet 6232 6233@item @code{-fallthrough} 6234 6235@item @code{@@fallthrough@@} 6236 6237@item @code{lint -fallthrough[ \t]*} 6238 6239@item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*} 6240 6241@end itemize 6242 6243@item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as 6244fallthrough comments, only attributes disable the warning. 6245 6246@end itemize 6247 6248The comment needs to be followed after optional whitespace and other comments 6249by @code{case} or @code{default} keywords or by a user label that precedes some 6250@code{case} or @code{default} label. 6251 6252@smallexample 6253@group 6254switch (cond) 6255 @{ 6256 case 1: 6257 bar (0); 6258 /* FALLTHRU */ 6259 default: 6260 @dots{} 6261 @} 6262@end group 6263@end smallexample 6264 6265The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}. 6266 6267@item -Wno-if-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)} 6268@opindex Wif-not-aligned 6269@opindex Wno-if-not-aligned 6270Control if warnings triggered by the @code{warn_if_not_aligned} attribute 6271should be issued. These warnings are enabled by default. 6272 6273@item -Wignored-qualifiers @r{(C and C++ only)} 6274@opindex Wignored-qualifiers 6275@opindex Wno-ignored-qualifiers 6276Warn if the return type of a function has a type qualifier 6277such as @code{const}. For ISO C such a type qualifier has no effect, 6278since the value returned by a function is not an lvalue. 6279For C++, the warning is only emitted for scalar types or @code{void}. 6280ISO C prohibits qualified @code{void} return types on function 6281definitions, so such return types always receive a warning 6282even without this option. 6283 6284This warning is also enabled by @option{-Wextra}. 6285 6286@item -Wno-ignored-attributes @r{(C and C++ only)} 6287@opindex Wignored-attributes 6288@opindex Wno-ignored-attributes 6289This option controls warnings when an attribute is ignored. 6290This is different from the 6291@option{-Wattributes} option in that it warns whenever the compiler decides 6292to drop an attribute, not that the attribute is either unknown, used in a 6293wrong place, etc. This warning is enabled by default. 6294 6295@item -Wmain 6296@opindex Wmain 6297@opindex Wno-main 6298Warn if the type of @code{main} is suspicious. @code{main} should be 6299a function with external linkage, returning int, taking either zero 6300arguments, two, or three arguments of appropriate types. This warning 6301is enabled by default in C++ and is enabled by either @option{-Wall} 6302or @option{-Wpedantic}. 6303 6304@item -Wmisleading-indentation @r{(C and C++ only)} 6305@opindex Wmisleading-indentation 6306@opindex Wno-misleading-indentation 6307Warn when the indentation of the code does not reflect the block structure. 6308Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and 6309@code{for} clauses with a guarded statement that does not use braces, 6310followed by an unguarded statement with the same indentation. 6311 6312In the following example, the call to ``bar'' is misleadingly indented as 6313if it were guarded by the ``if'' conditional. 6314 6315@smallexample 6316 if (some_condition ()) 6317 foo (); 6318 bar (); /* Gotcha: this is not guarded by the "if". */ 6319@end smallexample 6320 6321In the case of mixed tabs and spaces, the warning uses the 6322@option{-ftabstop=} option to determine if the statements line up 6323(defaulting to 8). 6324 6325The warning is not issued for code involving multiline preprocessor logic 6326such as the following example. 6327 6328@smallexample 6329 if (flagA) 6330 foo (0); 6331#if SOME_CONDITION_THAT_DOES_NOT_HOLD 6332 if (flagB) 6333#endif 6334 foo (1); 6335@end smallexample 6336 6337The warning is not issued after a @code{#line} directive, since this 6338typically indicates autogenerated code, and no assumptions can be made 6339about the layout of the file that the directive references. 6340 6341This warning is enabled by @option{-Wall} in C and C++. 6342 6343@item -Wmissing-attributes 6344@opindex Wmissing-attributes 6345@opindex Wno-missing-attributes 6346Warn when a declaration of a function is missing one or more attributes 6347that a related function is declared with and whose absence may adversely 6348affect the correctness or efficiency of generated code. For example, 6349the warning is issued for declarations of aliases that use attributes 6350to specify less restrictive requirements than those of their targets. 6351This typically represents a potential optimization opportunity. 6352By contrast, the @option{-Wattribute-alias=2} option controls warnings 6353issued when the alias is more restrictive than the target, which could 6354lead to incorrect code generation. 6355Attributes considered include @code{alloc_align}, @code{alloc_size}, 6356@code{cold}, @code{const}, @code{hot}, @code{leaf}, @code{malloc}, 6357@code{nonnull}, @code{noreturn}, @code{nothrow}, @code{pure}, 6358@code{returns_nonnull}, and @code{returns_twice}. 6359 6360In C++, the warning is issued when an explicit specialization of a primary 6361template declared with attribute @code{alloc_align}, @code{alloc_size}, 6362@code{assume_aligned}, @code{format}, @code{format_arg}, @code{malloc}, 6363or @code{nonnull} is declared without it. Attributes @code{deprecated}, 6364@code{error}, and @code{warning} suppress the warning. 6365(@pxref{Function Attributes}). 6366 6367You can use the @code{copy} attribute to apply the same 6368set of attributes to a declaration as that on another declaration without 6369explicitly enumerating the attributes. This attribute can be applied 6370to declarations of functions (@pxref{Common Function Attributes}), 6371variables (@pxref{Common Variable Attributes}), or types 6372(@pxref{Common Type Attributes}). 6373 6374@option{-Wmissing-attributes} is enabled by @option{-Wall}. 6375 6376For example, since the declaration of the primary function template 6377below makes use of both attribute @code{malloc} and @code{alloc_size} 6378the declaration of the explicit specialization of the template is 6379diagnosed because it is missing one of the attributes. 6380 6381@smallexample 6382template <class T> 6383T* __attribute__ ((malloc, alloc_size (1))) 6384allocate (size_t); 6385 6386template <> 6387void* __attribute__ ((malloc)) // missing alloc_size 6388allocate<void> (size_t); 6389@end smallexample 6390 6391@item -Wmissing-braces 6392@opindex Wmissing-braces 6393@opindex Wno-missing-braces 6394Warn if an aggregate or union initializer is not fully bracketed. In 6395the following example, the initializer for @code{a} is not fully 6396bracketed, but that for @code{b} is fully bracketed. 6397 6398@smallexample 6399int a[2][2] = @{ 0, 1, 2, 3 @}; 6400int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @}; 6401@end smallexample 6402 6403This warning is enabled by @option{-Wall}. 6404 6405@item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)} 6406@opindex Wmissing-include-dirs 6407@opindex Wno-missing-include-dirs 6408Warn if a user-supplied include directory does not exist. 6409 6410@item -Wno-missing-profile 6411@opindex Wmissing-profile 6412@opindex Wno-missing-profile 6413This option controls warnings if feedback profiles are missing when using the 6414@option{-fprofile-use} option. 6415This option diagnoses those cases where a new function or a new file is added 6416between compiling with @option{-fprofile-generate} and with 6417@option{-fprofile-use}, without regenerating the profiles. 6418In these cases, the profile feedback data files do not contain any 6419profile feedback information for 6420the newly added function or file respectively. Also, in the case when profile 6421count data (.gcda) files are removed, GCC cannot use any profile feedback 6422information. In all these cases, warnings are issued to inform you that a 6423profile generation step is due. 6424Ignoring the warning can result in poorly optimized code. 6425@option{-Wno-missing-profile} can be used to 6426disable the warning, but this is not recommended and should be done only 6427when non-existent profile data is justified. 6428 6429@item -Wno-mismatched-dealloc 6430@opindex Wmismatched-dealloc 6431@opindex Wno-mismatched-dealloc 6432 6433Warn for calls to deallocation functions with pointer arguments returned 6434from from allocations functions for which the former isn't a suitable 6435deallocator. A pair of functions can be associated as matching allocators 6436and deallocators by use of attribute @code{malloc}. Unless disabled by 6437the @option{-fno-builtin} option the standard functions @code{calloc}, 6438@code{malloc}, @code{realloc}, and @code{free}, as well as the corresponding 6439forms of C++ @code{operator new} and @code{operator delete} are implicitly 6440associated as matching allocators and deallocators. In the following 6441example @code{mydealloc} is the deallocator for pointers returned from 6442@code{myalloc}. 6443 6444@smallexample 6445void mydealloc (void*); 6446 6447__attribute__ ((malloc (mydealloc, 1))) void* 6448myalloc (size_t); 6449 6450void f (void) 6451@{ 6452 void *p = myalloc (32); 6453 // @dots{}use p@dots{} 6454 free (p); // warning: not a matching deallocator for myalloc 6455 mydealloc (p); // ok 6456@} 6457@end smallexample 6458 6459In C++, the related option @option{-Wmismatched-new-delete} diagnoses 6460mismatches involving either @code{operator new} or @code{operator delete}. 6461 6462Option @option{-Wmismatched-dealloc} is enabled by default. 6463 6464@item -Wmultistatement-macros 6465@opindex Wmultistatement-macros 6466@opindex Wno-multistatement-macros 6467Warn about unsafe multiple statement macros that appear to be guarded 6468by a clause such as @code{if}, @code{else}, @code{for}, @code{switch}, or 6469@code{while}, in which only the first statement is actually guarded after 6470the macro is expanded. 6471 6472For example: 6473 6474@smallexample 6475#define DOIT x++; y++ 6476if (c) 6477 DOIT; 6478@end smallexample 6479 6480will increment @code{y} unconditionally, not just when @code{c} holds. 6481The can usually be fixed by wrapping the macro in a do-while loop: 6482@smallexample 6483#define DOIT do @{ x++; y++; @} while (0) 6484if (c) 6485 DOIT; 6486@end smallexample 6487 6488This warning is enabled by @option{-Wall} in C and C++. 6489 6490@item -Wparentheses 6491@opindex Wparentheses 6492@opindex Wno-parentheses 6493Warn if parentheses are omitted in certain contexts, such 6494as when there is an assignment in a context where a truth value 6495is expected, or when operators are nested whose precedence people 6496often get confused about. 6497 6498Also warn if a comparison like @code{x<=y<=z} appears; this is 6499equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different 6500interpretation from that of ordinary mathematical notation. 6501 6502Also warn for dangerous uses of the GNU extension to 6503@code{?:} with omitted middle operand. When the condition 6504in the @code{?}: operator is a boolean expression, the omitted value is 6505always 1. Often programmers expect it to be a value computed 6506inside the conditional expression instead. 6507 6508For C++ this also warns for some cases of unnecessary parentheses in 6509declarations, which can indicate an attempt at a function call instead 6510of a declaration: 6511@smallexample 6512@{ 6513 // Declares a local variable called mymutex. 6514 std::unique_lock<std::mutex> (mymutex); 6515 // User meant std::unique_lock<std::mutex> lock (mymutex); 6516@} 6517@end smallexample 6518 6519This warning is enabled by @option{-Wall}. 6520 6521@item -Wsequence-point 6522@opindex Wsequence-point 6523@opindex Wno-sequence-point 6524Warn about code that may have undefined semantics because of violations 6525of sequence point rules in the C and C++ standards. 6526 6527The C and C++ standards define the order in which expressions in a C/C++ 6528program are evaluated in terms of @dfn{sequence points}, which represent 6529a partial ordering between the execution of parts of the program: those 6530executed before the sequence point, and those executed after it. These 6531occur after the evaluation of a full expression (one which is not part 6532of a larger expression), after the evaluation of the first operand of a 6533@code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a 6534function is called (but after the evaluation of its arguments and the 6535expression denoting the called function), and in certain other places. 6536Other than as expressed by the sequence point rules, the order of 6537evaluation of subexpressions of an expression is not specified. All 6538these rules describe only a partial order rather than a total order, 6539since, for example, if two functions are called within one expression 6540with no sequence point between them, the order in which the functions 6541are called is not specified. However, the standards committee have 6542ruled that function calls do not overlap. 6543 6544It is not specified when between sequence points modifications to the 6545values of objects take effect. Programs whose behavior depends on this 6546have undefined behavior; the C and C++ standards specify that ``Between 6547the previous and next sequence point an object shall have its stored 6548value modified at most once by the evaluation of an expression. 6549Furthermore, the prior value shall be read only to determine the value 6550to be stored.''. If a program breaks these rules, the results on any 6551particular implementation are entirely unpredictable. 6552 6553Examples of code with undefined behavior are @code{a = a++;}, @code{a[n] 6554= b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not 6555diagnosed by this option, and it may give an occasional false positive 6556result, but in general it has been found fairly effective at detecting 6557this sort of problem in programs. 6558 6559The C++17 standard will define the order of evaluation of operands in 6560more cases: in particular it requires that the right-hand side of an 6561assignment be evaluated before the left-hand side, so the above 6562examples are no longer undefined. But this option will still warn 6563about them, to help people avoid writing code that is undefined in C 6564and earlier revisions of C++. 6565 6566The standard is worded confusingly, therefore there is some debate 6567over the precise meaning of the sequence point rules in subtle cases. 6568Links to discussions of the problem, including proposed formal 6569definitions, may be found on the GCC readings page, at 6570@uref{http://gcc.gnu.org/@/readings.html}. 6571 6572This warning is enabled by @option{-Wall} for C and C++. 6573 6574@item -Wno-return-local-addr 6575@opindex Wno-return-local-addr 6576@opindex Wreturn-local-addr 6577Do not warn about returning a pointer (or in C++, a reference) to a 6578variable that goes out of scope after the function returns. 6579 6580@item -Wreturn-type 6581@opindex Wreturn-type 6582@opindex Wno-return-type 6583Warn whenever a function is defined with a return type that defaults 6584to @code{int}. Also warn about any @code{return} statement with no 6585return value in a function whose return type is not @code{void} 6586(falling off the end of the function body is considered returning 6587without a value). 6588 6589For C only, warn about a @code{return} statement with an expression in a 6590function whose return type is @code{void}, unless the expression type is 6591also @code{void}. As a GNU extension, the latter case is accepted 6592without a warning unless @option{-Wpedantic} is used. Attempting 6593to use the return value of a non-@code{void} function other than @code{main} 6594that flows off the end by reaching the closing curly brace that terminates 6595the function is undefined. 6596 6597Unlike in C, in C++, flowing off the end of a non-@code{void} function other 6598than @code{main} results in undefined behavior even when the value of 6599the function is not used. 6600 6601This warning is enabled by default in C++ and by @option{-Wall} otherwise. 6602 6603@item -Wno-shift-count-negative 6604@opindex Wshift-count-negative 6605@opindex Wno-shift-count-negative 6606Controls warnings if a shift count is negative. 6607This warning is enabled by default. 6608 6609@item -Wno-shift-count-overflow 6610@opindex Wshift-count-overflow 6611@opindex Wno-shift-count-overflow 6612Controls warnings if a shift count is greater than or equal to the bit width 6613of the type. This warning is enabled by default. 6614 6615@item -Wshift-negative-value 6616@opindex Wshift-negative-value 6617@opindex Wno-shift-negative-value 6618Warn if left shifting a negative value. This warning is enabled by 6619@option{-Wextra} in C99 and C++11 modes (and newer). 6620 6621@item -Wno-shift-overflow 6622@itemx -Wshift-overflow=@var{n} 6623@opindex Wshift-overflow 6624@opindex Wno-shift-overflow 6625These options control warnings about left shift overflows. 6626 6627@table @gcctabopt 6628@item -Wshift-overflow=1 6629This is the warning level of @option{-Wshift-overflow} and is enabled 6630by default in C99 and C++11 modes (and newer). This warning level does 6631not warn about left-shifting 1 into the sign bit. (However, in C, such 6632an overflow is still rejected in contexts where an integer constant expression 6633is required.) No warning is emitted in C++20 mode (and newer), as signed left 6634shifts always wrap. 6635 6636@item -Wshift-overflow=2 6637This warning level also warns about left-shifting 1 into the sign bit, 6638unless C++14 mode (or newer) is active. 6639@end table 6640 6641@item -Wswitch 6642@opindex Wswitch 6643@opindex Wno-switch 6644Warn whenever a @code{switch} statement has an index of enumerated type 6645and lacks a @code{case} for one or more of the named codes of that 6646enumeration. (The presence of a @code{default} label prevents this 6647warning.) @code{case} labels outside the enumeration range also 6648provoke warnings when this option is used (even if there is a 6649@code{default} label). 6650This warning is enabled by @option{-Wall}. 6651 6652@item -Wswitch-default 6653@opindex Wswitch-default 6654@opindex Wno-switch-default 6655Warn whenever a @code{switch} statement does not have a @code{default} 6656case. 6657 6658@item -Wswitch-enum 6659@opindex Wswitch-enum 6660@opindex Wno-switch-enum 6661Warn whenever a @code{switch} statement has an index of enumerated type 6662and lacks a @code{case} for one or more of the named codes of that 6663enumeration. @code{case} labels outside the enumeration range also 6664provoke warnings when this option is used. The only difference 6665between @option{-Wswitch} and this option is that this option gives a 6666warning about an omitted enumeration code even if there is a 6667@code{default} label. 6668 6669@item -Wno-switch-bool 6670@opindex Wswitch-bool 6671@opindex Wno-switch-bool 6672Do not warn when a @code{switch} statement has an index of boolean type 6673and the case values are outside the range of a boolean type. 6674It is possible to suppress this warning by casting the controlling 6675expression to a type other than @code{bool}. For example: 6676@smallexample 6677@group 6678switch ((int) (a == 4)) 6679 @{ 6680 @dots{} 6681 @} 6682@end group 6683@end smallexample 6684This warning is enabled by default for C and C++ programs. 6685 6686@item -Wno-switch-outside-range 6687@opindex Wswitch-outside-range 6688@opindex Wno-switch-outside-range 6689This option controls warnings when a @code{switch} case has a value 6690that is outside of its 6691respective type range. This warning is enabled by default for 6692C and C++ programs. 6693 6694@item -Wno-switch-unreachable 6695@opindex Wswitch-unreachable 6696@opindex Wno-switch-unreachable 6697Do not warn when a @code{switch} statement contains statements between the 6698controlling expression and the first case label, which will never be 6699executed. For example: 6700@smallexample 6701@group 6702switch (cond) 6703 @{ 6704 i = 15; 6705 @dots{} 6706 case 5: 6707 @dots{} 6708 @} 6709@end group 6710@end smallexample 6711@option{-Wswitch-unreachable} does not warn if the statement between the 6712controlling expression and the first case label is just a declaration: 6713@smallexample 6714@group 6715switch (cond) 6716 @{ 6717 int i; 6718 @dots{} 6719 case 5: 6720 i = 5; 6721 @dots{} 6722 @} 6723@end group 6724@end smallexample 6725This warning is enabled by default for C and C++ programs. 6726 6727@item -Wsync-nand @r{(C and C++ only)} 6728@opindex Wsync-nand 6729@opindex Wno-sync-nand 6730Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch} 6731built-in functions are used. These functions changed semantics in GCC 4.4. 6732 6733@item -Wunused-but-set-parameter 6734@opindex Wunused-but-set-parameter 6735@opindex Wno-unused-but-set-parameter 6736Warn whenever a function parameter is assigned to, but otherwise unused 6737(aside from its declaration). 6738 6739To suppress this warning use the @code{unused} attribute 6740(@pxref{Variable Attributes}). 6741 6742This warning is also enabled by @option{-Wunused} together with 6743@option{-Wextra}. 6744 6745@item -Wunused-but-set-variable 6746@opindex Wunused-but-set-variable 6747@opindex Wno-unused-but-set-variable 6748Warn whenever a local variable is assigned to, but otherwise unused 6749(aside from its declaration). 6750This warning is enabled by @option{-Wall}. 6751 6752To suppress this warning use the @code{unused} attribute 6753(@pxref{Variable Attributes}). 6754 6755This warning is also enabled by @option{-Wunused}, which is enabled 6756by @option{-Wall}. 6757 6758@item -Wunused-function 6759@opindex Wunused-function 6760@opindex Wno-unused-function 6761Warn whenever a static function is declared but not defined or a 6762non-inline static function is unused. 6763This warning is enabled by @option{-Wall}. 6764 6765@item -Wunused-label 6766@opindex Wunused-label 6767@opindex Wno-unused-label 6768Warn whenever a label is declared but not used. 6769This warning is enabled by @option{-Wall}. 6770 6771To suppress this warning use the @code{unused} attribute 6772(@pxref{Variable Attributes}). 6773 6774@item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)} 6775@opindex Wunused-local-typedefs 6776@opindex Wno-unused-local-typedefs 6777Warn when a typedef locally defined in a function is not used. 6778This warning is enabled by @option{-Wall}. 6779 6780@item -Wunused-parameter 6781@opindex Wunused-parameter 6782@opindex Wno-unused-parameter 6783Warn whenever a function parameter is unused aside from its declaration. 6784 6785To suppress this warning use the @code{unused} attribute 6786(@pxref{Variable Attributes}). 6787 6788@item -Wno-unused-result 6789@opindex Wunused-result 6790@opindex Wno-unused-result 6791Do not warn if a caller of a function marked with attribute 6792@code{warn_unused_result} (@pxref{Function Attributes}) does not use 6793its return value. The default is @option{-Wunused-result}. 6794 6795@item -Wunused-variable 6796@opindex Wunused-variable 6797@opindex Wno-unused-variable 6798Warn whenever a local or static variable is unused aside from its 6799declaration. This option implies @option{-Wunused-const-variable=1} for C, 6800but not for C++. This warning is enabled by @option{-Wall}. 6801 6802To suppress this warning use the @code{unused} attribute 6803(@pxref{Variable Attributes}). 6804 6805@item -Wunused-const-variable 6806@itemx -Wunused-const-variable=@var{n} 6807@opindex Wunused-const-variable 6808@opindex Wno-unused-const-variable 6809Warn whenever a constant static variable is unused aside from its declaration. 6810@option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable} 6811for C, but not for C++. In C this declares variable storage, but in C++ this 6812is not an error since const variables take the place of @code{#define}s. 6813 6814To suppress this warning use the @code{unused} attribute 6815(@pxref{Variable Attributes}). 6816 6817@table @gcctabopt 6818@item -Wunused-const-variable=1 6819This is the warning level that is enabled by @option{-Wunused-variable} for 6820C. It warns only about unused static const variables defined in the main 6821compilation unit, but not about static const variables declared in any 6822header included. 6823 6824@item -Wunused-const-variable=2 6825This warning level also warns for unused constant static variables in 6826headers (excluding system headers). This is the warning level of 6827@option{-Wunused-const-variable} and must be explicitly requested since 6828in C++ this isn't an error and in C it might be harder to clean up all 6829headers included. 6830@end table 6831 6832@item -Wunused-value 6833@opindex Wunused-value 6834@opindex Wno-unused-value 6835Warn whenever a statement computes a result that is explicitly not 6836used. To suppress this warning cast the unused expression to 6837@code{void}. This includes an expression-statement or the left-hand 6838side of a comma expression that contains no side effects. For example, 6839an expression such as @code{x[i,j]} causes a warning, while 6840@code{x[(void)i,j]} does not. 6841 6842This warning is enabled by @option{-Wall}. 6843 6844@item -Wunused 6845@opindex Wunused 6846@opindex Wno-unused 6847All the above @option{-Wunused} options combined. 6848 6849In order to get a warning about an unused function parameter, you must 6850either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies 6851@option{-Wunused}), or separately specify @option{-Wunused-parameter}. 6852 6853@item -Wuninitialized 6854@opindex Wuninitialized 6855@opindex Wno-uninitialized 6856Warn if an object with automatic or allocated storage duration is used 6857without having been initialized. In C++, also warn if a non-static 6858reference or non-static @code{const} member appears in a class without 6859constructors. 6860 6861In addition, passing a pointer (or in C++, a reference) to an uninitialized 6862object to a @code{const}-qualified argument of a built-in function known to 6863read the object is also diagnosed by this warning. 6864(@option{-Wmaybe-uninitialized} is issued for ordinary functions.) 6865 6866If you want to warn about code that uses the uninitialized value of the 6867variable in its own initializer, use the @option{-Winit-self} option. 6868 6869These warnings occur for individual uninitialized elements of 6870structure, union or array variables as well as for variables that are 6871uninitialized as a whole. They do not occur for variables or elements 6872declared @code{volatile}. Because these warnings depend on 6873optimization, the exact variables or elements for which there are 6874warnings depend on the precise optimization options and version of GCC 6875used. 6876 6877Note that there may be no warning about a variable that is used only 6878to compute a value that itself is never used, because such 6879computations may be deleted by data flow analysis before the warnings 6880are printed. 6881 6882@item -Wno-invalid-memory-model 6883@opindex Winvalid-memory-model 6884@opindex Wno-invalid-memory-model 6885This option controls warnings 6886for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins}, 6887and the C11 atomic generic functions with a memory consistency argument 6888that is either invalid for the operation or outside the range of values 6889of the @code{memory_order} enumeration. For example, since the 6890@code{__atomic_store} and @code{__atomic_store_n} built-ins are only 6891defined for the relaxed, release, and sequentially consistent memory 6892orders the following code is diagnosed: 6893 6894@smallexample 6895void store (int *i) 6896@{ 6897 __atomic_store_n (i, 0, memory_order_consume); 6898@} 6899@end smallexample 6900 6901@option{-Winvalid-memory-model} is enabled by default. 6902 6903@item -Wmaybe-uninitialized 6904@opindex Wmaybe-uninitialized 6905@opindex Wno-maybe-uninitialized 6906For an object with automatic or allocated storage duration, if there exists 6907a path from the function entry to a use of the object that is initialized, 6908but there exist some other paths for which the object is not initialized, 6909the compiler emits a warning if it cannot prove the uninitialized paths 6910are not executed at run time. 6911 6912In addition, passing a pointer (or in C++, a reference) to an uninitialized 6913object to a @code{const}-qualified function argument is also diagnosed by 6914this warning. (@option{-Wuninitialized} is issued for built-in functions 6915known to read the object.) Annotating the function with attribute 6916@code{access (none)} indicates that the argument isn't used to access 6917the object and avoids the warning (@pxref{Common Function Attributes}). 6918 6919These warnings are only possible in optimizing compilation, because otherwise 6920GCC does not keep track of the state of variables. 6921 6922These warnings are made optional because GCC may not be able to determine when 6923the code is correct in spite of appearing to have an error. Here is one 6924example of how this can happen: 6925 6926@smallexample 6927@group 6928@{ 6929 int x; 6930 switch (y) 6931 @{ 6932 case 1: x = 1; 6933 break; 6934 case 2: x = 4; 6935 break; 6936 case 3: x = 5; 6937 @} 6938 foo (x); 6939@} 6940@end group 6941@end smallexample 6942 6943@noindent 6944If the value of @code{y} is always 1, 2 or 3, then @code{x} is 6945always initialized, but GCC doesn't know this. To suppress the 6946warning, you need to provide a default case with assert(0) or 6947similar code. 6948 6949@cindex @code{longjmp} warnings 6950This option also warns when a non-volatile automatic variable might be 6951changed by a call to @code{longjmp}. 6952The compiler sees only the calls to @code{setjmp}. It cannot know 6953where @code{longjmp} will be called; in fact, a signal handler could 6954call it at any point in the code. As a result, you may get a warning 6955even when there is in fact no problem because @code{longjmp} cannot 6956in fact be called at the place that would cause a problem. 6957 6958Some spurious warnings can be avoided if you declare all the functions 6959you use that never return as @code{noreturn}. @xref{Function 6960Attributes}. 6961 6962This warning is enabled by @option{-Wall} or @option{-Wextra}. 6963 6964@item -Wunknown-pragmas 6965@opindex Wunknown-pragmas 6966@opindex Wno-unknown-pragmas 6967@cindex warning for unknown pragmas 6968@cindex unknown pragmas, warning 6969@cindex pragmas, warning of unknown 6970Warn when a @code{#pragma} directive is encountered that is not understood by 6971GCC@. If this command-line option is used, warnings are even issued 6972for unknown pragmas in system header files. This is not the case if 6973the warnings are only enabled by the @option{-Wall} command-line option. 6974 6975@item -Wno-pragmas 6976@opindex Wno-pragmas 6977@opindex Wpragmas 6978Do not warn about misuses of pragmas, such as incorrect parameters, 6979invalid syntax, or conflicts between pragmas. See also 6980@option{-Wunknown-pragmas}. 6981 6982@item -Wno-prio-ctor-dtor 6983@opindex Wno-prio-ctor-dtor 6984@opindex Wprio-ctor-dtor 6985Do not warn if a priority from 0 to 100 is used for constructor or destructor. 6986The use of constructor and destructor attributes allow you to assign a 6987priority to the constructor/destructor to control its order of execution 6988before @code{main} is called or after it returns. The priority values must be 6989greater than 100 as the compiler reserves priority values between 0--100 for 6990the implementation. 6991 6992@item -Wstrict-aliasing 6993@opindex Wstrict-aliasing 6994@opindex Wno-strict-aliasing 6995This option is only active when @option{-fstrict-aliasing} is active. 6996It warns about code that might break the strict aliasing rules that the 6997compiler is using for optimization. The warning does not catch all 6998cases, but does attempt to catch the more common pitfalls. It is 6999included in @option{-Wall}. 7000It is equivalent to @option{-Wstrict-aliasing=3} 7001 7002@item -Wstrict-aliasing=n 7003@opindex Wstrict-aliasing=n 7004This option is only active when @option{-fstrict-aliasing} is active. 7005It warns about code that might break the strict aliasing rules that the 7006compiler is using for optimization. 7007Higher levels correspond to higher accuracy (fewer false positives). 7008Higher levels also correspond to more effort, similar to the way @option{-O} 7009works. 7010@option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}. 7011 7012Level 1: Most aggressive, quick, least accurate. 7013Possibly useful when higher levels 7014do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few 7015false negatives. However, it has many false positives. 7016Warns for all pointer conversions between possibly incompatible types, 7017even if never dereferenced. Runs in the front end only. 7018 7019Level 2: Aggressive, quick, not too precise. 7020May still have many false positives (not as many as level 1 though), 7021and few false negatives (but possibly more than level 1). 7022Unlike level 1, it only warns when an address is taken. Warns about 7023incomplete types. Runs in the front end only. 7024 7025Level 3 (default for @option{-Wstrict-aliasing}): 7026Should have very few false positives and few false 7027negatives. Slightly slower than levels 1 or 2 when optimization is enabled. 7028Takes care of the common pun+dereference pattern in the front end: 7029@code{*(int*)&some_float}. 7030If optimization is enabled, it also runs in the back end, where it deals 7031with multiple statement cases using flow-sensitive points-to information. 7032Only warns when the converted pointer is dereferenced. 7033Does not warn about incomplete types. 7034 7035@item -Wstrict-overflow 7036@itemx -Wstrict-overflow=@var{n} 7037@opindex Wstrict-overflow 7038@opindex Wno-strict-overflow 7039This option is only active when signed overflow is undefined. 7040It warns about cases where the compiler optimizes based on the 7041assumption that signed overflow does not occur. Note that it does not 7042warn about all cases where the code might overflow: it only warns 7043about cases where the compiler implements some optimization. Thus 7044this warning depends on the optimization level. 7045 7046An optimization that assumes that signed overflow does not occur is 7047perfectly safe if the values of the variables involved are such that 7048overflow never does, in fact, occur. Therefore this warning can 7049easily give a false positive: a warning about code that is not 7050actually a problem. To help focus on important issues, several 7051warning levels are defined. No warnings are issued for the use of 7052undefined signed overflow when estimating how many iterations a loop 7053requires, in particular when determining whether a loop will be 7054executed at all. 7055 7056@table @gcctabopt 7057@item -Wstrict-overflow=1 7058Warn about cases that are both questionable and easy to avoid. For 7059example the compiler simplifies 7060@code{x + 1 > x} to @code{1}. This level of 7061@option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels 7062are not, and must be explicitly requested. 7063 7064@item -Wstrict-overflow=2 7065Also warn about other cases where a comparison is simplified to a 7066constant. For example: @code{abs (x) >= 0}. This can only be 7067simplified when signed integer overflow is undefined, because 7068@code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than 7069zero. @option{-Wstrict-overflow} (with no level) is the same as 7070@option{-Wstrict-overflow=2}. 7071 7072@item -Wstrict-overflow=3 7073Also warn about other cases where a comparison is simplified. For 7074example: @code{x + 1 > 1} is simplified to @code{x > 0}. 7075 7076@item -Wstrict-overflow=4 7077Also warn about other simplifications not covered by the above cases. 7078For example: @code{(x * 10) / 5} is simplified to @code{x * 2}. 7079 7080@item -Wstrict-overflow=5 7081Also warn about cases where the compiler reduces the magnitude of a 7082constant involved in a comparison. For example: @code{x + 2 > y} is 7083simplified to @code{x + 1 >= y}. This is reported only at the 7084highest warning level because this simplification applies to many 7085comparisons, so this warning level gives a very large number of 7086false positives. 7087@end table 7088 7089@item -Wstring-compare 7090@opindex Wstring-compare 7091@opindex Wno-string-compare 7092Warn for calls to @code{strcmp} and @code{strncmp} whose result is 7093determined to be either zero or non-zero in tests for such equality 7094owing to the length of one argument being greater than the size of 7095the array the other argument is stored in (or the bound in the case 7096of @code{strncmp}). Such calls could be mistakes. For example, 7097the call to @code{strcmp} below is diagnosed because its result is 7098necessarily non-zero irrespective of the contents of the array @code{a}. 7099 7100@smallexample 7101extern char a[4]; 7102void f (char *d) 7103@{ 7104 strcpy (d, "string"); 7105 @dots{} 7106 if (0 == strcmp (a, d)) // cannot be true 7107 puts ("a and d are the same"); 7108@} 7109@end smallexample 7110 7111@option{-Wstring-compare} is enabled by @option{-Wextra}. 7112 7113@item -Wno-stringop-overflow 7114@item -Wstringop-overflow 7115@itemx -Wstringop-overflow=@var{type} 7116@opindex Wstringop-overflow 7117@opindex Wno-stringop-overflow 7118Warn for calls to string manipulation functions such as @code{memcpy} and 7119@code{strcpy} that are determined to overflow the destination buffer. The 7120optional argument is one greater than the type of Object Size Checking to 7121perform to determine the size of the destination. @xref{Object Size Checking}. 7122The argument is meaningful only for functions that operate on character arrays 7123but not for raw memory functions like @code{memcpy} which always make use 7124of Object Size type-0. The option also warns for calls that specify a size 7125in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes. 7126The option produces the best results with optimization enabled but can detect 7127a small subset of simple buffer overflows even without optimization in 7128calls to the GCC built-in functions like @code{__builtin_memcpy} that 7129correspond to the standard functions. In any case, the option warns about 7130just a subset of buffer overflows detected by the corresponding overflow 7131checking built-ins. For example, the option issues a warning for 7132the @code{strcpy} call below because it copies at least 5 characters 7133(the string @code{"blue"} including the terminating NUL) into the buffer 7134of size 4. 7135 7136@smallexample 7137enum Color @{ blue, purple, yellow @}; 7138const char* f (enum Color clr) 7139@{ 7140 static char buf [4]; 7141 const char *str; 7142 switch (clr) 7143 @{ 7144 case blue: str = "blue"; break; 7145 case purple: str = "purple"; break; 7146 case yellow: str = "yellow"; break; 7147 @} 7148 7149 return strcpy (buf, str); // warning here 7150@} 7151@end smallexample 7152 7153Option @option{-Wstringop-overflow=2} is enabled by default. 7154 7155@table @gcctabopt 7156@item -Wstringop-overflow 7157@itemx -Wstringop-overflow=1 7158@opindex Wstringop-overflow 7159@opindex Wno-stringop-overflow 7160The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking 7161to determine the sizes of destination objects. At this setting the option 7162does not warn for writes past the end of subobjects of larger objects accessed 7163by pointers unless the size of the largest surrounding object is known. When 7164the destination may be one of several objects it is assumed to be the largest 7165one of them. On Linux systems, when optimization is enabled at this setting 7166the option warns for the same code as when the @code{_FORTIFY_SOURCE} macro 7167is defined to a non-zero value. 7168 7169@item -Wstringop-overflow=2 7170The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking 7171to determine the sizes of destination objects. At this setting the option 7172warns about overflows when writing to members of the largest complete 7173objects whose exact size is known. However, it does not warn for excessive 7174writes to the same members of unknown objects referenced by pointers since 7175they may point to arrays containing unknown numbers of elements. This is 7176the default setting of the option. 7177 7178@item -Wstringop-overflow=3 7179The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking 7180to determine the sizes of destination objects. At this setting the option 7181warns about overflowing the smallest object or data member. This is the 7182most restrictive setting of the option that may result in warnings for safe 7183code. 7184 7185@item -Wstringop-overflow=4 7186The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking 7187to determine the sizes of destination objects. At this setting the option 7188warns about overflowing any data members, and when the destination is 7189one of several objects it uses the size of the largest of them to decide 7190whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this 7191setting of the option may result in warnings for benign code. 7192@end table 7193 7194@item -Wno-stringop-overread 7195@opindex Wstringop-overread 7196@opindex Wno-stringop-overread 7197Warn for calls to string manipulation functions such as @code{memchr}, or 7198@code{strcpy} that are determined to read past the end of the source 7199sequence. 7200 7201Option @option{-Wstringop-overread} is enabled by default. 7202 7203@item -Wno-stringop-truncation 7204@opindex Wstringop-truncation 7205@opindex Wno-stringop-truncation 7206Do not warn for calls to bounded string manipulation functions 7207such as @code{strncat}, 7208@code{strncpy}, and @code{stpncpy} that may either truncate the copied string 7209or leave the destination unchanged. 7210 7211In the following example, the call to @code{strncat} specifies a bound that 7212is less than the length of the source string. As a result, the copy of 7213the source will be truncated and so the call is diagnosed. To avoid the 7214warning use @code{bufsize - strlen (buf) - 1)} as the bound. 7215 7216@smallexample 7217void append (char *buf, size_t bufsize) 7218@{ 7219 strncat (buf, ".txt", 3); 7220@} 7221@end smallexample 7222 7223As another example, the following call to @code{strncpy} results in copying 7224to @code{d} just the characters preceding the terminating NUL, without 7225appending the NUL to the end. Assuming the result of @code{strncpy} is 7226necessarily a NUL-terminated string is a common mistake, and so the call 7227is diagnosed. To avoid the warning when the result is not expected to be 7228NUL-terminated, call @code{memcpy} instead. 7229 7230@smallexample 7231void copy (char *d, const char *s) 7232@{ 7233 strncpy (d, s, strlen (s)); 7234@} 7235@end smallexample 7236 7237In the following example, the call to @code{strncpy} specifies the size 7238of the destination buffer as the bound. If the length of the source 7239string is equal to or greater than this size the result of the copy will 7240not be NUL-terminated. Therefore, the call is also diagnosed. To avoid 7241the warning, specify @code{sizeof buf - 1} as the bound and set the last 7242element of the buffer to @code{NUL}. 7243 7244@smallexample 7245void copy (const char *s) 7246@{ 7247 char buf[80]; 7248 strncpy (buf, s, sizeof buf); 7249 @dots{} 7250@} 7251@end smallexample 7252 7253In situations where a character array is intended to store a sequence 7254of bytes with no terminating @code{NUL} such an array may be annotated 7255with attribute @code{nonstring} to avoid this warning. Such arrays, 7256however, are not suitable arguments to functions that expect 7257@code{NUL}-terminated strings. To help detect accidental misuses of 7258such arrays GCC issues warnings unless it can prove that the use is 7259safe. @xref{Common Variable Attributes}. 7260 7261@item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}cold@r{|}malloc@r{]} 7262@opindex Wsuggest-attribute= 7263@opindex Wno-suggest-attribute= 7264Warn for cases where adding an attribute may be beneficial. The 7265attributes currently supported are listed below. 7266 7267@table @gcctabopt 7268@item -Wsuggest-attribute=pure 7269@itemx -Wsuggest-attribute=const 7270@itemx -Wsuggest-attribute=noreturn 7271@itemx -Wmissing-noreturn 7272@itemx -Wsuggest-attribute=malloc 7273@opindex Wsuggest-attribute=pure 7274@opindex Wno-suggest-attribute=pure 7275@opindex Wsuggest-attribute=const 7276@opindex Wno-suggest-attribute=const 7277@opindex Wsuggest-attribute=noreturn 7278@opindex Wno-suggest-attribute=noreturn 7279@opindex Wmissing-noreturn 7280@opindex Wno-missing-noreturn 7281@opindex Wsuggest-attribute=malloc 7282@opindex Wno-suggest-attribute=malloc 7283 7284Warn about functions that might be candidates for attributes 7285@code{pure}, @code{const} or @code{noreturn} or @code{malloc}. The compiler 7286only warns for functions visible in other compilation units or (in the case of 7287@code{pure} and @code{const}) if it cannot prove that the function returns 7288normally. A function returns normally if it doesn't contain an infinite loop or 7289return abnormally by throwing, calling @code{abort} or trapping. This analysis 7290requires option @option{-fipa-pure-const}, which is enabled by default at 7291@option{-O} and higher. Higher optimization levels improve the accuracy 7292of the analysis. 7293 7294@item -Wsuggest-attribute=format 7295@itemx -Wmissing-format-attribute 7296@opindex Wsuggest-attribute=format 7297@opindex Wmissing-format-attribute 7298@opindex Wno-suggest-attribute=format 7299@opindex Wno-missing-format-attribute 7300@opindex Wformat 7301@opindex Wno-format 7302 7303Warn about function pointers that might be candidates for @code{format} 7304attributes. Note these are only possible candidates, not absolute ones. 7305GCC guesses that function pointers with @code{format} attributes that 7306are used in assignment, initialization, parameter passing or return 7307statements should have a corresponding @code{format} attribute in the 7308resulting type. I.e.@: the left-hand side of the assignment or 7309initialization, the type of the parameter variable, or the return type 7310of the containing function respectively should also have a @code{format} 7311attribute to avoid the warning. 7312 7313GCC also warns about function definitions that might be 7314candidates for @code{format} attributes. Again, these are only 7315possible candidates. GCC guesses that @code{format} attributes 7316might be appropriate for any function that calls a function like 7317@code{vprintf} or @code{vscanf}, but this might not always be the 7318case, and some functions for which @code{format} attributes are 7319appropriate may not be detected. 7320 7321@item -Wsuggest-attribute=cold 7322@opindex Wsuggest-attribute=cold 7323@opindex Wno-suggest-attribute=cold 7324 7325Warn about functions that might be candidates for @code{cold} attribute. This 7326is based on static detection and generally only warns about functions which 7327always leads to a call to another @code{cold} function such as wrappers of 7328C++ @code{throw} or fatal error reporting functions leading to @code{abort}. 7329@end table 7330 7331@item -Walloc-zero 7332@opindex Wno-alloc-zero 7333@opindex Walloc-zero 7334Warn about calls to allocation functions decorated with attribute 7335@code{alloc_size} that specify zero bytes, including those to the built-in 7336forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc}, 7337@code{malloc}, and @code{realloc}. Because the behavior of these functions 7338when called with a zero size differs among implementations (and in the case 7339of @code{realloc} has been deprecated) relying on it may result in subtle 7340portability bugs and should be avoided. 7341 7342@item -Walloc-size-larger-than=@var{byte-size} 7343@opindex Walloc-size-larger-than= 7344@opindex Wno-alloc-size-larger-than 7345Warn about calls to functions decorated with attribute @code{alloc_size} 7346that attempt to allocate objects larger than the specified number of bytes, 7347or where the result of the size computation in an integer type with infinite 7348precision would exceed the value of @samp{PTRDIFF_MAX} on the target. 7349@option{-Walloc-size-larger-than=}@samp{PTRDIFF_MAX} is enabled by default. 7350Warnings controlled by the option can be disabled either by specifying 7351@var{byte-size} of @samp{SIZE_MAX} or more or by 7352@option{-Wno-alloc-size-larger-than}. 7353@xref{Function Attributes}. 7354 7355@item -Wno-alloc-size-larger-than 7356@opindex Wno-alloc-size-larger-than 7357Disable @option{-Walloc-size-larger-than=} warnings. The option is 7358equivalent to @option{-Walloc-size-larger-than=}@samp{SIZE_MAX} or 7359larger. 7360 7361@item -Walloca 7362@opindex Wno-alloca 7363@opindex Walloca 7364This option warns on all uses of @code{alloca} in the source. 7365 7366@item -Walloca-larger-than=@var{byte-size} 7367@opindex Walloca-larger-than= 7368@opindex Wno-alloca-larger-than 7369This option warns on calls to @code{alloca} with an integer argument whose 7370value is either zero, or that is not bounded by a controlling predicate 7371that limits its value to at most @var{byte-size}. It also warns for calls 7372to @code{alloca} where the bound value is unknown. Arguments of non-integer 7373types are considered unbounded even if they appear to be constrained to 7374the expected range. 7375 7376For example, a bounded case of @code{alloca} could be: 7377 7378@smallexample 7379void func (size_t n) 7380@{ 7381 void *p; 7382 if (n <= 1000) 7383 p = alloca (n); 7384 else 7385 p = malloc (n); 7386 f (p); 7387@} 7388@end smallexample 7389 7390In the above example, passing @code{-Walloca-larger-than=1000} would not 7391issue a warning because the call to @code{alloca} is known to be at most 73921000 bytes. However, if @code{-Walloca-larger-than=500} were passed, 7393the compiler would emit a warning. 7394 7395Unbounded uses, on the other hand, are uses of @code{alloca} with no 7396controlling predicate constraining its integer argument. For example: 7397 7398@smallexample 7399void func () 7400@{ 7401 void *p = alloca (n); 7402 f (p); 7403@} 7404@end smallexample 7405 7406If @code{-Walloca-larger-than=500} were passed, the above would trigger 7407a warning, but this time because of the lack of bounds checking. 7408 7409Note, that even seemingly correct code involving signed integers could 7410cause a warning: 7411 7412@smallexample 7413void func (signed int n) 7414@{ 7415 if (n < 500) 7416 @{ 7417 p = alloca (n); 7418 f (p); 7419 @} 7420@} 7421@end smallexample 7422 7423In the above example, @var{n} could be negative, causing a larger than 7424expected argument to be implicitly cast into the @code{alloca} call. 7425 7426This option also warns when @code{alloca} is used in a loop. 7427 7428@option{-Walloca-larger-than=}@samp{PTRDIFF_MAX} is enabled by default 7429but is usually only effective when @option{-ftree-vrp} is active (default 7430for @option{-O2} and above). 7431 7432See also @option{-Wvla-larger-than=}@samp{byte-size}. 7433 7434@item -Wno-alloca-larger-than 7435@opindex Wno-alloca-larger-than 7436Disable @option{-Walloca-larger-than=} warnings. The option is 7437equivalent to @option{-Walloca-larger-than=}@samp{SIZE_MAX} or larger. 7438 7439@item -Warith-conversion 7440@opindex Warith-conversion 7441@opindex Wno-arith-conversion 7442Do warn about implicit conversions from arithmetic operations even 7443when conversion of the operands to the same type cannot change their 7444values. This affects warnings from @option{-Wconversion}, 7445@option{-Wfloat-conversion}, and @option{-Wsign-conversion}. 7446 7447@smallexample 7448@group 7449void f (char c, int i) 7450@{ 7451 c = c + i; // warns with @option{-Wconversion} 7452 c = c + 1; // only warns with @option{-Warith-conversion} 7453@} 7454@end group 7455@end smallexample 7456 7457@item -Warray-bounds 7458@itemx -Warray-bounds=@var{n} 7459@opindex Wno-array-bounds 7460@opindex Warray-bounds 7461This option is only active when @option{-ftree-vrp} is active 7462(default for @option{-O2} and above). It warns about subscripts to arrays 7463that are always out of bounds. This warning is enabled by @option{-Wall}. 7464 7465@table @gcctabopt 7466@item -Warray-bounds=1 7467This is the warning level of @option{-Warray-bounds} and is enabled 7468by @option{-Wall}; higher levels are not, and must be explicitly requested. 7469 7470@item -Warray-bounds=2 7471This warning level also warns about out of bounds access for 7472arrays at the end of a struct and for arrays accessed through 7473pointers. This warning level may give a larger number of 7474false positives and is deactivated by default. 7475@end table 7476 7477@item -Warray-parameter 7478@itemx -Warray-parameter=@var{n} 7479@opindex Wno-array-parameter 7480Warn about redeclarations of functions involving arguments of array or 7481pointer types of inconsistent kinds or forms, and enable the detection 7482of out-of-bounds accesses to such parameters by warnings such as 7483@option{-Warray-bounds}. 7484 7485If the first function declaration uses the array form the bound specified 7486in the array is assumed to be the minimum number of elements expected to 7487be provided in calls to the function and the maximum number of elements 7488accessed by it. Failing to provide arguments of sufficient size or accessing 7489more than the maximum number of elements may be diagnosed by warnings such 7490as @option{-Warray-bounds}. At level 1 the warning diagnoses inconsistencies 7491involving array parameters declared using the @code{T[static N]} form. 7492 7493For example, the warning triggers for the following redeclarations because 7494the first one allows an array of any size to be passed to @code{f} while 7495the second one with the keyword @code{static} specifies that the array 7496argument must have at least four elements. 7497 7498@smallexample 7499void f (int[static 4]); 7500void f (int[]); // warning (inconsistent array form) 7501 7502void g (void) 7503@{ 7504 int *p = (int *)malloc (4); 7505 f (p); // warning (array too small) 7506 @dots{} 7507@} 7508@end smallexample 7509 7510At level 2 the warning also triggers for redeclarations involving any other 7511inconsistency in array or pointer argument forms denoting array sizes. 7512Pointers and arrays of unspecified bound are considered equivalent and do 7513not trigger a warning. 7514 7515@smallexample 7516void g (int*); 7517void g (int[]); // no warning 7518void g (int[8]); // warning (inconsistent array bound) 7519@end smallexample 7520 7521@option{-Warray-parameter=2} is included in @option{-Wall}. The 7522@option{-Wvla-parameter} option triggers warnings for similar inconsistencies 7523involving Variable Length Array arguments. 7524 7525@item -Wattribute-alias=@var{n} 7526@itemx -Wno-attribute-alias 7527@opindex Wattribute-alias 7528@opindex Wno-attribute-alias 7529Warn about declarations using the @code{alias} and similar attributes whose 7530target is incompatible with the type of the alias. 7531@xref{Function Attributes,,Declaring Attributes of Functions}. 7532 7533@table @gcctabopt 7534@item -Wattribute-alias=1 7535The default warning level of the @option{-Wattribute-alias} option diagnoses 7536incompatibilities between the type of the alias declaration and that of its 7537target. Such incompatibilities are typically indicative of bugs. 7538 7539@item -Wattribute-alias=2 7540 7541At this level @option{-Wattribute-alias} also diagnoses cases where 7542the attributes of the alias declaration are more restrictive than the 7543attributes applied to its target. These mismatches can potentially 7544result in incorrect code generation. In other cases they may be 7545benign and could be resolved simply by adding the missing attribute to 7546the target. For comparison, see the @option{-Wmissing-attributes} 7547option, which controls diagnostics when the alias declaration is less 7548restrictive than the target, rather than more restrictive. 7549 7550Attributes considered include @code{alloc_align}, @code{alloc_size}, 7551@code{cold}, @code{const}, @code{hot}, @code{leaf}, @code{malloc}, 7552@code{nonnull}, @code{noreturn}, @code{nothrow}, @code{pure}, 7553@code{returns_nonnull}, and @code{returns_twice}. 7554@end table 7555 7556@option{-Wattribute-alias} is equivalent to @option{-Wattribute-alias=1}. 7557This is the default. You can disable these warnings with either 7558@option{-Wno-attribute-alias} or @option{-Wattribute-alias=0}. 7559 7560@item -Wbool-compare 7561@opindex Wno-bool-compare 7562@opindex Wbool-compare 7563Warn about boolean expression compared with an integer value different from 7564@code{true}/@code{false}. For instance, the following comparison is 7565always false: 7566@smallexample 7567int n = 5; 7568@dots{} 7569if ((n > 1) == 2) @{ @dots{} @} 7570@end smallexample 7571This warning is enabled by @option{-Wall}. 7572 7573@item -Wbool-operation 7574@opindex Wno-bool-operation 7575@opindex Wbool-operation 7576Warn about suspicious operations on expressions of a boolean type. For 7577instance, bitwise negation of a boolean is very likely a bug in the program. 7578For C, this warning also warns about incrementing or decrementing a boolean, 7579which rarely makes sense. (In C++, decrementing a boolean is always invalid. 7580Incrementing a boolean is invalid in C++17, and deprecated otherwise.) 7581 7582This warning is enabled by @option{-Wall}. 7583 7584@item -Wduplicated-branches 7585@opindex Wno-duplicated-branches 7586@opindex Wduplicated-branches 7587Warn when an if-else has identical branches. This warning detects cases like 7588@smallexample 7589if (p != NULL) 7590 return 0; 7591else 7592 return 0; 7593@end smallexample 7594It doesn't warn when both branches contain just a null statement. This warning 7595also warn for conditional operators: 7596@smallexample 7597 int i = x ? *p : *p; 7598@end smallexample 7599 7600@item -Wduplicated-cond 7601@opindex Wno-duplicated-cond 7602@opindex Wduplicated-cond 7603Warn about duplicated conditions in an if-else-if chain. For instance, 7604warn for the following code: 7605@smallexample 7606if (p->q != NULL) @{ @dots{} @} 7607else if (p->q != NULL) @{ @dots{} @} 7608@end smallexample 7609 7610@item -Wframe-address 7611@opindex Wno-frame-address 7612@opindex Wframe-address 7613Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address} 7614is called with an argument greater than 0. Such calls may return indeterminate 7615values or crash the program. The warning is included in @option{-Wall}. 7616 7617@item -Wno-discarded-qualifiers @r{(C and Objective-C only)} 7618@opindex Wno-discarded-qualifiers 7619@opindex Wdiscarded-qualifiers 7620Do not warn if type qualifiers on pointers are being discarded. 7621Typically, the compiler warns if a @code{const char *} variable is 7622passed to a function that takes a @code{char *} parameter. This option 7623can be used to suppress such a warning. 7624 7625@item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)} 7626@opindex Wno-discarded-array-qualifiers 7627@opindex Wdiscarded-array-qualifiers 7628Do not warn if type qualifiers on arrays which are pointer targets 7629are being discarded. Typically, the compiler warns if a 7630@code{const int (*)[]} variable is passed to a function that 7631takes a @code{int (*)[]} parameter. This option can be used to 7632suppress such a warning. 7633 7634@item -Wno-incompatible-pointer-types @r{(C and Objective-C only)} 7635@opindex Wno-incompatible-pointer-types 7636@opindex Wincompatible-pointer-types 7637Do not warn when there is a conversion between pointers that have incompatible 7638types. This warning is for cases not covered by @option{-Wno-pointer-sign}, 7639which warns for pointer argument passing or assignment with different 7640signedness. 7641 7642@item -Wno-int-conversion @r{(C and Objective-C only)} 7643@opindex Wno-int-conversion 7644@opindex Wint-conversion 7645Do not warn about incompatible integer to pointer and pointer to integer 7646conversions. This warning is about implicit conversions; for explicit 7647conversions the warnings @option{-Wno-int-to-pointer-cast} and 7648@option{-Wno-pointer-to-int-cast} may be used. 7649 7650@item -Wzero-length-bounds 7651@opindex Wzero-length-bounds 7652@opindex Wzero-length-bounds 7653Warn about accesses to elements of zero-length array members that might 7654overlap other members of the same object. Declaring interior zero-length 7655arrays is discouraged because accesses to them are undefined. See 7656@xref{Zero Length}. 7657 7658For example, the first two stores in function @code{bad} are diagnosed 7659because the array elements overlap the subsequent members @code{b} and 7660@code{c}. The third store is diagnosed by @option{-Warray-bounds} 7661because it is beyond the bounds of the enclosing object. 7662 7663@smallexample 7664struct X @{ int a[0]; int b, c; @}; 7665struct X x; 7666 7667void bad (void) 7668@{ 7669 x.a[0] = 0; // -Wzero-length-bounds 7670 x.a[1] = 1; // -Wzero-length-bounds 7671 x.a[2] = 2; // -Warray-bounds 7672@} 7673@end smallexample 7674 7675Option @option{-Wzero-length-bounds} is enabled by @option{-Warray-bounds}. 7676 7677@item -Wno-div-by-zero 7678@opindex Wno-div-by-zero 7679@opindex Wdiv-by-zero 7680Do not warn about compile-time integer division by zero. Floating-point 7681division by zero is not warned about, as it can be a legitimate way of 7682obtaining infinities and NaNs. 7683 7684@item -Wsystem-headers 7685@opindex Wsystem-headers 7686@opindex Wno-system-headers 7687@cindex warnings from system headers 7688@cindex system headers, warnings from 7689Print warning messages for constructs found in system header files. 7690Warnings from system headers are normally suppressed, on the assumption 7691that they usually do not indicate real problems and would only make the 7692compiler output harder to read. Using this command-line option tells 7693GCC to emit warnings from system headers as if they occurred in user 7694code. However, note that using @option{-Wall} in conjunction with this 7695option does @emph{not} warn about unknown pragmas in system 7696headers---for that, @option{-Wunknown-pragmas} must also be used. 7697 7698@item -Wtautological-compare 7699@opindex Wtautological-compare 7700@opindex Wno-tautological-compare 7701Warn if a self-comparison always evaluates to true or false. This 7702warning detects various mistakes such as: 7703@smallexample 7704int i = 1; 7705@dots{} 7706if (i > i) @{ @dots{} @} 7707@end smallexample 7708 7709This warning also warns about bitwise comparisons that always evaluate 7710to true or false, for instance: 7711@smallexample 7712if ((a & 16) == 10) @{ @dots{} @} 7713@end smallexample 7714will always be false. 7715 7716This warning is enabled by @option{-Wall}. 7717 7718@item -Wtrampolines 7719@opindex Wtrampolines 7720@opindex Wno-trampolines 7721Warn about trampolines generated for pointers to nested functions. 7722A trampoline is a small piece of data or code that is created at run 7723time on the stack when the address of a nested function is taken, and is 7724used to call the nested function indirectly. For some targets, it is 7725made up of data only and thus requires no special treatment. But, for 7726most targets, it is made up of code and thus requires the stack to be 7727made executable in order for the program to work properly. 7728 7729@item -Wfloat-equal 7730@opindex Wfloat-equal 7731@opindex Wno-float-equal 7732Warn if floating-point values are used in equality comparisons. 7733 7734The idea behind this is that sometimes it is convenient (for the 7735programmer) to consider floating-point values as approximations to 7736infinitely precise real numbers. If you are doing this, then you need 7737to compute (by analyzing the code, or in some other way) the maximum or 7738likely maximum error that the computation introduces, and allow for it 7739when performing comparisons (and when producing output, but that's a 7740different problem). In particular, instead of testing for equality, you 7741should check to see whether the two values have ranges that overlap; and 7742this is done with the relational operators, so equality comparisons are 7743probably mistaken. 7744 7745@item -Wtraditional @r{(C and Objective-C only)} 7746@opindex Wtraditional 7747@opindex Wno-traditional 7748Warn about certain constructs that behave differently in traditional and 7749ISO C@. Also warn about ISO C constructs that have no traditional C 7750equivalent, and/or problematic constructs that should be avoided. 7751 7752@itemize @bullet 7753@item 7754Macro parameters that appear within string literals in the macro body. 7755In traditional C macro replacement takes place within string literals, 7756but in ISO C it does not. 7757 7758@item 7759In traditional C, some preprocessor directives did not exist. 7760Traditional preprocessors only considered a line to be a directive 7761if the @samp{#} appeared in column 1 on the line. Therefore 7762@option{-Wtraditional} warns about directives that traditional C 7763understands but ignores because the @samp{#} does not appear as the 7764first character on the line. It also suggests you hide directives like 7765@code{#pragma} not understood by traditional C by indenting them. Some 7766traditional implementations do not recognize @code{#elif}, so this option 7767suggests avoiding it altogether. 7768 7769@item 7770A function-like macro that appears without arguments. 7771 7772@item 7773The unary plus operator. 7774 7775@item 7776The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point 7777constant suffixes. (Traditional C does support the @samp{L} suffix on integer 7778constants.) Note, these suffixes appear in macros defined in the system 7779headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}. 7780Use of these macros in user code might normally lead to spurious 7781warnings, however GCC's integrated preprocessor has enough context to 7782avoid warning in these cases. 7783 7784@item 7785A function declared external in one block and then used after the end of 7786the block. 7787 7788@item 7789A @code{switch} statement has an operand of type @code{long}. 7790 7791@item 7792A non-@code{static} function declaration follows a @code{static} one. 7793This construct is not accepted by some traditional C compilers. 7794 7795@item 7796The ISO type of an integer constant has a different width or 7797signedness from its traditional type. This warning is only issued if 7798the base of the constant is ten. I.e.@: hexadecimal or octal values, which 7799typically represent bit patterns, are not warned about. 7800 7801@item 7802Usage of ISO string concatenation is detected. 7803 7804@item 7805Initialization of automatic aggregates. 7806 7807@item 7808Identifier conflicts with labels. Traditional C lacks a separate 7809namespace for labels. 7810 7811@item 7812Initialization of unions. If the initializer is zero, the warning is 7813omitted. This is done under the assumption that the zero initializer in 7814user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing 7815initializer warnings and relies on default initialization to zero in the 7816traditional C case. 7817 7818@item 7819Conversions by prototypes between fixed/floating-point values and vice 7820versa. The absence of these prototypes when compiling with traditional 7821C causes serious problems. This is a subset of the possible 7822conversion warnings; for the full set use @option{-Wtraditional-conversion}. 7823 7824@item 7825Use of ISO C style function definitions. This warning intentionally is 7826@emph{not} issued for prototype declarations or variadic functions 7827because these ISO C features appear in your code when using 7828libiberty's traditional C compatibility macros, @code{PARAMS} and 7829@code{VPARAMS}. This warning is also bypassed for nested functions 7830because that feature is already a GCC extension and thus not relevant to 7831traditional C compatibility. 7832@end itemize 7833 7834@item -Wtraditional-conversion @r{(C and Objective-C only)} 7835@opindex Wtraditional-conversion 7836@opindex Wno-traditional-conversion 7837Warn if a prototype causes a type conversion that is different from what 7838would happen to the same argument in the absence of a prototype. This 7839includes conversions of fixed point to floating and vice versa, and 7840conversions changing the width or signedness of a fixed-point argument 7841except when the same as the default promotion. 7842 7843@item -Wdeclaration-after-statement @r{(C and Objective-C only)} 7844@opindex Wdeclaration-after-statement 7845@opindex Wno-declaration-after-statement 7846Warn when a declaration is found after a statement in a block. This 7847construct, known from C++, was introduced with ISO C99 and is by default 7848allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Labels and Declarations}. 7849 7850@item -Wshadow 7851@opindex Wshadow 7852@opindex Wno-shadow 7853Warn whenever a local variable or type declaration shadows another 7854variable, parameter, type, class member (in C++), or instance variable 7855(in Objective-C) or whenever a built-in function is shadowed. Note 7856that in C++, the compiler warns if a local variable shadows an 7857explicit typedef, but not if it shadows a struct/class/enum. 7858If this warning is enabled, it includes also all instances of 7859local shadowing. This means that @option{-Wno-shadow=local} 7860and @option{-Wno-shadow=compatible-local} are ignored when 7861@option{-Wshadow} is used. 7862Same as @option{-Wshadow=global}. 7863 7864@item -Wno-shadow-ivar @r{(Objective-C only)} 7865@opindex Wno-shadow-ivar 7866@opindex Wshadow-ivar 7867Do not warn whenever a local variable shadows an instance variable in an 7868Objective-C method. 7869 7870@item -Wshadow=global 7871@opindex Wshadow=global 7872Warn for any shadowing. 7873Same as @option{-Wshadow}. 7874 7875@item -Wshadow=local 7876@opindex Wshadow=local 7877Warn when a local variable shadows another local variable or parameter. 7878 7879@item -Wshadow=compatible-local 7880@opindex Wshadow=compatible-local 7881Warn when a local variable shadows another local variable or parameter 7882whose type is compatible with that of the shadowing variable. In C++, 7883type compatibility here means the type of the shadowing variable can be 7884converted to that of the shadowed variable. The creation of this flag 7885(in addition to @option{-Wshadow=local}) is based on the idea that when 7886a local variable shadows another one of incompatible type, it is most 7887likely intentional, not a bug or typo, as shown in the following example: 7888 7889@smallexample 7890@group 7891for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i) 7892@{ 7893 for (int i = 0; i < N; ++i) 7894 @{ 7895 ... 7896 @} 7897 ... 7898@} 7899@end group 7900@end smallexample 7901 7902Since the two variable @code{i} in the example above have incompatible types, 7903enabling only @option{-Wshadow=compatible-local} does not emit a warning. 7904Because their types are incompatible, if a programmer accidentally uses one 7905in place of the other, type checking is expected to catch that and emit an 7906error or warning. Use of this flag instead of @option{-Wshadow=local} can 7907possibly reduce the number of warnings triggered by intentional shadowing. 7908Note that this also means that shadowing @code{const char *i} by 7909@code{char *i} does not emit a warning. 7910 7911This warning is also enabled by @option{-Wshadow=local}. 7912 7913@item -Wlarger-than=@var{byte-size} 7914@opindex Wlarger-than= 7915@opindex Wlarger-than-@var{byte-size} 7916Warn whenever an object is defined whose size exceeds @var{byte-size}. 7917@option{-Wlarger-than=}@samp{PTRDIFF_MAX} is enabled by default. 7918Warnings controlled by the option can be disabled either by specifying 7919@var{byte-size} of @samp{SIZE_MAX} or more or by @option{-Wno-larger-than}. 7920 7921Also warn for calls to bounded functions such as @code{memchr} or 7922@code{strnlen} that specify a bound greater than the largest possible 7923object, which is @samp{PTRDIFF_MAX} bytes by default. These warnings 7924can only be disabled by @option{-Wno-larger-than}. 7925 7926@item -Wno-larger-than 7927@opindex Wno-larger-than 7928Disable @option{-Wlarger-than=} warnings. The option is equivalent 7929to @option{-Wlarger-than=}@samp{SIZE_MAX} or larger. 7930 7931@item -Wframe-larger-than=@var{byte-size} 7932@opindex Wframe-larger-than= 7933@opindex Wno-frame-larger-than 7934Warn if the size of a function frame exceeds @var{byte-size}. 7935The computation done to determine the stack frame size is approximate 7936and not conservative. 7937The actual requirements may be somewhat greater than @var{byte-size} 7938even if you do not get a warning. In addition, any space allocated 7939via @code{alloca}, variable-length arrays, or related constructs 7940is not included by the compiler when determining 7941whether or not to issue a warning. 7942@option{-Wframe-larger-than=}@samp{PTRDIFF_MAX} is enabled by default. 7943Warnings controlled by the option can be disabled either by specifying 7944@var{byte-size} of @samp{SIZE_MAX} or more or by 7945@option{-Wno-frame-larger-than}. 7946 7947@item -Wno-frame-larger-than 7948@opindex Wno-frame-larger-than 7949Disable @option{-Wframe-larger-than=} warnings. The option is equivalent 7950to @option{-Wframe-larger-than=}@samp{SIZE_MAX} or larger. 7951 7952@item -Wno-free-nonheap-object 7953@opindex Wno-free-nonheap-object 7954@opindex Wfree-nonheap-object 7955Warn when attempting to deallocate an object that was either not allocated 7956on the heap, or by using a pointer that was not returned from a prior call 7957to the corresponding allocation function. For example, because the call 7958to @code{stpcpy} returns a pointer to the terminating nul character and 7959not to the begginning of the object, the call to @code{free} below is 7960diagnosed. 7961 7962@smallexample 7963void f (char *p) 7964@{ 7965 p = stpcpy (p, "abc"); 7966 // ... 7967 free (p); // warning 7968@} 7969@end smallexample 7970 7971@option{-Wfree-nonheap-object} is enabled by default. 7972 7973@item -Wstack-usage=@var{byte-size} 7974@opindex Wstack-usage 7975@opindex Wno-stack-usage 7976Warn if the stack usage of a function might exceed @var{byte-size}. 7977The computation done to determine the stack usage is conservative. 7978Any space allocated via @code{alloca}, variable-length arrays, or related 7979constructs is included by the compiler when determining whether or not to 7980issue a warning. 7981 7982The message is in keeping with the output of @option{-fstack-usage}. 7983 7984@itemize 7985@item 7986If the stack usage is fully static but exceeds the specified amount, it's: 7987 7988@smallexample 7989 warning: stack usage is 1120 bytes 7990@end smallexample 7991@item 7992If the stack usage is (partly) dynamic but bounded, it's: 7993 7994@smallexample 7995 warning: stack usage might be 1648 bytes 7996@end smallexample 7997@item 7998If the stack usage is (partly) dynamic and not bounded, it's: 7999 8000@smallexample 8001 warning: stack usage might be unbounded 8002@end smallexample 8003@end itemize 8004 8005@option{-Wstack-usage=}@samp{PTRDIFF_MAX} is enabled by default. 8006Warnings controlled by the option can be disabled either by specifying 8007@var{byte-size} of @samp{SIZE_MAX} or more or by 8008@option{-Wno-stack-usage}. 8009 8010@item -Wno-stack-usage 8011@opindex Wno-stack-usage 8012Disable @option{-Wstack-usage=} warnings. The option is equivalent 8013to @option{-Wstack-usage=}@samp{SIZE_MAX} or larger. 8014 8015@item -Wunsafe-loop-optimizations 8016@opindex Wunsafe-loop-optimizations 8017@opindex Wno-unsafe-loop-optimizations 8018Warn if the loop cannot be optimized because the compiler cannot 8019assume anything on the bounds of the loop indices. With 8020@option{-funsafe-loop-optimizations} warn if the compiler makes 8021such assumptions. 8022 8023@item -Wno-pedantic-ms-format @r{(MinGW targets only)} 8024@opindex Wno-pedantic-ms-format 8025@opindex Wpedantic-ms-format 8026When used in combination with @option{-Wformat} 8027and @option{-pedantic} without GNU extensions, this option 8028disables the warnings about non-ISO @code{printf} / @code{scanf} format 8029width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets, 8030which depend on the MS runtime. 8031 8032@item -Wpointer-arith 8033@opindex Wpointer-arith 8034@opindex Wno-pointer-arith 8035Warn about anything that depends on the ``size of'' a function type or 8036of @code{void}. GNU C assigns these types a size of 1, for 8037convenience in calculations with @code{void *} pointers and pointers 8038to functions. In C++, warn also when an arithmetic operation involves 8039@code{NULL}. This warning is also enabled by @option{-Wpedantic}. 8040 8041@item -Wno-pointer-compare 8042@opindex Wpointer-compare 8043@opindex Wno-pointer-compare 8044Do not warn if a pointer is compared with a zero character constant. 8045This usually 8046means that the pointer was meant to be dereferenced. For example: 8047 8048@smallexample 8049const char *p = foo (); 8050if (p == '\0') 8051 return 42; 8052@end smallexample 8053 8054Note that the code above is invalid in C++11. 8055 8056This warning is enabled by default. 8057 8058@item -Wtsan 8059@opindex Wtsan 8060@opindex Wno-tsan 8061Warn about unsupported features in ThreadSanitizer. 8062 8063ThreadSanitizer does not support @code{std::atomic_thread_fence} and 8064can report false positives. 8065 8066This warning is enabled by default. 8067 8068@item -Wtype-limits 8069@opindex Wtype-limits 8070@opindex Wno-type-limits 8071Warn if a comparison is always true or always false due to the limited 8072range of the data type, but do not warn for constant expressions. For 8073example, warn if an unsigned variable is compared against zero with 8074@code{<} or @code{>=}. This warning is also enabled by 8075@option{-Wextra}. 8076 8077@item -Wabsolute-value @r{(C and Objective-C only)} 8078@opindex Wabsolute-value 8079@opindex Wno-absolute-value 8080Warn for calls to standard functions that compute the absolute value 8081of an argument when a more appropriate standard function is available. 8082For example, calling @code{abs(3.14)} triggers the warning because the 8083appropriate function to call to compute the absolute value of a double 8084argument is @code{fabs}. The option also triggers warnings when the 8085argument in a call to such a function has an unsigned type. This 8086warning can be suppressed with an explicit type cast and it is also 8087enabled by @option{-Wextra}. 8088 8089@include cppwarnopts.texi 8090 8091@item -Wbad-function-cast @r{(C and Objective-C only)} 8092@opindex Wbad-function-cast 8093@opindex Wno-bad-function-cast 8094Warn when a function call is cast to a non-matching type. 8095For example, warn if a call to a function returning an integer type 8096is cast to a pointer type. 8097 8098@item -Wc90-c99-compat @r{(C and Objective-C only)} 8099@opindex Wc90-c99-compat 8100@opindex Wno-c90-c99-compat 8101Warn about features not present in ISO C90, but present in ISO C99. 8102For instance, warn about use of variable length arrays, @code{long long} 8103type, @code{bool} type, compound literals, designated initializers, and so 8104on. This option is independent of the standards mode. Warnings are disabled 8105in the expression that follows @code{__extension__}. 8106 8107@item -Wc99-c11-compat @r{(C and Objective-C only)} 8108@opindex Wc99-c11-compat 8109@opindex Wno-c99-c11-compat 8110Warn about features not present in ISO C99, but present in ISO C11. 8111For instance, warn about use of anonymous structures and unions, 8112@code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier, 8113@code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword, 8114and so on. This option is independent of the standards mode. Warnings are 8115disabled in the expression that follows @code{__extension__}. 8116 8117@item -Wc11-c2x-compat @r{(C and Objective-C only)} 8118@opindex Wc11-c2x-compat 8119@opindex Wno-c11-c2x-compat 8120Warn about features not present in ISO C11, but present in ISO C2X. 8121For instance, warn about omitting the string in @code{_Static_assert}, 8122use of @samp{[[]]} syntax for attributes, use of decimal 8123floating-point types, and so on. This option is independent of the 8124standards mode. Warnings are disabled in the expression that follows 8125@code{__extension__}. 8126 8127@item -Wc++-compat @r{(C and Objective-C only)} 8128@opindex Wc++-compat 8129@opindex Wno-c++-compat 8130Warn about ISO C constructs that are outside of the common subset of 8131ISO C and ISO C++, e.g.@: request for implicit conversion from 8132@code{void *} to a pointer to non-@code{void} type. 8133 8134@item -Wc++11-compat @r{(C++ and Objective-C++ only)} 8135@opindex Wc++11-compat 8136@opindex Wno-c++11-compat 8137Warn about C++ constructs whose meaning differs between ISO C++ 1998 8138and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords 8139in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is 8140enabled by @option{-Wall}. 8141 8142@item -Wc++14-compat @r{(C++ and Objective-C++ only)} 8143@opindex Wc++14-compat 8144@opindex Wno-c++14-compat 8145Warn about C++ constructs whose meaning differs between ISO C++ 2011 8146and ISO C++ 2014. This warning is enabled by @option{-Wall}. 8147 8148@item -Wc++17-compat @r{(C++ and Objective-C++ only)} 8149@opindex Wc++17-compat 8150@opindex Wno-c++17-compat 8151Warn about C++ constructs whose meaning differs between ISO C++ 2014 8152and ISO C++ 2017. This warning is enabled by @option{-Wall}. 8153 8154@item -Wc++20-compat @r{(C++ and Objective-C++ only)} 8155@opindex Wc++20-compat 8156@opindex Wno-c++20-compat 8157Warn about C++ constructs whose meaning differs between ISO C++ 2017 8158and ISO C++ 2020. This warning is enabled by @option{-Wall}. 8159 8160@item -Wcast-qual 8161@opindex Wcast-qual 8162@opindex Wno-cast-qual 8163Warn whenever a pointer is cast so as to remove a type qualifier from 8164the target type. For example, warn if a @code{const char *} is cast 8165to an ordinary @code{char *}. 8166 8167Also warn when making a cast that introduces a type qualifier in an 8168unsafe way. For example, casting @code{char **} to @code{const char **} 8169is unsafe, as in this example: 8170 8171@smallexample 8172 /* p is char ** value. */ 8173 const char **q = (const char **) p; 8174 /* Assignment of readonly string to const char * is OK. */ 8175 *q = "string"; 8176 /* Now char** pointer points to read-only memory. */ 8177 **p = 'b'; 8178@end smallexample 8179 8180@item -Wcast-align 8181@opindex Wcast-align 8182@opindex Wno-cast-align 8183Warn whenever a pointer is cast such that the required alignment of the 8184target is increased. For example, warn if a @code{char *} is cast to 8185an @code{int *} on machines where integers can only be accessed at 8186two- or four-byte boundaries. 8187 8188@item -Wcast-align=strict 8189@opindex Wcast-align=strict 8190Warn whenever a pointer is cast such that the required alignment of the 8191target is increased. For example, warn if a @code{char *} is cast to 8192an @code{int *} regardless of the target machine. 8193 8194@item -Wcast-function-type 8195@opindex Wcast-function-type 8196@opindex Wno-cast-function-type 8197Warn when a function pointer is cast to an incompatible function pointer. 8198In a cast involving function types with a variable argument list only 8199the types of initial arguments that are provided are considered. 8200Any parameter of pointer-type matches any other pointer-type. Any benign 8201differences in integral types are ignored, like @code{int} vs.@: @code{long} 8202on ILP32 targets. Likewise type qualifiers are ignored. The function 8203type @code{void (*) (void)} is special and matches everything, which can 8204be used to suppress this warning. 8205In a cast involving pointer to member types this warning warns whenever 8206the type cast is changing the pointer to member type. 8207This warning is enabled by @option{-Wextra}. 8208 8209@item -Wwrite-strings 8210@opindex Wwrite-strings 8211@opindex Wno-write-strings 8212When compiling C, give string constants the type @code{const 8213char[@var{length}]} so that copying the address of one into a 8214non-@code{const} @code{char *} pointer produces a warning. These 8215warnings help you find at compile time code that can try to write 8216into a string constant, but only if you have been very careful about 8217using @code{const} in declarations and prototypes. Otherwise, it is 8218just a nuisance. This is why we did not make @option{-Wall} request 8219these warnings. 8220 8221When compiling C++, warn about the deprecated conversion from string 8222literals to @code{char *}. This warning is enabled by default for C++ 8223programs. 8224 8225@item -Wclobbered 8226@opindex Wclobbered 8227@opindex Wno-clobbered 8228Warn for variables that might be changed by @code{longjmp} or 8229@code{vfork}. This warning is also enabled by @option{-Wextra}. 8230 8231@item -Wconversion 8232@opindex Wconversion 8233@opindex Wno-conversion 8234Warn for implicit conversions that may alter a value. This includes 8235conversions between real and integer, like @code{abs (x)} when 8236@code{x} is @code{double}; conversions between signed and unsigned, 8237like @code{unsigned ui = -1}; and conversions to smaller types, like 8238@code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs 8239((int) x)} and @code{ui = (unsigned) -1}, or if the value is not 8240changed by the conversion like in @code{abs (2.0)}. Warnings about 8241conversions between signed and unsigned integers can be disabled by 8242using @option{-Wno-sign-conversion}. 8243 8244For C++, also warn for confusing overload resolution for user-defined 8245conversions; and conversions that never use a type conversion 8246operator: conversions to @code{void}, the same type, a base class or a 8247reference to them. Warnings about conversions between signed and 8248unsigned integers are disabled by default in C++ unless 8249@option{-Wsign-conversion} is explicitly enabled. 8250 8251Warnings about conversion from arithmetic on a small type back to that 8252type are only given with @option{-Warith-conversion}. 8253 8254@item -Wdangling-else 8255@opindex Wdangling-else 8256@opindex Wno-dangling-else 8257Warn about constructions where there may be confusion to which 8258@code{if} statement an @code{else} branch belongs. Here is an example of 8259such a case: 8260 8261@smallexample 8262@group 8263@{ 8264 if (a) 8265 if (b) 8266 foo (); 8267 else 8268 bar (); 8269@} 8270@end group 8271@end smallexample 8272 8273In C/C++, every @code{else} branch belongs to the innermost possible 8274@code{if} statement, which in this example is @code{if (b)}. This is 8275often not what the programmer expected, as illustrated in the above 8276example by indentation the programmer chose. When there is the 8277potential for this confusion, GCC issues a warning when this flag 8278is specified. To eliminate the warning, add explicit braces around 8279the innermost @code{if} statement so there is no way the @code{else} 8280can belong to the enclosing @code{if}. The resulting code 8281looks like this: 8282 8283@smallexample 8284@group 8285@{ 8286 if (a) 8287 @{ 8288 if (b) 8289 foo (); 8290 else 8291 bar (); 8292 @} 8293@} 8294@end group 8295@end smallexample 8296 8297This warning is enabled by @option{-Wparentheses}. 8298 8299@item -Wdate-time 8300@opindex Wdate-time 8301@opindex Wno-date-time 8302Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__} 8303are encountered as they might prevent bit-wise-identical reproducible 8304compilations. 8305 8306@item -Wempty-body 8307@opindex Wempty-body 8308@opindex Wno-empty-body 8309Warn if an empty body occurs in an @code{if}, @code{else} or @code{do 8310while} statement. This warning is also enabled by @option{-Wextra}. 8311 8312@item -Wno-endif-labels 8313@opindex Wendif-labels 8314@opindex Wno-endif-labels 8315Do not warn about stray tokens after @code{#else} and @code{#endif}. 8316 8317@item -Wenum-compare 8318@opindex Wenum-compare 8319@opindex Wno-enum-compare 8320Warn about a comparison between values of different enumerated types. 8321In C++ enumerated type mismatches in conditional expressions are also 8322diagnosed and the warning is enabled by default. In C this warning is 8323enabled by @option{-Wall}. 8324 8325@item -Wenum-conversion 8326@opindex Wenum-conversion 8327@opindex Wno-enum-conversion 8328Warn when a value of enumerated type is implicitly converted to a 8329different enumerated type. This warning is enabled by @option{-Wextra} 8330in C@. 8331 8332@item -Wjump-misses-init @r{(C, Objective-C only)} 8333@opindex Wjump-misses-init 8334@opindex Wno-jump-misses-init 8335Warn if a @code{goto} statement or a @code{switch} statement jumps 8336forward across the initialization of a variable, or jumps backward to a 8337label after the variable has been initialized. This only warns about 8338variables that are initialized when they are declared. This warning is 8339only supported for C and Objective-C; in C++ this sort of branch is an 8340error in any case. 8341 8342@option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It 8343can be disabled with the @option{-Wno-jump-misses-init} option. 8344 8345@item -Wsign-compare 8346@opindex Wsign-compare 8347@opindex Wno-sign-compare 8348@cindex warning for comparison of signed and unsigned values 8349@cindex comparison of signed and unsigned values, warning 8350@cindex signed and unsigned values, comparison warning 8351Warn when a comparison between signed and unsigned values could produce 8352an incorrect result when the signed value is converted to unsigned. 8353In C++, this warning is also enabled by @option{-Wall}. In C, it is 8354also enabled by @option{-Wextra}. 8355 8356@item -Wsign-conversion 8357@opindex Wsign-conversion 8358@opindex Wno-sign-conversion 8359Warn for implicit conversions that may change the sign of an integer 8360value, like assigning a signed integer expression to an unsigned 8361integer variable. An explicit cast silences the warning. In C, this 8362option is enabled also by @option{-Wconversion}. 8363 8364@item -Wfloat-conversion 8365@opindex Wfloat-conversion 8366@opindex Wno-float-conversion 8367Warn for implicit conversions that reduce the precision of a real value. 8368This includes conversions from real to integer, and from higher precision 8369real to lower precision real values. This option is also enabled by 8370@option{-Wconversion}. 8371 8372@item -Wno-scalar-storage-order 8373@opindex Wno-scalar-storage-order 8374@opindex Wscalar-storage-order 8375Do not warn on suspicious constructs involving reverse scalar storage order. 8376 8377@item -Wsizeof-array-div 8378@opindex Wsizeof-array-div 8379@opindex Wno-sizeof-array-div 8380Warn about divisions of two sizeof operators when the first one is applied 8381to an array and the divisor does not equal the size of the array element. 8382In such a case, the computation will not yield the number of elements in the 8383array, which is likely what the user intended. This warning warns e.g. about 8384@smallexample 8385int fn () 8386@{ 8387 int arr[10]; 8388 return sizeof (arr) / sizeof (short); 8389@} 8390@end smallexample 8391 8392This warning is enabled by @option{-Wall}. 8393 8394@item -Wsizeof-pointer-div 8395@opindex Wsizeof-pointer-div 8396@opindex Wno-sizeof-pointer-div 8397Warn for suspicious divisions of two sizeof expressions that divide 8398the pointer size by the element size, which is the usual way to compute 8399the array size but won't work out correctly with pointers. This warning 8400warns e.g.@: about @code{sizeof (ptr) / sizeof (ptr[0])} if @code{ptr} is 8401not an array, but a pointer. This warning is enabled by @option{-Wall}. 8402 8403@item -Wsizeof-pointer-memaccess 8404@opindex Wsizeof-pointer-memaccess 8405@opindex Wno-sizeof-pointer-memaccess 8406Warn for suspicious length parameters to certain string and memory built-in 8407functions if the argument uses @code{sizeof}. This warning triggers for 8408example for @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not 8409an array, but a pointer, and suggests a possible fix, or about 8410@code{memcpy (&foo, ptr, sizeof (&foo));}. @option{-Wsizeof-pointer-memaccess} 8411also warns about calls to bounded string copy functions like @code{strncat} 8412or @code{strncpy} that specify as the bound a @code{sizeof} expression of 8413the source array. For example, in the following function the call to 8414@code{strncat} specifies the size of the source string as the bound. That 8415is almost certainly a mistake and so the call is diagnosed. 8416@smallexample 8417void make_file (const char *name) 8418@{ 8419 char path[PATH_MAX]; 8420 strncpy (path, name, sizeof path - 1); 8421 strncat (path, ".text", sizeof ".text"); 8422 @dots{} 8423@} 8424@end smallexample 8425 8426The @option{-Wsizeof-pointer-memaccess} option is enabled by @option{-Wall}. 8427 8428@item -Wno-sizeof-array-argument 8429@opindex Wsizeof-array-argument 8430@opindex Wno-sizeof-array-argument 8431Do not warn when the @code{sizeof} operator is applied to a parameter that is 8432declared as an array in a function definition. This warning is enabled by 8433default for C and C++ programs. 8434 8435@item -Wmemset-elt-size 8436@opindex Wmemset-elt-size 8437@opindex Wno-memset-elt-size 8438Warn for suspicious calls to the @code{memset} built-in function, if the 8439first argument references an array, and the third argument is a number 8440equal to the number of elements, but not equal to the size of the array 8441in memory. This indicates that the user has omitted a multiplication by 8442the element size. This warning is enabled by @option{-Wall}. 8443 8444@item -Wmemset-transposed-args 8445@opindex Wmemset-transposed-args 8446@opindex Wno-memset-transposed-args 8447Warn for suspicious calls to the @code{memset} built-in function where 8448the second argument is not zero and the third argument is zero. For 8449example, the call @code{memset (buf, sizeof buf, 0)} is diagnosed because 8450@code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostic 8451is only emitted if the third argument is a literal zero. Otherwise, if 8452it is an expression that is folded to zero, or a cast of zero to some 8453type, it is far less likely that the arguments have been mistakenly 8454transposed and no warning is emitted. This warning is enabled 8455by @option{-Wall}. 8456 8457@item -Waddress 8458@opindex Waddress 8459@opindex Wno-address 8460Warn about suspicious uses of memory addresses. These include using 8461the address of a function in a conditional expression, such as 8462@code{void func(void); if (func)}, and comparisons against the memory 8463address of a string literal, such as @code{if (x == "abc")}. Such 8464uses typically indicate a programmer error: the address of a function 8465always evaluates to true, so their use in a conditional usually 8466indicate that the programmer forgot the parentheses in a function 8467call; and comparisons against string literals result in unspecified 8468behavior and are not portable in C, so they usually indicate that the 8469programmer intended to use @code{strcmp}. This warning is enabled by 8470@option{-Wall}. 8471 8472@item -Wno-address-of-packed-member 8473@opindex Waddress-of-packed-member 8474@opindex Wno-address-of-packed-member 8475Do not warn when the address of packed member of struct or union is taken, 8476which usually results in an unaligned pointer value. This is 8477enabled by default. 8478 8479@item -Wlogical-op 8480@opindex Wlogical-op 8481@opindex Wno-logical-op 8482Warn about suspicious uses of logical operators in expressions. 8483This includes using logical operators in contexts where a 8484bit-wise operator is likely to be expected. Also warns when 8485the operands of a logical operator are the same: 8486@smallexample 8487extern int a; 8488if (a < 0 && a < 0) @{ @dots{} @} 8489@end smallexample 8490 8491@item -Wlogical-not-parentheses 8492@opindex Wlogical-not-parentheses 8493@opindex Wno-logical-not-parentheses 8494Warn about logical not used on the left hand side operand of a comparison. 8495This option does not warn if the right operand is considered to be a boolean 8496expression. Its purpose is to detect suspicious code like the following: 8497@smallexample 8498int a; 8499@dots{} 8500if (!a > 1) @{ @dots{} @} 8501@end smallexample 8502 8503It is possible to suppress the warning by wrapping the LHS into 8504parentheses: 8505@smallexample 8506if ((!a) > 1) @{ @dots{} @} 8507@end smallexample 8508 8509This warning is enabled by @option{-Wall}. 8510 8511@item -Waggregate-return 8512@opindex Waggregate-return 8513@opindex Wno-aggregate-return 8514Warn if any functions that return structures or unions are defined or 8515called. (In languages where you can return an array, this also elicits 8516a warning.) 8517 8518@item -Wno-aggressive-loop-optimizations 8519@opindex Wno-aggressive-loop-optimizations 8520@opindex Waggressive-loop-optimizations 8521Warn if in a loop with constant number of iterations the compiler detects 8522undefined behavior in some statement during one or more of the iterations. 8523 8524@item -Wno-attributes 8525@opindex Wno-attributes 8526@opindex Wattributes 8527Do not warn if an unexpected @code{__attribute__} is used, such as 8528unrecognized attributes, function attributes applied to variables, 8529etc. This does not stop errors for incorrect use of supported 8530attributes. 8531 8532@item -Wno-builtin-declaration-mismatch 8533@opindex Wno-builtin-declaration-mismatch 8534@opindex Wbuiltin-declaration-mismatch 8535Warn if a built-in function is declared with an incompatible signature 8536or as a non-function, or when a built-in function declared with a type 8537that does not include a prototype is called with arguments whose promoted 8538types do not match those expected by the function. When @option{-Wextra} 8539is specified, also warn when a built-in function that takes arguments is 8540declared without a prototype. The @option{-Wbuiltin-declaration-mismatch} 8541warning is enabled by default. To avoid the warning include the appropriate 8542header to bring the prototypes of built-in functions into scope. 8543 8544For example, the call to @code{memset} below is diagnosed by the warning 8545because the function expects a value of type @code{size_t} as its argument 8546but the type of @code{32} is @code{int}. With @option{-Wextra}, 8547the declaration of the function is diagnosed as well. 8548@smallexample 8549extern void* memset (); 8550void f (void *d) 8551@{ 8552 memset (d, '\0', 32); 8553@} 8554@end smallexample 8555 8556@item -Wno-builtin-macro-redefined 8557@opindex Wno-builtin-macro-redefined 8558@opindex Wbuiltin-macro-redefined 8559Do not warn if certain built-in macros are redefined. This suppresses 8560warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__}, 8561@code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}. 8562 8563@item -Wstrict-prototypes @r{(C and Objective-C only)} 8564@opindex Wstrict-prototypes 8565@opindex Wno-strict-prototypes 8566Warn if a function is declared or defined without specifying the 8567argument types. (An old-style function definition is permitted without 8568a warning if preceded by a declaration that specifies the argument 8569types.) 8570 8571@item -Wold-style-declaration @r{(C and Objective-C only)} 8572@opindex Wold-style-declaration 8573@opindex Wno-old-style-declaration 8574Warn for obsolescent usages, according to the C Standard, in a 8575declaration. For example, warn if storage-class specifiers like 8576@code{static} are not the first things in a declaration. This warning 8577is also enabled by @option{-Wextra}. 8578 8579@item -Wold-style-definition @r{(C and Objective-C only)} 8580@opindex Wold-style-definition 8581@opindex Wno-old-style-definition 8582Warn if an old-style function definition is used. A warning is given 8583even if there is a previous prototype. A definition using @samp{()} 8584is not considered an old-style definition in C2X mode, because it is 8585equivalent to @samp{(void)} in that case, but is considered an 8586old-style definition for older standards. 8587 8588@item -Wmissing-parameter-type @r{(C and Objective-C only)} 8589@opindex Wmissing-parameter-type 8590@opindex Wno-missing-parameter-type 8591A function parameter is declared without a type specifier in K&R-style 8592functions: 8593 8594@smallexample 8595void foo(bar) @{ @} 8596@end smallexample 8597 8598This warning is also enabled by @option{-Wextra}. 8599 8600@item -Wmissing-prototypes @r{(C and Objective-C only)} 8601@opindex Wmissing-prototypes 8602@opindex Wno-missing-prototypes 8603Warn if a global function is defined without a previous prototype 8604declaration. This warning is issued even if the definition itself 8605provides a prototype. Use this option to detect global functions 8606that do not have a matching prototype declaration in a header file. 8607This option is not valid for C++ because all function declarations 8608provide prototypes and a non-matching declaration declares an 8609overload rather than conflict with an earlier declaration. 8610Use @option{-Wmissing-declarations} to detect missing declarations in C++. 8611 8612@item -Wmissing-declarations 8613@opindex Wmissing-declarations 8614@opindex Wno-missing-declarations 8615Warn if a global function is defined without a previous declaration. 8616Do so even if the definition itself provides a prototype. 8617Use this option to detect global functions that are not declared in 8618header files. In C, no warnings are issued for functions with previous 8619non-prototype declarations; use @option{-Wmissing-prototypes} to detect 8620missing prototypes. In C++, no warnings are issued for function templates, 8621or for inline functions, or for functions in anonymous namespaces. 8622 8623@item -Wmissing-field-initializers 8624@opindex Wmissing-field-initializers 8625@opindex Wno-missing-field-initializers 8626@opindex W 8627@opindex Wextra 8628@opindex Wno-extra 8629Warn if a structure's initializer has some fields missing. For 8630example, the following code causes such a warning, because 8631@code{x.h} is implicitly zero: 8632 8633@smallexample 8634struct s @{ int f, g, h; @}; 8635struct s x = @{ 3, 4 @}; 8636@end smallexample 8637 8638This option does not warn about designated initializers, so the following 8639modification does not trigger a warning: 8640 8641@smallexample 8642struct s @{ int f, g, h; @}; 8643struct s x = @{ .f = 3, .g = 4 @}; 8644@end smallexample 8645 8646In C this option does not warn about the universal zero initializer 8647@samp{@{ 0 @}}: 8648 8649@smallexample 8650struct s @{ int f, g, h; @}; 8651struct s x = @{ 0 @}; 8652@end smallexample 8653 8654Likewise, in C++ this option does not warn about the empty @{ @} 8655initializer, for example: 8656 8657@smallexample 8658struct s @{ int f, g, h; @}; 8659s x = @{ @}; 8660@end smallexample 8661 8662This warning is included in @option{-Wextra}. To get other @option{-Wextra} 8663warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}. 8664 8665@item -Wno-multichar 8666@opindex Wno-multichar 8667@opindex Wmultichar 8668Do not warn if a multicharacter constant (@samp{'FOOF'}) is used. 8669Usually they indicate a typo in the user's code, as they have 8670implementation-defined values, and should not be used in portable code. 8671 8672@item -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} 8673@opindex Wnormalized= 8674@opindex Wnormalized 8675@opindex Wno-normalized 8676@cindex NFC 8677@cindex NFKC 8678@cindex character set, input normalization 8679In ISO C and ISO C++, two identifiers are different if they are 8680different sequences of characters. However, sometimes when characters 8681outside the basic ASCII character set are used, you can have two 8682different character sequences that look the same. To avoid confusion, 8683the ISO 10646 standard sets out some @dfn{normalization rules} which 8684when applied ensure that two sequences that look the same are turned into 8685the same sequence. GCC can warn you if you are using identifiers that 8686have not been normalized; this option controls that warning. 8687 8688There are four levels of warning supported by GCC@. The default is 8689@option{-Wnormalized=nfc}, which warns about any identifier that is 8690not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the 8691recommended form for most uses. It is equivalent to 8692@option{-Wnormalized}. 8693 8694Unfortunately, there are some characters allowed in identifiers by 8695ISO C and ISO C++ that, when turned into NFC, are not allowed in 8696identifiers. That is, there's no way to use these symbols in portable 8697ISO C or C++ and have all your identifiers in NFC@. 8698@option{-Wnormalized=id} suppresses the warning for these characters. 8699It is hoped that future versions of the standards involved will correct 8700this, which is why this option is not the default. 8701 8702You can switch the warning off for all characters by writing 8703@option{-Wnormalized=none} or @option{-Wno-normalized}. You should 8704only do this if you are using some other normalization scheme (like 8705``D''), because otherwise you can easily create bugs that are 8706literally impossible to see. 8707 8708Some characters in ISO 10646 have distinct meanings but look identical 8709in some fonts or display methodologies, especially once formatting has 8710been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL 8711LETTER N'', displays just like a regular @code{n} that has been 8712placed in a superscript. ISO 10646 defines the @dfn{NFKC} 8713normalization scheme to convert all these into a standard form as 8714well, and GCC warns if your code is not in NFKC if you use 8715@option{-Wnormalized=nfkc}. This warning is comparable to warning 8716about every identifier that contains the letter O because it might be 8717confused with the digit 0, and so is not the default, but may be 8718useful as a local coding convention if the programming environment 8719cannot be fixed to display these characters distinctly. 8720 8721@item -Wno-attribute-warning 8722@opindex Wno-attribute-warning 8723@opindex Wattribute-warning 8724Do not warn about usage of functions (@pxref{Function Attributes}) 8725declared with @code{warning} attribute. By default, this warning is 8726enabled. @option{-Wno-attribute-warning} can be used to disable the 8727warning or @option{-Wno-error=attribute-warning} can be used to 8728disable the error when compiled with @option{-Werror} flag. 8729 8730@item -Wno-deprecated 8731@opindex Wno-deprecated 8732@opindex Wdeprecated 8733Do not warn about usage of deprecated features. @xref{Deprecated Features}. 8734 8735@item -Wno-deprecated-declarations 8736@opindex Wno-deprecated-declarations 8737@opindex Wdeprecated-declarations 8738Do not warn about uses of functions (@pxref{Function Attributes}), 8739variables (@pxref{Variable Attributes}), and types (@pxref{Type 8740Attributes}) marked as deprecated by using the @code{deprecated} 8741attribute. 8742 8743@item -Wno-overflow 8744@opindex Wno-overflow 8745@opindex Woverflow 8746Do not warn about compile-time overflow in constant expressions. 8747 8748@item -Wno-odr 8749@opindex Wno-odr 8750@opindex Wodr 8751Warn about One Definition Rule violations during link-time optimization. 8752Enabled by default. 8753 8754@item -Wopenmp-simd 8755@opindex Wopenmp-simd 8756@opindex Wno-openmp-simd 8757Warn if the vectorizer cost model overrides the OpenMP 8758simd directive set by user. The @option{-fsimd-cost-model=unlimited} 8759option can be used to relax the cost model. 8760 8761@item -Woverride-init @r{(C and Objective-C only)} 8762@opindex Woverride-init 8763@opindex Wno-override-init 8764@opindex W 8765@opindex Wextra 8766@opindex Wno-extra 8767Warn if an initialized field without side effects is overridden when 8768using designated initializers (@pxref{Designated Inits, , Designated 8769Initializers}). 8770 8771This warning is included in @option{-Wextra}. To get other 8772@option{-Wextra} warnings without this one, use @option{-Wextra 8773-Wno-override-init}. 8774 8775@item -Wno-override-init-side-effects @r{(C and Objective-C only)} 8776@opindex Woverride-init-side-effects 8777@opindex Wno-override-init-side-effects 8778Do not warn if an initialized field with side effects is overridden when 8779using designated initializers (@pxref{Designated Inits, , Designated 8780Initializers}). This warning is enabled by default. 8781 8782@item -Wpacked 8783@opindex Wpacked 8784@opindex Wno-packed 8785Warn if a structure is given the packed attribute, but the packed 8786attribute has no effect on the layout or size of the structure. 8787Such structures may be mis-aligned for little benefit. For 8788instance, in this code, the variable @code{f.x} in @code{struct bar} 8789is misaligned even though @code{struct bar} does not itself 8790have the packed attribute: 8791 8792@smallexample 8793@group 8794struct foo @{ 8795 int x; 8796 char a, b, c, d; 8797@} __attribute__((packed)); 8798struct bar @{ 8799 char z; 8800 struct foo f; 8801@}; 8802@end group 8803@end smallexample 8804 8805@item -Wnopacked-bitfield-compat 8806@opindex Wpacked-bitfield-compat 8807@opindex Wno-packed-bitfield-compat 8808The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute 8809on bit-fields of type @code{char}. This was fixed in GCC 4.4 but 8810the change can lead to differences in the structure layout. GCC 8811informs you when the offset of such a field has changed in GCC 4.4. 8812For example there is no longer a 4-bit padding between field @code{a} 8813and @code{b} in this structure: 8814 8815@smallexample 8816struct foo 8817@{ 8818 char a:4; 8819 char b:8; 8820@} __attribute__ ((packed)); 8821@end smallexample 8822 8823This warning is enabled by default. Use 8824@option{-Wno-packed-bitfield-compat} to disable this warning. 8825 8826@item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)} 8827@opindex Wpacked-not-aligned 8828@opindex Wno-packed-not-aligned 8829Warn if a structure field with explicitly specified alignment in a 8830packed struct or union is misaligned. For example, a warning will 8831be issued on @code{struct S}, like, @code{warning: alignment 1 of 8832'struct S' is less than 8}, in this code: 8833 8834@smallexample 8835@group 8836struct __attribute__ ((aligned (8))) S8 @{ char a[8]; @}; 8837struct __attribute__ ((packed)) S @{ 8838 struct S8 s8; 8839@}; 8840@end group 8841@end smallexample 8842 8843This warning is enabled by @option{-Wall}. 8844 8845@item -Wpadded 8846@opindex Wpadded 8847@opindex Wno-padded 8848Warn if padding is included in a structure, either to align an element 8849of the structure or to align the whole structure. Sometimes when this 8850happens it is possible to rearrange the fields of the structure to 8851reduce the padding and so make the structure smaller. 8852 8853@item -Wredundant-decls 8854@opindex Wredundant-decls 8855@opindex Wno-redundant-decls 8856Warn if anything is declared more than once in the same scope, even in 8857cases where multiple declaration is valid and changes nothing. 8858 8859@item -Wrestrict 8860@opindex Wrestrict 8861@opindex Wno-restrict 8862Warn when an object referenced by a @code{restrict}-qualified parameter 8863(or, in C++, a @code{__restrict}-qualified parameter) is aliased by another 8864argument, or when copies between such objects overlap. For example, 8865the call to the @code{strcpy} function below attempts to truncate the string 8866by replacing its initial characters with the last four. However, because 8867the call writes the terminating NUL into @code{a[4]}, the copies overlap and 8868the call is diagnosed. 8869 8870@smallexample 8871void foo (void) 8872@{ 8873 char a[] = "abcd1234"; 8874 strcpy (a, a + 4); 8875 @dots{} 8876@} 8877@end smallexample 8878The @option{-Wrestrict} option detects some instances of simple overlap 8879even without optimization but works best at @option{-O2} and above. It 8880is included in @option{-Wall}. 8881 8882@item -Wnested-externs @r{(C and Objective-C only)} 8883@opindex Wnested-externs 8884@opindex Wno-nested-externs 8885Warn if an @code{extern} declaration is encountered within a function. 8886 8887@item -Winline 8888@opindex Winline 8889@opindex Wno-inline 8890Warn if a function that is declared as inline cannot be inlined. 8891Even with this option, the compiler does not warn about failures to 8892inline functions declared in system headers. 8893 8894The compiler uses a variety of heuristics to determine whether or not 8895to inline a function. For example, the compiler takes into account 8896the size of the function being inlined and the amount of inlining 8897that has already been done in the current function. Therefore, 8898seemingly insignificant changes in the source program can cause the 8899warnings produced by @option{-Winline} to appear or disappear. 8900 8901@item -Wint-in-bool-context 8902@opindex Wint-in-bool-context 8903@opindex Wno-int-in-bool-context 8904Warn for suspicious use of integer values where boolean values are expected, 8905such as conditional expressions (?:) using non-boolean integer constants in 8906boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed 8907integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise 8908for all kinds of multiplications regardless of the data type. 8909This warning is enabled by @option{-Wall}. 8910 8911@item -Wno-int-to-pointer-cast 8912@opindex Wno-int-to-pointer-cast 8913@opindex Wint-to-pointer-cast 8914Suppress warnings from casts to pointer type of an integer of a 8915different size. In C++, casting to a pointer type of smaller size is 8916an error. @option{Wint-to-pointer-cast} is enabled by default. 8917 8918 8919@item -Wno-pointer-to-int-cast @r{(C and Objective-C only)} 8920@opindex Wno-pointer-to-int-cast 8921@opindex Wpointer-to-int-cast 8922Suppress warnings from casts from a pointer to an integer type of a 8923different size. 8924 8925@item -Winvalid-pch 8926@opindex Winvalid-pch 8927@opindex Wno-invalid-pch 8928Warn if a precompiled header (@pxref{Precompiled Headers}) is found in 8929the search path but cannot be used. 8930 8931@item -Wlong-long 8932@opindex Wlong-long 8933@opindex Wno-long-long 8934Warn if @code{long long} type is used. This is enabled by either 8935@option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98 8936modes. To inhibit the warning messages, use @option{-Wno-long-long}. 8937 8938@item -Wvariadic-macros 8939@opindex Wvariadic-macros 8940@opindex Wno-variadic-macros 8941Warn if variadic macros are used in ISO C90 mode, or if the GNU 8942alternate syntax is used in ISO C99 mode. This is enabled by either 8943@option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning 8944messages, use @option{-Wno-variadic-macros}. 8945 8946@item -Wno-varargs 8947@opindex Wvarargs 8948@opindex Wno-varargs 8949Do not warn upon questionable usage of the macros used to handle variable 8950arguments like @code{va_start}. These warnings are enabled by default. 8951 8952@item -Wvector-operation-performance 8953@opindex Wvector-operation-performance 8954@opindex Wno-vector-operation-performance 8955Warn if vector operation is not implemented via SIMD capabilities of the 8956architecture. Mainly useful for the performance tuning. 8957Vector operation can be implemented @code{piecewise}, which means that the 8958scalar operation is performed on every vector element; 8959@code{in parallel}, which means that the vector operation is implemented 8960using scalars of wider type, which normally is more performance efficient; 8961and @code{as a single scalar}, which means that vector fits into a 8962scalar type. 8963 8964@item -Wvla 8965@opindex Wvla 8966@opindex Wno-vla 8967Warn if a variable-length array is used in the code. 8968@option{-Wno-vla} prevents the @option{-Wpedantic} warning of 8969the variable-length array. 8970 8971@item -Wvla-larger-than=@var{byte-size} 8972@opindex Wvla-larger-than= 8973@opindex Wno-vla-larger-than 8974If this option is used, the compiler warns for declarations of 8975variable-length arrays whose size is either unbounded, or bounded 8976by an argument that allows the array size to exceed @var{byte-size} 8977bytes. This is similar to how @option{-Walloca-larger-than=}@var{byte-size} 8978works, but with variable-length arrays. 8979 8980Note that GCC may optimize small variable-length arrays of a known 8981value into plain arrays, so this warning may not get triggered for 8982such arrays. 8983 8984@option{-Wvla-larger-than=}@samp{PTRDIFF_MAX} is enabled by default but 8985is typically only effective when @option{-ftree-vrp} is active (default 8986for @option{-O2} and above). 8987 8988See also @option{-Walloca-larger-than=@var{byte-size}}. 8989 8990@item -Wno-vla-larger-than 8991@opindex Wno-vla-larger-than 8992Disable @option{-Wvla-larger-than=} warnings. The option is equivalent 8993to @option{-Wvla-larger-than=}@samp{SIZE_MAX} or larger. 8994 8995@item -Wvla-parameter 8996@opindex Wno-vla-parameter 8997Warn about redeclarations of functions involving arguments of Variable 8998Length Array types of inconsistent kinds or forms, and enable the detection 8999of out-of-bounds accesses to such parameters by warnings such as 9000@option{-Warray-bounds}. 9001 9002If the first function declaration uses the VLA form the bound specified 9003in the array is assumed to be the minimum number of elements expected to 9004be provided in calls to the function and the maximum number of elements 9005accessed by it. Failing to provide arguments of sufficient size or 9006accessing more than the maximum number of elements may be diagnosed. 9007 9008For example, the warning triggers for the following redeclarations because 9009the first one allows an array of any size to be passed to @code{f} while 9010the second one specifies that the array argument must have at least @code{n} 9011elements. In addition, calling @code{f} with the assotiated VLA bound 9012parameter in excess of the actual VLA bound triggers a warning as well. 9013 9014@smallexample 9015void f (int n, int[n]); 9016void f (int, int[]); // warning: argument 2 previously declared as a VLA 9017 9018void g (int n) 9019@{ 9020 if (n > 4) 9021 return; 9022 int a[n]; 9023 f (sizeof a, a); // warning: access to a by f may be out of bounds 9024 @dots{} 9025@} 9026 9027@end smallexample 9028 9029@option{-Wvla-parameter} is included in @option{-Wall}. The 9030@option{-Warray-parameter} option triggers warnings for similar problems 9031involving ordinary array arguments. 9032 9033@item -Wvolatile-register-var 9034@opindex Wvolatile-register-var 9035@opindex Wno-volatile-register-var 9036Warn if a register variable is declared volatile. The volatile 9037modifier does not inhibit all optimizations that may eliminate reads 9038and/or writes to register variables. This warning is enabled by 9039@option{-Wall}. 9040 9041@item -Wdisabled-optimization 9042@opindex Wdisabled-optimization 9043@opindex Wno-disabled-optimization 9044Warn if a requested optimization pass is disabled. This warning does 9045not generally indicate that there is anything wrong with your code; it 9046merely indicates that GCC's optimizers are unable to handle the code 9047effectively. Often, the problem is that your code is too big or too 9048complex; GCC refuses to optimize programs when the optimization 9049itself is likely to take inordinate amounts of time. 9050 9051@item -Wpointer-sign @r{(C and Objective-C only)} 9052@opindex Wpointer-sign 9053@opindex Wno-pointer-sign 9054Warn for pointer argument passing or assignment with different signedness. 9055This option is only supported for C and Objective-C@. It is implied by 9056@option{-Wall} and by @option{-Wpedantic}, which can be disabled with 9057@option{-Wno-pointer-sign}. 9058 9059@item -Wstack-protector 9060@opindex Wstack-protector 9061@opindex Wno-stack-protector 9062This option is only active when @option{-fstack-protector} is active. It 9063warns about functions that are not protected against stack smashing. 9064 9065@item -Woverlength-strings 9066@opindex Woverlength-strings 9067@opindex Wno-overlength-strings 9068Warn about string constants that are longer than the ``minimum 9069maximum'' length specified in the C standard. Modern compilers 9070generally allow string constants that are much longer than the 9071standard's minimum limit, but very portable programs should avoid 9072using longer strings. 9073 9074The limit applies @emph{after} string constant concatenation, and does 9075not count the trailing NUL@. In C90, the limit was 509 characters; in 9076C99, it was raised to 4095. C++98 does not specify a normative 9077minimum maximum, so we do not diagnose overlength strings in C++@. 9078 9079This option is implied by @option{-Wpedantic}, and can be disabled with 9080@option{-Wno-overlength-strings}. 9081 9082@item -Wunsuffixed-float-constants @r{(C and Objective-C only)} 9083@opindex Wunsuffixed-float-constants 9084@opindex Wno-unsuffixed-float-constants 9085 9086Issue a warning for any floating constant that does not have 9087a suffix. When used together with @option{-Wsystem-headers} it 9088warns about such constants in system header files. This can be useful 9089when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma 9090from the decimal floating-point extension to C99. 9091 9092@item -Wno-lto-type-mismatch 9093@opindex Wlto-type-mismatch 9094@opindex Wno-lto-type-mismatch 9095 9096During the link-time optimization, do not warn about type mismatches in 9097global declarations from different compilation units. 9098Requires @option{-flto} to be enabled. Enabled by default. 9099 9100@item -Wno-designated-init @r{(C and Objective-C only)} 9101@opindex Wdesignated-init 9102@opindex Wno-designated-init 9103Suppress warnings when a positional initializer is used to initialize 9104a structure that has been marked with the @code{designated_init} 9105attribute. 9106 9107@end table 9108 9109@node Static Analyzer Options 9110@section Options That Control Static Analysis 9111 9112@table @gcctabopt 9113@item -fanalyzer 9114@opindex analyzer 9115@opindex fanalyzer 9116@opindex fno-analyzer 9117This option enables an static analysis of program flow which looks 9118for ``interesting'' interprocedural paths through the 9119code, and issues warnings for problems found on them. 9120 9121This analysis is much more expensive than other GCC warnings. 9122 9123Enabling this option effectively enables the following warnings: 9124 9125@gccoptlist{ @gol 9126-Wanalyzer-double-fclose @gol 9127-Wanalyzer-double-free @gol 9128-Wanalyzer-exposure-through-output-file @gol 9129-Wanalyzer-file-leak @gol 9130-Wanalyzer-free-of-non-heap @gol 9131-Wanalyzer-malloc-leak @gol 9132-Wanalyzer-mismatching-deallocation @gol 9133-Wanalyzer-possible-null-argument @gol 9134-Wanalyzer-possible-null-dereference @gol 9135-Wanalyzer-null-argument @gol 9136-Wanalyzer-null-dereference @gol 9137-Wanalyzer-shift-count-negative @gol 9138-Wanalyzer-shift-count-overflow @gol 9139-Wanalyzer-stale-setjmp-buffer @gol 9140-Wanalyzer-tainted-array-index @gol 9141-Wanalyzer-unsafe-call-within-signal-handler @gol 9142-Wanalyzer-use-after-free @gol 9143-Wanalyzer-use-of-pointer-in-stale-stack-frame @gol 9144-Wanalyzer-write-to-const @gol 9145-Wanalyzer-write-to-string-literal @gol 9146} 9147 9148This option is only available if GCC was configured with analyzer 9149support enabled. 9150 9151@item -Wanalyzer-too-complex 9152@opindex Wanalyzer-too-complex 9153@opindex Wno-analyzer-too-complex 9154If @option{-fanalyzer} is enabled, the analyzer uses various heuristics 9155to attempt to explore the control flow and data flow in the program, 9156but these can be defeated by sufficiently complicated code. 9157 9158By default, the analysis silently stops if the code is too 9159complicated for the analyzer to fully explore and it reaches an internal 9160limit. The @option{-Wanalyzer-too-complex} option warns if this occurs. 9161 9162@item -Wno-analyzer-double-fclose 9163@opindex Wanalyzer-double-fclose 9164@opindex Wno-analyzer-double-fclose 9165This warning requires @option{-fanalyzer}, which enables it; use 9166@option{-Wno-analyzer-double-fclose} to disable it. 9167 9168This diagnostic warns for paths through the code in which a @code{FILE *} 9169can have @code{fclose} called on it more than once. 9170 9171@item -Wno-analyzer-double-free 9172@opindex Wanalyzer-double-free 9173@opindex Wno-analyzer-double-free 9174This warning requires @option{-fanalyzer}, which enables it; use 9175@option{-Wno-analyzer-double-free} to disable it. 9176 9177This diagnostic warns for paths through the code in which a pointer 9178can have a deallocator called on it more than once, either @code{free}, 9179or a deallocator referenced by attribute @code{malloc}. 9180 9181@item -Wno-analyzer-exposure-through-output-file 9182@opindex Wanalyzer-exposure-through-output-file 9183@opindex Wno-analyzer-exposure-through-output-file 9184This warning requires @option{-fanalyzer}, which enables it; use 9185@option{-Wno-analyzer-exposure-through-output-file} 9186to disable it. 9187 9188This diagnostic warns for paths through the code in which a 9189security-sensitive value is written to an output file 9190(such as writing a password to a log file). 9191 9192@item -Wno-analyzer-file-leak 9193@opindex Wanalyzer-file-leak 9194@opindex Wno-analyzer-file-leak 9195This warning requires @option{-fanalyzer}, which enables it; use 9196@option{-Wno-analyzer-file-leak} 9197to disable it. 9198 9199This diagnostic warns for paths through the code in which a 9200@code{<stdio.h>} @code{FILE *} stream object is leaked. 9201 9202@item -Wno-analyzer-free-of-non-heap 9203@opindex Wanalyzer-free-of-non-heap 9204@opindex Wno-analyzer-free-of-non-heap 9205This warning requires @option{-fanalyzer}, which enables it; use 9206@option{-Wno-analyzer-free-of-non-heap} 9207to disable it. 9208 9209This diagnostic warns for paths through the code in which @code{free} 9210is called on a non-heap pointer (e.g. an on-stack buffer, or a global). 9211 9212@item -Wno-analyzer-malloc-leak 9213@opindex Wanalyzer-malloc-leak 9214@opindex Wno-analyzer-malloc-leak 9215This warning requires @option{-fanalyzer}, which enables it; use 9216@option{-Wno-analyzer-malloc-leak} 9217to disable it. 9218 9219This diagnostic warns for paths through the code in which a 9220pointer allocated via an allocator is leaked: either @code{malloc}, 9221or a function marked with attribute @code{malloc}. 9222 9223@item -Wno-analyzer-mismatching-deallocation 9224@opindex Wanalyzer-mismatching-deallocation 9225@opindex Wno-analyzer-mismatching-deallocation 9226This warning requires @option{-fanalyzer}, which enables it; use 9227@option{-Wno-analyzer-mismatching-deallocation} 9228to disable it. 9229 9230This diagnostic warns for paths through the code in which the 9231wrong deallocation function is called on a pointer value, based on 9232which function was used to allocate the pointer value. The diagnostic 9233will warn about mismatches between @code{free}, scalar @code{delete} 9234and vector @code{delete[]}, and those marked as allocator/deallocator 9235pairs using attribute @code{malloc}. 9236 9237@item -Wno-analyzer-possible-null-argument 9238@opindex Wanalyzer-possible-null-argument 9239@opindex Wno-analyzer-possible-null-argument 9240This warning requires @option{-fanalyzer}, which enables it; use 9241@option{-Wno-analyzer-possible-null-argument} to disable it. 9242 9243This diagnostic warns for paths through the code in which a 9244possibly-NULL value is passed to a function argument marked 9245with @code{__attribute__((nonnull))} as requiring a non-NULL 9246value. 9247 9248@item -Wno-analyzer-possible-null-dereference 9249@opindex Wanalyzer-possible-null-dereference 9250@opindex Wno-analyzer-possible-null-dereference 9251This warning requires @option{-fanalyzer}, which enables it; use 9252@option{-Wno-analyzer-possible-null-dereference} to disable it. 9253 9254This diagnostic warns for paths through the code in which a 9255possibly-NULL value is dereferenced. 9256 9257@item -Wno-analyzer-null-argument 9258@opindex Wanalyzer-null-argument 9259@opindex Wno-analyzer-null-argument 9260This warning requires @option{-fanalyzer}, which enables it; use 9261@option{-Wno-analyzer-null-argument} to disable it. 9262 9263This diagnostic warns for paths through the code in which a 9264value known to be NULL is passed to a function argument marked 9265with @code{__attribute__((nonnull))} as requiring a non-NULL 9266value. 9267 9268@item -Wno-analyzer-null-dereference 9269@opindex Wanalyzer-null-dereference 9270@opindex Wno-analyzer-null-dereference 9271This warning requires @option{-fanalyzer}, which enables it; use 9272@option{-Wno-analyzer-null-dereference} to disable it. 9273 9274This diagnostic warns for paths through the code in which a 9275value known to be NULL is dereferenced. 9276 9277@item -Wno-analyzer-shift-count-negative 9278@opindex Wanalyzer-shift-count-negative 9279@opindex Wno-analyzer-shift-count-negative 9280This warning requires @option{-fanalyzer}, which enables it; use 9281@option{-Wno-analyzer-shift-count-negative} to disable it. 9282 9283This diagnostic warns for paths through the code in which a 9284shift is attempted with a negative count. It is analogous to 9285the @option{-Wshift-count-negative} diagnostic implemented in 9286the C/C++ front ends, but is implemented based on analyzing 9287interprocedural paths, rather than merely parsing the syntax tree. 9288However, the analyzer does not prioritize detection of such paths, so 9289false negatives are more likely relative to other warnings. 9290 9291@item -Wno-analyzer-shift-count-overflow 9292@opindex Wanalyzer-shift-count-overflow 9293@opindex Wno-analyzer-shift-count-overflow 9294This warning requires @option{-fanalyzer}, which enables it; use 9295@option{-Wno-analyzer-shift-count-overflow} to disable it. 9296 9297This diagnostic warns for paths through the code in which a 9298shift is attempted with a count greater than or equal to the 9299precision of the operand's type. It is analogous to 9300the @option{-Wshift-count-overflow} diagnostic implemented in 9301the C/C++ front ends, but is implemented based on analyzing 9302interprocedural paths, rather than merely parsing the syntax tree. 9303However, the analyzer does not prioritize detection of such paths, so 9304false negatives are more likely relative to other warnings. 9305 9306@item -Wno-analyzer-stale-setjmp-buffer 9307@opindex Wanalyzer-stale-setjmp-buffer 9308@opindex Wno-analyzer-stale-setjmp-buffer 9309This warning requires @option{-fanalyzer}, which enables it; use 9310@option{-Wno-analyzer-stale-setjmp-buffer} to disable it. 9311 9312This diagnostic warns for paths through the code in which 9313@code{longjmp} is called to rewind to a @code{jmp_buf} relating 9314to a @code{setjmp} call in a function that has returned. 9315 9316When @code{setjmp} is called on a @code{jmp_buf} to record a rewind 9317location, it records the stack frame. The stack frame becomes invalid 9318when the function containing the @code{setjmp} call returns. Attempting 9319to rewind to it via @code{longjmp} would reference a stack frame that 9320no longer exists, and likely lead to a crash (or worse). 9321 9322@item -Wno-analyzer-tainted-array-index 9323@opindex Wanalyzer-tainted-array-index 9324@opindex Wno-analyzer-tainted-array-index 9325This warning requires both @option{-fanalyzer} and 9326@option{-fanalyzer-checker=taint} to enable it; 9327use @option{-Wno-analyzer-tainted-array-index} to disable it. 9328 9329This diagnostic warns for paths through the code in which a value 9330that could be under an attacker's control is used as the index 9331of an array access without being sanitized. 9332 9333@item -Wno-analyzer-unsafe-call-within-signal-handler 9334@opindex Wanalyzer-unsafe-call-within-signal-handler 9335@opindex Wno-analyzer-unsafe-call-within-signal-handler 9336This warning requires @option{-fanalyzer}, which enables it; use 9337@option{-Wno-analyzer-unsafe-call-within-signal-handler} to disable it. 9338 9339This diagnostic warns for paths through the code in which a 9340function known to be async-signal-unsafe (such as @code{fprintf}) is 9341called from a signal handler. 9342 9343@item -Wno-analyzer-use-after-free 9344@opindex Wanalyzer-use-after-free 9345@opindex Wno-analyzer-use-after-free 9346This warning requires @option{-fanalyzer}, which enables it; use 9347@option{-Wno-analyzer-use-after-free} to disable it. 9348 9349This diagnostic warns for paths through the code in which a 9350pointer is used after a deallocator is called on it: either @code{free}, 9351or a deallocator referenced by attribute @code{malloc}. 9352 9353@item -Wno-analyzer-use-of-pointer-in-stale-stack-frame 9354@opindex Wanalyzer-use-of-pointer-in-stale-stack-frame 9355@opindex Wno-analyzer-use-of-pointer-in-stale-stack-frame 9356This warning requires @option{-fanalyzer}, which enables it; use 9357@option{-Wno-analyzer-use-of-pointer-in-stale-stack-frame} 9358to disable it. 9359 9360This diagnostic warns for paths through the code in which a pointer 9361is dereferenced that points to a variable in a stale stack frame. 9362 9363@item -Wno-analyzer-write-to-const 9364@opindex Wanalyzer-write-to-const 9365@opindex Wno-analyzer-write-to-const 9366This warning requires @option{-fanalyzer}, which enables it; use 9367@option{-Wno-analyzer-write-to-const} 9368to disable it. 9369 9370This diagnostic warns for paths through the code in which the analyzer 9371detects an attempt to write through a pointer to a @code{const} object. 9372However, the analyzer does not prioritize detection of such paths, so 9373false negatives are more likely relative to other warnings. 9374 9375@item -Wno-analyzer-write-to-string-literal 9376@opindex Wanalyzer-write-to-string-literal 9377@opindex Wno-analyzer-write-to-string-literal 9378This warning requires @option{-fanalyzer}, which enables it; use 9379@option{-Wno-analyzer-write-to-string-literal} 9380to disable it. 9381 9382This diagnostic warns for paths through the code in which the analyzer 9383detects an attempt to write through a pointer to a string literal. 9384However, the analyzer does not prioritize detection of such paths, so 9385false negatives are more likely relative to other warnings. 9386 9387@end table 9388 9389Pertinent parameters for controlling the exploration are: 9390@option{--param analyzer-bb-explosion-factor=@var{value}}, 9391@option{--param analyzer-max-enodes-per-program-point=@var{value}}, 9392@option{--param analyzer-max-recursion-depth=@var{value}}, and 9393@option{--param analyzer-min-snodes-for-call-summary=@var{value}}. 9394 9395The following options control the analyzer. 9396 9397@table @gcctabopt 9398 9399@item -fanalyzer-call-summaries 9400@opindex fanalyzer-call-summaries 9401@opindex fno-analyzer-call-summaries 9402Simplify interprocedural analysis by computing the effect of certain calls, 9403rather than exploring all paths through the function from callsite to each 9404possible return. 9405 9406If enabled, call summaries are only used for functions with more than one 9407call site, and that are sufficiently complicated (as per 9408@option{--param analyzer-min-snodes-for-call-summary=@var{value}}). 9409 9410@item -fanalyzer-checker=@var{name} 9411@opindex fanalyzer-checker 9412Restrict the analyzer to run just the named checker, and enable it. 9413 9414Some checkers are disabled by default (even with @option{-fanalyzer}), 9415such as the @code{taint} checker that implements 9416@option{-Wanalyzer-tainted-array-index}, and this option is required 9417to enable them. 9418 9419@item -fno-analyzer-feasibility 9420@opindex fanalyzer-feasibility 9421@opindex fno-analyzer-feasibility 9422This option is intended for analyzer developers. 9423 9424By default the analyzer verifies that there is a feasible control flow path 9425for each diagnostic it emits: that the conditions that hold are not mutually 9426exclusive. Diagnostics for which no feasible path can be found are rejected. 9427This filtering can be suppressed with @option{-fno-analyzer-feasibility}, for 9428debugging issues in this code. 9429 9430@item -fanalyzer-fine-grained 9431@opindex fanalyzer-fine-grained 9432@opindex fno-analyzer-fine-grained 9433This option is intended for analyzer developers. 9434 9435Internally the analyzer builds an ``exploded graph'' that combines 9436control flow graphs with data flow information. 9437 9438By default, an edge in this graph can contain the effects of a run 9439of multiple statements within a basic block. With 9440@option{-fanalyzer-fine-grained}, each statement gets its own edge. 9441 9442@item -fanalyzer-show-duplicate-count 9443@opindex fanalyzer-show-duplicate-count 9444@opindex fno-analyzer-show-duplicate-count 9445This option is intended for analyzer developers: if multiple diagnostics 9446have been detected as being duplicates of each other, it emits a note when 9447reporting the best diagnostic, giving the number of additional diagnostics 9448that were suppressed by the deduplication logic. 9449 9450@item -fno-analyzer-state-merge 9451@opindex fanalyzer-state-merge 9452@opindex fno-analyzer-state-merge 9453This option is intended for analyzer developers. 9454 9455By default the analyzer attempts to simplify analysis by merging 9456sufficiently similar states at each program point as it builds its 9457``exploded graph''. With @option{-fno-analyzer-state-merge} this 9458merging can be suppressed, for debugging state-handling issues. 9459 9460@item -fno-analyzer-state-purge 9461@opindex fanalyzer-state-purge 9462@opindex fno-analyzer-state-purge 9463This option is intended for analyzer developers. 9464 9465By default the analyzer attempts to simplify analysis by purging 9466aspects of state at a program point that appear to no longer be relevant 9467e.g. the values of locals that aren't accessed later in the function 9468and which aren't relevant to leak analysis. 9469 9470With @option{-fno-analyzer-state-purge} this purging of state can 9471be suppressed, for debugging state-handling issues. 9472 9473@item -fanalyzer-transitivity 9474@opindex fanalyzer-transitivity 9475@opindex fno-analyzer-transitivity 9476This option enables transitivity of constraints within the analyzer. 9477 9478@item -fanalyzer-verbose-edges 9479This option is intended for analyzer developers. It enables more 9480verbose, lower-level detail in the descriptions of control flow 9481within diagnostic paths. 9482 9483@item -fanalyzer-verbose-state-changes 9484This option is intended for analyzer developers. It enables more 9485verbose, lower-level detail in the descriptions of events relating 9486to state machines within diagnostic paths. 9487 9488@item -fanalyzer-verbosity=@var{level} 9489This option controls the complexity of the control flow paths that are 9490emitted for analyzer diagnostics. 9491 9492The @var{level} can be one of: 9493 9494@table @samp 9495@item 0 9496At this level, interprocedural call and return events are displayed, 9497along with the most pertinent state-change events relating to 9498a diagnostic. For example, for a double-@code{free} diagnostic, 9499both calls to @code{free} will be shown. 9500 9501@item 1 9502As per the previous level, but also show events for the entry 9503to each function. 9504 9505@item 2 9506As per the previous level, but also show events relating to 9507control flow that are significant to triggering the issue 9508(e.g. ``true path taken'' at a conditional). 9509 9510This level is the default. 9511 9512@item 3 9513As per the previous level, but show all control flow events, not 9514just significant ones. 9515 9516@item 4 9517This level is intended for analyzer developers; it adds various 9518other events intended for debugging the analyzer. 9519 9520@end table 9521 9522@item -fdump-analyzer 9523@opindex fdump-analyzer 9524Dump internal details about what the analyzer is doing to 9525@file{@var{file}.analyzer.txt}. 9526This option is overridden by @option{-fdump-analyzer-stderr}. 9527 9528@item -fdump-analyzer-stderr 9529@opindex fdump-analyzer-stderr 9530Dump internal details about what the analyzer is doing to stderr. 9531This option overrides @option{-fdump-analyzer}. 9532 9533@item -fdump-analyzer-callgraph 9534@opindex fdump-analyzer-callgraph 9535Dump a representation of the call graph suitable for viewing with 9536GraphViz to @file{@var{file}.callgraph.dot}. 9537 9538@item -fdump-analyzer-exploded-graph 9539@opindex fdump-analyzer-exploded-graph 9540Dump a representation of the ``exploded graph'' suitable for viewing with 9541GraphViz to @file{@var{file}.eg.dot}. 9542Nodes are color-coded based on state-machine states to emphasize 9543state changes. 9544 9545@item -fdump-analyzer-exploded-nodes 9546@opindex dump-analyzer-exploded-nodes 9547Emit diagnostics showing where nodes in the ``exploded graph'' are 9548in relation to the program source. 9549 9550@item -fdump-analyzer-exploded-nodes-2 9551@opindex dump-analyzer-exploded-nodes-2 9552Dump a textual representation of the ``exploded graph'' to 9553@file{@var{file}.eg.txt}. 9554 9555@item -fdump-analyzer-exploded-nodes-3 9556@opindex dump-analyzer-exploded-nodes-3 9557Dump a textual representation of the ``exploded graph'' to 9558one dump file per node, to @file{@var{file}.eg-@var{id}.txt}. 9559This is typically a large number of dump files. 9560 9561@item -fdump-analyzer-feasibility 9562@opindex dump-analyzer-feasibility 9563Dump internal details about the analyzer's search for feasible paths. 9564The details are written in a form suitable for viewing with GraphViz 9565to filenames of the form @file{@var{file}.*.fg.dot} and 9566@file{@var{file}.*.tg.dot}. 9567 9568@item -fdump-analyzer-json 9569@opindex fdump-analyzer-json 9570Dump a compressed JSON representation of analyzer internals to 9571@file{@var{file}.analyzer.json.gz}. The precise format is subject 9572to change. 9573 9574@item -fdump-analyzer-state-purge 9575@opindex fdump-analyzer-state-purge 9576As per @option{-fdump-analyzer-supergraph}, dump a representation of the 9577``supergraph'' suitable for viewing with GraphViz, but annotate the 9578graph with information on what state will be purged at each node. 9579The graph is written to @file{@var{file}.state-purge.dot}. 9580 9581@item -fdump-analyzer-supergraph 9582@opindex fdump-analyzer-supergraph 9583Dump representations of the ``supergraph'' suitable for viewing with 9584GraphViz to @file{@var{file}.supergraph.dot} and to 9585@file{@var{file}.supergraph-eg.dot}. These show all of the 9586control flow graphs in the program, with interprocedural edges for 9587calls and returns. The second dump contains annotations showing nodes 9588in the ``exploded graph'' and diagnostics associated with them. 9589 9590@end table 9591 9592@node Debugging Options 9593@section Options for Debugging Your Program 9594@cindex options, debugging 9595@cindex debugging information options 9596 9597To tell GCC to emit extra information for use by a debugger, in almost 9598all cases you need only to add @option{-g} to your other options. 9599 9600GCC allows you to use @option{-g} with 9601@option{-O}. The shortcuts taken by optimized code may occasionally 9602be surprising: some variables you declared may not exist 9603at all; flow of control may briefly move where you did not expect it; 9604some statements may not be executed because they compute constant 9605results or their values are already at hand; some statements may 9606execute in different places because they have been moved out of loops. 9607Nevertheless it is possible to debug optimized output. This makes 9608it reasonable to use the optimizer for programs that might have bugs. 9609 9610If you are not using some other optimization option, consider 9611using @option{-Og} (@pxref{Optimize Options}) with @option{-g}. 9612With no @option{-O} option at all, some compiler passes that collect 9613information useful for debugging do not run at all, so that 9614@option{-Og} may result in a better debugging experience. 9615 9616@table @gcctabopt 9617@item -g 9618@opindex g 9619Produce debugging information in the operating system's native format 9620(stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging 9621information. 9622 9623On most systems that use stabs format, @option{-g} enables use of extra 9624debugging information that only GDB can use; this extra information 9625makes debugging work better in GDB but probably makes other debuggers 9626crash or 9627refuse to read the program. If you want to control for certain whether 9628to generate the extra information, use @option{-gstabs+}, @option{-gstabs}, 9629@option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below). 9630 9631@item -ggdb 9632@opindex ggdb 9633Produce debugging information for use by GDB@. This means to use the 9634most expressive format available (DWARF, stabs, or the native format 9635if neither of those are supported), including GDB extensions if at all 9636possible. 9637 9638@item -gdwarf 9639@itemx -gdwarf-@var{version} 9640@opindex gdwarf 9641Produce debugging information in DWARF format (if that is supported). 9642The value of @var{version} may be either 2, 3, 4 or 5; the default 9643version for most targets is 5 (with the exception of VxWorks, TPF and 9644Darwin/Mac OS X, which default to version 2, and AIX, which defaults 9645to version 4). 9646 9647Note that with DWARF Version 2, some ports require and always 9648use some non-conflicting DWARF 3 extensions in the unwind tables. 9649 9650Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments} 9651for maximum benefit. Version 5 requires GDB 8.0 or higher. 9652 9653GCC no longer supports DWARF Version 1, which is substantially 9654different than Version 2 and later. For historical reasons, some 9655other DWARF-related options such as 9656@option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2 9657in their names, but apply to all currently-supported versions of DWARF. 9658 9659@item -gstabs 9660@opindex gstabs 9661Produce debugging information in stabs format (if that is supported), 9662without GDB extensions. This is the format used by DBX on most BSD 9663systems. On MIPS, Alpha and System V Release 4 systems this option 9664produces stabs debugging output that is not understood by DBX@. 9665On System V Release 4 systems this option requires the GNU assembler. 9666 9667@item -gstabs+ 9668@opindex gstabs+ 9669Produce debugging information in stabs format (if that is supported), 9670using GNU extensions understood only by the GNU debugger (GDB)@. The 9671use of these extensions is likely to make other debuggers crash or 9672refuse to read the program. 9673 9674@item -gxcoff 9675@opindex gxcoff 9676Produce debugging information in XCOFF format (if that is supported). 9677This is the format used by the DBX debugger on IBM RS/6000 systems. 9678 9679@item -gxcoff+ 9680@opindex gxcoff+ 9681Produce debugging information in XCOFF format (if that is supported), 9682using GNU extensions understood only by the GNU debugger (GDB)@. The 9683use of these extensions is likely to make other debuggers crash or 9684refuse to read the program, and may cause assemblers other than the GNU 9685assembler (GAS) to fail with an error. 9686 9687@item -gvms 9688@opindex gvms 9689Produce debugging information in Alpha/VMS debug format (if that is 9690supported). This is the format used by DEBUG on Alpha/VMS systems. 9691 9692@item -g@var{level} 9693@itemx -ggdb@var{level} 9694@itemx -gstabs@var{level} 9695@itemx -gxcoff@var{level} 9696@itemx -gvms@var{level} 9697Request debugging information and also use @var{level} to specify how 9698much information. The default level is 2. 9699 9700Level 0 produces no debug information at all. Thus, @option{-g0} negates 9701@option{-g}. 9702 9703Level 1 produces minimal information, enough for making backtraces in 9704parts of the program that you don't plan to debug. This includes 9705descriptions of functions and external variables, and line number 9706tables, but no information about local variables. 9707 9708Level 3 includes extra information, such as all the macro definitions 9709present in the program. Some debuggers support macro expansion when 9710you use @option{-g3}. 9711 9712If you use multiple @option{-g} options, with or without level numbers, 9713the last such option is the one that is effective. 9714 9715@option{-gdwarf} does not accept a concatenated debug level, to avoid 9716confusion with @option{-gdwarf-@var{level}}. 9717Instead use an additional @option{-g@var{level}} option to change the 9718debug level for DWARF. 9719 9720@item -fno-eliminate-unused-debug-symbols 9721@opindex feliminate-unused-debug-symbols 9722@opindex fno-eliminate-unused-debug-symbols 9723By default, no debug information is produced for symbols that are not actually 9724used. Use this option if you want debug information for all symbols. 9725 9726@item -femit-class-debug-always 9727@opindex femit-class-debug-always 9728Instead of emitting debugging information for a C++ class in only one 9729object file, emit it in all object files using the class. This option 9730should be used only with debuggers that are unable to handle the way GCC 9731normally emits debugging information for classes because using this 9732option increases the size of debugging information by as much as a 9733factor of two. 9734 9735@item -fno-merge-debug-strings 9736@opindex fmerge-debug-strings 9737@opindex fno-merge-debug-strings 9738Direct the linker to not merge together strings in the debugging 9739information that are identical in different object files. Merging is 9740not supported by all assemblers or linkers. Merging decreases the size 9741of the debug information in the output file at the cost of increasing 9742link processing time. Merging is enabled by default. 9743 9744@item -fdebug-prefix-map=@var{old}=@var{new} 9745@opindex fdebug-prefix-map 9746When compiling files residing in directory @file{@var{old}}, record 9747debugging information describing them as if the files resided in 9748directory @file{@var{new}} instead. This can be used to replace a 9749build-time path with an install-time path in the debug info. It can 9750also be used to change an absolute path to a relative path by using 9751@file{.} for @var{new}. This can give more reproducible builds, which 9752are location independent, but may require an extra command to tell GDB 9753where to find the source files. See also @option{-ffile-prefix-map}. 9754 9755@item -fvar-tracking 9756@opindex fvar-tracking 9757Run variable tracking pass. It computes where variables are stored at each 9758position in code. Better debugging information is then generated 9759(if the debugging information format supports this information). 9760 9761It is enabled by default when compiling with optimization (@option{-Os}, 9762@option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and 9763the debug info format supports it. 9764 9765@item -fvar-tracking-assignments 9766@opindex fvar-tracking-assignments 9767@opindex fno-var-tracking-assignments 9768Annotate assignments to user variables early in the compilation and 9769attempt to carry the annotations over throughout the compilation all the 9770way to the end, in an attempt to improve debug information while 9771optimizing. Use of @option{-gdwarf-4} is recommended along with it. 9772 9773It can be enabled even if var-tracking is disabled, in which case 9774annotations are created and maintained, but discarded at the end. 9775By default, this flag is enabled together with @option{-fvar-tracking}, 9776except when selective scheduling is enabled. 9777 9778@item -gsplit-dwarf 9779@opindex gsplit-dwarf 9780If DWARF debugging information is enabled, separate as much debugging 9781information as possible into a separate output file with the extension 9782@file{.dwo}. This option allows the build system to avoid linking files with 9783debug information. To be useful, this option requires a debugger capable of 9784reading @file{.dwo} files. 9785 9786@item -gdwarf32 9787@itemx -gdwarf64 9788@opindex gdwarf32 9789@opindex gdwarf64 9790If DWARF debugging information is enabled, the @option{-gdwarf32} selects 9791the 32-bit DWARF format and the @option{-gdwarf64} selects the 64-bit 9792DWARF format. The default is target specific, on most targets it is 9793@option{-gdwarf32} though. The 32-bit DWARF format is smaller, but 9794can't support more than 2GiB of debug information in any of the DWARF 9795debug information sections. The 64-bit DWARF format allows larger debug 9796information and might not be well supported by all consumers yet. 9797 9798@item -gdescribe-dies 9799@opindex gdescribe-dies 9800Add description attributes to some DWARF DIEs that have no name attribute, 9801such as artificial variables, external references and call site 9802parameter DIEs. 9803 9804@item -gpubnames 9805@opindex gpubnames 9806Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections. 9807 9808@item -ggnu-pubnames 9809@opindex ggnu-pubnames 9810Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format 9811suitable for conversion into a GDB@ index. This option is only useful 9812with a linker that can produce GDB@ index version 7. 9813 9814@item -fdebug-types-section 9815@opindex fdebug-types-section 9816@opindex fno-debug-types-section 9817When using DWARF Version 4 or higher, type DIEs can be put into 9818their own @code{.debug_types} section instead of making them part of the 9819@code{.debug_info} section. It is more efficient to put them in a separate 9820comdat section since the linker can then remove duplicates. 9821But not all DWARF consumers support @code{.debug_types} sections yet 9822and on some objects @code{.debug_types} produces larger instead of smaller 9823debugging information. 9824 9825@item -grecord-gcc-switches 9826@itemx -gno-record-gcc-switches 9827@opindex grecord-gcc-switches 9828@opindex gno-record-gcc-switches 9829This switch causes the command-line options used to invoke the 9830compiler that may affect code generation to be appended to the 9831DW_AT_producer attribute in DWARF debugging information. The options 9832are concatenated with spaces separating them from each other and from 9833the compiler version. 9834It is enabled by default. 9835See also @option{-frecord-gcc-switches} for another 9836way of storing compiler options into the object file. 9837 9838@item -gstrict-dwarf 9839@opindex gstrict-dwarf 9840Disallow using extensions of later DWARF standard version than selected 9841with @option{-gdwarf-@var{version}}. On most targets using non-conflicting 9842DWARF extensions from later standard versions is allowed. 9843 9844@item -gno-strict-dwarf 9845@opindex gno-strict-dwarf 9846Allow using extensions of later DWARF standard version than selected with 9847@option{-gdwarf-@var{version}}. 9848 9849@item -gas-loc-support 9850@opindex gas-loc-support 9851Inform the compiler that the assembler supports @code{.loc} directives. 9852It may then use them for the assembler to generate DWARF2+ line number 9853tables. 9854 9855This is generally desirable, because assembler-generated line-number 9856tables are a lot more compact than those the compiler can generate 9857itself. 9858 9859This option will be enabled by default if, at GCC configure time, the 9860assembler was found to support such directives. 9861 9862@item -gno-as-loc-support 9863@opindex gno-as-loc-support 9864Force GCC to generate DWARF2+ line number tables internally, if DWARF2+ 9865line number tables are to be generated. 9866 9867@item -gas-locview-support 9868@opindex gas-locview-support 9869Inform the compiler that the assembler supports @code{view} assignment 9870and reset assertion checking in @code{.loc} directives. 9871 9872This option will be enabled by default if, at GCC configure time, the 9873assembler was found to support them. 9874 9875@item -gno-as-locview-support 9876Force GCC to assign view numbers internally, if 9877@option{-gvariable-location-views} are explicitly requested. 9878 9879@item -gcolumn-info 9880@itemx -gno-column-info 9881@opindex gcolumn-info 9882@opindex gno-column-info 9883Emit location column information into DWARF debugging information, rather 9884than just file and line. 9885This option is enabled by default. 9886 9887@item -gstatement-frontiers 9888@itemx -gno-statement-frontiers 9889@opindex gstatement-frontiers 9890@opindex gno-statement-frontiers 9891This option causes GCC to create markers in the internal representation 9892at the beginning of statements, and to keep them roughly in place 9893throughout compilation, using them to guide the output of @code{is_stmt} 9894markers in the line number table. This is enabled by default when 9895compiling with optimization (@option{-Os}, @option{-O}, @option{-O2}, 9896@dots{}), and outputting DWARF 2 debug information at the normal level. 9897 9898@item -gvariable-location-views 9899@itemx -gvariable-location-views=incompat5 9900@itemx -gno-variable-location-views 9901@opindex gvariable-location-views 9902@opindex gvariable-location-views=incompat5 9903@opindex gno-variable-location-views 9904Augment variable location lists with progressive view numbers implied 9905from the line number table. This enables debug information consumers to 9906inspect state at certain points of the program, even if no instructions 9907associated with the corresponding source locations are present at that 9908point. If the assembler lacks support for view numbers in line number 9909tables, this will cause the compiler to emit the line number table, 9910which generally makes them somewhat less compact. The augmented line 9911number tables and location lists are fully backward-compatible, so they 9912can be consumed by debug information consumers that are not aware of 9913these augmentations, but they won't derive any benefit from them either. 9914 9915This is enabled by default when outputting DWARF 2 debug information at 9916the normal level, as long as there is assembler support, 9917@option{-fvar-tracking-assignments} is enabled and 9918@option{-gstrict-dwarf} is not. When assembler support is not 9919available, this may still be enabled, but it will force GCC to output 9920internal line number tables, and if 9921@option{-ginternal-reset-location-views} is not enabled, that will most 9922certainly lead to silently mismatching location views. 9923 9924There is a proposed representation for view numbers that is not backward 9925compatible with the location list format introduced in DWARF 5, that can 9926be enabled with @option{-gvariable-location-views=incompat5}. This 9927option may be removed in the future, is only provided as a reference 9928implementation of the proposed representation. Debug information 9929consumers are not expected to support this extended format, and they 9930would be rendered unable to decode location lists using it. 9931 9932@item -ginternal-reset-location-views 9933@itemx -gno-internal-reset-location-views 9934@opindex ginternal-reset-location-views 9935@opindex gno-internal-reset-location-views 9936Attempt to determine location views that can be omitted from location 9937view lists. This requires the compiler to have very accurate insn 9938length estimates, which isn't always the case, and it may cause 9939incorrect view lists to be generated silently when using an assembler 9940that does not support location view lists. The GNU assembler will flag 9941any such error as a @code{view number mismatch}. This is only enabled 9942on ports that define a reliable estimation function. 9943 9944@item -ginline-points 9945@itemx -gno-inline-points 9946@opindex ginline-points 9947@opindex gno-inline-points 9948Generate extended debug information for inlined functions. Location 9949view tracking markers are inserted at inlined entry points, so that 9950address and view numbers can be computed and output in debug 9951information. This can be enabled independently of location views, in 9952which case the view numbers won't be output, but it can only be enabled 9953along with statement frontiers, and it is only enabled by default if 9954location views are enabled. 9955 9956@item -gz@r{[}=@var{type}@r{]} 9957@opindex gz 9958Produce compressed debug sections in DWARF format, if that is supported. 9959If @var{type} is not given, the default type depends on the capabilities 9960of the assembler and linker used. @var{type} may be one of 9961@samp{none} (don't compress debug sections), @samp{zlib} (use zlib 9962compression in ELF gABI format), or @samp{zlib-gnu} (use zlib 9963compression in traditional GNU format). If the linker doesn't support 9964writing compressed debug sections, the option is rejected. Otherwise, 9965if the assembler does not support them, @option{-gz} is silently ignored 9966when producing object files. 9967 9968@item -femit-struct-debug-baseonly 9969@opindex femit-struct-debug-baseonly 9970Emit debug information for struct-like types 9971only when the base name of the compilation source file 9972matches the base name of file in which the struct is defined. 9973 9974This option substantially reduces the size of debugging information, 9975but at significant potential loss in type information to the debugger. 9976See @option{-femit-struct-debug-reduced} for a less aggressive option. 9977See @option{-femit-struct-debug-detailed} for more detailed control. 9978 9979This option works only with DWARF debug output. 9980 9981@item -femit-struct-debug-reduced 9982@opindex femit-struct-debug-reduced 9983Emit debug information for struct-like types 9984only when the base name of the compilation source file 9985matches the base name of file in which the type is defined, 9986unless the struct is a template or defined in a system header. 9987 9988This option significantly reduces the size of debugging information, 9989with some potential loss in type information to the debugger. 9990See @option{-femit-struct-debug-baseonly} for a more aggressive option. 9991See @option{-femit-struct-debug-detailed} for more detailed control. 9992 9993This option works only with DWARF debug output. 9994 9995@item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} 9996@opindex femit-struct-debug-detailed 9997Specify the struct-like types 9998for which the compiler generates debug information. 9999The intent is to reduce duplicate struct debug information 10000between different object files within the same program. 10001 10002This option is a detailed version of 10003@option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly}, 10004which serves for most needs. 10005 10006A specification has the syntax@* 10007[@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none}) 10008 10009The optional first word limits the specification to 10010structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}). 10011A struct type is used directly when it is the type of a variable, member. 10012Indirect uses arise through pointers to structs. 10013That is, when use of an incomplete struct is valid, the use is indirect. 10014An example is 10015@samp{struct one direct; struct two * indirect;}. 10016 10017The optional second word limits the specification to 10018ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}). 10019Generic structs are a bit complicated to explain. 10020For C++, these are non-explicit specializations of template classes, 10021or non-template classes within the above. 10022Other programming languages have generics, 10023but @option{-femit-struct-debug-detailed} does not yet implement them. 10024 10025The third word specifies the source files for those 10026structs for which the compiler should emit debug information. 10027The values @samp{none} and @samp{any} have the normal meaning. 10028The value @samp{base} means that 10029the base of name of the file in which the type declaration appears 10030must match the base of the name of the main compilation file. 10031In practice, this means that when compiling @file{foo.c}, debug information 10032is generated for types declared in that file and @file{foo.h}, 10033but not other header files. 10034The value @samp{sys} means those types satisfying @samp{base} 10035or declared in system or compiler headers. 10036 10037You may need to experiment to determine the best settings for your application. 10038 10039The default is @option{-femit-struct-debug-detailed=all}. 10040 10041This option works only with DWARF debug output. 10042 10043@item -fno-dwarf2-cfi-asm 10044@opindex fdwarf2-cfi-asm 10045@opindex fno-dwarf2-cfi-asm 10046Emit DWARF unwind info as compiler generated @code{.eh_frame} section 10047instead of using GAS @code{.cfi_*} directives. 10048 10049@item -fno-eliminate-unused-debug-types 10050@opindex feliminate-unused-debug-types 10051@opindex fno-eliminate-unused-debug-types 10052Normally, when producing DWARF output, GCC avoids producing debug symbol 10053output for types that are nowhere used in the source file being compiled. 10054Sometimes it is useful to have GCC emit debugging 10055information for all types declared in a compilation 10056unit, regardless of whether or not they are actually used 10057in that compilation unit, for example 10058if, in the debugger, you want to cast a value to a type that is 10059not actually used in your program (but is declared). More often, 10060however, this results in a significant amount of wasted space. 10061@end table 10062 10063@node Optimize Options 10064@section Options That Control Optimization 10065@cindex optimize options 10066@cindex options, optimization 10067 10068These options control various sorts of optimizations. 10069 10070Without any optimization option, the compiler's goal is to reduce the 10071cost of compilation and to make debugging produce the expected 10072results. Statements are independent: if you stop the program with a 10073breakpoint between statements, you can then assign a new value to any 10074variable or change the program counter to any other statement in the 10075function and get exactly the results you expect from the source 10076code. 10077 10078Turning on optimization flags makes the compiler attempt to improve 10079the performance and/or code size at the expense of compilation time 10080and possibly the ability to debug the program. 10081 10082The compiler performs optimization based on the knowledge it has of the 10083program. Compiling multiple files at once to a single output file mode allows 10084the compiler to use information gained from all of the files when compiling 10085each of them. 10086 10087Not all optimizations are controlled directly by a flag. Only 10088optimizations that have a flag are listed in this section. 10089 10090Most optimizations are completely disabled at @option{-O0} or if an 10091@option{-O} level is not set on the command line, even if individual 10092optimization flags are specified. Similarly, @option{-Og} suppresses 10093many optimization passes. 10094 10095Depending on the target and how GCC was configured, a slightly different 10096set of optimizations may be enabled at each @option{-O} level than 10097those listed here. You can invoke GCC with @option{-Q --help=optimizers} 10098to find out the exact set of optimizations that are enabled at each level. 10099@xref{Overall Options}, for examples. 10100 10101@table @gcctabopt 10102@item -O 10103@itemx -O1 10104@opindex O 10105@opindex O1 10106Optimize. Optimizing compilation takes somewhat more time, and a lot 10107more memory for a large function. 10108 10109With @option{-O}, the compiler tries to reduce code size and execution 10110time, without performing any optimizations that take a great deal of 10111compilation time. 10112 10113@c Note that in addition to the default_options_table list in opts.c, 10114@c several optimization flags default to true but control optimization 10115@c passes that are explicitly disabled at -O0. 10116 10117@option{-O} turns on the following optimization flags: 10118 10119@c Please keep the following list alphabetized. 10120@gccoptlist{-fauto-inc-dec @gol 10121-fbranch-count-reg @gol 10122-fcombine-stack-adjustments @gol 10123-fcompare-elim @gol 10124-fcprop-registers @gol 10125-fdce @gol 10126-fdefer-pop @gol 10127-fdelayed-branch @gol 10128-fdse @gol 10129-fforward-propagate @gol 10130-fguess-branch-probability @gol 10131-fif-conversion @gol 10132-fif-conversion2 @gol 10133-finline-functions-called-once @gol 10134-fipa-modref @gol 10135-fipa-profile @gol 10136-fipa-pure-const @gol 10137-fipa-reference @gol 10138-fipa-reference-addressable @gol 10139-fmerge-constants @gol 10140-fmove-loop-invariants @gol 10141-fomit-frame-pointer @gol 10142-freorder-blocks @gol 10143-fshrink-wrap @gol 10144-fshrink-wrap-separate @gol 10145-fsplit-wide-types @gol 10146-fssa-backprop @gol 10147-fssa-phiopt @gol 10148-ftree-bit-ccp @gol 10149-ftree-ccp @gol 10150-ftree-ch @gol 10151-ftree-coalesce-vars @gol 10152-ftree-copy-prop @gol 10153-ftree-dce @gol 10154-ftree-dominator-opts @gol 10155-ftree-dse @gol 10156-ftree-forwprop @gol 10157-ftree-fre @gol 10158-ftree-phiprop @gol 10159-ftree-pta @gol 10160-ftree-scev-cprop @gol 10161-ftree-sink @gol 10162-ftree-slsr @gol 10163-ftree-sra @gol 10164-ftree-ter @gol 10165-funit-at-a-time} 10166 10167@item -O2 10168@opindex O2 10169Optimize even more. GCC performs nearly all supported optimizations 10170that do not involve a space-speed tradeoff. 10171As compared to @option{-O}, this option increases both compilation time 10172and the performance of the generated code. 10173 10174@option{-O2} turns on all optimization flags specified by @option{-O}. It 10175also turns on the following optimization flags: 10176 10177@c Please keep the following list alphabetized! 10178@gccoptlist{-falign-functions -falign-jumps @gol 10179-falign-labels -falign-loops @gol 10180-fcaller-saves @gol 10181-fcode-hoisting @gol 10182-fcrossjumping @gol 10183-fcse-follow-jumps -fcse-skip-blocks @gol 10184-fdelete-null-pointer-checks @gol 10185-fdevirtualize -fdevirtualize-speculatively @gol 10186-fexpensive-optimizations @gol 10187-ffinite-loops @gol 10188-fgcse -fgcse-lm @gol 10189-fhoist-adjacent-loads @gol 10190-finline-functions @gol 10191-finline-small-functions @gol 10192-findirect-inlining @gol 10193-fipa-bit-cp -fipa-cp -fipa-icf @gol 10194-fipa-ra -fipa-sra -fipa-vrp @gol 10195-fisolate-erroneous-paths-dereference @gol 10196-flra-remat @gol 10197-foptimize-sibling-calls @gol 10198-foptimize-strlen @gol 10199-fpartial-inlining @gol 10200-fpeephole2 @gol 10201-freorder-blocks-algorithm=stc @gol 10202-freorder-blocks-and-partition -freorder-functions @gol 10203-frerun-cse-after-loop @gol 10204-fschedule-insns -fschedule-insns2 @gol 10205-fsched-interblock -fsched-spec @gol 10206-fstore-merging @gol 10207-fstrict-aliasing @gol 10208-fthread-jumps @gol 10209-ftree-builtin-call-dce @gol 10210-ftree-pre @gol 10211-ftree-switch-conversion -ftree-tail-merge @gol 10212-ftree-vrp} 10213 10214Please note the warning under @option{-fgcse} about 10215invoking @option{-O2} on programs that use computed gotos. 10216 10217@item -O3 10218@opindex O3 10219Optimize yet more. @option{-O3} turns on all optimizations specified 10220by @option{-O2} and also turns on the following optimization flags: 10221 10222@c Please keep the following list alphabetized! 10223@gccoptlist{-fgcse-after-reload @gol 10224-fipa-cp-clone 10225-floop-interchange @gol 10226-floop-unroll-and-jam @gol 10227-fpeel-loops @gol 10228-fpredictive-commoning @gol 10229-fsplit-loops @gol 10230-fsplit-paths @gol 10231-ftree-loop-distribution @gol 10232-ftree-loop-vectorize @gol 10233-ftree-partial-pre @gol 10234-ftree-slp-vectorize @gol 10235-funswitch-loops @gol 10236-fvect-cost-model @gol 10237-fvect-cost-model=dynamic @gol 10238-fversion-loops-for-strides} 10239 10240@item -O0 10241@opindex O0 10242Reduce compilation time and make debugging produce the expected 10243results. This is the default. 10244 10245@item -Os 10246@opindex Os 10247Optimize for size. @option{-Os} enables all @option{-O2} optimizations 10248except those that often increase code size: 10249 10250@gccoptlist{-falign-functions -falign-jumps @gol 10251-falign-labels -falign-loops @gol 10252-fprefetch-loop-arrays -freorder-blocks-algorithm=stc} 10253 10254It also enables @option{-finline-functions}, causes the compiler to tune for 10255code size rather than execution speed, and performs further optimizations 10256designed to reduce code size. 10257 10258@item -Ofast 10259@opindex Ofast 10260Disregard strict standards compliance. @option{-Ofast} enables all 10261@option{-O3} optimizations. It also enables optimizations that are not 10262valid for all standard-compliant programs. 10263It turns on @option{-ffast-math}, @option{-fallow-store-data-races} 10264and the Fortran-specific @option{-fstack-arrays}, unless 10265@option{-fmax-stack-var-size} is specified, and @option{-fno-protect-parens}. 10266 10267@item -Og 10268@opindex Og 10269Optimize debugging experience. @option{-Og} should be the optimization 10270level of choice for the standard edit-compile-debug cycle, offering 10271a reasonable level of optimization while maintaining fast compilation 10272and a good debugging experience. It is a better choice than @option{-O0} 10273for producing debuggable code because some compiler passes 10274that collect debug information are disabled at @option{-O0}. 10275 10276Like @option{-O0}, @option{-Og} completely disables a number of 10277optimization passes so that individual options controlling them have 10278no effect. Otherwise @option{-Og} enables all @option{-O1} 10279optimization flags except for those that may interfere with debugging: 10280 10281@gccoptlist{-fbranch-count-reg -fdelayed-branch @gol 10282-fdse -fif-conversion -fif-conversion2 @gol 10283-finline-functions-called-once @gol 10284-fmove-loop-invariants -fssa-phiopt @gol 10285-ftree-bit-ccp -ftree-dse -ftree-pta -ftree-sra} 10286 10287@end table 10288 10289If you use multiple @option{-O} options, with or without level numbers, 10290the last such option is the one that is effective. 10291 10292Options of the form @option{-f@var{flag}} specify machine-independent 10293flags. Most flags have both positive and negative forms; the negative 10294form of @option{-ffoo} is @option{-fno-foo}. In the table 10295below, only one of the forms is listed---the one you typically 10296use. You can figure out the other form by either removing @samp{no-} 10297or adding it. 10298 10299The following options control specific optimizations. They are either 10300activated by @option{-O} options or are related to ones that are. You 10301can use the following flags in the rare cases when ``fine-tuning'' of 10302optimizations to be performed is desired. 10303 10304@table @gcctabopt 10305@item -fno-defer-pop 10306@opindex fno-defer-pop 10307@opindex fdefer-pop 10308For machines that must pop arguments after a function call, always pop 10309the arguments as soon as each function returns. 10310At levels @option{-O1} and higher, @option{-fdefer-pop} is the default; 10311this allows the compiler to let arguments accumulate on the stack for several 10312function calls and pop them all at once. 10313 10314@item -fforward-propagate 10315@opindex fforward-propagate 10316Perform a forward propagation pass on RTL@. The pass tries to combine two 10317instructions and checks if the result can be simplified. If loop unrolling 10318is active, two passes are performed and the second is scheduled after 10319loop unrolling. 10320 10321This option is enabled by default at optimization levels @option{-O}, 10322@option{-O2}, @option{-O3}, @option{-Os}. 10323 10324@item -ffp-contract=@var{style} 10325@opindex ffp-contract 10326@option{-ffp-contract=off} disables floating-point expression contraction. 10327@option{-ffp-contract=fast} enables floating-point expression contraction 10328such as forming of fused multiply-add operations if the target has 10329native support for them. 10330@option{-ffp-contract=on} enables floating-point expression contraction 10331if allowed by the language standard. This is currently not implemented 10332and treated equal to @option{-ffp-contract=off}. 10333 10334The default is @option{-ffp-contract=fast}. 10335 10336@item -fomit-frame-pointer 10337@opindex fomit-frame-pointer 10338Omit the frame pointer in functions that don't need one. This avoids the 10339instructions to save, set up and restore the frame pointer; on many targets 10340it also makes an extra register available. 10341 10342On some targets this flag has no effect because the standard calling sequence 10343always uses a frame pointer, so it cannot be omitted. 10344 10345Note that @option{-fno-omit-frame-pointer} doesn't guarantee the frame pointer 10346is used in all functions. Several targets always omit the frame pointer in 10347leaf functions. 10348 10349Enabled by default at @option{-O} and higher. 10350 10351@item -foptimize-sibling-calls 10352@opindex foptimize-sibling-calls 10353Optimize sibling and tail recursive calls. 10354 10355Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10356 10357@item -foptimize-strlen 10358@opindex foptimize-strlen 10359Optimize various standard C string functions (e.g.@: @code{strlen}, 10360@code{strchr} or @code{strcpy}) and 10361their @code{_FORTIFY_SOURCE} counterparts into faster alternatives. 10362 10363Enabled at levels @option{-O2}, @option{-O3}. 10364 10365@item -fno-inline 10366@opindex fno-inline 10367@opindex finline 10368Do not expand any functions inline apart from those marked with 10369the @code{always_inline} attribute. This is the default when not 10370optimizing. 10371 10372Single functions can be exempted from inlining by marking them 10373with the @code{noinline} attribute. 10374 10375@item -finline-small-functions 10376@opindex finline-small-functions 10377Integrate functions into their callers when their body is smaller than expected 10378function call code (so overall size of program gets smaller). The compiler 10379heuristically decides which functions are simple enough to be worth integrating 10380in this way. This inlining applies to all functions, even those not declared 10381inline. 10382 10383Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10384 10385@item -findirect-inlining 10386@opindex findirect-inlining 10387Inline also indirect calls that are discovered to be known at compile 10388time thanks to previous inlining. This option has any effect only 10389when inlining itself is turned on by the @option{-finline-functions} 10390or @option{-finline-small-functions} options. 10391 10392Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10393 10394@item -finline-functions 10395@opindex finline-functions 10396Consider all functions for inlining, even if they are not declared inline. 10397The compiler heuristically decides which functions are worth integrating 10398in this way. 10399 10400If all calls to a given function are integrated, and the function is 10401declared @code{static}, then the function is normally not output as 10402assembler code in its own right. 10403 10404Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. Also enabled 10405by @option{-fprofile-use} and @option{-fauto-profile}. 10406 10407@item -finline-functions-called-once 10408@opindex finline-functions-called-once 10409Consider all @code{static} functions called once for inlining into their 10410caller even if they are not marked @code{inline}. If a call to a given 10411function is integrated, then the function is not output as assembler code 10412in its own right. 10413 10414Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}, 10415but not @option{-Og}. 10416 10417@item -fearly-inlining 10418@opindex fearly-inlining 10419Inline functions marked by @code{always_inline} and functions whose body seems 10420smaller than the function call overhead early before doing 10421@option{-fprofile-generate} instrumentation and real inlining pass. Doing so 10422makes profiling significantly cheaper and usually inlining faster on programs 10423having large chains of nested wrapper functions. 10424 10425Enabled by default. 10426 10427@item -fipa-sra 10428@opindex fipa-sra 10429Perform interprocedural scalar replacement of aggregates, removal of 10430unused parameters and replacement of parameters passed by reference 10431by parameters passed by value. 10432 10433Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}. 10434 10435@item -finline-limit=@var{n} 10436@opindex finline-limit 10437By default, GCC limits the size of functions that can be inlined. This flag 10438allows coarse control of this limit. @var{n} is the size of functions that 10439can be inlined in number of pseudo instructions. 10440 10441Inlining is actually controlled by a number of parameters, which may be 10442specified individually by using @option{--param @var{name}=@var{value}}. 10443The @option{-finline-limit=@var{n}} option sets some of these parameters 10444as follows: 10445 10446@table @gcctabopt 10447@item max-inline-insns-single 10448is set to @var{n}/2. 10449@item max-inline-insns-auto 10450is set to @var{n}/2. 10451@end table 10452 10453See below for a documentation of the individual 10454parameters controlling inlining and for the defaults of these parameters. 10455 10456@emph{Note:} there may be no value to @option{-finline-limit} that results 10457in default behavior. 10458 10459@emph{Note:} pseudo instruction represents, in this particular context, an 10460abstract measurement of function's size. In no way does it represent a count 10461of assembly instructions and as such its exact meaning might change from one 10462release to an another. 10463 10464@item -fno-keep-inline-dllexport 10465@opindex fno-keep-inline-dllexport 10466@opindex fkeep-inline-dllexport 10467This is a more fine-grained version of @option{-fkeep-inline-functions}, 10468which applies only to functions that are declared using the @code{dllexport} 10469attribute or declspec. @xref{Function Attributes,,Declaring Attributes of 10470Functions}. 10471 10472@item -fkeep-inline-functions 10473@opindex fkeep-inline-functions 10474In C, emit @code{static} functions that are declared @code{inline} 10475into the object file, even if the function has been inlined into all 10476of its callers. This switch does not affect functions using the 10477@code{extern inline} extension in GNU C90@. In C++, emit any and all 10478inline functions into the object file. 10479 10480@item -fkeep-static-functions 10481@opindex fkeep-static-functions 10482Emit @code{static} functions into the object file, even if the function 10483is never used. 10484 10485@item -fkeep-static-consts 10486@opindex fkeep-static-consts 10487Emit variables declared @code{static const} when optimization isn't turned 10488on, even if the variables aren't referenced. 10489 10490GCC enables this option by default. If you want to force the compiler to 10491check if a variable is referenced, regardless of whether or not 10492optimization is turned on, use the @option{-fno-keep-static-consts} option. 10493 10494@item -fmerge-constants 10495@opindex fmerge-constants 10496Attempt to merge identical constants (string constants and floating-point 10497constants) across compilation units. 10498 10499This option is the default for optimized compilation if the assembler and 10500linker support it. Use @option{-fno-merge-constants} to inhibit this 10501behavior. 10502 10503Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 10504 10505@item -fmerge-all-constants 10506@opindex fmerge-all-constants 10507Attempt to merge identical constants and identical variables. 10508 10509This option implies @option{-fmerge-constants}. In addition to 10510@option{-fmerge-constants} this considers e.g.@: even constant initialized 10511arrays or initialized constant variables with integral or floating-point 10512types. Languages like C or C++ require each variable, including multiple 10513instances of the same variable in recursive calls, to have distinct locations, 10514so using this option results in non-conforming 10515behavior. 10516 10517@item -fmodulo-sched 10518@opindex fmodulo-sched 10519Perform swing modulo scheduling immediately before the first scheduling 10520pass. This pass looks at innermost loops and reorders their 10521instructions by overlapping different iterations. 10522 10523@item -fmodulo-sched-allow-regmoves 10524@opindex fmodulo-sched-allow-regmoves 10525Perform more aggressive SMS-based modulo scheduling with register moves 10526allowed. By setting this flag certain anti-dependences edges are 10527deleted, which triggers the generation of reg-moves based on the 10528life-range analysis. This option is effective only with 10529@option{-fmodulo-sched} enabled. 10530 10531@item -fno-branch-count-reg 10532@opindex fno-branch-count-reg 10533@opindex fbranch-count-reg 10534Disable the optimization pass that scans for opportunities to use 10535``decrement and branch'' instructions on a count register instead of 10536instruction sequences that decrement a register, compare it against zero, and 10537then branch based upon the result. This option is only meaningful on 10538architectures that support such instructions, which include x86, PowerPC, 10539IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option 10540doesn't remove the decrement and branch instructions from the generated 10541instruction stream introduced by other optimization passes. 10542 10543The default is @option{-fbranch-count-reg} at @option{-O1} and higher, 10544except for @option{-Og}. 10545 10546@item -fno-function-cse 10547@opindex fno-function-cse 10548@opindex ffunction-cse 10549Do not put function addresses in registers; make each instruction that 10550calls a constant function contain the function's address explicitly. 10551 10552This option results in less efficient code, but some strange hacks 10553that alter the assembler output may be confused by the optimizations 10554performed when this option is not used. 10555 10556The default is @option{-ffunction-cse} 10557 10558@item -fno-zero-initialized-in-bss 10559@opindex fno-zero-initialized-in-bss 10560@opindex fzero-initialized-in-bss 10561If the target supports a BSS section, GCC by default puts variables that 10562are initialized to zero into BSS@. This can save space in the resulting 10563code. 10564 10565This option turns off this behavior because some programs explicitly 10566rely on variables going to the data section---e.g., so that the 10567resulting executable can find the beginning of that section and/or make 10568assumptions based on that. 10569 10570The default is @option{-fzero-initialized-in-bss}. 10571 10572@item -fthread-jumps 10573@opindex fthread-jumps 10574Perform optimizations that check to see if a jump branches to a 10575location where another comparison subsumed by the first is found. If 10576so, the first branch is redirected to either the destination of the 10577second branch or a point immediately following it, depending on whether 10578the condition is known to be true or false. 10579 10580Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10581 10582@item -fsplit-wide-types 10583@opindex fsplit-wide-types 10584When using a type that occupies multiple registers, such as @code{long 10585long} on a 32-bit system, split the registers apart and allocate them 10586independently. This normally generates better code for those types, 10587but may make debugging more difficult. 10588 10589Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, 10590@option{-Os}. 10591 10592@item -fsplit-wide-types-early 10593@opindex fsplit-wide-types-early 10594Fully split wide types early, instead of very late. 10595This option has no effect unless @option{-fsplit-wide-types} is turned on. 10596 10597This is the default on some targets. 10598 10599@item -fcse-follow-jumps 10600@opindex fcse-follow-jumps 10601In common subexpression elimination (CSE), scan through jump instructions 10602when the target of the jump is not reached by any other path. For 10603example, when CSE encounters an @code{if} statement with an 10604@code{else} clause, CSE follows the jump when the condition 10605tested is false. 10606 10607Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10608 10609@item -fcse-skip-blocks 10610@opindex fcse-skip-blocks 10611This is similar to @option{-fcse-follow-jumps}, but causes CSE to 10612follow jumps that conditionally skip over blocks. When CSE 10613encounters a simple @code{if} statement with no else clause, 10614@option{-fcse-skip-blocks} causes CSE to follow the jump around the 10615body of the @code{if}. 10616 10617Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10618 10619@item -frerun-cse-after-loop 10620@opindex frerun-cse-after-loop 10621Re-run common subexpression elimination after loop optimizations are 10622performed. 10623 10624Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10625 10626@item -fgcse 10627@opindex fgcse 10628Perform a global common subexpression elimination pass. 10629This pass also performs global constant and copy propagation. 10630 10631@emph{Note:} When compiling a program using computed gotos, a GCC 10632extension, you may get better run-time performance if you disable 10633the global common subexpression elimination pass by adding 10634@option{-fno-gcse} to the command line. 10635 10636Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10637 10638@item -fgcse-lm 10639@opindex fgcse-lm 10640When @option{-fgcse-lm} is enabled, global common subexpression elimination 10641attempts to move loads that are only killed by stores into themselves. This 10642allows a loop containing a load/store sequence to be changed to a load outside 10643the loop, and a copy/store within the loop. 10644 10645Enabled by default when @option{-fgcse} is enabled. 10646 10647@item -fgcse-sm 10648@opindex fgcse-sm 10649When @option{-fgcse-sm} is enabled, a store motion pass is run after 10650global common subexpression elimination. This pass attempts to move 10651stores out of loops. When used in conjunction with @option{-fgcse-lm}, 10652loops containing a load/store sequence can be changed to a load before 10653the loop and a store after the loop. 10654 10655Not enabled at any optimization level. 10656 10657@item -fgcse-las 10658@opindex fgcse-las 10659When @option{-fgcse-las} is enabled, the global common subexpression 10660elimination pass eliminates redundant loads that come after stores to the 10661same memory location (both partial and full redundancies). 10662 10663Not enabled at any optimization level. 10664 10665@item -fgcse-after-reload 10666@opindex fgcse-after-reload 10667When @option{-fgcse-after-reload} is enabled, a redundant load elimination 10668pass is performed after reload. The purpose of this pass is to clean up 10669redundant spilling. 10670 10671Enabled by @option{-fprofile-use} and @option{-fauto-profile}. 10672 10673@item -faggressive-loop-optimizations 10674@opindex faggressive-loop-optimizations 10675This option tells the loop optimizer to use language constraints to 10676derive bounds for the number of iterations of a loop. This assumes that 10677loop code does not invoke undefined behavior by for example causing signed 10678integer overflows or out-of-bound array accesses. The bounds for the 10679number of iterations of a loop are used to guide loop unrolling and peeling 10680and loop exit test optimizations. 10681This option is enabled by default. 10682 10683@item -funconstrained-commons 10684@opindex funconstrained-commons 10685This option tells the compiler that variables declared in common blocks 10686(e.g.@: Fortran) may later be overridden with longer trailing arrays. This 10687prevents certain optimizations that depend on knowing the array bounds. 10688 10689@item -fcrossjumping 10690@opindex fcrossjumping 10691Perform cross-jumping transformation. 10692This transformation unifies equivalent code and saves code size. The 10693resulting code may or may not perform better than without cross-jumping. 10694 10695Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10696 10697@item -fauto-inc-dec 10698@opindex fauto-inc-dec 10699Combine increments or decrements of addresses with memory accesses. 10700This pass is always skipped on architectures that do not have 10701instructions to support this. Enabled by default at @option{-O} and 10702higher on architectures that support this. 10703 10704@item -fdce 10705@opindex fdce 10706Perform dead code elimination (DCE) on RTL@. 10707Enabled by default at @option{-O} and higher. 10708 10709@item -fdse 10710@opindex fdse 10711Perform dead store elimination (DSE) on RTL@. 10712Enabled by default at @option{-O} and higher. 10713 10714@item -fif-conversion 10715@opindex fif-conversion 10716Attempt to transform conditional jumps into branch-less equivalents. This 10717includes use of conditional moves, min, max, set flags and abs instructions, and 10718some tricks doable by standard arithmetics. The use of conditional execution 10719on chips where it is available is controlled by @option{-fif-conversion2}. 10720 10721Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}, but 10722not with @option{-Og}. 10723 10724@item -fif-conversion2 10725@opindex fif-conversion2 10726Use conditional execution (where available) to transform conditional jumps into 10727branch-less equivalents. 10728 10729Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}, but 10730not with @option{-Og}. 10731 10732@item -fdeclone-ctor-dtor 10733@opindex fdeclone-ctor-dtor 10734The C++ ABI requires multiple entry points for constructors and 10735destructors: one for a base subobject, one for a complete object, and 10736one for a virtual destructor that calls operator delete afterwards. 10737For a hierarchy with virtual bases, the base and complete variants are 10738clones, which means two copies of the function. With this option, the 10739base and complete variants are changed to be thunks that call a common 10740implementation. 10741 10742Enabled by @option{-Os}. 10743 10744@item -fdelete-null-pointer-checks 10745@opindex fdelete-null-pointer-checks 10746Assume that programs cannot safely dereference null pointers, and that 10747no code or data element resides at address zero. 10748This option enables simple constant 10749folding optimizations at all optimization levels. In addition, other 10750optimization passes in GCC use this flag to control global dataflow 10751analyses that eliminate useless checks for null pointers; these assume 10752that a memory access to address zero always results in a trap, so 10753that if a pointer is checked after it has already been dereferenced, 10754it cannot be null. 10755 10756Note however that in some environments this assumption is not true. 10757Use @option{-fno-delete-null-pointer-checks} to disable this optimization 10758for programs that depend on that behavior. 10759 10760This option is enabled by default on most targets. On Nios II ELF, it 10761defaults to off. On AVR, CR16, and MSP430, this option is completely disabled. 10762 10763Passes that use the dataflow information 10764are enabled independently at different optimization levels. 10765 10766@item -fdevirtualize 10767@opindex fdevirtualize 10768Attempt to convert calls to virtual functions to direct calls. This 10769is done both within a procedure and interprocedurally as part of 10770indirect inlining (@option{-findirect-inlining}) and interprocedural constant 10771propagation (@option{-fipa-cp}). 10772Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10773 10774@item -fdevirtualize-speculatively 10775@opindex fdevirtualize-speculatively 10776Attempt to convert calls to virtual functions to speculative direct calls. 10777Based on the analysis of the type inheritance graph, determine for a given call 10778the set of likely targets. If the set is small, preferably of size 1, change 10779the call into a conditional deciding between direct and indirect calls. The 10780speculative calls enable more optimizations, such as inlining. When they seem 10781useless after further optimization, they are converted back into original form. 10782 10783@item -fdevirtualize-at-ltrans 10784@opindex fdevirtualize-at-ltrans 10785Stream extra information needed for aggressive devirtualization when running 10786the link-time optimizer in local transformation mode. 10787This option enables more devirtualization but 10788significantly increases the size of streamed data. For this reason it is 10789disabled by default. 10790 10791@item -fexpensive-optimizations 10792@opindex fexpensive-optimizations 10793Perform a number of minor optimizations that are relatively expensive. 10794 10795Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10796 10797@item -free 10798@opindex free 10799Attempt to remove redundant extension instructions. This is especially 10800helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit 10801registers after writing to their lower 32-bit half. 10802 10803Enabled for Alpha, AArch64 and x86 at levels @option{-O2}, 10804@option{-O3}, @option{-Os}. 10805 10806@item -fno-lifetime-dse 10807@opindex fno-lifetime-dse 10808@opindex flifetime-dse 10809In C++ the value of an object is only affected by changes within its 10810lifetime: when the constructor begins, the object has an indeterminate 10811value, and any changes during the lifetime of the object are dead when 10812the object is destroyed. Normally dead store elimination will take 10813advantage of this; if your code relies on the value of the object 10814storage persisting beyond the lifetime of the object, you can use this 10815flag to disable this optimization. To preserve stores before the 10816constructor starts (e.g.@: because your operator new clears the object 10817storage) but still treat the object as dead after the destructor, you 10818can use @option{-flifetime-dse=1}. The default behavior can be 10819explicitly selected with @option{-flifetime-dse=2}. 10820@option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}. 10821 10822@item -flive-range-shrinkage 10823@opindex flive-range-shrinkage 10824Attempt to decrease register pressure through register live range 10825shrinkage. This is helpful for fast processors with small or moderate 10826size register sets. 10827 10828@item -fira-algorithm=@var{algorithm} 10829@opindex fira-algorithm 10830Use the specified coloring algorithm for the integrated register 10831allocator. The @var{algorithm} argument can be @samp{priority}, which 10832specifies Chow's priority coloring, or @samp{CB}, which specifies 10833Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented 10834for all architectures, but for those targets that do support it, it is 10835the default because it generates better code. 10836 10837@item -fira-region=@var{region} 10838@opindex fira-region 10839Use specified regions for the integrated register allocator. The 10840@var{region} argument should be one of the following: 10841 10842@table @samp 10843 10844@item all 10845Use all loops as register allocation regions. 10846This can give the best results for machines with a small and/or 10847irregular register set. 10848 10849@item mixed 10850Use all loops except for loops with small register pressure 10851as the regions. This value usually gives 10852the best results in most cases and for most architectures, 10853and is enabled by default when compiling with optimization for speed 10854(@option{-O}, @option{-O2}, @dots{}). 10855 10856@item one 10857Use all functions as a single region. 10858This typically results in the smallest code size, and is enabled by default for 10859@option{-Os} or @option{-O0}. 10860 10861@end table 10862 10863@item -fira-hoist-pressure 10864@opindex fira-hoist-pressure 10865Use IRA to evaluate register pressure in the code hoisting pass for 10866decisions to hoist expressions. This option usually results in smaller 10867code, but it can slow the compiler down. 10868 10869This option is enabled at level @option{-Os} for all targets. 10870 10871@item -fira-loop-pressure 10872@opindex fira-loop-pressure 10873Use IRA to evaluate register pressure in loops for decisions to move 10874loop invariants. This option usually results in generation 10875of faster and smaller code on machines with large register files (>= 32 10876registers), but it can slow the compiler down. 10877 10878This option is enabled at level @option{-O3} for some targets. 10879 10880@item -fno-ira-share-save-slots 10881@opindex fno-ira-share-save-slots 10882@opindex fira-share-save-slots 10883Disable sharing of stack slots used for saving call-used hard 10884registers living through a call. Each hard register gets a 10885separate stack slot, and as a result function stack frames are 10886larger. 10887 10888@item -fno-ira-share-spill-slots 10889@opindex fno-ira-share-spill-slots 10890@opindex fira-share-spill-slots 10891Disable sharing of stack slots allocated for pseudo-registers. Each 10892pseudo-register that does not get a hard register gets a separate 10893stack slot, and as a result function stack frames are larger. 10894 10895@item -flra-remat 10896@opindex flra-remat 10897Enable CFG-sensitive rematerialization in LRA. Instead of loading 10898values of spilled pseudos, LRA tries to rematerialize (recalculate) 10899values if it is profitable. 10900 10901Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10902 10903@item -fdelayed-branch 10904@opindex fdelayed-branch 10905If supported for the target machine, attempt to reorder instructions 10906to exploit instruction slots available after delayed branch 10907instructions. 10908 10909Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}, 10910but not at @option{-Og}. 10911 10912@item -fschedule-insns 10913@opindex fschedule-insns 10914If supported for the target machine, attempt to reorder instructions to 10915eliminate execution stalls due to required data being unavailable. This 10916helps machines that have slow floating point or memory load instructions 10917by allowing other instructions to be issued until the result of the load 10918or floating-point instruction is required. 10919 10920Enabled at levels @option{-O2}, @option{-O3}. 10921 10922@item -fschedule-insns2 10923@opindex fschedule-insns2 10924Similar to @option{-fschedule-insns}, but requests an additional pass of 10925instruction scheduling after register allocation has been done. This is 10926especially useful on machines with a relatively small number of 10927registers and where memory load instructions take more than one cycle. 10928 10929Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 10930 10931@item -fno-sched-interblock 10932@opindex fno-sched-interblock 10933@opindex fsched-interblock 10934Disable instruction scheduling across basic blocks, which 10935is normally enabled when scheduling before register allocation, i.e.@: 10936with @option{-fschedule-insns} or at @option{-O2} or higher. 10937 10938@item -fno-sched-spec 10939@opindex fno-sched-spec 10940@opindex fsched-spec 10941Disable speculative motion of non-load instructions, which 10942is normally enabled when scheduling before register allocation, i.e.@: 10943with @option{-fschedule-insns} or at @option{-O2} or higher. 10944 10945@item -fsched-pressure 10946@opindex fsched-pressure 10947Enable register pressure sensitive insn scheduling before register 10948allocation. This only makes sense when scheduling before register 10949allocation is enabled, i.e.@: with @option{-fschedule-insns} or at 10950@option{-O2} or higher. Usage of this option can improve the 10951generated code and decrease its size by preventing register pressure 10952increase above the number of available hard registers and subsequent 10953spills in register allocation. 10954 10955@item -fsched-spec-load 10956@opindex fsched-spec-load 10957Allow speculative motion of some load instructions. This only makes 10958sense when scheduling before register allocation, i.e.@: with 10959@option{-fschedule-insns} or at @option{-O2} or higher. 10960 10961@item -fsched-spec-load-dangerous 10962@opindex fsched-spec-load-dangerous 10963Allow speculative motion of more load instructions. This only makes 10964sense when scheduling before register allocation, i.e.@: with 10965@option{-fschedule-insns} or at @option{-O2} or higher. 10966 10967@item -fsched-stalled-insns 10968@itemx -fsched-stalled-insns=@var{n} 10969@opindex fsched-stalled-insns 10970Define how many insns (if any) can be moved prematurely from the queue 10971of stalled insns into the ready list during the second scheduling pass. 10972@option{-fno-sched-stalled-insns} means that no insns are moved 10973prematurely, @option{-fsched-stalled-insns=0} means there is no limit 10974on how many queued insns can be moved prematurely. 10975@option{-fsched-stalled-insns} without a value is equivalent to 10976@option{-fsched-stalled-insns=1}. 10977 10978@item -fsched-stalled-insns-dep 10979@itemx -fsched-stalled-insns-dep=@var{n} 10980@opindex fsched-stalled-insns-dep 10981Define how many insn groups (cycles) are examined for a dependency 10982on a stalled insn that is a candidate for premature removal from the queue 10983of stalled insns. This has an effect only during the second scheduling pass, 10984and only if @option{-fsched-stalled-insns} is used. 10985@option{-fno-sched-stalled-insns-dep} is equivalent to 10986@option{-fsched-stalled-insns-dep=0}. 10987@option{-fsched-stalled-insns-dep} without a value is equivalent to 10988@option{-fsched-stalled-insns-dep=1}. 10989 10990@item -fsched2-use-superblocks 10991@opindex fsched2-use-superblocks 10992When scheduling after register allocation, use superblock scheduling. 10993This allows motion across basic block boundaries, 10994resulting in faster schedules. This option is experimental, as not all machine 10995descriptions used by GCC model the CPU closely enough to avoid unreliable 10996results from the algorithm. 10997 10998This only makes sense when scheduling after register allocation, i.e.@: with 10999@option{-fschedule-insns2} or at @option{-O2} or higher. 11000 11001@item -fsched-group-heuristic 11002@opindex fsched-group-heuristic 11003Enable the group heuristic in the scheduler. This heuristic favors 11004the instruction that belongs to a schedule group. This is enabled 11005by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns} 11006or @option{-fschedule-insns2} or at @option{-O2} or higher. 11007 11008@item -fsched-critical-path-heuristic 11009@opindex fsched-critical-path-heuristic 11010Enable the critical-path heuristic in the scheduler. This heuristic favors 11011instructions on the critical path. This is enabled by default when 11012scheduling is enabled, i.e.@: with @option{-fschedule-insns} 11013or @option{-fschedule-insns2} or at @option{-O2} or higher. 11014 11015@item -fsched-spec-insn-heuristic 11016@opindex fsched-spec-insn-heuristic 11017Enable the speculative instruction heuristic in the scheduler. This 11018heuristic favors speculative instructions with greater dependency weakness. 11019This is enabled by default when scheduling is enabled, i.e.@: 11020with @option{-fschedule-insns} or @option{-fschedule-insns2} 11021or at @option{-O2} or higher. 11022 11023@item -fsched-rank-heuristic 11024@opindex fsched-rank-heuristic 11025Enable the rank heuristic in the scheduler. This heuristic favors 11026the instruction belonging to a basic block with greater size or frequency. 11027This is enabled by default when scheduling is enabled, i.e.@: 11028with @option{-fschedule-insns} or @option{-fschedule-insns2} or 11029at @option{-O2} or higher. 11030 11031@item -fsched-last-insn-heuristic 11032@opindex fsched-last-insn-heuristic 11033Enable the last-instruction heuristic in the scheduler. This heuristic 11034favors the instruction that is less dependent on the last instruction 11035scheduled. This is enabled by default when scheduling is enabled, 11036i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or 11037at @option{-O2} or higher. 11038 11039@item -fsched-dep-count-heuristic 11040@opindex fsched-dep-count-heuristic 11041Enable the dependent-count heuristic in the scheduler. This heuristic 11042favors the instruction that has more instructions depending on it. 11043This is enabled by default when scheduling is enabled, i.e.@: 11044with @option{-fschedule-insns} or @option{-fschedule-insns2} or 11045at @option{-O2} or higher. 11046 11047@item -freschedule-modulo-scheduled-loops 11048@opindex freschedule-modulo-scheduled-loops 11049Modulo scheduling is performed before traditional scheduling. If a loop 11050is modulo scheduled, later scheduling passes may change its schedule. 11051Use this option to control that behavior. 11052 11053@item -fselective-scheduling 11054@opindex fselective-scheduling 11055Schedule instructions using selective scheduling algorithm. Selective 11056scheduling runs instead of the first scheduler pass. 11057 11058@item -fselective-scheduling2 11059@opindex fselective-scheduling2 11060Schedule instructions using selective scheduling algorithm. Selective 11061scheduling runs instead of the second scheduler pass. 11062 11063@item -fsel-sched-pipelining 11064@opindex fsel-sched-pipelining 11065Enable software pipelining of innermost loops during selective scheduling. 11066This option has no effect unless one of @option{-fselective-scheduling} or 11067@option{-fselective-scheduling2} is turned on. 11068 11069@item -fsel-sched-pipelining-outer-loops 11070@opindex fsel-sched-pipelining-outer-loops 11071When pipelining loops during selective scheduling, also pipeline outer loops. 11072This option has no effect unless @option{-fsel-sched-pipelining} is turned on. 11073 11074@item -fsemantic-interposition 11075@opindex fsemantic-interposition 11076Some object formats, like ELF, allow interposing of symbols by the 11077dynamic linker. 11078This means that for symbols exported from the DSO, the compiler cannot perform 11079interprocedural propagation, inlining and other optimizations in anticipation 11080that the function or variable in question may change. While this feature is 11081useful, for example, to rewrite memory allocation functions by a debugging 11082implementation, it is expensive in the terms of code quality. 11083With @option{-fno-semantic-interposition} the compiler assumes that 11084if interposition happens for functions the overwriting function will have 11085precisely the same semantics (and side effects). 11086Similarly if interposition happens 11087for variables, the constructor of the variable will be the same. The flag 11088has no effect for functions explicitly declared inline 11089(where it is never allowed for interposition to change semantics) 11090and for symbols explicitly declared weak. 11091 11092@item -fshrink-wrap 11093@opindex fshrink-wrap 11094Emit function prologues only before parts of the function that need it, 11095rather than at the top of the function. This flag is enabled by default at 11096@option{-O} and higher. 11097 11098@item -fshrink-wrap-separate 11099@opindex fshrink-wrap-separate 11100Shrink-wrap separate parts of the prologue and epilogue separately, so that 11101those parts are only executed when needed. 11102This option is on by default, but has no effect unless @option{-fshrink-wrap} 11103is also turned on and the target supports this. 11104 11105@item -fcaller-saves 11106@opindex fcaller-saves 11107Enable allocation of values to registers that are clobbered by 11108function calls, by emitting extra instructions to save and restore the 11109registers around such calls. Such allocation is done only when it 11110seems to result in better code. 11111 11112This option is always enabled by default on certain machines, usually 11113those which have no call-preserved registers to use instead. 11114 11115Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 11116 11117@item -fcombine-stack-adjustments 11118@opindex fcombine-stack-adjustments 11119Tracks stack adjustments (pushes and pops) and stack memory references 11120and then tries to find ways to combine them. 11121 11122Enabled by default at @option{-O1} and higher. 11123 11124@item -fipa-ra 11125@opindex fipa-ra 11126Use caller save registers for allocation if those registers are not used by 11127any called function. In that case it is not necessary to save and restore 11128them around calls. This is only possible if called functions are part of 11129same compilation unit as current function and they are compiled before it. 11130 11131Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option 11132is disabled if generated code will be instrumented for profiling 11133(@option{-p}, or @option{-pg}) or if callee's register usage cannot be known 11134exactly (this happens on targets that do not expose prologues 11135and epilogues in RTL). 11136 11137@item -fconserve-stack 11138@opindex fconserve-stack 11139Attempt to minimize stack usage. The compiler attempts to use less 11140stack space, even if that makes the program slower. This option 11141implies setting the @option{large-stack-frame} parameter to 100 11142and the @option{large-stack-frame-growth} parameter to 400. 11143 11144@item -ftree-reassoc 11145@opindex ftree-reassoc 11146Perform reassociation on trees. This flag is enabled by default 11147at @option{-O} and higher. 11148 11149@item -fcode-hoisting 11150@opindex fcode-hoisting 11151Perform code hoisting. Code hoisting tries to move the 11152evaluation of expressions executed on all paths to the function exit 11153as early as possible. This is especially useful as a code size 11154optimization, but it often helps for code speed as well. 11155This flag is enabled by default at @option{-O2} and higher. 11156 11157@item -ftree-pre 11158@opindex ftree-pre 11159Perform partial redundancy elimination (PRE) on trees. This flag is 11160enabled by default at @option{-O2} and @option{-O3}. 11161 11162@item -ftree-partial-pre 11163@opindex ftree-partial-pre 11164Make partial redundancy elimination (PRE) more aggressive. This flag is 11165enabled by default at @option{-O3}. 11166 11167@item -ftree-forwprop 11168@opindex ftree-forwprop 11169Perform forward propagation on trees. This flag is enabled by default 11170at @option{-O} and higher. 11171 11172@item -ftree-fre 11173@opindex ftree-fre 11174Perform full redundancy elimination (FRE) on trees. The difference 11175between FRE and PRE is that FRE only considers expressions 11176that are computed on all paths leading to the redundant computation. 11177This analysis is faster than PRE, though it exposes fewer redundancies. 11178This flag is enabled by default at @option{-O} and higher. 11179 11180@item -ftree-phiprop 11181@opindex ftree-phiprop 11182Perform hoisting of loads from conditional pointers on trees. This 11183pass is enabled by default at @option{-O} and higher. 11184 11185@item -fhoist-adjacent-loads 11186@opindex fhoist-adjacent-loads 11187Speculatively hoist loads from both branches of an if-then-else if the 11188loads are from adjacent locations in the same structure and the target 11189architecture has a conditional move instruction. This flag is enabled 11190by default at @option{-O2} and higher. 11191 11192@item -ftree-copy-prop 11193@opindex ftree-copy-prop 11194Perform copy propagation on trees. This pass eliminates unnecessary 11195copy operations. This flag is enabled by default at @option{-O} and 11196higher. 11197 11198@item -fipa-pure-const 11199@opindex fipa-pure-const 11200Discover which functions are pure or constant. 11201Enabled by default at @option{-O} and higher. 11202 11203@item -fipa-reference 11204@opindex fipa-reference 11205Discover which static variables do not escape the 11206compilation unit. 11207Enabled by default at @option{-O} and higher. 11208 11209@item -fipa-reference-addressable 11210@opindex fipa-reference-addressable 11211Discover read-only, write-only and non-addressable static variables. 11212Enabled by default at @option{-O} and higher. 11213 11214@item -fipa-stack-alignment 11215@opindex fipa-stack-alignment 11216Reduce stack alignment on call sites if possible. 11217Enabled by default. 11218 11219@item -fipa-pta 11220@opindex fipa-pta 11221Perform interprocedural pointer analysis and interprocedural modification 11222and reference analysis. This option can cause excessive memory and 11223compile-time usage on large compilation units. It is not enabled by 11224default at any optimization level. 11225 11226@item -fipa-profile 11227@opindex fipa-profile 11228Perform interprocedural profile propagation. The functions called only from 11229cold functions are marked as cold. Also functions executed once (such as 11230@code{cold}, @code{noreturn}, static constructors or destructors) are 11231identified. Cold functions and loop less parts of functions executed once are 11232then optimized for size. 11233Enabled by default at @option{-O} and higher. 11234 11235@item -fipa-modref 11236@opindex fipa-modref 11237Perform interprocedural mod/ref analysis. This optimization analyzes the side 11238effects of functions (memory locations that are modified or referenced) and 11239enables better optimization across the function call boundary. This flag is 11240enabled by default at @option{-O} and higher. 11241 11242@item -fipa-cp 11243@opindex fipa-cp 11244Perform interprocedural constant propagation. 11245This optimization analyzes the program to determine when values passed 11246to functions are constants and then optimizes accordingly. 11247This optimization can substantially increase performance 11248if the application has constants passed to functions. 11249This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}. 11250It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}. 11251 11252@item -fipa-cp-clone 11253@opindex fipa-cp-clone 11254Perform function cloning to make interprocedural constant propagation stronger. 11255When enabled, interprocedural constant propagation performs function cloning 11256when externally visible function can be called with constant arguments. 11257Because this optimization can create multiple copies of functions, 11258it may significantly increase code size 11259(see @option{--param ipa-cp-unit-growth=@var{value}}). 11260This flag is enabled by default at @option{-O3}. 11261It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}. 11262 11263@item -fipa-bit-cp 11264@opindex fipa-bit-cp 11265When enabled, perform interprocedural bitwise constant 11266propagation. This flag is enabled by default at @option{-O2} and 11267by @option{-fprofile-use} and @option{-fauto-profile}. 11268It requires that @option{-fipa-cp} is enabled. 11269 11270@item -fipa-vrp 11271@opindex fipa-vrp 11272When enabled, perform interprocedural propagation of value 11273ranges. This flag is enabled by default at @option{-O2}. It requires 11274that @option{-fipa-cp} is enabled. 11275 11276@item -fipa-icf 11277@opindex fipa-icf 11278Perform Identical Code Folding for functions and read-only variables. 11279The optimization reduces code size and may disturb unwind stacks by replacing 11280a function by equivalent one with a different name. The optimization works 11281more effectively with link-time optimization enabled. 11282 11283Although the behavior is similar to the Gold Linker's ICF optimization, GCC ICF 11284works on different levels and thus the optimizations are not same - there are 11285equivalences that are found only by GCC and equivalences found only by Gold. 11286 11287This flag is enabled by default at @option{-O2} and @option{-Os}. 11288 11289@item -flive-patching=@var{level} 11290@opindex flive-patching 11291Control GCC's optimizations to produce output suitable for live-patching. 11292 11293If the compiler's optimization uses a function's body or information extracted 11294from its body to optimize/change another function, the latter is called an 11295impacted function of the former. If a function is patched, its impacted 11296functions should be patched too. 11297 11298The impacted functions are determined by the compiler's interprocedural 11299optimizations. For example, a caller is impacted when inlining a function 11300into its caller, 11301cloning a function and changing its caller to call this new clone, 11302or extracting a function's pureness/constness information to optimize 11303its direct or indirect callers, etc. 11304 11305Usually, the more IPA optimizations enabled, the larger the number of 11306impacted functions for each function. In order to control the number of 11307impacted functions and more easily compute the list of impacted function, 11308IPA optimizations can be partially enabled at two different levels. 11309 11310The @var{level} argument should be one of the following: 11311 11312@table @samp 11313 11314@item inline-clone 11315 11316Only enable inlining and cloning optimizations, which includes inlining, 11317cloning, interprocedural scalar replacement of aggregates and partial inlining. 11318As a result, when patching a function, all its callers and its clones' 11319callers are impacted, therefore need to be patched as well. 11320 11321@option{-flive-patching=inline-clone} disables the following optimization flags: 11322@gccoptlist{-fwhole-program -fipa-pta -fipa-reference -fipa-ra @gol 11323-fipa-icf -fipa-icf-functions -fipa-icf-variables @gol 11324-fipa-bit-cp -fipa-vrp -fipa-pure-const -fipa-reference-addressable @gol 11325-fipa-stack-alignment -fipa-modref} 11326 11327@item inline-only-static 11328 11329Only enable inlining of static functions. 11330As a result, when patching a static function, all its callers are impacted 11331and so need to be patched as well. 11332 11333In addition to all the flags that @option{-flive-patching=inline-clone} 11334disables, 11335@option{-flive-patching=inline-only-static} disables the following additional 11336optimization flags: 11337@gccoptlist{-fipa-cp-clone -fipa-sra -fpartial-inlining -fipa-cp} 11338 11339@end table 11340 11341When @option{-flive-patching} is specified without any value, the default value 11342is @var{inline-clone}. 11343 11344This flag is disabled by default. 11345 11346Note that @option{-flive-patching} is not supported with link-time optimization 11347(@option{-flto}). 11348 11349@item -fisolate-erroneous-paths-dereference 11350@opindex fisolate-erroneous-paths-dereference 11351Detect paths that trigger erroneous or undefined behavior due to 11352dereferencing a null pointer. Isolate those paths from the main control 11353flow and turn the statement with erroneous or undefined behavior into a trap. 11354This flag is enabled by default at @option{-O2} and higher and depends on 11355@option{-fdelete-null-pointer-checks} also being enabled. 11356 11357@item -fisolate-erroneous-paths-attribute 11358@opindex fisolate-erroneous-paths-attribute 11359Detect paths that trigger erroneous or undefined behavior due to a null value 11360being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull} 11361attribute. Isolate those paths from the main control flow and turn the 11362statement with erroneous or undefined behavior into a trap. This is not 11363currently enabled, but may be enabled by @option{-O2} in the future. 11364 11365@item -ftree-sink 11366@opindex ftree-sink 11367Perform forward store motion on trees. This flag is 11368enabled by default at @option{-O} and higher. 11369 11370@item -ftree-bit-ccp 11371@opindex ftree-bit-ccp 11372Perform sparse conditional bit constant propagation on trees and propagate 11373pointer alignment information. 11374This pass only operates on local scalar variables and is enabled by default 11375at @option{-O1} and higher, except for @option{-Og}. 11376It requires that @option{-ftree-ccp} is enabled. 11377 11378@item -ftree-ccp 11379@opindex ftree-ccp 11380Perform sparse conditional constant propagation (CCP) on trees. This 11381pass only operates on local scalar variables and is enabled by default 11382at @option{-O} and higher. 11383 11384@item -fssa-backprop 11385@opindex fssa-backprop 11386Propagate information about uses of a value up the definition chain 11387in order to simplify the definitions. For example, this pass strips 11388sign operations if the sign of a value never matters. The flag is 11389enabled by default at @option{-O} and higher. 11390 11391@item -fssa-phiopt 11392@opindex fssa-phiopt 11393Perform pattern matching on SSA PHI nodes to optimize conditional 11394code. This pass is enabled by default at @option{-O1} and higher, 11395except for @option{-Og}. 11396 11397@item -ftree-switch-conversion 11398@opindex ftree-switch-conversion 11399Perform conversion of simple initializations in a switch to 11400initializations from a scalar array. This flag is enabled by default 11401at @option{-O2} and higher. 11402 11403@item -ftree-tail-merge 11404@opindex ftree-tail-merge 11405Look for identical code sequences. When found, replace one with a jump to the 11406other. This optimization is known as tail merging or cross jumping. This flag 11407is enabled by default at @option{-O2} and higher. The compilation time 11408in this pass can 11409be limited using @option{max-tail-merge-comparisons} parameter and 11410@option{max-tail-merge-iterations} parameter. 11411 11412@item -ftree-dce 11413@opindex ftree-dce 11414Perform dead code elimination (DCE) on trees. This flag is enabled by 11415default at @option{-O} and higher. 11416 11417@item -ftree-builtin-call-dce 11418@opindex ftree-builtin-call-dce 11419Perform conditional dead code elimination (DCE) for calls to built-in functions 11420that may set @code{errno} but are otherwise free of side effects. This flag is 11421enabled by default at @option{-O2} and higher if @option{-Os} is not also 11422specified. 11423 11424@item -ffinite-loops 11425@opindex ffinite-loops 11426@opindex fno-finite-loops 11427Assume that a loop with an exit will eventually take the exit and not loop 11428indefinitely. This allows the compiler to remove loops that otherwise have 11429no side-effects, not considering eventual endless looping as such. 11430 11431This option is enabled by default at @option{-O2} for C++ with -std=c++11 11432or higher. 11433 11434@item -ftree-dominator-opts 11435@opindex ftree-dominator-opts 11436Perform a variety of simple scalar cleanups (constant/copy 11437propagation, redundancy elimination, range propagation and expression 11438simplification) based on a dominator tree traversal. This also 11439performs jump threading (to reduce jumps to jumps). This flag is 11440enabled by default at @option{-O} and higher. 11441 11442@item -ftree-dse 11443@opindex ftree-dse 11444Perform dead store elimination (DSE) on trees. A dead store is a store into 11445a memory location that is later overwritten by another store without 11446any intervening loads. In this case the earlier store can be deleted. This 11447flag is enabled by default at @option{-O} and higher. 11448 11449@item -ftree-ch 11450@opindex ftree-ch 11451Perform loop header copying on trees. This is beneficial since it increases 11452effectiveness of code motion optimizations. It also saves one jump. This flag 11453is enabled by default at @option{-O} and higher. It is not enabled 11454for @option{-Os}, since it usually increases code size. 11455 11456@item -ftree-loop-optimize 11457@opindex ftree-loop-optimize 11458Perform loop optimizations on trees. This flag is enabled by default 11459at @option{-O} and higher. 11460 11461@item -ftree-loop-linear 11462@itemx -floop-strip-mine 11463@itemx -floop-block 11464@opindex ftree-loop-linear 11465@opindex floop-strip-mine 11466@opindex floop-block 11467Perform loop nest optimizations. Same as 11468@option{-floop-nest-optimize}. To use this code transformation, GCC has 11469to be configured with @option{--with-isl} to enable the Graphite loop 11470transformation infrastructure. 11471 11472@item -fgraphite-identity 11473@opindex fgraphite-identity 11474Enable the identity transformation for graphite. For every SCoP we generate 11475the polyhedral representation and transform it back to gimple. Using 11476@option{-fgraphite-identity} we can check the costs or benefits of the 11477GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations 11478are also performed by the code generator isl, like index splitting and 11479dead code elimination in loops. 11480 11481@item -floop-nest-optimize 11482@opindex floop-nest-optimize 11483Enable the isl based loop nest optimizer. This is a generic loop nest 11484optimizer based on the Pluto optimization algorithms. It calculates a loop 11485structure optimized for data-locality and parallelism. This option 11486is experimental. 11487 11488@item -floop-parallelize-all 11489@opindex floop-parallelize-all 11490Use the Graphite data dependence analysis to identify loops that can 11491be parallelized. Parallelize all the loops that can be analyzed to 11492not contain loop carried dependences without checking that it is 11493profitable to parallelize the loops. 11494 11495@item -ftree-coalesce-vars 11496@opindex ftree-coalesce-vars 11497While transforming the program out of the SSA representation, attempt to 11498reduce copying by coalescing versions of different user-defined 11499variables, instead of just compiler temporaries. This may severely 11500limit the ability to debug an optimized program compiled with 11501@option{-fno-var-tracking-assignments}. In the negated form, this flag 11502prevents SSA coalescing of user variables. This option is enabled by 11503default if optimization is enabled, and it does very little otherwise. 11504 11505@item -ftree-loop-if-convert 11506@opindex ftree-loop-if-convert 11507Attempt to transform conditional jumps in the innermost loops to 11508branch-less equivalents. The intent is to remove control-flow from 11509the innermost loops in order to improve the ability of the 11510vectorization pass to handle these loops. This is enabled by default 11511if vectorization is enabled. 11512 11513@item -ftree-loop-distribution 11514@opindex ftree-loop-distribution 11515Perform loop distribution. This flag can improve cache performance on 11516big loop bodies and allow further loop optimizations, like 11517parallelization or vectorization, to take place. For example, the loop 11518@smallexample 11519DO I = 1, N 11520 A(I) = B(I) + C 11521 D(I) = E(I) * F 11522ENDDO 11523@end smallexample 11524is transformed to 11525@smallexample 11526DO I = 1, N 11527 A(I) = B(I) + C 11528ENDDO 11529DO I = 1, N 11530 D(I) = E(I) * F 11531ENDDO 11532@end smallexample 11533This flag is enabled by default at @option{-O3}. 11534It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}. 11535 11536@item -ftree-loop-distribute-patterns 11537@opindex ftree-loop-distribute-patterns 11538Perform loop distribution of patterns that can be code generated with 11539calls to a library. This flag is enabled by default at @option{-O2} and 11540higher, and by @option{-fprofile-use} and @option{-fauto-profile}. 11541 11542This pass distributes the initialization loops and generates a call to 11543memset zero. For example, the loop 11544@smallexample 11545DO I = 1, N 11546 A(I) = 0 11547 B(I) = A(I) + I 11548ENDDO 11549@end smallexample 11550is transformed to 11551@smallexample 11552DO I = 1, N 11553 A(I) = 0 11554ENDDO 11555DO I = 1, N 11556 B(I) = A(I) + I 11557ENDDO 11558@end smallexample 11559and the initialization loop is transformed into a call to memset zero. 11560This flag is enabled by default at @option{-O3}. 11561It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}. 11562 11563@item -floop-interchange 11564@opindex floop-interchange 11565Perform loop interchange outside of graphite. This flag can improve cache 11566performance on loop nest and allow further loop optimizations, like 11567vectorization, to take place. For example, the loop 11568@smallexample 11569for (int i = 0; i < N; i++) 11570 for (int j = 0; j < N; j++) 11571 for (int k = 0; k < N; k++) 11572 c[i][j] = c[i][j] + a[i][k]*b[k][j]; 11573@end smallexample 11574is transformed to 11575@smallexample 11576for (int i = 0; i < N; i++) 11577 for (int k = 0; k < N; k++) 11578 for (int j = 0; j < N; j++) 11579 c[i][j] = c[i][j] + a[i][k]*b[k][j]; 11580@end smallexample 11581This flag is enabled by default at @option{-O3}. 11582It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}. 11583 11584@item -floop-unroll-and-jam 11585@opindex floop-unroll-and-jam 11586Apply unroll and jam transformations on feasible loops. In a loop 11587nest this unrolls the outer loop by some factor and fuses the resulting 11588multiple inner loops. This flag is enabled by default at @option{-O3}. 11589It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}. 11590 11591@item -ftree-loop-im 11592@opindex ftree-loop-im 11593Perform loop invariant motion on trees. This pass moves only invariants that 11594are hard to handle at RTL level (function calls, operations that expand to 11595nontrivial sequences of insns). With @option{-funswitch-loops} it also moves 11596operands of conditions that are invariant out of the loop, so that we can use 11597just trivial invariantness analysis in loop unswitching. The pass also includes 11598store motion. 11599 11600@item -ftree-loop-ivcanon 11601@opindex ftree-loop-ivcanon 11602Create a canonical counter for number of iterations in loops for which 11603determining number of iterations requires complicated analysis. Later 11604optimizations then may determine the number easily. Useful especially 11605in connection with unrolling. 11606 11607@item -ftree-scev-cprop 11608@opindex ftree-scev-cprop 11609Perform final value replacement. If a variable is modified in a loop 11610in such a way that its value when exiting the loop can be determined using 11611only its initial value and the number of loop iterations, replace uses of 11612the final value by such a computation, provided it is sufficiently cheap. 11613This reduces data dependencies and may allow further simplifications. 11614Enabled by default at @option{-O} and higher. 11615 11616@item -fivopts 11617@opindex fivopts 11618Perform induction variable optimizations (strength reduction, induction 11619variable merging and induction variable elimination) on trees. 11620 11621@item -ftree-parallelize-loops=n 11622@opindex ftree-parallelize-loops 11623Parallelize loops, i.e., split their iteration space to run in n threads. 11624This is only possible for loops whose iterations are independent 11625and can be arbitrarily reordered. The optimization is only 11626profitable on multiprocessor machines, for loops that are CPU-intensive, 11627rather than constrained e.g.@: by memory bandwidth. This option 11628implies @option{-pthread}, and thus is only supported on targets 11629that have support for @option{-pthread}. 11630 11631@item -ftree-pta 11632@opindex ftree-pta 11633Perform function-local points-to analysis on trees. This flag is 11634enabled by default at @option{-O1} and higher, except for @option{-Og}. 11635 11636@item -ftree-sra 11637@opindex ftree-sra 11638Perform scalar replacement of aggregates. This pass replaces structure 11639references with scalars to prevent committing structures to memory too 11640early. This flag is enabled by default at @option{-O1} and higher, 11641except for @option{-Og}. 11642 11643@item -fstore-merging 11644@opindex fstore-merging 11645Perform merging of narrow stores to consecutive memory addresses. This pass 11646merges contiguous stores of immediate values narrower than a word into fewer 11647wider stores to reduce the number of instructions. This is enabled by default 11648at @option{-O2} and higher as well as @option{-Os}. 11649 11650@item -ftree-ter 11651@opindex ftree-ter 11652Perform temporary expression replacement during the SSA->normal phase. Single 11653use/single def temporaries are replaced at their use location with their 11654defining expression. This results in non-GIMPLE code, but gives the expanders 11655much more complex trees to work on resulting in better RTL generation. This is 11656enabled by default at @option{-O} and higher. 11657 11658@item -ftree-slsr 11659@opindex ftree-slsr 11660Perform straight-line strength reduction on trees. This recognizes related 11661expressions involving multiplications and replaces them by less expensive 11662calculations when possible. This is enabled by default at @option{-O} and 11663higher. 11664 11665@item -ftree-vectorize 11666@opindex ftree-vectorize 11667Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize} 11668and @option{-ftree-slp-vectorize} if not explicitly specified. 11669 11670@item -ftree-loop-vectorize 11671@opindex ftree-loop-vectorize 11672Perform loop vectorization on trees. This flag is enabled by default at 11673@option{-O3} and by @option{-ftree-vectorize}, @option{-fprofile-use}, 11674and @option{-fauto-profile}. 11675 11676@item -ftree-slp-vectorize 11677@opindex ftree-slp-vectorize 11678Perform basic block vectorization on trees. This flag is enabled by default at 11679@option{-O3} and by @option{-ftree-vectorize}, @option{-fprofile-use}, 11680and @option{-fauto-profile}. 11681 11682@item -fvect-cost-model=@var{model} 11683@opindex fvect-cost-model 11684Alter the cost model used for vectorization. The @var{model} argument 11685should be one of @samp{unlimited}, @samp{dynamic}, @samp{cheap} or 11686@samp{very-cheap}. 11687With the @samp{unlimited} model the vectorized code-path is assumed 11688to be profitable while with the @samp{dynamic} model a runtime check 11689guards the vectorized code-path to enable it only for iteration 11690counts that will likely execute faster than when executing the original 11691scalar loop. The @samp{cheap} model disables vectorization of 11692loops where doing so would be cost prohibitive for example due to 11693required runtime checks for data dependence or alignment but otherwise 11694is equal to the @samp{dynamic} model. The @samp{very-cheap} model only 11695allows vectorization if the vector code would entirely replace the 11696scalar code that is being vectorized. For example, if each iteration 11697of a vectorized loop would only be able to handle exactly four iterations 11698of the scalar loop, the @samp{very-cheap} model would only allow 11699vectorization if the scalar iteration count is known to be a multiple 11700of four. 11701 11702The default cost model depends on other optimization flags and is 11703either @samp{dynamic} or @samp{cheap}. 11704 11705@item -fsimd-cost-model=@var{model} 11706@opindex fsimd-cost-model 11707Alter the cost model used for vectorization of loops marked with the OpenMP 11708simd directive. The @var{model} argument should be one of 11709@samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model} 11710have the same meaning as described in @option{-fvect-cost-model} and by 11711default a cost model defined with @option{-fvect-cost-model} is used. 11712 11713@item -ftree-vrp 11714@opindex ftree-vrp 11715Perform Value Range Propagation on trees. This is similar to the 11716constant propagation pass, but instead of values, ranges of values are 11717propagated. This allows the optimizers to remove unnecessary range 11718checks like array bound checks and null pointer checks. This is 11719enabled by default at @option{-O2} and higher. Null pointer check 11720elimination is only done if @option{-fdelete-null-pointer-checks} is 11721enabled. 11722 11723@item -fsplit-paths 11724@opindex fsplit-paths 11725Split paths leading to loop backedges. This can improve dead code 11726elimination and common subexpression elimination. This is enabled by 11727default at @option{-O3} and above. 11728 11729@item -fsplit-ivs-in-unroller 11730@opindex fsplit-ivs-in-unroller 11731Enables expression of values of induction variables in later iterations 11732of the unrolled loop using the value in the first iteration. This breaks 11733long dependency chains, thus improving efficiency of the scheduling passes. 11734 11735A combination of @option{-fweb} and CSE is often sufficient to obtain the 11736same effect. However, that is not reliable in cases where the loop body 11737is more complicated than a single basic block. It also does not work at all 11738on some architectures due to restrictions in the CSE pass. 11739 11740This optimization is enabled by default. 11741 11742@item -fvariable-expansion-in-unroller 11743@opindex fvariable-expansion-in-unroller 11744With this option, the compiler creates multiple copies of some 11745local variables when unrolling a loop, which can result in superior code. 11746 11747This optimization is enabled by default for PowerPC targets, but disabled 11748by default otherwise. 11749 11750@item -fpartial-inlining 11751@opindex fpartial-inlining 11752Inline parts of functions. This option has any effect only 11753when inlining itself is turned on by the @option{-finline-functions} 11754or @option{-finline-small-functions} options. 11755 11756Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 11757 11758@item -fpredictive-commoning 11759@opindex fpredictive-commoning 11760Perform predictive commoning optimization, i.e., reusing computations 11761(especially memory loads and stores) performed in previous 11762iterations of loops. 11763 11764This option is enabled at level @option{-O3}. 11765It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}. 11766 11767@item -fprefetch-loop-arrays 11768@opindex fprefetch-loop-arrays 11769If supported by the target machine, generate instructions to prefetch 11770memory to improve the performance of loops that access large arrays. 11771 11772This option may generate better or worse code; results are highly 11773dependent on the structure of loops within the source code. 11774 11775Disabled at level @option{-Os}. 11776 11777@item -fno-printf-return-value 11778@opindex fno-printf-return-value 11779@opindex fprintf-return-value 11780Do not substitute constants for known return value of formatted output 11781functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and 11782@code{vsnprintf} (but not @code{printf} of @code{fprintf}). This 11783transformation allows GCC to optimize or even eliminate branches based 11784on the known return value of these functions called with arguments that 11785are either constant, or whose values are known to be in a range that 11786makes determining the exact return value possible. For example, when 11787@option{-fprintf-return-value} is in effect, both the branch and the 11788body of the @code{if} statement (but not the call to @code{snprint}) 11789can be optimized away when @code{i} is a 32-bit or smaller integer 11790because the return value is guaranteed to be at most 8. 11791 11792@smallexample 11793char buf[9]; 11794if (snprintf (buf, "%08x", i) >= sizeof buf) 11795 @dots{} 11796@end smallexample 11797 11798The @option{-fprintf-return-value} option relies on other optimizations 11799and yields best results with @option{-O2} and above. It works in tandem 11800with the @option{-Wformat-overflow} and @option{-Wformat-truncation} 11801options. The @option{-fprintf-return-value} option is enabled by default. 11802 11803@item -fno-peephole 11804@itemx -fno-peephole2 11805@opindex fno-peephole 11806@opindex fpeephole 11807@opindex fno-peephole2 11808@opindex fpeephole2 11809Disable any machine-specific peephole optimizations. The difference 11810between @option{-fno-peephole} and @option{-fno-peephole2} is in how they 11811are implemented in the compiler; some targets use one, some use the 11812other, a few use both. 11813 11814@option{-fpeephole} is enabled by default. 11815@option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 11816 11817@item -fno-guess-branch-probability 11818@opindex fno-guess-branch-probability 11819@opindex fguess-branch-probability 11820Do not guess branch probabilities using heuristics. 11821 11822GCC uses heuristics to guess branch probabilities if they are 11823not provided by profiling feedback (@option{-fprofile-arcs}). These 11824heuristics are based on the control flow graph. If some branch probabilities 11825are specified by @code{__builtin_expect}, then the heuristics are 11826used to guess branch probabilities for the rest of the control flow graph, 11827taking the @code{__builtin_expect} info into account. The interactions 11828between the heuristics and @code{__builtin_expect} can be complex, and in 11829some cases, it may be useful to disable the heuristics so that the effects 11830of @code{__builtin_expect} are easier to understand. 11831 11832It is also possible to specify expected probability of the expression 11833with @code{__builtin_expect_with_probability} built-in function. 11834 11835The default is @option{-fguess-branch-probability} at levels 11836@option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 11837 11838@item -freorder-blocks 11839@opindex freorder-blocks 11840Reorder basic blocks in the compiled function in order to reduce number of 11841taken branches and improve code locality. 11842 11843Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 11844 11845@item -freorder-blocks-algorithm=@var{algorithm} 11846@opindex freorder-blocks-algorithm 11847Use the specified algorithm for basic block reordering. The 11848@var{algorithm} argument can be @samp{simple}, which does not increase 11849code size (except sometimes due to secondary effects like alignment), 11850or @samp{stc}, the ``software trace cache'' algorithm, which tries to 11851put all often executed code together, minimizing the number of branches 11852executed by making extra copies of code. 11853 11854The default is @samp{simple} at levels @option{-O}, @option{-Os}, and 11855@samp{stc} at levels @option{-O2}, @option{-O3}. 11856 11857@item -freorder-blocks-and-partition 11858@opindex freorder-blocks-and-partition 11859In addition to reordering basic blocks in the compiled function, in order 11860to reduce number of taken branches, partitions hot and cold basic blocks 11861into separate sections of the assembly and @file{.o} files, to improve 11862paging and cache locality performance. 11863 11864This optimization is automatically turned off in the presence of 11865exception handling or unwind tables (on targets using setjump/longjump or target specific scheme), for linkonce sections, for functions with a user-defined 11866section attribute and on any architecture that does not support named 11867sections. When @option{-fsplit-stack} is used this option is not 11868enabled by default (to avoid linker errors), but may be enabled 11869explicitly (if using a working linker). 11870 11871Enabled for x86 at levels @option{-O2}, @option{-O3}, @option{-Os}. 11872 11873@item -freorder-functions 11874@opindex freorder-functions 11875Reorder functions in the object file in order to 11876improve code locality. This is implemented by using special 11877subsections @code{.text.hot} for most frequently executed functions and 11878@code{.text.unlikely} for unlikely executed functions. Reordering is done by 11879the linker so object file format must support named sections and linker must 11880place them in a reasonable way. 11881 11882This option isn't effective unless you either provide profile feedback 11883(see @option{-fprofile-arcs} for details) or manually annotate functions with 11884@code{hot} or @code{cold} attributes (@pxref{Common Function Attributes}). 11885 11886Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 11887 11888@item -fstrict-aliasing 11889@opindex fstrict-aliasing 11890Allow the compiler to assume the strictest aliasing rules applicable to 11891the language being compiled. For C (and C++), this activates 11892optimizations based on the type of expressions. In particular, an 11893object of one type is assumed never to reside at the same address as an 11894object of a different type, unless the types are almost the same. For 11895example, an @code{unsigned int} can alias an @code{int}, but not a 11896@code{void*} or a @code{double}. A character type may alias any other 11897type. 11898 11899@anchor{Type-punning}Pay special attention to code like this: 11900@smallexample 11901union a_union @{ 11902 int i; 11903 double d; 11904@}; 11905 11906int f() @{ 11907 union a_union t; 11908 t.d = 3.0; 11909 return t.i; 11910@} 11911@end smallexample 11912The practice of reading from a different union member than the one most 11913recently written to (called ``type-punning'') is common. Even with 11914@option{-fstrict-aliasing}, type-punning is allowed, provided the memory 11915is accessed through the union type. So, the code above works as 11916expected. @xref{Structures unions enumerations and bit-fields 11917implementation}. However, this code might not: 11918@smallexample 11919int f() @{ 11920 union a_union t; 11921 int* ip; 11922 t.d = 3.0; 11923 ip = &t.i; 11924 return *ip; 11925@} 11926@end smallexample 11927 11928Similarly, access by taking the address, casting the resulting pointer 11929and dereferencing the result has undefined behavior, even if the cast 11930uses a union type, e.g.: 11931@smallexample 11932int f() @{ 11933 double d = 3.0; 11934 return ((union a_union *) &d)->i; 11935@} 11936@end smallexample 11937 11938The @option{-fstrict-aliasing} option is enabled at levels 11939@option{-O2}, @option{-O3}, @option{-Os}. 11940 11941@item -falign-functions 11942@itemx -falign-functions=@var{n} 11943@itemx -falign-functions=@var{n}:@var{m} 11944@itemx -falign-functions=@var{n}:@var{m}:@var{n2} 11945@itemx -falign-functions=@var{n}:@var{m}:@var{n2}:@var{m2} 11946@opindex falign-functions 11947Align the start of functions to the next power-of-two greater than or 11948equal to @var{n}, skipping up to @var{m}-1 bytes. This ensures that at 11949least the first @var{m} bytes of the function can be fetched by the CPU 11950without crossing an @var{n}-byte alignment boundary. 11951 11952If @var{m} is not specified, it defaults to @var{n}. 11953 11954Examples: @option{-falign-functions=32} aligns functions to the next 1195532-byte boundary, @option{-falign-functions=24} aligns to the next 1195632-byte boundary only if this can be done by skipping 23 bytes or less, 11957@option{-falign-functions=32:7} aligns to the next 1195832-byte boundary only if this can be done by skipping 6 bytes or less. 11959 11960The second pair of @var{n2}:@var{m2} values allows you to specify 11961a secondary alignment: @option{-falign-functions=64:7:32:3} aligns to 11962the next 64-byte boundary if this can be done by skipping 6 bytes or less, 11963otherwise aligns to the next 32-byte boundary if this can be done 11964by skipping 2 bytes or less. 11965If @var{m2} is not specified, it defaults to @var{n2}. 11966 11967Some assemblers only support this flag when @var{n} is a power of two; 11968in that case, it is rounded up. 11969 11970@option{-fno-align-functions} and @option{-falign-functions=1} are 11971equivalent and mean that functions are not aligned. 11972 11973If @var{n} is not specified or is zero, use a machine-dependent default. 11974The maximum allowed @var{n} option value is 65536. 11975 11976Enabled at levels @option{-O2}, @option{-O3}. 11977 11978@item -flimit-function-alignment 11979If this option is enabled, the compiler tries to avoid unnecessarily 11980overaligning functions. It attempts to instruct the assembler to align 11981by the amount specified by @option{-falign-functions}, but not to 11982skip more bytes than the size of the function. 11983 11984@item -falign-labels 11985@itemx -falign-labels=@var{n} 11986@itemx -falign-labels=@var{n}:@var{m} 11987@itemx -falign-labels=@var{n}:@var{m}:@var{n2} 11988@itemx -falign-labels=@var{n}:@var{m}:@var{n2}:@var{m2} 11989@opindex falign-labels 11990Align all branch targets to a power-of-two boundary. 11991 11992Parameters of this option are analogous to the @option{-falign-functions} option. 11993@option{-fno-align-labels} and @option{-falign-labels=1} are 11994equivalent and mean that labels are not aligned. 11995 11996If @option{-falign-loops} or @option{-falign-jumps} are applicable and 11997are greater than this value, then their values are used instead. 11998 11999If @var{n} is not specified or is zero, use a machine-dependent default 12000which is very likely to be @samp{1}, meaning no alignment. 12001The maximum allowed @var{n} option value is 65536. 12002 12003Enabled at levels @option{-O2}, @option{-O3}. 12004 12005@item -falign-loops 12006@itemx -falign-loops=@var{n} 12007@itemx -falign-loops=@var{n}:@var{m} 12008@itemx -falign-loops=@var{n}:@var{m}:@var{n2} 12009@itemx -falign-loops=@var{n}:@var{m}:@var{n2}:@var{m2} 12010@opindex falign-loops 12011Align loops to a power-of-two boundary. If the loops are executed 12012many times, this makes up for any execution of the dummy padding 12013instructions. 12014 12015If @option{-falign-labels} is greater than this value, then its value 12016is used instead. 12017 12018Parameters of this option are analogous to the @option{-falign-functions} option. 12019@option{-fno-align-loops} and @option{-falign-loops=1} are 12020equivalent and mean that loops are not aligned. 12021The maximum allowed @var{n} option value is 65536. 12022 12023If @var{n} is not specified or is zero, use a machine-dependent default. 12024 12025Enabled at levels @option{-O2}, @option{-O3}. 12026 12027@item -falign-jumps 12028@itemx -falign-jumps=@var{n} 12029@itemx -falign-jumps=@var{n}:@var{m} 12030@itemx -falign-jumps=@var{n}:@var{m}:@var{n2} 12031@itemx -falign-jumps=@var{n}:@var{m}:@var{n2}:@var{m2} 12032@opindex falign-jumps 12033Align branch targets to a power-of-two boundary, for branch targets 12034where the targets can only be reached by jumping. In this case, 12035no dummy operations need be executed. 12036 12037If @option{-falign-labels} is greater than this value, then its value 12038is used instead. 12039 12040Parameters of this option are analogous to the @option{-falign-functions} option. 12041@option{-fno-align-jumps} and @option{-falign-jumps=1} are 12042equivalent and mean that loops are not aligned. 12043 12044If @var{n} is not specified or is zero, use a machine-dependent default. 12045The maximum allowed @var{n} option value is 65536. 12046 12047Enabled at levels @option{-O2}, @option{-O3}. 12048 12049@item -fno-allocation-dce 12050@opindex fno-allocation-dce 12051Do not remove unused C++ allocations in dead code elimination. 12052 12053@item -fallow-store-data-races 12054@opindex fallow-store-data-races 12055Allow the compiler to perform optimizations that may introduce new data races 12056on stores, without proving that the variable cannot be concurrently accessed 12057by other threads. Does not affect optimization of local data. It is safe to 12058use this option if it is known that global data will not be accessed by 12059multiple threads. 12060 12061Examples of optimizations enabled by @option{-fallow-store-data-races} include 12062hoisting or if-conversions that may cause a value that was already in memory 12063to be re-written with that same value. Such re-writing is safe in a single 12064threaded context but may be unsafe in a multi-threaded context. Note that on 12065some processors, if-conversions may be required in order to enable 12066vectorization. 12067 12068Enabled at level @option{-Ofast}. 12069 12070@item -funit-at-a-time 12071@opindex funit-at-a-time 12072This option is left for compatibility reasons. @option{-funit-at-a-time} 12073has no effect, while @option{-fno-unit-at-a-time} implies 12074@option{-fno-toplevel-reorder} and @option{-fno-section-anchors}. 12075 12076Enabled by default. 12077 12078@item -fno-toplevel-reorder 12079@opindex fno-toplevel-reorder 12080@opindex ftoplevel-reorder 12081Do not reorder top-level functions, variables, and @code{asm} 12082statements. Output them in the same order that they appear in the 12083input file. When this option is used, unreferenced static variables 12084are not removed. This option is intended to support existing code 12085that relies on a particular ordering. For new code, it is better to 12086use attributes when possible. 12087 12088@option{-ftoplevel-reorder} is the default at @option{-O1} and higher, and 12089also at @option{-O0} if @option{-fsection-anchors} is explicitly requested. 12090Additionally @option{-fno-toplevel-reorder} implies 12091@option{-fno-section-anchors}. 12092 12093@item -fweb 12094@opindex fweb 12095Constructs webs as commonly used for register allocation purposes and assign 12096each web individual pseudo register. This allows the register allocation pass 12097to operate on pseudos directly, but also strengthens several other optimization 12098passes, such as CSE, loop optimizer and trivial dead code remover. It can, 12099however, make debugging impossible, since variables no longer stay in a 12100``home register''. 12101 12102Enabled by default with @option{-funroll-loops}. 12103 12104@item -fwhole-program 12105@opindex fwhole-program 12106Assume that the current compilation unit represents the whole program being 12107compiled. All public functions and variables with the exception of @code{main} 12108and those merged by attribute @code{externally_visible} become static functions 12109and in effect are optimized more aggressively by interprocedural optimizers. 12110 12111This option should not be used in combination with @option{-flto}. 12112Instead relying on a linker plugin should provide safer and more precise 12113information. 12114 12115@item -flto[=@var{n}] 12116@opindex flto 12117This option runs the standard link-time optimizer. When invoked 12118with source code, it generates GIMPLE (one of GCC's internal 12119representations) and writes it to special ELF sections in the object 12120file. When the object files are linked together, all the function 12121bodies are read from these ELF sections and instantiated as if they 12122had been part of the same translation unit. 12123 12124To use the link-time optimizer, @option{-flto} and optimization 12125options should be specified at compile time and during the final link. 12126It is recommended that you compile all the files participating in the 12127same link with the same options and also specify those options at 12128link time. 12129For example: 12130 12131@smallexample 12132gcc -c -O2 -flto foo.c 12133gcc -c -O2 -flto bar.c 12134gcc -o myprog -flto -O2 foo.o bar.o 12135@end smallexample 12136 12137The first two invocations to GCC save a bytecode representation 12138of GIMPLE into special ELF sections inside @file{foo.o} and 12139@file{bar.o}. The final invocation reads the GIMPLE bytecode from 12140@file{foo.o} and @file{bar.o}, merges the two files into a single 12141internal image, and compiles the result as usual. Since both 12142@file{foo.o} and @file{bar.o} are merged into a single image, this 12143causes all the interprocedural analyses and optimizations in GCC to 12144work across the two files as if they were a single one. This means, 12145for example, that the inliner is able to inline functions in 12146@file{bar.o} into functions in @file{foo.o} and vice-versa. 12147 12148Another (simpler) way to enable link-time optimization is: 12149 12150@smallexample 12151gcc -o myprog -flto -O2 foo.c bar.c 12152@end smallexample 12153 12154The above generates bytecode for @file{foo.c} and @file{bar.c}, 12155merges them together into a single GIMPLE representation and optimizes 12156them as usual to produce @file{myprog}. 12157 12158The important thing to keep in mind is that to enable link-time 12159optimizations you need to use the GCC driver to perform the link step. 12160GCC automatically performs link-time optimization if any of the 12161objects involved were compiled with the @option{-flto} command-line option. 12162You can always override 12163the automatic decision to do link-time optimization 12164by passing @option{-fno-lto} to the link command. 12165 12166To make whole program optimization effective, it is necessary to make 12167certain whole program assumptions. The compiler needs to know 12168what functions and variables can be accessed by libraries and runtime 12169outside of the link-time optimized unit. When supported by the linker, 12170the linker plugin (see @option{-fuse-linker-plugin}) passes information 12171to the compiler about used and externally visible symbols. When 12172the linker plugin is not available, @option{-fwhole-program} should be 12173used to allow the compiler to make these assumptions, which leads 12174to more aggressive optimization decisions. 12175 12176When a file is compiled with @option{-flto} without 12177@option{-fuse-linker-plugin}, the generated object file is larger than 12178a regular object file because it contains GIMPLE bytecodes and the usual 12179final code (see @option{-ffat-lto-objects}). This means that 12180object files with LTO information can be linked as normal object 12181files; if @option{-fno-lto} is passed to the linker, no 12182interprocedural optimizations are applied. Note that when 12183@option{-fno-fat-lto-objects} is enabled the compile stage is faster 12184but you cannot perform a regular, non-LTO link on them. 12185 12186When producing the final binary, GCC only 12187applies link-time optimizations to those files that contain bytecode. 12188Therefore, you can mix and match object files and libraries with 12189GIMPLE bytecodes and final object code. GCC automatically selects 12190which files to optimize in LTO mode and which files to link without 12191further processing. 12192 12193Generally, options specified at link time override those 12194specified at compile time, although in some cases GCC attempts to infer 12195link-time options from the settings used to compile the input files. 12196 12197If you do not specify an optimization level option @option{-O} at 12198link time, then GCC uses the highest optimization level 12199used when compiling the object files. Note that it is generally 12200ineffective to specify an optimization level option only at link time and 12201not at compile time, for two reasons. First, compiling without 12202optimization suppresses compiler passes that gather information 12203needed for effective optimization at link time. Second, some early 12204optimization passes can be performed only at compile time and 12205not at link time. 12206 12207There are some code generation flags preserved by GCC when 12208generating bytecodes, as they need to be used during the final link. 12209Currently, the following options and their settings are taken from 12210the first object file that explicitly specifies them: 12211@option{-fcommon}, @option{-fexceptions}, @option{-fnon-call-exceptions}, 12212@option{-fgnu-tm} and all the @option{-m} target flags. 12213 12214The following options @option{-fPIC}, @option{-fpic}, @option{-fpie} and 12215@option{-fPIE} are combined based on the following scheme: 12216 12217@smallexample 12218@option{-fPIC} + @option{-fpic} = @option{-fpic} 12219@option{-fPIC} + @option{-fno-pic} = @option{-fno-pic} 12220@option{-fpic/-fPIC} + (no option) = (no option) 12221@option{-fPIC} + @option{-fPIE} = @option{-fPIE} 12222@option{-fpic} + @option{-fPIE} = @option{-fpie} 12223@option{-fPIC/-fpic} + @option{-fpie} = @option{-fpie} 12224@end smallexample 12225 12226Certain ABI-changing flags are required to match in all compilation units, 12227and trying to override this at link time with a conflicting value 12228is ignored. This includes options such as @option{-freg-struct-return} 12229and @option{-fpcc-struct-return}. 12230 12231Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow}, 12232@option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing} 12233are passed through to the link stage and merged conservatively for 12234conflicting translation units. Specifically 12235@option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take 12236precedence; and for example @option{-ffp-contract=off} takes precedence 12237over @option{-ffp-contract=fast}. You can override them at link time. 12238 12239Diagnostic options such as @option{-Wstringop-overflow} are passed 12240through to the link stage and their setting matches that of the 12241compile-step at function granularity. Note that this matters only 12242for diagnostics emitted during optimization. Note that code 12243transforms such as inlining can lead to warnings being enabled 12244or disabled for regions if code not consistent with the setting 12245at compile time. 12246 12247When you need to pass options to the assembler via @option{-Wa} or 12248@option{-Xassembler} make sure to either compile such translation 12249units with @option{-fno-lto} or consistently use the same assembler 12250options on all translation units. You can alternatively also 12251specify assembler options at LTO link time. 12252 12253To enable debug info generation you need to supply @option{-g} at 12254compile time. If any of the input files at link time were built 12255with debug info generation enabled the link will enable debug info 12256generation as well. Any elaborate debug info settings 12257like the dwarf level @option{-gdwarf-5} need to be explicitly repeated 12258at the linker command line and mixing different settings in different 12259translation units is discouraged. 12260 12261If LTO encounters objects with C linkage declared with incompatible 12262types in separate translation units to be linked together (undefined 12263behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be 12264issued. The behavior is still undefined at run time. Similar 12265diagnostics may be raised for other languages. 12266 12267Another feature of LTO is that it is possible to apply interprocedural 12268optimizations on files written in different languages: 12269 12270@smallexample 12271gcc -c -flto foo.c 12272g++ -c -flto bar.cc 12273gfortran -c -flto baz.f90 12274g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran 12275@end smallexample 12276 12277Notice that the final link is done with @command{g++} to get the C++ 12278runtime libraries and @option{-lgfortran} is added to get the Fortran 12279runtime libraries. In general, when mixing languages in LTO mode, you 12280should use the same link command options as when mixing languages in a 12281regular (non-LTO) compilation. 12282 12283If object files containing GIMPLE bytecode are stored in a library archive, say 12284@file{libfoo.a}, it is possible to extract and use them in an LTO link if you 12285are using a linker with plugin support. To create static libraries suitable 12286for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar} 12287and @command{ranlib}; 12288to show the symbols of object files with GIMPLE bytecode, use 12289@command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib} 12290and @command{nm} have been compiled with plugin support. At link time, use the 12291flag @option{-fuse-linker-plugin} to ensure that the library participates in 12292the LTO optimization process: 12293 12294@smallexample 12295gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo 12296@end smallexample 12297 12298With the linker plugin enabled, the linker extracts the needed 12299GIMPLE files from @file{libfoo.a} and passes them on to the running GCC 12300to make them part of the aggregated GIMPLE image to be optimized. 12301 12302If you are not using a linker with plugin support and/or do not 12303enable the linker plugin, then the objects inside @file{libfoo.a} 12304are extracted and linked as usual, but they do not participate 12305in the LTO optimization process. In order to make a static library suitable 12306for both LTO optimization and usual linkage, compile its object files with 12307@option{-flto} @option{-ffat-lto-objects}. 12308 12309Link-time optimizations do not require the presence of the whole program to 12310operate. If the program does not require any symbols to be exported, it is 12311possible to combine @option{-flto} and @option{-fwhole-program} to allow 12312the interprocedural optimizers to use more aggressive assumptions which may 12313lead to improved optimization opportunities. 12314Use of @option{-fwhole-program} is not needed when linker plugin is 12315active (see @option{-fuse-linker-plugin}). 12316 12317The current implementation of LTO makes no 12318attempt to generate bytecode that is portable between different 12319types of hosts. The bytecode files are versioned and there is a 12320strict version check, so bytecode files generated in one version of 12321GCC do not work with an older or newer version of GCC. 12322 12323Link-time optimization does not work well with generation of debugging 12324information on systems other than those using a combination of ELF and 12325DWARF. 12326 12327If you specify the optional @var{n}, the optimization and code 12328generation done at link time is executed in parallel using @var{n} 12329parallel jobs by utilizing an installed @command{make} program. The 12330environment variable @env{MAKE} may be used to override the program 12331used. 12332 12333You can also specify @option{-flto=jobserver} to use GNU make's 12334job server mode to determine the number of parallel jobs. This 12335is useful when the Makefile calling GCC is already executing in parallel. 12336You must prepend a @samp{+} to the command recipe in the parent Makefile 12337for this to work. This option likely only works if @env{MAKE} is 12338GNU make. Even without the option value, GCC tries to automatically 12339detect a running GNU make's job server. 12340 12341Use @option{-flto=auto} to use GNU make's job server, if available, 12342or otherwise fall back to autodetection of the number of CPU threads 12343present in your system. 12344 12345@item -flto-partition=@var{alg} 12346@opindex flto-partition 12347Specify the partitioning algorithm used by the link-time optimizer. 12348The value is either @samp{1to1} to specify a partitioning mirroring 12349the original source files or @samp{balanced} to specify partitioning 12350into equally sized chunks (whenever possible) or @samp{max} to create 12351new partition for every symbol where possible. Specifying @samp{none} 12352as an algorithm disables partitioning and streaming completely. 12353The default value is @samp{balanced}. While @samp{1to1} can be used 12354as an workaround for various code ordering issues, the @samp{max} 12355partitioning is intended for internal testing only. 12356The value @samp{one} specifies that exactly one partition should be 12357used while the value @samp{none} bypasses partitioning and executes 12358the link-time optimization step directly from the WPA phase. 12359 12360@item -flto-compression-level=@var{n} 12361@opindex flto-compression-level 12362This option specifies the level of compression used for intermediate 12363language written to LTO object files, and is only meaningful in 12364conjunction with LTO mode (@option{-flto}). GCC currently supports two 12365LTO compression algorithms. For zstd, valid values are 0 (no compression) 12366to 19 (maximum compression), while zlib supports values from 0 to 9. 12367Values outside this range are clamped to either minimum or maximum 12368of the supported values. If the option is not given, 12369a default balanced compression setting is used. 12370 12371@item -fuse-linker-plugin 12372@opindex fuse-linker-plugin 12373Enables the use of a linker plugin during link-time optimization. This 12374option relies on plugin support in the linker, which is available in gold 12375or in GNU ld 2.21 or newer. 12376 12377This option enables the extraction of object files with GIMPLE bytecode out 12378of library archives. This improves the quality of optimization by exposing 12379more code to the link-time optimizer. This information specifies what 12380symbols can be accessed externally (by non-LTO object or during dynamic 12381linking). Resulting code quality improvements on binaries (and shared 12382libraries that use hidden visibility) are similar to @option{-fwhole-program}. 12383See @option{-flto} for a description of the effect of this flag and how to 12384use it. 12385 12386This option is enabled by default when LTO support in GCC is enabled 12387and GCC was configured for use with 12388a linker supporting plugins (GNU ld 2.21 or newer or gold). 12389 12390@item -ffat-lto-objects 12391@opindex ffat-lto-objects 12392Fat LTO objects are object files that contain both the intermediate language 12393and the object code. This makes them usable for both LTO linking and normal 12394linking. This option is effective only when compiling with @option{-flto} 12395and is ignored at link time. 12396 12397@option{-fno-fat-lto-objects} improves compilation time over plain LTO, but 12398requires the complete toolchain to be aware of LTO. It requires a linker with 12399linker plugin support for basic functionality. Additionally, 12400@command{nm}, @command{ar} and @command{ranlib} 12401need to support linker plugins to allow a full-featured build environment 12402(capable of building static libraries etc). GCC provides the @command{gcc-ar}, 12403@command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options 12404to these tools. With non fat LTO makefiles need to be modified to use them. 12405 12406Note that modern binutils provide plugin auto-load mechanism. 12407Installing the linker plugin into @file{$libdir/bfd-plugins} has the same 12408effect as usage of the command wrappers (@command{gcc-ar}, @command{gcc-nm} and 12409@command{gcc-ranlib}). 12410 12411The default is @option{-fno-fat-lto-objects} on targets with linker plugin 12412support. 12413 12414@item -fcompare-elim 12415@opindex fcompare-elim 12416After register allocation and post-register allocation instruction splitting, 12417identify arithmetic instructions that compute processor flags similar to a 12418comparison operation based on that arithmetic. If possible, eliminate the 12419explicit comparison operation. 12420 12421This pass only applies to certain targets that cannot explicitly represent 12422the comparison operation before register allocation is complete. 12423 12424Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 12425 12426@item -fcprop-registers 12427@opindex fcprop-registers 12428After register allocation and post-register allocation instruction splitting, 12429perform a copy-propagation pass to try to reduce scheduling dependencies 12430and occasionally eliminate the copy. 12431 12432Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}. 12433 12434@item -fprofile-correction 12435@opindex fprofile-correction 12436Profiles collected using an instrumented binary for multi-threaded programs may 12437be inconsistent due to missed counter updates. When this option is specified, 12438GCC uses heuristics to correct or smooth out such inconsistencies. By 12439default, GCC emits an error message when an inconsistent profile is detected. 12440 12441This option is enabled by @option{-fauto-profile}. 12442 12443@item -fprofile-partial-training 12444@opindex fprofile-partial-training 12445With @code{-fprofile-use} all portions of programs not executed during train 12446run are optimized agressively for size rather than speed. In some cases it is 12447not practical to train all possible hot paths in the program. (For 12448example, program may contain functions specific for a given hardware and 12449trianing may not cover all hardware configurations program is run on.) With 12450@code{-fprofile-partial-training} profile feedback will be ignored for all 12451functions not executed during the train run leading them to be optimized as if 12452they were compiled without profile feedback. This leads to better performance 12453when train run is not representative but also leads to significantly bigger 12454code. 12455 12456@item -fprofile-use 12457@itemx -fprofile-use=@var{path} 12458@opindex fprofile-use 12459Enable profile feedback-directed optimizations, 12460and the following optimizations, many of which 12461are generally profitable only with profile feedback available: 12462 12463@gccoptlist{-fbranch-probabilities -fprofile-values @gol 12464-funroll-loops -fpeel-loops -ftracer -fvpt @gol 12465-finline-functions -fipa-cp -fipa-cp-clone -fipa-bit-cp @gol 12466-fpredictive-commoning -fsplit-loops -funswitch-loops @gol 12467-fgcse-after-reload -ftree-loop-vectorize -ftree-slp-vectorize @gol 12468-fvect-cost-model=dynamic -ftree-loop-distribute-patterns @gol 12469-fprofile-reorder-functions} 12470 12471Before you can use this option, you must first generate profiling information. 12472@xref{Instrumentation Options}, for information about the 12473@option{-fprofile-generate} option. 12474 12475By default, GCC emits an error message if the feedback profiles do not 12476match the source code. This error can be turned into a warning by using 12477@option{-Wno-error=coverage-mismatch}. Note this may result in poorly 12478optimized code. Additionally, by default, GCC also emits a warning message if 12479the feedback profiles do not exist (see @option{-Wmissing-profile}). 12480 12481If @var{path} is specified, GCC looks at the @var{path} to find 12482the profile feedback data files. See @option{-fprofile-dir}. 12483 12484@item -fauto-profile 12485@itemx -fauto-profile=@var{path} 12486@opindex fauto-profile 12487Enable sampling-based feedback-directed optimizations, 12488and the following optimizations, 12489many of which are generally profitable only with profile feedback available: 12490 12491@gccoptlist{-fbranch-probabilities -fprofile-values @gol 12492-funroll-loops -fpeel-loops -ftracer -fvpt @gol 12493-finline-functions -fipa-cp -fipa-cp-clone -fipa-bit-cp @gol 12494-fpredictive-commoning -fsplit-loops -funswitch-loops @gol 12495-fgcse-after-reload -ftree-loop-vectorize -ftree-slp-vectorize @gol 12496-fvect-cost-model=dynamic -ftree-loop-distribute-patterns @gol 12497-fprofile-correction} 12498 12499@var{path} is the name of a file containing AutoFDO profile information. 12500If omitted, it defaults to @file{fbdata.afdo} in the current directory. 12501 12502Producing an AutoFDO profile data file requires running your program 12503with the @command{perf} utility on a supported GNU/Linux target system. 12504For more information, see @uref{https://perf.wiki.kernel.org/}. 12505 12506E.g. 12507@smallexample 12508perf record -e br_inst_retired:near_taken -b -o perf.data \ 12509 -- your_program 12510@end smallexample 12511 12512Then use the @command{create_gcov} tool to convert the raw profile data 12513to a format that can be used by GCC.@ You must also supply the 12514unstripped binary for your program to this tool. 12515See @uref{https://github.com/google/autofdo}. 12516 12517E.g. 12518@smallexample 12519create_gcov --binary=your_program.unstripped --profile=perf.data \ 12520 --gcov=profile.afdo 12521@end smallexample 12522@end table 12523 12524The following options control compiler behavior regarding floating-point 12525arithmetic. These options trade off between speed and 12526correctness. All must be specifically enabled. 12527 12528@table @gcctabopt 12529@item -ffloat-store 12530@opindex ffloat-store 12531Do not store floating-point variables in registers, and inhibit other 12532options that might change whether a floating-point value is taken from a 12533register or memory. 12534 12535@cindex floating-point precision 12536This option prevents undesirable excess precision on machines such as 12537the 68000 where the floating registers (of the 68881) keep more 12538precision than a @code{double} is supposed to have. Similarly for the 12539x86 architecture. For most programs, the excess precision does only 12540good, but a few programs rely on the precise definition of IEEE floating 12541point. Use @option{-ffloat-store} for such programs, after modifying 12542them to store all pertinent intermediate computations into variables. 12543 12544@item -fexcess-precision=@var{style} 12545@opindex fexcess-precision 12546This option allows further control over excess precision on machines 12547where floating-point operations occur in a format with more precision or 12548range than the IEEE standard and interchange floating-point types. By 12549default, @option{-fexcess-precision=fast} is in effect; this means that 12550operations may be carried out in a wider precision than the types specified 12551in the source if that would result in faster code, and it is unpredictable 12552when rounding to the types specified in the source code takes place. 12553When compiling C, if @option{-fexcess-precision=standard} is specified then 12554excess precision follows the rules specified in ISO C99; in particular, 12555both casts and assignments cause values to be rounded to their 12556semantic types (whereas @option{-ffloat-store} only affects 12557assignments). This option is enabled by default for C if a strict 12558conformance option such as @option{-std=c99} is used. 12559@option{-ffast-math} enables @option{-fexcess-precision=fast} by default 12560regardless of whether a strict conformance option is used. 12561 12562@opindex mfpmath 12563@option{-fexcess-precision=standard} is not implemented for languages 12564other than C. On the x86, it has no effect if @option{-mfpmath=sse} 12565or @option{-mfpmath=sse+387} is specified; in the former case, IEEE 12566semantics apply without excess precision, and in the latter, rounding 12567is unpredictable. 12568 12569@item -ffast-math 12570@opindex ffast-math 12571Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, 12572@option{-ffinite-math-only}, @option{-fno-rounding-math}, 12573@option{-fno-signaling-nans}, @option{-fcx-limited-range} and 12574@option{-fexcess-precision=fast}. 12575 12576This option causes the preprocessor macro @code{__FAST_MATH__} to be defined. 12577 12578This option is not turned on by any @option{-O} option besides 12579@option{-Ofast} since it can result in incorrect output for programs 12580that depend on an exact implementation of IEEE or ISO rules/specifications 12581for math functions. It may, however, yield faster code for programs 12582that do not require the guarantees of these specifications. 12583 12584@item -fno-math-errno 12585@opindex fno-math-errno 12586@opindex fmath-errno 12587Do not set @code{errno} after calling math functions that are executed 12588with a single instruction, e.g., @code{sqrt}. A program that relies on 12589IEEE exceptions for math error handling may want to use this flag 12590for speed while maintaining IEEE arithmetic compatibility. 12591 12592This option is not turned on by any @option{-O} option since 12593it can result in incorrect output for programs that depend on 12594an exact implementation of IEEE or ISO rules/specifications for 12595math functions. It may, however, yield faster code for programs 12596that do not require the guarantees of these specifications. 12597 12598The default is @option{-fmath-errno}. 12599 12600On Darwin systems, the math library never sets @code{errno}. There is 12601therefore no reason for the compiler to consider the possibility that 12602it might, and @option{-fno-math-errno} is the default. 12603 12604@item -funsafe-math-optimizations 12605@opindex funsafe-math-optimizations 12606 12607Allow optimizations for floating-point arithmetic that (a) assume 12608that arguments and results are valid and (b) may violate IEEE or 12609ANSI standards. When used at link time, it may include libraries 12610or startup files that change the default FPU control word or other 12611similar optimizations. 12612 12613This option is not turned on by any @option{-O} option since 12614it can result in incorrect output for programs that depend on 12615an exact implementation of IEEE or ISO rules/specifications for 12616math functions. It may, however, yield faster code for programs 12617that do not require the guarantees of these specifications. 12618Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math}, 12619@option{-fassociative-math} and @option{-freciprocal-math}. 12620 12621The default is @option{-fno-unsafe-math-optimizations}. 12622 12623@item -fassociative-math 12624@opindex fassociative-math 12625 12626Allow re-association of operands in series of floating-point operations. 12627This violates the ISO C and C++ language standard by possibly changing 12628computation result. NOTE: re-ordering may change the sign of zero as 12629well as ignore NaNs and inhibit or create underflow or overflow (and 12630thus cannot be used on code that relies on rounding behavior like 12631@code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons 12632and thus may not be used when ordered comparisons are required. 12633This option requires that both @option{-fno-signed-zeros} and 12634@option{-fno-trapping-math} be in effect. Moreover, it doesn't make 12635much sense with @option{-frounding-math}. For Fortran the option 12636is automatically enabled when both @option{-fno-signed-zeros} and 12637@option{-fno-trapping-math} are in effect. 12638 12639The default is @option{-fno-associative-math}. 12640 12641@item -freciprocal-math 12642@opindex freciprocal-math 12643 12644Allow the reciprocal of a value to be used instead of dividing by 12645the value if this enables optimizations. For example @code{x / y} 12646can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)} 12647is subject to common subexpression elimination. Note that this loses 12648precision and increases the number of flops operating on the value. 12649 12650The default is @option{-fno-reciprocal-math}. 12651 12652@item -ffinite-math-only 12653@opindex ffinite-math-only 12654Allow optimizations for floating-point arithmetic that assume 12655that arguments and results are not NaNs or +-Infs. 12656 12657This option is not turned on by any @option{-O} option since 12658it can result in incorrect output for programs that depend on 12659an exact implementation of IEEE or ISO rules/specifications for 12660math functions. It may, however, yield faster code for programs 12661that do not require the guarantees of these specifications. 12662 12663The default is @option{-fno-finite-math-only}. 12664 12665@item -fno-signed-zeros 12666@opindex fno-signed-zeros 12667@opindex fsigned-zeros 12668Allow optimizations for floating-point arithmetic that ignore the 12669signedness of zero. IEEE arithmetic specifies the behavior of 12670distinct +0.0 and @minus{}0.0 values, which then prohibits simplification 12671of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}). 12672This option implies that the sign of a zero result isn't significant. 12673 12674The default is @option{-fsigned-zeros}. 12675 12676@item -fno-trapping-math 12677@opindex fno-trapping-math 12678@opindex ftrapping-math 12679Compile code assuming that floating-point operations cannot generate 12680user-visible traps. These traps include division by zero, overflow, 12681underflow, inexact result and invalid operation. This option requires 12682that @option{-fno-signaling-nans} be in effect. Setting this option may 12683allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example. 12684 12685This option should never be turned on by any @option{-O} option since 12686it can result in incorrect output for programs that depend on 12687an exact implementation of IEEE or ISO rules/specifications for 12688math functions. 12689 12690The default is @option{-ftrapping-math}. 12691 12692@item -frounding-math 12693@opindex frounding-math 12694Disable transformations and optimizations that assume default floating-point 12695rounding behavior. This is round-to-zero for all floating point 12696to integer conversions, and round-to-nearest for all other arithmetic 12697truncations. This option should be specified for programs that change 12698the FP rounding mode dynamically, or that may be executed with a 12699non-default rounding mode. This option disables constant folding of 12700floating-point expressions at compile time (which may be affected by 12701rounding mode) and arithmetic transformations that are unsafe in the 12702presence of sign-dependent rounding modes. 12703 12704The default is @option{-fno-rounding-math}. 12705 12706This option is experimental and does not currently guarantee to 12707disable all GCC optimizations that are affected by rounding mode. 12708Future versions of GCC may provide finer control of this setting 12709using C99's @code{FENV_ACCESS} pragma. This command-line option 12710will be used to specify the default state for @code{FENV_ACCESS}. 12711 12712@item -fsignaling-nans 12713@opindex fsignaling-nans 12714Compile code assuming that IEEE signaling NaNs may generate user-visible 12715traps during floating-point operations. Setting this option disables 12716optimizations that may change the number of exceptions visible with 12717signaling NaNs. This option implies @option{-ftrapping-math}. 12718 12719This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to 12720be defined. 12721 12722The default is @option{-fno-signaling-nans}. 12723 12724This option is experimental and does not currently guarantee to 12725disable all GCC optimizations that affect signaling NaN behavior. 12726 12727@item -fno-fp-int-builtin-inexact 12728@opindex fno-fp-int-builtin-inexact 12729@opindex ffp-int-builtin-inexact 12730Do not allow the built-in functions @code{ceil}, @code{floor}, 12731@code{round} and @code{trunc}, and their @code{float} and @code{long 12732double} variants, to generate code that raises the ``inexact'' 12733floating-point exception for noninteger arguments. ISO C99 and C11 12734allow these functions to raise the ``inexact'' exception, but ISO/IEC 12735TS 18661-1:2014, the C bindings to IEEE 754-2008, as integrated into 12736ISO C2X, does not allow these functions to do so. 12737 12738The default is @option{-ffp-int-builtin-inexact}, allowing the 12739exception to be raised, unless C2X or a later C standard is selected. 12740This option does nothing unless @option{-ftrapping-math} is in effect. 12741 12742Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions 12743generate a call to a library function then the ``inexact'' exception 12744may be raised if the library implementation does not follow TS 18661. 12745 12746@item -fsingle-precision-constant 12747@opindex fsingle-precision-constant 12748Treat floating-point constants as single precision instead of 12749implicitly converting them to double-precision constants. 12750 12751@item -fcx-limited-range 12752@opindex fcx-limited-range 12753When enabled, this option states that a range reduction step is not 12754needed when performing complex division. Also, there is no checking 12755whether the result of a complex multiplication or division is @code{NaN 12756+ I*NaN}, with an attempt to rescue the situation in that case. The 12757default is @option{-fno-cx-limited-range}, but is enabled by 12758@option{-ffast-math}. 12759 12760This option controls the default setting of the ISO C99 12761@code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to 12762all languages. 12763 12764@item -fcx-fortran-rules 12765@opindex fcx-fortran-rules 12766Complex multiplication and division follow Fortran rules. Range 12767reduction is done as part of complex division, but there is no checking 12768whether the result of a complex multiplication or division is @code{NaN 12769+ I*NaN}, with an attempt to rescue the situation in that case. 12770 12771The default is @option{-fno-cx-fortran-rules}. 12772 12773@end table 12774 12775The following options control optimizations that may improve 12776performance, but are not enabled by any @option{-O} options. This 12777section includes experimental options that may produce broken code. 12778 12779@table @gcctabopt 12780@item -fbranch-probabilities 12781@opindex fbranch-probabilities 12782After running a program compiled with @option{-fprofile-arcs} 12783(@pxref{Instrumentation Options}), 12784you can compile it a second time using 12785@option{-fbranch-probabilities}, to improve optimizations based on 12786the number of times each branch was taken. When a program 12787compiled with @option{-fprofile-arcs} exits, it saves arc execution 12788counts to a file called @file{@var{sourcename}.gcda} for each source 12789file. The information in this data file is very dependent on the 12790structure of the generated code, so you must use the same source code 12791and the same optimization options for both compilations. 12792 12793With @option{-fbranch-probabilities}, GCC puts a 12794@samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}. 12795These can be used to improve optimization. Currently, they are only 12796used in one place: in @file{reorg.c}, instead of guessing which path a 12797branch is most likely to take, the @samp{REG_BR_PROB} values are used to 12798exactly determine which path is taken more often. 12799 12800Enabled by @option{-fprofile-use} and @option{-fauto-profile}. 12801 12802@item -fprofile-values 12803@opindex fprofile-values 12804If combined with @option{-fprofile-arcs}, it adds code so that some 12805data about values of expressions in the program is gathered. 12806 12807With @option{-fbranch-probabilities}, it reads back the data gathered 12808from profiling values of expressions for usage in optimizations. 12809 12810Enabled by @option{-fprofile-generate}, @option{-fprofile-use}, and 12811@option{-fauto-profile}. 12812 12813@item -fprofile-reorder-functions 12814@opindex fprofile-reorder-functions 12815Function reordering based on profile instrumentation collects 12816first time of execution of a function and orders these functions 12817in ascending order. 12818 12819Enabled with @option{-fprofile-use}. 12820 12821@item -fvpt 12822@opindex fvpt 12823If combined with @option{-fprofile-arcs}, this option instructs the compiler 12824to add code to gather information about values of expressions. 12825 12826With @option{-fbranch-probabilities}, it reads back the data gathered 12827and actually performs the optimizations based on them. 12828Currently the optimizations include specialization of division operations 12829using the knowledge about the value of the denominator. 12830 12831Enabled with @option{-fprofile-use} and @option{-fauto-profile}. 12832 12833@item -frename-registers 12834@opindex frename-registers 12835Attempt to avoid false dependencies in scheduled code by making use 12836of registers left over after register allocation. This optimization 12837most benefits processors with lots of registers. Depending on the 12838debug information format adopted by the target, however, it can 12839make debugging impossible, since variables no longer stay in 12840a ``home register''. 12841 12842Enabled by default with @option{-funroll-loops}. 12843 12844@item -fschedule-fusion 12845@opindex fschedule-fusion 12846Performs a target dependent pass over the instruction stream to schedule 12847instructions of same type together because target machine can execute them 12848more efficiently if they are adjacent to each other in the instruction flow. 12849 12850Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}. 12851 12852@item -ftracer 12853@opindex ftracer 12854Perform tail duplication to enlarge superblock size. This transformation 12855simplifies the control flow of the function allowing other optimizations to do 12856a better job. 12857 12858Enabled by @option{-fprofile-use} and @option{-fauto-profile}. 12859 12860@item -funroll-loops 12861@opindex funroll-loops 12862Unroll loops whose number of iterations can be determined at compile time or 12863upon entry to the loop. @option{-funroll-loops} implies 12864@option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}. 12865It also turns on complete loop peeling (i.e.@: complete removal of loops with 12866a small constant number of iterations). This option makes code larger, and may 12867or may not make it run faster. 12868 12869Enabled by @option{-fprofile-use} and @option{-fauto-profile}. 12870 12871@item -funroll-all-loops 12872@opindex funroll-all-loops 12873Unroll all loops, even if their number of iterations is uncertain when 12874the loop is entered. This usually makes programs run more slowly. 12875@option{-funroll-all-loops} implies the same options as 12876@option{-funroll-loops}. 12877 12878@item -fpeel-loops 12879@opindex fpeel-loops 12880Peels loops for which there is enough information that they do not 12881roll much (from profile feedback or static analysis). It also turns on 12882complete loop peeling (i.e.@: complete removal of loops with small constant 12883number of iterations). 12884 12885Enabled by @option{-O3}, @option{-fprofile-use}, and @option{-fauto-profile}. 12886 12887@item -fmove-loop-invariants 12888@opindex fmove-loop-invariants 12889Enables the loop invariant motion pass in the RTL loop optimizer. Enabled 12890at level @option{-O1} and higher, except for @option{-Og}. 12891 12892@item -fsplit-loops 12893@opindex fsplit-loops 12894Split a loop into two if it contains a condition that's always true 12895for one side of the iteration space and false for the other. 12896 12897Enabled by @option{-fprofile-use} and @option{-fauto-profile}. 12898 12899@item -funswitch-loops 12900@opindex funswitch-loops 12901Move branches with loop invariant conditions out of the loop, with duplicates 12902of the loop on both branches (modified according to result of the condition). 12903 12904Enabled by @option{-fprofile-use} and @option{-fauto-profile}. 12905 12906@item -fversion-loops-for-strides 12907@opindex fversion-loops-for-strides 12908If a loop iterates over an array with a variable stride, create another 12909version of the loop that assumes the stride is always one. For example: 12910 12911@smallexample 12912for (int i = 0; i < n; ++i) 12913 x[i * stride] = @dots{}; 12914@end smallexample 12915 12916becomes: 12917 12918@smallexample 12919if (stride == 1) 12920 for (int i = 0; i < n; ++i) 12921 x[i] = @dots{}; 12922else 12923 for (int i = 0; i < n; ++i) 12924 x[i * stride] = @dots{}; 12925@end smallexample 12926 12927This is particularly useful for assumed-shape arrays in Fortran where 12928(for example) it allows better vectorization assuming contiguous accesses. 12929This flag is enabled by default at @option{-O3}. 12930It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}. 12931 12932@item -ffunction-sections 12933@itemx -fdata-sections 12934@opindex ffunction-sections 12935@opindex fdata-sections 12936Place each function or data item into its own section in the output 12937file if the target supports arbitrary sections. The name of the 12938function or the name of the data item determines the section's name 12939in the output file. 12940 12941Use these options on systems where the linker can perform optimizations to 12942improve locality of reference in the instruction space. Most systems using the 12943ELF object format have linkers with such optimizations. On AIX, the linker 12944rearranges sections (CSECTs) based on the call graph. The performance impact 12945varies. 12946 12947Together with a linker garbage collection (linker @option{--gc-sections} 12948option) these options may lead to smaller statically-linked executables (after 12949stripping). 12950 12951On ELF/DWARF systems these options do not degenerate the quality of the debug 12952information. There could be issues with other object files/debug info formats. 12953 12954Only use these options when there are significant benefits from doing so. When 12955you specify these options, the assembler and linker create larger object and 12956executable files and are also slower. These options affect code generation. 12957They prevent optimizations by the compiler and assembler using relative 12958locations inside a translation unit since the locations are unknown until 12959link time. An example of such an optimization is relaxing calls to short call 12960instructions. 12961 12962@item -fstdarg-opt 12963@opindex fstdarg-opt 12964Optimize the prologue of variadic argument functions with respect to usage of 12965those arguments. 12966 12967@item -fsection-anchors 12968@opindex fsection-anchors 12969Try to reduce the number of symbolic address calculations by using 12970shared ``anchor'' symbols to address nearby objects. This transformation 12971can help to reduce the number of GOT entries and GOT accesses on some 12972targets. 12973 12974For example, the implementation of the following function @code{foo}: 12975 12976@smallexample 12977static int a, b, c; 12978int foo (void) @{ return a + b + c; @} 12979@end smallexample 12980 12981@noindent 12982usually calculates the addresses of all three variables, but if you 12983compile it with @option{-fsection-anchors}, it accesses the variables 12984from a common anchor point instead. The effect is similar to the 12985following pseudocode (which isn't valid C): 12986 12987@smallexample 12988int foo (void) 12989@{ 12990 register int *xr = &x; 12991 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x]; 12992@} 12993@end smallexample 12994 12995Not all targets support this option. 12996 12997@item -fzero-call-used-regs=@var{choice} 12998@opindex fzero-call-used-regs 12999Zero call-used registers at function return to increase program 13000security by either mitigating Return-Oriented Programming (ROP) 13001attacks or preventing information leakage through registers. 13002 13003The possible values of @var{choice} are the same as for the 13004@code{zero_call_used_regs} attribute (@pxref{Function Attributes}). 13005The default is @samp{skip}. 13006 13007You can control this behavior for a specific function by using the function 13008attribute @code{zero_call_used_regs} (@pxref{Function Attributes}). 13009 13010@item --param @var{name}=@var{value} 13011@opindex param 13012In some places, GCC uses various constants to control the amount of 13013optimization that is done. For example, GCC does not inline functions 13014that contain more than a certain number of instructions. You can 13015control some of these constants on the command line using the 13016@option{--param} option. 13017 13018The names of specific parameters, and the meaning of the values, are 13019tied to the internals of the compiler, and are subject to change 13020without notice in future releases. 13021 13022In order to get minimal, maximal and default value of a parameter, 13023one can use @option{--help=param -Q} options. 13024 13025In each case, the @var{value} is an integer. The following choices 13026of @var{name} are recognized for all targets: 13027 13028@table @gcctabopt 13029@item predictable-branch-outcome 13030When branch is predicted to be taken with probability lower than this threshold 13031(in percent), then it is considered well predictable. 13032 13033@item max-rtl-if-conversion-insns 13034RTL if-conversion tries to remove conditional branches around a block and 13035replace them with conditionally executed instructions. This parameter 13036gives the maximum number of instructions in a block which should be 13037considered for if-conversion. The compiler will 13038also use other heuristics to decide whether if-conversion is likely to be 13039profitable. 13040 13041@item max-rtl-if-conversion-predictable-cost 13042RTL if-conversion will try to remove conditional branches around a block 13043and replace them with conditionally executed instructions. These parameters 13044give the maximum permissible cost for the sequence that would be generated 13045by if-conversion depending on whether the branch is statically determined 13046to be predictable or not. The units for this parameter are the same as 13047those for the GCC internal seq_cost metric. The compiler will try to 13048provide a reasonable default for this parameter using the BRANCH_COST 13049target macro. 13050 13051@item max-crossjump-edges 13052The maximum number of incoming edges to consider for cross-jumping. 13053The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in 13054the number of edges incoming to each block. Increasing values mean 13055more aggressive optimization, making the compilation time increase with 13056probably small improvement in executable size. 13057 13058@item min-crossjump-insns 13059The minimum number of instructions that must be matched at the end 13060of two blocks before cross-jumping is performed on them. This 13061value is ignored in the case where all instructions in the block being 13062cross-jumped from are matched. 13063 13064@item max-grow-copy-bb-insns 13065The maximum code size expansion factor when copying basic blocks 13066instead of jumping. The expansion is relative to a jump instruction. 13067 13068@item max-goto-duplication-insns 13069The maximum number of instructions to duplicate to a block that jumps 13070to a computed goto. To avoid @math{O(N^2)} behavior in a number of 13071passes, GCC factors computed gotos early in the compilation process, 13072and unfactors them as late as possible. Only computed jumps at the 13073end of a basic blocks with no more than max-goto-duplication-insns are 13074unfactored. 13075 13076@item max-delay-slot-insn-search 13077The maximum number of instructions to consider when looking for an 13078instruction to fill a delay slot. If more than this arbitrary number of 13079instructions are searched, the time savings from filling the delay slot 13080are minimal, so stop searching. Increasing values mean more 13081aggressive optimization, making the compilation time increase with probably 13082small improvement in execution time. 13083 13084@item max-delay-slot-live-search 13085When trying to fill delay slots, the maximum number of instructions to 13086consider when searching for a block with valid live register 13087information. Increasing this arbitrarily chosen value means more 13088aggressive optimization, increasing the compilation time. This parameter 13089should be removed when the delay slot code is rewritten to maintain the 13090control-flow graph. 13091 13092@item max-gcse-memory 13093The approximate maximum amount of memory in @code{kB} that can be allocated in 13094order to perform the global common subexpression elimination 13095optimization. If more memory than specified is required, the 13096optimization is not done. 13097 13098@item max-gcse-insertion-ratio 13099If the ratio of expression insertions to deletions is larger than this value 13100for any expression, then RTL PRE inserts or removes the expression and thus 13101leaves partially redundant computations in the instruction stream. 13102 13103@item max-pending-list-length 13104The maximum number of pending dependencies scheduling allows 13105before flushing the current state and starting over. Large functions 13106with few branches or calls can create excessively large lists which 13107needlessly consume memory and resources. 13108 13109@item max-modulo-backtrack-attempts 13110The maximum number of backtrack attempts the scheduler should make 13111when modulo scheduling a loop. Larger values can exponentially increase 13112compilation time. 13113 13114@item max-inline-insns-single 13115Several parameters control the tree inliner used in GCC@. This number sets the 13116maximum number of instructions (counted in GCC's internal representation) in a 13117single function that the tree inliner considers for inlining. This only 13118affects functions declared inline and methods implemented in a class 13119declaration (C++). 13120 13121 13122@item max-inline-insns-auto 13123When you use @option{-finline-functions} (included in @option{-O3}), 13124a lot of functions that would otherwise not be considered for inlining 13125by the compiler are investigated. To those functions, a different 13126(more restrictive) limit compared to functions declared inline can 13127be applied (@option{--param max-inline-insns-auto}). 13128 13129@item max-inline-insns-small 13130This is bound applied to calls which are considered relevant with 13131@option{-finline-small-functions}. 13132 13133@item max-inline-insns-size 13134This is bound applied to calls which are optimized for size. Small growth 13135may be desirable to anticipate optimization oppurtunities exposed by inlining. 13136 13137@item uninlined-function-insns 13138Number of instructions accounted by inliner for function overhead such as 13139function prologue and epilogue. 13140 13141@item uninlined-function-time 13142Extra time accounted by inliner for function overhead such as time needed to 13143execute function prologue and epilogue 13144 13145@item inline-heuristics-hint-percent 13146The scale (in percents) applied to @option{inline-insns-single}, 13147@option{inline-insns-single-O2}, @option{inline-insns-auto} 13148when inline heuristics hints that inlining is 13149very profitable (will enable later optimizations). 13150 13151@item uninlined-thunk-insns 13152@item uninlined-thunk-time 13153Same as @option{--param uninlined-function-insns} and 13154@option{--param uninlined-function-time} but applied to function thunks 13155 13156@item inline-min-speedup 13157When estimated performance improvement of caller + callee runtime exceeds this 13158threshold (in percent), the function can be inlined regardless of the limit on 13159@option{--param max-inline-insns-single} and @option{--param 13160max-inline-insns-auto}. 13161 13162@item large-function-insns 13163The limit specifying really large functions. For functions larger than this 13164limit after inlining, inlining is constrained by 13165@option{--param large-function-growth}. This parameter is useful primarily 13166to avoid extreme compilation time caused by non-linear algorithms used by the 13167back end. 13168 13169@item large-function-growth 13170Specifies maximal growth of large function caused by inlining in percents. 13171For example, parameter value 100 limits large function growth to 2.0 times 13172the original size. 13173 13174@item large-unit-insns 13175The limit specifying large translation unit. Growth caused by inlining of 13176units larger than this limit is limited by @option{--param inline-unit-growth}. 13177For small units this might be too tight. 13178For example, consider a unit consisting of function A 13179that is inline and B that just calls A three times. If B is small relative to 13180A, the growth of unit is 300\% and yet such inlining is very sane. For very 13181large units consisting of small inlineable functions, however, the overall unit 13182growth limit is needed to avoid exponential explosion of code size. Thus for 13183smaller units, the size is increased to @option{--param large-unit-insns} 13184before applying @option{--param inline-unit-growth}. 13185 13186@item lazy-modules 13187Maximum number of concurrently open C++ module files when lazy loading. 13188 13189@item inline-unit-growth 13190Specifies maximal overall growth of the compilation unit caused by inlining. 13191For example, parameter value 20 limits unit growth to 1.2 times the original 13192size. Cold functions (either marked cold via an attribute or by profile 13193feedback) are not accounted into the unit size. 13194 13195@item ipa-cp-unit-growth 13196Specifies maximal overall growth of the compilation unit caused by 13197interprocedural constant propagation. For example, parameter value 10 limits 13198unit growth to 1.1 times the original size. 13199 13200@item ipa-cp-large-unit-insns 13201The size of translation unit that IPA-CP pass considers large. 13202 13203@item large-stack-frame 13204The limit specifying large stack frames. While inlining the algorithm is trying 13205to not grow past this limit too much. 13206 13207@item large-stack-frame-growth 13208Specifies maximal growth of large stack frames caused by inlining in percents. 13209For example, parameter value 1000 limits large stack frame growth to 11 times 13210the original size. 13211 13212@item max-inline-insns-recursive 13213@itemx max-inline-insns-recursive-auto 13214Specifies the maximum number of instructions an out-of-line copy of a 13215self-recursive inline 13216function can grow into by performing recursive inlining. 13217 13218@option{--param max-inline-insns-recursive} applies to functions 13219declared inline. 13220For functions not declared inline, recursive inlining 13221happens only when @option{-finline-functions} (included in @option{-O3}) is 13222enabled; @option{--param max-inline-insns-recursive-auto} applies instead. 13223 13224@item max-inline-recursive-depth 13225@itemx max-inline-recursive-depth-auto 13226Specifies the maximum recursion depth used for recursive inlining. 13227 13228@option{--param max-inline-recursive-depth} applies to functions 13229declared inline. For functions not declared inline, recursive inlining 13230happens only when @option{-finline-functions} (included in @option{-O3}) is 13231enabled; @option{--param max-inline-recursive-depth-auto} applies instead. 13232 13233@item min-inline-recursive-probability 13234Recursive inlining is profitable only for function having deep recursion 13235in average and can hurt for function having little recursion depth by 13236increasing the prologue size or complexity of function body to other 13237optimizers. 13238 13239When profile feedback is available (see @option{-fprofile-generate}) the actual 13240recursion depth can be guessed from the probability that function recurses 13241via a given call expression. This parameter limits inlining only to call 13242expressions whose probability exceeds the given threshold (in percents). 13243 13244@item early-inlining-insns 13245Specify growth that the early inliner can make. In effect it increases 13246the amount of inlining for code having a large abstraction penalty. 13247 13248@item max-early-inliner-iterations 13249Limit of iterations of the early inliner. This basically bounds 13250the number of nested indirect calls the early inliner can resolve. 13251Deeper chains are still handled by late inlining. 13252 13253@item comdat-sharing-probability 13254Probability (in percent) that C++ inline function with comdat visibility 13255are shared across multiple compilation units. 13256 13257@item modref-max-bases 13258@item modref-max-refs 13259@item modref-max-accesses 13260Specifies the maximal number of base pointers, references and accesses stored 13261for a single function by mod/ref analysis. 13262 13263@item modref-max-tests 13264Specifies the maxmal number of tests alias oracle can perform to disambiguate 13265memory locations using the mod/ref information. This parameter ought to be 13266bigger than @option{--param modref-max-bases} and @option{--param 13267modref-max-refs}. 13268 13269@item modref-max-depth 13270Specifies the maximum depth of DFS walk used by modref escape analysis. 13271Setting to 0 disables the analysis completely. 13272 13273@item modref-max-escape-points 13274Specifies the maximum number of escape points tracked by modref per SSA-name. 13275 13276@item profile-func-internal-id 13277A parameter to control whether to use function internal id in profile 13278database lookup. If the value is 0, the compiler uses an id that 13279is based on function assembler name and filename, which makes old profile 13280data more tolerant to source changes such as function reordering etc. 13281 13282@item min-vect-loop-bound 13283The minimum number of iterations under which loops are not vectorized 13284when @option{-ftree-vectorize} is used. The number of iterations after 13285vectorization needs to be greater than the value specified by this option 13286to allow vectorization. 13287 13288@item gcse-cost-distance-ratio 13289Scaling factor in calculation of maximum distance an expression 13290can be moved by GCSE optimizations. This is currently supported only in the 13291code hoisting pass. The bigger the ratio, the more aggressive code hoisting 13292is with simple expressions, i.e., the expressions that have cost 13293less than @option{gcse-unrestricted-cost}. Specifying 0 disables 13294hoisting of simple expressions. 13295 13296@item gcse-unrestricted-cost 13297Cost, roughly measured as the cost of a single typical machine 13298instruction, at which GCSE optimizations do not constrain 13299the distance an expression can travel. This is currently 13300supported only in the code hoisting pass. The lesser the cost, 13301the more aggressive code hoisting is. Specifying 0 13302allows all expressions to travel unrestricted distances. 13303 13304@item max-hoist-depth 13305The depth of search in the dominator tree for expressions to hoist. 13306This is used to avoid quadratic behavior in hoisting algorithm. 13307The value of 0 does not limit on the search, but may slow down compilation 13308of huge functions. 13309 13310@item max-tail-merge-comparisons 13311The maximum amount of similar bbs to compare a bb with. This is used to 13312avoid quadratic behavior in tree tail merging. 13313 13314@item max-tail-merge-iterations 13315The maximum amount of iterations of the pass over the function. This is used to 13316limit compilation time in tree tail merging. 13317 13318@item store-merging-allow-unaligned 13319Allow the store merging pass to introduce unaligned stores if it is legal to 13320do so. 13321 13322@item max-stores-to-merge 13323The maximum number of stores to attempt to merge into wider stores in the store 13324merging pass. 13325 13326@item max-store-chains-to-track 13327The maximum number of store chains to track at the same time in the attempt 13328to merge them into wider stores in the store merging pass. 13329 13330@item max-stores-to-track 13331The maximum number of stores to track at the same time in the attemt to 13332to merge them into wider stores in the store merging pass. 13333 13334@item max-unrolled-insns 13335The maximum number of instructions that a loop may have to be unrolled. 13336If a loop is unrolled, this parameter also determines how many times 13337the loop code is unrolled. 13338 13339@item max-average-unrolled-insns 13340The maximum number of instructions biased by probabilities of their execution 13341that a loop may have to be unrolled. If a loop is unrolled, 13342this parameter also determines how many times the loop code is unrolled. 13343 13344@item max-unroll-times 13345The maximum number of unrollings of a single loop. 13346 13347@item max-peeled-insns 13348The maximum number of instructions that a loop may have to be peeled. 13349If a loop is peeled, this parameter also determines how many times 13350the loop code is peeled. 13351 13352@item max-peel-times 13353The maximum number of peelings of a single loop. 13354 13355@item max-peel-branches 13356The maximum number of branches on the hot path through the peeled sequence. 13357 13358@item max-completely-peeled-insns 13359The maximum number of insns of a completely peeled loop. 13360 13361@item max-completely-peel-times 13362The maximum number of iterations of a loop to be suitable for complete peeling. 13363 13364@item max-completely-peel-loop-nest-depth 13365The maximum depth of a loop nest suitable for complete peeling. 13366 13367@item max-unswitch-insns 13368The maximum number of insns of an unswitched loop. 13369 13370@item max-unswitch-level 13371The maximum number of branches unswitched in a single loop. 13372 13373@item lim-expensive 13374The minimum cost of an expensive expression in the loop invariant motion. 13375 13376@item min-loop-cond-split-prob 13377When FDO profile information is available, @option{min-loop-cond-split-prob} 13378specifies minimum threshold for probability of semi-invariant condition 13379statement to trigger loop split. 13380 13381@item iv-consider-all-candidates-bound 13382Bound on number of candidates for induction variables, below which 13383all candidates are considered for each use in induction variable 13384optimizations. If there are more candidates than this, 13385only the most relevant ones are considered to avoid quadratic time complexity. 13386 13387@item iv-max-considered-uses 13388The induction variable optimizations give up on loops that contain more 13389induction variable uses. 13390 13391@item iv-always-prune-cand-set-bound 13392If the number of candidates in the set is smaller than this value, 13393always try to remove unnecessary ivs from the set 13394when adding a new one. 13395 13396@item avg-loop-niter 13397Average number of iterations of a loop. 13398 13399@item dse-max-object-size 13400Maximum size (in bytes) of objects tracked bytewise by dead store elimination. 13401Larger values may result in larger compilation times. 13402 13403@item dse-max-alias-queries-per-store 13404Maximum number of queries into the alias oracle per store. 13405Larger values result in larger compilation times and may result in more 13406removed dead stores. 13407 13408@item scev-max-expr-size 13409Bound on size of expressions used in the scalar evolutions analyzer. 13410Large expressions slow the analyzer. 13411 13412@item scev-max-expr-complexity 13413Bound on the complexity of the expressions in the scalar evolutions analyzer. 13414Complex expressions slow the analyzer. 13415 13416@item max-tree-if-conversion-phi-args 13417Maximum number of arguments in a PHI supported by TREE if conversion 13418unless the loop is marked with simd pragma. 13419 13420@item vect-max-version-for-alignment-checks 13421The maximum number of run-time checks that can be performed when 13422doing loop versioning for alignment in the vectorizer. 13423 13424@item vect-max-version-for-alias-checks 13425The maximum number of run-time checks that can be performed when 13426doing loop versioning for alias in the vectorizer. 13427 13428@item vect-max-peeling-for-alignment 13429The maximum number of loop peels to enhance access alignment 13430for vectorizer. Value -1 means no limit. 13431 13432@item max-iterations-to-track 13433The maximum number of iterations of a loop the brute-force algorithm 13434for analysis of the number of iterations of the loop tries to evaluate. 13435 13436@item hot-bb-count-fraction 13437The denominator n of fraction 1/n of the maximal execution count of a 13438basic block in the entire program that a basic block needs to at least 13439have in order to be considered hot. The default is 10000, which means 13440that a basic block is considered hot if its execution count is greater 13441than 1/10000 of the maximal execution count. 0 means that it is never 13442considered hot. Used in non-LTO mode. 13443 13444@item hot-bb-count-ws-permille 13445The number of most executed permilles, ranging from 0 to 1000, of the 13446profiled execution of the entire program to which the execution count 13447of a basic block must be part of in order to be considered hot. The 13448default is 990, which means that a basic block is considered hot if 13449its execution count contributes to the upper 990 permilles, or 99.0%, 13450of the profiled execution of the entire program. 0 means that it is 13451never considered hot. Used in LTO mode. 13452 13453@item hot-bb-frequency-fraction 13454The denominator n of fraction 1/n of the execution frequency of the 13455entry block of a function that a basic block of this function needs 13456to at least have in order to be considered hot. The default is 1000, 13457which means that a basic block is considered hot in a function if it 13458is executed more frequently than 1/1000 of the frequency of the entry 13459block of the function. 0 means that it is never considered hot. 13460 13461@item unlikely-bb-count-fraction 13462The denominator n of fraction 1/n of the number of profiled runs of 13463the entire program below which the execution count of a basic block 13464must be in order for the basic block to be considered unlikely executed. 13465The default is 20, which means that a basic block is considered unlikely 13466executed if it is executed in fewer than 1/20, or 5%, of the runs of 13467the program. 0 means that it is always considered unlikely executed. 13468 13469@item max-predicted-iterations 13470The maximum number of loop iterations we predict statically. This is useful 13471in cases where a function contains a single loop with known bound and 13472another loop with unknown bound. 13473The known number of iterations is predicted correctly, while 13474the unknown number of iterations average to roughly 10. This means that the 13475loop without bounds appears artificially cold relative to the other one. 13476 13477@item builtin-expect-probability 13478Control the probability of the expression having the specified value. This 13479parameter takes a percentage (i.e.@: 0 ... 100) as input. 13480 13481@item builtin-string-cmp-inline-length 13482The maximum length of a constant string for a builtin string cmp call 13483eligible for inlining. 13484 13485@item align-threshold 13486 13487Select fraction of the maximal frequency of executions of a basic block in 13488a function to align the basic block. 13489 13490@item align-loop-iterations 13491 13492A loop expected to iterate at least the selected number of iterations is 13493aligned. 13494 13495@item tracer-dynamic-coverage 13496@itemx tracer-dynamic-coverage-feedback 13497 13498This value is used to limit superblock formation once the given percentage of 13499executed instructions is covered. This limits unnecessary code size 13500expansion. 13501 13502The @option{tracer-dynamic-coverage-feedback} parameter 13503is used only when profile 13504feedback is available. The real profiles (as opposed to statically estimated 13505ones) are much less balanced allowing the threshold to be larger value. 13506 13507@item tracer-max-code-growth 13508Stop tail duplication once code growth has reached given percentage. This is 13509a rather artificial limit, as most of the duplicates are eliminated later in 13510cross jumping, so it may be set to much higher values than is the desired code 13511growth. 13512 13513@item tracer-min-branch-ratio 13514 13515Stop reverse growth when the reverse probability of best edge is less than this 13516threshold (in percent). 13517 13518@item tracer-min-branch-probability 13519@itemx tracer-min-branch-probability-feedback 13520 13521Stop forward growth if the best edge has probability lower than this 13522threshold. 13523 13524Similarly to @option{tracer-dynamic-coverage} two parameters are 13525provided. @option{tracer-min-branch-probability-feedback} is used for 13526compilation with profile feedback and @option{tracer-min-branch-probability} 13527compilation without. The value for compilation with profile feedback 13528needs to be more conservative (higher) in order to make tracer 13529effective. 13530 13531@item stack-clash-protection-guard-size 13532Specify the size of the operating system provided stack guard as 135332 raised to @var{num} bytes. Higher values may reduce the 13534number of explicit probes, but a value larger than the operating system 13535provided guard will leave code vulnerable to stack clash style attacks. 13536 13537@item stack-clash-protection-probe-interval 13538Stack clash protection involves probing stack space as it is allocated. This 13539param controls the maximum distance between probes into the stack as 2 raised 13540to @var{num} bytes. Higher values may reduce the number of explicit probes, but a value 13541larger than the operating system provided guard will leave code vulnerable to 13542stack clash style attacks. 13543 13544@item max-cse-path-length 13545 13546The maximum number of basic blocks on path that CSE considers. 13547 13548@item max-cse-insns 13549The maximum number of instructions CSE processes before flushing. 13550 13551@item ggc-min-expand 13552 13553GCC uses a garbage collector to manage its own memory allocation. This 13554parameter specifies the minimum percentage by which the garbage 13555collector's heap should be allowed to expand between collections. 13556Tuning this may improve compilation speed; it has no effect on code 13557generation. 13558 13559The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when 13560RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is 13561the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If 13562GCC is not able to calculate RAM on a particular platform, the lower 13563bound of 30% is used. Setting this parameter and 13564@option{ggc-min-heapsize} to zero causes a full collection to occur at 13565every opportunity. This is extremely slow, but can be useful for 13566debugging. 13567 13568@item ggc-min-heapsize 13569 13570Minimum size of the garbage collector's heap before it begins bothering 13571to collect garbage. The first collection occurs after the heap expands 13572by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again, 13573tuning this may improve compilation speed, and has no effect on code 13574generation. 13575 13576The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that 13577tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but 13578with a lower bound of 4096 (four megabytes) and an upper bound of 13579131072 (128 megabytes). If GCC is not able to calculate RAM on a 13580particular platform, the lower bound is used. Setting this parameter 13581very large effectively disables garbage collection. Setting this 13582parameter and @option{ggc-min-expand} to zero causes a full collection 13583to occur at every opportunity. 13584 13585@item max-reload-search-insns 13586The maximum number of instruction reload should look backward for equivalent 13587register. Increasing values mean more aggressive optimization, making the 13588compilation time increase with probably slightly better performance. 13589 13590@item max-cselib-memory-locations 13591The maximum number of memory locations cselib should take into account. 13592Increasing values mean more aggressive optimization, making the compilation time 13593increase with probably slightly better performance. 13594 13595@item max-sched-ready-insns 13596The maximum number of instructions ready to be issued the scheduler should 13597consider at any given time during the first scheduling pass. Increasing 13598values mean more thorough searches, making the compilation time increase 13599with probably little benefit. 13600 13601@item max-sched-region-blocks 13602The maximum number of blocks in a region to be considered for 13603interblock scheduling. 13604 13605@item max-pipeline-region-blocks 13606The maximum number of blocks in a region to be considered for 13607pipelining in the selective scheduler. 13608 13609@item max-sched-region-insns 13610The maximum number of insns in a region to be considered for 13611interblock scheduling. 13612 13613@item max-pipeline-region-insns 13614The maximum number of insns in a region to be considered for 13615pipelining in the selective scheduler. 13616 13617@item min-spec-prob 13618The minimum probability (in percents) of reaching a source block 13619for interblock speculative scheduling. 13620 13621@item max-sched-extend-regions-iters 13622The maximum number of iterations through CFG to extend regions. 13623A value of 0 disables region extensions. 13624 13625@item max-sched-insn-conflict-delay 13626The maximum conflict delay for an insn to be considered for speculative motion. 13627 13628@item sched-spec-prob-cutoff 13629The minimal probability of speculation success (in percents), so that 13630speculative insns are scheduled. 13631 13632@item sched-state-edge-prob-cutoff 13633The minimum probability an edge must have for the scheduler to save its 13634state across it. 13635 13636@item sched-mem-true-dep-cost 13637Minimal distance (in CPU cycles) between store and load targeting same 13638memory locations. 13639 13640@item selsched-max-lookahead 13641The maximum size of the lookahead window of selective scheduling. It is a 13642depth of search for available instructions. 13643 13644@item selsched-max-sched-times 13645The maximum number of times that an instruction is scheduled during 13646selective scheduling. This is the limit on the number of iterations 13647through which the instruction may be pipelined. 13648 13649@item selsched-insns-to-rename 13650The maximum number of best instructions in the ready list that are considered 13651for renaming in the selective scheduler. 13652 13653@item sms-min-sc 13654The minimum value of stage count that swing modulo scheduler 13655generates. 13656 13657@item max-last-value-rtl 13658The maximum size measured as number of RTLs that can be recorded in an expression 13659in combiner for a pseudo register as last known value of that register. 13660 13661@item max-combine-insns 13662The maximum number of instructions the RTL combiner tries to combine. 13663 13664@item integer-share-limit 13665Small integer constants can use a shared data structure, reducing the 13666compiler's memory usage and increasing its speed. This sets the maximum 13667value of a shared integer constant. 13668 13669@item ssp-buffer-size 13670The minimum size of buffers (i.e.@: arrays) that receive stack smashing 13671protection when @option{-fstack-protection} is used. 13672 13673@item min-size-for-stack-sharing 13674The minimum size of variables taking part in stack slot sharing when not 13675optimizing. 13676 13677@item max-jump-thread-duplication-stmts 13678Maximum number of statements allowed in a block that needs to be 13679duplicated when threading jumps. 13680 13681@item max-fields-for-field-sensitive 13682Maximum number of fields in a structure treated in 13683a field sensitive manner during pointer analysis. 13684 13685@item prefetch-latency 13686Estimate on average number of instructions that are executed before 13687prefetch finishes. The distance prefetched ahead is proportional 13688to this constant. Increasing this number may also lead to less 13689streams being prefetched (see @option{simultaneous-prefetches}). 13690 13691@item simultaneous-prefetches 13692Maximum number of prefetches that can run at the same time. 13693 13694@item l1-cache-line-size 13695The size of cache line in L1 data cache, in bytes. 13696 13697@item l1-cache-size 13698The size of L1 data cache, in kilobytes. 13699 13700@item l2-cache-size 13701The size of L2 data cache, in kilobytes. 13702 13703@item prefetch-dynamic-strides 13704Whether the loop array prefetch pass should issue software prefetch hints 13705for strides that are non-constant. In some cases this may be 13706beneficial, though the fact the stride is non-constant may make it 13707hard to predict when there is clear benefit to issuing these hints. 13708 13709Set to 1 if the prefetch hints should be issued for non-constant 13710strides. Set to 0 if prefetch hints should be issued only for strides that 13711are known to be constant and below @option{prefetch-minimum-stride}. 13712 13713@item prefetch-minimum-stride 13714Minimum constant stride, in bytes, to start using prefetch hints for. If 13715the stride is less than this threshold, prefetch hints will not be issued. 13716 13717This setting is useful for processors that have hardware prefetchers, in 13718which case there may be conflicts between the hardware prefetchers and 13719the software prefetchers. If the hardware prefetchers have a maximum 13720stride they can handle, it should be used here to improve the use of 13721software prefetchers. 13722 13723A value of -1 means we don't have a threshold and therefore 13724prefetch hints can be issued for any constant stride. 13725 13726This setting is only useful for strides that are known and constant. 13727 13728@item loop-interchange-max-num-stmts 13729The maximum number of stmts in a loop to be interchanged. 13730 13731@item loop-interchange-stride-ratio 13732The minimum ratio between stride of two loops for interchange to be profitable. 13733 13734@item min-insn-to-prefetch-ratio 13735The minimum ratio between the number of instructions and the 13736number of prefetches to enable prefetching in a loop. 13737 13738@item prefetch-min-insn-to-mem-ratio 13739The minimum ratio between the number of instructions and the 13740number of memory references to enable prefetching in a loop. 13741 13742@item use-canonical-types 13743Whether the compiler should use the ``canonical'' type system. 13744Should always be 1, which uses a more efficient internal 13745mechanism for comparing types in C++ and Objective-C++. However, if 13746bugs in the canonical type system are causing compilation failures, 13747set this value to 0 to disable canonical types. 13748 13749@item switch-conversion-max-branch-ratio 13750Switch initialization conversion refuses to create arrays that are 13751bigger than @option{switch-conversion-max-branch-ratio} times the number of 13752branches in the switch. 13753 13754@item max-partial-antic-length 13755Maximum length of the partial antic set computed during the tree 13756partial redundancy elimination optimization (@option{-ftree-pre}) when 13757optimizing at @option{-O3} and above. For some sorts of source code 13758the enhanced partial redundancy elimination optimization can run away, 13759consuming all of the memory available on the host machine. This 13760parameter sets a limit on the length of the sets that are computed, 13761which prevents the runaway behavior. Setting a value of 0 for 13762this parameter allows an unlimited set length. 13763 13764@item rpo-vn-max-loop-depth 13765Maximum loop depth that is value-numbered optimistically. 13766When the limit hits the innermost 13767@var{rpo-vn-max-loop-depth} loops and the outermost loop in the 13768loop nest are value-numbered optimistically and the remaining ones not. 13769 13770@item sccvn-max-alias-queries-per-access 13771Maximum number of alias-oracle queries we perform when looking for 13772redundancies for loads and stores. If this limit is hit the search 13773is aborted and the load or store is not considered redundant. The 13774number of queries is algorithmically limited to the number of 13775stores on all paths from the load to the function entry. 13776 13777@item ira-max-loops-num 13778IRA uses regional register allocation by default. If a function 13779contains more loops than the number given by this parameter, only at most 13780the given number of the most frequently-executed loops form regions 13781for regional register allocation. 13782 13783@item ira-max-conflict-table-size 13784Although IRA uses a sophisticated algorithm to compress the conflict 13785table, the table can still require excessive amounts of memory for 13786huge functions. If the conflict table for a function could be more 13787than the size in MB given by this parameter, the register allocator 13788instead uses a faster, simpler, and lower-quality 13789algorithm that does not require building a pseudo-register conflict table. 13790 13791@item ira-loop-reserved-regs 13792IRA can be used to evaluate more accurate register pressure in loops 13793for decisions to move loop invariants (see @option{-O3}). The number 13794of available registers reserved for some other purposes is given 13795by this parameter. Default of the parameter 13796is the best found from numerous experiments. 13797 13798@item lra-inheritance-ebb-probability-cutoff 13799LRA tries to reuse values reloaded in registers in subsequent insns. 13800This optimization is called inheritance. EBB is used as a region to 13801do this optimization. The parameter defines a minimal fall-through 13802edge probability in percentage used to add BB to inheritance EBB in 13803LRA. The default value was chosen 13804from numerous runs of SPEC2000 on x86-64. 13805 13806@item loop-invariant-max-bbs-in-loop 13807Loop invariant motion can be very expensive, both in compilation time and 13808in amount of needed compile-time memory, with very large loops. Loops 13809with more basic blocks than this parameter won't have loop invariant 13810motion optimization performed on them. 13811 13812@item loop-max-datarefs-for-datadeps 13813Building data dependencies is expensive for very large loops. This 13814parameter limits the number of data references in loops that are 13815considered for data dependence analysis. These large loops are no 13816handled by the optimizations using loop data dependencies. 13817 13818@item max-vartrack-size 13819Sets a maximum number of hash table slots to use during variable 13820tracking dataflow analysis of any function. If this limit is exceeded 13821with variable tracking at assignments enabled, analysis for that 13822function is retried without it, after removing all debug insns from 13823the function. If the limit is exceeded even without debug insns, var 13824tracking analysis is completely disabled for the function. Setting 13825the parameter to zero makes it unlimited. 13826 13827@item max-vartrack-expr-depth 13828Sets a maximum number of recursion levels when attempting to map 13829variable names or debug temporaries to value expressions. This trades 13830compilation time for more complete debug information. If this is set too 13831low, value expressions that are available and could be represented in 13832debug information may end up not being used; setting this higher may 13833enable the compiler to find more complex debug expressions, but compile 13834time and memory use may grow. 13835 13836@item max-debug-marker-count 13837Sets a threshold on the number of debug markers (e.g.@: begin stmt 13838markers) to avoid complexity explosion at inlining or expanding to RTL. 13839If a function has more such gimple stmts than the set limit, such stmts 13840will be dropped from the inlined copy of a function, and from its RTL 13841expansion. 13842 13843@item min-nondebug-insn-uid 13844Use uids starting at this parameter for nondebug insns. The range below 13845the parameter is reserved exclusively for debug insns created by 13846@option{-fvar-tracking-assignments}, but debug insns may get 13847(non-overlapping) uids above it if the reserved range is exhausted. 13848 13849@item ipa-sra-ptr-growth-factor 13850IPA-SRA replaces a pointer to an aggregate with one or more new 13851parameters only when their cumulative size is less or equal to 13852@option{ipa-sra-ptr-growth-factor} times the size of the original 13853pointer parameter. 13854 13855@item ipa-sra-max-replacements 13856Maximum pieces of an aggregate that IPA-SRA tracks. As a 13857consequence, it is also the maximum number of replacements of a formal 13858parameter. 13859 13860@item sra-max-scalarization-size-Ospeed 13861@itemx sra-max-scalarization-size-Osize 13862The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to 13863replace scalar parts of aggregates with uses of independent scalar 13864variables. These parameters control the maximum size, in storage units, 13865of aggregate which is considered for replacement when compiling for 13866speed 13867(@option{sra-max-scalarization-size-Ospeed}) or size 13868(@option{sra-max-scalarization-size-Osize}) respectively. 13869 13870@item sra-max-propagations 13871The maximum number of artificial accesses that Scalar Replacement of 13872Aggregates (SRA) will track, per one local variable, in order to 13873facilitate copy propagation. 13874 13875@item tm-max-aggregate-size 13876When making copies of thread-local variables in a transaction, this 13877parameter specifies the size in bytes after which variables are 13878saved with the logging functions as opposed to save/restore code 13879sequence pairs. This option only applies when using 13880@option{-fgnu-tm}. 13881 13882@item graphite-max-nb-scop-params 13883To avoid exponential effects in the Graphite loop transforms, the 13884number of parameters in a Static Control Part (SCoP) is bounded. 13885A value of zero can be used to lift 13886the bound. A variable whose value is unknown at compilation time and 13887defined outside a SCoP is a parameter of the SCoP. 13888 13889@item loop-block-tile-size 13890Loop blocking or strip mining transforms, enabled with 13891@option{-floop-block} or @option{-floop-strip-mine}, strip mine each 13892loop in the loop nest by a given number of iterations. The strip 13893length can be changed using the @option{loop-block-tile-size} 13894parameter. 13895 13896@item ipa-jump-function-lookups 13897Specifies number of statements visited during jump function offset discovery. 13898 13899@item ipa-cp-value-list-size 13900IPA-CP attempts to track all possible values and types passed to a function's 13901parameter in order to propagate them and perform devirtualization. 13902@option{ipa-cp-value-list-size} is the maximum number of values and types it 13903stores per one formal parameter of a function. 13904 13905@item ipa-cp-eval-threshold 13906IPA-CP calculates its own score of cloning profitability heuristics 13907and performs those cloning opportunities with scores that exceed 13908@option{ipa-cp-eval-threshold}. 13909 13910@item ipa-cp-max-recursive-depth 13911Maximum depth of recursive cloning for self-recursive function. 13912 13913@item ipa-cp-min-recursive-probability 13914Recursive cloning only when the probability of call being executed exceeds 13915the parameter. 13916 13917@item ipa-cp-recursion-penalty 13918Percentage penalty the recursive functions will receive when they 13919are evaluated for cloning. 13920 13921@item ipa-cp-single-call-penalty 13922Percentage penalty functions containing a single call to another 13923function will receive when they are evaluated for cloning. 13924 13925@item ipa-max-agg-items 13926IPA-CP is also capable to propagate a number of scalar values passed 13927in an aggregate. @option{ipa-max-agg-items} controls the maximum 13928number of such values per one parameter. 13929 13930@item ipa-cp-loop-hint-bonus 13931When IPA-CP determines that a cloning candidate would make the number 13932of iterations of a loop known, it adds a bonus of 13933@option{ipa-cp-loop-hint-bonus} to the profitability score of 13934the candidate. 13935 13936@item ipa-max-loop-predicates 13937The maximum number of different predicates IPA will use to describe when 13938loops in a function have known properties. 13939 13940@item ipa-max-aa-steps 13941During its analysis of function bodies, IPA-CP employs alias analysis 13942in order to track values pointed to by function parameters. In order 13943not spend too much time analyzing huge functions, it gives up and 13944consider all memory clobbered after examining 13945@option{ipa-max-aa-steps} statements modifying memory. 13946 13947@item ipa-max-switch-predicate-bounds 13948Maximal number of boundary endpoints of case ranges of switch statement. 13949For switch exceeding this limit, IPA-CP will not construct cloning cost 13950predicate, which is used to estimate cloning benefit, for default case 13951of the switch statement. 13952 13953@item ipa-max-param-expr-ops 13954IPA-CP will analyze conditional statement that references some function 13955parameter to estimate benefit for cloning upon certain constant value. 13956But if number of operations in a parameter expression exceeds 13957@option{ipa-max-param-expr-ops}, the expression is treated as complicated 13958one, and is not handled by IPA analysis. 13959 13960@item lto-partitions 13961Specify desired number of partitions produced during WHOPR compilation. 13962The number of partitions should exceed the number of CPUs used for compilation. 13963 13964@item lto-min-partition 13965Size of minimal partition for WHOPR (in estimated instructions). 13966This prevents expenses of splitting very small programs into too many 13967partitions. 13968 13969@item lto-max-partition 13970Size of max partition for WHOPR (in estimated instructions). 13971to provide an upper bound for individual size of partition. 13972Meant to be used only with balanced partitioning. 13973 13974@item lto-max-streaming-parallelism 13975Maximal number of parallel processes used for LTO streaming. 13976 13977@item cxx-max-namespaces-for-diagnostic-help 13978The maximum number of namespaces to consult for suggestions when C++ 13979name lookup fails for an identifier. 13980 13981@item sink-frequency-threshold 13982The maximum relative execution frequency (in percents) of the target block 13983relative to a statement's original block to allow statement sinking of a 13984statement. Larger numbers result in more aggressive statement sinking. 13985A small positive adjustment is applied for 13986statements with memory operands as those are even more profitable so sink. 13987 13988@item max-stores-to-sink 13989The maximum number of conditional store pairs that can be sunk. Set to 0 13990if either vectorization (@option{-ftree-vectorize}) or if-conversion 13991(@option{-ftree-loop-if-convert}) is disabled. 13992 13993@item case-values-threshold 13994The smallest number of different values for which it is best to use a 13995jump-table instead of a tree of conditional branches. If the value is 139960, use the default for the machine. 13997 13998@item jump-table-max-growth-ratio-for-size 13999The maximum code size growth ratio when expanding 14000into a jump table (in percent). The parameter is used when 14001optimizing for size. 14002 14003@item jump-table-max-growth-ratio-for-speed 14004The maximum code size growth ratio when expanding 14005into a jump table (in percent). The parameter is used when 14006optimizing for speed. 14007 14008@item tree-reassoc-width 14009Set the maximum number of instructions executed in parallel in 14010reassociated tree. This parameter overrides target dependent 14011heuristics used by default if has non zero value. 14012 14013@item sched-pressure-algorithm 14014Choose between the two available implementations of 14015@option{-fsched-pressure}. Algorithm 1 is the original implementation 14016and is the more likely to prevent instructions from being reordered. 14017Algorithm 2 was designed to be a compromise between the relatively 14018conservative approach taken by algorithm 1 and the rather aggressive 14019approach taken by the default scheduler. It relies more heavily on 14020having a regular register file and accurate register pressure classes. 14021See @file{haifa-sched.c} in the GCC sources for more details. 14022 14023The default choice depends on the target. 14024 14025@item max-slsr-cand-scan 14026Set the maximum number of existing candidates that are considered when 14027seeking a basis for a new straight-line strength reduction candidate. 14028 14029@item asan-globals 14030Enable buffer overflow detection for global objects. This kind 14031of protection is enabled by default if you are using 14032@option{-fsanitize=address} option. 14033To disable global objects protection use @option{--param asan-globals=0}. 14034 14035@item asan-stack 14036Enable buffer overflow detection for stack objects. This kind of 14037protection is enabled by default when using @option{-fsanitize=address}. 14038To disable stack protection use @option{--param asan-stack=0} option. 14039 14040@item asan-instrument-reads 14041Enable buffer overflow detection for memory reads. This kind of 14042protection is enabled by default when using @option{-fsanitize=address}. 14043To disable memory reads protection use 14044@option{--param asan-instrument-reads=0}. 14045 14046@item asan-instrument-writes 14047Enable buffer overflow detection for memory writes. This kind of 14048protection is enabled by default when using @option{-fsanitize=address}. 14049To disable memory writes protection use 14050@option{--param asan-instrument-writes=0} option. 14051 14052@item asan-memintrin 14053Enable detection for built-in functions. This kind of protection 14054is enabled by default when using @option{-fsanitize=address}. 14055To disable built-in functions protection use 14056@option{--param asan-memintrin=0}. 14057 14058@item asan-use-after-return 14059Enable detection of use-after-return. This kind of protection 14060is enabled by default when using the @option{-fsanitize=address} option. 14061To disable it use @option{--param asan-use-after-return=0}. 14062 14063Note: By default the check is disabled at run time. To enable it, 14064add @code{detect_stack_use_after_return=1} to the environment variable 14065@env{ASAN_OPTIONS}. 14066 14067@item asan-instrumentation-with-call-threshold 14068If number of memory accesses in function being instrumented 14069is greater or equal to this number, use callbacks instead of inline checks. 14070E.g. to disable inline code use 14071@option{--param asan-instrumentation-with-call-threshold=0}. 14072 14073@item hwasan-instrument-stack 14074Enable hwasan instrumentation of statically sized stack-allocated variables. 14075This kind of instrumentation is enabled by default when using 14076@option{-fsanitize=hwaddress} and disabled by default when using 14077@option{-fsanitize=kernel-hwaddress}. 14078To disable stack instrumentation use 14079@option{--param hwasan-instrument-stack=0}, and to enable it use 14080@option{--param hwasan-instrument-stack=1}. 14081 14082@item hwasan-random-frame-tag 14083When using stack instrumentation, decide tags for stack variables using a 14084deterministic sequence beginning at a random tag for each frame. With this 14085parameter unset tags are chosen using the same sequence but beginning from 1. 14086This is enabled by default for @option{-fsanitize=hwaddress} and unavailable 14087for @option{-fsanitize=kernel-hwaddress}. 14088To disable it use @option{--param hwasan-random-frame-tag=0}. 14089 14090@item hwasan-instrument-allocas 14091Enable hwasan instrumentation of dynamically sized stack-allocated variables. 14092This kind of instrumentation is enabled by default when using 14093@option{-fsanitize=hwaddress} and disabled by default when using 14094@option{-fsanitize=kernel-hwaddress}. 14095To disable instrumentation of such variables use 14096@option{--param hwasan-instrument-allocas=0}, and to enable it use 14097@option{--param hwasan-instrument-allocas=1}. 14098 14099@item hwasan-instrument-reads 14100Enable hwasan checks on memory reads. Instrumentation of reads is enabled by 14101default for both @option{-fsanitize=hwaddress} and 14102@option{-fsanitize=kernel-hwaddress}. 14103To disable checking memory reads use 14104@option{--param hwasan-instrument-reads=0}. 14105 14106@item hwasan-instrument-writes 14107Enable hwasan checks on memory writes. Instrumentation of writes is enabled by 14108default for both @option{-fsanitize=hwaddress} and 14109@option{-fsanitize=kernel-hwaddress}. 14110To disable checking memory writes use 14111@option{--param hwasan-instrument-writes=0}. 14112 14113@item hwasan-instrument-mem-intrinsics 14114Enable hwasan instrumentation of builtin functions. Instrumentation of these 14115builtin functions is enabled by default for both @option{-fsanitize=hwaddress} 14116and @option{-fsanitize=kernel-hwaddress}. 14117To disable instrumentation of builtin functions use 14118@option{--param hwasan-instrument-mem-intrinsics=0}. 14119 14120@item use-after-scope-direct-emission-threshold 14121If the size of a local variable in bytes is smaller or equal to this 14122number, directly poison (or unpoison) shadow memory instead of using 14123run-time callbacks. 14124 14125@item tsan-distinguish-volatile 14126Emit special instrumentation for accesses to volatiles. 14127 14128@item tsan-instrument-func-entry-exit 14129Emit instrumentation calls to __tsan_func_entry() and __tsan_func_exit(). 14130 14131@item max-fsm-thread-path-insns 14132Maximum number of instructions to copy when duplicating blocks on a 14133finite state automaton jump thread path. 14134 14135@item max-fsm-thread-length 14136Maximum number of basic blocks on a finite state automaton jump thread 14137path. 14138 14139@item max-fsm-thread-paths 14140Maximum number of new jump thread paths to create for a finite state 14141automaton. 14142 14143@item parloops-chunk-size 14144Chunk size of omp schedule for loops parallelized by parloops. 14145 14146@item parloops-schedule 14147Schedule type of omp schedule for loops parallelized by parloops (static, 14148dynamic, guided, auto, runtime). 14149 14150@item parloops-min-per-thread 14151The minimum number of iterations per thread of an innermost parallelized 14152loop for which the parallelized variant is preferred over the single threaded 14153one. Note that for a parallelized loop nest the 14154minimum number of iterations of the outermost loop per thread is two. 14155 14156@item max-ssa-name-query-depth 14157Maximum depth of recursion when querying properties of SSA names in things 14158like fold routines. One level of recursion corresponds to following a 14159use-def chain. 14160 14161@item max-speculative-devirt-maydefs 14162The maximum number of may-defs we analyze when looking for a must-def 14163specifying the dynamic type of an object that invokes a virtual call 14164we may be able to devirtualize speculatively. 14165 14166@item max-vrp-switch-assertions 14167The maximum number of assertions to add along the default edge of a switch 14168statement during VRP. 14169 14170@item evrp-mode 14171Specifies the mode Early VRP should operate in. 14172 14173@item unroll-jam-min-percent 14174The minimum percentage of memory references that must be optimized 14175away for the unroll-and-jam transformation to be considered profitable. 14176 14177@item unroll-jam-max-unroll 14178The maximum number of times the outer loop should be unrolled by 14179the unroll-and-jam transformation. 14180 14181@item max-rtl-if-conversion-unpredictable-cost 14182Maximum permissible cost for the sequence that would be generated 14183by the RTL if-conversion pass for a branch that is considered unpredictable. 14184 14185@item max-variable-expansions-in-unroller 14186If @option{-fvariable-expansion-in-unroller} is used, the maximum number 14187of times that an individual variable will be expanded during loop unrolling. 14188 14189@item tracer-min-branch-probability-feedback 14190Stop forward growth if the probability of best edge is less than 14191this threshold (in percent). Used when profile feedback is available. 14192 14193@item partial-inlining-entry-probability 14194Maximum probability of the entry BB of split region 14195(in percent relative to entry BB of the function) 14196to make partial inlining happen. 14197 14198@item max-tracked-strlens 14199Maximum number of strings for which strlen optimization pass will 14200track string lengths. 14201 14202@item gcse-after-reload-partial-fraction 14203The threshold ratio for performing partial redundancy 14204elimination after reload. 14205 14206@item gcse-after-reload-critical-fraction 14207The threshold ratio of critical edges execution count that 14208permit performing redundancy elimination after reload. 14209 14210@item max-loop-header-insns 14211The maximum number of insns in loop header duplicated 14212by the copy loop headers pass. 14213 14214@item vect-epilogues-nomask 14215Enable loop epilogue vectorization using smaller vector size. 14216 14217@item vect-partial-vector-usage 14218Controls when the loop vectorizer considers using partial vector loads 14219and stores as an alternative to falling back to scalar code. 0 stops 14220the vectorizer from ever using partial vector loads and stores. 1 allows 14221partial vector loads and stores if vectorization removes the need for the 14222code to iterate. 2 allows partial vector loads and stores in all loops. 14223The parameter only has an effect on targets that support partial 14224vector loads and stores. 14225 14226@item avoid-fma-max-bits 14227Maximum number of bits for which we avoid creating FMAs. 14228 14229@item sms-loop-average-count-threshold 14230A threshold on the average loop count considered by the swing modulo scheduler. 14231 14232@item sms-dfa-history 14233The number of cycles the swing modulo scheduler considers when checking 14234conflicts using DFA. 14235 14236@item max-inline-insns-recursive-auto 14237The maximum number of instructions non-inline function 14238can grow to via recursive inlining. 14239 14240@item graphite-allow-codegen-errors 14241Whether codegen errors should be ICEs when @option{-fchecking}. 14242 14243@item sms-max-ii-factor 14244A factor for tuning the upper bound that swing modulo scheduler 14245uses for scheduling a loop. 14246 14247@item lra-max-considered-reload-pseudos 14248The max number of reload pseudos which are considered during 14249spilling a non-reload pseudo. 14250 14251@item max-pow-sqrt-depth 14252Maximum depth of sqrt chains to use when synthesizing exponentiation 14253by a real constant. 14254 14255@item max-dse-active-local-stores 14256Maximum number of active local stores in RTL dead store elimination. 14257 14258@item asan-instrument-allocas 14259Enable asan allocas/VLAs protection. 14260 14261@item max-iterations-computation-cost 14262Bound on the cost of an expression to compute the number of iterations. 14263 14264@item max-isl-operations 14265Maximum number of isl operations, 0 means unlimited. 14266 14267@item graphite-max-arrays-per-scop 14268Maximum number of arrays per scop. 14269 14270@item max-vartrack-reverse-op-size 14271Max. size of loc list for which reverse ops should be added. 14272 14273@item tracer-dynamic-coverage-feedback 14274The percentage of function, weighted by execution frequency, 14275that must be covered by trace formation. 14276Used when profile feedback is available. 14277 14278@item max-inline-recursive-depth-auto 14279The maximum depth of recursive inlining for non-inline functions. 14280 14281@item fsm-scale-path-stmts 14282Scale factor to apply to the number of statements in a threading path 14283when comparing to the number of (scaled) blocks. 14284 14285@item fsm-maximum-phi-arguments 14286Maximum number of arguments a PHI may have before the FSM threader 14287will not try to thread through its block. 14288 14289@item uninit-control-dep-attempts 14290Maximum number of nested calls to search for control dependencies 14291during uninitialized variable analysis. 14292 14293@item sra-max-scalarization-size-Osize 14294Maximum size, in storage units, of an aggregate 14295which should be considered for scalarization when compiling for size. 14296 14297@item fsm-scale-path-blocks 14298Scale factor to apply to the number of blocks in a threading path 14299when comparing to the number of (scaled) statements. 14300 14301@item sched-autopref-queue-depth 14302Hardware autoprefetcher scheduler model control flag. 14303Number of lookahead cycles the model looks into; at ' 14304' only enable instruction sorting heuristic. 14305 14306@item loop-versioning-max-inner-insns 14307The maximum number of instructions that an inner loop can have 14308before the loop versioning pass considers it too big to copy. 14309 14310@item loop-versioning-max-outer-insns 14311The maximum number of instructions that an outer loop can have 14312before the loop versioning pass considers it too big to copy, 14313discounting any instructions in inner loops that directly benefit 14314from versioning. 14315 14316@item ssa-name-def-chain-limit 14317The maximum number of SSA_NAME assignments to follow in determining 14318a property of a variable such as its value. This limits the number 14319of iterations or recursive calls GCC performs when optimizing certain 14320statements or when determining their validity prior to issuing 14321diagnostics. 14322 14323@item store-merging-max-size 14324Maximum size of a single store merging region in bytes. 14325 14326@item hash-table-verification-limit 14327The number of elements for which hash table verification is done 14328for each searched element. 14329 14330@item max-find-base-term-values 14331Maximum number of VALUEs handled during a single find_base_term call. 14332 14333@item analyzer-max-enodes-per-program-point 14334The maximum number of exploded nodes per program point within 14335the analyzer, before terminating analysis of that point. 14336 14337@item analyzer-max-constraints 14338The maximum number of constraints per state. 14339 14340@item analyzer-min-snodes-for-call-summary 14341The minimum number of supernodes within a function for the 14342analyzer to consider summarizing its effects at call sites. 14343 14344@item analyzer-max-enodes-for-full-dump 14345The maximum depth of exploded nodes that should appear in a dot dump 14346before switching to a less verbose format. 14347 14348@item analyzer-max-recursion-depth 14349The maximum number of times a callsite can appear in a call stack 14350within the analyzer, before terminating analysis of a call that would 14351recurse deeper. 14352 14353@item analyzer-max-svalue-depth 14354The maximum depth of a symbolic value, before approximating 14355the value as unknown. 14356 14357@item analyzer-max-infeasible-edges 14358The maximum number of infeasible edges to reject before declaring 14359a diagnostic as infeasible. 14360 14361@item gimple-fe-computed-hot-bb-threshold 14362The number of executions of a basic block which is considered hot. 14363The parameter is used only in GIMPLE FE. 14364 14365@item analyzer-bb-explosion-factor 14366The maximum number of 'after supernode' exploded nodes within the analyzer 14367per supernode, before terminating analysis. 14368 14369@item ranger-logical-depth 14370Maximum depth of logical expression evaluation ranger will look through 14371when evaluating outgoing edge ranges. 14372 14373@item openacc-kernels 14374Specify mode of OpenACC `kernels' constructs handling. 14375With @option{--param=openacc-kernels=decompose}, OpenACC `kernels' 14376constructs are decomposed into parts, a sequence of compute 14377constructs, each then handled individually. 14378This is work in progress. 14379With @option{--param=openacc-kernels=parloops}, OpenACC `kernels' 14380constructs are handled by the @samp{parloops} pass, en bloc. 14381This is the current default. 14382 14383@end table 14384 14385The following choices of @var{name} are available on AArch64 targets: 14386 14387@table @gcctabopt 14388@item aarch64-sve-compare-costs 14389When vectorizing for SVE, consider using ``unpacked'' vectors for 14390smaller elements and use the cost model to pick the cheapest approach. 14391Also use the cost model to choose between SVE and Advanced SIMD vectorization. 14392 14393Using unpacked vectors includes storing smaller elements in larger 14394containers and accessing elements with extending loads and truncating 14395stores. 14396 14397@item aarch64-float-recp-precision 14398The number of Newton iterations for calculating the reciprocal for float type. 14399The precision of division is proportional to this param when division 14400approximation is enabled. The default value is 1. 14401 14402@item aarch64-double-recp-precision 14403The number of Newton iterations for calculating the reciprocal for double type. 14404The precision of division is propotional to this param when division 14405approximation is enabled. The default value is 2. 14406 14407@item aarch64-autovec-preference 14408Force an ISA selection strategy for auto-vectorization. Accepts values from 144090 to 4, inclusive. 14410@table @samp 14411@item 0 14412Use the default heuristics. 14413@item 1 14414Use only Advanced SIMD for auto-vectorization. 14415@item 2 14416Use only SVE for auto-vectorization. 14417@item 3 14418Use both Advanced SIMD and SVE. Prefer Advanced SIMD when the costs are 14419deemed equal. 14420@item 4 14421Use both Advanced SIMD and SVE. Prefer SVE when the costs are deemed equal. 14422@end table 14423The default value is 0. 14424 14425@item aarch64-loop-vect-issue-rate-niters 14426The tuning for some AArch64 CPUs tries to take both latencies and issue 14427rates into account when deciding whether a loop should be vectorized 14428using SVE, vectorized using Advanced SIMD, or not vectorized at all. 14429If this parameter is set to @var{n}, GCC will not use this heuristic 14430for loops that are known to execute in fewer than @var{n} Advanced 14431SIMD iterations. 14432 14433@end table 14434 14435@end table 14436 14437@node Instrumentation Options 14438@section Program Instrumentation Options 14439@cindex instrumentation options 14440@cindex program instrumentation options 14441@cindex run-time error checking options 14442@cindex profiling options 14443@cindex options, program instrumentation 14444@cindex options, run-time error checking 14445@cindex options, profiling 14446 14447GCC supports a number of command-line options that control adding 14448run-time instrumentation to the code it normally generates. 14449For example, one purpose of instrumentation is collect profiling 14450statistics for use in finding program hot spots, code coverage 14451analysis, or profile-guided optimizations. 14452Another class of program instrumentation is adding run-time checking 14453to detect programming errors like invalid pointer 14454dereferences or out-of-bounds array accesses, as well as deliberately 14455hostile attacks such as stack smashing or C++ vtable hijacking. 14456There is also a general hook which can be used to implement other 14457forms of tracing or function-level instrumentation for debug or 14458program analysis purposes. 14459 14460@table @gcctabopt 14461@cindex @command{prof} 14462@cindex @command{gprof} 14463@item -p 14464@itemx -pg 14465@opindex p 14466@opindex pg 14467Generate extra code to write profile information suitable for the 14468analysis program @command{prof} (for @option{-p}) or @command{gprof} 14469(for @option{-pg}). You must use this option when compiling 14470the source files you want data about, and you must also use it when 14471linking. 14472 14473You can use the function attribute @code{no_instrument_function} to 14474suppress profiling of individual functions when compiling with these options. 14475@xref{Common Function Attributes}. 14476 14477@item -fprofile-arcs 14478@opindex fprofile-arcs 14479Add code so that program flow @dfn{arcs} are instrumented. During 14480execution the program records how many times each branch and call is 14481executed and how many times it is taken or returns. On targets that support 14482constructors with priority support, profiling properly handles constructors, 14483destructors and C++ constructors (and destructors) of classes which are used 14484as a type of a global variable. 14485 14486When the compiled 14487program exits it saves this data to a file called 14488@file{@var{auxname}.gcda} for each source file. The data may be used for 14489profile-directed optimizations (@option{-fbranch-probabilities}), or for 14490test coverage analysis (@option{-ftest-coverage}). Each object file's 14491@var{auxname} is generated from the name of the output file, if 14492explicitly specified and it is not the final executable, otherwise it is 14493the basename of the source file. In both cases any suffix is removed 14494(e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or 14495@file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}). 14496@xref{Cross-profiling}. 14497 14498@cindex @command{gcov} 14499@item --coverage 14500@opindex coverage 14501 14502This option is used to compile and link code instrumented for coverage 14503analysis. The option is a synonym for @option{-fprofile-arcs} 14504@option{-ftest-coverage} (when compiling) and @option{-lgcov} (when 14505linking). See the documentation for those options for more details. 14506 14507@itemize 14508 14509@item 14510Compile the source files with @option{-fprofile-arcs} plus optimization 14511and code generation options. For test coverage analysis, use the 14512additional @option{-ftest-coverage} option. You do not need to profile 14513every source file in a program. 14514 14515@item 14516Compile the source files additionally with @option{-fprofile-abs-path} 14517to create absolute path names in the @file{.gcno} files. This allows 14518@command{gcov} to find the correct sources in projects where compilations 14519occur with different working directories. 14520 14521@item 14522Link your object files with @option{-lgcov} or @option{-fprofile-arcs} 14523(the latter implies the former). 14524 14525@item 14526Run the program on a representative workload to generate the arc profile 14527information. This may be repeated any number of times. You can run 14528concurrent instances of your program, and provided that the file system 14529supports locking, the data files will be correctly updated. Unless 14530a strict ISO C dialect option is in effect, @code{fork} calls are 14531detected and correctly handled without double counting. 14532 14533@item 14534For profile-directed optimizations, compile the source files again with 14535the same optimization and code generation options plus 14536@option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that 14537Control Optimization}). 14538 14539@item 14540For test coverage analysis, use @command{gcov} to produce human readable 14541information from the @file{.gcno} and @file{.gcda} files. Refer to the 14542@command{gcov} documentation for further information. 14543 14544@end itemize 14545 14546With @option{-fprofile-arcs}, for each function of your program GCC 14547creates a program flow graph, then finds a spanning tree for the graph. 14548Only arcs that are not on the spanning tree have to be instrumented: the 14549compiler adds code to count the number of times that these arcs are 14550executed. When an arc is the only exit or only entrance to a block, the 14551instrumentation code can be added to the block; otherwise, a new basic 14552block must be created to hold the instrumentation code. 14553 14554@need 2000 14555@item -ftest-coverage 14556@opindex ftest-coverage 14557Produce a notes file that the @command{gcov} code-coverage utility 14558(@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to 14559show program coverage. Each source file's note file is called 14560@file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option 14561above for a description of @var{auxname} and instructions on how to 14562generate test coverage data. Coverage data matches the source files 14563more closely if you do not optimize. 14564 14565@item -fprofile-abs-path 14566@opindex fprofile-abs-path 14567Automatically convert relative source file names to absolute path names 14568in the @file{.gcno} files. This allows @command{gcov} to find the correct 14569sources in projects where compilations occur with different working 14570directories. 14571 14572@item -fprofile-dir=@var{path} 14573@opindex fprofile-dir 14574 14575Set the directory to search for the profile data files in to @var{path}. 14576This option affects only the profile data generated by 14577@option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs} 14578and used by @option{-fprofile-use} and @option{-fbranch-probabilities} 14579and its related options. Both absolute and relative paths can be used. 14580By default, GCC uses the current directory as @var{path}, thus the 14581profile data file appears in the same directory as the object file. 14582In order to prevent the file name clashing, if the object file name is 14583not an absolute path, we mangle the absolute path of the 14584@file{@var{sourcename}.gcda} file and use it as the file name of a 14585@file{.gcda} file. See similar option @option{-fprofile-note}. 14586 14587When an executable is run in a massive parallel environment, it is recommended 14588to save profile to different folders. That can be done with variables 14589in @var{path} that are exported during run-time: 14590 14591@table @gcctabopt 14592 14593@item %p 14594process ID. 14595 14596@item %q@{VAR@} 14597value of environment variable @var{VAR} 14598 14599@end table 14600 14601@item -fprofile-generate 14602@itemx -fprofile-generate=@var{path} 14603@opindex fprofile-generate 14604 14605Enable options usually used for instrumenting application to produce 14606profile useful for later recompilation with profile feedback based 14607optimization. You must use @option{-fprofile-generate} both when 14608compiling and when linking your program. 14609 14610The following options are enabled: 14611@option{-fprofile-arcs}, @option{-fprofile-values}, 14612@option{-finline-functions}, and @option{-fipa-bit-cp}. 14613 14614If @var{path} is specified, GCC looks at the @var{path} to find 14615the profile feedback data files. See @option{-fprofile-dir}. 14616 14617To optimize the program based on the collected profile information, use 14618@option{-fprofile-use}. @xref{Optimize Options}, for more information. 14619 14620@item -fprofile-info-section 14621@itemx -fprofile-info-section=@var{name} 14622@opindex fprofile-info-section 14623 14624Register the profile information in the specified section instead of using a 14625constructor/destructor. The section name is @var{name} if it is specified, 14626otherwise the section name defaults to @code{.gcov_info}. A pointer to the 14627profile information generated by @option{-fprofile-arcs} or 14628@option{-ftest-coverage} is placed in the specified section for each 14629translation unit. This option disables the profile information registration 14630through a constructor and it disables the profile information processing 14631through a destructor. This option is not intended to be used in hosted 14632environments such as GNU/Linux. It targets systems with limited resources 14633which do not support constructors and destructors. The linker could collect 14634the input sections in a continuous memory block and define start and end 14635symbols. The runtime support could dump the profiling information registered 14636in this linker set during program termination to a serial line for example. A 14637GNU linker script example which defines a linker output section follows: 14638 14639@smallexample 14640 .gcov_info : 14641 @{ 14642 PROVIDE (__gcov_info_start = .); 14643 KEEP (*(.gcov_info)) 14644 PROVIDE (__gcov_info_end = .); 14645 @} 14646@end smallexample 14647 14648@item -fprofile-note=@var{path} 14649@opindex fprofile-note 14650 14651If @var{path} is specified, GCC saves @file{.gcno} file into @var{path} 14652location. If you combine the option with multiple source files, 14653the @file{.gcno} file will be overwritten. 14654 14655@item -fprofile-prefix-path=@var{path} 14656@opindex fprofile-prefix-path 14657 14658This option can be used in combination with 14659@option{profile-generate=}@var{profile_dir} and 14660@option{profile-use=}@var{profile_dir} to inform GCC where is the base 14661directory of built source tree. By default @var{profile_dir} will contain 14662files with mangled absolute paths of all object files in the built project. 14663This is not desirable when directory used to build the instrumented binary 14664differs from the directory used to build the binary optimized with profile 14665feedback because the profile data will not be found during the optimized build. 14666In such setups @option{-fprofile-prefix-path=}@var{path} with @var{path} 14667pointing to the base directory of the build can be used to strip the irrelevant 14668part of the path and keep all file names relative to the main build directory. 14669 14670@item -fprofile-update=@var{method} 14671@opindex fprofile-update 14672 14673Alter the update method for an application instrumented for profile 14674feedback based optimization. The @var{method} argument should be one of 14675@samp{single}, @samp{atomic} or @samp{prefer-atomic}. 14676The first one is useful for single-threaded applications, 14677while the second one prevents profile corruption by emitting thread-safe code. 14678 14679@strong{Warning:} When an application does not properly join all threads 14680(or creates an detached thread), a profile file can be still corrupted. 14681 14682Using @samp{prefer-atomic} would be transformed either to @samp{atomic}, 14683when supported by a target, or to @samp{single} otherwise. The GCC driver 14684automatically selects @samp{prefer-atomic} when @option{-pthread} 14685is present in the command line. 14686 14687@item -fprofile-filter-files=@var{regex} 14688@opindex fprofile-filter-files 14689 14690Instrument only functions from files whose name matches 14691any of the regular expressions (separated by semi-colons). 14692 14693For example, @option{-fprofile-filter-files=main\.c;module.*\.c} will instrument 14694only @file{main.c} and all C files starting with 'module'. 14695 14696@item -fprofile-exclude-files=@var{regex} 14697@opindex fprofile-exclude-files 14698 14699Instrument only functions from files whose name does not match 14700any of the regular expressions (separated by semi-colons). 14701 14702For example, @option{-fprofile-exclude-files=/usr/.*} will prevent instrumentation 14703of all files that are located in the @file{/usr/} folder. 14704 14705@item -fprofile-reproducible=@r{[}multithreaded@r{|}parallel-runs@r{|}serial@r{]} 14706@opindex fprofile-reproducible 14707Control level of reproducibility of profile gathered by 14708@code{-fprofile-generate}. This makes it possible to rebuild program 14709with same outcome which is useful, for example, for distribution 14710packages. 14711 14712With @option{-fprofile-reproducible=serial} the profile gathered by 14713@option{-fprofile-generate} is reproducible provided the trained program 14714behaves the same at each invocation of the train run, it is not 14715multi-threaded and profile data streaming is always done in the same 14716order. Note that profile streaming happens at the end of program run but 14717also before @code{fork} function is invoked. 14718 14719Note that it is quite common that execution counts of some part of 14720programs depends, for example, on length of temporary file names or 14721memory space randomization (that may affect hash-table collision rate). 14722Such non-reproducible part of programs may be annotated by 14723@code{no_instrument_function} function attribute. @command{gcov-dump} with 14724@option{-l} can be used to dump gathered data and verify that they are 14725indeed reproducible. 14726 14727With @option{-fprofile-reproducible=parallel-runs} collected profile 14728stays reproducible regardless the order of streaming of the data into 14729gcda files. This setting makes it possible to run multiple instances of 14730instrumented program in parallel (such as with @code{make -j}). This 14731reduces quality of gathered data, in particular of indirect call 14732profiling. 14733 14734@item -fsanitize=address 14735@opindex fsanitize=address 14736Enable AddressSanitizer, a fast memory error detector. 14737Memory access instructions are instrumented to detect 14738out-of-bounds and use-after-free bugs. 14739The option enables @option{-fsanitize-address-use-after-scope}. 14740See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for 14741more details. The run-time behavior can be influenced using the 14742@env{ASAN_OPTIONS} environment variable. When set to @code{help=1}, 14743the available options are shown at startup of the instrumented program. See 14744@url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags} 14745for a list of supported options. 14746The option cannot be combined with @option{-fsanitize=thread} or 14747@option{-fsanitize=hwaddress}. Note that the only target 14748@option{-fsanitize=hwaddress} is currently supported on is AArch64. 14749 14750@item -fsanitize=kernel-address 14751@opindex fsanitize=kernel-address 14752Enable AddressSanitizer for Linux kernel. 14753See @uref{https://github.com/google/kasan} for more details. 14754 14755@item -fsanitize=hwaddress 14756@opindex fsanitize=hwaddress 14757Enable Hardware-assisted AddressSanitizer, which uses a hardware ability to 14758ignore the top byte of a pointer to allow the detection of memory errors with 14759a low memory overhead. 14760Memory access instructions are instrumented to detect out-of-bounds and 14761use-after-free bugs. 14762The option enables @option{-fsanitize-address-use-after-scope}. 14763See 14764@uref{https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html} 14765for more details. The run-time behavior can be influenced using the 14766@env{HWASAN_OPTIONS} environment variable. When set to @code{help=1}, 14767the available options are shown at startup of the instrumented program. 14768The option cannot be combined with @option{-fsanitize=thread} or 14769@option{-fsanitize=address}, and is currently only available on AArch64. 14770 14771@item -fsanitize=kernel-hwaddress 14772@opindex fsanitize=kernel-hwaddress 14773Enable Hardware-assisted AddressSanitizer for compilation of the Linux kernel. 14774Similar to @option{-fsanitize=kernel-address} but using an alternate 14775instrumentation method, and similar to @option{-fsanitize=hwaddress} but with 14776instrumentation differences necessary for compiling the Linux kernel. 14777These differences are to avoid hwasan library initialization calls and to 14778account for the stack pointer having a different value in its top byte. 14779 14780@emph{Note:} This option has different defaults to the @option{-fsanitize=hwaddress}. 14781Instrumenting the stack and alloca calls are not on by default but are still 14782possible by specifying the command-line options 14783@option{--param hwasan-instrument-stack=1} and 14784@option{--param hwasan-instrument-allocas=1} respectively. Using a random frame 14785tag is not implemented for kernel instrumentation. 14786 14787@item -fsanitize=pointer-compare 14788@opindex fsanitize=pointer-compare 14789Instrument comparison operation (<, <=, >, >=) with pointer operands. 14790The option must be combined with either @option{-fsanitize=kernel-address} or 14791@option{-fsanitize=address} 14792The option cannot be combined with @option{-fsanitize=thread}. 14793Note: By default the check is disabled at run time. To enable it, 14794add @code{detect_invalid_pointer_pairs=2} to the environment variable 14795@env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects 14796invalid operation only when both pointers are non-null. 14797 14798@item -fsanitize=pointer-subtract 14799@opindex fsanitize=pointer-subtract 14800Instrument subtraction with pointer operands. 14801The option must be combined with either @option{-fsanitize=kernel-address} or 14802@option{-fsanitize=address} 14803The option cannot be combined with @option{-fsanitize=thread}. 14804Note: By default the check is disabled at run time. To enable it, 14805add @code{detect_invalid_pointer_pairs=2} to the environment variable 14806@env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects 14807invalid operation only when both pointers are non-null. 14808 14809@item -fsanitize=thread 14810@opindex fsanitize=thread 14811Enable ThreadSanitizer, a fast data race detector. 14812Memory access instructions are instrumented to detect 14813data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more 14814details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS} 14815environment variable; see 14816@url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of 14817supported options. 14818The option cannot be combined with @option{-fsanitize=address}, 14819@option{-fsanitize=leak}. 14820 14821Note that sanitized atomic builtins cannot throw exceptions when 14822operating on invalid memory addresses with non-call exceptions 14823(@option{-fnon-call-exceptions}). 14824 14825@item -fsanitize=leak 14826@opindex fsanitize=leak 14827Enable LeakSanitizer, a memory leak detector. 14828This option only matters for linking of executables and 14829the executable is linked against a library that overrides @code{malloc} 14830and other allocator functions. See 14831@uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more 14832details. The run-time behavior can be influenced using the 14833@env{LSAN_OPTIONS} environment variable. 14834The option cannot be combined with @option{-fsanitize=thread}. 14835 14836@item -fsanitize=undefined 14837@opindex fsanitize=undefined 14838Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector. 14839Various computations are instrumented to detect undefined behavior 14840at runtime. Current suboptions are: 14841 14842@table @gcctabopt 14843 14844@item -fsanitize=shift 14845@opindex fsanitize=shift 14846This option enables checking that the result of a shift operation is 14847not undefined. Note that what exactly is considered undefined differs 14848slightly between C and C++, as well as between ISO C90 and C99, etc. 14849This option has two suboptions, @option{-fsanitize=shift-base} and 14850@option{-fsanitize=shift-exponent}. 14851 14852@item -fsanitize=shift-exponent 14853@opindex fsanitize=shift-exponent 14854This option enables checking that the second argument of a shift operation 14855is not negative and is smaller than the precision of the promoted first 14856argument. 14857 14858@item -fsanitize=shift-base 14859@opindex fsanitize=shift-base 14860If the second argument of a shift operation is within range, check that the 14861result of a shift operation is not undefined. Note that what exactly is 14862considered undefined differs slightly between C and C++, as well as between 14863ISO C90 and C99, etc. 14864 14865@item -fsanitize=integer-divide-by-zero 14866@opindex fsanitize=integer-divide-by-zero 14867Detect integer division by zero as well as @code{INT_MIN / -1} division. 14868 14869@item -fsanitize=unreachable 14870@opindex fsanitize=unreachable 14871With this option, the compiler turns the @code{__builtin_unreachable} 14872call into a diagnostics message call instead. When reaching the 14873@code{__builtin_unreachable} call, the behavior is undefined. 14874 14875@item -fsanitize=vla-bound 14876@opindex fsanitize=vla-bound 14877This option instructs the compiler to check that the size of a variable 14878length array is positive. 14879 14880@item -fsanitize=null 14881@opindex fsanitize=null 14882This option enables pointer checking. Particularly, the application 14883built with this option turned on will issue an error message when it 14884tries to dereference a NULL pointer, or if a reference (possibly an 14885rvalue reference) is bound to a NULL pointer, or if a method is invoked 14886on an object pointed by a NULL pointer. 14887 14888@item -fsanitize=return 14889@opindex fsanitize=return 14890This option enables return statement checking. Programs 14891built with this option turned on will issue an error message 14892when the end of a non-void function is reached without actually 14893returning a value. This option works in C++ only. 14894 14895@item -fsanitize=signed-integer-overflow 14896@opindex fsanitize=signed-integer-overflow 14897This option enables signed integer overflow checking. We check that 14898the result of @code{+}, @code{*}, and both unary and binary @code{-} 14899does not overflow in the signed arithmetics. Note, integer promotion 14900rules must be taken into account. That is, the following is not an 14901overflow: 14902@smallexample 14903signed char a = SCHAR_MAX; 14904a++; 14905@end smallexample 14906 14907@item -fsanitize=bounds 14908@opindex fsanitize=bounds 14909This option enables instrumentation of array bounds. Various out of bounds 14910accesses are detected. Flexible array members, flexible array member-like 14911arrays, and initializers of variables with static storage are not instrumented. 14912 14913@item -fsanitize=bounds-strict 14914@opindex fsanitize=bounds-strict 14915This option enables strict instrumentation of array bounds. Most out of bounds 14916accesses are detected, including flexible array members and flexible array 14917member-like arrays. Initializers of variables with static storage are not 14918instrumented. 14919 14920@item -fsanitize=alignment 14921@opindex fsanitize=alignment 14922 14923This option enables checking of alignment of pointers when they are 14924dereferenced, or when a reference is bound to insufficiently aligned target, 14925or when a method or constructor is invoked on insufficiently aligned object. 14926 14927@item -fsanitize=object-size 14928@opindex fsanitize=object-size 14929This option enables instrumentation of memory references using the 14930@code{__builtin_object_size} function. Various out of bounds pointer 14931accesses are detected. 14932 14933@item -fsanitize=float-divide-by-zero 14934@opindex fsanitize=float-divide-by-zero 14935Detect floating-point division by zero. Unlike other similar options, 14936@option{-fsanitize=float-divide-by-zero} is not enabled by 14937@option{-fsanitize=undefined}, since floating-point division by zero can 14938be a legitimate way of obtaining infinities and NaNs. 14939 14940@item -fsanitize=float-cast-overflow 14941@opindex fsanitize=float-cast-overflow 14942This option enables floating-point type to integer conversion checking. 14943We check that the result of the conversion does not overflow. 14944Unlike other similar options, @option{-fsanitize=float-cast-overflow} is 14945not enabled by @option{-fsanitize=undefined}. 14946This option does not work well with @code{FE_INVALID} exceptions enabled. 14947 14948@item -fsanitize=nonnull-attribute 14949@opindex fsanitize=nonnull-attribute 14950 14951This option enables instrumentation of calls, checking whether null values 14952are not passed to arguments marked as requiring a non-null value by the 14953@code{nonnull} function attribute. 14954 14955@item -fsanitize=returns-nonnull-attribute 14956@opindex fsanitize=returns-nonnull-attribute 14957 14958This option enables instrumentation of return statements in functions 14959marked with @code{returns_nonnull} function attribute, to detect returning 14960of null values from such functions. 14961 14962@item -fsanitize=bool 14963@opindex fsanitize=bool 14964 14965This option enables instrumentation of loads from bool. If a value other 14966than 0/1 is loaded, a run-time error is issued. 14967 14968@item -fsanitize=enum 14969@opindex fsanitize=enum 14970 14971This option enables instrumentation of loads from an enum type. If 14972a value outside the range of values for the enum type is loaded, 14973a run-time error is issued. 14974 14975@item -fsanitize=vptr 14976@opindex fsanitize=vptr 14977 14978This option enables instrumentation of C++ member function calls, member 14979accesses and some conversions between pointers to base and derived classes, 14980to verify the referenced object has the correct dynamic type. 14981 14982@item -fsanitize=pointer-overflow 14983@opindex fsanitize=pointer-overflow 14984 14985This option enables instrumentation of pointer arithmetics. If the pointer 14986arithmetics overflows, a run-time error is issued. 14987 14988@item -fsanitize=builtin 14989@opindex fsanitize=builtin 14990 14991This option enables instrumentation of arguments to selected builtin 14992functions. If an invalid value is passed to such arguments, a run-time 14993error is issued. E.g.@ passing 0 as the argument to @code{__builtin_ctz} 14994or @code{__builtin_clz} invokes undefined behavior and is diagnosed 14995by this option. 14996 14997@end table 14998 14999While @option{-ftrapv} causes traps for signed overflows to be emitted, 15000@option{-fsanitize=undefined} gives a diagnostic message. 15001This currently works only for the C family of languages. 15002 15003@item -fno-sanitize=all 15004@opindex fno-sanitize=all 15005 15006This option disables all previously enabled sanitizers. 15007@option{-fsanitize=all} is not allowed, as some sanitizers cannot be used 15008together. 15009 15010@item -fasan-shadow-offset=@var{number} 15011@opindex fasan-shadow-offset 15012This option forces GCC to use custom shadow offset in AddressSanitizer checks. 15013It is useful for experimenting with different shadow memory layouts in 15014Kernel AddressSanitizer. 15015 15016@item -fsanitize-sections=@var{s1},@var{s2},... 15017@opindex fsanitize-sections 15018Sanitize global variables in selected user-defined sections. @var{si} may 15019contain wildcards. 15020 15021@item -fsanitize-recover@r{[}=@var{opts}@r{]} 15022@opindex fsanitize-recover 15023@opindex fno-sanitize-recover 15024@option{-fsanitize-recover=} controls error recovery mode for sanitizers 15025mentioned in comma-separated list of @var{opts}. Enabling this option 15026for a sanitizer component causes it to attempt to continue 15027running the program as if no error happened. This means multiple 15028runtime errors can be reported in a single program run, and the exit 15029code of the program may indicate success even when errors 15030have been reported. The @option{-fno-sanitize-recover=} option 15031can be used to alter 15032this behavior: only the first detected error is reported 15033and program then exits with a non-zero exit code. 15034 15035Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions 15036except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}), 15037@option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero}, 15038@option{-fsanitize=bounds-strict}, 15039@option{-fsanitize=kernel-address} and @option{-fsanitize=address}. 15040For these sanitizers error recovery is turned on by default, 15041except @option{-fsanitize=address}, for which this feature is experimental. 15042@option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also 15043accepted, the former enables recovery for all sanitizers that support it, 15044the latter disables recovery for all sanitizers that support it. 15045 15046Even if a recovery mode is turned on the compiler side, it needs to be also 15047enabled on the runtime library side, otherwise the failures are still fatal. 15048The runtime library defaults to @code{halt_on_error=0} for 15049ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for 15050AddressSanitizer is @code{halt_on_error=1}. This can be overridden through 15051setting the @code{halt_on_error} flag in the corresponding environment variable. 15052 15053Syntax without an explicit @var{opts} parameter is deprecated. It is 15054equivalent to specifying an @var{opts} list of: 15055 15056@smallexample 15057undefined,float-cast-overflow,float-divide-by-zero,bounds-strict 15058@end smallexample 15059 15060@item -fsanitize-address-use-after-scope 15061@opindex fsanitize-address-use-after-scope 15062Enable sanitization of local variables to detect use-after-scope bugs. 15063The option sets @option{-fstack-reuse} to @samp{none}. 15064 15065@item -fsanitize-undefined-trap-on-error 15066@opindex fsanitize-undefined-trap-on-error 15067The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to 15068report undefined behavior using @code{__builtin_trap} rather than 15069a @code{libubsan} library routine. The advantage of this is that the 15070@code{libubsan} library is not needed and is not linked in, so this 15071is usable even in freestanding environments. 15072 15073@item -fsanitize-coverage=trace-pc 15074@opindex fsanitize-coverage=trace-pc 15075Enable coverage-guided fuzzing code instrumentation. 15076Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block. 15077 15078@item -fsanitize-coverage=trace-cmp 15079@opindex fsanitize-coverage=trace-cmp 15080Enable dataflow guided fuzzing code instrumentation. 15081Inserts a call to @code{__sanitizer_cov_trace_cmp1}, 15082@code{__sanitizer_cov_trace_cmp2}, @code{__sanitizer_cov_trace_cmp4} or 15083@code{__sanitizer_cov_trace_cmp8} for integral comparison with both operands 15084variable or @code{__sanitizer_cov_trace_const_cmp1}, 15085@code{__sanitizer_cov_trace_const_cmp2}, 15086@code{__sanitizer_cov_trace_const_cmp4} or 15087@code{__sanitizer_cov_trace_const_cmp8} for integral comparison with one 15088operand constant, @code{__sanitizer_cov_trace_cmpf} or 15089@code{__sanitizer_cov_trace_cmpd} for float or double comparisons and 15090@code{__sanitizer_cov_trace_switch} for switch statements. 15091 15092@item -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{|}check@r{]} 15093@opindex fcf-protection 15094Enable code instrumentation of control-flow transfers to increase 15095program security by checking that target addresses of control-flow 15096transfer instructions (such as indirect function call, function return, 15097indirect jump) are valid. This prevents diverting the flow of control 15098to an unexpected target. This is intended to protect against such 15099threats as Return-oriented Programming (ROP), and similarly 15100call/jmp-oriented programming (COP/JOP). 15101 15102The value @code{branch} tells the compiler to implement checking of 15103validity of control-flow transfer at the point of indirect branch 15104instructions, i.e.@: call/jmp instructions. The value @code{return} 15105implements checking of validity at the point of returning from a 15106function. The value @code{full} is an alias for specifying both 15107@code{branch} and @code{return}. The value @code{none} turns off 15108instrumentation. 15109 15110The value @code{check} is used for the final link with link-time 15111optimization (LTO). An error is issued if LTO object files are 15112compiled with different @option{-fcf-protection} values. The 15113value @code{check} is ignored at the compile time. 15114 15115The macro @code{__CET__} is defined when @option{-fcf-protection} is 15116used. The first bit of @code{__CET__} is set to 1 for the value 15117@code{branch} and the second bit of @code{__CET__} is set to 1 for 15118the @code{return}. 15119 15120You can also use the @code{nocf_check} attribute to identify 15121which functions and calls should be skipped from instrumentation 15122(@pxref{Function Attributes}). 15123 15124Currently the x86 GNU/Linux target provides an implementation based 15125on Intel Control-flow Enforcement Technology (CET). 15126 15127@item -fstack-protector 15128@opindex fstack-protector 15129Emit extra code to check for buffer overflows, such as stack smashing 15130attacks. This is done by adding a guard variable to functions with 15131vulnerable objects. This includes functions that call @code{alloca}, and 15132functions with buffers larger than or equal to 8 bytes. The guards are 15133initialized when a function is entered and then checked when the function 15134exits. If a guard check fails, an error message is printed and the program 15135exits. Only variables that are actually allocated on the stack are 15136considered, optimized away variables or variables allocated in registers 15137don't count. 15138 15139@item -fstack-protector-all 15140@opindex fstack-protector-all 15141Like @option{-fstack-protector} except that all functions are protected. 15142 15143@item -fstack-protector-strong 15144@opindex fstack-protector-strong 15145Like @option{-fstack-protector} but includes additional functions to 15146be protected --- those that have local array definitions, or have 15147references to local frame addresses. Only variables that are actually 15148allocated on the stack are considered, optimized away variables or variables 15149allocated in registers don't count. 15150 15151@item -fstack-protector-explicit 15152@opindex fstack-protector-explicit 15153Like @option{-fstack-protector} but only protects those functions which 15154have the @code{stack_protect} attribute. 15155 15156@item -fstack-check 15157@opindex fstack-check 15158Generate code to verify that you do not go beyond the boundary of the 15159stack. You should specify this flag if you are running in an 15160environment with multiple threads, but you only rarely need to specify it in 15161a single-threaded environment since stack overflow is automatically 15162detected on nearly all systems if there is only one stack. 15163 15164Note that this switch does not actually cause checking to be done; the 15165operating system or the language runtime must do that. The switch causes 15166generation of code to ensure that they see the stack being extended. 15167 15168You can additionally specify a string parameter: @samp{no} means no 15169checking, @samp{generic} means force the use of old-style checking, 15170@samp{specific} means use the best checking method and is equivalent 15171to bare @option{-fstack-check}. 15172 15173Old-style checking is a generic mechanism that requires no specific 15174target support in the compiler but comes with the following drawbacks: 15175 15176@enumerate 15177@item 15178Modified allocation strategy for large objects: they are always 15179allocated dynamically if their size exceeds a fixed threshold. Note this 15180may change the semantics of some code. 15181 15182@item 15183Fixed limit on the size of the static frame of functions: when it is 15184topped by a particular function, stack checking is not reliable and 15185a warning is issued by the compiler. 15186 15187@item 15188Inefficiency: because of both the modified allocation strategy and the 15189generic implementation, code performance is hampered. 15190@end enumerate 15191 15192Note that old-style stack checking is also the fallback method for 15193@samp{specific} if no target support has been added in the compiler. 15194 15195@samp{-fstack-check=} is designed for Ada's needs to detect infinite recursion 15196and stack overflows. @samp{specific} is an excellent choice when compiling 15197Ada code. It is not generally sufficient to protect against stack-clash 15198attacks. To protect against those you want @samp{-fstack-clash-protection}. 15199 15200@item -fstack-clash-protection 15201@opindex fstack-clash-protection 15202Generate code to prevent stack clash style attacks. When this option is 15203enabled, the compiler will only allocate one page of stack space at a time 15204and each page is accessed immediately after allocation. Thus, it prevents 15205allocations from jumping over any stack guard page provided by the 15206operating system. 15207 15208Most targets do not fully support stack clash protection. However, on 15209those targets @option{-fstack-clash-protection} will protect dynamic stack 15210allocations. @option{-fstack-clash-protection} may also provide limited 15211protection for static stack allocations if the target supports 15212@option{-fstack-check=specific}. 15213 15214@item -fstack-limit-register=@var{reg} 15215@itemx -fstack-limit-symbol=@var{sym} 15216@itemx -fno-stack-limit 15217@opindex fstack-limit-register 15218@opindex fstack-limit-symbol 15219@opindex fno-stack-limit 15220Generate code to ensure that the stack does not grow beyond a certain value, 15221either the value of a register or the address of a symbol. If a larger 15222stack is required, a signal is raised at run time. For most targets, 15223the signal is raised before the stack overruns the boundary, so 15224it is possible to catch the signal without taking special precautions. 15225 15226For instance, if the stack starts at absolute address @samp{0x80000000} 15227and grows downwards, you can use the flags 15228@option{-fstack-limit-symbol=__stack_limit} and 15229@option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit 15230of 128KB@. Note that this may only work with the GNU linker. 15231 15232You can locally override stack limit checking by using the 15233@code{no_stack_limit} function attribute (@pxref{Function Attributes}). 15234 15235@item -fsplit-stack 15236@opindex fsplit-stack 15237Generate code to automatically split the stack before it overflows. 15238The resulting program has a discontiguous stack which can only 15239overflow if the program is unable to allocate any more memory. This 15240is most useful when running threaded programs, as it is no longer 15241necessary to calculate a good stack size to use for each thread. This 15242is currently only implemented for the x86 targets running 15243GNU/Linux. 15244 15245When code compiled with @option{-fsplit-stack} calls code compiled 15246without @option{-fsplit-stack}, there may not be much stack space 15247available for the latter code to run. If compiling all code, 15248including library code, with @option{-fsplit-stack} is not an option, 15249then the linker can fix up these calls so that the code compiled 15250without @option{-fsplit-stack} always has a large stack. Support for 15251this is implemented in the gold linker in GNU binutils release 2.21 15252and later. 15253 15254@item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} 15255@opindex fvtable-verify 15256This option is only available when compiling C++ code. 15257It turns on (or off, if using @option{-fvtable-verify=none}) the security 15258feature that verifies at run time, for every virtual call, that 15259the vtable pointer through which the call is made is valid for the type of 15260the object, and has not been corrupted or overwritten. If an invalid vtable 15261pointer is detected at run time, an error is reported and execution of the 15262program is immediately halted. 15263 15264This option causes run-time data structures to be built at program startup, 15265which are used for verifying the vtable pointers. 15266The options @samp{std} and @samp{preinit} 15267control the timing of when these data structures are built. In both cases the 15268data structures are built before execution reaches @code{main}. Using 15269@option{-fvtable-verify=std} causes the data structures to be built after 15270shared libraries have been loaded and initialized. 15271@option{-fvtable-verify=preinit} causes them to be built before shared 15272libraries have been loaded and initialized. 15273 15274If this option appears multiple times in the command line with different 15275values specified, @samp{none} takes highest priority over both @samp{std} and 15276@samp{preinit}; @samp{preinit} takes priority over @samp{std}. 15277 15278@item -fvtv-debug 15279@opindex fvtv-debug 15280When used in conjunction with @option{-fvtable-verify=std} or 15281@option{-fvtable-verify=preinit}, causes debug versions of the 15282runtime functions for the vtable verification feature to be called. 15283This flag also causes the compiler to log information about which 15284vtable pointers it finds for each class. 15285This information is written to a file named @file{vtv_set_ptr_data.log} 15286in the directory named by the environment variable @env{VTV_LOGS_DIR} 15287if that is defined or the current working directory otherwise. 15288 15289Note: This feature @emph{appends} data to the log file. If you want a fresh log 15290file, be sure to delete any existing one. 15291 15292@item -fvtv-counts 15293@opindex fvtv-counts 15294This is a debugging flag. When used in conjunction with 15295@option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this 15296causes the compiler to keep track of the total number of virtual calls 15297it encounters and the number of verifications it inserts. It also 15298counts the number of calls to certain run-time library functions 15299that it inserts and logs this information for each compilation unit. 15300The compiler writes this information to a file named 15301@file{vtv_count_data.log} in the directory named by the environment 15302variable @env{VTV_LOGS_DIR} if that is defined or the current working 15303directory otherwise. It also counts the size of the vtable pointer sets 15304for each class, and writes this information to @file{vtv_class_set_sizes.log} 15305in the same directory. 15306 15307Note: This feature @emph{appends} data to the log files. To get fresh log 15308files, be sure to delete any existing ones. 15309 15310@item -finstrument-functions 15311@opindex finstrument-functions 15312Generate instrumentation calls for entry and exit to functions. Just 15313after function entry and just before function exit, the following 15314profiling functions are called with the address of the current 15315function and its call site. (On some platforms, 15316@code{__builtin_return_address} does not work beyond the current 15317function, so the call site information may not be available to the 15318profiling functions otherwise.) 15319 15320@smallexample 15321void __cyg_profile_func_enter (void *this_fn, 15322 void *call_site); 15323void __cyg_profile_func_exit (void *this_fn, 15324 void *call_site); 15325@end smallexample 15326 15327The first argument is the address of the start of the current function, 15328which may be looked up exactly in the symbol table. 15329 15330This instrumentation is also done for functions expanded inline in other 15331functions. The profiling calls indicate where, conceptually, the 15332inline function is entered and exited. This means that addressable 15333versions of such functions must be available. If all your uses of a 15334function are expanded inline, this may mean an additional expansion of 15335code size. If you use @code{extern inline} in your C code, an 15336addressable version of such functions must be provided. (This is 15337normally the case anyway, but if you get lucky and the optimizer always 15338expands the functions inline, you might have gotten away without 15339providing static copies.) 15340 15341A function may be given the attribute @code{no_instrument_function}, in 15342which case this instrumentation is not done. This can be used, for 15343example, for the profiling functions listed above, high-priority 15344interrupt routines, and any functions from which the profiling functions 15345cannot safely be called (perhaps signal handlers, if the profiling 15346routines generate output or allocate memory). 15347@xref{Common Function Attributes}. 15348 15349@item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} 15350@opindex finstrument-functions-exclude-file-list 15351 15352Set the list of functions that are excluded from instrumentation (see 15353the description of @option{-finstrument-functions}). If the file that 15354contains a function definition matches with one of @var{file}, then 15355that function is not instrumented. The match is done on substrings: 15356if the @var{file} parameter is a substring of the file name, it is 15357considered to be a match. 15358 15359For example: 15360 15361@smallexample 15362-finstrument-functions-exclude-file-list=/bits/stl,include/sys 15363@end smallexample 15364 15365@noindent 15366excludes any inline function defined in files whose pathnames 15367contain @file{/bits/stl} or @file{include/sys}. 15368 15369If, for some reason, you want to include letter @samp{,} in one of 15370@var{sym}, write @samp{\,}. For example, 15371@option{-finstrument-functions-exclude-file-list='\,\,tmp'} 15372(note the single quote surrounding the option). 15373 15374@item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} 15375@opindex finstrument-functions-exclude-function-list 15376 15377This is similar to @option{-finstrument-functions-exclude-file-list}, 15378but this option sets the list of function names to be excluded from 15379instrumentation. The function name to be matched is its user-visible 15380name, such as @code{vector<int> blah(const vector<int> &)}, not the 15381internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The 15382match is done on substrings: if the @var{sym} parameter is a substring 15383of the function name, it is considered to be a match. For C99 and C++ 15384extended identifiers, the function name must be given in UTF-8, not 15385using universal character names. 15386 15387@item -fpatchable-function-entry=@var{N}[,@var{M}] 15388@opindex fpatchable-function-entry 15389Generate @var{N} NOPs right at the beginning 15390of each function, with the function entry point before the @var{M}th NOP. 15391If @var{M} is omitted, it defaults to @code{0} so the 15392function entry points to the address just at the first NOP. 15393The NOP instructions reserve extra space which can be used to patch in 15394any desired instrumentation at run time, provided that the code segment 15395is writable. The amount of space is controllable indirectly via 15396the number of NOPs; the NOP instruction used corresponds to the instruction 15397emitted by the internal GCC back-end interface @code{gen_nop}. This behavior 15398is target-specific and may also depend on the architecture variant and/or 15399other compilation options. 15400 15401For run-time identification, the starting addresses of these areas, 15402which correspond to their respective function entries minus @var{M}, 15403are additionally collected in the @code{__patchable_function_entries} 15404section of the resulting binary. 15405 15406Note that the value of @code{__attribute__ ((patchable_function_entry 15407(N,M)))} takes precedence over command-line option 15408@option{-fpatchable-function-entry=N,M}. This can be used to increase 15409the area size or to remove it completely on a single function. 15410If @code{N=0}, no pad location is recorded. 15411 15412The NOP instructions are inserted at---and maybe before, depending on 15413@var{M}---the function entry address, even before the prologue. 15414 15415The maximum value of @var{N} and @var{M} is 65535. 15416@end table 15417 15418 15419@node Preprocessor Options 15420@section Options Controlling the Preprocessor 15421@cindex preprocessor options 15422@cindex options, preprocessor 15423 15424These options control the C preprocessor, which is run on each C source 15425file before actual compilation. 15426 15427If you use the @option{-E} option, nothing is done except preprocessing. 15428Some of these options make sense only together with @option{-E} because 15429they cause the preprocessor output to be unsuitable for actual 15430compilation. 15431 15432In addition to the options listed here, there are a number of options 15433to control search paths for include files documented in 15434@ref{Directory Options}. 15435Options to control preprocessor diagnostics are listed in 15436@ref{Warning Options}. 15437 15438@table @gcctabopt 15439@include cppopts.texi 15440 15441@item -Wp,@var{option} 15442@opindex Wp 15443You can use @option{-Wp,@var{option}} to bypass the compiler driver 15444and pass @var{option} directly through to the preprocessor. If 15445@var{option} contains commas, it is split into multiple options at the 15446commas. However, many options are modified, translated or interpreted 15447by the compiler driver before being passed to the preprocessor, and 15448@option{-Wp} forcibly bypasses this phase. The preprocessor's direct 15449interface is undocumented and subject to change, so whenever possible 15450you should avoid using @option{-Wp} and let the driver handle the 15451options instead. 15452 15453@item -Xpreprocessor @var{option} 15454@opindex Xpreprocessor 15455Pass @var{option} as an option to the preprocessor. You can use this to 15456supply system-specific preprocessor options that GCC does not 15457recognize. 15458 15459If you want to pass an option that takes an argument, you must use 15460@option{-Xpreprocessor} twice, once for the option and once for the argument. 15461 15462@item -no-integrated-cpp 15463@opindex no-integrated-cpp 15464Perform preprocessing as a separate pass before compilation. 15465By default, GCC performs preprocessing as an integrated part of 15466input tokenization and parsing. 15467If this option is provided, the appropriate language front end 15468(@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++, 15469and Objective-C, respectively) is instead invoked twice, 15470once for preprocessing only and once for actual compilation 15471of the preprocessed input. 15472This option may be useful in conjunction with the @option{-B} or 15473@option{-wrapper} options to specify an alternate preprocessor or 15474perform additional processing of the program source between 15475normal preprocessing and compilation. 15476 15477@item -flarge-source-files 15478@opindex flarge-source-files 15479Adjust GCC to expect large source files, at the expense of slower 15480compilation and higher memory usage. 15481 15482Specifically, GCC normally tracks both column numbers and line numbers 15483within source files and it normally prints both of these numbers in 15484diagnostics. However, once it has processed a certain number of source 15485lines, it stops tracking column numbers and only tracks line numbers. 15486This means that diagnostics for later lines do not include column numbers. 15487It also means that options like @option{-Wmisleading-indentation} cease to work 15488at that point, although the compiler prints a note if this happens. 15489Passing @option{-flarge-source-files} significantly increases the number 15490of source lines that GCC can process before it stops tracking columns. 15491 15492@end table 15493 15494@node Assembler Options 15495@section Passing Options to the Assembler 15496 15497@c prevent bad page break with this line 15498You can pass options to the assembler. 15499 15500@table @gcctabopt 15501@item -Wa,@var{option} 15502@opindex Wa 15503Pass @var{option} as an option to the assembler. If @var{option} 15504contains commas, it is split into multiple options at the commas. 15505 15506@item -Xassembler @var{option} 15507@opindex Xassembler 15508Pass @var{option} as an option to the assembler. You can use this to 15509supply system-specific assembler options that GCC does not 15510recognize. 15511 15512If you want to pass an option that takes an argument, you must use 15513@option{-Xassembler} twice, once for the option and once for the argument. 15514 15515@end table 15516 15517@node Link Options 15518@section Options for Linking 15519@cindex link options 15520@cindex options, linking 15521 15522These options come into play when the compiler links object files into 15523an executable output file. They are meaningless if the compiler is 15524not doing a link step. 15525 15526@table @gcctabopt 15527@cindex file names 15528@item @var{object-file-name} 15529A file name that does not end in a special recognized suffix is 15530considered to name an object file or library. (Object files are 15531distinguished from libraries by the linker according to the file 15532contents.) If linking is done, these object files are used as input 15533to the linker. 15534 15535@item -c 15536@itemx -S 15537@itemx -E 15538@opindex c 15539@opindex S 15540@opindex E 15541If any of these options is used, then the linker is not run, and 15542object file names should not be used as arguments. @xref{Overall 15543Options}. 15544 15545@item -flinker-output=@var{type} 15546@opindex flinker-output 15547This option controls code generation of the link-time optimizer. By 15548default the linker output is automatically determined by the linker 15549plugin. For debugging the compiler and if incremental linking with a 15550non-LTO object file is desired, it may be useful to control the type 15551manually. 15552 15553If @var{type} is @samp{exec}, code generation produces a static 15554binary. In this case @option{-fpic} and @option{-fpie} are both 15555disabled. 15556 15557If @var{type} is @samp{dyn}, code generation produces a shared 15558library. In this case @option{-fpic} or @option{-fPIC} is preserved, 15559but not enabled automatically. This allows to build shared libraries 15560without position-independent code on architectures where this is 15561possible, i.e.@: on x86. 15562 15563If @var{type} is @samp{pie}, code generation produces an @option{-fpie} 15564executable. This results in similar optimizations as @samp{exec} 15565except that @option{-fpie} is not disabled if specified at compilation 15566time. 15567 15568If @var{type} is @samp{rel}, the compiler assumes that incremental linking is 15569done. The sections containing intermediate code for link-time optimization are 15570merged, pre-optimized, and output to the resulting object file. In addition, if 15571@option{-ffat-lto-objects} is specified, binary code is produced for future 15572non-LTO linking. The object file produced by incremental linking is smaller 15573than a static library produced from the same object files. At link time the 15574result of incremental linking also loads faster than a static 15575library assuming that the majority of objects in the library are used. 15576 15577Finally @samp{nolto-rel} configures the compiler for incremental linking where 15578code generation is forced, a final binary is produced, and the intermediate 15579code for later link-time optimization is stripped. When multiple object files 15580are linked together the resulting code is better optimized than with 15581link-time optimizations disabled (for example, cross-module inlining 15582happens), but most of benefits of whole program optimizations are lost. 15583 15584During the incremental link (by @option{-r}) the linker plugin defaults to 15585@option{rel}. With current interfaces to GNU Binutils it is however not 15586possible to incrementally link LTO objects and non-LTO objects into a single 15587mixed object file. If any of object files in incremental link cannot 15588be used for link-time optimization, the linker plugin issues a warning and 15589uses @samp{nolto-rel}. To maintain whole program optimization, it is 15590recommended to link such objects into static library instead. Alternatively it 15591is possible to use H.J. Lu's binutils with support for mixed objects. 15592 15593@item -fuse-ld=bfd 15594@opindex fuse-ld=bfd 15595Use the @command{bfd} linker instead of the default linker. 15596 15597@item -fuse-ld=gold 15598@opindex fuse-ld=gold 15599Use the @command{gold} linker instead of the default linker. 15600 15601@item -fuse-ld=lld 15602@opindex fuse-ld=lld 15603Use the LLVM @command{lld} linker instead of the default linker. 15604 15605@cindex Libraries 15606@item -l@var{library} 15607@itemx -l @var{library} 15608@opindex l 15609Search the library named @var{library} when linking. (The second 15610alternative with the library as a separate argument is only for 15611POSIX compliance and is not recommended.) 15612 15613The @option{-l} option is passed directly to the linker by GCC. Refer 15614to your linker documentation for exact details. The general 15615description below applies to the GNU linker. 15616 15617The linker searches a standard list of directories for the library. 15618The directories searched include several standard system directories 15619plus any that you specify with @option{-L}. 15620 15621Static libraries are archives of object files, and have file names 15622like @file{lib@var{library}.a}. Some targets also support shared 15623libraries, which typically have names like @file{lib@var{library}.so}. 15624If both static and shared libraries are found, the linker gives 15625preference to linking with the shared library unless the 15626@option{-static} option is used. 15627 15628It makes a difference where in the command you write this option; the 15629linker searches and processes libraries and object files in the order they 15630are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z} 15631after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers 15632to functions in @samp{z}, those functions may not be loaded. 15633 15634@item -lobjc 15635@opindex lobjc 15636You need this special case of the @option{-l} option in order to 15637link an Objective-C or Objective-C++ program. 15638 15639@item -nostartfiles 15640@opindex nostartfiles 15641Do not use the standard system startup files when linking. 15642The standard system libraries are used normally, unless @option{-nostdlib}, 15643@option{-nolibc}, or @option{-nodefaultlibs} is used. 15644 15645@item -nodefaultlibs 15646@opindex nodefaultlibs 15647Do not use the standard system libraries when linking. 15648Only the libraries you specify are passed to the linker, and options 15649specifying linkage of the system libraries, such as @option{-static-libgcc} 15650or @option{-shared-libgcc}, are ignored. 15651The standard startup files are used normally, unless @option{-nostartfiles} 15652is used. 15653 15654The compiler may generate calls to @code{memcmp}, 15655@code{memset}, @code{memcpy} and @code{memmove}. 15656These entries are usually resolved by entries in 15657libc. These entry points should be supplied through some other 15658mechanism when this option is specified. 15659 15660@item -nolibc 15661@opindex nolibc 15662Do not use the C library or system libraries tightly coupled with it when 15663linking. Still link with the startup files, @file{libgcc} or toolchain 15664provided language support libraries such as @file{libgnat}, @file{libgfortran} 15665or @file{libstdc++} unless options preventing their inclusion are used as 15666well. This typically removes @option{-lc} from the link command line, as well 15667as system libraries that normally go with it and become meaningless when 15668absence of a C library is assumed, for example @option{-lpthread} or 15669@option{-lm} in some configurations. This is intended for bare-board 15670targets when there is indeed no C library available. 15671 15672@item -nostdlib 15673@opindex nostdlib 15674Do not use the standard system startup files or libraries when linking. 15675No startup files and only the libraries you specify are passed to 15676the linker, and options specifying linkage of the system libraries, such as 15677@option{-static-libgcc} or @option{-shared-libgcc}, are ignored. 15678 15679The compiler may generate calls to @code{memcmp}, @code{memset}, 15680@code{memcpy} and @code{memmove}. 15681These entries are usually resolved by entries in 15682libc. These entry points should be supplied through some other 15683mechanism when this option is specified. 15684 15685@cindex @option{-lgcc}, use with @option{-nostdlib} 15686@cindex @option{-nostdlib} and unresolved references 15687@cindex unresolved references and @option{-nostdlib} 15688@cindex @option{-lgcc}, use with @option{-nodefaultlibs} 15689@cindex @option{-nodefaultlibs} and unresolved references 15690@cindex unresolved references and @option{-nodefaultlibs} 15691One of the standard libraries bypassed by @option{-nostdlib} and 15692@option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines 15693which GCC uses to overcome shortcomings of particular machines, or special 15694needs for some languages. 15695(@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler 15696Collection (GCC) Internals}, 15697for more discussion of @file{libgcc.a}.) 15698In most cases, you need @file{libgcc.a} even when you want to avoid 15699other standard libraries. In other words, when you specify @option{-nostdlib} 15700or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well. 15701This ensures that you have no unresolved references to internal GCC 15702library subroutines. 15703(An example of such an internal subroutine is @code{__main}, used to ensure C++ 15704constructors are called; @pxref{Collect2,,@code{collect2}, gccint, 15705GNU Compiler Collection (GCC) Internals}.) 15706 15707@item -e @var{entry} 15708@itemx --entry=@var{entry} 15709@opindex e 15710@opindex entry 15711 15712Specify that the program entry point is @var{entry}. The argument is 15713interpreted by the linker; the GNU linker accepts either a symbol name 15714or an address. 15715 15716@item -pie 15717@opindex pie 15718Produce a dynamically linked position independent executable on targets 15719that support it. For predictable results, you must also specify the same 15720set of options used for compilation (@option{-fpie}, @option{-fPIE}, 15721or model suboptions) when you specify this linker option. 15722 15723@item -no-pie 15724@opindex no-pie 15725Don't produce a dynamically linked position independent executable. 15726 15727@item -static-pie 15728@opindex static-pie 15729Produce a static position independent executable on targets that support 15730it. A static position independent executable is similar to a static 15731executable, but can be loaded at any address without a dynamic linker. 15732For predictable results, you must also specify the same set of options 15733used for compilation (@option{-fpie}, @option{-fPIE}, or model 15734suboptions) when you specify this linker option. 15735 15736@item -pthread 15737@opindex pthread 15738Link with the POSIX threads library. This option is supported on 15739GNU/Linux targets, most other Unix derivatives, and also on 15740x86 Cygwin and MinGW targets. On some targets this option also sets 15741flags for the preprocessor, so it should be used consistently for both 15742compilation and linking. 15743 15744@item -r 15745@opindex r 15746Produce a relocatable object as output. This is also known as partial 15747linking. 15748 15749@item -rdynamic 15750@opindex rdynamic 15751Pass the flag @option{-export-dynamic} to the ELF linker, on targets 15752that support it. This instructs the linker to add all symbols, not 15753only used ones, to the dynamic symbol table. This option is needed 15754for some uses of @code{dlopen} or to allow obtaining backtraces 15755from within a program. 15756 15757@item -s 15758@opindex s 15759Remove all symbol table and relocation information from the executable. 15760 15761@item -static 15762@opindex static 15763On systems that support dynamic linking, this overrides @option{-pie} 15764and prevents linking with the shared libraries. On other systems, this 15765option has no effect. 15766 15767@item -shared 15768@opindex shared 15769Produce a shared object which can then be linked with other objects to 15770form an executable. Not all systems support this option. For predictable 15771results, you must also specify the same set of options used for compilation 15772(@option{-fpic}, @option{-fPIC}, or model suboptions) when 15773you specify this linker option.@footnote{On some systems, @samp{gcc -shared} 15774needs to build supplementary stub code for constructors to work. On 15775multi-libbed systems, @samp{gcc -shared} must select the correct support 15776libraries to link against. Failing to supply the correct flags may lead 15777to subtle defects. Supplying them in cases where they are not necessary 15778is innocuous.} 15779 15780@item -shared-libgcc 15781@itemx -static-libgcc 15782@opindex shared-libgcc 15783@opindex static-libgcc 15784On systems that provide @file{libgcc} as a shared library, these options 15785force the use of either the shared or static version, respectively. 15786If no shared version of @file{libgcc} was built when the compiler was 15787configured, these options have no effect. 15788 15789There are several situations in which an application should use the 15790shared @file{libgcc} instead of the static version. The most common 15791of these is when the application wishes to throw and catch exceptions 15792across different shared libraries. In that case, each of the libraries 15793as well as the application itself should use the shared @file{libgcc}. 15794 15795Therefore, the G++ driver automatically adds @option{-shared-libgcc} 15796whenever you build a shared library or a main executable, because C++ 15797programs typically use exceptions, so this is the right thing to do. 15798 15799If, instead, you use the GCC driver to create shared libraries, you may 15800find that they are not always linked with the shared @file{libgcc}. 15801If GCC finds, at its configuration time, that you have a non-GNU linker 15802or a GNU linker that does not support option @option{--eh-frame-hdr}, 15803it links the shared version of @file{libgcc} into shared libraries 15804by default. Otherwise, it takes advantage of the linker and optimizes 15805away the linking with the shared version of @file{libgcc}, linking with 15806the static version of libgcc by default. This allows exceptions to 15807propagate through such shared libraries, without incurring relocation 15808costs at library load time. 15809 15810However, if a library or main executable is supposed to throw or catch 15811exceptions, you must link it using the G++ driver, or using the option 15812@option{-shared-libgcc}, such that it is linked with the shared 15813@file{libgcc}. 15814 15815@item -static-libasan 15816@opindex static-libasan 15817When the @option{-fsanitize=address} option is used to link a program, 15818the GCC driver automatically links against @option{libasan}. If 15819@file{libasan} is available as a shared library, and the @option{-static} 15820option is not used, then this links against the shared version of 15821@file{libasan}. The @option{-static-libasan} option directs the GCC 15822driver to link @file{libasan} statically, without necessarily linking 15823other libraries statically. 15824 15825@item -static-libtsan 15826@opindex static-libtsan 15827When the @option{-fsanitize=thread} option is used to link a program, 15828the GCC driver automatically links against @option{libtsan}. If 15829@file{libtsan} is available as a shared library, and the @option{-static} 15830option is not used, then this links against the shared version of 15831@file{libtsan}. The @option{-static-libtsan} option directs the GCC 15832driver to link @file{libtsan} statically, without necessarily linking 15833other libraries statically. 15834 15835@item -static-liblsan 15836@opindex static-liblsan 15837When the @option{-fsanitize=leak} option is used to link a program, 15838the GCC driver automatically links against @option{liblsan}. If 15839@file{liblsan} is available as a shared library, and the @option{-static} 15840option is not used, then this links against the shared version of 15841@file{liblsan}. The @option{-static-liblsan} option directs the GCC 15842driver to link @file{liblsan} statically, without necessarily linking 15843other libraries statically. 15844 15845@item -static-libubsan 15846@opindex static-libubsan 15847When the @option{-fsanitize=undefined} option is used to link a program, 15848the GCC driver automatically links against @option{libubsan}. If 15849@file{libubsan} is available as a shared library, and the @option{-static} 15850option is not used, then this links against the shared version of 15851@file{libubsan}. The @option{-static-libubsan} option directs the GCC 15852driver to link @file{libubsan} statically, without necessarily linking 15853other libraries statically. 15854 15855@item -static-libstdc++ 15856@opindex static-libstdc++ 15857When the @command{g++} program is used to link a C++ program, it 15858normally automatically links against @option{libstdc++}. If 15859@file{libstdc++} is available as a shared library, and the 15860@option{-static} option is not used, then this links against the 15861shared version of @file{libstdc++}. That is normally fine. However, it 15862is sometimes useful to freeze the version of @file{libstdc++} used by 15863the program without going all the way to a fully static link. The 15864@option{-static-libstdc++} option directs the @command{g++} driver to 15865link @file{libstdc++} statically, without necessarily linking other 15866libraries statically. 15867 15868@item -symbolic 15869@opindex symbolic 15870Bind references to global symbols when building a shared object. Warn 15871about any unresolved references (unless overridden by the link editor 15872option @option{-Xlinker -z -Xlinker defs}). Only a few systems support 15873this option. 15874 15875@item -T @var{script} 15876@opindex T 15877@cindex linker script 15878Use @var{script} as the linker script. This option is supported by most 15879systems using the GNU linker. On some targets, such as bare-board 15880targets without an operating system, the @option{-T} option may be required 15881when linking to avoid references to undefined symbols. 15882 15883@item -Xlinker @var{option} 15884@opindex Xlinker 15885Pass @var{option} as an option to the linker. You can use this to 15886supply system-specific linker options that GCC does not recognize. 15887 15888If you want to pass an option that takes a separate argument, you must use 15889@option{-Xlinker} twice, once for the option and once for the argument. 15890For example, to pass @option{-assert definitions}, you must write 15891@option{-Xlinker -assert -Xlinker definitions}. It does not work to write 15892@option{-Xlinker "-assert definitions"}, because this passes the entire 15893string as a single argument, which is not what the linker expects. 15894 15895When using the GNU linker, it is usually more convenient to pass 15896arguments to linker options using the @option{@var{option}=@var{value}} 15897syntax than as separate arguments. For example, you can specify 15898@option{-Xlinker -Map=output.map} rather than 15899@option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support 15900this syntax for command-line options. 15901 15902@item -Wl,@var{option} 15903@opindex Wl 15904Pass @var{option} as an option to the linker. If @var{option} contains 15905commas, it is split into multiple options at the commas. You can use this 15906syntax to pass an argument to the option. 15907For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the 15908linker. When using the GNU linker, you can also get the same effect with 15909@option{-Wl,-Map=output.map}. 15910 15911@item -u @var{symbol} 15912@opindex u 15913Pretend the symbol @var{symbol} is undefined, to force linking of 15914library modules to define it. You can use @option{-u} multiple times with 15915different symbols to force loading of additional library modules. 15916 15917@item -z @var{keyword} 15918@opindex z 15919@option{-z} is passed directly on to the linker along with the keyword 15920@var{keyword}. See the section in the documentation of your linker for 15921permitted values and their meanings. 15922@end table 15923 15924@node Directory Options 15925@section Options for Directory Search 15926@cindex directory options 15927@cindex options, directory search 15928@cindex search path 15929 15930These options specify directories to search for header files, for 15931libraries and for parts of the compiler: 15932 15933@table @gcctabopt 15934@include cppdiropts.texi 15935 15936@item -iplugindir=@var{dir} 15937@opindex iplugindir= 15938Set the directory to search for plugins that are passed 15939by @option{-fplugin=@var{name}} instead of 15940@option{-fplugin=@var{path}/@var{name}.so}. This option is not meant 15941to be used by the user, but only passed by the driver. 15942 15943@item -L@var{dir} 15944@opindex L 15945Add directory @var{dir} to the list of directories to be searched 15946for @option{-l}. 15947 15948@item -B@var{prefix} 15949@opindex B 15950This option specifies where to find the executables, libraries, 15951include files, and data files of the compiler itself. 15952 15953The compiler driver program runs one or more of the subprograms 15954@command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries 15955@var{prefix} as a prefix for each program it tries to run, both with and 15956without @samp{@var{machine}/@var{version}/} for the corresponding target 15957machine and compiler version. 15958 15959For each subprogram to be run, the compiler driver first tries the 15960@option{-B} prefix, if any. If that name is not found, or if @option{-B} 15961is not specified, the driver tries two standard prefixes, 15962@file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of 15963those results in a file name that is found, the unmodified program 15964name is searched for using the directories specified in your 15965@env{PATH} environment variable. 15966 15967The compiler checks to see if the path provided by @option{-B} 15968refers to a directory, and if necessary it adds a directory 15969separator character at the end of the path. 15970 15971@option{-B} prefixes that effectively specify directory names also apply 15972to libraries in the linker, because the compiler translates these 15973options into @option{-L} options for the linker. They also apply to 15974include files in the preprocessor, because the compiler translates these 15975options into @option{-isystem} options for the preprocessor. In this case, 15976the compiler appends @samp{include} to the prefix. 15977 15978The runtime support file @file{libgcc.a} can also be searched for using 15979the @option{-B} prefix, if needed. If it is not found there, the two 15980standard prefixes above are tried, and that is all. The file is left 15981out of the link if it is not found by those means. 15982 15983Another way to specify a prefix much like the @option{-B} prefix is to use 15984the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment 15985Variables}. 15986 15987As a special kludge, if the path provided by @option{-B} is 15988@file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to 159899, then it is replaced by @file{[dir/]include}. This is to help 15990with boot-strapping the compiler. 15991 15992@item -no-canonical-prefixes 15993@opindex no-canonical-prefixes 15994Do not expand any symbolic links, resolve references to @samp{/../} 15995or @samp{/./}, or make the path absolute when generating a relative 15996prefix. 15997 15998@item --sysroot=@var{dir} 15999@opindex sysroot 16000Use @var{dir} as the logical root directory for headers and libraries. 16001For example, if the compiler normally searches for headers in 16002@file{/usr/include} and libraries in @file{/usr/lib}, it instead 16003searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}. 16004 16005If you use both this option and the @option{-isysroot} option, then 16006the @option{--sysroot} option applies to libraries, but the 16007@option{-isysroot} option applies to header files. 16008 16009The GNU linker (beginning with version 2.16) has the necessary support 16010for this option. If your linker does not support this option, the 16011header file aspect of @option{--sysroot} still works, but the 16012library aspect does not. 16013 16014@item --no-sysroot-suffix 16015@opindex no-sysroot-suffix 16016For some targets, a suffix is added to the root directory specified 16017with @option{--sysroot}, depending on the other options used, so that 16018headers may for example be found in 16019@file{@var{dir}/@var{suffix}/usr/include} instead of 16020@file{@var{dir}/usr/include}. This option disables the addition of 16021such a suffix. 16022 16023@end table 16024 16025@node Code Gen Options 16026@section Options for Code Generation Conventions 16027@cindex code generation conventions 16028@cindex options, code generation 16029@cindex run-time options 16030 16031These machine-independent options control the interface conventions 16032used in code generation. 16033 16034Most of them have both positive and negative forms; the negative form 16035of @option{-ffoo} is @option{-fno-foo}. In the table below, only 16036one of the forms is listed---the one that is not the default. You 16037can figure out the other form by either removing @samp{no-} or adding 16038it. 16039 16040@table @gcctabopt 16041@item -fstack-reuse=@var{reuse-level} 16042@opindex fstack_reuse 16043This option controls stack space reuse for user declared local/auto variables 16044and compiler generated temporaries. @var{reuse_level} can be @samp{all}, 16045@samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all 16046local variables and temporaries, @samp{named_vars} enables the reuse only for 16047user defined local variables with names, and @samp{none} disables stack reuse 16048completely. The default value is @samp{all}. The option is needed when the 16049program extends the lifetime of a scoped local variable or a compiler generated 16050temporary beyond the end point defined by the language. When a lifetime of 16051a variable ends, and if the variable lives in memory, the optimizing compiler 16052has the freedom to reuse its stack space with other temporaries or scoped 16053local variables whose live range does not overlap with it. Legacy code extending 16054local lifetime is likely to break with the stack reuse optimization. 16055 16056For example, 16057 16058@smallexample 16059 int *p; 16060 @{ 16061 int local1; 16062 16063 p = &local1; 16064 local1 = 10; 16065 .... 16066 @} 16067 @{ 16068 int local2; 16069 local2 = 20; 16070 ... 16071 @} 16072 16073 if (*p == 10) // out of scope use of local1 16074 @{ 16075 16076 @} 16077@end smallexample 16078 16079Another example: 16080@smallexample 16081 16082 struct A 16083 @{ 16084 A(int k) : i(k), j(k) @{ @} 16085 int i; 16086 int j; 16087 @}; 16088 16089 A *ap; 16090 16091 void foo(const A& ar) 16092 @{ 16093 ap = &ar; 16094 @} 16095 16096 void bar() 16097 @{ 16098 foo(A(10)); // temp object's lifetime ends when foo returns 16099 16100 @{ 16101 A a(20); 16102 .... 16103 @} 16104 ap->i+= 10; // ap references out of scope temp whose space 16105 // is reused with a. What is the value of ap->i? 16106 @} 16107 16108@end smallexample 16109 16110The lifetime of a compiler generated temporary is well defined by the C++ 16111standard. When a lifetime of a temporary ends, and if the temporary lives 16112in memory, the optimizing compiler has the freedom to reuse its stack 16113space with other temporaries or scoped local variables whose live range 16114does not overlap with it. However some of the legacy code relies on 16115the behavior of older compilers in which temporaries' stack space is 16116not reused, the aggressive stack reuse can lead to runtime errors. This 16117option is used to control the temporary stack reuse optimization. 16118 16119@item -ftrapv 16120@opindex ftrapv 16121This option generates traps for signed overflow on addition, subtraction, 16122multiplication operations. 16123The options @option{-ftrapv} and @option{-fwrapv} override each other, so using 16124@option{-ftrapv} @option{-fwrapv} on the command-line results in 16125@option{-fwrapv} being effective. Note that only active options override, so 16126using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line 16127results in @option{-ftrapv} being effective. 16128 16129@item -fwrapv 16130@opindex fwrapv 16131This option instructs the compiler to assume that signed arithmetic 16132overflow of addition, subtraction and multiplication wraps around 16133using twos-complement representation. This flag enables some optimizations 16134and disables others. 16135The options @option{-ftrapv} and @option{-fwrapv} override each other, so using 16136@option{-ftrapv} @option{-fwrapv} on the command-line results in 16137@option{-fwrapv} being effective. Note that only active options override, so 16138using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line 16139results in @option{-ftrapv} being effective. 16140 16141@item -fwrapv-pointer 16142@opindex fwrapv-pointer 16143This option instructs the compiler to assume that pointer arithmetic 16144overflow on addition and subtraction wraps around using twos-complement 16145representation. This flag disables some optimizations which assume 16146pointer overflow is invalid. 16147 16148@item -fstrict-overflow 16149@opindex fstrict-overflow 16150This option implies @option{-fno-wrapv} @option{-fno-wrapv-pointer} and when 16151negated implies @option{-fwrapv} @option{-fwrapv-pointer}. 16152 16153@item -fexceptions 16154@opindex fexceptions 16155Enable exception handling. Generates extra code needed to propagate 16156exceptions. For some targets, this implies GCC generates frame 16157unwind information for all functions, which can produce significant data 16158size overhead, although it does not affect execution. If you do not 16159specify this option, GCC enables it by default for languages like 16160C++ that normally require exception handling, and disables it for 16161languages like C that do not normally require it. However, you may need 16162to enable this option when compiling C code that needs to interoperate 16163properly with exception handlers written in C++. You may also wish to 16164disable this option if you are compiling older C++ programs that don't 16165use exception handling. 16166 16167@item -fnon-call-exceptions 16168@opindex fnon-call-exceptions 16169Generate code that allows trapping instructions to throw exceptions. 16170Note that this requires platform-specific runtime support that does 16171not exist everywhere. Moreover, it only allows @emph{trapping} 16172instructions to throw exceptions, i.e.@: memory references or floating-point 16173instructions. It does not allow exceptions to be thrown from 16174arbitrary signal handlers such as @code{SIGALRM}. 16175 16176@item -fdelete-dead-exceptions 16177@opindex fdelete-dead-exceptions 16178Consider that instructions that may throw exceptions but don't otherwise 16179contribute to the execution of the program can be optimized away. 16180This option is enabled by default for the Ada compiler, as permitted by 16181the Ada language specification. 16182Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels. 16183 16184@item -funwind-tables 16185@opindex funwind-tables 16186Similar to @option{-fexceptions}, except that it just generates any needed 16187static data, but does not affect the generated code in any other way. 16188You normally do not need to enable this option; instead, a language processor 16189that needs this handling enables it on your behalf. 16190 16191@item -fasynchronous-unwind-tables 16192@opindex fasynchronous-unwind-tables 16193Generate unwind table in DWARF format, if supported by target machine. The 16194table is exact at each instruction boundary, so it can be used for stack 16195unwinding from asynchronous events (such as debugger or garbage collector). 16196 16197@item -fno-gnu-unique 16198@opindex fno-gnu-unique 16199@opindex fgnu-unique 16200On systems with recent GNU assembler and C library, the C++ compiler 16201uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions 16202of template static data members and static local variables in inline 16203functions are unique even in the presence of @code{RTLD_LOCAL}; this 16204is necessary to avoid problems with a library used by two different 16205@code{RTLD_LOCAL} plugins depending on a definition in one of them and 16206therefore disagreeing with the other one about the binding of the 16207symbol. But this causes @code{dlclose} to be ignored for affected 16208DSOs; if your program relies on reinitialization of a DSO via 16209@code{dlclose} and @code{dlopen}, you can use 16210@option{-fno-gnu-unique}. 16211 16212@item -fpcc-struct-return 16213@opindex fpcc-struct-return 16214Return ``short'' @code{struct} and @code{union} values in memory like 16215longer ones, rather than in registers. This convention is less 16216efficient, but it has the advantage of allowing intercallability between 16217GCC-compiled files and files compiled with other compilers, particularly 16218the Portable C Compiler (pcc). 16219 16220The precise convention for returning structures in memory depends 16221on the target configuration macros. 16222 16223Short structures and unions are those whose size and alignment match 16224that of some integer type. 16225 16226@strong{Warning:} code compiled with the @option{-fpcc-struct-return} 16227switch is not binary compatible with code compiled with the 16228@option{-freg-struct-return} switch. 16229Use it to conform to a non-default application binary interface. 16230 16231@item -freg-struct-return 16232@opindex freg-struct-return 16233Return @code{struct} and @code{union} values in registers when possible. 16234This is more efficient for small structures than 16235@option{-fpcc-struct-return}. 16236 16237If you specify neither @option{-fpcc-struct-return} nor 16238@option{-freg-struct-return}, GCC defaults to whichever convention is 16239standard for the target. If there is no standard convention, GCC 16240defaults to @option{-fpcc-struct-return}, except on targets where GCC is 16241the principal compiler. In those cases, we can choose the standard, and 16242we chose the more efficient register return alternative. 16243 16244@strong{Warning:} code compiled with the @option{-freg-struct-return} 16245switch is not binary compatible with code compiled with the 16246@option{-fpcc-struct-return} switch. 16247Use it to conform to a non-default application binary interface. 16248 16249@item -fshort-enums 16250@opindex fshort-enums 16251Allocate to an @code{enum} type only as many bytes as it needs for the 16252declared range of possible values. Specifically, the @code{enum} type 16253is equivalent to the smallest integer type that has enough room. 16254 16255@strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate 16256code that is not binary compatible with code generated without that switch. 16257Use it to conform to a non-default application binary interface. 16258 16259@item -fshort-wchar 16260@opindex fshort-wchar 16261Override the underlying type for @code{wchar_t} to be @code{short 16262unsigned int} instead of the default for the target. This option is 16263useful for building programs to run under WINE@. 16264 16265@strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate 16266code that is not binary compatible with code generated without that switch. 16267Use it to conform to a non-default application binary interface. 16268 16269@item -fcommon 16270@opindex fcommon 16271@opindex fno-common 16272@cindex tentative definitions 16273In C code, this option controls the placement of global variables 16274defined without an initializer, known as @dfn{tentative definitions} 16275in the C standard. Tentative definitions are distinct from declarations 16276of a variable with the @code{extern} keyword, which do not allocate storage. 16277 16278The default is @option{-fno-common}, which specifies that the compiler places 16279uninitialized global variables in the BSS section of the object file. 16280This inhibits the merging of tentative definitions by the linker so you get a 16281multiple-definition error if the same variable is accidentally defined in more 16282than one compilation unit. 16283 16284The @option{-fcommon} places uninitialized global variables in a common block. 16285This allows the linker to resolve all tentative definitions of the same variable 16286in different compilation units to the same object, or to a non-tentative 16287definition. This behavior is inconsistent with C++, and on many targets implies 16288a speed and code size penalty on global variable references. It is mainly 16289useful to enable legacy code to link without errors. 16290 16291@item -fno-ident 16292@opindex fno-ident 16293@opindex fident 16294Ignore the @code{#ident} directive. 16295 16296@item -finhibit-size-directive 16297@opindex finhibit-size-directive 16298Don't output a @code{.size} assembler directive, or anything else that 16299would cause trouble if the function is split in the middle, and the 16300two halves are placed at locations far apart in memory. This option is 16301used when compiling @file{crtstuff.c}; you should not need to use it 16302for anything else. 16303 16304@item -fverbose-asm 16305@opindex fverbose-asm 16306Put extra commentary information in the generated assembly code to 16307make it more readable. This option is generally only of use to those 16308who actually need to read the generated assembly code (perhaps while 16309debugging the compiler itself). 16310 16311@option{-fno-verbose-asm}, the default, causes the 16312extra information to be omitted and is useful when comparing two assembler 16313files. 16314 16315The added comments include: 16316 16317@itemize @bullet 16318 16319@item 16320information on the compiler version and command-line options, 16321 16322@item 16323the source code lines associated with the assembly instructions, 16324in the form FILENAME:LINENUMBER:CONTENT OF LINE, 16325 16326@item 16327hints on which high-level expressions correspond to 16328the various assembly instruction operands. 16329 16330@end itemize 16331 16332For example, given this C source file: 16333 16334@smallexample 16335int test (int n) 16336@{ 16337 int i; 16338 int total = 0; 16339 16340 for (i = 0; i < n; i++) 16341 total += i * i; 16342 16343 return total; 16344@} 16345@end smallexample 16346 16347compiling to (x86_64) assembly via @option{-S} and emitting the result 16348direct to stdout via @option{-o} @option{-} 16349 16350@smallexample 16351gcc -S test.c -fverbose-asm -Os -o - 16352@end smallexample 16353 16354gives output similar to this: 16355 16356@smallexample 16357 .file "test.c" 16358# GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu) 16359 [...snip...] 16360# options passed: 16361 [...snip...] 16362 16363 .text 16364 .globl test 16365 .type test, @@function 16366test: 16367.LFB0: 16368 .cfi_startproc 16369# test.c:4: int total = 0; 16370 xorl %eax, %eax # <retval> 16371# test.c:6: for (i = 0; i < n; i++) 16372 xorl %edx, %edx # i 16373.L2: 16374# test.c:6: for (i = 0; i < n; i++) 16375 cmpl %edi, %edx # n, i 16376 jge .L5 #, 16377# test.c:7: total += i * i; 16378 movl %edx, %ecx # i, tmp92 16379 imull %edx, %ecx # i, tmp92 16380# test.c:6: for (i = 0; i < n; i++) 16381 incl %edx # i 16382# test.c:7: total += i * i; 16383 addl %ecx, %eax # tmp92, <retval> 16384 jmp .L2 # 16385.L5: 16386# test.c:10: @} 16387 ret 16388 .cfi_endproc 16389.LFE0: 16390 .size test, .-test 16391 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)" 16392 .section .note.GNU-stack,"",@@progbits 16393@end smallexample 16394 16395The comments are intended for humans rather than machines and hence the 16396precise format of the comments is subject to change. 16397 16398@item -frecord-gcc-switches 16399@opindex frecord-gcc-switches 16400This switch causes the command line used to invoke the 16401compiler to be recorded into the object file that is being created. 16402This switch is only implemented on some targets and the exact format 16403of the recording is target and binary file format dependent, but it 16404usually takes the form of a section containing ASCII text. This 16405switch is related to the @option{-fverbose-asm} switch, but that 16406switch only records information in the assembler output file as 16407comments, so it never reaches the object file. 16408See also @option{-grecord-gcc-switches} for another 16409way of storing compiler options into the object file. 16410 16411@item -fpic 16412@opindex fpic 16413@cindex global offset table 16414@cindex PIC 16415Generate position-independent code (PIC) suitable for use in a shared 16416library, if supported for the target machine. Such code accesses all 16417constant addresses through a global offset table (GOT)@. The dynamic 16418loader resolves the GOT entries when the program starts (the dynamic 16419loader is not part of GCC; it is part of the operating system). If 16420the GOT size for the linked executable exceeds a machine-specific 16421maximum size, you get an error message from the linker indicating that 16422@option{-fpic} does not work; in that case, recompile with @option{-fPIC} 16423instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k 16424on the m68k and RS/6000. The x86 has no such limit.) 16425 16426Position-independent code requires special support, and therefore works 16427only on certain machines. For the x86, GCC supports PIC for System V 16428but not for the Sun 386i. Code generated for the IBM RS/6000 is always 16429position-independent. 16430 16431When this flag is set, the macros @code{__pic__} and @code{__PIC__} 16432are defined to 1. 16433 16434@item -fPIC 16435@opindex fPIC 16436If supported for the target machine, emit position-independent code, 16437suitable for dynamic linking and avoiding any limit on the size of the 16438global offset table. This option makes a difference on AArch64, m68k, 16439PowerPC and SPARC@. 16440 16441Position-independent code requires special support, and therefore works 16442only on certain machines. 16443 16444When this flag is set, the macros @code{__pic__} and @code{__PIC__} 16445are defined to 2. 16446 16447@item -fpie 16448@itemx -fPIE 16449@opindex fpie 16450@opindex fPIE 16451These options are similar to @option{-fpic} and @option{-fPIC}, but the 16452generated position-independent code can be only linked into executables. 16453Usually these options are used to compile code that will be linked using 16454the @option{-pie} GCC option. 16455 16456@option{-fpie} and @option{-fPIE} both define the macros 16457@code{__pie__} and @code{__PIE__}. The macros have the value 1 16458for @option{-fpie} and 2 for @option{-fPIE}. 16459 16460@item -fno-plt 16461@opindex fno-plt 16462@opindex fplt 16463Do not use the PLT for external function calls in position-independent code. 16464Instead, load the callee address at call sites from the GOT and branch to it. 16465This leads to more efficient code by eliminating PLT stubs and exposing 16466GOT loads to optimizations. On architectures such as 32-bit x86 where 16467PLT stubs expect the GOT pointer in a specific register, this gives more 16468register allocation freedom to the compiler. 16469Lazy binding requires use of the PLT; 16470with @option{-fno-plt} all external symbols are resolved at load time. 16471 16472Alternatively, the function attribute @code{noplt} can be used to avoid calls 16473through the PLT for specific external functions. 16474 16475In position-dependent code, a few targets also convert calls to 16476functions that are marked to not use the PLT to use the GOT instead. 16477 16478@item -fno-jump-tables 16479@opindex fno-jump-tables 16480@opindex fjump-tables 16481Do not use jump tables for switch statements even where it would be 16482more efficient than other code generation strategies. This option is 16483of use in conjunction with @option{-fpic} or @option{-fPIC} for 16484building code that forms part of a dynamic linker and cannot 16485reference the address of a jump table. On some targets, jump tables 16486do not require a GOT and this option is not needed. 16487 16488@item -fno-bit-tests 16489@opindex fno-bit-tests 16490@opindex fbit-tests 16491Do not use bit tests for switch statements even where it would be 16492more efficient than other code generation strategies. 16493 16494@item -ffixed-@var{reg} 16495@opindex ffixed 16496Treat the register named @var{reg} as a fixed register; generated code 16497should never refer to it (except perhaps as a stack pointer, frame 16498pointer or in some other fixed role). 16499 16500@var{reg} must be the name of a register. The register names accepted 16501are machine-specific and are defined in the @code{REGISTER_NAMES} 16502macro in the machine description macro file. 16503 16504This flag does not have a negative form, because it specifies a 16505three-way choice. 16506 16507@item -fcall-used-@var{reg} 16508@opindex fcall-used 16509Treat the register named @var{reg} as an allocable register that is 16510clobbered by function calls. It may be allocated for temporaries or 16511variables that do not live across a call. Functions compiled this way 16512do not save and restore the register @var{reg}. 16513 16514It is an error to use this flag with the frame pointer or stack pointer. 16515Use of this flag for other registers that have fixed pervasive roles in 16516the machine's execution model produces disastrous results. 16517 16518This flag does not have a negative form, because it specifies a 16519three-way choice. 16520 16521@item -fcall-saved-@var{reg} 16522@opindex fcall-saved 16523Treat the register named @var{reg} as an allocable register saved by 16524functions. It may be allocated even for temporaries or variables that 16525live across a call. Functions compiled this way save and restore 16526the register @var{reg} if they use it. 16527 16528It is an error to use this flag with the frame pointer or stack pointer. 16529Use of this flag for other registers that have fixed pervasive roles in 16530the machine's execution model produces disastrous results. 16531 16532A different sort of disaster results from the use of this flag for 16533a register in which function values may be returned. 16534 16535This flag does not have a negative form, because it specifies a 16536three-way choice. 16537 16538@item -fpack-struct[=@var{n}] 16539@opindex fpack-struct 16540Without a value specified, pack all structure members together without 16541holes. When a value is specified (which must be a small power of two), pack 16542structure members according to this value, representing the maximum 16543alignment (that is, objects with default alignment requirements larger than 16544this are output potentially unaligned at the next fitting location. 16545 16546@strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate 16547code that is not binary compatible with code generated without that switch. 16548Additionally, it makes the code suboptimal. 16549Use it to conform to a non-default application binary interface. 16550 16551@item -fleading-underscore 16552@opindex fleading-underscore 16553This option and its counterpart, @option{-fno-leading-underscore}, forcibly 16554change the way C symbols are represented in the object file. One use 16555is to help link with legacy assembly code. 16556 16557@strong{Warning:} the @option{-fleading-underscore} switch causes GCC to 16558generate code that is not binary compatible with code generated without that 16559switch. Use it to conform to a non-default application binary interface. 16560Not all targets provide complete support for this switch. 16561 16562@item -ftls-model=@var{model} 16563@opindex ftls-model 16564Alter the thread-local storage model to be used (@pxref{Thread-Local}). 16565The @var{model} argument should be one of @samp{global-dynamic}, 16566@samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}. 16567Note that the choice is subject to optimization: the compiler may use 16568a more efficient model for symbols not visible outside of the translation 16569unit, or if @option{-fpic} is not given on the command line. 16570 16571The default without @option{-fpic} is @samp{initial-exec}; with 16572@option{-fpic} the default is @samp{global-dynamic}. 16573 16574@item -ftrampolines 16575@opindex ftrampolines 16576For targets that normally need trampolines for nested functions, always 16577generate them instead of using descriptors. Otherwise, for targets that 16578do not need them, like for example HP-PA or IA-64, do nothing. 16579 16580A trampoline is a small piece of code that is created at run time on the 16581stack when the address of a nested function is taken, and is used to call 16582the nested function indirectly. Therefore, it requires the stack to be 16583made executable in order for the program to work properly. 16584 16585@option{-fno-trampolines} is enabled by default on a language by language 16586basis to let the compiler avoid generating them, if it computes that this 16587is safe, and replace them with descriptors. Descriptors are made up of data 16588only, but the generated code must be prepared to deal with them. As of this 16589writing, @option{-fno-trampolines} is enabled by default only for Ada. 16590 16591Moreover, code compiled with @option{-ftrampolines} and code compiled with 16592@option{-fno-trampolines} are not binary compatible if nested functions are 16593present. This option must therefore be used on a program-wide basis and be 16594manipulated with extreme care. 16595 16596@item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} 16597@opindex fvisibility 16598Set the default ELF image symbol visibility to the specified option---all 16599symbols are marked with this unless overridden within the code. 16600Using this feature can very substantially improve linking and 16601load times of shared object libraries, produce more optimized 16602code, provide near-perfect API export and prevent symbol clashes. 16603It is @strong{strongly} recommended that you use this in any shared objects 16604you distribute. 16605 16606Despite the nomenclature, @samp{default} always means public; i.e., 16607available to be linked against from outside the shared object. 16608@samp{protected} and @samp{internal} are pretty useless in real-world 16609usage so the only other commonly used option is @samp{hidden}. 16610The default if @option{-fvisibility} isn't specified is 16611@samp{default}, i.e., make every symbol public. 16612 16613A good explanation of the benefits offered by ensuring ELF 16614symbols have the correct visibility is given by ``How To Write 16615Shared Libraries'' by Ulrich Drepper (which can be found at 16616@w{@uref{https://www.akkadia.org/drepper/}})---however a superior 16617solution made possible by this option to marking things hidden when 16618the default is public is to make the default hidden and mark things 16619public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden} 16620and @code{__attribute__ ((visibility("default")))} instead of 16621@code{__declspec(dllexport)} you get almost identical semantics with 16622identical syntax. This is a great boon to those working with 16623cross-platform projects. 16624 16625For those adding visibility support to existing code, you may find 16626@code{#pragma GCC visibility} of use. This works by you enclosing 16627the declarations you wish to set visibility for with (for example) 16628@code{#pragma GCC visibility push(hidden)} and 16629@code{#pragma GCC visibility pop}. 16630Bear in mind that symbol visibility should be viewed @strong{as 16631part of the API interface contract} and thus all new code should 16632always specify visibility when it is not the default; i.e., declarations 16633only for use within the local DSO should @strong{always} be marked explicitly 16634as hidden as so to avoid PLT indirection overheads---making this 16635abundantly clear also aids readability and self-documentation of the code. 16636Note that due to ISO C++ specification requirements, @code{operator new} and 16637@code{operator delete} must always be of default visibility. 16638 16639Be aware that headers from outside your project, in particular system 16640headers and headers from any other library you use, may not be 16641expecting to be compiled with visibility other than the default. You 16642may need to explicitly say @code{#pragma GCC visibility push(default)} 16643before including any such headers. 16644 16645@code{extern} declarations are not affected by @option{-fvisibility}, so 16646a lot of code can be recompiled with @option{-fvisibility=hidden} with 16647no modifications. However, this means that calls to @code{extern} 16648functions with no explicit visibility use the PLT, so it is more 16649effective to use @code{__attribute ((visibility))} and/or 16650@code{#pragma GCC visibility} to tell the compiler which @code{extern} 16651declarations should be treated as hidden. 16652 16653Note that @option{-fvisibility} does affect C++ vague linkage 16654entities. This means that, for instance, an exception class that is 16655be thrown between DSOs must be explicitly marked with default 16656visibility so that the @samp{type_info} nodes are unified between 16657the DSOs. 16658 16659An overview of these techniques, their benefits and how to use them 16660is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}. 16661 16662@item -fstrict-volatile-bitfields 16663@opindex fstrict-volatile-bitfields 16664This option should be used if accesses to volatile bit-fields (or other 16665structure fields, although the compiler usually honors those types 16666anyway) should use a single access of the width of the 16667field's type, aligned to a natural alignment if possible. For 16668example, targets with memory-mapped peripheral registers might require 16669all such accesses to be 16 bits wide; with this flag you can 16670declare all peripheral bit-fields as @code{unsigned short} (assuming short 16671is 16 bits on these targets) to force GCC to use 16-bit accesses 16672instead of, perhaps, a more efficient 32-bit access. 16673 16674If this option is disabled, the compiler uses the most efficient 16675instruction. In the previous example, that might be a 32-bit load 16676instruction, even though that accesses bytes that do not contain 16677any portion of the bit-field, or memory-mapped registers unrelated to 16678the one being updated. 16679 16680In some cases, such as when the @code{packed} attribute is applied to a 16681structure field, it may not be possible to access the field with a single 16682read or write that is correctly aligned for the target machine. In this 16683case GCC falls back to generating multiple accesses rather than code that 16684will fault or truncate the result at run time. 16685 16686Note: Due to restrictions of the C/C++11 memory model, write accesses are 16687not allowed to touch non bit-field members. It is therefore recommended 16688to define all bits of the field's type as bit-field members. 16689 16690The default value of this option is determined by the application binary 16691interface for the target processor. 16692 16693@item -fsync-libcalls 16694@opindex fsync-libcalls 16695This option controls whether any out-of-line instance of the @code{__sync} 16696family of functions may be used to implement the C++11 @code{__atomic} 16697family of functions. 16698 16699The default value of this option is enabled, thus the only useful form 16700of the option is @option{-fno-sync-libcalls}. This option is used in 16701the implementation of the @file{libatomic} runtime library. 16702 16703@end table 16704 16705@node Developer Options 16706@section GCC Developer Options 16707@cindex developer options 16708@cindex debugging GCC 16709@cindex debug dump options 16710@cindex dump options 16711@cindex compilation statistics 16712 16713This section describes command-line options that are primarily of 16714interest to GCC developers, including options to support compiler 16715testing and investigation of compiler bugs and compile-time 16716performance problems. This includes options that produce debug dumps 16717at various points in the compilation; that print statistics such as 16718memory use and execution time; and that print information about GCC's 16719configuration, such as where it searches for libraries. You should 16720rarely need to use any of these options for ordinary compilation and 16721linking tasks. 16722 16723Many developer options that cause GCC to dump output to a file take an 16724optional @samp{=@var{filename}} suffix. You can specify @samp{stdout} 16725or @samp{-} to dump to standard output, and @samp{stderr} for standard 16726error. 16727 16728If @samp{=@var{filename}} is omitted, a default dump file name is 16729constructed by concatenating the base dump file name, a pass number, 16730phase letter, and pass name. The base dump file name is the name of 16731output file produced by the compiler if explicitly specified and not 16732an executable; otherwise it is the source file name. 16733The pass number is determined by the order passes are registered with 16734the compiler's pass manager. 16735This is generally the same as the order of execution, but passes 16736registered by plugins, target-specific passes, or passes that are 16737otherwise registered late are numbered higher than the pass named 16738@samp{final}, even if they are executed earlier. The phase letter is 16739one of @samp{i} (inter-procedural analysis), @samp{l} 16740(language-specific), @samp{r} (RTL), or @samp{t} (tree). 16741The files are created in the directory of the output file. 16742 16743@table @gcctabopt 16744 16745@item -fcallgraph-info 16746@itemx -fcallgraph-info=@var{MARKERS} 16747@opindex fcallgraph-info 16748Makes the compiler output callgraph information for the program, on a 16749per-object-file basis. The information is generated in the common VCG 16750format. It can be decorated with additional, per-node and/or per-edge 16751information, if a list of comma-separated markers is additionally 16752specified. When the @code{su} marker is specified, the callgraph is 16753decorated with stack usage information; it is equivalent to 16754@option{-fstack-usage}. When the @code{da} marker is specified, the 16755callgraph is decorated with information about dynamically allocated 16756objects. 16757 16758When compiling with @option{-flto}, no callgraph information is output 16759along with the object file. At LTO link time, @option{-fcallgraph-info} 16760may generate multiple callgraph information files next to intermediate 16761LTO output files. 16762 16763@item -d@var{letters} 16764@itemx -fdump-rtl-@var{pass} 16765@itemx -fdump-rtl-@var{pass}=@var{filename} 16766@opindex d 16767@opindex fdump-rtl-@var{pass} 16768Says to make debugging dumps during compilation at times specified by 16769@var{letters}. This is used for debugging the RTL-based passes of the 16770compiler. 16771 16772Some @option{-d@var{letters}} switches have different meaning when 16773@option{-E} is used for preprocessing. @xref{Preprocessor Options}, 16774for information about preprocessor-specific dump options. 16775 16776Debug dumps can be enabled with a @option{-fdump-rtl} switch or some 16777@option{-d} option @var{letters}. Here are the possible 16778letters for use in @var{pass} and @var{letters}, and their meanings: 16779 16780@table @gcctabopt 16781 16782@item -fdump-rtl-alignments 16783@opindex fdump-rtl-alignments 16784Dump after branch alignments have been computed. 16785 16786@item -fdump-rtl-asmcons 16787@opindex fdump-rtl-asmcons 16788Dump after fixing rtl statements that have unsatisfied in/out constraints. 16789 16790@item -fdump-rtl-auto_inc_dec 16791@opindex fdump-rtl-auto_inc_dec 16792Dump after auto-inc-dec discovery. This pass is only run on 16793architectures that have auto inc or auto dec instructions. 16794 16795@item -fdump-rtl-barriers 16796@opindex fdump-rtl-barriers 16797Dump after cleaning up the barrier instructions. 16798 16799@item -fdump-rtl-bbpart 16800@opindex fdump-rtl-bbpart 16801Dump after partitioning hot and cold basic blocks. 16802 16803@item -fdump-rtl-bbro 16804@opindex fdump-rtl-bbro 16805Dump after block reordering. 16806 16807@item -fdump-rtl-btl1 16808@itemx -fdump-rtl-btl2 16809@opindex fdump-rtl-btl2 16810@opindex fdump-rtl-btl2 16811@option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping 16812after the two branch 16813target load optimization passes. 16814 16815@item -fdump-rtl-bypass 16816@opindex fdump-rtl-bypass 16817Dump after jump bypassing and control flow optimizations. 16818 16819@item -fdump-rtl-combine 16820@opindex fdump-rtl-combine 16821Dump after the RTL instruction combination pass. 16822 16823@item -fdump-rtl-compgotos 16824@opindex fdump-rtl-compgotos 16825Dump after duplicating the computed gotos. 16826 16827@item -fdump-rtl-ce1 16828@itemx -fdump-rtl-ce2 16829@itemx -fdump-rtl-ce3 16830@opindex fdump-rtl-ce1 16831@opindex fdump-rtl-ce2 16832@opindex fdump-rtl-ce3 16833@option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and 16834@option{-fdump-rtl-ce3} enable dumping after the three 16835if conversion passes. 16836 16837@item -fdump-rtl-cprop_hardreg 16838@opindex fdump-rtl-cprop_hardreg 16839Dump after hard register copy propagation. 16840 16841@item -fdump-rtl-csa 16842@opindex fdump-rtl-csa 16843Dump after combining stack adjustments. 16844 16845@item -fdump-rtl-cse1 16846@itemx -fdump-rtl-cse2 16847@opindex fdump-rtl-cse1 16848@opindex fdump-rtl-cse2 16849@option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after 16850the two common subexpression elimination passes. 16851 16852@item -fdump-rtl-dce 16853@opindex fdump-rtl-dce 16854Dump after the standalone dead code elimination passes. 16855 16856@item -fdump-rtl-dbr 16857@opindex fdump-rtl-dbr 16858Dump after delayed branch scheduling. 16859 16860@item -fdump-rtl-dce1 16861@itemx -fdump-rtl-dce2 16862@opindex fdump-rtl-dce1 16863@opindex fdump-rtl-dce2 16864@option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after 16865the two dead store elimination passes. 16866 16867@item -fdump-rtl-eh 16868@opindex fdump-rtl-eh 16869Dump after finalization of EH handling code. 16870 16871@item -fdump-rtl-eh_ranges 16872@opindex fdump-rtl-eh_ranges 16873Dump after conversion of EH handling range regions. 16874 16875@item -fdump-rtl-expand 16876@opindex fdump-rtl-expand 16877Dump after RTL generation. 16878 16879@item -fdump-rtl-fwprop1 16880@itemx -fdump-rtl-fwprop2 16881@opindex fdump-rtl-fwprop1 16882@opindex fdump-rtl-fwprop2 16883@option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable 16884dumping after the two forward propagation passes. 16885 16886@item -fdump-rtl-gcse1 16887@itemx -fdump-rtl-gcse2 16888@opindex fdump-rtl-gcse1 16889@opindex fdump-rtl-gcse2 16890@option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping 16891after global common subexpression elimination. 16892 16893@item -fdump-rtl-init-regs 16894@opindex fdump-rtl-init-regs 16895Dump after the initialization of the registers. 16896 16897@item -fdump-rtl-initvals 16898@opindex fdump-rtl-initvals 16899Dump after the computation of the initial value sets. 16900 16901@item -fdump-rtl-into_cfglayout 16902@opindex fdump-rtl-into_cfglayout 16903Dump after converting to cfglayout mode. 16904 16905@item -fdump-rtl-ira 16906@opindex fdump-rtl-ira 16907Dump after iterated register allocation. 16908 16909@item -fdump-rtl-jump 16910@opindex fdump-rtl-jump 16911Dump after the second jump optimization. 16912 16913@item -fdump-rtl-loop2 16914@opindex fdump-rtl-loop2 16915@option{-fdump-rtl-loop2} enables dumping after the rtl 16916loop optimization passes. 16917 16918@item -fdump-rtl-mach 16919@opindex fdump-rtl-mach 16920Dump after performing the machine dependent reorganization pass, if that 16921pass exists. 16922 16923@item -fdump-rtl-mode_sw 16924@opindex fdump-rtl-mode_sw 16925Dump after removing redundant mode switches. 16926 16927@item -fdump-rtl-rnreg 16928@opindex fdump-rtl-rnreg 16929Dump after register renumbering. 16930 16931@item -fdump-rtl-outof_cfglayout 16932@opindex fdump-rtl-outof_cfglayout 16933Dump after converting from cfglayout mode. 16934 16935@item -fdump-rtl-peephole2 16936@opindex fdump-rtl-peephole2 16937Dump after the peephole pass. 16938 16939@item -fdump-rtl-postreload 16940@opindex fdump-rtl-postreload 16941Dump after post-reload optimizations. 16942 16943@item -fdump-rtl-pro_and_epilogue 16944@opindex fdump-rtl-pro_and_epilogue 16945Dump after generating the function prologues and epilogues. 16946 16947@item -fdump-rtl-sched1 16948@itemx -fdump-rtl-sched2 16949@opindex fdump-rtl-sched1 16950@opindex fdump-rtl-sched2 16951@option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping 16952after the basic block scheduling passes. 16953 16954@item -fdump-rtl-ree 16955@opindex fdump-rtl-ree 16956Dump after sign/zero extension elimination. 16957 16958@item -fdump-rtl-seqabstr 16959@opindex fdump-rtl-seqabstr 16960Dump after common sequence discovery. 16961 16962@item -fdump-rtl-shorten 16963@opindex fdump-rtl-shorten 16964Dump after shortening branches. 16965 16966@item -fdump-rtl-sibling 16967@opindex fdump-rtl-sibling 16968Dump after sibling call optimizations. 16969 16970@item -fdump-rtl-split1 16971@itemx -fdump-rtl-split2 16972@itemx -fdump-rtl-split3 16973@itemx -fdump-rtl-split4 16974@itemx -fdump-rtl-split5 16975@opindex fdump-rtl-split1 16976@opindex fdump-rtl-split2 16977@opindex fdump-rtl-split3 16978@opindex fdump-rtl-split4 16979@opindex fdump-rtl-split5 16980These options enable dumping after five rounds of 16981instruction splitting. 16982 16983@item -fdump-rtl-sms 16984@opindex fdump-rtl-sms 16985Dump after modulo scheduling. This pass is only run on some 16986architectures. 16987 16988@item -fdump-rtl-stack 16989@opindex fdump-rtl-stack 16990Dump after conversion from GCC's ``flat register file'' registers to the 16991x87's stack-like registers. This pass is only run on x86 variants. 16992 16993@item -fdump-rtl-subreg1 16994@itemx -fdump-rtl-subreg2 16995@opindex fdump-rtl-subreg1 16996@opindex fdump-rtl-subreg2 16997@option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after 16998the two subreg expansion passes. 16999 17000@item -fdump-rtl-unshare 17001@opindex fdump-rtl-unshare 17002Dump after all rtl has been unshared. 17003 17004@item -fdump-rtl-vartrack 17005@opindex fdump-rtl-vartrack 17006Dump after variable tracking. 17007 17008@item -fdump-rtl-vregs 17009@opindex fdump-rtl-vregs 17010Dump after converting virtual registers to hard registers. 17011 17012@item -fdump-rtl-web 17013@opindex fdump-rtl-web 17014Dump after live range splitting. 17015 17016@item -fdump-rtl-regclass 17017@itemx -fdump-rtl-subregs_of_mode_init 17018@itemx -fdump-rtl-subregs_of_mode_finish 17019@itemx -fdump-rtl-dfinit 17020@itemx -fdump-rtl-dfinish 17021@opindex fdump-rtl-regclass 17022@opindex fdump-rtl-subregs_of_mode_init 17023@opindex fdump-rtl-subregs_of_mode_finish 17024@opindex fdump-rtl-dfinit 17025@opindex fdump-rtl-dfinish 17026These dumps are defined but always produce empty files. 17027 17028@item -da 17029@itemx -fdump-rtl-all 17030@opindex da 17031@opindex fdump-rtl-all 17032Produce all the dumps listed above. 17033 17034@item -dA 17035@opindex dA 17036Annotate the assembler output with miscellaneous debugging information. 17037 17038@item -dD 17039@opindex dD 17040Dump all macro definitions, at the end of preprocessing, in addition to 17041normal output. 17042 17043@item -dH 17044@opindex dH 17045Produce a core dump whenever an error occurs. 17046 17047@item -dp 17048@opindex dp 17049Annotate the assembler output with a comment indicating which 17050pattern and alternative is used. The length and cost of each instruction are 17051also printed. 17052 17053@item -dP 17054@opindex dP 17055Dump the RTL in the assembler output as a comment before each instruction. 17056Also turns on @option{-dp} annotation. 17057 17058@item -dx 17059@opindex dx 17060Just generate RTL for a function instead of compiling it. Usually used 17061with @option{-fdump-rtl-expand}. 17062@end table 17063 17064@item -fdump-debug 17065@opindex fdump-debug 17066Dump debugging information generated during the debug 17067generation phase. 17068 17069@item -fdump-earlydebug 17070@opindex fdump-earlydebug 17071Dump debugging information generated during the early debug 17072generation phase. 17073 17074@item -fdump-noaddr 17075@opindex fdump-noaddr 17076When doing debugging dumps, suppress address output. This makes it more 17077feasible to use diff on debugging dumps for compiler invocations with 17078different compiler binaries and/or different 17079text / bss / data / heap / stack / dso start locations. 17080 17081@item -freport-bug 17082@opindex freport-bug 17083Collect and dump debug information into a temporary file if an 17084internal compiler error (ICE) occurs. 17085 17086@item -fdump-unnumbered 17087@opindex fdump-unnumbered 17088When doing debugging dumps, suppress instruction numbers and address output. 17089This makes it more feasible to use diff on debugging dumps for compiler 17090invocations with different options, in particular with and without 17091@option{-g}. 17092 17093@item -fdump-unnumbered-links 17094@opindex fdump-unnumbered-links 17095When doing debugging dumps (see @option{-d} option above), suppress 17096instruction numbers for the links to the previous and next instructions 17097in a sequence. 17098 17099@item -fdump-ipa-@var{switch} 17100@itemx -fdump-ipa-@var{switch}-@var{options} 17101@opindex fdump-ipa 17102Control the dumping at various stages of inter-procedural analysis 17103language tree to a file. The file name is generated by appending a 17104switch specific suffix to the source file name, and the file is created 17105in the same directory as the output file. The following dumps are 17106possible: 17107 17108@table @samp 17109@item all 17110Enables all inter-procedural analysis dumps. 17111 17112@item cgraph 17113Dumps information about call-graph optimization, unused function removal, 17114and inlining decisions. 17115 17116@item inline 17117Dump after function inlining. 17118 17119@end table 17120 17121Additionally, the options @option{-optimized}, @option{-missed}, 17122@option{-note}, and @option{-all} can be provided, with the same meaning 17123as for @option{-fopt-info}, defaulting to @option{-optimized}. 17124 17125For example, @option{-fdump-ipa-inline-optimized-missed} will emit 17126information on callsites that were inlined, along with callsites 17127that were not inlined. 17128 17129By default, the dump will contain messages about successful 17130optimizations (equivalent to @option{-optimized}) together with 17131low-level details about the analysis. 17132 17133@item -fdump-lang 17134@opindex fdump-lang 17135Dump language-specific information. The file name is made by appending 17136@file{.lang} to the source file name. 17137 17138@item -fdump-lang-all 17139@itemx -fdump-lang-@var{switch} 17140@itemx -fdump-lang-@var{switch}-@var{options} 17141@itemx -fdump-lang-@var{switch}-@var{options}=@var{filename} 17142@opindex fdump-lang-all 17143@opindex fdump-lang 17144Control the dumping of language-specific information. The @var{options} 17145and @var{filename} portions behave as described in the 17146@option{-fdump-tree} option. The following @var{switch} values are 17147accepted: 17148 17149@table @samp 17150@item all 17151 17152Enable all language-specific dumps. 17153 17154@item class 17155Dump class hierarchy information. Virtual table information is emitted 17156unless '@option{slim}' is specified. This option is applicable to C++ only. 17157 17158@item module 17159Dump module information. Options @option{lineno} (locations), 17160@option{graph} (reachability), @option{blocks} (clusters), 17161@option{uid} (serialization), @option{alias} (mergeable), 17162@option{asmname} (Elrond), @option{eh} (mapper) & @option{vops} 17163(macros) may provide additional information. This option is 17164applicable to C++ only. 17165 17166@item raw 17167Dump the raw internal tree data. This option is applicable to C++ only. 17168 17169@end table 17170 17171@item -fdump-passes 17172@opindex fdump-passes 17173Print on @file{stderr} the list of optimization passes that are turned 17174on and off by the current command-line options. 17175 17176@item -fdump-statistics-@var{option} 17177@opindex fdump-statistics 17178Enable and control dumping of pass statistics in a separate file. The 17179file name is generated by appending a suffix ending in 17180@samp{.statistics} to the source file name, and the file is created in 17181the same directory as the output file. If the @samp{-@var{option}} 17182form is used, @samp{-stats} causes counters to be summed over the 17183whole compilation unit while @samp{-details} dumps every event as 17184the passes generate them. The default with no option is to sum 17185counters for each function compiled. 17186 17187@item -fdump-tree-all 17188@itemx -fdump-tree-@var{switch} 17189@itemx -fdump-tree-@var{switch}-@var{options} 17190@itemx -fdump-tree-@var{switch}-@var{options}=@var{filename} 17191@opindex fdump-tree-all 17192@opindex fdump-tree 17193Control the dumping at various stages of processing the intermediate 17194language tree to a file. If the @samp{-@var{options}} 17195form is used, @var{options} is a list of @samp{-} separated options 17196which control the details of the dump. Not all options are applicable 17197to all dumps; those that are not meaningful are ignored. The 17198following options are available 17199 17200@table @samp 17201@item address 17202Print the address of each node. Usually this is not meaningful as it 17203changes according to the environment and source file. Its primary use 17204is for tying up a dump file with a debug environment. 17205@item asmname 17206If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that 17207in the dump instead of @code{DECL_NAME}. Its primary use is ease of 17208use working backward from mangled names in the assembly file. 17209@item slim 17210When dumping front-end intermediate representations, inhibit dumping 17211of members of a scope or body of a function merely because that scope 17212has been reached. Only dump such items when they are directly reachable 17213by some other path. 17214 17215When dumping pretty-printed trees, this option inhibits dumping the 17216bodies of control structures. 17217 17218When dumping RTL, print the RTL in slim (condensed) form instead of 17219the default LISP-like representation. 17220@item raw 17221Print a raw representation of the tree. By default, trees are 17222pretty-printed into a C-like representation. 17223@item details 17224Enable more detailed dumps (not honored by every dump option). Also 17225include information from the optimization passes. 17226@item stats 17227Enable dumping various statistics about the pass (not honored by every dump 17228option). 17229@item blocks 17230Enable showing basic block boundaries (disabled in raw dumps). 17231@item graph 17232For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}), 17233dump a representation of the control flow graph suitable for viewing with 17234GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in 17235the file is pretty-printed as a subgraph, so that GraphViz can render them 17236all in a single plot. 17237 17238This option currently only works for RTL dumps, and the RTL is always 17239dumped in slim form. 17240@item vops 17241Enable showing virtual operands for every statement. 17242@item lineno 17243Enable showing line numbers for statements. 17244@item uid 17245Enable showing the unique ID (@code{DECL_UID}) for each variable. 17246@item verbose 17247Enable showing the tree dump for each statement. 17248@item eh 17249Enable showing the EH region number holding each statement. 17250@item scev 17251Enable showing scalar evolution analysis details. 17252@item optimized 17253Enable showing optimization information (only available in certain 17254passes). 17255@item missed 17256Enable showing missed optimization information (only available in certain 17257passes). 17258@item note 17259Enable other detailed optimization information (only available in 17260certain passes). 17261@item all 17262Turn on all options, except @option{raw}, @option{slim}, @option{verbose} 17263and @option{lineno}. 17264@item optall 17265Turn on all optimization options, i.e., @option{optimized}, 17266@option{missed}, and @option{note}. 17267@end table 17268 17269To determine what tree dumps are available or find the dump for a pass 17270of interest follow the steps below. 17271 17272@enumerate 17273@item 17274Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output 17275look for a code that corresponds to the pass you are interested in. 17276For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and 17277@code{tree-vrp2} correspond to the three Value Range Propagation passes. 17278The number at the end distinguishes distinct invocations of the same pass. 17279@item 17280To enable the creation of the dump file, append the pass code to 17281the @option{-fdump-} option prefix and invoke GCC with it. For example, 17282to enable the dump from the Early Value Range Propagation pass, invoke 17283GCC with the @option{-fdump-tree-evrp} option. Optionally, you may 17284specify the name of the dump file. If you don't specify one, GCC 17285creates as described below. 17286@item 17287Find the pass dump in a file whose name is composed of three components 17288separated by a period: the name of the source file GCC was invoked to 17289compile, a numeric suffix indicating the pass number followed by the 17290letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes), 17291and finally the pass code. For example, the Early VRP pass dump might 17292be in a file named @file{myfile.c.038t.evrp} in the current working 17293directory. Note that the numeric codes are not stable and may change 17294from one version of GCC to another. 17295@end enumerate 17296 17297@item -fopt-info 17298@itemx -fopt-info-@var{options} 17299@itemx -fopt-info-@var{options}=@var{filename} 17300@opindex fopt-info 17301Controls optimization dumps from various optimization passes. If the 17302@samp{-@var{options}} form is used, @var{options} is a list of 17303@samp{-} separated option keywords to select the dump details and 17304optimizations. 17305 17306The @var{options} can be divided into three groups: 17307@enumerate 17308@item 17309options describing what kinds of messages should be emitted, 17310@item 17311options describing the verbosity of the dump, and 17312@item 17313options describing which optimizations should be included. 17314@end enumerate 17315The options from each group can be freely mixed as they are 17316non-overlapping. However, in case of any conflicts, 17317the later options override the earlier options on the command 17318line. 17319 17320The following options control which kinds of messages should be emitted: 17321 17322@table @samp 17323@item optimized 17324Print information when an optimization is successfully applied. It is 17325up to a pass to decide which information is relevant. For example, the 17326vectorizer passes print the source location of loops which are 17327successfully vectorized. 17328@item missed 17329Print information about missed optimizations. Individual passes 17330control which information to include in the output. 17331@item note 17332Print verbose information about optimizations, such as certain 17333transformations, more detailed messages about decisions etc. 17334@item all 17335Print detailed optimization information. This includes 17336@samp{optimized}, @samp{missed}, and @samp{note}. 17337@end table 17338 17339The following option controls the dump verbosity: 17340 17341@table @samp 17342@item internals 17343By default, only ``high-level'' messages are emitted. This option enables 17344additional, more detailed, messages, which are likely to only be of interest 17345to GCC developers. 17346@end table 17347 17348One or more of the following option keywords can be used to describe a 17349group of optimizations: 17350 17351@table @samp 17352@item ipa 17353Enable dumps from all interprocedural optimizations. 17354@item loop 17355Enable dumps from all loop optimizations. 17356@item inline 17357Enable dumps from all inlining optimizations. 17358@item omp 17359Enable dumps from all OMP (Offloading and Multi Processing) optimizations. 17360@item vec 17361Enable dumps from all vectorization optimizations. 17362@item optall 17363Enable dumps from all optimizations. This is a superset of 17364the optimization groups listed above. 17365@end table 17366 17367If @var{options} is 17368omitted, it defaults to @samp{optimized-optall}, which means to dump messages 17369about successful optimizations from all the passes, omitting messages 17370that are treated as ``internals''. 17371 17372If the @var{filename} is provided, then the dumps from all the 17373applicable optimizations are concatenated into the @var{filename}. 17374Otherwise the dump is output onto @file{stderr}. Though multiple 17375@option{-fopt-info} options are accepted, only one of them can include 17376a @var{filename}. If other filenames are provided then all but the 17377first such option are ignored. 17378 17379Note that the output @var{filename} is overwritten 17380in case of multiple translation units. If a combined output from 17381multiple translation units is desired, @file{stderr} should be used 17382instead. 17383 17384In the following example, the optimization info is output to 17385@file{stderr}: 17386 17387@smallexample 17388gcc -O3 -fopt-info 17389@end smallexample 17390 17391This example: 17392@smallexample 17393gcc -O3 -fopt-info-missed=missed.all 17394@end smallexample 17395 17396@noindent 17397outputs missed optimization report from all the passes into 17398@file{missed.all}, and this one: 17399 17400@smallexample 17401gcc -O2 -ftree-vectorize -fopt-info-vec-missed 17402@end smallexample 17403 17404@noindent 17405prints information about missed optimization opportunities from 17406vectorization passes on @file{stderr}. 17407Note that @option{-fopt-info-vec-missed} is equivalent to 17408@option{-fopt-info-missed-vec}. The order of the optimization group 17409names and message types listed after @option{-fopt-info} does not matter. 17410 17411As another example, 17412@smallexample 17413gcc -O3 -fopt-info-inline-optimized-missed=inline.txt 17414@end smallexample 17415 17416@noindent 17417outputs information about missed optimizations as well as 17418optimized locations from all the inlining passes into 17419@file{inline.txt}. 17420 17421Finally, consider: 17422 17423@smallexample 17424gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt 17425@end smallexample 17426 17427@noindent 17428Here the two output filenames @file{vec.miss} and @file{loop.opt} are 17429in conflict since only one output file is allowed. In this case, only 17430the first option takes effect and the subsequent options are 17431ignored. Thus only @file{vec.miss} is produced which contains 17432dumps from the vectorizer about missed opportunities. 17433 17434@item -fsave-optimization-record 17435@opindex fsave-optimization-record 17436Write a SRCFILE.opt-record.json.gz file detailing what optimizations 17437were performed, for those optimizations that support @option{-fopt-info}. 17438 17439This option is experimental and the format of the data within the 17440compressed JSON file is subject to change. 17441 17442It is roughly equivalent to a machine-readable version of 17443@option{-fopt-info-all}, as a collection of messages with source file, 17444line number and column number, with the following additional data for 17445each message: 17446 17447@itemize @bullet 17448 17449@item 17450the execution count of the code being optimized, along with metadata about 17451whether this was from actual profile data, or just an estimate, allowing 17452consumers to prioritize messages by code hotness, 17453 17454@item 17455the function name of the code being optimized, where applicable, 17456 17457@item 17458the ``inlining chain'' for the code being optimized, so that when 17459a function is inlined into several different places (which might 17460themselves be inlined), the reader can distinguish between the copies, 17461 17462@item 17463objects identifying those parts of the message that refer to expressions, 17464statements or symbol-table nodes, which of these categories they are, and, 17465when available, their source code location, 17466 17467@item 17468the GCC pass that emitted the message, and 17469 17470@item 17471the location in GCC's own code from which the message was emitted 17472 17473@end itemize 17474 17475Additionally, some messages are logically nested within other 17476messages, reflecting implementation details of the optimization 17477passes. 17478 17479@item -fsched-verbose=@var{n} 17480@opindex fsched-verbose 17481On targets that use instruction scheduling, this option controls the 17482amount of debugging output the scheduler prints to the dump files. 17483 17484For @var{n} greater than zero, @option{-fsched-verbose} outputs the 17485same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}. 17486For @var{n} greater than one, it also output basic block probabilities, 17487detailed ready list information and unit/insn info. For @var{n} greater 17488than two, it includes RTL at abort point, control-flow and regions info. 17489And for @var{n} over four, @option{-fsched-verbose} also includes 17490dependence info. 17491 17492 17493 17494@item -fenable-@var{kind}-@var{pass} 17495@itemx -fdisable-@var{kind}-@var{pass}=@var{range-list} 17496@opindex fdisable- 17497@opindex fenable- 17498 17499This is a set of options that are used to explicitly disable/enable 17500optimization passes. These options are intended for use for debugging GCC. 17501Compiler users should use regular options for enabling/disabling 17502passes instead. 17503 17504@table @gcctabopt 17505 17506@item -fdisable-ipa-@var{pass} 17507Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is 17508statically invoked in the compiler multiple times, the pass name should be 17509appended with a sequential number starting from 1. 17510 17511@item -fdisable-rtl-@var{pass} 17512@itemx -fdisable-rtl-@var{pass}=@var{range-list} 17513Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is 17514statically invoked in the compiler multiple times, the pass name should be 17515appended with a sequential number starting from 1. @var{range-list} is a 17516comma-separated list of function ranges or assembler names. Each range is a number 17517pair separated by a colon. The range is inclusive in both ends. If the range 17518is trivial, the number pair can be simplified as a single number. If the 17519function's call graph node's @var{uid} falls within one of the specified ranges, 17520the @var{pass} is disabled for that function. The @var{uid} is shown in the 17521function header of a dump file, and the pass names can be dumped by using 17522option @option{-fdump-passes}. 17523 17524@item -fdisable-tree-@var{pass} 17525@itemx -fdisable-tree-@var{pass}=@var{range-list} 17526Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of 17527option arguments. 17528 17529@item -fenable-ipa-@var{pass} 17530Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is 17531statically invoked in the compiler multiple times, the pass name should be 17532appended with a sequential number starting from 1. 17533 17534@item -fenable-rtl-@var{pass} 17535@itemx -fenable-rtl-@var{pass}=@var{range-list} 17536Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument 17537description and examples. 17538 17539@item -fenable-tree-@var{pass} 17540@itemx -fenable-tree-@var{pass}=@var{range-list} 17541Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description 17542of option arguments. 17543 17544@end table 17545 17546Here are some examples showing uses of these options. 17547 17548@smallexample 17549 17550# disable ccp1 for all functions 17551 -fdisable-tree-ccp1 17552# disable complete unroll for function whose cgraph node uid is 1 17553 -fenable-tree-cunroll=1 17554# disable gcse2 for functions at the following ranges [1,1], 17555# [300,400], and [400,1000] 17556# disable gcse2 for functions foo and foo2 17557 -fdisable-rtl-gcse2=foo,foo2 17558# disable early inlining 17559 -fdisable-tree-einline 17560# disable ipa inlining 17561 -fdisable-ipa-inline 17562# enable tree full unroll 17563 -fenable-tree-unroll 17564 17565@end smallexample 17566 17567@item -fchecking 17568@itemx -fchecking=@var{n} 17569@opindex fchecking 17570@opindex fno-checking 17571Enable internal consistency checking. The default depends on 17572the compiler configuration. @option{-fchecking=2} enables further 17573internal consistency checking that might affect code generation. 17574 17575@item -frandom-seed=@var{string} 17576@opindex frandom-seed 17577This option provides a seed that GCC uses in place of 17578random numbers in generating certain symbol names 17579that have to be different in every compiled file. It is also used to 17580place unique stamps in coverage data files and the object files that 17581produce them. You can use the @option{-frandom-seed} option to produce 17582reproducibly identical object files. 17583 17584The @var{string} can either be a number (decimal, octal or hex) or an 17585arbitrary string (in which case it's converted to a number by 17586computing CRC32). 17587 17588The @var{string} should be different for every file you compile. 17589 17590@item -save-temps 17591@opindex save-temps 17592Store the usual ``temporary'' intermediate files permanently; name them 17593as auxiliary output files, as specified described under 17594@option{-dumpbase} and @option{-dumpdir}. 17595 17596When used in combination with the @option{-x} command-line option, 17597@option{-save-temps} is sensible enough to avoid overwriting an 17598input source file with the same extension as an intermediate file. 17599The corresponding intermediate file may be obtained by renaming the 17600source file before using @option{-save-temps}. 17601 17602@item -save-temps=cwd 17603@opindex save-temps=cwd 17604Equivalent to @option{-save-temps -dumpdir ./}. 17605 17606@item -save-temps=obj 17607@opindex save-temps=obj 17608Equivalent to @option{-save-temps -dumpdir @file{outdir/}}, where 17609@file{outdir/} is the directory of the output file specified after the 17610@option{-o} option, including any directory separators. If the 17611@option{-o} option is not used, the @option{-save-temps=obj} switch 17612behaves like @option{-save-temps=cwd}. 17613 17614@item -time@r{[}=@var{file}@r{]} 17615@opindex time 17616Report the CPU time taken by each subprocess in the compilation 17617sequence. For C source files, this is the compiler proper and assembler 17618(plus the linker if linking is done). 17619 17620Without the specification of an output file, the output looks like this: 17621 17622@smallexample 17623# cc1 0.12 0.01 17624# as 0.00 0.01 17625@end smallexample 17626 17627The first number on each line is the ``user time'', that is time spent 17628executing the program itself. The second number is ``system time'', 17629time spent executing operating system routines on behalf of the program. 17630Both numbers are in seconds. 17631 17632With the specification of an output file, the output is appended to the 17633named file, and it looks like this: 17634 17635@smallexample 176360.12 0.01 cc1 @var{options} 176370.00 0.01 as @var{options} 17638@end smallexample 17639 17640The ``user time'' and the ``system time'' are moved before the program 17641name, and the options passed to the program are displayed, so that one 17642can later tell what file was being compiled, and with which options. 17643 17644@item -fdump-final-insns@r{[}=@var{file}@r{]} 17645@opindex fdump-final-insns 17646Dump the final internal representation (RTL) to @var{file}. If the 17647optional argument is omitted (or if @var{file} is @code{.}), the name 17648of the dump file is determined by appending @code{.gkd} to the 17649dump base name, see @option{-dumpbase}. 17650 17651@item -fcompare-debug@r{[}=@var{opts}@r{]} 17652@opindex fcompare-debug 17653@opindex fno-compare-debug 17654If no error occurs during compilation, run the compiler a second time, 17655adding @var{opts} and @option{-fcompare-debug-second} to the arguments 17656passed to the second compilation. Dump the final internal 17657representation in both compilations, and print an error if they differ. 17658 17659If the equal sign is omitted, the default @option{-gtoggle} is used. 17660 17661The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty 17662and nonzero, implicitly enables @option{-fcompare-debug}. If 17663@env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash, 17664then it is used for @var{opts}, otherwise the default @option{-gtoggle} 17665is used. 17666 17667@option{-fcompare-debug=}, with the equal sign but without @var{opts}, 17668is equivalent to @option{-fno-compare-debug}, which disables the dumping 17669of the final representation and the second compilation, preventing even 17670@env{GCC_COMPARE_DEBUG} from taking effect. 17671 17672To verify full coverage during @option{-fcompare-debug} testing, set 17673@env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden}, 17674which GCC rejects as an invalid option in any actual compilation 17675(rather than preprocessing, assembly or linking). To get just a 17676warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug 17677not overridden} will do. 17678 17679@item -fcompare-debug-second 17680@opindex fcompare-debug-second 17681This option is implicitly passed to the compiler for the second 17682compilation requested by @option{-fcompare-debug}, along with options to 17683silence warnings, and omitting other options that would cause the compiler 17684to produce output to files or to standard output as a side effect. Dump 17685files and preserved temporary files are renamed so as to contain the 17686@code{.gk} additional extension during the second compilation, to avoid 17687overwriting those generated by the first. 17688 17689When this option is passed to the compiler driver, it causes the 17690@emph{first} compilation to be skipped, which makes it useful for little 17691other than debugging the compiler proper. 17692 17693@item -gtoggle 17694@opindex gtoggle 17695Turn off generation of debug info, if leaving out this option 17696generates it, or turn it on at level 2 otherwise. The position of this 17697argument in the command line does not matter; it takes effect after all 17698other options are processed, and it does so only once, no matter how 17699many times it is given. This is mainly intended to be used with 17700@option{-fcompare-debug}. 17701 17702@item -fvar-tracking-assignments-toggle 17703@opindex fvar-tracking-assignments-toggle 17704@opindex fno-var-tracking-assignments-toggle 17705Toggle @option{-fvar-tracking-assignments}, in the same way that 17706@option{-gtoggle} toggles @option{-g}. 17707 17708@item -Q 17709@opindex Q 17710Makes the compiler print out each function name as it is compiled, and 17711print some statistics about each pass when it finishes. 17712 17713@item -ftime-report 17714@opindex ftime-report 17715Makes the compiler print some statistics about the time consumed by each 17716pass when it finishes. 17717 17718@item -ftime-report-details 17719@opindex ftime-report-details 17720Record the time consumed by infrastructure parts separately for each pass. 17721 17722@item -fira-verbose=@var{n} 17723@opindex fira-verbose 17724Control the verbosity of the dump file for the integrated register allocator. 17725The default value is 5. If the value @var{n} is greater or equal to 10, 17726the dump output is sent to stderr using the same format as @var{n} minus 10. 17727 17728@item -flto-report 17729@opindex flto-report 17730Prints a report with internal details on the workings of the link-time 17731optimizer. The contents of this report vary from version to version. 17732It is meant to be useful to GCC developers when processing object 17733files in LTO mode (via @option{-flto}). 17734 17735Disabled by default. 17736 17737@item -flto-report-wpa 17738@opindex flto-report-wpa 17739Like @option{-flto-report}, but only print for the WPA phase of link-time 17740optimization. 17741 17742@item -fmem-report 17743@opindex fmem-report 17744Makes the compiler print some statistics about permanent memory 17745allocation when it finishes. 17746 17747@item -fmem-report-wpa 17748@opindex fmem-report-wpa 17749Makes the compiler print some statistics about permanent memory 17750allocation for the WPA phase only. 17751 17752@item -fpre-ipa-mem-report 17753@opindex fpre-ipa-mem-report 17754@item -fpost-ipa-mem-report 17755@opindex fpost-ipa-mem-report 17756Makes the compiler print some statistics about permanent memory 17757allocation before or after interprocedural optimization. 17758 17759@item -fprofile-report 17760@opindex fprofile-report 17761Makes the compiler print some statistics about consistency of the 17762(estimated) profile and effect of individual passes. 17763 17764@item -fstack-usage 17765@opindex fstack-usage 17766Makes the compiler output stack usage information for the program, on a 17767per-function basis. The filename for the dump is made by appending 17768@file{.su} to the @var{auxname}. @var{auxname} is generated from the name of 17769the output file, if explicitly specified and it is not an executable, 17770otherwise it is the basename of the source file. An entry is made up 17771of three fields: 17772 17773@itemize 17774@item 17775The name of the function. 17776@item 17777A number of bytes. 17778@item 17779One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}. 17780@end itemize 17781 17782The qualifier @code{static} means that the function manipulates the stack 17783statically: a fixed number of bytes are allocated for the frame on function 17784entry and released on function exit; no stack adjustments are otherwise made 17785in the function. The second field is this fixed number of bytes. 17786 17787The qualifier @code{dynamic} means that the function manipulates the stack 17788dynamically: in addition to the static allocation described above, stack 17789adjustments are made in the body of the function, for example to push/pop 17790arguments around function calls. If the qualifier @code{bounded} is also 17791present, the amount of these adjustments is bounded at compile time and 17792the second field is an upper bound of the total amount of stack used by 17793the function. If it is not present, the amount of these adjustments is 17794not bounded at compile time and the second field only represents the 17795bounded part. 17796 17797@item -fstats 17798@opindex fstats 17799Emit statistics about front-end processing at the end of the compilation. 17800This option is supported only by the C++ front end, and 17801the information is generally only useful to the G++ development team. 17802 17803@item -fdbg-cnt-list 17804@opindex fdbg-cnt-list 17805Print the name and the counter upper bound for all debug counters. 17806 17807 17808@item -fdbg-cnt=@var{counter-value-list} 17809@opindex fdbg-cnt 17810Set the internal debug counter lower and upper bound. @var{counter-value-list} 17811is a comma-separated list of @var{name}:@var{lower_bound1}-@var{upper_bound1} 17812[:@var{lower_bound2}-@var{upper_bound2}...] tuples which sets 17813the name of the counter and list of closed intervals. 17814The @var{lower_bound} is optional and is zero 17815initialized if not set. 17816For example, with @option{-fdbg-cnt=dce:2-4:10-11,tail_call:10}, 17817@code{dbg_cnt(dce)} returns true only for second, third, fourth, tenth and 17818eleventh invocation. 17819For @code{dbg_cnt(tail_call)} true is returned for first 10 invocations. 17820 17821@item -print-file-name=@var{library} 17822@opindex print-file-name 17823Print the full absolute name of the library file @var{library} that 17824would be used when linking---and don't do anything else. With this 17825option, GCC does not compile or link anything; it just prints the 17826file name. 17827 17828@item -print-multi-directory 17829@opindex print-multi-directory 17830Print the directory name corresponding to the multilib selected by any 17831other switches present in the command line. This directory is supposed 17832to exist in @env{GCC_EXEC_PREFIX}. 17833 17834@item -print-multi-lib 17835@opindex print-multi-lib 17836Print the mapping from multilib directory names to compiler switches 17837that enable them. The directory name is separated from the switches by 17838@samp{;}, and each switch starts with an @samp{@@} instead of the 17839@samp{-}, without spaces between multiple switches. This is supposed to 17840ease shell processing. 17841 17842@item -print-multi-os-directory 17843@opindex print-multi-os-directory 17844Print the path to OS libraries for the selected 17845multilib, relative to some @file{lib} subdirectory. If OS libraries are 17846present in the @file{lib} subdirectory and no multilibs are used, this is 17847usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}} 17848sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or 17849@file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}} 17850subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}. 17851 17852@item -print-multiarch 17853@opindex print-multiarch 17854Print the path to OS libraries for the selected multiarch, 17855relative to some @file{lib} subdirectory. 17856 17857@item -print-prog-name=@var{program} 17858@opindex print-prog-name 17859Like @option{-print-file-name}, but searches for a program such as @command{cpp}. 17860 17861@item -print-libgcc-file-name 17862@opindex print-libgcc-file-name 17863Same as @option{-print-file-name=libgcc.a}. 17864 17865This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs} 17866but you do want to link with @file{libgcc.a}. You can do: 17867 17868@smallexample 17869gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name` 17870@end smallexample 17871 17872@item -print-search-dirs 17873@opindex print-search-dirs 17874Print the name of the configured installation directory and a list of 17875program and library directories @command{gcc} searches---and don't do anything else. 17876 17877This is useful when @command{gcc} prints the error message 17878@samp{installation problem, cannot exec cpp0: No such file or directory}. 17879To resolve this you either need to put @file{cpp0} and the other compiler 17880components where @command{gcc} expects to find them, or you can set the environment 17881variable @env{GCC_EXEC_PREFIX} to the directory where you installed them. 17882Don't forget the trailing @samp{/}. 17883@xref{Environment Variables}. 17884 17885@item -print-sysroot 17886@opindex print-sysroot 17887Print the target sysroot directory that is used during 17888compilation. This is the target sysroot specified either at configure 17889time or using the @option{--sysroot} option, possibly with an extra 17890suffix that depends on compilation options. If no target sysroot is 17891specified, the option prints nothing. 17892 17893@item -print-sysroot-headers-suffix 17894@opindex print-sysroot-headers-suffix 17895Print the suffix added to the target sysroot when searching for 17896headers, or give an error if the compiler is not configured with such 17897a suffix---and don't do anything else. 17898 17899@item -dumpmachine 17900@opindex dumpmachine 17901Print the compiler's target machine (for example, 17902@samp{i686-pc-linux-gnu})---and don't do anything else. 17903 17904@item -dumpversion 17905@opindex dumpversion 17906Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do 17907anything else. This is the compiler version used in filesystem paths and 17908specs. Depending on how the compiler has been configured it can be just 17909a single number (major version), two numbers separated by a dot (major and 17910minor version) or three numbers separated by dots (major, minor and patchlevel 17911version). 17912 17913@item -dumpfullversion 17914@opindex dumpfullversion 17915Print the full compiler version---and don't do anything else. The output is 17916always three numbers separated by dots, major, minor and patchlevel version. 17917 17918@item -dumpspecs 17919@opindex dumpspecs 17920Print the compiler's built-in specs---and don't do anything else. (This 17921is used when GCC itself is being built.) @xref{Spec Files}. 17922@end table 17923 17924@node Submodel Options 17925@section Machine-Dependent Options 17926@cindex submodel options 17927@cindex specifying hardware config 17928@cindex hardware models and configurations, specifying 17929@cindex target-dependent options 17930@cindex machine-dependent options 17931 17932Each target machine supported by GCC can have its own options---for 17933example, to allow you to compile for a particular processor variant or 17934ABI, or to control optimizations specific to that machine. By 17935convention, the names of machine-specific options start with 17936@samp{-m}. 17937 17938Some configurations of the compiler also support additional target-specific 17939options, usually for compatibility with other compilers on the same 17940platform. 17941 17942@c This list is ordered alphanumerically by subsection name. 17943@c It should be the same order and spelling as these options are listed 17944@c in Machine Dependent Options 17945 17946@menu 17947* AArch64 Options:: 17948* Adapteva Epiphany Options:: 17949* AMD GCN Options:: 17950* ARC Options:: 17951* ARM Options:: 17952* AVR Options:: 17953* Blackfin Options:: 17954* C6X Options:: 17955* CRIS Options:: 17956* CR16 Options:: 17957* C-SKY Options:: 17958* Darwin Options:: 17959* DEC Alpha Options:: 17960* eBPF Options:: 17961* FR30 Options:: 17962* FT32 Options:: 17963* FRV Options:: 17964* GNU/Linux Options:: 17965* H8/300 Options:: 17966* HPPA Options:: 17967* IA-64 Options:: 17968* LM32 Options:: 17969* M32C Options:: 17970* M32R/D Options:: 17971* M680x0 Options:: 17972* MCore Options:: 17973* MeP Options:: 17974* MicroBlaze Options:: 17975* MIPS Options:: 17976* MMIX Options:: 17977* MN10300 Options:: 17978* Moxie Options:: 17979* MSP430 Options:: 17980* NDS32 Options:: 17981* Nios II Options:: 17982* Nvidia PTX Options:: 17983* OpenRISC Options:: 17984* PDP-11 Options:: 17985* picoChip Options:: 17986* PowerPC Options:: 17987* PRU Options:: 17988* RISC-V Options:: 17989* RL78 Options:: 17990* RS/6000 and PowerPC Options:: 17991* RX Options:: 17992* S/390 and zSeries Options:: 17993* Score Options:: 17994* SH Options:: 17995* Solaris 2 Options:: 17996* SPARC Options:: 17997* System V Options:: 17998* TILE-Gx Options:: 17999* TILEPro Options:: 18000* V850 Options:: 18001* VAX Options:: 18002* Visium Options:: 18003* VMS Options:: 18004* VxWorks Options:: 18005* x86 Options:: 18006* x86 Windows Options:: 18007* Xstormy16 Options:: 18008* Xtensa Options:: 18009* zSeries Options:: 18010@end menu 18011 18012@node AArch64 Options 18013@subsection AArch64 Options 18014@cindex AArch64 Options 18015 18016These options are defined for AArch64 implementations: 18017 18018@table @gcctabopt 18019 18020@item -mabi=@var{name} 18021@opindex mabi 18022Generate code for the specified data model. Permissible values 18023are @samp{ilp32} for SysV-like data model where int, long int and pointers 18024are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits, 18025but long int and pointers are 64 bits. 18026 18027The default depends on the specific target configuration. Note that 18028the LP64 and ILP32 ABIs are not link-compatible; you must compile your 18029entire program with the same ABI, and link with a compatible set of libraries. 18030 18031@item -mbig-endian 18032@opindex mbig-endian 18033Generate big-endian code. This is the default when GCC is configured for an 18034@samp{aarch64_be-*-*} target. 18035 18036@item -mgeneral-regs-only 18037@opindex mgeneral-regs-only 18038Generate code which uses only the general-purpose registers. This will prevent 18039the compiler from using floating-point and Advanced SIMD registers but will not 18040impose any restrictions on the assembler. 18041 18042@item -mlittle-endian 18043@opindex mlittle-endian 18044Generate little-endian code. This is the default when GCC is configured for an 18045@samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target. 18046 18047@item -mcmodel=tiny 18048@opindex mcmodel=tiny 18049Generate code for the tiny code model. The program and its statically defined 18050symbols must be within 1MB of each other. Programs can be statically or 18051dynamically linked. 18052 18053@item -mcmodel=small 18054@opindex mcmodel=small 18055Generate code for the small code model. The program and its statically defined 18056symbols must be within 4GB of each other. Programs can be statically or 18057dynamically linked. This is the default code model. 18058 18059@item -mcmodel=large 18060@opindex mcmodel=large 18061Generate code for the large code model. This makes no assumptions about 18062addresses and sizes of sections. Programs can be statically linked only. The 18063@option{-mcmodel=large} option is incompatible with @option{-mabi=ilp32}, 18064@option{-fpic} and @option{-fPIC}. 18065 18066@item -mstrict-align 18067@itemx -mno-strict-align 18068@opindex mstrict-align 18069@opindex mno-strict-align 18070Avoid or allow generating memory accesses that may not be aligned on a natural 18071object boundary as described in the architecture specification. 18072 18073@item -momit-leaf-frame-pointer 18074@itemx -mno-omit-leaf-frame-pointer 18075@opindex momit-leaf-frame-pointer 18076@opindex mno-omit-leaf-frame-pointer 18077Omit or keep the frame pointer in leaf functions. The former behavior is the 18078default. 18079 18080@item -mstack-protector-guard=@var{guard} 18081@itemx -mstack-protector-guard-reg=@var{reg} 18082@itemx -mstack-protector-guard-offset=@var{offset} 18083@opindex mstack-protector-guard 18084@opindex mstack-protector-guard-reg 18085@opindex mstack-protector-guard-offset 18086Generate stack protection code using canary at @var{guard}. Supported 18087locations are @samp{global} for a global canary or @samp{sysreg} for a 18088canary in an appropriate system register. 18089 18090With the latter choice the options 18091@option{-mstack-protector-guard-reg=@var{reg}} and 18092@option{-mstack-protector-guard-offset=@var{offset}} furthermore specify 18093which system register to use as base register for reading the canary, 18094and from what offset from that base register. There is no default 18095register or offset as this is entirely for use within the Linux 18096kernel. 18097 18098@item -mtls-dialect=desc 18099@opindex mtls-dialect=desc 18100Use TLS descriptors as the thread-local storage mechanism for dynamic accesses 18101of TLS variables. This is the default. 18102 18103@item -mtls-dialect=traditional 18104@opindex mtls-dialect=traditional 18105Use traditional TLS as the thread-local storage mechanism for dynamic accesses 18106of TLS variables. 18107 18108@item -mtls-size=@var{size} 18109@opindex mtls-size 18110Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48. 18111This option requires binutils 2.26 or newer. 18112 18113@item -mfix-cortex-a53-835769 18114@itemx -mno-fix-cortex-a53-835769 18115@opindex mfix-cortex-a53-835769 18116@opindex mno-fix-cortex-a53-835769 18117Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769. 18118This involves inserting a NOP instruction between memory instructions and 1811964-bit integer multiply-accumulate instructions. 18120 18121@item -mfix-cortex-a53-843419 18122@itemx -mno-fix-cortex-a53-843419 18123@opindex mfix-cortex-a53-843419 18124@opindex mno-fix-cortex-a53-843419 18125Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419. 18126This erratum workaround is made at link time and this will only pass the 18127corresponding flag to the linker. 18128 18129@item -mlow-precision-recip-sqrt 18130@itemx -mno-low-precision-recip-sqrt 18131@opindex mlow-precision-recip-sqrt 18132@opindex mno-low-precision-recip-sqrt 18133Enable or disable the reciprocal square root approximation. 18134This option only has an effect if @option{-ffast-math} or 18135@option{-funsafe-math-optimizations} is used as well. Enabling this reduces 18136precision of reciprocal square root results to about 16 bits for 18137single precision and to 32 bits for double precision. 18138 18139@item -mlow-precision-sqrt 18140@itemx -mno-low-precision-sqrt 18141@opindex mlow-precision-sqrt 18142@opindex mno-low-precision-sqrt 18143Enable or disable the square root approximation. 18144This option only has an effect if @option{-ffast-math} or 18145@option{-funsafe-math-optimizations} is used as well. Enabling this reduces 18146precision of square root results to about 16 bits for 18147single precision and to 32 bits for double precision. 18148If enabled, it implies @option{-mlow-precision-recip-sqrt}. 18149 18150@item -mlow-precision-div 18151@itemx -mno-low-precision-div 18152@opindex mlow-precision-div 18153@opindex mno-low-precision-div 18154Enable or disable the division approximation. 18155This option only has an effect if @option{-ffast-math} or 18156@option{-funsafe-math-optimizations} is used as well. Enabling this reduces 18157precision of division results to about 16 bits for 18158single precision and to 32 bits for double precision. 18159 18160@item -mtrack-speculation 18161@itemx -mno-track-speculation 18162Enable or disable generation of additional code to track speculative 18163execution through conditional branches. The tracking state can then 18164be used by the compiler when expanding calls to 18165@code{__builtin_speculation_safe_copy} to permit a more efficient code 18166sequence to be generated. 18167 18168@item -moutline-atomics 18169@itemx -mno-outline-atomics 18170Enable or disable calls to out-of-line helpers to implement atomic operations. 18171These helpers will, at runtime, determine if the LSE instructions from 18172ARMv8.1-A can be used; if not, they will use the load/store-exclusive 18173instructions that are present in the base ARMv8.0 ISA. 18174 18175This option is only applicable when compiling for the base ARMv8.0 18176instruction set. If using a later revision, e.g. @option{-march=armv8.1-a} 18177or @option{-march=armv8-a+lse}, the ARMv8.1-Atomics instructions will be 18178used directly. The same applies when using @option{-mcpu=} when the 18179selected cpu supports the @samp{lse} feature. 18180This option is on by default. 18181 18182@item -march=@var{name} 18183@opindex march 18184Specify the name of the target architecture and, optionally, one or 18185more feature modifiers. This option has the form 18186@option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}. 18187 18188The table below summarizes the permissible values for @var{arch} 18189and the features that they enable by default: 18190 18191@multitable @columnfractions 0.20 0.20 0.60 18192@headitem @var{arch} value @tab Architecture @tab Includes by default 18193@item @samp{armv8-a} @tab Armv8-A @tab @samp{+fp}, @samp{+simd} 18194@item @samp{armv8.1-a} @tab Armv8.1-A @tab @samp{armv8-a}, @samp{+crc}, @samp{+lse}, @samp{+rdma} 18195@item @samp{armv8.2-a} @tab Armv8.2-A @tab @samp{armv8.1-a} 18196@item @samp{armv8.3-a} @tab Armv8.3-A @tab @samp{armv8.2-a}, @samp{+pauth} 18197@item @samp{armv8.4-a} @tab Armv8.4-A @tab @samp{armv8.3-a}, @samp{+flagm}, @samp{+fp16fml}, @samp{+dotprod} 18198@item @samp{armv8.5-a} @tab Armv8.5-A @tab @samp{armv8.4-a}, @samp{+sb}, @samp{+ssbs}, @samp{+predres} 18199@item @samp{armv8.6-a} @tab Armv8.6-A @tab @samp{armv8.5-a}, @samp{+bf16}, @samp{+i8mm} 18200@item @samp{armv8-r} @tab Armv8-R @tab @samp{armv8-r} 18201@end multitable 18202 18203The value @samp{native} is available on native AArch64 GNU/Linux and 18204causes the compiler to pick the architecture of the host system. This 18205option has no effect if the compiler is unable to recognize the 18206architecture of the host system, 18207 18208The permissible values for @var{feature} are listed in the sub-section 18209on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu} 18210Feature Modifiers}. Where conflicting feature modifiers are 18211specified, the right-most feature is used. 18212 18213GCC uses @var{name} to determine what kind of instructions it can emit 18214when generating assembly code. If @option{-march} is specified 18215without either of @option{-mtune} or @option{-mcpu} also being 18216specified, the code is tuned to perform well across a range of target 18217processors implementing the target architecture. 18218 18219@item -mtune=@var{name} 18220@opindex mtune 18221Specify the name of the target processor for which GCC should tune the 18222performance of the code. Permissible values for this option are: 18223@samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, 18224@samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, 18225@samp{cortex-a76}, @samp{cortex-a76ae}, @samp{cortex-a77}, 18226@samp{cortex-a65}, @samp{cortex-a65ae}, @samp{cortex-a34}, 18227@samp{cortex-a78}, @samp{cortex-a78ae}, @samp{cortex-a78c}, 18228@samp{ares}, @samp{exynos-m1}, @samp{emag}, @samp{falkor}, 18229@samp{neoverse-512tvb}, @samp{neoverse-e1}, @samp{neoverse-n1}, 18230@samp{neoverse-n2}, @samp{neoverse-v1}, @samp{qdf24xx}, 18231@samp{saphira}, @samp{phecda}, @samp{xgene1}, @samp{vulcan}, 18232@samp{octeontx}, @samp{octeontx81}, @samp{octeontx83}, 18233@samp{octeontx2}, @samp{octeontx2t98}, @samp{octeontx2t96} 18234@samp{octeontx2t93}, @samp{octeontx2f95}, @samp{octeontx2f95n}, 18235@samp{octeontx2f95mm}, 18236@samp{a64fx}, 18237@samp{thunderx}, @samp{thunderxt88}, 18238@samp{thunderxt88p1}, @samp{thunderxt81}, @samp{tsv110}, 18239@samp{thunderxt83}, @samp{thunderx2t99}, @samp{thunderx3t110}, @samp{zeus}, 18240@samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}, 18241@samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53}, 18242@samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55}, 18243@samp{cortex-r82}, @samp{cortex-x1}, @samp{native}. 18244 18245The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}, 18246@samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53}, 18247@samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55} specify that GCC 18248should tune for a big.LITTLE system. 18249 18250The value @samp{neoverse-512tvb} specifies that GCC should tune 18251for Neoverse cores that (a) implement SVE and (b) have a total vector 18252bandwidth of 512 bits per cycle. In other words, the option tells GCC to 18253tune for Neoverse cores that can execute 4 128-bit Advanced SIMD arithmetic 18254instructions a cycle and that can execute an equivalent number of SVE 18255arithmetic instructions per cycle (2 for 256-bit SVE, 4 for 128-bit SVE). 18256This is more general than tuning for a specific core like Neoverse V1 18257but is more specific than the default tuning described below. 18258 18259Additionally on native AArch64 GNU/Linux systems the value 18260@samp{native} tunes performance to the host system. This option has no effect 18261if the compiler is unable to recognize the processor of the host system. 18262 18263Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=} 18264are specified, the code is tuned to perform well across a range 18265of target processors. 18266 18267This option cannot be suffixed by feature modifiers. 18268 18269@item -mcpu=@var{name} 18270@opindex mcpu 18271Specify the name of the target processor, optionally suffixed by one 18272or more feature modifiers. This option has the form 18273@option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where 18274the permissible values for @var{cpu} are the same as those available 18275for @option{-mtune}. The permissible values for @var{feature} are 18276documented in the sub-section on 18277@ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu} 18278Feature Modifiers}. Where conflicting feature modifiers are 18279specified, the right-most feature is used. 18280 18281GCC uses @var{name} to determine what kind of instructions it can emit when 18282generating assembly code (as if by @option{-march}) and to determine 18283the target processor for which to tune for performance (as if 18284by @option{-mtune}). Where this option is used in conjunction 18285with @option{-march} or @option{-mtune}, those options take precedence 18286over the appropriate part of this option. 18287 18288@option{-mcpu=neoverse-512tvb} is special in that it does not refer 18289to a specific core, but instead refers to all Neoverse cores that 18290(a) implement SVE and (b) have a total vector bandwidth of 512 bits 18291a cycle. Unless overridden by @option{-march}, 18292@option{-mcpu=neoverse-512tvb} generates code that can run on a 18293Neoverse V1 core, since Neoverse V1 is the first Neoverse core with 18294these properties. Unless overridden by @option{-mtune}, 18295@option{-mcpu=neoverse-512tvb} tunes code in the same way as for 18296@option{-mtune=neoverse-512tvb}. 18297 18298@item -moverride=@var{string} 18299@opindex moverride 18300Override tuning decisions made by the back-end in response to a 18301@option{-mtune=} switch. The syntax, semantics, and accepted values 18302for @var{string} in this option are not guaranteed to be consistent 18303across releases. 18304 18305This option is only intended to be useful when developing GCC. 18306 18307@item -mverbose-cost-dump 18308@opindex mverbose-cost-dump 18309Enable verbose cost model dumping in the debug dump files. This option is 18310provided for use in debugging the compiler. 18311 18312@item -mpc-relative-literal-loads 18313@itemx -mno-pc-relative-literal-loads 18314@opindex mpc-relative-literal-loads 18315@opindex mno-pc-relative-literal-loads 18316Enable or disable PC-relative literal loads. With this option literal pools are 18317accessed using a single instruction and emitted after each function. This 18318limits the maximum size of functions to 1MB. This is enabled by default for 18319@option{-mcmodel=tiny}. 18320 18321@item -msign-return-address=@var{scope} 18322@opindex msign-return-address 18323Select the function scope on which return address signing will be applied. 18324Permissible values are @samp{none}, which disables return address signing, 18325@samp{non-leaf}, which enables pointer signing for functions which are not leaf 18326functions, and @samp{all}, which enables pointer signing for all functions. The 18327default value is @samp{none}. This option has been deprecated by 18328-mbranch-protection. 18329 18330@item -mbranch-protection=@var{none}|@var{standard}|@var{pac-ret}[+@var{leaf}+@var{b-key}]|@var{bti} 18331@opindex mbranch-protection 18332Select the branch protection features to use. 18333@samp{none} is the default and turns off all types of branch protection. 18334@samp{standard} turns on all types of branch protection features. If a feature 18335has additional tuning options, then @samp{standard} sets it to its standard 18336level. 18337@samp{pac-ret[+@var{leaf}]} turns on return address signing to its standard 18338level: signing functions that save the return address to memory (non-leaf 18339functions will practically always do this) using the a-key. The optional 18340argument @samp{leaf} can be used to extend the signing to include leaf 18341functions. The optional argument @samp{b-key} can be used to sign the functions 18342with the B-key instead of the A-key. 18343@samp{bti} turns on branch target identification mechanism. 18344 18345@item -mharden-sls=@var{opts} 18346@opindex mharden-sls 18347Enable compiler hardening against straight line speculation (SLS). 18348@var{opts} is a comma-separated list of the following options: 18349@table @samp 18350@item retbr 18351@item blr 18352@end table 18353In addition, @samp{-mharden-sls=all} enables all SLS hardening while 18354@samp{-mharden-sls=none} disables all SLS hardening. 18355 18356@item -msve-vector-bits=@var{bits} 18357@opindex msve-vector-bits 18358Specify the number of bits in an SVE vector register. This option only has 18359an effect when SVE is enabled. 18360 18361GCC supports two forms of SVE code generation: ``vector-length 18362agnostic'' output that works with any size of vector register and 18363``vector-length specific'' output that allows GCC to make assumptions 18364about the vector length when it is useful for optimization reasons. 18365The possible values of @samp{bits} are: @samp{scalable}, @samp{128}, 18366@samp{256}, @samp{512}, @samp{1024} and @samp{2048}. 18367Specifying @samp{scalable} selects vector-length agnostic 18368output. At present @samp{-msve-vector-bits=128} also generates vector-length 18369agnostic output for big-endian targets. All other values generate 18370vector-length specific code. The behavior of these values may change 18371in future releases and no value except @samp{scalable} should be 18372relied on for producing code that is portable across different 18373hardware SVE vector lengths. 18374 18375The default is @samp{-msve-vector-bits=scalable}, which produces 18376vector-length agnostic code. 18377@end table 18378 18379@subsubsection @option{-march} and @option{-mcpu} Feature Modifiers 18380@anchor{aarch64-feature-modifiers} 18381@cindex @option{-march} feature modifiers 18382@cindex @option{-mcpu} feature modifiers 18383Feature modifiers used with @option{-march} and @option{-mcpu} can be any of 18384the following and their inverses @option{no@var{feature}}: 18385 18386@table @samp 18387@item crc 18388Enable CRC extension. This is on by default for 18389@option{-march=armv8.1-a}. 18390@item crypto 18391Enable Crypto extension. This also enables Advanced SIMD and floating-point 18392instructions. 18393@item fp 18394Enable floating-point instructions. This is on by default for all possible 18395values for options @option{-march} and @option{-mcpu}. 18396@item simd 18397Enable Advanced SIMD instructions. This also enables floating-point 18398instructions. This is on by default for all possible values for options 18399@option{-march} and @option{-mcpu}. 18400@item sve 18401Enable Scalable Vector Extension instructions. This also enables Advanced 18402SIMD and floating-point instructions. 18403@item lse 18404Enable Large System Extension instructions. This is on by default for 18405@option{-march=armv8.1-a}. 18406@item rdma 18407Enable Round Double Multiply Accumulate instructions. This is on by default 18408for @option{-march=armv8.1-a}. 18409@item fp16 18410Enable FP16 extension. This also enables floating-point instructions. 18411@item fp16fml 18412Enable FP16 fmla extension. This also enables FP16 extensions and 18413floating-point instructions. This option is enabled by default for @option{-march=armv8.4-a}. Use of this option with architectures prior to Armv8.2-A is not supported. 18414 18415@item rcpc 18416Enable the RcPc extension. This does not change code generation from GCC, 18417but is passed on to the assembler, enabling inline asm statements to use 18418instructions from the RcPc extension. 18419@item dotprod 18420Enable the Dot Product extension. This also enables Advanced SIMD instructions. 18421@item aes 18422Enable the Armv8-a aes and pmull crypto extension. This also enables Advanced 18423SIMD instructions. 18424@item sha2 18425Enable the Armv8-a sha2 crypto extension. This also enables Advanced SIMD instructions. 18426@item sha3 18427Enable the sha512 and sha3 crypto extension. This also enables Advanced SIMD 18428instructions. Use of this option with architectures prior to Armv8.2-A is not supported. 18429@item sm4 18430Enable the sm3 and sm4 crypto extension. This also enables Advanced SIMD instructions. 18431Use of this option with architectures prior to Armv8.2-A is not supported. 18432@item profile 18433Enable the Statistical Profiling extension. This option is only to enable the 18434extension at the assembler level and does not affect code generation. 18435@item rng 18436Enable the Armv8.5-a Random Number instructions. This option is only to 18437enable the extension at the assembler level and does not affect code 18438generation. 18439@item memtag 18440Enable the Armv8.5-a Memory Tagging Extensions. 18441Use of this option with architectures prior to Armv8.5-A is not supported. 18442@item sb 18443Enable the Armv8-a Speculation Barrier instruction. This option is only to 18444enable the extension at the assembler level and does not affect code 18445generation. This option is enabled by default for @option{-march=armv8.5-a}. 18446@item ssbs 18447Enable the Armv8-a Speculative Store Bypass Safe instruction. This option 18448is only to enable the extension at the assembler level and does not affect code 18449generation. This option is enabled by default for @option{-march=armv8.5-a}. 18450@item predres 18451Enable the Armv8-a Execution and Data Prediction Restriction instructions. 18452This option is only to enable the extension at the assembler level and does 18453not affect code generation. This option is enabled by default for 18454@option{-march=armv8.5-a}. 18455@item sve2 18456Enable the Armv8-a Scalable Vector Extension 2. This also enables SVE 18457instructions. 18458@item sve2-bitperm 18459Enable SVE2 bitperm instructions. This also enables SVE2 instructions. 18460@item sve2-sm4 18461Enable SVE2 sm4 instructions. This also enables SVE2 instructions. 18462@item sve2-aes 18463Enable SVE2 aes instructions. This also enables SVE2 instructions. 18464@item sve2-sha3 18465Enable SVE2 sha3 instructions. This also enables SVE2 instructions. 18466@item tme 18467Enable the Transactional Memory Extension. 18468@item i8mm 18469Enable 8-bit Integer Matrix Multiply instructions. This also enables 18470Advanced SIMD and floating-point instructions. This option is enabled by 18471default for @option{-march=armv8.6-a}. Use of this option with architectures 18472prior to Armv8.2-A is not supported. 18473@item f32mm 18474Enable 32-bit Floating point Matrix Multiply instructions. This also enables 18475SVE instructions. Use of this option with architectures prior to Armv8.2-A is 18476not supported. 18477@item f64mm 18478Enable 64-bit Floating point Matrix Multiply instructions. This also enables 18479SVE instructions. Use of this option with architectures prior to Armv8.2-A is 18480not supported. 18481@item bf16 18482Enable brain half-precision floating-point instructions. This also enables 18483Advanced SIMD and floating-point instructions. This option is enabled by 18484default for @option{-march=armv8.6-a}. Use of this option with architectures 18485prior to Armv8.2-A is not supported. 18486@item flagm 18487Enable the Flag Manipulation instructions Extension. 18488@item pauth 18489Enable the Pointer Authentication Extension. 18490 18491@end table 18492 18493Feature @option{crypto} implies @option{aes}, @option{sha2}, and @option{simd}, 18494which implies @option{fp}. 18495Conversely, @option{nofp} implies @option{nosimd}, which implies 18496@option{nocrypto}, @option{noaes} and @option{nosha2}. 18497 18498@node Adapteva Epiphany Options 18499@subsection Adapteva Epiphany Options 18500 18501These @samp{-m} options are defined for Adapteva Epiphany: 18502 18503@table @gcctabopt 18504@item -mhalf-reg-file 18505@opindex mhalf-reg-file 18506Don't allocate any register in the range @code{r32}@dots{}@code{r63}. 18507That allows code to run on hardware variants that lack these registers. 18508 18509@item -mprefer-short-insn-regs 18510@opindex mprefer-short-insn-regs 18511Preferentially allocate registers that allow short instruction generation. 18512This can result in increased instruction count, so this may either reduce or 18513increase overall code size. 18514 18515@item -mbranch-cost=@var{num} 18516@opindex mbranch-cost 18517Set the cost of branches to roughly @var{num} ``simple'' instructions. 18518This cost is only a heuristic and is not guaranteed to produce 18519consistent results across releases. 18520 18521@item -mcmove 18522@opindex mcmove 18523Enable the generation of conditional moves. 18524 18525@item -mnops=@var{num} 18526@opindex mnops 18527Emit @var{num} NOPs before every other generated instruction. 18528 18529@item -mno-soft-cmpsf 18530@opindex mno-soft-cmpsf 18531@opindex msoft-cmpsf 18532For single-precision floating-point comparisons, emit an @code{fsub} instruction 18533and test the flags. This is faster than a software comparison, but can 18534get incorrect results in the presence of NaNs, or when two different small 18535numbers are compared such that their difference is calculated as zero. 18536The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant, 18537software comparisons. 18538 18539@item -mstack-offset=@var{num} 18540@opindex mstack-offset 18541Set the offset between the top of the stack and the stack pointer. 18542E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7} 18543can be used by leaf functions without stack allocation. 18544Values other than @samp{8} or @samp{16} are untested and unlikely to work. 18545Note also that this option changes the ABI; compiling a program with a 18546different stack offset than the libraries have been compiled with 18547generally does not work. 18548This option can be useful if you want to evaluate if a different stack 18549offset would give you better code, but to actually use a different stack 18550offset to build working programs, it is recommended to configure the 18551toolchain with the appropriate @option{--with-stack-offset=@var{num}} option. 18552 18553@item -mno-round-nearest 18554@opindex mno-round-nearest 18555@opindex mround-nearest 18556Make the scheduler assume that the rounding mode has been set to 18557truncating. The default is @option{-mround-nearest}. 18558 18559@item -mlong-calls 18560@opindex mlong-calls 18561If not otherwise specified by an attribute, assume all calls might be beyond 18562the offset range of the @code{b} / @code{bl} instructions, and therefore load the 18563function address into a register before performing a (otherwise direct) call. 18564This is the default. 18565 18566@item -mshort-calls 18567@opindex short-calls 18568If not otherwise specified by an attribute, assume all direct calls are 18569in the range of the @code{b} / @code{bl} instructions, so use these instructions 18570for direct calls. The default is @option{-mlong-calls}. 18571 18572@item -msmall16 18573@opindex msmall16 18574Assume addresses can be loaded as 16-bit unsigned values. This does not 18575apply to function addresses for which @option{-mlong-calls} semantics 18576are in effect. 18577 18578@item -mfp-mode=@var{mode} 18579@opindex mfp-mode 18580Set the prevailing mode of the floating-point unit. 18581This determines the floating-point mode that is provided and expected 18582at function call and return time. Making this mode match the mode you 18583predominantly need at function start can make your programs smaller and 18584faster by avoiding unnecessary mode switches. 18585 18586@var{mode} can be set to one the following values: 18587 18588@table @samp 18589@item caller 18590Any mode at function entry is valid, and retained or restored when 18591the function returns, and when it calls other functions. 18592This mode is useful for compiling libraries or other compilation units 18593you might want to incorporate into different programs with different 18594prevailing FPU modes, and the convenience of being able to use a single 18595object file outweighs the size and speed overhead for any extra 18596mode switching that might be needed, compared with what would be needed 18597with a more specific choice of prevailing FPU mode. 18598 18599@item truncate 18600This is the mode used for floating-point calculations with 18601truncating (i.e.@: round towards zero) rounding mode. That includes 18602conversion from floating point to integer. 18603 18604@item round-nearest 18605This is the mode used for floating-point calculations with 18606round-to-nearest-or-even rounding mode. 18607 18608@item int 18609This is the mode used to perform integer calculations in the FPU, e.g.@: 18610integer multiply, or integer multiply-and-accumulate. 18611@end table 18612 18613The default is @option{-mfp-mode=caller} 18614 18615@item -mno-split-lohi 18616@itemx -mno-postinc 18617@itemx -mno-postmodify 18618@opindex mno-split-lohi 18619@opindex msplit-lohi 18620@opindex mno-postinc 18621@opindex mpostinc 18622@opindex mno-postmodify 18623@opindex mpostmodify 18624Code generation tweaks that disable, respectively, splitting of 32-bit 18625loads, generation of post-increment addresses, and generation of 18626post-modify addresses. The defaults are @option{msplit-lohi}, 18627@option{-mpost-inc}, and @option{-mpost-modify}. 18628 18629@item -mnovect-double 18630@opindex mno-vect-double 18631@opindex mvect-double 18632Change the preferred SIMD mode to SImode. The default is 18633@option{-mvect-double}, which uses DImode as preferred SIMD mode. 18634 18635@item -max-vect-align=@var{num} 18636@opindex max-vect-align 18637The maximum alignment for SIMD vector mode types. 18638@var{num} may be 4 or 8. The default is 8. 18639Note that this is an ABI change, even though many library function 18640interfaces are unaffected if they don't use SIMD vector modes 18641in places that affect size and/or alignment of relevant types. 18642 18643@item -msplit-vecmove-early 18644@opindex msplit-vecmove-early 18645Split vector moves into single word moves before reload. In theory this 18646can give better register allocation, but so far the reverse seems to be 18647generally the case. 18648 18649@item -m1reg-@var{reg} 18650@opindex m1reg- 18651Specify a register to hold the constant @minus{}1, which makes loading small negative 18652constants and certain bitmasks faster. 18653Allowable values for @var{reg} are @samp{r43} and @samp{r63}, 18654which specify use of that register as a fixed register, 18655and @samp{none}, which means that no register is used for this 18656purpose. The default is @option{-m1reg-none}. 18657 18658@end table 18659 18660@node AMD GCN Options 18661@subsection AMD GCN Options 18662@cindex AMD GCN Options 18663 18664These options are defined specifically for the AMD GCN port. 18665 18666@table @gcctabopt 18667 18668@item -march=@var{gpu} 18669@opindex march 18670@itemx -mtune=@var{gpu} 18671@opindex mtune 18672Set architecture type or tuning for @var{gpu}. Supported values for @var{gpu} 18673are 18674 18675@table @samp 18676@opindex fiji 18677@item fiji 18678Compile for GCN3 Fiji devices (gfx803). 18679 18680@item gfx900 18681Compile for GCN5 Vega 10 devices (gfx900). 18682 18683@item gfx906 18684Compile for GCN5 Vega 20 devices (gfx906). 18685 18686@end table 18687 18688@item -mstack-size=@var{bytes} 18689@opindex mstack-size 18690Specify how many @var{bytes} of stack space will be requested for each GPU 18691thread (wave-front). Beware that there may be many threads and limited memory 18692available. The size of the stack allocation may also have an impact on 18693run-time performance. The default is 32KB when using OpenACC or OpenMP, and 186941MB otherwise. 18695 18696@end table 18697 18698@node ARC Options 18699@subsection ARC Options 18700@cindex ARC options 18701 18702The following options control the architecture variant for which code 18703is being compiled: 18704 18705@c architecture variants 18706@table @gcctabopt 18707 18708@item -mbarrel-shifter 18709@opindex mbarrel-shifter 18710Generate instructions supported by barrel shifter. This is the default 18711unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect. 18712 18713@item -mjli-always 18714@opindex mjli-alawys 18715Force to call a function using jli_s instruction. This option is 18716valid only for ARCv2 architecture. 18717 18718@item -mcpu=@var{cpu} 18719@opindex mcpu 18720Set architecture type, register usage, and instruction scheduling 18721parameters for @var{cpu}. There are also shortcut alias options 18722available for backward compatibility and convenience. Supported 18723values for @var{cpu} are 18724 18725@table @samp 18726@opindex mA6 18727@opindex mARC600 18728@item arc600 18729Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}. 18730 18731@item arc601 18732@opindex mARC601 18733Compile for ARC601. Alias: @option{-mARC601}. 18734 18735@item arc700 18736@opindex mA7 18737@opindex mARC700 18738Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}. 18739This is the default when configured with @option{--with-cpu=arc700}@. 18740 18741@item arcem 18742Compile for ARC EM. 18743 18744@item archs 18745Compile for ARC HS. 18746 18747@item em 18748Compile for ARC EM CPU with no hardware extensions. 18749 18750@item em4 18751Compile for ARC EM4 CPU. 18752 18753@item em4_dmips 18754Compile for ARC EM4 DMIPS CPU. 18755 18756@item em4_fpus 18757Compile for ARC EM4 DMIPS CPU with the single-precision floating-point 18758extension. 18759 18760@item em4_fpuda 18761Compile for ARC EM4 DMIPS CPU with single-precision floating-point and 18762double assist instructions. 18763 18764@item hs 18765Compile for ARC HS CPU with no hardware extensions except the atomic 18766instructions. 18767 18768@item hs34 18769Compile for ARC HS34 CPU. 18770 18771@item hs38 18772Compile for ARC HS38 CPU. 18773 18774@item hs38_linux 18775Compile for ARC HS38 CPU with all hardware extensions on. 18776 18777@item arc600_norm 18778Compile for ARC 600 CPU with @code{norm} instructions enabled. 18779 18780@item arc600_mul32x16 18781Compile for ARC 600 CPU with @code{norm} and 32x16-bit multiply 18782instructions enabled. 18783 18784@item arc600_mul64 18785Compile for ARC 600 CPU with @code{norm} and @code{mul64}-family 18786instructions enabled. 18787 18788@item arc601_norm 18789Compile for ARC 601 CPU with @code{norm} instructions enabled. 18790 18791@item arc601_mul32x16 18792Compile for ARC 601 CPU with @code{norm} and 32x16-bit multiply 18793instructions enabled. 18794 18795@item arc601_mul64 18796Compile for ARC 601 CPU with @code{norm} and @code{mul64}-family 18797instructions enabled. 18798 18799@item nps400 18800Compile for ARC 700 on NPS400 chip. 18801 18802@item em_mini 18803Compile for ARC EM minimalist configuration featuring reduced register 18804set. 18805 18806@end table 18807 18808@item -mdpfp 18809@opindex mdpfp 18810@itemx -mdpfp-compact 18811@opindex mdpfp-compact 18812Generate double-precision FPX instructions, tuned for the compact 18813implementation. 18814 18815@item -mdpfp-fast 18816@opindex mdpfp-fast 18817Generate double-precision FPX instructions, tuned for the fast 18818implementation. 18819 18820@item -mno-dpfp-lrsr 18821@opindex mno-dpfp-lrsr 18822Disable @code{lr} and @code{sr} instructions from using FPX extension 18823aux registers. 18824 18825@item -mea 18826@opindex mea 18827Generate extended arithmetic instructions. Currently only 18828@code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are 18829supported. Only valid for @option{-mcpu=ARC700}. 18830 18831@item -mno-mpy 18832@opindex mno-mpy 18833@opindex mmpy 18834Do not generate @code{mpy}-family instructions for ARC700. This option is 18835deprecated. 18836 18837@item -mmul32x16 18838@opindex mmul32x16 18839Generate 32x16-bit multiply and multiply-accumulate instructions. 18840 18841@item -mmul64 18842@opindex mmul64 18843Generate @code{mul64} and @code{mulu64} instructions. 18844Only valid for @option{-mcpu=ARC600}. 18845 18846@item -mnorm 18847@opindex mnorm 18848Generate @code{norm} instructions. This is the default if @option{-mcpu=ARC700} 18849is in effect. 18850 18851@item -mspfp 18852@opindex mspfp 18853@itemx -mspfp-compact 18854@opindex mspfp-compact 18855Generate single-precision FPX instructions, tuned for the compact 18856implementation. 18857 18858@item -mspfp-fast 18859@opindex mspfp-fast 18860Generate single-precision FPX instructions, tuned for the fast 18861implementation. 18862 18863@item -msimd 18864@opindex msimd 18865Enable generation of ARC SIMD instructions via target-specific 18866builtins. Only valid for @option{-mcpu=ARC700}. 18867 18868@item -msoft-float 18869@opindex msoft-float 18870This option ignored; it is provided for compatibility purposes only. 18871Software floating-point code is emitted by default, and this default 18872can overridden by FPX options; @option{-mspfp}, @option{-mspfp-compact}, or 18873@option{-mspfp-fast} for single precision, and @option{-mdpfp}, 18874@option{-mdpfp-compact}, or @option{-mdpfp-fast} for double precision. 18875 18876@item -mswap 18877@opindex mswap 18878Generate @code{swap} instructions. 18879 18880@item -matomic 18881@opindex matomic 18882This enables use of the locked load/store conditional extension to implement 18883atomic memory built-in functions. Not available for ARC 6xx or ARC 18884EM cores. 18885 18886@item -mdiv-rem 18887@opindex mdiv-rem 18888Enable @code{div} and @code{rem} instructions for ARCv2 cores. 18889 18890@item -mcode-density 18891@opindex mcode-density 18892Enable code density instructions for ARC EM. 18893This option is on by default for ARC HS. 18894 18895@item -mll64 18896@opindex mll64 18897Enable double load/store operations for ARC HS cores. 18898 18899@item -mtp-regno=@var{regno} 18900@opindex mtp-regno 18901Specify thread pointer register number. 18902 18903@item -mmpy-option=@var{multo} 18904@opindex mmpy-option 18905Compile ARCv2 code with a multiplier design option. You can specify 18906the option using either a string or numeric value for @var{multo}. 18907@samp{wlh1} is the default value. The recognized values are: 18908 18909@table @samp 18910@item 0 18911@itemx none 18912No multiplier available. 18913 18914@item 1 18915@itemx w 1891616x16 multiplier, fully pipelined. 18917The following instructions are enabled: @code{mpyw} and @code{mpyuw}. 18918 18919@item 2 18920@itemx wlh1 1892132x32 multiplier, fully 18922pipelined (1 stage). The following instructions are additionally 18923enabled: @code{mpy}, @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}. 18924 18925@item 3 18926@itemx wlh2 1892732x32 multiplier, fully pipelined 18928(2 stages). The following instructions are additionally enabled: @code{mpy}, 18929@code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}. 18930 18931@item 4 18932@itemx wlh3 18933Two 16x16 multipliers, blocking, 18934sequential. The following instructions are additionally enabled: @code{mpy}, 18935@code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}. 18936 18937@item 5 18938@itemx wlh4 18939One 16x16 multiplier, blocking, 18940sequential. The following instructions are additionally enabled: @code{mpy}, 18941@code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}. 18942 18943@item 6 18944@itemx wlh5 18945One 32x4 multiplier, blocking, 18946sequential. The following instructions are additionally enabled: @code{mpy}, 18947@code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}. 18948 18949@item 7 18950@itemx plus_dmpy 18951ARC HS SIMD support. 18952 18953@item 8 18954@itemx plus_macd 18955ARC HS SIMD support. 18956 18957@item 9 18958@itemx plus_qmacw 18959ARC HS SIMD support. 18960 18961@end table 18962 18963This option is only available for ARCv2 cores@. 18964 18965@item -mfpu=@var{fpu} 18966@opindex mfpu 18967Enables support for specific floating-point hardware extensions for ARCv2 18968cores. Supported values for @var{fpu} are: 18969 18970@table @samp 18971 18972@item fpus 18973Enables support for single-precision floating-point hardware 18974extensions@. 18975 18976@item fpud 18977Enables support for double-precision floating-point hardware 18978extensions. The single-precision floating-point extension is also 18979enabled. Not available for ARC EM@. 18980 18981@item fpuda 18982Enables support for double-precision floating-point hardware 18983extensions using double-precision assist instructions. The single-precision 18984floating-point extension is also enabled. This option is 18985only available for ARC EM@. 18986 18987@item fpuda_div 18988Enables support for double-precision floating-point hardware 18989extensions using double-precision assist instructions. 18990The single-precision floating-point, square-root, and divide 18991extensions are also enabled. This option is 18992only available for ARC EM@. 18993 18994@item fpuda_fma 18995Enables support for double-precision floating-point hardware 18996extensions using double-precision assist instructions. 18997The single-precision floating-point and fused multiply and add 18998hardware extensions are also enabled. This option is 18999only available for ARC EM@. 19000 19001@item fpuda_all 19002Enables support for double-precision floating-point hardware 19003extensions using double-precision assist instructions. 19004All single-precision floating-point hardware extensions are also 19005enabled. This option is only available for ARC EM@. 19006 19007@item fpus_div 19008Enables support for single-precision floating-point, square-root and divide 19009hardware extensions@. 19010 19011@item fpud_div 19012Enables support for double-precision floating-point, square-root and divide 19013hardware extensions. This option 19014includes option @samp{fpus_div}. Not available for ARC EM@. 19015 19016@item fpus_fma 19017Enables support for single-precision floating-point and 19018fused multiply and add hardware extensions@. 19019 19020@item fpud_fma 19021Enables support for double-precision floating-point and 19022fused multiply and add hardware extensions. This option 19023includes option @samp{fpus_fma}. Not available for ARC EM@. 19024 19025@item fpus_all 19026Enables support for all single-precision floating-point hardware 19027extensions@. 19028 19029@item fpud_all 19030Enables support for all single- and double-precision floating-point 19031hardware extensions. Not available for ARC EM@. 19032 19033@end table 19034 19035@item -mirq-ctrl-saved=@var{register-range}, @var{blink}, @var{lp_count} 19036@opindex mirq-ctrl-saved 19037Specifies general-purposes registers that the processor automatically 19038saves/restores on interrupt entry and exit. @var{register-range} is 19039specified as two registers separated by a dash. The register range 19040always starts with @code{r0}, the upper limit is @code{fp} register. 19041@var{blink} and @var{lp_count} are optional. This option is only 19042valid for ARC EM and ARC HS cores. 19043 19044@item -mrgf-banked-regs=@var{number} 19045@opindex mrgf-banked-regs 19046Specifies the number of registers replicated in second register bank 19047on entry to fast interrupt. Fast interrupts are interrupts with the 19048highest priority level P0. These interrupts save only PC and STATUS32 19049registers to avoid memory transactions during interrupt entry and exit 19050sequences. Use this option when you are using fast interrupts in an 19051ARC V2 family processor. Permitted values are 4, 8, 16, and 32. 19052 19053@item -mlpc-width=@var{width} 19054@opindex mlpc-width 19055Specify the width of the @code{lp_count} register. Valid values for 19056@var{width} are 8, 16, 20, 24, 28 and 32 bits. The default width is 19057fixed to 32 bits. If the width is less than 32, the compiler does not 19058attempt to transform loops in your program to use the zero-delay loop 19059mechanism unless it is known that the @code{lp_count} register can 19060hold the required loop-counter value. Depending on the width 19061specified, the compiler and run-time library might continue to use the 19062loop mechanism for various needs. This option defines macro 19063@code{__ARC_LPC_WIDTH__} with the value of @var{width}. 19064 19065@item -mrf16 19066@opindex mrf16 19067This option instructs the compiler to generate code for a 16-entry 19068register file. This option defines the @code{__ARC_RF16__} 19069preprocessor macro. 19070 19071@item -mbranch-index 19072@opindex mbranch-index 19073Enable use of @code{bi} or @code{bih} instructions to implement jump 19074tables. 19075 19076@end table 19077 19078The following options are passed through to the assembler, and also 19079define preprocessor macro symbols. 19080 19081@c Flags used by the assembler, but for which we define preprocessor 19082@c macro symbols as well. 19083@table @gcctabopt 19084@item -mdsp-packa 19085@opindex mdsp-packa 19086Passed down to the assembler to enable the DSP Pack A extensions. 19087Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is 19088deprecated. 19089 19090@item -mdvbf 19091@opindex mdvbf 19092Passed down to the assembler to enable the dual Viterbi butterfly 19093extension. Also sets the preprocessor symbol @code{__Xdvbf}. This 19094option is deprecated. 19095 19096@c ARC700 4.10 extension instruction 19097@item -mlock 19098@opindex mlock 19099Passed down to the assembler to enable the locked load/store 19100conditional extension. Also sets the preprocessor symbol 19101@code{__Xlock}. 19102 19103@item -mmac-d16 19104@opindex mmac-d16 19105Passed down to the assembler. Also sets the preprocessor symbol 19106@code{__Xxmac_d16}. This option is deprecated. 19107 19108@item -mmac-24 19109@opindex mmac-24 19110Passed down to the assembler. Also sets the preprocessor symbol 19111@code{__Xxmac_24}. This option is deprecated. 19112 19113@c ARC700 4.10 extension instruction 19114@item -mrtsc 19115@opindex mrtsc 19116Passed down to the assembler to enable the 64-bit time-stamp counter 19117extension instruction. Also sets the preprocessor symbol 19118@code{__Xrtsc}. This option is deprecated. 19119 19120@c ARC700 4.10 extension instruction 19121@item -mswape 19122@opindex mswape 19123Passed down to the assembler to enable the swap byte ordering 19124extension instruction. Also sets the preprocessor symbol 19125@code{__Xswape}. 19126 19127@item -mtelephony 19128@opindex mtelephony 19129Passed down to the assembler to enable dual- and single-operand 19130instructions for telephony. Also sets the preprocessor symbol 19131@code{__Xtelephony}. This option is deprecated. 19132 19133@item -mxy 19134@opindex mxy 19135Passed down to the assembler to enable the XY memory extension. Also 19136sets the preprocessor symbol @code{__Xxy}. 19137 19138@end table 19139 19140The following options control how the assembly code is annotated: 19141 19142@c Assembly annotation options 19143@table @gcctabopt 19144@item -misize 19145@opindex misize 19146Annotate assembler instructions with estimated addresses. 19147 19148@item -mannotate-align 19149@opindex mannotate-align 19150Explain what alignment considerations lead to the decision to make an 19151instruction short or long. 19152 19153@end table 19154 19155The following options are passed through to the linker: 19156 19157@c options passed through to the linker 19158@table @gcctabopt 19159@item -marclinux 19160@opindex marclinux 19161Passed through to the linker, to specify use of the @code{arclinux} emulation. 19162This option is enabled by default in tool chains built for 19163@w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets 19164when profiling is not requested. 19165 19166@item -marclinux_prof 19167@opindex marclinux_prof 19168Passed through to the linker, to specify use of the 19169@code{arclinux_prof} emulation. This option is enabled by default in 19170tool chains built for @w{@code{arc-linux-uclibc}} and 19171@w{@code{arceb-linux-uclibc}} targets when profiling is requested. 19172 19173@end table 19174 19175The following options control the semantics of generated code: 19176 19177@c semantically relevant code generation options 19178@table @gcctabopt 19179@item -mlong-calls 19180@opindex mlong-calls 19181Generate calls as register indirect calls, thus providing access 19182to the full 32-bit address range. 19183 19184@item -mmedium-calls 19185@opindex mmedium-calls 19186Don't use less than 25-bit addressing range for calls, which is the 19187offset available for an unconditional branch-and-link 19188instruction. Conditional execution of function calls is suppressed, to 19189allow use of the 25-bit range, rather than the 21-bit range with 19190conditional branch-and-link. This is the default for tool chains built 19191for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets. 19192 19193@item -G @var{num} 19194@opindex G 19195Put definitions of externally-visible data in a small data section if 19196that data is no bigger than @var{num} bytes. The default value of 19197@var{num} is 4 for any ARC configuration, or 8 when we have double 19198load/store operations. 19199 19200@item -mno-sdata 19201@opindex mno-sdata 19202@opindex msdata 19203Do not generate sdata references. This is the default for tool chains 19204built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} 19205targets. 19206 19207@item -mvolatile-cache 19208@opindex mvolatile-cache 19209Use ordinarily cached memory accesses for volatile references. This is the 19210default. 19211 19212@item -mno-volatile-cache 19213@opindex mno-volatile-cache 19214@opindex mvolatile-cache 19215Enable cache bypass for volatile references. 19216 19217@end table 19218 19219The following options fine tune code generation: 19220@c code generation tuning options 19221@table @gcctabopt 19222@item -malign-call 19223@opindex malign-call 19224Do alignment optimizations for call instructions. 19225 19226@item -mauto-modify-reg 19227@opindex mauto-modify-reg 19228Enable the use of pre/post modify with register displacement. 19229 19230@item -mbbit-peephole 19231@opindex mbbit-peephole 19232Enable bbit peephole2. 19233 19234@item -mno-brcc 19235@opindex mno-brcc 19236This option disables a target-specific pass in @file{arc_reorg} to 19237generate compare-and-branch (@code{br@var{cc}}) instructions. 19238It has no effect on 19239generation of these instructions driven by the combiner pass. 19240 19241@item -mcase-vector-pcrel 19242@opindex mcase-vector-pcrel 19243Use PC-relative switch case tables to enable case table shortening. 19244This is the default for @option{-Os}. 19245 19246@item -mcompact-casesi 19247@opindex mcompact-casesi 19248Enable compact @code{casesi} pattern. This is the default for @option{-Os}, 19249and only available for ARCv1 cores. This option is deprecated. 19250 19251@item -mno-cond-exec 19252@opindex mno-cond-exec 19253Disable the ARCompact-specific pass to generate conditional 19254execution instructions. 19255 19256Due to delay slot scheduling and interactions between operand numbers, 19257literal sizes, instruction lengths, and the support for conditional execution, 19258the target-independent pass to generate conditional execution is often lacking, 19259so the ARC port has kept a special pass around that tries to find more 19260conditional execution generation opportunities after register allocation, 19261branch shortening, and delay slot scheduling have been done. This pass 19262generally, but not always, improves performance and code size, at the cost of 19263extra compilation time, which is why there is an option to switch it off. 19264If you have a problem with call instructions exceeding their allowable 19265offset range because they are conditionalized, you should consider using 19266@option{-mmedium-calls} instead. 19267 19268@item -mearly-cbranchsi 19269@opindex mearly-cbranchsi 19270Enable pre-reload use of the @code{cbranchsi} pattern. 19271 19272@item -mexpand-adddi 19273@opindex mexpand-adddi 19274Expand @code{adddi3} and @code{subdi3} at RTL generation time into 19275@code{add.f}, @code{adc} etc. This option is deprecated. 19276 19277@item -mindexed-loads 19278@opindex mindexed-loads 19279Enable the use of indexed loads. This can be problematic because some 19280optimizers then assume that indexed stores exist, which is not 19281the case. 19282 19283@item -mlra 19284@opindex mlra 19285Enable Local Register Allocation. This is still experimental for ARC, 19286so by default the compiler uses standard reload 19287(i.e.@: @option{-mno-lra}). 19288 19289@item -mlra-priority-none 19290@opindex mlra-priority-none 19291Don't indicate any priority for target registers. 19292 19293@item -mlra-priority-compact 19294@opindex mlra-priority-compact 19295Indicate target register priority for r0..r3 / r12..r15. 19296 19297@item -mlra-priority-noncompact 19298@opindex mlra-priority-noncompact 19299Reduce target register priority for r0..r3 / r12..r15. 19300 19301@item -mmillicode 19302@opindex mmillicode 19303When optimizing for size (using @option{-Os}), prologues and epilogues 19304that have to save or restore a large number of registers are often 19305shortened by using call to a special function in libgcc; this is 19306referred to as a @emph{millicode} call. As these calls can pose 19307performance issues, and/or cause linking issues when linking in a 19308nonstandard way, this option is provided to turn on or off millicode 19309call generation. 19310 19311@item -mcode-density-frame 19312@opindex mcode-density-frame 19313This option enable the compiler to emit @code{enter} and @code{leave} 19314instructions. These instructions are only valid for CPUs with 19315code-density feature. 19316 19317@item -mmixed-code 19318@opindex mmixed-code 19319Tweak register allocation to help 16-bit instruction generation. 19320This generally has the effect of decreasing the average instruction size 19321while increasing the instruction count. 19322 19323@item -mq-class 19324@opindex mq-class 19325Ths option is deprecated. Enable @samp{q} instruction alternatives. 19326This is the default for @option{-Os}. 19327 19328@item -mRcq 19329@opindex mRcq 19330Enable @samp{Rcq} constraint handling. 19331Most short code generation depends on this. 19332This is the default. 19333 19334@item -mRcw 19335@opindex mRcw 19336Enable @samp{Rcw} constraint handling. 19337Most ccfsm condexec mostly depends on this. 19338This is the default. 19339 19340@item -msize-level=@var{level} 19341@opindex msize-level 19342Fine-tune size optimization with regards to instruction lengths and alignment. 19343The recognized values for @var{level} are: 19344@table @samp 19345@item 0 19346No size optimization. This level is deprecated and treated like @samp{1}. 19347 19348@item 1 19349Short instructions are used opportunistically. 19350 19351@item 2 19352In addition, alignment of loops and of code after barriers are dropped. 19353 19354@item 3 19355In addition, optional data alignment is dropped, and the option @option{Os} is enabled. 19356 19357@end table 19358 19359This defaults to @samp{3} when @option{-Os} is in effect. Otherwise, 19360the behavior when this is not set is equivalent to level @samp{1}. 19361 19362@item -mtune=@var{cpu} 19363@opindex mtune 19364Set instruction scheduling parameters for @var{cpu}, overriding any implied 19365by @option{-mcpu=}. 19366 19367Supported values for @var{cpu} are 19368 19369@table @samp 19370@item ARC600 19371Tune for ARC600 CPU. 19372 19373@item ARC601 19374Tune for ARC601 CPU. 19375 19376@item ARC700 19377Tune for ARC700 CPU with standard multiplier block. 19378 19379@item ARC700-xmac 19380Tune for ARC700 CPU with XMAC block. 19381 19382@item ARC725D 19383Tune for ARC725D CPU. 19384 19385@item ARC750D 19386Tune for ARC750D CPU. 19387 19388@end table 19389 19390@item -mmultcost=@var{num} 19391@opindex mmultcost 19392Cost to assume for a multiply instruction, with @samp{4} being equal to a 19393normal instruction. 19394 19395@item -munalign-prob-threshold=@var{probability} 19396@opindex munalign-prob-threshold 19397Set probability threshold for unaligning branches. 19398When tuning for @samp{ARC700} and optimizing for speed, branches without 19399filled delay slot are preferably emitted unaligned and long, unless 19400profiling indicates that the probability for the branch to be taken 19401is below @var{probability}. @xref{Cross-profiling}. 19402The default is (REG_BR_PROB_BASE/2), i.e.@: 5000. 19403 19404@end table 19405 19406The following options are maintained for backward compatibility, but 19407are now deprecated and will be removed in a future release: 19408 19409@c Deprecated options 19410@table @gcctabopt 19411 19412@item -margonaut 19413@opindex margonaut 19414Obsolete FPX. 19415 19416@item -mbig-endian 19417@opindex mbig-endian 19418@itemx -EB 19419@opindex EB 19420Compile code for big-endian targets. Use of these options is now 19421deprecated. Big-endian code is supported by configuring GCC to build 19422@w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets, 19423for which big endian is the default. 19424 19425@item -mlittle-endian 19426@opindex mlittle-endian 19427@itemx -EL 19428@opindex EL 19429Compile code for little-endian targets. Use of these options is now 19430deprecated. Little-endian code is supported by configuring GCC to build 19431@w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets, 19432for which little endian is the default. 19433 19434@item -mbarrel_shifter 19435@opindex mbarrel_shifter 19436Replaced by @option{-mbarrel-shifter}. 19437 19438@item -mdpfp_compact 19439@opindex mdpfp_compact 19440Replaced by @option{-mdpfp-compact}. 19441 19442@item -mdpfp_fast 19443@opindex mdpfp_fast 19444Replaced by @option{-mdpfp-fast}. 19445 19446@item -mdsp_packa 19447@opindex mdsp_packa 19448Replaced by @option{-mdsp-packa}. 19449 19450@item -mEA 19451@opindex mEA 19452Replaced by @option{-mea}. 19453 19454@item -mmac_24 19455@opindex mmac_24 19456Replaced by @option{-mmac-24}. 19457 19458@item -mmac_d16 19459@opindex mmac_d16 19460Replaced by @option{-mmac-d16}. 19461 19462@item -mspfp_compact 19463@opindex mspfp_compact 19464Replaced by @option{-mspfp-compact}. 19465 19466@item -mspfp_fast 19467@opindex mspfp_fast 19468Replaced by @option{-mspfp-fast}. 19469 19470@item -mtune=@var{cpu} 19471@opindex mtune 19472Values @samp{arc600}, @samp{arc601}, @samp{arc700} and 19473@samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600}, 19474@samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively. 19475 19476@item -multcost=@var{num} 19477@opindex multcost 19478Replaced by @option{-mmultcost}. 19479 19480@end table 19481 19482@node ARM Options 19483@subsection ARM Options 19484@cindex ARM options 19485 19486These @samp{-m} options are defined for the ARM port: 19487 19488@table @gcctabopt 19489@item -mabi=@var{name} 19490@opindex mabi 19491Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu}, 19492@samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}. 19493 19494@item -mapcs-frame 19495@opindex mapcs-frame 19496Generate a stack frame that is compliant with the ARM Procedure Call 19497Standard for all functions, even if this is not strictly necessary for 19498correct execution of the code. Specifying @option{-fomit-frame-pointer} 19499with this option causes the stack frames not to be generated for 19500leaf functions. The default is @option{-mno-apcs-frame}. 19501This option is deprecated. 19502 19503@item -mapcs 19504@opindex mapcs 19505This is a synonym for @option{-mapcs-frame} and is deprecated. 19506 19507@ignore 19508@c not currently implemented 19509@item -mapcs-stack-check 19510@opindex mapcs-stack-check 19511Generate code to check the amount of stack space available upon entry to 19512every function (that actually uses some stack space). If there is 19513insufficient space available then either the function 19514@code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is 19515called, depending upon the amount of stack space required. The runtime 19516system is required to provide these functions. The default is 19517@option{-mno-apcs-stack-check}, since this produces smaller code. 19518 19519@c not currently implemented 19520@item -mapcs-reentrant 19521@opindex mapcs-reentrant 19522Generate reentrant, position-independent code. The default is 19523@option{-mno-apcs-reentrant}. 19524@end ignore 19525 19526@item -mthumb-interwork 19527@opindex mthumb-interwork 19528Generate code that supports calling between the ARM and Thumb 19529instruction sets. Without this option, on pre-v5 architectures, the 19530two instruction sets cannot be reliably used inside one program. The 19531default is @option{-mno-thumb-interwork}, since slightly larger code 19532is generated when @option{-mthumb-interwork} is specified. In AAPCS 19533configurations this option is meaningless. 19534 19535@item -mno-sched-prolog 19536@opindex mno-sched-prolog 19537@opindex msched-prolog 19538Prevent the reordering of instructions in the function prologue, or the 19539merging of those instruction with the instructions in the function's 19540body. This means that all functions start with a recognizable set 19541of instructions (or in fact one of a choice from a small set of 19542different function prologues), and this information can be used to 19543locate the start of functions inside an executable piece of code. The 19544default is @option{-msched-prolog}. 19545 19546@item -mfloat-abi=@var{name} 19547@opindex mfloat-abi 19548Specifies which floating-point ABI to use. Permissible values 19549are: @samp{soft}, @samp{softfp} and @samp{hard}. 19550 19551Specifying @samp{soft} causes GCC to generate output containing 19552library calls for floating-point operations. 19553@samp{softfp} allows the generation of code using hardware floating-point 19554instructions, but still uses the soft-float calling conventions. 19555@samp{hard} allows generation of floating-point instructions 19556and uses FPU-specific calling conventions. 19557 19558The default depends on the specific target configuration. Note that 19559the hard-float and soft-float ABIs are not link-compatible; you must 19560compile your entire program with the same ABI, and link with a 19561compatible set of libraries. 19562 19563@item -mgeneral-regs-only 19564@opindex mgeneral-regs-only 19565Generate code which uses only the general-purpose registers. This will prevent 19566the compiler from using floating-point and Advanced SIMD registers but will not 19567impose any restrictions on the assembler. 19568 19569@item -mlittle-endian 19570@opindex mlittle-endian 19571Generate code for a processor running in little-endian mode. This is 19572the default for all standard configurations. 19573 19574@item -mbig-endian 19575@opindex mbig-endian 19576Generate code for a processor running in big-endian mode; the default is 19577to compile code for a little-endian processor. 19578 19579@item -mbe8 19580@itemx -mbe32 19581@opindex mbe8 19582When linking a big-endian image select between BE8 and BE32 formats. 19583The option has no effect for little-endian images and is ignored. The 19584default is dependent on the selected target architecture. For ARMv6 19585and later architectures the default is BE8, for older architectures 19586the default is BE32. BE32 format has been deprecated by ARM. 19587 19588@item -march=@var{name}@r{[}+extension@dots{}@r{]} 19589@opindex march 19590This specifies the name of the target ARM architecture. GCC uses this 19591name to determine what kind of instructions it can emit when generating 19592assembly code. This option can be used in conjunction with or instead 19593of the @option{-mcpu=} option. 19594 19595Permissible names are: 19596@samp{armv4t}, 19597@samp{armv5t}, @samp{armv5te}, 19598@samp{armv6}, @samp{armv6j}, @samp{armv6k}, @samp{armv6kz}, @samp{armv6t2}, 19599@samp{armv6z}, @samp{armv6zk}, 19600@samp{armv7}, @samp{armv7-a}, @samp{armv7ve}, 19601@samp{armv8-a}, @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a}, 19602@samp{armv8.4-a}, 19603@samp{armv8.5-a}, 19604@samp{armv8.6-a}, 19605@samp{armv7-r}, 19606@samp{armv8-r}, 19607@samp{armv6-m}, @samp{armv6s-m}, 19608@samp{armv7-m}, @samp{armv7e-m}, 19609@samp{armv8-m.base}, @samp{armv8-m.main}, 19610@samp{armv8.1-m.main}, 19611@samp{iwmmxt} and @samp{iwmmxt2}. 19612 19613Additionally, the following architectures, which lack support for the 19614Thumb execution state, are recognized but support is deprecated: @samp{armv4}. 19615 19616Many of the architectures support extensions. These can be added by 19617appending @samp{+@var{extension}} to the architecture name. Extension 19618options are processed in order and capabilities accumulate. An extension 19619will also enable any necessary base extensions 19620upon which it depends. For example, the @samp{+crypto} extension 19621will always enable the @samp{+simd} extension. The exception to the 19622additive construction is for extensions that are prefixed with 19623@samp{+no@dots{}}: these extensions disable the specified option and 19624any other extensions that may depend on the presence of that 19625extension. 19626 19627For example, @samp{-march=armv7-a+simd+nofp+vfpv4} is equivalent to 19628writing @samp{-march=armv7-a+vfpv4} since the @samp{+simd} option is 19629entirely disabled by the @samp{+nofp} option that follows it. 19630 19631Most extension names are generically named, but have an effect that is 19632dependent upon the architecture to which it is applied. For example, 19633the @samp{+simd} option can be applied to both @samp{armv7-a} and 19634@samp{armv8-a} architectures, but will enable the original ARMv7-A 19635Advanced SIMD (Neon) extensions for @samp{armv7-a} and the ARMv8-A 19636variant for @samp{armv8-a}. 19637 19638The table below lists the supported extensions for each architecture. 19639Architectures not mentioned do not support any extensions. 19640 19641@table @samp 19642@item armv5te 19643@itemx armv6 19644@itemx armv6j 19645@itemx armv6k 19646@itemx armv6kz 19647@itemx armv6t2 19648@itemx armv6z 19649@itemx armv6zk 19650@table @samp 19651@item +fp 19652The VFPv2 floating-point instructions. The extension @samp{+vfpv2} can be 19653used as an alias for this extension. 19654 19655@item +nofp 19656Disable the floating-point instructions. 19657@end table 19658 19659@item armv7 19660The common subset of the ARMv7-A, ARMv7-R and ARMv7-M architectures. 19661@table @samp 19662@item +fp 19663The VFPv3 floating-point instructions, with 16 double-precision 19664registers. The extension @samp{+vfpv3-d16} can be used as an alias 19665for this extension. Note that floating-point is not supported by the 19666base ARMv7-M architecture, but is compatible with both the ARMv7-A and 19667ARMv7-R architectures. 19668 19669@item +nofp 19670Disable the floating-point instructions. 19671@end table 19672 19673@item armv7-a 19674@table @samp 19675@item +mp 19676The multiprocessing extension. 19677 19678@item +sec 19679The security extension. 19680 19681@item +fp 19682The VFPv3 floating-point instructions, with 16 double-precision 19683registers. The extension @samp{+vfpv3-d16} can be used as an alias 19684for this extension. 19685 19686@item +simd 19687The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions. 19688The extensions @samp{+neon} and @samp{+neon-vfpv3} can be used as aliases 19689for this extension. 19690 19691@item +vfpv3 19692The VFPv3 floating-point instructions, with 32 double-precision 19693registers. 19694 19695@item +vfpv3-d16-fp16 19696The VFPv3 floating-point instructions, with 16 double-precision 19697registers and the half-precision floating-point conversion operations. 19698 19699@item +vfpv3-fp16 19700The VFPv3 floating-point instructions, with 32 double-precision 19701registers and the half-precision floating-point conversion operations. 19702 19703@item +vfpv4-d16 19704The VFPv4 floating-point instructions, with 16 double-precision 19705registers. 19706 19707@item +vfpv4 19708The VFPv4 floating-point instructions, with 32 double-precision 19709registers. 19710 19711@item +neon-fp16 19712The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with 19713the half-precision floating-point conversion operations. 19714 19715@item +neon-vfpv4 19716The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. 19717 19718@item +nosimd 19719Disable the Advanced SIMD instructions (does not disable floating point). 19720 19721@item +nofp 19722Disable the floating-point and Advanced SIMD instructions. 19723@end table 19724 19725@item armv7ve 19726The extended version of the ARMv7-A architecture with support for 19727virtualization. 19728@table @samp 19729@item +fp 19730The VFPv4 floating-point instructions, with 16 double-precision registers. 19731The extension @samp{+vfpv4-d16} can be used as an alias for this extension. 19732 19733@item +simd 19734The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. The 19735extension @samp{+neon-vfpv4} can be used as an alias for this extension. 19736 19737@item +vfpv3-d16 19738The VFPv3 floating-point instructions, with 16 double-precision 19739registers. 19740 19741@item +vfpv3 19742The VFPv3 floating-point instructions, with 32 double-precision 19743registers. 19744 19745@item +vfpv3-d16-fp16 19746The VFPv3 floating-point instructions, with 16 double-precision 19747registers and the half-precision floating-point conversion operations. 19748 19749@item +vfpv3-fp16 19750The VFPv3 floating-point instructions, with 32 double-precision 19751registers and the half-precision floating-point conversion operations. 19752 19753@item +vfpv4-d16 19754The VFPv4 floating-point instructions, with 16 double-precision 19755registers. 19756 19757@item +vfpv4 19758The VFPv4 floating-point instructions, with 32 double-precision 19759registers. 19760 19761@item +neon 19762The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions. 19763The extension @samp{+neon-vfpv3} can be used as an alias for this extension. 19764 19765@item +neon-fp16 19766The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with 19767the half-precision floating-point conversion operations. 19768 19769@item +nosimd 19770Disable the Advanced SIMD instructions (does not disable floating point). 19771 19772@item +nofp 19773Disable the floating-point and Advanced SIMD instructions. 19774@end table 19775 19776@item armv8-a 19777@table @samp 19778@item +crc 19779The Cyclic Redundancy Check (CRC) instructions. 19780@item +simd 19781The ARMv8-A Advanced SIMD and floating-point instructions. 19782@item +crypto 19783The cryptographic instructions. 19784@item +nocrypto 19785Disable the cryptographic instructions. 19786@item +nofp 19787Disable the floating-point, Advanced SIMD and cryptographic instructions. 19788@item +sb 19789Speculation Barrier Instruction. 19790@item +predres 19791Execution and Data Prediction Restriction Instructions. 19792@end table 19793 19794@item armv8.1-a 19795@table @samp 19796@item +simd 19797The ARMv8.1-A Advanced SIMD and floating-point instructions. 19798 19799@item +crypto 19800The cryptographic instructions. This also enables the Advanced SIMD and 19801floating-point instructions. 19802 19803@item +nocrypto 19804Disable the cryptographic instructions. 19805 19806@item +nofp 19807Disable the floating-point, Advanced SIMD and cryptographic instructions. 19808 19809@item +sb 19810Speculation Barrier Instruction. 19811 19812@item +predres 19813Execution and Data Prediction Restriction Instructions. 19814@end table 19815 19816@item armv8.2-a 19817@itemx armv8.3-a 19818@table @samp 19819@item +fp16 19820The half-precision floating-point data processing instructions. 19821This also enables the Advanced SIMD and floating-point instructions. 19822 19823@item +fp16fml 19824The half-precision floating-point fmla extension. This also enables 19825the half-precision floating-point extension and Advanced SIMD and 19826floating-point instructions. 19827 19828@item +simd 19829The ARMv8.1-A Advanced SIMD and floating-point instructions. 19830 19831@item +crypto 19832The cryptographic instructions. This also enables the Advanced SIMD and 19833floating-point instructions. 19834 19835@item +dotprod 19836Enable the Dot Product extension. This also enables Advanced SIMD instructions. 19837 19838@item +nocrypto 19839Disable the cryptographic extension. 19840 19841@item +nofp 19842Disable the floating-point, Advanced SIMD and cryptographic instructions. 19843 19844@item +sb 19845Speculation Barrier Instruction. 19846 19847@item +predres 19848Execution and Data Prediction Restriction Instructions. 19849 19850@item +i8mm 198518-bit Integer Matrix Multiply instructions. 19852This also enables Advanced SIMD and floating-point instructions. 19853 19854@item +bf16 19855Brain half-precision floating-point instructions. 19856This also enables Advanced SIMD and floating-point instructions. 19857@end table 19858 19859@item armv8.4-a 19860@table @samp 19861@item +fp16 19862The half-precision floating-point data processing instructions. 19863This also enables the Advanced SIMD and floating-point instructions as well 19864as the Dot Product extension and the half-precision floating-point fmla 19865extension. 19866 19867@item +simd 19868The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the 19869Dot Product extension. 19870 19871@item +crypto 19872The cryptographic instructions. This also enables the Advanced SIMD and 19873floating-point instructions as well as the Dot Product extension. 19874 19875@item +nocrypto 19876Disable the cryptographic extension. 19877 19878@item +nofp 19879Disable the floating-point, Advanced SIMD and cryptographic instructions. 19880 19881@item +sb 19882Speculation Barrier Instruction. 19883 19884@item +predres 19885Execution and Data Prediction Restriction Instructions. 19886 19887@item +i8mm 198888-bit Integer Matrix Multiply instructions. 19889This also enables Advanced SIMD and floating-point instructions. 19890 19891@item +bf16 19892Brain half-precision floating-point instructions. 19893This also enables Advanced SIMD and floating-point instructions. 19894@end table 19895 19896@item armv8.5-a 19897@table @samp 19898@item +fp16 19899The half-precision floating-point data processing instructions. 19900This also enables the Advanced SIMD and floating-point instructions as well 19901as the Dot Product extension and the half-precision floating-point fmla 19902extension. 19903 19904@item +simd 19905The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the 19906Dot Product extension. 19907 19908@item +crypto 19909The cryptographic instructions. This also enables the Advanced SIMD and 19910floating-point instructions as well as the Dot Product extension. 19911 19912@item +nocrypto 19913Disable the cryptographic extension. 19914 19915@item +nofp 19916Disable the floating-point, Advanced SIMD and cryptographic instructions. 19917 19918@item +i8mm 199198-bit Integer Matrix Multiply instructions. 19920This also enables Advanced SIMD and floating-point instructions. 19921 19922@item +bf16 19923Brain half-precision floating-point instructions. 19924This also enables Advanced SIMD and floating-point instructions. 19925@end table 19926 19927@item armv8.6-a 19928@table @samp 19929@item +fp16 19930The half-precision floating-point data processing instructions. 19931This also enables the Advanced SIMD and floating-point instructions as well 19932as the Dot Product extension and the half-precision floating-point fmla 19933extension. 19934 19935@item +simd 19936The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the 19937Dot Product extension. 19938 19939@item +crypto 19940The cryptographic instructions. This also enables the Advanced SIMD and 19941floating-point instructions as well as the Dot Product extension. 19942 19943@item +nocrypto 19944Disable the cryptographic extension. 19945 19946@item +nofp 19947Disable the floating-point, Advanced SIMD and cryptographic instructions. 19948 19949@item +i8mm 199508-bit Integer Matrix Multiply instructions. 19951This also enables Advanced SIMD and floating-point instructions. 19952 19953@item +bf16 19954Brain half-precision floating-point instructions. 19955This also enables Advanced SIMD and floating-point instructions. 19956@end table 19957 19958@item armv7-r 19959@table @samp 19960@item +fp.sp 19961The single-precision VFPv3 floating-point instructions. The extension 19962@samp{+vfpv3xd} can be used as an alias for this extension. 19963 19964@item +fp 19965The VFPv3 floating-point instructions with 16 double-precision registers. 19966The extension +vfpv3-d16 can be used as an alias for this extension. 19967 19968@item +vfpv3xd-d16-fp16 19969The single-precision VFPv3 floating-point instructions with 16 double-precision 19970registers and the half-precision floating-point conversion operations. 19971 19972@item +vfpv3-d16-fp16 19973The VFPv3 floating-point instructions with 16 double-precision 19974registers and the half-precision floating-point conversion operations. 19975 19976@item +nofp 19977Disable the floating-point extension. 19978 19979@item +idiv 19980The ARM-state integer division instructions. 19981 19982@item +noidiv 19983Disable the ARM-state integer division extension. 19984@end table 19985 19986@item armv7e-m 19987@table @samp 19988@item +fp 19989The single-precision VFPv4 floating-point instructions. 19990 19991@item +fpv5 19992The single-precision FPv5 floating-point instructions. 19993 19994@item +fp.dp 19995The single- and double-precision FPv5 floating-point instructions. 19996 19997@item +nofp 19998Disable the floating-point extensions. 19999@end table 20000 20001@item armv8.1-m.main 20002@table @samp 20003 20004@item +dsp 20005The DSP instructions. 20006 20007@item +mve 20008The M-Profile Vector Extension (MVE) integer instructions. 20009 20010@item +mve.fp 20011The M-Profile Vector Extension (MVE) integer and single precision 20012floating-point instructions. 20013 20014@item +fp 20015The single-precision floating-point instructions. 20016 20017@item +fp.dp 20018The single- and double-precision floating-point instructions. 20019 20020@item +nofp 20021Disable the floating-point extension. 20022 20023@item +cdecp0, +cdecp1, ... , +cdecp7 20024Enable the Custom Datapath Extension (CDE) on selected coprocessors according 20025to the numbers given in the options in the range 0 to 7. 20026@end table 20027 20028@item armv8-m.main 20029@table @samp 20030@item +dsp 20031The DSP instructions. 20032 20033@item +nodsp 20034Disable the DSP extension. 20035 20036@item +fp 20037The single-precision floating-point instructions. 20038 20039@item +fp.dp 20040The single- and double-precision floating-point instructions. 20041 20042@item +nofp 20043Disable the floating-point extension. 20044 20045@item +cdecp0, +cdecp1, ... , +cdecp7 20046Enable the Custom Datapath Extension (CDE) on selected coprocessors according 20047to the numbers given in the options in the range 0 to 7. 20048@end table 20049 20050@item armv8-r 20051@table @samp 20052@item +crc 20053The Cyclic Redundancy Check (CRC) instructions. 20054@item +fp.sp 20055The single-precision FPv5 floating-point instructions. 20056@item +simd 20057The ARMv8-A Advanced SIMD and floating-point instructions. 20058@item +crypto 20059The cryptographic instructions. 20060@item +nocrypto 20061Disable the cryptographic instructions. 20062@item +nofp 20063Disable the floating-point, Advanced SIMD and cryptographic instructions. 20064@end table 20065 20066@end table 20067 20068@option{-march=native} causes the compiler to auto-detect the architecture 20069of the build computer. At present, this feature is only supported on 20070GNU/Linux, and not all architectures are recognized. If the auto-detect 20071is unsuccessful the option has no effect. 20072 20073@item -mtune=@var{name} 20074@opindex mtune 20075This option specifies the name of the target ARM processor for 20076which GCC should tune the performance of the code. 20077For some ARM implementations better performance can be obtained by using 20078this option. 20079Permissible names are: @samp{arm7tdmi}, @samp{arm7tdmi-s}, @samp{arm710t}, 20080@samp{arm720t}, @samp{arm740t}, @samp{strongarm}, @samp{strongarm110}, 20081@samp{strongarm1100}, 0@samp{strongarm1110}, @samp{arm8}, @samp{arm810}, 20082@samp{arm9}, @samp{arm9e}, @samp{arm920}, @samp{arm920t}, @samp{arm922t}, 20083@samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm926ej-s}, 20084@samp{arm940t}, @samp{arm9tdmi}, @samp{arm10tdmi}, @samp{arm1020t}, 20085@samp{arm1026ej-s}, @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e}, 20086@samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp}, 20087@samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s}, 20088@samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, 20089@samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17}, 20090@samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, 20091@samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, 20092@samp{cortex-a76}, @samp{cortex-a76ae}, @samp{cortex-a77}, 20093@samp{cortex-a78}, @samp{cortex-a78ae}, @samp{cortex-a78c}, 20094@samp{ares}, @samp{cortex-r4}, @samp{cortex-r4f}, 20095@samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8}, @samp{cortex-r52}, 20096@samp{cortex-m0}, @samp{cortex-m0plus}, @samp{cortex-m1}, @samp{cortex-m3}, 20097@samp{cortex-m4}, @samp{cortex-m7}, @samp{cortex-m23}, @samp{cortex-m33}, 20098@samp{cortex-m35p}, @samp{cortex-m55}, @samp{cortex-x1}, 20099@samp{cortex-m1.small-multiply}, @samp{cortex-m0.small-multiply}, 20100@samp{cortex-m0plus.small-multiply}, @samp{exynos-m1}, @samp{marvell-pj4}, 20101@samp{neoverse-n1}, @samp{neoverse-n2}, @samp{neoverse-v1}, @samp{xscale}, 20102@samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}, @samp{fa526}, @samp{fa626}, 20103@samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}, @samp{xgene1}. 20104 20105Additionally, this option can specify that GCC should tune the performance 20106of the code for a big.LITTLE system. Permissible names are: 20107@samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7}, 20108@samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}, 20109@samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53}, 20110@samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55}. 20111 20112@option{-mtune=generic-@var{arch}} specifies that GCC should tune the 20113performance for a blend of processors within architecture @var{arch}. 20114The aim is to generate code that run well on the current most popular 20115processors, balancing between optimizations that benefit some CPUs in the 20116range, and avoiding performance pitfalls of other CPUs. The effects of 20117this option may change in future GCC versions as CPU models come and go. 20118 20119@option{-mtune} permits the same extension options as @option{-mcpu}, but 20120the extension options do not affect the tuning of the generated code. 20121 20122@option{-mtune=native} causes the compiler to auto-detect the CPU 20123of the build computer. At present, this feature is only supported on 20124GNU/Linux, and not all architectures are recognized. If the auto-detect is 20125unsuccessful the option has no effect. 20126 20127@item -mcpu=@var{name}@r{[}+extension@dots{}@r{]} 20128@opindex mcpu 20129This specifies the name of the target ARM processor. GCC uses this name 20130to derive the name of the target ARM architecture (as if specified 20131by @option{-march}) and the ARM processor type for which to tune for 20132performance (as if specified by @option{-mtune}). Where this option 20133is used in conjunction with @option{-march} or @option{-mtune}, 20134those options take precedence over the appropriate part of this option. 20135 20136Many of the supported CPUs implement optional architectural 20137extensions. Where this is so the architectural extensions are 20138normally enabled by default. If implementations that lack the 20139extension exist, then the extension syntax can be used to disable 20140those extensions that have been omitted. For floating-point and 20141Advanced SIMD (Neon) instructions, the settings of the options 20142@option{-mfloat-abi} and @option{-mfpu} must also be considered: 20143floating-point and Advanced SIMD instructions will only be used if 20144@option{-mfloat-abi} is not set to @samp{soft}; and any setting of 20145@option{-mfpu} other than @samp{auto} will override the available 20146floating-point and SIMD extension instructions. 20147 20148For example, @samp{cortex-a9} can be found in three major 20149configurations: integer only, with just a floating-point unit or with 20150floating-point and Advanced SIMD. The default is to enable all the 20151instructions, but the extensions @samp{+nosimd} and @samp{+nofp} can 20152be used to disable just the SIMD or both the SIMD and floating-point 20153instructions respectively. 20154 20155Permissible names for this option are the same as those for 20156@option{-mtune}. 20157 20158The following extension options are common to the listed CPUs: 20159 20160@table @samp 20161@item +nodsp 20162Disable the DSP instructions on @samp{cortex-m33}, @samp{cortex-m35p}. 20163 20164@item +nofp 20165Disables the floating-point instructions on @samp{arm9e}, 20166@samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm10e}, 20167@samp{arm1020e}, @samp{arm1022e}, @samp{arm926ej-s}, 20168@samp{arm1026ej-s}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8}, 20169@samp{cortex-m4}, @samp{cortex-m7}, @samp{cortex-m33} and @samp{cortex-m35p}. 20170Disables the floating-point and SIMD instructions on 20171@samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, 20172@samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a12}, 20173@samp{cortex-a15}, @samp{cortex-a17}, @samp{cortex-a15.cortex-a7}, 20174@samp{cortex-a17.cortex-a7}, @samp{cortex-a32}, @samp{cortex-a35}, 20175@samp{cortex-a53} and @samp{cortex-a55}. 20176 20177@item +nofp.dp 20178Disables the double-precision component of the floating-point instructions 20179on @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8}, @samp{cortex-r52} and 20180@samp{cortex-m7}. 20181 20182@item +nosimd 20183Disables the SIMD (but not floating-point) instructions on 20184@samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7} 20185and @samp{cortex-a9}. 20186 20187@item +crypto 20188Enables the cryptographic instructions on @samp{cortex-a32}, 20189@samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, @samp{cortex-a57}, 20190@samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, @samp{exynos-m1}, 20191@samp{xgene1}, @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}, 20192@samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53} and 20193@samp{cortex-a75.cortex-a55}. 20194@end table 20195 20196Additionally the @samp{generic-armv7-a} pseudo target defaults to 20197VFPv3 with 16 double-precision registers. It supports the following 20198extension options: @samp{mp}, @samp{sec}, @samp{vfpv3-d16}, 20199@samp{vfpv3}, @samp{vfpv3-d16-fp16}, @samp{vfpv3-fp16}, 20200@samp{vfpv4-d16}, @samp{vfpv4}, @samp{neon}, @samp{neon-vfpv3}, 20201@samp{neon-fp16}, @samp{neon-vfpv4}. The meanings are the same as for 20202the extensions to @option{-march=armv7-a}. 20203 20204@option{-mcpu=generic-@var{arch}} is also permissible, and is 20205equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}. 20206See @option{-mtune} for more information. 20207 20208@option{-mcpu=native} causes the compiler to auto-detect the CPU 20209of the build computer. At present, this feature is only supported on 20210GNU/Linux, and not all architectures are recognized. If the auto-detect 20211is unsuccessful the option has no effect. 20212 20213@item -mfpu=@var{name} 20214@opindex mfpu 20215This specifies what floating-point hardware (or hardware emulation) is 20216available on the target. Permissible names are: @samp{auto}, @samp{vfpv2}, 20217@samp{vfpv3}, 20218@samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, 20219@samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4}, 20220@samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4}, 20221@samp{fpv5-d16}, @samp{fpv5-sp-d16}, 20222@samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}. 20223Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp} 20224is an alias for @samp{vfpv2}. 20225 20226The setting @samp{auto} is the default and is special. It causes the 20227compiler to select the floating-point and Advanced SIMD instructions 20228based on the settings of @option{-mcpu} and @option{-march}. 20229 20230If the selected floating-point hardware includes the NEON extension 20231(e.g.@: @option{-mfpu=neon}), note that floating-point 20232operations are not generated by GCC's auto-vectorization pass unless 20233@option{-funsafe-math-optimizations} is also specified. This is 20234because NEON hardware does not fully implement the IEEE 754 standard for 20235floating-point arithmetic (in particular denormal values are treated as 20236zero), so the use of NEON instructions may lead to a loss of precision. 20237 20238You can also set the fpu name at function level by using the @code{target("fpu=")} function attributes (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}). 20239 20240@item -mfp16-format=@var{name} 20241@opindex mfp16-format 20242Specify the format of the @code{__fp16} half-precision floating-point type. 20243Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative}; 20244the default is @samp{none}, in which case the @code{__fp16} type is not 20245defined. @xref{Half-Precision}, for more information. 20246 20247@item -mstructure-size-boundary=@var{n} 20248@opindex mstructure-size-boundary 20249The sizes of all structures and unions are rounded up to a multiple 20250of the number of bits set by this option. Permissible values are 8, 32 20251and 64. The default value varies for different toolchains. For the COFF 20252targeted toolchain the default value is 8. A value of 64 is only allowed 20253if the underlying ABI supports it. 20254 20255Specifying a larger number can produce faster, more efficient code, but 20256can also increase the size of the program. Different values are potentially 20257incompatible. Code compiled with one value cannot necessarily expect to 20258work with code or libraries compiled with another value, if they exchange 20259information using structures or unions. 20260 20261This option is deprecated. 20262 20263@item -mabort-on-noreturn 20264@opindex mabort-on-noreturn 20265Generate a call to the function @code{abort} at the end of a 20266@code{noreturn} function. It is executed if the function tries to 20267return. 20268 20269@item -mlong-calls 20270@itemx -mno-long-calls 20271@opindex mlong-calls 20272@opindex mno-long-calls 20273Tells the compiler to perform function calls by first loading the 20274address of the function into a register and then performing a subroutine 20275call on this register. This switch is needed if the target function 20276lies outside of the 64-megabyte addressing range of the offset-based 20277version of subroutine call instruction. 20278 20279Even if this switch is enabled, not all function calls are turned 20280into long calls. The heuristic is that static functions, functions 20281that have the @code{short_call} attribute, functions that are inside 20282the scope of a @code{#pragma no_long_calls} directive, and functions whose 20283definitions have already been compiled within the current compilation 20284unit are not turned into long calls. The exceptions to this rule are 20285that weak function definitions, functions with the @code{long_call} 20286attribute or the @code{section} attribute, and functions that are within 20287the scope of a @code{#pragma long_calls} directive are always 20288turned into long calls. 20289 20290This feature is not enabled by default. Specifying 20291@option{-mno-long-calls} restores the default behavior, as does 20292placing the function calls within the scope of a @code{#pragma 20293long_calls_off} directive. Note these switches have no effect on how 20294the compiler generates code to handle function calls via function 20295pointers. 20296 20297@item -msingle-pic-base 20298@opindex msingle-pic-base 20299Treat the register used for PIC addressing as read-only, rather than 20300loading it in the prologue for each function. The runtime system is 20301responsible for initializing this register with an appropriate value 20302before execution begins. 20303 20304@item -mpic-register=@var{reg} 20305@opindex mpic-register 20306Specify the register to be used for PIC addressing. 20307For standard PIC base case, the default is any suitable register 20308determined by compiler. For single PIC base case, the default is 20309@samp{R9} if target is EABI based or stack-checking is enabled, 20310otherwise the default is @samp{R10}. 20311 20312@item -mpic-data-is-text-relative 20313@opindex mpic-data-is-text-relative 20314Assume that the displacement between the text and data segments is fixed 20315at static link time. This permits using PC-relative addressing 20316operations to access data known to be in the data segment. For 20317non-VxWorks RTP targets, this option is enabled by default. When 20318disabled on such targets, it will enable @option{-msingle-pic-base} by 20319default. 20320 20321@item -mpoke-function-name 20322@opindex mpoke-function-name 20323Write the name of each function into the text section, directly 20324preceding the function prologue. The generated code is similar to this: 20325 20326@smallexample 20327 t0 20328 .ascii "arm_poke_function_name", 0 20329 .align 20330 t1 20331 .word 0xff000000 + (t1 - t0) 20332 arm_poke_function_name 20333 mov ip, sp 20334 stmfd sp!, @{fp, ip, lr, pc@} 20335 sub fp, ip, #4 20336@end smallexample 20337 20338When performing a stack backtrace, code can inspect the value of 20339@code{pc} stored at @code{fp + 0}. If the trace function then looks at 20340location @code{pc - 12} and the top 8 bits are set, then we know that 20341there is a function name embedded immediately preceding this location 20342and has length @code{((pc[-3]) & 0xff000000)}. 20343 20344@item -mthumb 20345@itemx -marm 20346@opindex marm 20347@opindex mthumb 20348 20349Select between generating code that executes in ARM and Thumb 20350states. The default for most configurations is to generate code 20351that executes in ARM state, but the default can be changed by 20352configuring GCC with the @option{--with-mode=}@var{state} 20353configure option. 20354 20355You can also override the ARM and Thumb mode for each function 20356by using the @code{target("thumb")} and @code{target("arm")} function attributes 20357(@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}). 20358 20359@item -mflip-thumb 20360@opindex mflip-thumb 20361Switch ARM/Thumb modes on alternating functions. 20362This option is provided for regression testing of mixed Thumb/ARM code 20363generation, and is not intended for ordinary use in compiling code. 20364 20365@item -mtpcs-frame 20366@opindex mtpcs-frame 20367Generate a stack frame that is compliant with the Thumb Procedure Call 20368Standard for all non-leaf functions. (A leaf function is one that does 20369not call any other functions.) The default is @option{-mno-tpcs-frame}. 20370 20371@item -mtpcs-leaf-frame 20372@opindex mtpcs-leaf-frame 20373Generate a stack frame that is compliant with the Thumb Procedure Call 20374Standard for all leaf functions. (A leaf function is one that does 20375not call any other functions.) The default is @option{-mno-apcs-leaf-frame}. 20376 20377@item -mcallee-super-interworking 20378@opindex mcallee-super-interworking 20379Gives all externally visible functions in the file being compiled an ARM 20380instruction set header which switches to Thumb mode before executing the 20381rest of the function. This allows these functions to be called from 20382non-interworking code. This option is not valid in AAPCS configurations 20383because interworking is enabled by default. 20384 20385@item -mcaller-super-interworking 20386@opindex mcaller-super-interworking 20387Allows calls via function pointers (including virtual functions) to 20388execute correctly regardless of whether the target code has been 20389compiled for interworking or not. There is a small overhead in the cost 20390of executing a function pointer if this option is enabled. This option 20391is not valid in AAPCS configurations because interworking is enabled 20392by default. 20393 20394@item -mtp=@var{name} 20395@opindex mtp 20396Specify the access model for the thread local storage pointer. The valid 20397models are @samp{soft}, which generates calls to @code{__aeabi_read_tp}, 20398@samp{cp15}, which fetches the thread pointer from @code{cp15} directly 20399(supported in the arm6k architecture), and @samp{auto}, which uses the 20400best available method for the selected processor. The default setting is 20401@samp{auto}. 20402 20403@item -mtls-dialect=@var{dialect} 20404@opindex mtls-dialect 20405Specify the dialect to use for accessing thread local storage. Two 20406@var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The 20407@samp{gnu} dialect selects the original GNU scheme for supporting 20408local and global dynamic TLS models. The @samp{gnu2} dialect 20409selects the GNU descriptor scheme, which provides better performance 20410for shared libraries. The GNU descriptor scheme is compatible with 20411the original scheme, but does require new assembler, linker and 20412library support. Initial and local exec TLS models are unaffected by 20413this option and always use the original scheme. 20414 20415@item -mword-relocations 20416@opindex mword-relocations 20417Only generate absolute relocations on word-sized values (i.e.@: R_ARM_ABS32). 20418This is enabled by default on targets (uClinux, SymbianOS) where the runtime 20419loader imposes this restriction, and when @option{-fpic} or @option{-fPIC} 20420is specified. This option conflicts with @option{-mslow-flash-data}. 20421 20422@item -mfix-cortex-m3-ldrd 20423@opindex mfix-cortex-m3-ldrd 20424Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions 20425with overlapping destination and base registers are used. This option avoids 20426generating these instructions. This option is enabled by default when 20427@option{-mcpu=cortex-m3} is specified. 20428 20429@item -munaligned-access 20430@itemx -mno-unaligned-access 20431@opindex munaligned-access 20432@opindex mno-unaligned-access 20433Enables (or disables) reading and writing of 16- and 32- bit values 20434from addresses that are not 16- or 32- bit aligned. By default 20435unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for 20436ARMv8-M Baseline architectures, and enabled for all other 20437architectures. If unaligned access is not enabled then words in packed 20438data structures are accessed a byte at a time. 20439 20440The ARM attribute @code{Tag_CPU_unaligned_access} is set in the 20441generated object file to either true or false, depending upon the 20442setting of this option. If unaligned access is enabled then the 20443preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also 20444defined. 20445 20446@item -mneon-for-64bits 20447@opindex mneon-for-64bits 20448This option is deprecated and has no effect. 20449 20450@item -mslow-flash-data 20451@opindex mslow-flash-data 20452Assume loading data from flash is slower than fetching instruction. 20453Therefore literal load is minimized for better performance. 20454This option is only supported when compiling for ARMv7 M-profile and 20455off by default. It conflicts with @option{-mword-relocations}. 20456 20457@item -masm-syntax-unified 20458@opindex masm-syntax-unified 20459Assume inline assembler is using unified asm syntax. The default is 20460currently off which implies divided syntax. This option has no impact 20461on Thumb2. However, this may change in future releases of GCC. 20462Divided syntax should be considered deprecated. 20463 20464@item -mrestrict-it 20465@opindex mrestrict-it 20466Restricts generation of IT blocks to conform to the rules of ARMv8-A. 20467IT blocks can only contain a single 16-bit instruction from a select 20468set of instructions. This option is on by default for ARMv8-A Thumb mode. 20469 20470@item -mprint-tune-info 20471@opindex mprint-tune-info 20472Print CPU tuning information as comment in assembler file. This is 20473an option used only for regression testing of the compiler and not 20474intended for ordinary use in compiling code. This option is disabled 20475by default. 20476 20477@item -mverbose-cost-dump 20478@opindex mverbose-cost-dump 20479Enable verbose cost model dumping in the debug dump files. This option is 20480provided for use in debugging the compiler. 20481 20482@item -mpure-code 20483@opindex mpure-code 20484Do not allow constant data to be placed in code sections. 20485Additionally, when compiling for ELF object format give all text sections the 20486ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option 20487is only available when generating non-pic code for M-profile targets. 20488 20489@item -mcmse 20490@opindex mcmse 20491Generate secure code as per the "ARMv8-M Security Extensions: Requirements on 20492Development Tools Engineering Specification", which can be found on 20493@url{https://developer.arm.com/documentation/ecm0359818/latest/}. 20494 20495@item -mfix-cmse-cve-2021-35465 20496@opindex mfix-cmse-cve-2021-35465 20497Mitigate against a potential security issue with the @code{VLLDM} instruction 20498in some M-profile devices when using CMSE (CVE-2021-365465). This option is 20499enabled by default when the option @option{-mcpu=} is used with 20500@code{cortex-m33}, @code{cortex-m35p} or @code{cortex-m55}. The option 20501@option{-mno-fix-cmse-cve-2021-35465} can be used to disable the mitigation. 20502 20503@item -mfdpic 20504@itemx -mno-fdpic 20505@opindex mfdpic 20506@opindex mno-fdpic 20507Select the FDPIC ABI, which uses 64-bit function descriptors to 20508represent pointers to functions. When the compiler is configured for 20509@code{arm-*-uclinuxfdpiceabi} targets, this option is on by default 20510and implies @option{-fPIE} if none of the PIC/PIE-related options is 20511provided. On other targets, it only enables the FDPIC-specific code 20512generation features, and the user should explicitly provide the 20513PIC/PIE-related options as needed. 20514 20515Note that static linking is not supported because it would still 20516involve the dynamic linker when the program self-relocates. If such 20517behavior is acceptable, use -static and -Wl,-dynamic-linker options. 20518 20519The opposite @option{-mno-fdpic} option is useful (and required) to 20520build the Linux kernel using the same (@code{arm-*-uclinuxfdpiceabi}) 20521toolchain as the one used to build the userland programs. 20522 20523@end table 20524 20525@node AVR Options 20526@subsection AVR Options 20527@cindex AVR Options 20528 20529These options are defined for AVR implementations: 20530 20531@table @gcctabopt 20532@item -mmcu=@var{mcu} 20533@opindex mmcu 20534Specify Atmel AVR instruction set architectures (ISA) or MCU type. 20535 20536The default for this option is@tie{}@samp{avr2}. 20537 20538GCC supports the following AVR devices and ISAs: 20539 20540@include avr-mmcu.texi 20541 20542@item -mabsdata 20543@opindex mabsdata 20544 20545Assume that all data in static storage can be accessed by LDS / STS 20546instructions. This option has only an effect on reduced Tiny devices like 20547ATtiny40. See also the @code{absdata} 20548@ref{AVR Variable Attributes,variable attribute}. 20549 20550@item -maccumulate-args 20551@opindex maccumulate-args 20552Accumulate outgoing function arguments and acquire/release the needed 20553stack space for outgoing function arguments once in function 20554prologue/epilogue. Without this option, outgoing arguments are pushed 20555before calling a function and popped afterwards. 20556 20557Popping the arguments after the function call can be expensive on 20558AVR so that accumulating the stack space might lead to smaller 20559executables because arguments need not be removed from the 20560stack after such a function call. 20561 20562This option can lead to reduced code size for functions that perform 20563several calls to functions that get their arguments on the stack like 20564calls to printf-like functions. 20565 20566@item -mbranch-cost=@var{cost} 20567@opindex mbranch-cost 20568Set the branch costs for conditional branch instructions to 20569@var{cost}. Reasonable values for @var{cost} are small, non-negative 20570integers. The default branch cost is 0. 20571 20572@item -mcall-prologues 20573@opindex mcall-prologues 20574Functions prologues/epilogues are expanded as calls to appropriate 20575subroutines. Code size is smaller. 20576 20577@item -mdouble=@var{bits} 20578@itemx -mlong-double=@var{bits} 20579@opindex mdouble 20580@opindex mlong-double 20581Set the size (in bits) of the @code{double} or @code{long double} type, 20582respectively. Possible values for @var{bits} are 32 and 64. 20583Whether or not a specific value for @var{bits} is allowed depends on 20584the @code{--with-double=} and @code{--with-long-double=} 20585@w{@uref{https://gcc.gnu.org/install/configure.html#avr,configure options}}, 20586and the same applies for the default values of the options. 20587 20588@item -mgas-isr-prologues 20589@opindex mgas-isr-prologues 20590Interrupt service routines (ISRs) may use the @code{__gcc_isr} pseudo 20591instruction supported by GNU Binutils. 20592If this option is on, the feature can still be disabled for individual 20593ISRs by means of the @ref{AVR Function Attributes,,@code{no_gccisr}} 20594function attribute. This feature is activated per default 20595if optimization is on (but not with @option{-Og}, @pxref{Optimize Options}), 20596and if GNU Binutils support @w{@uref{https://sourceware.org/PR21683,PR21683}}. 20597 20598@item -mint8 20599@opindex mint8 20600Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a 20601@code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes, 20602and @code{long long} is 4 bytes. Please note that this option does not 20603conform to the C standards, but it results in smaller code 20604size. 20605 20606@item -mmain-is-OS_task 20607@opindex mmain-is-OS_task 20608Do not save registers in @code{main}. The effect is the same like 20609attaching attribute @ref{AVR Function Attributes,,@code{OS_task}} 20610to @code{main}. It is activated per default if optimization is on. 20611 20612@item -mn-flash=@var{num} 20613@opindex mn-flash 20614Assume that the flash memory has a size of 20615@var{num} times 64@tie{}KiB. 20616 20617@item -mno-interrupts 20618@opindex mno-interrupts 20619Generated code is not compatible with hardware interrupts. 20620Code size is smaller. 20621 20622@item -mrelax 20623@opindex mrelax 20624Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter 20625@code{RCALL} resp.@: @code{RJMP} instruction if applicable. 20626Setting @option{-mrelax} just adds the @option{--mlink-relax} option to 20627the assembler's command line and the @option{--relax} option to the 20628linker's command line. 20629 20630Jump relaxing is performed by the linker because jump offsets are not 20631known before code is located. Therefore, the assembler code generated by the 20632compiler is the same, but the instructions in the executable may 20633differ from instructions in the assembler code. 20634 20635Relaxing must be turned on if linker stubs are needed, see the 20636section on @code{EIND} and linker stubs below. 20637 20638@item -mrmw 20639@opindex mrmw 20640Assume that the device supports the Read-Modify-Write 20641instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}. 20642 20643@item -mshort-calls 20644@opindex mshort-calls 20645 20646Assume that @code{RJMP} and @code{RCALL} can target the whole 20647program memory. 20648 20649This option is used internally for multilib selection. It is 20650not an optimization option, and you don't need to set it by hand. 20651 20652@item -msp8 20653@opindex msp8 20654Treat the stack pointer register as an 8-bit register, 20655i.e.@: assume the high byte of the stack pointer is zero. 20656In general, you don't need to set this option by hand. 20657 20658This option is used internally by the compiler to select and 20659build multilibs for architectures @code{avr2} and @code{avr25}. 20660These architectures mix devices with and without @code{SPH}. 20661For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25} 20662the compiler driver adds or removes this option from the compiler 20663proper's command line, because the compiler then knows if the device 20664or architecture has an 8-bit stack pointer and thus no @code{SPH} 20665register or not. 20666 20667@item -mstrict-X 20668@opindex mstrict-X 20669Use address register @code{X} in a way proposed by the hardware. This means 20670that @code{X} is only used in indirect, post-increment or 20671pre-decrement addressing. 20672 20673Without this option, the @code{X} register may be used in the same way 20674as @code{Y} or @code{Z} which then is emulated by additional 20675instructions. 20676For example, loading a value with @code{X+const} addressing with a 20677small non-negative @code{const < 64} to a register @var{Rn} is 20678performed as 20679 20680@example 20681adiw r26, const ; X += const 20682ld @var{Rn}, X ; @var{Rn} = *X 20683sbiw r26, const ; X -= const 20684@end example 20685 20686@item -mtiny-stack 20687@opindex mtiny-stack 20688Only change the lower 8@tie{}bits of the stack pointer. 20689 20690@item -mfract-convert-truncate 20691@opindex mfract-convert-truncate 20692Allow to use truncation instead of rounding towards zero for fractional fixed-point types. 20693 20694@item -nodevicelib 20695@opindex nodevicelib 20696Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}. 20697 20698@item -nodevicespecs 20699@opindex nodevicespecs 20700Don't add @option{-specs=device-specs/specs-@var{mcu}} to the compiler driver's 20701command line. The user takes responsibility for supplying the sub-processes 20702like compiler proper, assembler and linker with appropriate command line 20703options. This means that the user has to supply her private device specs 20704file by means of @option{-specs=@var{path-to-specs-file}}. There is no 20705more need for option @option{-mmcu=@var{mcu}}. 20706 20707This option can also serve as a replacement for the older way of 20708specifying custom device-specs files that needed @option{-B @var{some-path}} to point to a directory 20709which contains a folder named @code{device-specs} which contains a specs file named 20710@code{specs-@var{mcu}}, where @var{mcu} was specified by @option{-mmcu=@var{mcu}}. 20711 20712@item -Waddr-space-convert 20713@opindex Waddr-space-convert 20714@opindex Wno-addr-space-convert 20715Warn about conversions between address spaces in the case where the 20716resulting address space is not contained in the incoming address space. 20717 20718@item -Wmisspelled-isr 20719@opindex Wmisspelled-isr 20720@opindex Wno-misspelled-isr 20721Warn if the ISR is misspelled, i.e.@: without __vector prefix. 20722Enabled by default. 20723@end table 20724 20725@subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash 20726@cindex @code{EIND} 20727Pointers in the implementation are 16@tie{}bits wide. 20728The address of a function or label is represented as word address so 20729that indirect jumps and calls can target any code address in the 20730range of 64@tie{}Ki words. 20731 20732In order to facilitate indirect jump on devices with more than 128@tie{}Ki 20733bytes of program memory space, there is a special function register called 20734@code{EIND} that serves as most significant part of the target address 20735when @code{EICALL} or @code{EIJMP} instructions are used. 20736 20737Indirect jumps and calls on these devices are handled as follows by 20738the compiler and are subject to some limitations: 20739 20740@itemize @bullet 20741 20742@item 20743The compiler never sets @code{EIND}. 20744 20745@item 20746The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP} 20747instructions or might read @code{EIND} directly in order to emulate an 20748indirect call/jump by means of a @code{RET} instruction. 20749 20750@item 20751The compiler assumes that @code{EIND} never changes during the startup 20752code or during the application. In particular, @code{EIND} is not 20753saved/restored in function or interrupt service routine 20754prologue/epilogue. 20755 20756@item 20757For indirect calls to functions and computed goto, the linker 20758generates @emph{stubs}. Stubs are jump pads sometimes also called 20759@emph{trampolines}. Thus, the indirect call/jump jumps to such a stub. 20760The stub contains a direct jump to the desired address. 20761 20762@item 20763Linker relaxation must be turned on so that the linker generates 20764the stubs correctly in all situations. See the compiler option 20765@option{-mrelax} and the linker option @option{--relax}. 20766There are corner cases where the linker is supposed to generate stubs 20767but aborts without relaxation and without a helpful error message. 20768 20769@item 20770The default linker script is arranged for code with @code{EIND = 0}. 20771If code is supposed to work for a setup with @code{EIND != 0}, a custom 20772linker script has to be used in order to place the sections whose 20773name start with @code{.trampolines} into the segment where @code{EIND} 20774points to. 20775 20776@item 20777The startup code from libgcc never sets @code{EIND}. 20778Notice that startup code is a blend of code from libgcc and AVR-LibC. 20779For the impact of AVR-LibC on @code{EIND}, see the 20780@w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}. 20781 20782@item 20783It is legitimate for user-specific startup code to set up @code{EIND} 20784early, for example by means of initialization code located in 20785section @code{.init3}. Such code runs prior to general startup code 20786that initializes RAM and calls constructors, but after the bit 20787of startup code from AVR-LibC that sets @code{EIND} to the segment 20788where the vector table is located. 20789@example 20790#include <avr/io.h> 20791 20792static void 20793__attribute__((section(".init3"),naked,used,no_instrument_function)) 20794init3_set_eind (void) 20795@{ 20796 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t" 20797 "out %i0,r24" :: "n" (&EIND) : "r24","memory"); 20798@} 20799@end example 20800 20801@noindent 20802The @code{__trampolines_start} symbol is defined in the linker script. 20803 20804@item 20805Stubs are generated automatically by the linker if 20806the following two conditions are met: 20807@itemize @minus 20808 20809@item The address of a label is taken by means of the @code{gs} modifier 20810(short for @emph{generate stubs}) like so: 20811@example 20812LDI r24, lo8(gs(@var{func})) 20813LDI r25, hi8(gs(@var{func})) 20814@end example 20815@item The final location of that label is in a code segment 20816@emph{outside} the segment where the stubs are located. 20817@end itemize 20818 20819@item 20820The compiler emits such @code{gs} modifiers for code labels in the 20821following situations: 20822@itemize @minus 20823@item Taking address of a function or code label. 20824@item Computed goto. 20825@item If prologue-save function is used, see @option{-mcall-prologues} 20826command-line option. 20827@item Switch/case dispatch tables. If you do not want such dispatch 20828tables you can specify the @option{-fno-jump-tables} command-line option. 20829@item C and C++ constructors/destructors called during startup/shutdown. 20830@item If the tools hit a @code{gs()} modifier explained above. 20831@end itemize 20832 20833@item 20834Jumping to non-symbolic addresses like so is @emph{not} supported: 20835 20836@example 20837int main (void) 20838@{ 20839 /* Call function at word address 0x2 */ 20840 return ((int(*)(void)) 0x2)(); 20841@} 20842@end example 20843 20844Instead, a stub has to be set up, i.e.@: the function has to be called 20845through a symbol (@code{func_4} in the example): 20846 20847@example 20848int main (void) 20849@{ 20850 extern int func_4 (void); 20851 20852 /* Call function at byte address 0x4 */ 20853 return func_4(); 20854@} 20855@end example 20856 20857and the application be linked with @option{-Wl,--defsym,func_4=0x4}. 20858Alternatively, @code{func_4} can be defined in the linker script. 20859@end itemize 20860 20861@subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers 20862@cindex @code{RAMPD} 20863@cindex @code{RAMPX} 20864@cindex @code{RAMPY} 20865@cindex @code{RAMPZ} 20866Some AVR devices support memories larger than the 64@tie{}KiB range 20867that can be accessed with 16-bit pointers. To access memory locations 20868outside this 64@tie{}KiB range, the content of a @code{RAMP} 20869register is used as high part of the address: 20870The @code{X}, @code{Y}, @code{Z} address register is concatenated 20871with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function 20872register, respectively, to get a wide address. Similarly, 20873@code{RAMPD} is used together with direct addressing. 20874 20875@itemize 20876@item 20877The startup code initializes the @code{RAMP} special function 20878registers with zero. 20879 20880@item 20881If a @ref{AVR Named Address Spaces,named address space} other than 20882generic or @code{__flash} is used, then @code{RAMPZ} is set 20883as needed before the operation. 20884 20885@item 20886If the device supports RAM larger than 64@tie{}KiB and the compiler 20887needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ} 20888is reset to zero after the operation. 20889 20890@item 20891If the device comes with a specific @code{RAMP} register, the ISR 20892prologue/epilogue saves/restores that SFR and initializes it with 20893zero in case the ISR code might (implicitly) use it. 20894 20895@item 20896RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets. 20897If you use inline assembler to read from locations outside the 2089816-bit address range and change one of the @code{RAMP} registers, 20899you must reset it to zero after the access. 20900 20901@end itemize 20902 20903@subsubsection AVR Built-in Macros 20904 20905GCC defines several built-in macros so that the user code can test 20906for the presence or absence of features. Almost any of the following 20907built-in macros are deduced from device capabilities and thus 20908triggered by the @option{-mmcu=} command-line option. 20909 20910For even more AVR-specific built-in macros see 20911@ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}. 20912 20913@table @code 20914 20915@item __AVR_ARCH__ 20916Build-in macro that resolves to a decimal number that identifies the 20917architecture and depends on the @option{-mmcu=@var{mcu}} option. 20918Possible values are: 20919 20920@code{2}, @code{25}, @code{3}, @code{31}, @code{35}, 20921@code{4}, @code{5}, @code{51}, @code{6} 20922 20923for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31}, 20924@code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6}, 20925 20926respectively and 20927 20928@code{100}, 20929@code{102}, @code{103}, @code{104}, 20930@code{105}, @code{106}, @code{107} 20931 20932for @var{mcu}=@code{avrtiny}, 20933@code{avrxmega2}, @code{avrxmega3}, @code{avrxmega4}, 20934@code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively. 20935If @var{mcu} specifies a device, this built-in macro is set 20936accordingly. For example, with @option{-mmcu=atmega8} the macro is 20937defined to @code{4}. 20938 20939@item __AVR_@var{Device}__ 20940Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects 20941the device's name. For example, @option{-mmcu=atmega8} defines the 20942built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines 20943@code{__AVR_ATtiny261A__}, etc. 20944 20945The built-in macros' names follow 20946the scheme @code{__AVR_@var{Device}__} where @var{Device} is 20947the device name as from the AVR user manual. The difference between 20948@var{Device} in the built-in macro and @var{device} in 20949@option{-mmcu=@var{device}} is that the latter is always lowercase. 20950 20951If @var{device} is not a device but only a core architecture like 20952@samp{avr51}, this macro is not defined. 20953 20954@item __AVR_DEVICE_NAME__ 20955Setting @option{-mmcu=@var{device}} defines this built-in macro to 20956the device's name. For example, with @option{-mmcu=atmega8} the macro 20957is defined to @code{atmega8}. 20958 20959If @var{device} is not a device but only a core architecture like 20960@samp{avr51}, this macro is not defined. 20961 20962@item __AVR_XMEGA__ 20963The device / architecture belongs to the XMEGA family of devices. 20964 20965@item __AVR_HAVE_ELPM__ 20966The device has the @code{ELPM} instruction. 20967 20968@item __AVR_HAVE_ELPMX__ 20969The device has the @code{ELPM R@var{n},Z} and @code{ELPM 20970R@var{n},Z+} instructions. 20971 20972@item __AVR_HAVE_MOVW__ 20973The device has the @code{MOVW} instruction to perform 16-bit 20974register-register moves. 20975 20976@item __AVR_HAVE_LPMX__ 20977The device has the @code{LPM R@var{n},Z} and 20978@code{LPM R@var{n},Z+} instructions. 20979 20980@item __AVR_HAVE_MUL__ 20981The device has a hardware multiplier. 20982 20983@item __AVR_HAVE_JMP_CALL__ 20984The device has the @code{JMP} and @code{CALL} instructions. 20985This is the case for devices with more than 8@tie{}KiB of program 20986memory. 20987 20988@item __AVR_HAVE_EIJMP_EICALL__ 20989@itemx __AVR_3_BYTE_PC__ 20990The device has the @code{EIJMP} and @code{EICALL} instructions. 20991This is the case for devices with more than 128@tie{}KiB of program memory. 20992This also means that the program counter 20993(PC) is 3@tie{}bytes wide. 20994 20995@item __AVR_2_BYTE_PC__ 20996The program counter (PC) is 2@tie{}bytes wide. This is the case for devices 20997with up to 128@tie{}KiB of program memory. 20998 20999@item __AVR_HAVE_8BIT_SP__ 21000@itemx __AVR_HAVE_16BIT_SP__ 21001The stack pointer (SP) register is treated as 8-bit respectively 2100216-bit register by the compiler. 21003The definition of these macros is affected by @option{-mtiny-stack}. 21004 21005@item __AVR_HAVE_SPH__ 21006@itemx __AVR_SP8__ 21007The device has the SPH (high part of stack pointer) special function 21008register or has an 8-bit stack pointer, respectively. 21009The definition of these macros is affected by @option{-mmcu=} and 21010in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also 21011by @option{-msp8}. 21012 21013@item __AVR_HAVE_RAMPD__ 21014@itemx __AVR_HAVE_RAMPX__ 21015@itemx __AVR_HAVE_RAMPY__ 21016@itemx __AVR_HAVE_RAMPZ__ 21017The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY}, 21018@code{RAMPZ} special function register, respectively. 21019 21020@item __NO_INTERRUPTS__ 21021This macro reflects the @option{-mno-interrupts} command-line option. 21022 21023@item __AVR_ERRATA_SKIP__ 21024@itemx __AVR_ERRATA_SKIP_JMP_CALL__ 21025Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit 21026instructions because of a hardware erratum. Skip instructions are 21027@code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}. 21028The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also 21029set. 21030 21031@item __AVR_ISA_RMW__ 21032The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT). 21033 21034@item __AVR_SFR_OFFSET__=@var{offset} 21035Instructions that can address I/O special function registers directly 21036like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different 21037address as if addressed by an instruction to access RAM like @code{LD} 21038or @code{STS}. This offset depends on the device architecture and has 21039to be subtracted from the RAM address in order to get the 21040respective I/O@tie{}address. 21041 21042@item __AVR_SHORT_CALLS__ 21043The @option{-mshort-calls} command line option is set. 21044 21045@item __AVR_PM_BASE_ADDRESS__=@var{addr} 21046Some devices support reading from flash memory by means of @code{LD*} 21047instructions. The flash memory is seen in the data address space 21048at an offset of @code{__AVR_PM_BASE_ADDRESS__}. If this macro 21049is not defined, this feature is not available. If defined, 21050the address space is linear and there is no need to put 21051@code{.rodata} into RAM. This is handled by the default linker 21052description file, and is currently available for 21053@code{avrtiny} and @code{avrxmega3}. Even more convenient, 21054there is no need to use address spaces like @code{__flash} or 21055features like attribute @code{progmem} and @code{pgm_read_*}. 21056 21057@item __WITH_AVRLIBC__ 21058The compiler is configured to be used together with AVR-Libc. 21059See the @option{--with-avrlibc} configure option. 21060 21061@item __HAVE_DOUBLE_MULTILIB__ 21062Defined if @option{-mdouble=} acts as a multilib option. 21063 21064@item __HAVE_DOUBLE32__ 21065@itemx __HAVE_DOUBLE64__ 21066Defined if the compiler supports 32-bit double resp. 64-bit double. 21067The actual layout is specified by option @option{-mdouble=}. 21068 21069@item __DEFAULT_DOUBLE__ 21070The size in bits of @code{double} if @option{-mdouble=} is not set. 21071To test the layout of @code{double} in a program, use the built-in 21072macro @code{__SIZEOF_DOUBLE__}. 21073 21074@item __HAVE_LONG_DOUBLE32__ 21075@itemx __HAVE_LONG_DOUBLE64__ 21076@itemx __HAVE_LONG_DOUBLE_MULTILIB__ 21077@itemx __DEFAULT_LONG_DOUBLE__ 21078Same as above, but for @code{long double} instead of @code{double}. 21079 21080@item __WITH_DOUBLE_COMPARISON__ 21081Reflects the @code{--with-double-comparison=@{tristate|bool|libf7@}} 21082@w{@uref{https://gcc.gnu.org/install/configure.html#avr,configure option}} 21083and is defined to @code{2} or @code{3}. 21084 21085@item __WITH_LIBF7_LIBGCC__ 21086@itemx __WITH_LIBF7_MATH__ 21087@itemx __WITH_LIBF7_MATH_SYMBOLS__ 21088Reflects the @code{--with-libf7=@{libgcc|math|math-symbols@}} 21089@w{@uref{https://gcc.gnu.org/install/configure.html#avr,configure option}}. 21090 21091@end table 21092 21093@node Blackfin Options 21094@subsection Blackfin Options 21095@cindex Blackfin Options 21096 21097@table @gcctabopt 21098@item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} 21099@opindex mcpu= 21100Specifies the name of the target Blackfin processor. Currently, @var{cpu} 21101can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518}, 21102@samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526}, 21103@samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533}, 21104@samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539}, 21105@samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549}, 21106@samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m}, 21107@samp{bf561}, @samp{bf592}. 21108 21109The optional @var{sirevision} specifies the silicon revision of the target 21110Blackfin processor. Any workarounds available for the targeted silicon revision 21111are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled. 21112If @var{sirevision} is @samp{any}, all workarounds for the targeted processor 21113are enabled. The @code{__SILICON_REVISION__} macro is defined to two 21114hexadecimal digits representing the major and minor numbers in the silicon 21115revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__} 21116is not defined. If @var{sirevision} is @samp{any}, the 21117@code{__SILICON_REVISION__} is defined to be @code{0xffff}. 21118If this optional @var{sirevision} is not used, GCC assumes the latest known 21119silicon revision of the targeted Blackfin processor. 21120 21121GCC defines a preprocessor macro for the specified @var{cpu}. 21122For the @samp{bfin-elf} toolchain, this option causes the hardware BSP 21123provided by libgloss to be linked in if @option{-msim} is not given. 21124 21125Without this option, @samp{bf532} is used as the processor by default. 21126 21127Note that support for @samp{bf561} is incomplete. For @samp{bf561}, 21128only the preprocessor macro is defined. 21129 21130@item -msim 21131@opindex msim 21132Specifies that the program will be run on the simulator. This causes 21133the simulator BSP provided by libgloss to be linked in. This option 21134has effect only for @samp{bfin-elf} toolchain. 21135Certain other options, such as @option{-mid-shared-library} and 21136@option{-mfdpic}, imply @option{-msim}. 21137 21138@item -momit-leaf-frame-pointer 21139@opindex momit-leaf-frame-pointer 21140Don't keep the frame pointer in a register for leaf functions. This 21141avoids the instructions to save, set up and restore frame pointers and 21142makes an extra register available in leaf functions. 21143 21144@item -mspecld-anomaly 21145@opindex mspecld-anomaly 21146When enabled, the compiler ensures that the generated code does not 21147contain speculative loads after jump instructions. If this option is used, 21148@code{__WORKAROUND_SPECULATIVE_LOADS} is defined. 21149 21150@item -mno-specld-anomaly 21151@opindex mno-specld-anomaly 21152@opindex mspecld-anomaly 21153Don't generate extra code to prevent speculative loads from occurring. 21154 21155@item -mcsync-anomaly 21156@opindex mcsync-anomaly 21157When enabled, the compiler ensures that the generated code does not 21158contain CSYNC or SSYNC instructions too soon after conditional branches. 21159If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined. 21160 21161@item -mno-csync-anomaly 21162@opindex mno-csync-anomaly 21163@opindex mcsync-anomaly 21164Don't generate extra code to prevent CSYNC or SSYNC instructions from 21165occurring too soon after a conditional branch. 21166 21167@item -mlow64k 21168@opindex mlow64k 21169When enabled, the compiler is free to take advantage of the knowledge that 21170the entire program fits into the low 64k of memory. 21171 21172@item -mno-low64k 21173@opindex mno-low64k 21174Assume that the program is arbitrarily large. This is the default. 21175 21176@item -mstack-check-l1 21177@opindex mstack-check-l1 21178Do stack checking using information placed into L1 scratchpad memory by the 21179uClinux kernel. 21180 21181@item -mid-shared-library 21182@opindex mid-shared-library 21183Generate code that supports shared libraries via the library ID method. 21184This allows for execute in place and shared libraries in an environment 21185without virtual memory management. This option implies @option{-fPIC}. 21186With a @samp{bfin-elf} target, this option implies @option{-msim}. 21187 21188@item -mno-id-shared-library 21189@opindex mno-id-shared-library 21190@opindex mid-shared-library 21191Generate code that doesn't assume ID-based shared libraries are being used. 21192This is the default. 21193 21194@item -mleaf-id-shared-library 21195@opindex mleaf-id-shared-library 21196Generate code that supports shared libraries via the library ID method, 21197but assumes that this library or executable won't link against any other 21198ID shared libraries. That allows the compiler to use faster code for jumps 21199and calls. 21200 21201@item -mno-leaf-id-shared-library 21202@opindex mno-leaf-id-shared-library 21203@opindex mleaf-id-shared-library 21204Do not assume that the code being compiled won't link against any ID shared 21205libraries. Slower code is generated for jump and call insns. 21206 21207@item -mshared-library-id=n 21208@opindex mshared-library-id 21209Specifies the identification number of the ID-based shared library being 21210compiled. Specifying a value of 0 generates more compact code; specifying 21211other values forces the allocation of that number to the current 21212library but is no more space- or time-efficient than omitting this option. 21213 21214@item -msep-data 21215@opindex msep-data 21216Generate code that allows the data segment to be located in a different 21217area of memory from the text segment. This allows for execute in place in 21218an environment without virtual memory management by eliminating relocations 21219against the text section. 21220 21221@item -mno-sep-data 21222@opindex mno-sep-data 21223@opindex msep-data 21224Generate code that assumes that the data segment follows the text segment. 21225This is the default. 21226 21227@item -mlong-calls 21228@itemx -mno-long-calls 21229@opindex mlong-calls 21230@opindex mno-long-calls 21231Tells the compiler to perform function calls by first loading the 21232address of the function into a register and then performing a subroutine 21233call on this register. This switch is needed if the target function 21234lies outside of the 24-bit addressing range of the offset-based 21235version of subroutine call instruction. 21236 21237This feature is not enabled by default. Specifying 21238@option{-mno-long-calls} restores the default behavior. Note these 21239switches have no effect on how the compiler generates code to handle 21240function calls via function pointers. 21241 21242@item -mfast-fp 21243@opindex mfast-fp 21244Link with the fast floating-point library. This library relaxes some of 21245the IEEE floating-point standard's rules for checking inputs against 21246Not-a-Number (NAN), in the interest of performance. 21247 21248@item -minline-plt 21249@opindex minline-plt 21250Enable inlining of PLT entries in function calls to functions that are 21251not known to bind locally. It has no effect without @option{-mfdpic}. 21252 21253@item -mmulticore 21254@opindex mmulticore 21255Build a standalone application for multicore Blackfin processors. 21256This option causes proper start files and link scripts supporting 21257multicore to be used, and defines the macro @code{__BFIN_MULTICORE}. 21258It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. 21259 21260This option can be used with @option{-mcorea} or @option{-mcoreb}, which 21261selects the one-application-per-core programming model. Without 21262@option{-mcorea} or @option{-mcoreb}, the single-application/dual-core 21263programming model is used. In this model, the main function of Core B 21264should be named as @code{coreb_main}. 21265 21266If this option is not used, the single-core application programming 21267model is used. 21268 21269@item -mcorea 21270@opindex mcorea 21271Build a standalone application for Core A of BF561 when using 21272the one-application-per-core programming model. Proper start files 21273and link scripts are used to support Core A, and the macro 21274@code{__BFIN_COREA} is defined. 21275This option can only be used in conjunction with @option{-mmulticore}. 21276 21277@item -mcoreb 21278@opindex mcoreb 21279Build a standalone application for Core B of BF561 when using 21280the one-application-per-core programming model. Proper start files 21281and link scripts are used to support Core B, and the macro 21282@code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main} 21283should be used instead of @code{main}. 21284This option can only be used in conjunction with @option{-mmulticore}. 21285 21286@item -msdram 21287@opindex msdram 21288Build a standalone application for SDRAM. Proper start files and 21289link scripts are used to put the application into SDRAM, and the macro 21290@code{__BFIN_SDRAM} is defined. 21291The loader should initialize SDRAM before loading the application. 21292 21293@item -micplb 21294@opindex micplb 21295Assume that ICPLBs are enabled at run time. This has an effect on certain 21296anomaly workarounds. For Linux targets, the default is to assume ICPLBs 21297are enabled; for standalone applications the default is off. 21298@end table 21299 21300@node C6X Options 21301@subsection C6X Options 21302@cindex C6X Options 21303 21304@table @gcctabopt 21305@item -march=@var{name} 21306@opindex march 21307This specifies the name of the target architecture. GCC uses this 21308name to determine what kind of instructions it can emit when generating 21309assembly code. Permissible names are: @samp{c62x}, 21310@samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}. 21311 21312@item -mbig-endian 21313@opindex mbig-endian 21314Generate code for a big-endian target. 21315 21316@item -mlittle-endian 21317@opindex mlittle-endian 21318Generate code for a little-endian target. This is the default. 21319 21320@item -msim 21321@opindex msim 21322Choose startup files and linker script suitable for the simulator. 21323 21324@item -msdata=default 21325@opindex msdata=default 21326Put small global and static data in the @code{.neardata} section, 21327which is pointed to by register @code{B14}. Put small uninitialized 21328global and static data in the @code{.bss} section, which is adjacent 21329to the @code{.neardata} section. Put small read-only data into the 21330@code{.rodata} section. The corresponding sections used for large 21331pieces of data are @code{.fardata}, @code{.far} and @code{.const}. 21332 21333@item -msdata=all 21334@opindex msdata=all 21335Put all data, not just small objects, into the sections reserved for 21336small data, and use addressing relative to the @code{B14} register to 21337access them. 21338 21339@item -msdata=none 21340@opindex msdata=none 21341Make no use of the sections reserved for small data, and use absolute 21342addresses to access all data. Put all initialized global and static 21343data in the @code{.fardata} section, and all uninitialized data in the 21344@code{.far} section. Put all constant data into the @code{.const} 21345section. 21346@end table 21347 21348@node CRIS Options 21349@subsection CRIS Options 21350@cindex CRIS Options 21351 21352These options are defined specifically for the CRIS ports. 21353 21354@table @gcctabopt 21355@item -march=@var{architecture-type} 21356@itemx -mcpu=@var{architecture-type} 21357@opindex march 21358@opindex mcpu 21359Generate code for the specified architecture. The choices for 21360@var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for 21361respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@. 21362Default is @samp{v0} except for cris-axis-linux-gnu, where the default is 21363@samp{v10}. 21364 21365@item -mtune=@var{architecture-type} 21366@opindex mtune 21367Tune to @var{architecture-type} everything applicable about the generated 21368code, except for the ABI and the set of available instructions. The 21369choices for @var{architecture-type} are the same as for 21370@option{-march=@var{architecture-type}}. 21371 21372@item -mmax-stack-frame=@var{n} 21373@opindex mmax-stack-frame 21374Warn when the stack frame of a function exceeds @var{n} bytes. 21375 21376@item -metrax4 21377@itemx -metrax100 21378@opindex metrax4 21379@opindex metrax100 21380The options @option{-metrax4} and @option{-metrax100} are synonyms for 21381@option{-march=v3} and @option{-march=v8} respectively. 21382 21383@item -mmul-bug-workaround 21384@itemx -mno-mul-bug-workaround 21385@opindex mmul-bug-workaround 21386@opindex mno-mul-bug-workaround 21387Work around a bug in the @code{muls} and @code{mulu} instructions for CPU 21388models where it applies. This option is active by default. 21389 21390@item -mpdebug 21391@opindex mpdebug 21392Enable CRIS-specific verbose debug-related information in the assembly 21393code. This option also has the effect of turning off the @samp{#NO_APP} 21394formatted-code indicator to the assembler at the beginning of the 21395assembly file. 21396 21397@item -mcc-init 21398@opindex mcc-init 21399Do not use condition-code results from previous instruction; always emit 21400compare and test instructions before use of condition codes. 21401 21402@item -mno-side-effects 21403@opindex mno-side-effects 21404@opindex mside-effects 21405Do not emit instructions with side effects in addressing modes other than 21406post-increment. 21407 21408@item -mstack-align 21409@itemx -mno-stack-align 21410@itemx -mdata-align 21411@itemx -mno-data-align 21412@itemx -mconst-align 21413@itemx -mno-const-align 21414@opindex mstack-align 21415@opindex mno-stack-align 21416@opindex mdata-align 21417@opindex mno-data-align 21418@opindex mconst-align 21419@opindex mno-const-align 21420These options (@samp{no-} options) arrange (eliminate arrangements) for the 21421stack frame, individual data and constants to be aligned for the maximum 21422single data access size for the chosen CPU model. The default is to 21423arrange for 32-bit alignment. ABI details such as structure layout are 21424not affected by these options. 21425 21426@item -m32-bit 21427@itemx -m16-bit 21428@itemx -m8-bit 21429@opindex m32-bit 21430@opindex m16-bit 21431@opindex m8-bit 21432Similar to the stack- data- and const-align options above, these options 21433arrange for stack frame, writable data and constants to all be 32-bit, 2143416-bit or 8-bit aligned. The default is 32-bit alignment. 21435 21436@item -mno-prologue-epilogue 21437@itemx -mprologue-epilogue 21438@opindex mno-prologue-epilogue 21439@opindex mprologue-epilogue 21440With @option{-mno-prologue-epilogue}, the normal function prologue and 21441epilogue which set up the stack frame are omitted and no return 21442instructions or return sequences are generated in the code. Use this 21443option only together with visual inspection of the compiled code: no 21444warnings or errors are generated when call-saved registers must be saved, 21445or storage for local variables needs to be allocated. 21446 21447@item -mno-gotplt 21448@itemx -mgotplt 21449@opindex mno-gotplt 21450@opindex mgotplt 21451With @option{-fpic} and @option{-fPIC}, don't generate (do generate) 21452instruction sequences that load addresses for functions from the PLT part 21453of the GOT rather than (traditional on other architectures) calls to the 21454PLT@. The default is @option{-mgotplt}. 21455 21456@item -melf 21457@opindex melf 21458Legacy no-op option only recognized with the cris-axis-elf and 21459cris-axis-linux-gnu targets. 21460 21461@item -mlinux 21462@opindex mlinux 21463Legacy no-op option only recognized with the cris-axis-linux-gnu target. 21464 21465@item -sim 21466@opindex sim 21467This option, recognized for the cris-axis-elf, arranges 21468to link with input-output functions from a simulator library. Code, 21469initialized data and zero-initialized data are allocated consecutively. 21470 21471@item -sim2 21472@opindex sim2 21473Like @option{-sim}, but pass linker options to locate initialized data at 214740x40000000 and zero-initialized data at 0x80000000. 21475@end table 21476 21477@node CR16 Options 21478@subsection CR16 Options 21479@cindex CR16 Options 21480 21481These options are defined specifically for the CR16 ports. 21482 21483@table @gcctabopt 21484 21485@item -mmac 21486@opindex mmac 21487Enable the use of multiply-accumulate instructions. Disabled by default. 21488 21489@item -mcr16cplus 21490@itemx -mcr16c 21491@opindex mcr16cplus 21492@opindex mcr16c 21493Generate code for CR16C or CR16C+ architecture. CR16C+ architecture 21494is default. 21495 21496@item -msim 21497@opindex msim 21498Links the library libsim.a which is in compatible with simulator. Applicable 21499to ELF compiler only. 21500 21501@item -mint32 21502@opindex mint32 21503Choose integer type as 32-bit wide. 21504 21505@item -mbit-ops 21506@opindex mbit-ops 21507Generates @code{sbit}/@code{cbit} instructions for bit manipulations. 21508 21509@item -mdata-model=@var{model} 21510@opindex mdata-model 21511Choose a data model. The choices for @var{model} are @samp{near}, 21512@samp{far} or @samp{medium}. @samp{medium} is default. 21513However, @samp{far} is not valid with @option{-mcr16c}, as the 21514CR16C architecture does not support the far data model. 21515@end table 21516 21517@node C-SKY Options 21518@subsection C-SKY Options 21519@cindex C-SKY Options 21520 21521GCC supports these options when compiling for C-SKY V2 processors. 21522 21523@table @gcctabopt 21524 21525@item -march=@var{arch} 21526@opindex march= 21527Specify the C-SKY target architecture. Valid values for @var{arch} are: 21528@samp{ck801}, @samp{ck802}, @samp{ck803}, @samp{ck807}, and @samp{ck810}. 21529The default is @samp{ck810}. 21530 21531@item -mcpu=@var{cpu} 21532@opindex mcpu= 21533Specify the C-SKY target processor. Valid values for @var{cpu} are: 21534@samp{ck801}, @samp{ck801t}, 21535@samp{ck802}, @samp{ck802t}, @samp{ck802j}, 21536@samp{ck803}, @samp{ck803h}, @samp{ck803t}, @samp{ck803ht}, 21537@samp{ck803f}, @samp{ck803fh}, @samp{ck803e}, @samp{ck803eh}, 21538@samp{ck803et}, @samp{ck803eht}, @samp{ck803ef}, @samp{ck803efh}, 21539@samp{ck803ft}, @samp{ck803eft}, @samp{ck803efht}, @samp{ck803r1}, 21540@samp{ck803hr1}, @samp{ck803tr1}, @samp{ck803htr1}, @samp{ck803fr1}, 21541@samp{ck803fhr1}, @samp{ck803er1}, @samp{ck803ehr1}, @samp{ck803etr1}, 21542@samp{ck803ehtr1}, @samp{ck803efr1}, @samp{ck803efhr1}, @samp{ck803ftr1}, 21543@samp{ck803eftr1}, @samp{ck803efhtr1}, 21544@samp{ck803s}, @samp{ck803st}, @samp{ck803se}, @samp{ck803sf}, 21545@samp{ck803sef}, @samp{ck803seft}, 21546@samp{ck807e}, @samp{ck807ef}, @samp{ck807}, @samp{ck807f}, 21547@samp{ck810e}, @samp{ck810et}, @samp{ck810ef}, @samp{ck810eft}, 21548@samp{ck810}, @samp{ck810v}, @samp{ck810f}, @samp{ck810t}, @samp{ck810fv}, 21549@samp{ck810tv}, @samp{ck810ft}, and @samp{ck810ftv}. 21550 21551@item -mbig-endian 21552@opindex mbig-endian 21553@itemx -EB 21554@opindex EB 21555@itemx -mlittle-endian 21556@opindex mlittle-endian 21557@itemx -EL 21558@opindex EL 21559 21560Select big- or little-endian code. The default is little-endian. 21561 21562@item -mfloat-abi=@var{name} 21563@opindex mfloat-abi 21564Specifies which floating-point ABI to use. Permissible values 21565are: @samp{soft}, @samp{softfp} and @samp{hard}. 21566 21567Specifying @samp{soft} causes GCC to generate output containing 21568library calls for floating-point operations. 21569@samp{softfp} allows the generation of code using hardware floating-point 21570instructions, but still uses the soft-float calling conventions. 21571@samp{hard} allows generation of floating-point instructions 21572and uses FPU-specific calling conventions. 21573 21574The default depends on the specific target configuration. Note that 21575the hard-float and soft-float ABIs are not link-compatible; you must 21576compile your entire program with the same ABI, and link with a 21577compatible set of libraries. 21578 21579@item -mhard-float 21580@opindex mhard-float 21581@itemx -msoft-float 21582@opindex msoft-float 21583 21584Select hardware or software floating-point implementations. 21585The default is soft float. 21586 21587@item -mdouble-float 21588@itemx -mno-double-float 21589@opindex mdouble-float 21590When @option{-mhard-float} is in effect, enable generation of 21591double-precision float instructions. This is the default except 21592when compiling for CK803. 21593 21594@item -mfdivdu 21595@itemx -mno-fdivdu 21596@opindex mfdivdu 21597When @option{-mhard-float} is in effect, enable generation of 21598@code{frecipd}, @code{fsqrtd}, and @code{fdivd} instructions. 21599This is the default except when compiling for CK803. 21600 21601@item -mfpu=@var{fpu} 21602@opindex mfpu= 21603Select the floating-point processor. This option can only be used with 21604@option{-mhard-float}. 21605Values for @var{fpu} are 21606@samp{fpv2_sf} (equivalent to @samp{-mno-double-float -mno-fdivdu}), 21607@samp{fpv2} (@samp{-mdouble-float -mno-divdu}), and 21608@samp{fpv2_divd} (@samp{-mdouble-float -mdivdu}). 21609 21610@item -melrw 21611@itemx -mno-elrw 21612@opindex melrw 21613Enable the extended @code{lrw} instruction. This option defaults to on 21614for CK801 and off otherwise. 21615 21616@item -mistack 21617@itemx -mno-istack 21618@opindex mistack 21619Enable interrupt stack instructions; the default is off. 21620 21621The @option{-mistack} option is required to handle the 21622@code{interrupt} and @code{isr} function attributes 21623(@pxref{C-SKY Function Attributes}). 21624 21625@item -mmp 21626@opindex mmp 21627Enable multiprocessor instructions; the default is off. 21628 21629@item -mcp 21630@opindex mcp 21631Enable coprocessor instructions; the default is off. 21632 21633@item -mcache 21634@opindex mcache 21635Enable coprocessor instructions; the default is off. 21636 21637@item -msecurity 21638@opindex msecurity 21639Enable C-SKY security instructions; the default is off. 21640 21641@item -mtrust 21642@opindex mtrust 21643Enable C-SKY trust instructions; the default is off. 21644 21645@item -mdsp 21646@opindex mdsp 21647@itemx -medsp 21648@opindex medsp 21649@itemx -mvdsp 21650@opindex mvdsp 21651Enable C-SKY DSP, Enhanced DSP, or Vector DSP instructions, respectively. 21652All of these options default to off. 21653 21654@item -mdiv 21655@itemx -mno-div 21656@opindex mdiv 21657Generate divide instructions. Default is off. 21658 21659@item -msmart 21660@itemx -mno-smart 21661@opindex msmart 21662Generate code for Smart Mode, using only registers numbered 0-7 to allow 21663use of 16-bit instructions. This option is ignored for CK801 where this 21664is the required behavior, and it defaults to on for CK802. 21665For other targets, the default is off. 21666 21667@item -mhigh-registers 21668@itemx -mno-high-registers 21669@opindex mhigh-registers 21670Generate code using the high registers numbered 16-31. This option 21671is not supported on CK801, CK802, or CK803, and is enabled by default 21672for other processors. 21673 21674@item -manchor 21675@itemx -mno-anchor 21676@opindex manchor 21677Generate code using global anchor symbol addresses. 21678 21679@item -mpushpop 21680@itemx -mno-pushpop 21681@opindex mpushpop 21682Generate code using @code{push} and @code{pop} instructions. This option 21683defaults to on. 21684 21685@item -mmultiple-stld 21686@itemx -mstm 21687@itemx -mno-multiple-stld 21688@itemx -mno-stm 21689@opindex mmultiple-stld 21690Generate code using @code{stm} and @code{ldm} instructions. This option 21691isn't supported on CK801 but is enabled by default on other processors. 21692 21693@item -mconstpool 21694@itemx -mno-constpool 21695@opindex mconstpool 21696Create constant pools in the compiler instead of deferring it to the 21697assembler. This option is the default and required for correct code 21698generation on CK801 and CK802, and is optional on other processors. 21699 21700@item -mstack-size 21701@item -mno-stack-size 21702@opindex mstack-size 21703Emit @code{.stack_size} directives for each function in the assembly 21704output. This option defaults to off. 21705 21706@item -mccrt 21707@itemx -mno-ccrt 21708@opindex mccrt 21709Generate code for the C-SKY compiler runtime instead of libgcc. This 21710option defaults to off. 21711 21712@item -mbranch-cost=@var{n} 21713@opindex mbranch-cost= 21714Set the branch costs to roughly @code{n} instructions. The default is 1. 21715 21716@item -msched-prolog 21717@itemx -mno-sched-prolog 21718@opindex msched-prolog 21719Permit scheduling of function prologue and epilogue sequences. Using 21720this option can result in code that is not compliant with the C-SKY V2 ABI 21721prologue requirements and that cannot be debugged or backtraced. 21722It is disabled by default. 21723 21724@item -msim 21725@opindex msim 21726Links the library libsemi.a which is in compatible with simulator. Applicable 21727to ELF compiler only. 21728 21729@end table 21730 21731@node Darwin Options 21732@subsection Darwin Options 21733@cindex Darwin options 21734 21735These options are defined for all architectures running the Darwin operating 21736system. 21737 21738FSF GCC on Darwin does not create ``fat'' object files; it creates 21739an object file for the single architecture that GCC was built to 21740target. Apple's GCC on Darwin does create ``fat'' files if multiple 21741@option{-arch} options are used; it does so by running the compiler or 21742linker multiple times and joining the results together with 21743@file{lipo}. 21744 21745The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or 21746@samp{i686}) is determined by the flags that specify the ISA 21747that GCC is targeting, like @option{-mcpu} or @option{-march}. The 21748@option{-force_cpusubtype_ALL} option can be used to override this. 21749 21750The Darwin tools vary in their behavior when presented with an ISA 21751mismatch. The assembler, @file{as}, only permits instructions to 21752be used that are valid for the subtype of the file it is generating, 21753so you cannot put 64-bit instructions in a @samp{ppc750} object file. 21754The linker for shared libraries, @file{/usr/bin/libtool}, fails 21755and prints an error if asked to create a shared library with a less 21756restrictive subtype than its input files (for instance, trying to put 21757a @samp{ppc970} object file in a @samp{ppc7400} library). The linker 21758for executables, @command{ld}, quietly gives the executable the most 21759restrictive subtype of any of its input files. 21760 21761@table @gcctabopt 21762@item -F@var{dir} 21763@opindex F 21764Add the framework directory @var{dir} to the head of the list of 21765directories to be searched for header files. These directories are 21766interleaved with those specified by @option{-I} options and are 21767scanned in a left-to-right order. 21768 21769A framework directory is a directory with frameworks in it. A 21770framework is a directory with a @file{Headers} and/or 21771@file{PrivateHeaders} directory contained directly in it that ends 21772in @file{.framework}. The name of a framework is the name of this 21773directory excluding the @file{.framework}. Headers associated with 21774the framework are found in one of those two directories, with 21775@file{Headers} being searched first. A subframework is a framework 21776directory that is in a framework's @file{Frameworks} directory. 21777Includes of subframework headers can only appear in a header of a 21778framework that contains the subframework, or in a sibling subframework 21779header. Two subframeworks are siblings if they occur in the same 21780framework. A subframework should not have the same name as a 21781framework; a warning is issued if this is violated. Currently a 21782subframework cannot have subframeworks; in the future, the mechanism 21783may be extended to support this. The standard frameworks can be found 21784in @file{/System/Library/Frameworks} and 21785@file{/Library/Frameworks}. An example include looks like 21786@code{#include <Framework/header.h>}, where @file{Framework} denotes 21787the name of the framework and @file{header.h} is found in the 21788@file{PrivateHeaders} or @file{Headers} directory. 21789 21790@item -iframework@var{dir} 21791@opindex iframework 21792Like @option{-F} except the directory is a treated as a system 21793directory. The main difference between this @option{-iframework} and 21794@option{-F} is that with @option{-iframework} the compiler does not 21795warn about constructs contained within header files found via 21796@var{dir}. This option is valid only for the C family of languages. 21797 21798@item -gused 21799@opindex gused 21800Emit debugging information for symbols that are used. For stabs 21801debugging format, this enables @option{-feliminate-unused-debug-symbols}. 21802This is by default ON@. 21803 21804@item -gfull 21805@opindex gfull 21806Emit debugging information for all symbols and types. 21807 21808@item -mmacosx-version-min=@var{version} 21809The earliest version of MacOS X that this executable will run on 21810is @var{version}. Typical values of @var{version} include @code{10.1}, 21811@code{10.2}, and @code{10.3.9}. 21812 21813If the compiler was built to use the system's headers by default, 21814then the default for this option is the system version on which the 21815compiler is running, otherwise the default is to make choices that 21816are compatible with as many systems and code bases as possible. 21817 21818@item -mkernel 21819@opindex mkernel 21820Enable kernel development mode. The @option{-mkernel} option sets 21821@option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit}, 21822@option{-fno-exceptions}, @option{-fno-non-call-exceptions}, 21823@option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where 21824applicable. This mode also sets @option{-mno-altivec}, 21825@option{-msoft-float}, @option{-fno-builtin} and 21826@option{-mlong-branch} for PowerPC targets. 21827 21828@item -mone-byte-bool 21829@opindex mone-byte-bool 21830Override the defaults for @code{bool} so that @code{sizeof(bool)==1}. 21831By default @code{sizeof(bool)} is @code{4} when compiling for 21832Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this 21833option has no effect on x86. 21834 21835@strong{Warning:} The @option{-mone-byte-bool} switch causes GCC 21836to generate code that is not binary compatible with code generated 21837without that switch. Using this switch may require recompiling all 21838other modules in a program, including system libraries. Use this 21839switch to conform to a non-default data model. 21840 21841@item -mfix-and-continue 21842@itemx -ffix-and-continue 21843@itemx -findirect-data 21844@opindex mfix-and-continue 21845@opindex ffix-and-continue 21846@opindex findirect-data 21847Generate code suitable for fast turnaround development, such as to 21848allow GDB to dynamically load @file{.o} files into already-running 21849programs. @option{-findirect-data} and @option{-ffix-and-continue} 21850are provided for backwards compatibility. 21851 21852@item -all_load 21853@opindex all_load 21854Loads all members of static archive libraries. 21855See man ld(1) for more information. 21856 21857@item -arch_errors_fatal 21858@opindex arch_errors_fatal 21859Cause the errors having to do with files that have the wrong architecture 21860to be fatal. 21861 21862@item -bind_at_load 21863@opindex bind_at_load 21864Causes the output file to be marked such that the dynamic linker will 21865bind all undefined references when the file is loaded or launched. 21866 21867@item -bundle 21868@opindex bundle 21869Produce a Mach-o bundle format file. 21870See man ld(1) for more information. 21871 21872@item -bundle_loader @var{executable} 21873@opindex bundle_loader 21874This option specifies the @var{executable} that will load the build 21875output file being linked. See man ld(1) for more information. 21876 21877@item -dynamiclib 21878@opindex dynamiclib 21879When passed this option, GCC produces a dynamic library instead of 21880an executable when linking, using the Darwin @file{libtool} command. 21881 21882@item -force_cpusubtype_ALL 21883@opindex force_cpusubtype_ALL 21884This causes GCC's output file to have the @samp{ALL} subtype, instead of 21885one controlled by the @option{-mcpu} or @option{-march} option. 21886 21887@item -allowable_client @var{client_name} 21888@itemx -client_name 21889@itemx -compatibility_version 21890@itemx -current_version 21891@itemx -dead_strip 21892@itemx -dependency-file 21893@itemx -dylib_file 21894@itemx -dylinker_install_name 21895@itemx -dynamic 21896@itemx -exported_symbols_list 21897@itemx -filelist 21898@need 800 21899@itemx -flat_namespace 21900@itemx -force_flat_namespace 21901@itemx -headerpad_max_install_names 21902@itemx -image_base 21903@itemx -init 21904@itemx -install_name 21905@itemx -keep_private_externs 21906@itemx -multi_module 21907@itemx -multiply_defined 21908@itemx -multiply_defined_unused 21909@need 800 21910@itemx -noall_load 21911@itemx -no_dead_strip_inits_and_terms 21912@itemx -nofixprebinding 21913@itemx -nomultidefs 21914@itemx -noprebind 21915@itemx -noseglinkedit 21916@itemx -pagezero_size 21917@itemx -prebind 21918@itemx -prebind_all_twolevel_modules 21919@itemx -private_bundle 21920@need 800 21921@itemx -read_only_relocs 21922@itemx -sectalign 21923@itemx -sectobjectsymbols 21924@itemx -whyload 21925@itemx -seg1addr 21926@itemx -sectcreate 21927@itemx -sectobjectsymbols 21928@itemx -sectorder 21929@itemx -segaddr 21930@itemx -segs_read_only_addr 21931@need 800 21932@itemx -segs_read_write_addr 21933@itemx -seg_addr_table 21934@itemx -seg_addr_table_filename 21935@itemx -seglinkedit 21936@itemx -segprot 21937@itemx -segs_read_only_addr 21938@itemx -segs_read_write_addr 21939@itemx -single_module 21940@itemx -static 21941@itemx -sub_library 21942@need 800 21943@itemx -sub_umbrella 21944@itemx -twolevel_namespace 21945@itemx -umbrella 21946@itemx -undefined 21947@itemx -unexported_symbols_list 21948@itemx -weak_reference_mismatches 21949@itemx -whatsloaded 21950@opindex allowable_client 21951@opindex client_name 21952@opindex compatibility_version 21953@opindex current_version 21954@opindex dead_strip 21955@opindex dependency-file 21956@opindex dylib_file 21957@opindex dylinker_install_name 21958@opindex dynamic 21959@opindex exported_symbols_list 21960@opindex filelist 21961@opindex flat_namespace 21962@opindex force_flat_namespace 21963@opindex headerpad_max_install_names 21964@opindex image_base 21965@opindex init 21966@opindex install_name 21967@opindex keep_private_externs 21968@opindex multi_module 21969@opindex multiply_defined 21970@opindex multiply_defined_unused 21971@opindex noall_load 21972@opindex no_dead_strip_inits_and_terms 21973@opindex nofixprebinding 21974@opindex nomultidefs 21975@opindex noprebind 21976@opindex noseglinkedit 21977@opindex pagezero_size 21978@opindex prebind 21979@opindex prebind_all_twolevel_modules 21980@opindex private_bundle 21981@opindex read_only_relocs 21982@opindex sectalign 21983@opindex sectobjectsymbols 21984@opindex whyload 21985@opindex seg1addr 21986@opindex sectcreate 21987@opindex sectobjectsymbols 21988@opindex sectorder 21989@opindex segaddr 21990@opindex segs_read_only_addr 21991@opindex segs_read_write_addr 21992@opindex seg_addr_table 21993@opindex seg_addr_table_filename 21994@opindex seglinkedit 21995@opindex segprot 21996@opindex segs_read_only_addr 21997@opindex segs_read_write_addr 21998@opindex single_module 21999@opindex static 22000@opindex sub_library 22001@opindex sub_umbrella 22002@opindex twolevel_namespace 22003@opindex umbrella 22004@opindex undefined 22005@opindex unexported_symbols_list 22006@opindex weak_reference_mismatches 22007@opindex whatsloaded 22008These options are passed to the Darwin linker. The Darwin linker man page 22009describes them in detail. 22010@end table 22011 22012@node DEC Alpha Options 22013@subsection DEC Alpha Options 22014 22015These @samp{-m} options are defined for the DEC Alpha implementations: 22016 22017@table @gcctabopt 22018@item -mno-soft-float 22019@itemx -msoft-float 22020@opindex mno-soft-float 22021@opindex msoft-float 22022Use (do not use) the hardware floating-point instructions for 22023floating-point operations. When @option{-msoft-float} is specified, 22024functions in @file{libgcc.a} are used to perform floating-point 22025operations. Unless they are replaced by routines that emulate the 22026floating-point operations, or compiled in such a way as to call such 22027emulations routines, these routines issue floating-point 22028operations. If you are compiling for an Alpha without floating-point 22029operations, you must ensure that the library is built so as not to call 22030them. 22031 22032Note that Alpha implementations without floating-point operations are 22033required to have floating-point registers. 22034 22035@item -mfp-reg 22036@itemx -mno-fp-regs 22037@opindex mfp-reg 22038@opindex mno-fp-regs 22039Generate code that uses (does not use) the floating-point register set. 22040@option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point 22041register set is not used, floating-point operands are passed in integer 22042registers as if they were integers and floating-point results are passed 22043in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence, 22044so any function with a floating-point argument or return value called by code 22045compiled with @option{-mno-fp-regs} must also be compiled with that 22046option. 22047 22048A typical use of this option is building a kernel that does not use, 22049and hence need not save and restore, any floating-point registers. 22050 22051@item -mieee 22052@opindex mieee 22053The Alpha architecture implements floating-point hardware optimized for 22054maximum performance. It is mostly compliant with the IEEE floating-point 22055standard. However, for full compliance, software assistance is 22056required. This option generates code fully IEEE-compliant code 22057@emph{except} that the @var{inexact-flag} is not maintained (see below). 22058If this option is turned on, the preprocessor macro @code{_IEEE_FP} is 22059defined during compilation. The resulting code is less efficient but is 22060able to correctly support denormalized numbers and exceptional IEEE 22061values such as not-a-number and plus/minus infinity. Other Alpha 22062compilers call this option @option{-ieee_with_no_inexact}. 22063 22064@item -mieee-with-inexact 22065@opindex mieee-with-inexact 22066This is like @option{-mieee} except the generated code also maintains 22067the IEEE @var{inexact-flag}. Turning on this option causes the 22068generated code to implement fully-compliant IEEE math. In addition to 22069@code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor 22070macro. On some Alpha implementations the resulting code may execute 22071significantly slower than the code generated by default. Since there is 22072very little code that depends on the @var{inexact-flag}, you should 22073normally not specify this option. Other Alpha compilers call this 22074option @option{-ieee_with_inexact}. 22075 22076@item -mfp-trap-mode=@var{trap-mode} 22077@opindex mfp-trap-mode 22078This option controls what floating-point related traps are enabled. 22079Other Alpha compilers call this option @option{-fptm @var{trap-mode}}. 22080The trap mode can be set to one of four values: 22081 22082@table @samp 22083@item n 22084This is the default (normal) setting. The only traps that are enabled 22085are the ones that cannot be disabled in software (e.g., division by zero 22086trap). 22087 22088@item u 22089In addition to the traps enabled by @samp{n}, underflow traps are enabled 22090as well. 22091 22092@item su 22093Like @samp{u}, but the instructions are marked to be safe for software 22094completion (see Alpha architecture manual for details). 22095 22096@item sui 22097Like @samp{su}, but inexact traps are enabled as well. 22098@end table 22099 22100@item -mfp-rounding-mode=@var{rounding-mode} 22101@opindex mfp-rounding-mode 22102Selects the IEEE rounding mode. Other Alpha compilers call this option 22103@option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one 22104of: 22105 22106@table @samp 22107@item n 22108Normal IEEE rounding mode. Floating-point numbers are rounded towards 22109the nearest machine number or towards the even machine number in case 22110of a tie. 22111 22112@item m 22113Round towards minus infinity. 22114 22115@item c 22116Chopped rounding mode. Floating-point numbers are rounded towards zero. 22117 22118@item d 22119Dynamic rounding mode. A field in the floating-point control register 22120(@var{fpcr}, see Alpha architecture reference manual) controls the 22121rounding mode in effect. The C library initializes this register for 22122rounding towards plus infinity. Thus, unless your program modifies the 22123@var{fpcr}, @samp{d} corresponds to round towards plus infinity. 22124@end table 22125 22126@item -mtrap-precision=@var{trap-precision} 22127@opindex mtrap-precision 22128In the Alpha architecture, floating-point traps are imprecise. This 22129means without software assistance it is impossible to recover from a 22130floating trap and program execution normally needs to be terminated. 22131GCC can generate code that can assist operating system trap handlers 22132in determining the exact location that caused a floating-point trap. 22133Depending on the requirements of an application, different levels of 22134precisions can be selected: 22135 22136@table @samp 22137@item p 22138Program precision. This option is the default and means a trap handler 22139can only identify which program caused a floating-point exception. 22140 22141@item f 22142Function precision. The trap handler can determine the function that 22143caused a floating-point exception. 22144 22145@item i 22146Instruction precision. The trap handler can determine the exact 22147instruction that caused a floating-point exception. 22148@end table 22149 22150Other Alpha compilers provide the equivalent options called 22151@option{-scope_safe} and @option{-resumption_safe}. 22152 22153@item -mieee-conformant 22154@opindex mieee-conformant 22155This option marks the generated code as IEEE conformant. You must not 22156use this option unless you also specify @option{-mtrap-precision=i} and either 22157@option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect 22158is to emit the line @samp{.eflag 48} in the function prologue of the 22159generated assembly file. 22160 22161@item -mbuild-constants 22162@opindex mbuild-constants 22163Normally GCC examines a 32- or 64-bit integer constant to 22164see if it can construct it from smaller constants in two or three 22165instructions. If it cannot, it outputs the constant as a literal and 22166generates code to load it from the data segment at run time. 22167 22168Use this option to require GCC to construct @emph{all} integer constants 22169using code, even if it takes more instructions (the maximum is six). 22170 22171You typically use this option to build a shared library dynamic 22172loader. Itself a shared library, it must relocate itself in memory 22173before it can find the variables and constants in its own data segment. 22174 22175@item -mbwx 22176@itemx -mno-bwx 22177@itemx -mcix 22178@itemx -mno-cix 22179@itemx -mfix 22180@itemx -mno-fix 22181@itemx -mmax 22182@itemx -mno-max 22183@opindex mbwx 22184@opindex mno-bwx 22185@opindex mcix 22186@opindex mno-cix 22187@opindex mfix 22188@opindex mno-fix 22189@opindex mmax 22190@opindex mno-max 22191Indicate whether GCC should generate code to use the optional BWX, 22192CIX, FIX and MAX instruction sets. The default is to use the instruction 22193sets supported by the CPU type specified via @option{-mcpu=} option or that 22194of the CPU on which GCC was built if none is specified. 22195 22196@item -mfloat-vax 22197@itemx -mfloat-ieee 22198@opindex mfloat-vax 22199@opindex mfloat-ieee 22200Generate code that uses (does not use) VAX F and G floating-point 22201arithmetic instead of IEEE single and double precision. 22202 22203@item -mexplicit-relocs 22204@itemx -mno-explicit-relocs 22205@opindex mexplicit-relocs 22206@opindex mno-explicit-relocs 22207Older Alpha assemblers provided no way to generate symbol relocations 22208except via assembler macros. Use of these macros does not allow 22209optimal instruction scheduling. GNU binutils as of version 2.12 22210supports a new syntax that allows the compiler to explicitly mark 22211which relocations should apply to which instructions. This option 22212is mostly useful for debugging, as GCC detects the capabilities of 22213the assembler when it is built and sets the default accordingly. 22214 22215@item -msmall-data 22216@itemx -mlarge-data 22217@opindex msmall-data 22218@opindex mlarge-data 22219When @option{-mexplicit-relocs} is in effect, static data is 22220accessed via @dfn{gp-relative} relocations. When @option{-msmall-data} 22221is used, objects 8 bytes long or smaller are placed in a @dfn{small data area} 22222(the @code{.sdata} and @code{.sbss} sections) and are accessed via 2222316-bit relocations off of the @code{$gp} register. This limits the 22224size of the small data area to 64KB, but allows the variables to be 22225directly accessed via a single instruction. 22226 22227The default is @option{-mlarge-data}. With this option the data area 22228is limited to just below 2GB@. Programs that require more than 2GB of 22229data must use @code{malloc} or @code{mmap} to allocate the data in the 22230heap instead of in the program's data segment. 22231 22232When generating code for shared libraries, @option{-fpic} implies 22233@option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}. 22234 22235@item -msmall-text 22236@itemx -mlarge-text 22237@opindex msmall-text 22238@opindex mlarge-text 22239When @option{-msmall-text} is used, the compiler assumes that the 22240code of the entire program (or shared library) fits in 4MB, and is 22241thus reachable with a branch instruction. When @option{-msmall-data} 22242is used, the compiler can assume that all local symbols share the 22243same @code{$gp} value, and thus reduce the number of instructions 22244required for a function call from 4 to 1. 22245 22246The default is @option{-mlarge-text}. 22247 22248@item -mcpu=@var{cpu_type} 22249@opindex mcpu 22250Set the instruction set and instruction scheduling parameters for 22251machine type @var{cpu_type}. You can specify either the @samp{EV} 22252style name or the corresponding chip number. GCC supports scheduling 22253parameters for the EV4, EV5 and EV6 family of processors and 22254chooses the default values for the instruction set from the processor 22255you specify. If you do not specify a processor type, GCC defaults 22256to the processor on which the compiler was built. 22257 22258Supported values for @var{cpu_type} are 22259 22260@table @samp 22261@item ev4 22262@itemx ev45 22263@itemx 21064 22264Schedules as an EV4 and has no instruction set extensions. 22265 22266@item ev5 22267@itemx 21164 22268Schedules as an EV5 and has no instruction set extensions. 22269 22270@item ev56 22271@itemx 21164a 22272Schedules as an EV5 and supports the BWX extension. 22273 22274@item pca56 22275@itemx 21164pc 22276@itemx 21164PC 22277Schedules as an EV5 and supports the BWX and MAX extensions. 22278 22279@item ev6 22280@itemx 21264 22281Schedules as an EV6 and supports the BWX, FIX, and MAX extensions. 22282 22283@item ev67 22284@itemx 21264a 22285Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions. 22286@end table 22287 22288Native toolchains also support the value @samp{native}, 22289which selects the best architecture option for the host processor. 22290@option{-mcpu=native} has no effect if GCC does not recognize 22291the processor. 22292 22293@item -mtune=@var{cpu_type} 22294@opindex mtune 22295Set only the instruction scheduling parameters for machine type 22296@var{cpu_type}. The instruction set is not changed. 22297 22298Native toolchains also support the value @samp{native}, 22299which selects the best architecture option for the host processor. 22300@option{-mtune=native} has no effect if GCC does not recognize 22301the processor. 22302 22303@item -mmemory-latency=@var{time} 22304@opindex mmemory-latency 22305Sets the latency the scheduler should assume for typical memory 22306references as seen by the application. This number is highly 22307dependent on the memory access patterns used by the application 22308and the size of the external cache on the machine. 22309 22310Valid options for @var{time} are 22311 22312@table @samp 22313@item @var{number} 22314A decimal number representing clock cycles. 22315 22316@item L1 22317@itemx L2 22318@itemx L3 22319@itemx main 22320The compiler contains estimates of the number of clock cycles for 22321``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches 22322(also called Dcache, Scache, and Bcache), as well as to main memory. 22323Note that L3 is only valid for EV5. 22324 22325@end table 22326@end table 22327 22328@node eBPF Options 22329@subsection eBPF Options 22330@cindex eBPF Options 22331 22332@table @gcctabopt 22333@item -mframe-limit=@var{bytes} 22334This specifies the hard limit for frame sizes, in bytes. Currently, 22335the value that can be specified should be less than or equal to 22336@samp{32767}. Defaults to whatever limit is imposed by the version of 22337the Linux kernel targeted. 22338 22339@item -mkernel=@var{version} 22340@opindex mkernel 22341This specifies the minimum version of the kernel that will run the 22342compiled program. GCC uses this version to determine which 22343instructions to use, what kernel helpers to allow, etc. Currently, 22344@var{version} can be one of @samp{4.0}, @samp{4.1}, @samp{4.2}, 22345@samp{4.3}, @samp{4.4}, @samp{4.5}, @samp{4.6}, @samp{4.7}, 22346@samp{4.8}, @samp{4.9}, @samp{4.10}, @samp{4.11}, @samp{4.12}, 22347@samp{4.13}, @samp{4.14}, @samp{4.15}, @samp{4.16}, @samp{4.17}, 22348@samp{4.18}, @samp{4.19}, @samp{4.20}, @samp{5.0}, @samp{5.1}, 22349@samp{5.2}, @samp{latest} and @samp{native}. 22350 22351@item -mbig-endian 22352@opindex mbig-endian 22353Generate code for a big-endian target. 22354 22355@item -mlittle-endian 22356@opindex mlittle-endian 22357Generate code for a little-endian target. This is the default. 22358 22359@item -mxbpf 22360Generate code for an expanded version of BPF, which relaxes some of 22361the restrictions imposed by the BPF architecture: 22362@itemize @minus 22363@item Save and restore callee-saved registers at function entry and 22364exit, respectively. 22365@end itemize 22366@end table 22367 22368@node FR30 Options 22369@subsection FR30 Options 22370@cindex FR30 Options 22371 22372These options are defined specifically for the FR30 port. 22373 22374@table @gcctabopt 22375 22376@item -msmall-model 22377@opindex msmall-model 22378Use the small address space model. This can produce smaller code, but 22379it does assume that all symbolic values and addresses fit into a 2238020-bit range. 22381 22382@item -mno-lsim 22383@opindex mno-lsim 22384Assume that runtime support has been provided and so there is no need 22385to include the simulator library (@file{libsim.a}) on the linker 22386command line. 22387 22388@end table 22389 22390@node FT32 Options 22391@subsection FT32 Options 22392@cindex FT32 Options 22393 22394These options are defined specifically for the FT32 port. 22395 22396@table @gcctabopt 22397 22398@item -msim 22399@opindex msim 22400Specifies that the program will be run on the simulator. This causes 22401an alternate runtime startup and library to be linked. 22402You must not use this option when generating programs that will run on 22403real hardware; you must provide your own runtime library for whatever 22404I/O functions are needed. 22405 22406@item -mlra 22407@opindex mlra 22408Enable Local Register Allocation. This is still experimental for FT32, 22409so by default the compiler uses standard reload. 22410 22411@item -mnodiv 22412@opindex mnodiv 22413Do not use div and mod instructions. 22414 22415@item -mft32b 22416@opindex mft32b 22417Enable use of the extended instructions of the FT32B processor. 22418 22419@item -mcompress 22420@opindex mcompress 22421Compress all code using the Ft32B code compression scheme. 22422 22423@item -mnopm 22424@opindex mnopm 22425Do not generate code that reads program memory. 22426 22427@end table 22428 22429@node FRV Options 22430@subsection FRV Options 22431@cindex FRV Options 22432 22433@table @gcctabopt 22434@item -mgpr-32 22435@opindex mgpr-32 22436 22437Only use the first 32 general-purpose registers. 22438 22439@item -mgpr-64 22440@opindex mgpr-64 22441 22442Use all 64 general-purpose registers. 22443 22444@item -mfpr-32 22445@opindex mfpr-32 22446 22447Use only the first 32 floating-point registers. 22448 22449@item -mfpr-64 22450@opindex mfpr-64 22451 22452Use all 64 floating-point registers. 22453 22454@item -mhard-float 22455@opindex mhard-float 22456 22457Use hardware instructions for floating-point operations. 22458 22459@item -msoft-float 22460@opindex msoft-float 22461 22462Use library routines for floating-point operations. 22463 22464@item -malloc-cc 22465@opindex malloc-cc 22466 22467Dynamically allocate condition code registers. 22468 22469@item -mfixed-cc 22470@opindex mfixed-cc 22471 22472Do not try to dynamically allocate condition code registers, only 22473use @code{icc0} and @code{fcc0}. 22474 22475@item -mdword 22476@opindex mdword 22477 22478Change ABI to use double word insns. 22479 22480@item -mno-dword 22481@opindex mno-dword 22482@opindex mdword 22483 22484Do not use double word instructions. 22485 22486@item -mdouble 22487@opindex mdouble 22488 22489Use floating-point double instructions. 22490 22491@item -mno-double 22492@opindex mno-double 22493 22494Do not use floating-point double instructions. 22495 22496@item -mmedia 22497@opindex mmedia 22498 22499Use media instructions. 22500 22501@item -mno-media 22502@opindex mno-media 22503 22504Do not use media instructions. 22505 22506@item -mmuladd 22507@opindex mmuladd 22508 22509Use multiply and add/subtract instructions. 22510 22511@item -mno-muladd 22512@opindex mno-muladd 22513 22514Do not use multiply and add/subtract instructions. 22515 22516@item -mfdpic 22517@opindex mfdpic 22518 22519Select the FDPIC ABI, which uses function descriptors to represent 22520pointers to functions. Without any PIC/PIE-related options, it 22521implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it 22522assumes GOT entries and small data are within a 12-bit range from the 22523GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets 22524are computed with 32 bits. 22525With a @samp{bfin-elf} target, this option implies @option{-msim}. 22526 22527@item -minline-plt 22528@opindex minline-plt 22529 22530Enable inlining of PLT entries in function calls to functions that are 22531not known to bind locally. It has no effect without @option{-mfdpic}. 22532It's enabled by default if optimizing for speed and compiling for 22533shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an 22534optimization option such as @option{-O3} or above is present in the 22535command line. 22536 22537@item -mTLS 22538@opindex mTLS 22539 22540Assume a large TLS segment when generating thread-local code. 22541 22542@item -mtls 22543@opindex mtls 22544 22545Do not assume a large TLS segment when generating thread-local code. 22546 22547@item -mgprel-ro 22548@opindex mgprel-ro 22549 22550Enable the use of @code{GPREL} relocations in the FDPIC ABI for data 22551that is known to be in read-only sections. It's enabled by default, 22552except for @option{-fpic} or @option{-fpie}: even though it may help 22553make the global offset table smaller, it trades 1 instruction for 4. 22554With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4, 22555one of which may be shared by multiple symbols, and it avoids the need 22556for a GOT entry for the referenced symbol, so it's more likely to be a 22557win. If it is not, @option{-mno-gprel-ro} can be used to disable it. 22558 22559@item -multilib-library-pic 22560@opindex multilib-library-pic 22561 22562Link with the (library, not FD) pic libraries. It's implied by 22563@option{-mlibrary-pic}, as well as by @option{-fPIC} and 22564@option{-fpic} without @option{-mfdpic}. You should never have to use 22565it explicitly. 22566 22567@item -mlinked-fp 22568@opindex mlinked-fp 22569 22570Follow the EABI requirement of always creating a frame pointer whenever 22571a stack frame is allocated. This option is enabled by default and can 22572be disabled with @option{-mno-linked-fp}. 22573 22574@item -mlong-calls 22575@opindex mlong-calls 22576 22577Use indirect addressing to call functions outside the current 22578compilation unit. This allows the functions to be placed anywhere 22579within the 32-bit address space. 22580 22581@item -malign-labels 22582@opindex malign-labels 22583 22584Try to align labels to an 8-byte boundary by inserting NOPs into the 22585previous packet. This option only has an effect when VLIW packing 22586is enabled. It doesn't create new packets; it merely adds NOPs to 22587existing ones. 22588 22589@item -mlibrary-pic 22590@opindex mlibrary-pic 22591 22592Generate position-independent EABI code. 22593 22594@item -macc-4 22595@opindex macc-4 22596 22597Use only the first four media accumulator registers. 22598 22599@item -macc-8 22600@opindex macc-8 22601 22602Use all eight media accumulator registers. 22603 22604@item -mpack 22605@opindex mpack 22606 22607Pack VLIW instructions. 22608 22609@item -mno-pack 22610@opindex mno-pack 22611 22612Do not pack VLIW instructions. 22613 22614@item -mno-eflags 22615@opindex mno-eflags 22616 22617Do not mark ABI switches in e_flags. 22618 22619@item -mcond-move 22620@opindex mcond-move 22621 22622Enable the use of conditional-move instructions (default). 22623 22624This switch is mainly for debugging the compiler and will likely be removed 22625in a future version. 22626 22627@item -mno-cond-move 22628@opindex mno-cond-move 22629 22630Disable the use of conditional-move instructions. 22631 22632This switch is mainly for debugging the compiler and will likely be removed 22633in a future version. 22634 22635@item -mscc 22636@opindex mscc 22637 22638Enable the use of conditional set instructions (default). 22639 22640This switch is mainly for debugging the compiler and will likely be removed 22641in a future version. 22642 22643@item -mno-scc 22644@opindex mno-scc 22645 22646Disable the use of conditional set instructions. 22647 22648This switch is mainly for debugging the compiler and will likely be removed 22649in a future version. 22650 22651@item -mcond-exec 22652@opindex mcond-exec 22653 22654Enable the use of conditional execution (default). 22655 22656This switch is mainly for debugging the compiler and will likely be removed 22657in a future version. 22658 22659@item -mno-cond-exec 22660@opindex mno-cond-exec 22661 22662Disable the use of conditional execution. 22663 22664This switch is mainly for debugging the compiler and will likely be removed 22665in a future version. 22666 22667@item -mvliw-branch 22668@opindex mvliw-branch 22669 22670Run a pass to pack branches into VLIW instructions (default). 22671 22672This switch is mainly for debugging the compiler and will likely be removed 22673in a future version. 22674 22675@item -mno-vliw-branch 22676@opindex mno-vliw-branch 22677 22678Do not run a pass to pack branches into VLIW instructions. 22679 22680This switch is mainly for debugging the compiler and will likely be removed 22681in a future version. 22682 22683@item -mmulti-cond-exec 22684@opindex mmulti-cond-exec 22685 22686Enable optimization of @code{&&} and @code{||} in conditional execution 22687(default). 22688 22689This switch is mainly for debugging the compiler and will likely be removed 22690in a future version. 22691 22692@item -mno-multi-cond-exec 22693@opindex mno-multi-cond-exec 22694 22695Disable optimization of @code{&&} and @code{||} in conditional execution. 22696 22697This switch is mainly for debugging the compiler and will likely be removed 22698in a future version. 22699 22700@item -mnested-cond-exec 22701@opindex mnested-cond-exec 22702 22703Enable nested conditional execution optimizations (default). 22704 22705This switch is mainly for debugging the compiler and will likely be removed 22706in a future version. 22707 22708@item -mno-nested-cond-exec 22709@opindex mno-nested-cond-exec 22710 22711Disable nested conditional execution optimizations. 22712 22713This switch is mainly for debugging the compiler and will likely be removed 22714in a future version. 22715 22716@item -moptimize-membar 22717@opindex moptimize-membar 22718 22719This switch removes redundant @code{membar} instructions from the 22720compiler-generated code. It is enabled by default. 22721 22722@item -mno-optimize-membar 22723@opindex mno-optimize-membar 22724@opindex moptimize-membar 22725 22726This switch disables the automatic removal of redundant @code{membar} 22727instructions from the generated code. 22728 22729@item -mtomcat-stats 22730@opindex mtomcat-stats 22731 22732Cause gas to print out tomcat statistics. 22733 22734@item -mcpu=@var{cpu} 22735@opindex mcpu 22736 22737Select the processor type for which to generate code. Possible values are 22738@samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450}, 22739@samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}. 22740 22741@end table 22742 22743@node GNU/Linux Options 22744@subsection GNU/Linux Options 22745 22746These @samp{-m} options are defined for GNU/Linux targets: 22747 22748@table @gcctabopt 22749@item -mglibc 22750@opindex mglibc 22751Use the GNU C library. This is the default except 22752on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and 22753@samp{*-*-linux-*android*} targets. 22754 22755@item -muclibc 22756@opindex muclibc 22757Use uClibc C library. This is the default on 22758@samp{*-*-linux-*uclibc*} targets. 22759 22760@item -mmusl 22761@opindex mmusl 22762Use the musl C library. This is the default on 22763@samp{*-*-linux-*musl*} targets. 22764 22765@item -mbionic 22766@opindex mbionic 22767Use Bionic C library. This is the default on 22768@samp{*-*-linux-*android*} targets. 22769 22770@item -mandroid 22771@opindex mandroid 22772Compile code compatible with Android platform. This is the default on 22773@samp{*-*-linux-*android*} targets. 22774 22775When compiling, this option enables @option{-mbionic}, @option{-fPIC}, 22776@option{-fno-exceptions} and @option{-fno-rtti} by default. When linking, 22777this option makes the GCC driver pass Android-specific options to the linker. 22778Finally, this option causes the preprocessor macro @code{__ANDROID__} 22779to be defined. 22780 22781@item -tno-android-cc 22782@opindex tno-android-cc 22783Disable compilation effects of @option{-mandroid}, i.e., do not enable 22784@option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and 22785@option{-fno-rtti} by default. 22786 22787@item -tno-android-ld 22788@opindex tno-android-ld 22789Disable linking effects of @option{-mandroid}, i.e., pass standard Linux 22790linking options to the linker. 22791 22792@end table 22793 22794@node H8/300 Options 22795@subsection H8/300 Options 22796 22797These @samp{-m} options are defined for the H8/300 implementations: 22798 22799@table @gcctabopt 22800@item -mrelax 22801@opindex mrelax 22802Shorten some address references at link time, when possible; uses the 22803linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300, 22804ld, Using ld}, for a fuller description. 22805 22806@item -mh 22807@opindex mh 22808Generate code for the H8/300H@. 22809 22810@item -ms 22811@opindex ms 22812Generate code for the H8S@. 22813 22814@item -mn 22815@opindex mn 22816Generate code for the H8S and H8/300H in the normal mode. This switch 22817must be used either with @option{-mh} or @option{-ms}. 22818 22819@item -ms2600 22820@opindex ms2600 22821Generate code for the H8S/2600. This switch must be used with @option{-ms}. 22822 22823@item -mexr 22824@opindex mexr 22825Extended registers are stored on stack before execution of function 22826with monitor attribute. Default option is @option{-mexr}. 22827This option is valid only for H8S targets. 22828 22829@item -mno-exr 22830@opindex mno-exr 22831@opindex mexr 22832Extended registers are not stored on stack before execution of function 22833with monitor attribute. Default option is @option{-mno-exr}. 22834This option is valid only for H8S targets. 22835 22836@item -mint32 22837@opindex mint32 22838Make @code{int} data 32 bits by default. 22839 22840@item -malign-300 22841@opindex malign-300 22842On the H8/300H and H8S, use the same alignment rules as for the H8/300. 22843The default for the H8/300H and H8S is to align longs and floats on 228444-byte boundaries. 22845@option{-malign-300} causes them to be aligned on 2-byte boundaries. 22846This option has no effect on the H8/300. 22847@end table 22848 22849@node HPPA Options 22850@subsection HPPA Options 22851@cindex HPPA Options 22852 22853These @samp{-m} options are defined for the HPPA family of computers: 22854 22855@table @gcctabopt 22856@item -march=@var{architecture-type} 22857@opindex march 22858Generate code for the specified architecture. The choices for 22859@var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA 228601.1, and @samp{2.0} for PA 2.0 processors. Refer to 22861@file{/usr/lib/sched.models} on an HP-UX system to determine the proper 22862architecture option for your machine. Code compiled for lower numbered 22863architectures runs on higher numbered architectures, but not the 22864other way around. 22865 22866@item -mpa-risc-1-0 22867@itemx -mpa-risc-1-1 22868@itemx -mpa-risc-2-0 22869@opindex mpa-risc-1-0 22870@opindex mpa-risc-1-1 22871@opindex mpa-risc-2-0 22872Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively. 22873 22874@item -mcaller-copies 22875@opindex mcaller-copies 22876The caller copies function arguments passed by hidden reference. This 22877option should be used with care as it is not compatible with the default 2287832-bit runtime. However, only aggregates larger than eight bytes are 22879passed by hidden reference and the option provides better compatibility 22880with OpenMP. 22881 22882@item -mjump-in-delay 22883@opindex mjump-in-delay 22884This option is ignored and provided for compatibility purposes only. 22885 22886@item -mdisable-fpregs 22887@opindex mdisable-fpregs 22888Prevent floating-point registers from being used in any manner. This is 22889necessary for compiling kernels that perform lazy context switching of 22890floating-point registers. If you use this option and attempt to perform 22891floating-point operations, the compiler aborts. 22892 22893@item -mdisable-indexing 22894@opindex mdisable-indexing 22895Prevent the compiler from using indexing address modes. This avoids some 22896rather obscure problems when compiling MIG generated code under MACH@. 22897 22898@item -mno-space-regs 22899@opindex mno-space-regs 22900@opindex mspace-regs 22901Generate code that assumes the target has no space registers. This allows 22902GCC to generate faster indirect calls and use unscaled index address modes. 22903 22904Such code is suitable for level 0 PA systems and kernels. 22905 22906@item -mfast-indirect-calls 22907@opindex mfast-indirect-calls 22908Generate code that assumes calls never cross space boundaries. This 22909allows GCC to emit code that performs faster indirect calls. 22910 22911This option does not work in the presence of shared libraries or nested 22912functions. 22913 22914@item -mfixed-range=@var{register-range} 22915@opindex mfixed-range 22916Generate code treating the given register range as fixed registers. 22917A fixed register is one that the register allocator cannot use. This is 22918useful when compiling kernel code. A register range is specified as 22919two registers separated by a dash. Multiple register ranges can be 22920specified separated by a comma. 22921 22922@item -mlong-load-store 22923@opindex mlong-load-store 22924Generate 3-instruction load and store sequences as sometimes required by 22925the HP-UX 10 linker. This is equivalent to the @samp{+k} option to 22926the HP compilers. 22927 22928@item -mportable-runtime 22929@opindex mportable-runtime 22930Use the portable calling conventions proposed by HP for ELF systems. 22931 22932@item -mgas 22933@opindex mgas 22934Enable the use of assembler directives only GAS understands. 22935 22936@item -mschedule=@var{cpu-type} 22937@opindex mschedule 22938Schedule code according to the constraints for the machine type 22939@var{cpu-type}. The choices for @var{cpu-type} are @samp{700} 22940@samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer 22941to @file{/usr/lib/sched.models} on an HP-UX system to determine the 22942proper scheduling option for your machine. The default scheduling is 22943@samp{8000}. 22944 22945@item -mlinker-opt 22946@opindex mlinker-opt 22947Enable the optimization pass in the HP-UX linker. Note this makes symbolic 22948debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9 22949linkers in which they give bogus error messages when linking some programs. 22950 22951@item -msoft-float 22952@opindex msoft-float 22953Generate output containing library calls for floating point. 22954@strong{Warning:} the requisite libraries are not available for all HPPA 22955targets. Normally the facilities of the machine's usual C compiler are 22956used, but this cannot be done directly in cross-compilation. You must make 22957your own arrangements to provide suitable library functions for 22958cross-compilation. 22959 22960@option{-msoft-float} changes the calling convention in the output file; 22961therefore, it is only useful if you compile @emph{all} of a program with 22962this option. In particular, you need to compile @file{libgcc.a}, the 22963library that comes with GCC, with @option{-msoft-float} in order for 22964this to work. 22965 22966@item -msio 22967@opindex msio 22968Generate the predefine, @code{_SIO}, for server IO@. The default is 22969@option{-mwsio}. This generates the predefines, @code{__hp9000s700}, 22970@code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These 22971options are available under HP-UX and HI-UX@. 22972 22973@item -mgnu-ld 22974@opindex mgnu-ld 22975Use options specific to GNU @command{ld}. 22976This passes @option{-shared} to @command{ld} when 22977building a shared library. It is the default when GCC is configured, 22978explicitly or implicitly, with the GNU linker. This option does not 22979affect which @command{ld} is called; it only changes what parameters 22980are passed to that @command{ld}. 22981The @command{ld} that is called is determined by the 22982@option{--with-ld} configure option, GCC's program search path, and 22983finally by the user's @env{PATH}. The linker used by GCC can be printed 22984using @samp{which `gcc -print-prog-name=ld`}. This option is only available 22985on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}. 22986 22987@item -mhp-ld 22988@opindex mhp-ld 22989Use options specific to HP @command{ld}. 22990This passes @option{-b} to @command{ld} when building 22991a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all 22992links. It is the default when GCC is configured, explicitly or 22993implicitly, with the HP linker. This option does not affect 22994which @command{ld} is called; it only changes what parameters are passed to that 22995@command{ld}. 22996The @command{ld} that is called is determined by the @option{--with-ld} 22997configure option, GCC's program search path, and finally by the user's 22998@env{PATH}. The linker used by GCC can be printed using @samp{which 22999`gcc -print-prog-name=ld`}. This option is only available on the 64-bit 23000HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}. 23001 23002@item -mlong-calls 23003@opindex mno-long-calls 23004@opindex mlong-calls 23005Generate code that uses long call sequences. This ensures that a call 23006is always able to reach linker generated stubs. The default is to generate 23007long calls only when the distance from the call site to the beginning 23008of the function or translation unit, as the case may be, exceeds a 23009predefined limit set by the branch type being used. The limits for 23010normal calls are 7,600,000 and 240,000 bytes, respectively for the 23011PA 2.0 and PA 1.X architectures. Sibcalls are always limited at 23012240,000 bytes. 23013 23014Distances are measured from the beginning of functions when using the 23015@option{-ffunction-sections} option, or when using the @option{-mgas} 23016and @option{-mno-portable-runtime} options together under HP-UX with 23017the SOM linker. 23018 23019It is normally not desirable to use this option as it degrades 23020performance. However, it may be useful in large applications, 23021particularly when partial linking is used to build the application. 23022 23023The types of long calls used depends on the capabilities of the 23024assembler and linker, and the type of code being generated. The 23025impact on systems that support long absolute calls, and long pic 23026symbol-difference or pc-relative calls should be relatively small. 23027However, an indirect call is used on 32-bit ELF systems in pic code 23028and it is quite long. 23029 23030@item -munix=@var{unix-std} 23031@opindex march 23032Generate compiler predefines and select a startfile for the specified 23033UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95} 23034and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95} 23035is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX 2303611.11 and later. The default values are @samp{93} for HP-UX 10.00, 23037@samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11 23038and later. 23039 23040@option{-munix=93} provides the same predefines as GCC 3.3 and 3.4. 23041@option{-munix=95} provides additional predefines for @code{XOPEN_UNIX} 23042and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}. 23043@option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX}, 23044@code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and 23045@code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}. 23046 23047It is @emph{important} to note that this option changes the interfaces 23048for various library routines. It also affects the operational behavior 23049of the C library. Thus, @emph{extreme} care is needed in using this 23050option. 23051 23052Library code that is intended to operate with more than one UNIX 23053standard must test, set and restore the variable @code{__xpg4_extended_mask} 23054as appropriate. Most GNU software doesn't provide this capability. 23055 23056@item -nolibdld 23057@opindex nolibdld 23058Suppress the generation of link options to search libdld.sl when the 23059@option{-static} option is specified on HP-UX 10 and later. 23060 23061@item -static 23062@opindex static 23063The HP-UX implementation of setlocale in libc has a dependency on 23064libdld.sl. There isn't an archive version of libdld.sl. Thus, 23065when the @option{-static} option is specified, special link options 23066are needed to resolve this dependency. 23067 23068On HP-UX 10 and later, the GCC driver adds the necessary options to 23069link with libdld.sl when the @option{-static} option is specified. 23070This causes the resulting binary to be dynamic. On the 64-bit port, 23071the linkers generate dynamic binaries by default in any case. The 23072@option{-nolibdld} option can be used to prevent the GCC driver from 23073adding these link options. 23074 23075@item -threads 23076@opindex threads 23077Add support for multithreading with the @dfn{dce thread} library 23078under HP-UX@. This option sets flags for both the preprocessor and 23079linker. 23080@end table 23081 23082@node IA-64 Options 23083@subsection IA-64 Options 23084@cindex IA-64 Options 23085 23086These are the @samp{-m} options defined for the Intel IA-64 architecture. 23087 23088@table @gcctabopt 23089@item -mbig-endian 23090@opindex mbig-endian 23091Generate code for a big-endian target. This is the default for HP-UX@. 23092 23093@item -mlittle-endian 23094@opindex mlittle-endian 23095Generate code for a little-endian target. This is the default for AIX5 23096and GNU/Linux. 23097 23098@item -mgnu-as 23099@itemx -mno-gnu-as 23100@opindex mgnu-as 23101@opindex mno-gnu-as 23102Generate (or don't) code for the GNU assembler. This is the default. 23103@c Also, this is the default if the configure option @option{--with-gnu-as} 23104@c is used. 23105 23106@item -mgnu-ld 23107@itemx -mno-gnu-ld 23108@opindex mgnu-ld 23109@opindex mno-gnu-ld 23110Generate (or don't) code for the GNU linker. This is the default. 23111@c Also, this is the default if the configure option @option{--with-gnu-ld} 23112@c is used. 23113 23114@item -mno-pic 23115@opindex mno-pic 23116Generate code that does not use a global pointer register. The result 23117is not position independent code, and violates the IA-64 ABI@. 23118 23119@item -mvolatile-asm-stop 23120@itemx -mno-volatile-asm-stop 23121@opindex mvolatile-asm-stop 23122@opindex mno-volatile-asm-stop 23123Generate (or don't) a stop bit immediately before and after volatile asm 23124statements. 23125 23126@item -mregister-names 23127@itemx -mno-register-names 23128@opindex mregister-names 23129@opindex mno-register-names 23130Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for 23131the stacked registers. This may make assembler output more readable. 23132 23133@item -mno-sdata 23134@itemx -msdata 23135@opindex mno-sdata 23136@opindex msdata 23137Disable (or enable) optimizations that use the small data section. This may 23138be useful for working around optimizer bugs. 23139 23140@item -mconstant-gp 23141@opindex mconstant-gp 23142Generate code that uses a single constant global pointer value. This is 23143useful when compiling kernel code. 23144 23145@item -mauto-pic 23146@opindex mauto-pic 23147Generate code that is self-relocatable. This implies @option{-mconstant-gp}. 23148This is useful when compiling firmware code. 23149 23150@item -minline-float-divide-min-latency 23151@opindex minline-float-divide-min-latency 23152Generate code for inline divides of floating-point values 23153using the minimum latency algorithm. 23154 23155@item -minline-float-divide-max-throughput 23156@opindex minline-float-divide-max-throughput 23157Generate code for inline divides of floating-point values 23158using the maximum throughput algorithm. 23159 23160@item -mno-inline-float-divide 23161@opindex mno-inline-float-divide 23162Do not generate inline code for divides of floating-point values. 23163 23164@item -minline-int-divide-min-latency 23165@opindex minline-int-divide-min-latency 23166Generate code for inline divides of integer values 23167using the minimum latency algorithm. 23168 23169@item -minline-int-divide-max-throughput 23170@opindex minline-int-divide-max-throughput 23171Generate code for inline divides of integer values 23172using the maximum throughput algorithm. 23173 23174@item -mno-inline-int-divide 23175@opindex mno-inline-int-divide 23176@opindex minline-int-divide 23177Do not generate inline code for divides of integer values. 23178 23179@item -minline-sqrt-min-latency 23180@opindex minline-sqrt-min-latency 23181Generate code for inline square roots 23182using the minimum latency algorithm. 23183 23184@item -minline-sqrt-max-throughput 23185@opindex minline-sqrt-max-throughput 23186Generate code for inline square roots 23187using the maximum throughput algorithm. 23188 23189@item -mno-inline-sqrt 23190@opindex mno-inline-sqrt 23191Do not generate inline code for @code{sqrt}. 23192 23193@item -mfused-madd 23194@itemx -mno-fused-madd 23195@opindex mfused-madd 23196@opindex mno-fused-madd 23197Do (don't) generate code that uses the fused multiply/add or multiply/subtract 23198instructions. The default is to use these instructions. 23199 23200@item -mno-dwarf2-asm 23201@itemx -mdwarf2-asm 23202@opindex mno-dwarf2-asm 23203@opindex mdwarf2-asm 23204Don't (or do) generate assembler code for the DWARF line number debugging 23205info. This may be useful when not using the GNU assembler. 23206 23207@item -mearly-stop-bits 23208@itemx -mno-early-stop-bits 23209@opindex mearly-stop-bits 23210@opindex mno-early-stop-bits 23211Allow stop bits to be placed earlier than immediately preceding the 23212instruction that triggered the stop bit. This can improve instruction 23213scheduling, but does not always do so. 23214 23215@item -mfixed-range=@var{register-range} 23216@opindex mfixed-range 23217Generate code treating the given register range as fixed registers. 23218A fixed register is one that the register allocator cannot use. This is 23219useful when compiling kernel code. A register range is specified as 23220two registers separated by a dash. Multiple register ranges can be 23221specified separated by a comma. 23222 23223@item -mtls-size=@var{tls-size} 23224@opindex mtls-size 23225Specify bit size of immediate TLS offsets. Valid values are 14, 22, and 2322664. 23227 23228@item -mtune=@var{cpu-type} 23229@opindex mtune 23230Tune the instruction scheduling for a particular CPU, Valid values are 23231@samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2}, 23232and @samp{mckinley}. 23233 23234@item -milp32 23235@itemx -mlp64 23236@opindex milp32 23237@opindex mlp64 23238Generate code for a 32-bit or 64-bit environment. 23239The 32-bit environment sets int, long and pointer to 32 bits. 23240The 64-bit environment sets int to 32 bits and long and pointer 23241to 64 bits. These are HP-UX specific flags. 23242 23243@item -mno-sched-br-data-spec 23244@itemx -msched-br-data-spec 23245@opindex mno-sched-br-data-spec 23246@opindex msched-br-data-spec 23247(Dis/En)able data speculative scheduling before reload. 23248This results in generation of @code{ld.a} instructions and 23249the corresponding check instructions (@code{ld.c} / @code{chk.a}). 23250The default setting is disabled. 23251 23252@item -msched-ar-data-spec 23253@itemx -mno-sched-ar-data-spec 23254@opindex msched-ar-data-spec 23255@opindex mno-sched-ar-data-spec 23256(En/Dis)able data speculative scheduling after reload. 23257This results in generation of @code{ld.a} instructions and 23258the corresponding check instructions (@code{ld.c} / @code{chk.a}). 23259The default setting is enabled. 23260 23261@item -mno-sched-control-spec 23262@itemx -msched-control-spec 23263@opindex mno-sched-control-spec 23264@opindex msched-control-spec 23265(Dis/En)able control speculative scheduling. This feature is 23266available only during region scheduling (i.e.@: before reload). 23267This results in generation of the @code{ld.s} instructions and 23268the corresponding check instructions @code{chk.s}. 23269The default setting is disabled. 23270 23271@item -msched-br-in-data-spec 23272@itemx -mno-sched-br-in-data-spec 23273@opindex msched-br-in-data-spec 23274@opindex mno-sched-br-in-data-spec 23275(En/Dis)able speculative scheduling of the instructions that 23276are dependent on the data speculative loads before reload. 23277This is effective only with @option{-msched-br-data-spec} enabled. 23278The default setting is enabled. 23279 23280@item -msched-ar-in-data-spec 23281@itemx -mno-sched-ar-in-data-spec 23282@opindex msched-ar-in-data-spec 23283@opindex mno-sched-ar-in-data-spec 23284(En/Dis)able speculative scheduling of the instructions that 23285are dependent on the data speculative loads after reload. 23286This is effective only with @option{-msched-ar-data-spec} enabled. 23287The default setting is enabled. 23288 23289@item -msched-in-control-spec 23290@itemx -mno-sched-in-control-spec 23291@opindex msched-in-control-spec 23292@opindex mno-sched-in-control-spec 23293(En/Dis)able speculative scheduling of the instructions that 23294are dependent on the control speculative loads. 23295This is effective only with @option{-msched-control-spec} enabled. 23296The default setting is enabled. 23297 23298@item -mno-sched-prefer-non-data-spec-insns 23299@itemx -msched-prefer-non-data-spec-insns 23300@opindex mno-sched-prefer-non-data-spec-insns 23301@opindex msched-prefer-non-data-spec-insns 23302If enabled, data-speculative instructions are chosen for schedule 23303only if there are no other choices at the moment. This makes 23304the use of the data speculation much more conservative. 23305The default setting is disabled. 23306 23307@item -mno-sched-prefer-non-control-spec-insns 23308@itemx -msched-prefer-non-control-spec-insns 23309@opindex mno-sched-prefer-non-control-spec-insns 23310@opindex msched-prefer-non-control-spec-insns 23311If enabled, control-speculative instructions are chosen for schedule 23312only if there are no other choices at the moment. This makes 23313the use of the control speculation much more conservative. 23314The default setting is disabled. 23315 23316@item -mno-sched-count-spec-in-critical-path 23317@itemx -msched-count-spec-in-critical-path 23318@opindex mno-sched-count-spec-in-critical-path 23319@opindex msched-count-spec-in-critical-path 23320If enabled, speculative dependencies are considered during 23321computation of the instructions priorities. This makes the use of the 23322speculation a bit more conservative. 23323The default setting is disabled. 23324 23325@item -msched-spec-ldc 23326@opindex msched-spec-ldc 23327Use a simple data speculation check. This option is on by default. 23328 23329@item -msched-control-spec-ldc 23330@opindex msched-spec-ldc 23331Use a simple check for control speculation. This option is on by default. 23332 23333@item -msched-stop-bits-after-every-cycle 23334@opindex msched-stop-bits-after-every-cycle 23335Place a stop bit after every cycle when scheduling. This option is on 23336by default. 23337 23338@item -msched-fp-mem-deps-zero-cost 23339@opindex msched-fp-mem-deps-zero-cost 23340Assume that floating-point stores and loads are not likely to cause a conflict 23341when placed into the same instruction group. This option is disabled by 23342default. 23343 23344@item -msel-sched-dont-check-control-spec 23345@opindex msel-sched-dont-check-control-spec 23346Generate checks for control speculation in selective scheduling. 23347This flag is disabled by default. 23348 23349@item -msched-max-memory-insns=@var{max-insns} 23350@opindex msched-max-memory-insns 23351Limit on the number of memory insns per instruction group, giving lower 23352priority to subsequent memory insns attempting to schedule in the same 23353instruction group. Frequently useful to prevent cache bank conflicts. 23354The default value is 1. 23355 23356@item -msched-max-memory-insns-hard-limit 23357@opindex msched-max-memory-insns-hard-limit 23358Makes the limit specified by @option{msched-max-memory-insns} a hard limit, 23359disallowing more than that number in an instruction group. 23360Otherwise, the limit is ``soft'', meaning that non-memory operations 23361are preferred when the limit is reached, but memory operations may still 23362be scheduled. 23363 23364@end table 23365 23366@node LM32 Options 23367@subsection LM32 Options 23368@cindex LM32 options 23369 23370These @option{-m} options are defined for the LatticeMico32 architecture: 23371 23372@table @gcctabopt 23373@item -mbarrel-shift-enabled 23374@opindex mbarrel-shift-enabled 23375Enable barrel-shift instructions. 23376 23377@item -mdivide-enabled 23378@opindex mdivide-enabled 23379Enable divide and modulus instructions. 23380 23381@item -mmultiply-enabled 23382@opindex multiply-enabled 23383Enable multiply instructions. 23384 23385@item -msign-extend-enabled 23386@opindex msign-extend-enabled 23387Enable sign extend instructions. 23388 23389@item -muser-enabled 23390@opindex muser-enabled 23391Enable user-defined instructions. 23392 23393@end table 23394 23395@node M32C Options 23396@subsection M32C Options 23397@cindex M32C options 23398 23399@table @gcctabopt 23400@item -mcpu=@var{name} 23401@opindex mcpu= 23402Select the CPU for which code is generated. @var{name} may be one of 23403@samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to 23404/60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for 23405the M32C/80 series. 23406 23407@item -msim 23408@opindex msim 23409Specifies that the program will be run on the simulator. This causes 23410an alternate runtime library to be linked in which supports, for 23411example, file I/O@. You must not use this option when generating 23412programs that will run on real hardware; you must provide your own 23413runtime library for whatever I/O functions are needed. 23414 23415@item -memregs=@var{number} 23416@opindex memregs= 23417Specifies the number of memory-based pseudo-registers GCC uses 23418during code generation. These pseudo-registers are used like real 23419registers, so there is a tradeoff between GCC's ability to fit the 23420code into available registers, and the performance penalty of using 23421memory instead of registers. Note that all modules in a program must 23422be compiled with the same value for this option. Because of that, you 23423must not use this option with GCC's default runtime libraries. 23424 23425@end table 23426 23427@node M32R/D Options 23428@subsection M32R/D Options 23429@cindex M32R/D options 23430 23431These @option{-m} options are defined for Renesas M32R/D architectures: 23432 23433@table @gcctabopt 23434@item -m32r2 23435@opindex m32r2 23436Generate code for the M32R/2@. 23437 23438@item -m32rx 23439@opindex m32rx 23440Generate code for the M32R/X@. 23441 23442@item -m32r 23443@opindex m32r 23444Generate code for the M32R@. This is the default. 23445 23446@item -mmodel=small 23447@opindex mmodel=small 23448Assume all objects live in the lower 16MB of memory (so that their addresses 23449can be loaded with the @code{ld24} instruction), and assume all subroutines 23450are reachable with the @code{bl} instruction. 23451This is the default. 23452 23453The addressability of a particular object can be set with the 23454@code{model} attribute. 23455 23456@item -mmodel=medium 23457@opindex mmodel=medium 23458Assume objects may be anywhere in the 32-bit address space (the compiler 23459generates @code{seth/add3} instructions to load their addresses), and 23460assume all subroutines are reachable with the @code{bl} instruction. 23461 23462@item -mmodel=large 23463@opindex mmodel=large 23464Assume objects may be anywhere in the 32-bit address space (the compiler 23465generates @code{seth/add3} instructions to load their addresses), and 23466assume subroutines may not be reachable with the @code{bl} instruction 23467(the compiler generates the much slower @code{seth/add3/jl} 23468instruction sequence). 23469 23470@item -msdata=none 23471@opindex msdata=none 23472Disable use of the small data area. Variables are put into 23473one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the 23474@code{section} attribute has been specified). 23475This is the default. 23476 23477The small data area consists of sections @code{.sdata} and @code{.sbss}. 23478Objects may be explicitly put in the small data area with the 23479@code{section} attribute using one of these sections. 23480 23481@item -msdata=sdata 23482@opindex msdata=sdata 23483Put small global and static data in the small data area, but do not 23484generate special code to reference them. 23485 23486@item -msdata=use 23487@opindex msdata=use 23488Put small global and static data in the small data area, and generate 23489special instructions to reference them. 23490 23491@item -G @var{num} 23492@opindex G 23493@cindex smaller data references 23494Put global and static objects less than or equal to @var{num} bytes 23495into the small data or BSS sections instead of the normal data or BSS 23496sections. The default value of @var{num} is 8. 23497The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use} 23498for this option to have any effect. 23499 23500All modules should be compiled with the same @option{-G @var{num}} value. 23501Compiling with different values of @var{num} may or may not work; if it 23502doesn't the linker gives an error message---incorrect code is not 23503generated. 23504 23505@item -mdebug 23506@opindex mdebug 23507Makes the M32R-specific code in the compiler display some statistics 23508that might help in debugging programs. 23509 23510@item -malign-loops 23511@opindex malign-loops 23512Align all loops to a 32-byte boundary. 23513 23514@item -mno-align-loops 23515@opindex mno-align-loops 23516Do not enforce a 32-byte alignment for loops. This is the default. 23517 23518@item -missue-rate=@var{number} 23519@opindex missue-rate=@var{number} 23520Issue @var{number} instructions per cycle. @var{number} can only be 1 23521or 2. 23522 23523@item -mbranch-cost=@var{number} 23524@opindex mbranch-cost=@var{number} 23525@var{number} can only be 1 or 2. If it is 1 then branches are 23526preferred over conditional code, if it is 2, then the opposite applies. 23527 23528@item -mflush-trap=@var{number} 23529@opindex mflush-trap=@var{number} 23530Specifies the trap number to use to flush the cache. The default is 2353112. Valid numbers are between 0 and 15 inclusive. 23532 23533@item -mno-flush-trap 23534@opindex mno-flush-trap 23535Specifies that the cache cannot be flushed by using a trap. 23536 23537@item -mflush-func=@var{name} 23538@opindex mflush-func=@var{name} 23539Specifies the name of the operating system function to call to flush 23540the cache. The default is @samp{_flush_cache}, but a function call 23541is only used if a trap is not available. 23542 23543@item -mno-flush-func 23544@opindex mno-flush-func 23545Indicates that there is no OS function for flushing the cache. 23546 23547@end table 23548 23549@node M680x0 Options 23550@subsection M680x0 Options 23551@cindex M680x0 options 23552 23553These are the @samp{-m} options defined for M680x0 and ColdFire processors. 23554The default settings depend on which architecture was selected when 23555the compiler was configured; the defaults for the most common choices 23556are given below. 23557 23558@table @gcctabopt 23559@item -march=@var{arch} 23560@opindex march 23561Generate code for a specific M680x0 or ColdFire instruction set 23562architecture. Permissible values of @var{arch} for M680x0 23563architectures are: @samp{68000}, @samp{68010}, @samp{68020}, 23564@samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire 23565architectures are selected according to Freescale's ISA classification 23566and the permissible values are: @samp{isaa}, @samp{isaaplus}, 23567@samp{isab} and @samp{isac}. 23568 23569GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating 23570code for a ColdFire target. The @var{arch} in this macro is one of the 23571@option{-march} arguments given above. 23572 23573When used together, @option{-march} and @option{-mtune} select code 23574that runs on a family of similar processors but that is optimized 23575for a particular microarchitecture. 23576 23577@item -mcpu=@var{cpu} 23578@opindex mcpu 23579Generate code for a specific M680x0 or ColdFire processor. 23580The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020}, 23581@samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332} 23582and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table 23583below, which also classifies the CPUs into families: 23584 23585@multitable @columnfractions 0.20 0.80 23586@item @strong{Family} @tab @strong{@samp{-mcpu} arguments} 23587@item @samp{51} @tab @samp{51} @samp{51ac} @samp{51ag} @samp{51cn} @samp{51em} @samp{51je} @samp{51jf} @samp{51jg} @samp{51jm} @samp{51mm} @samp{51qe} @samp{51qm} 23588@item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206} 23589@item @samp{5206e} @tab @samp{5206e} 23590@item @samp{5208} @tab @samp{5207} @samp{5208} 23591@item @samp{5211a} @tab @samp{5210a} @samp{5211a} 23592@item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213} 23593@item @samp{5216} @tab @samp{5214} @samp{5216} 23594@item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235} 23595@item @samp{5225} @tab @samp{5224} @samp{5225} 23596@item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259} 23597@item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x} 23598@item @samp{5249} @tab @samp{5249} 23599@item @samp{5250} @tab @samp{5250} 23600@item @samp{5271} @tab @samp{5270} @samp{5271} 23601@item @samp{5272} @tab @samp{5272} 23602@item @samp{5275} @tab @samp{5274} @samp{5275} 23603@item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x} 23604@item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017} 23605@item @samp{5307} @tab @samp{5307} 23606@item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x} 23607@item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x} 23608@item @samp{5407} @tab @samp{5407} 23609@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} 23610@end multitable 23611 23612@option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if 23613@var{arch} is compatible with @var{cpu}. Other combinations of 23614@option{-mcpu} and @option{-march} are rejected. 23615 23616GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target 23617@var{cpu} is selected. It also defines @code{__mcf_family_@var{family}}, 23618where the value of @var{family} is given by the table above. 23619 23620@item -mtune=@var{tune} 23621@opindex mtune 23622Tune the code for a particular microarchitecture within the 23623constraints set by @option{-march} and @option{-mcpu}. 23624The M680x0 microarchitectures are: @samp{68000}, @samp{68010}, 23625@samp{68020}, @samp{68030}, @samp{68040}, @samp{68060} 23626and @samp{cpu32}. The ColdFire microarchitectures 23627are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}. 23628 23629You can also use @option{-mtune=68020-40} for code that needs 23630to run relatively well on 68020, 68030 and 68040 targets. 23631@option{-mtune=68020-60} is similar but includes 68060 targets 23632as well. These two options select the same tuning decisions as 23633@option{-m68020-40} and @option{-m68020-60} respectively. 23634 23635GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__} 23636when tuning for 680x0 architecture @var{arch}. It also defines 23637@code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std} 23638option is used. If GCC is tuning for a range of architectures, 23639as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60}, 23640it defines the macros for every architecture in the range. 23641 23642GCC also defines the macro @code{__m@var{uarch}__} when tuning for 23643ColdFire microarchitecture @var{uarch}, where @var{uarch} is one 23644of the arguments given above. 23645 23646@item -m68000 23647@itemx -mc68000 23648@opindex m68000 23649@opindex mc68000 23650Generate output for a 68000. This is the default 23651when the compiler is configured for 68000-based systems. 23652It is equivalent to @option{-march=68000}. 23653 23654Use this option for microcontrollers with a 68000 or EC000 core, 23655including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356. 23656 23657@item -m68010 23658@opindex m68010 23659Generate output for a 68010. This is the default 23660when the compiler is configured for 68010-based systems. 23661It is equivalent to @option{-march=68010}. 23662 23663@item -m68020 23664@itemx -mc68020 23665@opindex m68020 23666@opindex mc68020 23667Generate output for a 68020. This is the default 23668when the compiler is configured for 68020-based systems. 23669It is equivalent to @option{-march=68020}. 23670 23671@item -m68030 23672@opindex m68030 23673Generate output for a 68030. This is the default when the compiler is 23674configured for 68030-based systems. It is equivalent to 23675@option{-march=68030}. 23676 23677@item -m68040 23678@opindex m68040 23679Generate output for a 68040. This is the default when the compiler is 23680configured for 68040-based systems. It is equivalent to 23681@option{-march=68040}. 23682 23683This option inhibits the use of 68881/68882 instructions that have to be 23684emulated by software on the 68040. Use this option if your 68040 does not 23685have code to emulate those instructions. 23686 23687@item -m68060 23688@opindex m68060 23689Generate output for a 68060. This is the default when the compiler is 23690configured for 68060-based systems. It is equivalent to 23691@option{-march=68060}. 23692 23693This option inhibits the use of 68020 and 68881/68882 instructions that 23694have to be emulated by software on the 68060. Use this option if your 68060 23695does not have code to emulate those instructions. 23696 23697@item -mcpu32 23698@opindex mcpu32 23699Generate output for a CPU32. This is the default 23700when the compiler is configured for CPU32-based systems. 23701It is equivalent to @option{-march=cpu32}. 23702 23703Use this option for microcontrollers with a 23704CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334, 2370568336, 68340, 68341, 68349 and 68360. 23706 23707@item -m5200 23708@opindex m5200 23709Generate output for a 520X ColdFire CPU@. This is the default 23710when the compiler is configured for 520X-based systems. 23711It is equivalent to @option{-mcpu=5206}, and is now deprecated 23712in favor of that option. 23713 23714Use this option for microcontroller with a 5200 core, including 23715the MCF5202, MCF5203, MCF5204 and MCF5206. 23716 23717@item -m5206e 23718@opindex m5206e 23719Generate output for a 5206e ColdFire CPU@. The option is now 23720deprecated in favor of the equivalent @option{-mcpu=5206e}. 23721 23722@item -m528x 23723@opindex m528x 23724Generate output for a member of the ColdFire 528X family. 23725The option is now deprecated in favor of the equivalent 23726@option{-mcpu=528x}. 23727 23728@item -m5307 23729@opindex m5307 23730Generate output for a ColdFire 5307 CPU@. The option is now deprecated 23731in favor of the equivalent @option{-mcpu=5307}. 23732 23733@item -m5407 23734@opindex m5407 23735Generate output for a ColdFire 5407 CPU@. The option is now deprecated 23736in favor of the equivalent @option{-mcpu=5407}. 23737 23738@item -mcfv4e 23739@opindex mcfv4e 23740Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x). 23741This includes use of hardware floating-point instructions. 23742The option is equivalent to @option{-mcpu=547x}, and is now 23743deprecated in favor of that option. 23744 23745@item -m68020-40 23746@opindex m68020-40 23747Generate output for a 68040, without using any of the new instructions. 23748This results in code that can run relatively efficiently on either a 2374968020/68881 or a 68030 or a 68040. The generated code does use the 2375068881 instructions that are emulated on the 68040. 23751 23752The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}. 23753 23754@item -m68020-60 23755@opindex m68020-60 23756Generate output for a 68060, without using any of the new instructions. 23757This results in code that can run relatively efficiently on either a 2375868020/68881 or a 68030 or a 68040. The generated code does use the 2375968881 instructions that are emulated on the 68060. 23760 23761The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}. 23762 23763@item -mhard-float 23764@itemx -m68881 23765@opindex mhard-float 23766@opindex m68881 23767Generate floating-point instructions. This is the default for 68020 23768and above, and for ColdFire devices that have an FPU@. It defines the 23769macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__} 23770on ColdFire targets. 23771 23772@item -msoft-float 23773@opindex msoft-float 23774Do not generate floating-point instructions; use library calls instead. 23775This is the default for 68000, 68010, and 68832 targets. It is also 23776the default for ColdFire devices that have no FPU. 23777 23778@item -mdiv 23779@itemx -mno-div 23780@opindex mdiv 23781@opindex mno-div 23782Generate (do not generate) ColdFire hardware divide and remainder 23783instructions. If @option{-march} is used without @option{-mcpu}, 23784the default is ``on'' for ColdFire architectures and ``off'' for M680x0 23785architectures. Otherwise, the default is taken from the target CPU 23786(either the default CPU, or the one specified by @option{-mcpu}). For 23787example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for 23788@option{-mcpu=5206e}. 23789 23790GCC defines the macro @code{__mcfhwdiv__} when this option is enabled. 23791 23792@item -mshort 23793@opindex mshort 23794Consider type @code{int} to be 16 bits wide, like @code{short int}. 23795Additionally, parameters passed on the stack are also aligned to a 2379616-bit boundary even on targets whose API mandates promotion to 32-bit. 23797 23798@item -mno-short 23799@opindex mno-short 23800Do not consider type @code{int} to be 16 bits wide. This is the default. 23801 23802@item -mnobitfield 23803@itemx -mno-bitfield 23804@opindex mnobitfield 23805@opindex mno-bitfield 23806Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32} 23807and @option{-m5200} options imply @w{@option{-mnobitfield}}. 23808 23809@item -mbitfield 23810@opindex mbitfield 23811Do use the bit-field instructions. The @option{-m68020} option implies 23812@option{-mbitfield}. This is the default if you use a configuration 23813designed for a 68020. 23814 23815@item -mrtd 23816@opindex mrtd 23817Use a different function-calling convention, in which functions 23818that take a fixed number of arguments return with the @code{rtd} 23819instruction, which pops their arguments while returning. This 23820saves one instruction in the caller since there is no need to pop 23821the arguments there. 23822 23823This calling convention is incompatible with the one normally 23824used on Unix, so you cannot use it if you need to call libraries 23825compiled with the Unix compiler. 23826 23827Also, you must provide function prototypes for all functions that 23828take variable numbers of arguments (including @code{printf}); 23829otherwise incorrect code is generated for calls to those 23830functions. 23831 23832In addition, seriously incorrect code results if you call a 23833function with too many arguments. (Normally, extra arguments are 23834harmlessly ignored.) 23835 23836The @code{rtd} instruction is supported by the 68010, 68020, 68030, 2383768040, 68060 and CPU32 processors, but not by the 68000 or 5200. 23838 23839The default is @option{-mno-rtd}. 23840 23841@item -malign-int 23842@itemx -mno-align-int 23843@opindex malign-int 23844@opindex mno-align-int 23845Control whether GCC aligns @code{int}, @code{long}, @code{long long}, 23846@code{float}, @code{double}, and @code{long double} variables on a 32-bit 23847boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}). 23848Aligning variables on 32-bit boundaries produces code that runs somewhat 23849faster on processors with 32-bit busses at the expense of more memory. 23850 23851@strong{Warning:} if you use the @option{-malign-int} switch, GCC 23852aligns structures containing the above types differently than 23853most published application binary interface specifications for the m68k. 23854 23855@opindex mpcrel 23856Use the pc-relative addressing mode of the 68000 directly, instead of 23857using a global offset table. At present, this option implies @option{-fpic}, 23858allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is 23859not presently supported with @option{-mpcrel}, though this could be supported for 2386068020 and higher processors. 23861 23862@item -mno-strict-align 23863@itemx -mstrict-align 23864@opindex mno-strict-align 23865@opindex mstrict-align 23866Do not (do) assume that unaligned memory references are handled by 23867the system. 23868 23869@item -msep-data 23870Generate code that allows the data segment to be located in a different 23871area of memory from the text segment. This allows for execute-in-place in 23872an environment without virtual memory management. This option implies 23873@option{-fPIC}. 23874 23875@item -mno-sep-data 23876Generate code that assumes that the data segment follows the text segment. 23877This is the default. 23878 23879@item -mid-shared-library 23880Generate code that supports shared libraries via the library ID method. 23881This allows for execute-in-place and shared libraries in an environment 23882without virtual memory management. This option implies @option{-fPIC}. 23883 23884@item -mno-id-shared-library 23885Generate code that doesn't assume ID-based shared libraries are being used. 23886This is the default. 23887 23888@item -mshared-library-id=n 23889Specifies the identification number of the ID-based shared library being 23890compiled. Specifying a value of 0 generates more compact code; specifying 23891other values forces the allocation of that number to the current 23892library, but is no more space- or time-efficient than omitting this option. 23893 23894@item -mxgot 23895@itemx -mno-xgot 23896@opindex mxgot 23897@opindex mno-xgot 23898When generating position-independent code for ColdFire, generate code 23899that works if the GOT has more than 8192 entries. This code is 23900larger and slower than code generated without this option. On M680x0 23901processors, this option is not needed; @option{-fPIC} suffices. 23902 23903GCC normally uses a single instruction to load values from the GOT@. 23904While this is relatively efficient, it only works if the GOT 23905is smaller than about 64k. Anything larger causes the linker 23906to report an error such as: 23907 23908@cindex relocation truncated to fit (ColdFire) 23909@smallexample 23910relocation truncated to fit: R_68K_GOT16O foobar 23911@end smallexample 23912 23913If this happens, you should recompile your code with @option{-mxgot}. 23914It should then work with very large GOTs. However, code generated with 23915@option{-mxgot} is less efficient, since it takes 4 instructions to fetch 23916the value of a global symbol. 23917 23918Note that some linkers, including newer versions of the GNU linker, 23919can create multiple GOTs and sort GOT entries. If you have such a linker, 23920you should only need to use @option{-mxgot} when compiling a single 23921object file that accesses more than 8192 GOT entries. Very few do. 23922 23923These options have no effect unless GCC is generating 23924position-independent code. 23925 23926@item -mlong-jump-table-offsets 23927@opindex mlong-jump-table-offsets 23928Use 32-bit offsets in @code{switch} tables. The default is to use 2392916-bit offsets. 23930 23931@end table 23932 23933@node MCore Options 23934@subsection MCore Options 23935@cindex MCore options 23936 23937These are the @samp{-m} options defined for the Motorola M*Core 23938processors. 23939 23940@table @gcctabopt 23941 23942@item -mhardlit 23943@itemx -mno-hardlit 23944@opindex mhardlit 23945@opindex mno-hardlit 23946Inline constants into the code stream if it can be done in two 23947instructions or less. 23948 23949@item -mdiv 23950@itemx -mno-div 23951@opindex mdiv 23952@opindex mno-div 23953Use the divide instruction. (Enabled by default). 23954 23955@item -mrelax-immediate 23956@itemx -mno-relax-immediate 23957@opindex mrelax-immediate 23958@opindex mno-relax-immediate 23959Allow arbitrary-sized immediates in bit operations. 23960 23961@item -mwide-bitfields 23962@itemx -mno-wide-bitfields 23963@opindex mwide-bitfields 23964@opindex mno-wide-bitfields 23965Always treat bit-fields as @code{int}-sized. 23966 23967@item -m4byte-functions 23968@itemx -mno-4byte-functions 23969@opindex m4byte-functions 23970@opindex mno-4byte-functions 23971Force all functions to be aligned to a 4-byte boundary. 23972 23973@item -mcallgraph-data 23974@itemx -mno-callgraph-data 23975@opindex mcallgraph-data 23976@opindex mno-callgraph-data 23977Emit callgraph information. 23978 23979@item -mslow-bytes 23980@itemx -mno-slow-bytes 23981@opindex mslow-bytes 23982@opindex mno-slow-bytes 23983Prefer word access when reading byte quantities. 23984 23985@item -mlittle-endian 23986@itemx -mbig-endian 23987@opindex mlittle-endian 23988@opindex mbig-endian 23989Generate code for a little-endian target. 23990 23991@item -m210 23992@itemx -m340 23993@opindex m210 23994@opindex m340 23995Generate code for the 210 processor. 23996 23997@item -mno-lsim 23998@opindex mno-lsim 23999Assume that runtime support has been provided and so omit the 24000simulator library (@file{libsim.a)} from the linker command line. 24001 24002@item -mstack-increment=@var{size} 24003@opindex mstack-increment 24004Set the maximum amount for a single stack increment operation. Large 24005values can increase the speed of programs that contain functions 24006that need a large amount of stack space, but they can also trigger a 24007segmentation fault if the stack is extended too much. The default 24008value is 0x1000. 24009 24010@end table 24011 24012@node MeP Options 24013@subsection MeP Options 24014@cindex MeP options 24015 24016@table @gcctabopt 24017 24018@item -mabsdiff 24019@opindex mabsdiff 24020Enables the @code{abs} instruction, which is the absolute difference 24021between two registers. 24022 24023@item -mall-opts 24024@opindex mall-opts 24025Enables all the optional instructions---average, multiply, divide, bit 24026operations, leading zero, absolute difference, min/max, clip, and 24027saturation. 24028 24029 24030@item -maverage 24031@opindex maverage 24032Enables the @code{ave} instruction, which computes the average of two 24033registers. 24034 24035@item -mbased=@var{n} 24036@opindex mbased= 24037Variables of size @var{n} bytes or smaller are placed in the 24038@code{.based} section by default. Based variables use the @code{$tp} 24039register as a base register, and there is a 128-byte limit to the 24040@code{.based} section. 24041 24042@item -mbitops 24043@opindex mbitops 24044Enables the bit operation instructions---bit test (@code{btstm}), set 24045(@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and 24046test-and-set (@code{tas}). 24047 24048@item -mc=@var{name} 24049@opindex mc= 24050Selects which section constant data is placed in. @var{name} may 24051be @samp{tiny}, @samp{near}, or @samp{far}. 24052 24053@item -mclip 24054@opindex mclip 24055Enables the @code{clip} instruction. Note that @option{-mclip} is not 24056useful unless you also provide @option{-mminmax}. 24057 24058@item -mconfig=@var{name} 24059@opindex mconfig= 24060Selects one of the built-in core configurations. Each MeP chip has 24061one or more modules in it; each module has a core CPU and a variety of 24062coprocessors, optional instructions, and peripherals. The 24063@code{MeP-Integrator} tool, not part of GCC, provides these 24064configurations through this option; using this option is the same as 24065using all the corresponding command-line options. The default 24066configuration is @samp{default}. 24067 24068@item -mcop 24069@opindex mcop 24070Enables the coprocessor instructions. By default, this is a 32-bit 24071coprocessor. Note that the coprocessor is normally enabled via the 24072@option{-mconfig=} option. 24073 24074@item -mcop32 24075@opindex mcop32 24076Enables the 32-bit coprocessor's instructions. 24077 24078@item -mcop64 24079@opindex mcop64 24080Enables the 64-bit coprocessor's instructions. 24081 24082@item -mivc2 24083@opindex mivc2 24084Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor. 24085 24086@item -mdc 24087@opindex mdc 24088Causes constant variables to be placed in the @code{.near} section. 24089 24090@item -mdiv 24091@opindex mdiv 24092Enables the @code{div} and @code{divu} instructions. 24093 24094@item -meb 24095@opindex meb 24096Generate big-endian code. 24097 24098@item -mel 24099@opindex mel 24100Generate little-endian code. 24101 24102@item -mio-volatile 24103@opindex mio-volatile 24104Tells the compiler that any variable marked with the @code{io} 24105attribute is to be considered volatile. 24106 24107@item -ml 24108@opindex ml 24109Causes variables to be assigned to the @code{.far} section by default. 24110 24111@item -mleadz 24112@opindex mleadz 24113Enables the @code{leadz} (leading zero) instruction. 24114 24115@item -mm 24116@opindex mm 24117Causes variables to be assigned to the @code{.near} section by default. 24118 24119@item -mminmax 24120@opindex mminmax 24121Enables the @code{min} and @code{max} instructions. 24122 24123@item -mmult 24124@opindex mmult 24125Enables the multiplication and multiply-accumulate instructions. 24126 24127@item -mno-opts 24128@opindex mno-opts 24129Disables all the optional instructions enabled by @option{-mall-opts}. 24130 24131@item -mrepeat 24132@opindex mrepeat 24133Enables the @code{repeat} and @code{erepeat} instructions, used for 24134low-overhead looping. 24135 24136@item -ms 24137@opindex ms 24138Causes all variables to default to the @code{.tiny} section. Note 24139that there is a 65536-byte limit to this section. Accesses to these 24140variables use the @code{%gp} base register. 24141 24142@item -msatur 24143@opindex msatur 24144Enables the saturation instructions. Note that the compiler does not 24145currently generate these itself, but this option is included for 24146compatibility with other tools, like @code{as}. 24147 24148@item -msdram 24149@opindex msdram 24150Link the SDRAM-based runtime instead of the default ROM-based runtime. 24151 24152@item -msim 24153@opindex msim 24154Link the simulator run-time libraries. 24155 24156@item -msimnovec 24157@opindex msimnovec 24158Link the simulator runtime libraries, excluding built-in support 24159for reset and exception vectors and tables. 24160 24161@item -mtf 24162@opindex mtf 24163Causes all functions to default to the @code{.far} section. Without 24164this option, functions default to the @code{.near} section. 24165 24166@item -mtiny=@var{n} 24167@opindex mtiny= 24168Variables that are @var{n} bytes or smaller are allocated to the 24169@code{.tiny} section. These variables use the @code{$gp} base 24170register. The default for this option is 4, but note that there's a 2417165536-byte limit to the @code{.tiny} section. 24172 24173@end table 24174 24175@node MicroBlaze Options 24176@subsection MicroBlaze Options 24177@cindex MicroBlaze Options 24178 24179@table @gcctabopt 24180 24181@item -msoft-float 24182@opindex msoft-float 24183Use software emulation for floating point (default). 24184 24185@item -mhard-float 24186@opindex mhard-float 24187Use hardware floating-point instructions. 24188 24189@item -mmemcpy 24190@opindex mmemcpy 24191Do not optimize block moves, use @code{memcpy}. 24192 24193@item -mno-clearbss 24194@opindex mno-clearbss 24195This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead. 24196 24197@item -mcpu=@var{cpu-type} 24198@opindex mcpu= 24199Use features of, and schedule code for, the given CPU. 24200Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}}, 24201where @var{X} is a major version, @var{YY} is the minor version, and 24202@var{Z} is compatibility code. Example values are @samp{v3.00.a}, 24203@samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v6.00.a}. 24204 24205@item -mxl-soft-mul 24206@opindex mxl-soft-mul 24207Use software multiply emulation (default). 24208 24209@item -mxl-soft-div 24210@opindex mxl-soft-div 24211Use software emulation for divides (default). 24212 24213@item -mxl-barrel-shift 24214@opindex mxl-barrel-shift 24215Use the hardware barrel shifter. 24216 24217@item -mxl-pattern-compare 24218@opindex mxl-pattern-compare 24219Use pattern compare instructions. 24220 24221@item -msmall-divides 24222@opindex msmall-divides 24223Use table lookup optimization for small signed integer divisions. 24224 24225@item -mxl-stack-check 24226@opindex mxl-stack-check 24227This option is deprecated. Use @option{-fstack-check} instead. 24228 24229@item -mxl-gp-opt 24230@opindex mxl-gp-opt 24231Use GP-relative @code{.sdata}/@code{.sbss} sections. 24232 24233@item -mxl-multiply-high 24234@opindex mxl-multiply-high 24235Use multiply high instructions for high part of 32x32 multiply. 24236 24237@item -mxl-float-convert 24238@opindex mxl-float-convert 24239Use hardware floating-point conversion instructions. 24240 24241@item -mxl-float-sqrt 24242@opindex mxl-float-sqrt 24243Use hardware floating-point square root instruction. 24244 24245@item -mbig-endian 24246@opindex mbig-endian 24247Generate code for a big-endian target. 24248 24249@item -mlittle-endian 24250@opindex mlittle-endian 24251Generate code for a little-endian target. 24252 24253@item -mxl-reorder 24254@opindex mxl-reorder 24255Use reorder instructions (swap and byte reversed load/store). 24256 24257@item -mxl-mode-@var{app-model} 24258Select application model @var{app-model}. Valid models are 24259@table @samp 24260@item executable 24261normal executable (default), uses startup code @file{crt0.o}. 24262 24263@item -mpic-data-is-text-relative 24264@opindex mpic-data-is-text-relative 24265Assume that the displacement between the text and data segments is fixed 24266at static link time. This allows data to be referenced by offset from start of 24267text address instead of GOT since PC-relative addressing is not supported. 24268 24269@item xmdstub 24270for use with Xilinx Microprocessor Debugger (XMD) based 24271software intrusive debug agent called xmdstub. This uses startup file 24272@file{crt1.o} and sets the start address of the program to 0x800. 24273 24274@item bootstrap 24275for applications that are loaded using a bootloader. 24276This model uses startup file @file{crt2.o} which does not contain a processor 24277reset vector handler. This is suitable for transferring control on a 24278processor reset to the bootloader rather than the application. 24279 24280@item novectors 24281for applications that do not require any of the 24282MicroBlaze vectors. This option may be useful for applications running 24283within a monitoring application. This model uses @file{crt3.o} as a startup file. 24284@end table 24285 24286Option @option{-xl-mode-@var{app-model}} is a deprecated alias for 24287@option{-mxl-mode-@var{app-model}}. 24288 24289@end table 24290 24291@node MIPS Options 24292@subsection MIPS Options 24293@cindex MIPS options 24294 24295@table @gcctabopt 24296 24297@item -EB 24298@opindex EB 24299Generate big-endian code. 24300 24301@item -EL 24302@opindex EL 24303Generate little-endian code. This is the default for @samp{mips*el-*-*} 24304configurations. 24305 24306@item -march=@var{arch} 24307@opindex march 24308Generate code that runs on @var{arch}, which can be the name of a 24309generic MIPS ISA, or the name of a particular processor. 24310The ISA names are: 24311@samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4}, 24312@samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5}, 24313@samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3}, 24314@samp{mips64r5} and @samp{mips64r6}. 24315The processor names are: 24316@samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc}, 24317@samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd}, 24318@samp{5kc}, @samp{5kf}, 24319@samp{20kc}, 24320@samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1}, 24321@samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1}, 24322@samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn}, 24323@samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2}, 24324@samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1}, 24325@samp{i6400}, @samp{i6500}, 24326@samp{interaptiv}, 24327@samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a}, @samp{gs464}, 24328@samp{gs464e}, @samp{gs264e}, 24329@samp{m4k}, 24330@samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec}, 24331@samp{m5100}, @samp{m5101}, 24332@samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3}, 24333@samp{orion}, 24334@samp{p5600}, @samp{p6600}, 24335@samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400}, 24336@samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r5900}, 24337@samp{r6000}, @samp{r8000}, 24338@samp{rm7000}, @samp{rm9000}, 24339@samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000}, 24340@samp{sb1}, 24341@samp{sr71000}, 24342@samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300}, 24343@samp{vr5000}, @samp{vr5400}, @samp{vr5500}, 24344@samp{xlr} and @samp{xlp}. 24345The special value @samp{from-abi} selects the 24346most compatible architecture for the selected ABI (that is, 24347@samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@. 24348 24349The native Linux/GNU toolchain also supports the value @samp{native}, 24350which selects the best architecture option for the host processor. 24351@option{-march=native} has no effect if GCC does not recognize 24352the processor. 24353 24354In processor names, a final @samp{000} can be abbreviated as @samp{k} 24355(for example, @option{-march=r2k}). Prefixes are optional, and 24356@samp{vr} may be written @samp{r}. 24357 24358Names of the form @samp{@var{n}f2_1} refer to processors with 24359FPUs clocked at half the rate of the core, names of the form 24360@samp{@var{n}f1_1} refer to processors with FPUs clocked at the same 24361rate as the core, and names of the form @samp{@var{n}f3_2} refer to 24362processors with FPUs clocked a ratio of 3:2 with respect to the core. 24363For compatibility reasons, @samp{@var{n}f} is accepted as a synonym 24364for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are 24365accepted as synonyms for @samp{@var{n}f1_1}. 24366 24367GCC defines two macros based on the value of this option. The first 24368is @code{_MIPS_ARCH}, which gives the name of target architecture, as 24369a string. The second has the form @code{_MIPS_ARCH_@var{foo}}, 24370where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@. 24371For example, @option{-march=r2000} sets @code{_MIPS_ARCH} 24372to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}. 24373 24374Note that the @code{_MIPS_ARCH} macro uses the processor names given 24375above. In other words, it has the full prefix and does not 24376abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi}, 24377the macro names the resolved architecture (either @code{"mips1"} or 24378@code{"mips3"}). It names the default architecture when no 24379@option{-march} option is given. 24380 24381@item -mtune=@var{arch} 24382@opindex mtune 24383Optimize for @var{arch}. Among other things, this option controls 24384the way instructions are scheduled, and the perceived cost of arithmetic 24385operations. The list of @var{arch} values is the same as for 24386@option{-march}. 24387 24388When this option is not used, GCC optimizes for the processor 24389specified by @option{-march}. By using @option{-march} and 24390@option{-mtune} together, it is possible to generate code that 24391runs on a family of processors, but optimize the code for one 24392particular member of that family. 24393 24394@option{-mtune} defines the macros @code{_MIPS_TUNE} and 24395@code{_MIPS_TUNE_@var{foo}}, which work in the same way as the 24396@option{-march} ones described above. 24397 24398@item -mips1 24399@opindex mips1 24400Equivalent to @option{-march=mips1}. 24401 24402@item -mips2 24403@opindex mips2 24404Equivalent to @option{-march=mips2}. 24405 24406@item -mips3 24407@opindex mips3 24408Equivalent to @option{-march=mips3}. 24409 24410@item -mips4 24411@opindex mips4 24412Equivalent to @option{-march=mips4}. 24413 24414@item -mips32 24415@opindex mips32 24416Equivalent to @option{-march=mips32}. 24417 24418@item -mips32r3 24419@opindex mips32r3 24420Equivalent to @option{-march=mips32r3}. 24421 24422@item -mips32r5 24423@opindex mips32r5 24424Equivalent to @option{-march=mips32r5}. 24425 24426@item -mips32r6 24427@opindex mips32r6 24428Equivalent to @option{-march=mips32r6}. 24429 24430@item -mips64 24431@opindex mips64 24432Equivalent to @option{-march=mips64}. 24433 24434@item -mips64r2 24435@opindex mips64r2 24436Equivalent to @option{-march=mips64r2}. 24437 24438@item -mips64r3 24439@opindex mips64r3 24440Equivalent to @option{-march=mips64r3}. 24441 24442@item -mips64r5 24443@opindex mips64r5 24444Equivalent to @option{-march=mips64r5}. 24445 24446@item -mips64r6 24447@opindex mips64r6 24448Equivalent to @option{-march=mips64r6}. 24449 24450@item -mips16 24451@itemx -mno-mips16 24452@opindex mips16 24453@opindex mno-mips16 24454Generate (do not generate) MIPS16 code. If GCC is targeting a 24455MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@. 24456 24457MIPS16 code generation can also be controlled on a per-function basis 24458by means of @code{mips16} and @code{nomips16} attributes. 24459@xref{Function Attributes}, for more information. 24460 24461@item -mflip-mips16 24462@opindex mflip-mips16 24463Generate MIPS16 code on alternating functions. This option is provided 24464for regression testing of mixed MIPS16/non-MIPS16 code generation, and is 24465not intended for ordinary use in compiling user code. 24466 24467@item -minterlink-compressed 24468@itemx -mno-interlink-compressed 24469@opindex minterlink-compressed 24470@opindex mno-interlink-compressed 24471Require (do not require) that code using the standard (uncompressed) MIPS ISA 24472be link-compatible with MIPS16 and microMIPS code, and vice versa. 24473 24474For example, code using the standard ISA encoding cannot jump directly 24475to MIPS16 or microMIPS code; it must either use a call or an indirect jump. 24476@option{-minterlink-compressed} therefore disables direct jumps unless GCC 24477knows that the target of the jump is not compressed. 24478 24479@item -minterlink-mips16 24480@itemx -mno-interlink-mips16 24481@opindex minterlink-mips16 24482@opindex mno-interlink-mips16 24483Aliases of @option{-minterlink-compressed} and 24484@option{-mno-interlink-compressed}. These options predate the microMIPS ASE 24485and are retained for backwards compatibility. 24486 24487@item -mabi=32 24488@itemx -mabi=o64 24489@itemx -mabi=n32 24490@itemx -mabi=64 24491@itemx -mabi=eabi 24492@opindex mabi=32 24493@opindex mabi=o64 24494@opindex mabi=n32 24495@opindex mabi=64 24496@opindex mabi=eabi 24497Generate code for the given ABI@. 24498 24499Note that the EABI has a 32-bit and a 64-bit variant. GCC normally 24500generates 64-bit code when you select a 64-bit architecture, but you 24501can use @option{-mgp32} to get 32-bit code instead. 24502 24503For information about the O64 ABI, see 24504@uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}. 24505 24506GCC supports a variant of the o32 ABI in which floating-point registers 24507are 64 rather than 32 bits wide. You can select this combination with 24508@option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1} 24509and @code{mfhc1} instructions and is therefore only supported for 24510MIPS32R2, MIPS32R3 and MIPS32R5 processors. 24511 24512The register assignments for arguments and return values remain the 24513same, but each scalar value is passed in a single 64-bit register 24514rather than a pair of 32-bit registers. For example, scalar 24515floating-point values are returned in @samp{$f0} only, not a 24516@samp{$f0}/@samp{$f1} pair. The set of call-saved registers also 24517remains the same in that the even-numbered double-precision registers 24518are saved. 24519 24520Two additional variants of the o32 ABI are supported to enable 24521a transition from 32-bit to 64-bit registers. These are FPXX 24522(@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}). 24523The FPXX extension mandates that all code must execute correctly 24524when run using 32-bit or 64-bit registers. The code can be interlinked 24525with either FP32 or FP64, but not both. 24526The FP64A extension is similar to the FP64 extension but forbids the 24527use of odd-numbered single-precision registers. This can be used 24528in conjunction with the @code{FRE} mode of FPUs in MIPS32R5 24529processors and allows both FP32 and FP64A code to interlink and 24530run in the same process without changing FPU modes. 24531 24532@item -mabicalls 24533@itemx -mno-abicalls 24534@opindex mabicalls 24535@opindex mno-abicalls 24536Generate (do not generate) code that is suitable for SVR4-style 24537dynamic objects. @option{-mabicalls} is the default for SVR4-based 24538systems. 24539 24540@item -mshared 24541@itemx -mno-shared 24542Generate (do not generate) code that is fully position-independent, 24543and that can therefore be linked into shared libraries. This option 24544only affects @option{-mabicalls}. 24545 24546All @option{-mabicalls} code has traditionally been position-independent, 24547regardless of options like @option{-fPIC} and @option{-fpic}. However, 24548as an extension, the GNU toolchain allows executables to use absolute 24549accesses for locally-binding symbols. It can also use shorter GP 24550initialization sequences and generate direct calls to locally-defined 24551functions. This mode is selected by @option{-mno-shared}. 24552 24553@option{-mno-shared} depends on binutils 2.16 or higher and generates 24554objects that can only be linked by the GNU linker. However, the option 24555does not affect the ABI of the final executable; it only affects the ABI 24556of relocatable objects. Using @option{-mno-shared} generally makes 24557executables both smaller and quicker. 24558 24559@option{-mshared} is the default. 24560 24561@item -mplt 24562@itemx -mno-plt 24563@opindex mplt 24564@opindex mno-plt 24565Assume (do not assume) that the static and dynamic linkers 24566support PLTs and copy relocations. This option only affects 24567@option{-mno-shared -mabicalls}. For the n64 ABI, this option 24568has no effect without @option{-msym32}. 24569 24570You can make @option{-mplt} the default by configuring 24571GCC with @option{--with-mips-plt}. The default is 24572@option{-mno-plt} otherwise. 24573 24574@item -mxgot 24575@itemx -mno-xgot 24576@opindex mxgot 24577@opindex mno-xgot 24578Lift (do not lift) the usual restrictions on the size of the global 24579offset table. 24580 24581GCC normally uses a single instruction to load values from the GOT@. 24582While this is relatively efficient, it only works if the GOT 24583is smaller than about 64k. Anything larger causes the linker 24584to report an error such as: 24585 24586@cindex relocation truncated to fit (MIPS) 24587@smallexample 24588relocation truncated to fit: R_MIPS_GOT16 foobar 24589@end smallexample 24590 24591If this happens, you should recompile your code with @option{-mxgot}. 24592This works with very large GOTs, although the code is also 24593less efficient, since it takes three instructions to fetch the 24594value of a global symbol. 24595 24596Note that some linkers can create multiple GOTs. If you have such a 24597linker, you should only need to use @option{-mxgot} when a single object 24598file accesses more than 64k's worth of GOT entries. Very few do. 24599 24600These options have no effect unless GCC is generating position 24601independent code. 24602 24603@item -mgp32 24604@opindex mgp32 24605Assume that general-purpose registers are 32 bits wide. 24606 24607@item -mgp64 24608@opindex mgp64 24609Assume that general-purpose registers are 64 bits wide. 24610 24611@item -mfp32 24612@opindex mfp32 24613Assume that floating-point registers are 32 bits wide. 24614 24615@item -mfp64 24616@opindex mfp64 24617Assume that floating-point registers are 64 bits wide. 24618 24619@item -mfpxx 24620@opindex mfpxx 24621Do not assume the width of floating-point registers. 24622 24623@item -mhard-float 24624@opindex mhard-float 24625Use floating-point coprocessor instructions. 24626 24627@item -msoft-float 24628@opindex msoft-float 24629Do not use floating-point coprocessor instructions. Implement 24630floating-point calculations using library calls instead. 24631 24632@item -mno-float 24633@opindex mno-float 24634Equivalent to @option{-msoft-float}, but additionally asserts that the 24635program being compiled does not perform any floating-point operations. 24636This option is presently supported only by some bare-metal MIPS 24637configurations, where it may select a special set of libraries 24638that lack all floating-point support (including, for example, the 24639floating-point @code{printf} formats). 24640If code compiled with @option{-mno-float} accidentally contains 24641floating-point operations, it is likely to suffer a link-time 24642or run-time failure. 24643 24644@item -msingle-float 24645@opindex msingle-float 24646Assume that the floating-point coprocessor only supports single-precision 24647operations. 24648 24649@item -mdouble-float 24650@opindex mdouble-float 24651Assume that the floating-point coprocessor supports double-precision 24652operations. This is the default. 24653 24654@item -modd-spreg 24655@itemx -mno-odd-spreg 24656@opindex modd-spreg 24657@opindex mno-odd-spreg 24658Enable the use of odd-numbered single-precision floating-point registers 24659for the o32 ABI. This is the default for processors that are known to 24660support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg} 24661is set by default. 24662 24663@item -mabs=2008 24664@itemx -mabs=legacy 24665@opindex mabs=2008 24666@opindex mabs=legacy 24667These options control the treatment of the special not-a-number (NaN) 24668IEEE 754 floating-point data with the @code{abs.@i{fmt}} and 24669@code{neg.@i{fmt}} machine instructions. 24670 24671By default or when @option{-mabs=legacy} is used the legacy 24672treatment is selected. In this case these instructions are considered 24673arithmetic and avoided where correct operation is required and the 24674input operand might be a NaN. A longer sequence of instructions that 24675manipulate the sign bit of floating-point datum manually is used 24676instead unless the @option{-ffinite-math-only} option has also been 24677specified. 24678 24679The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In 24680this case these instructions are considered non-arithmetic and therefore 24681operating correctly in all cases, including in particular where the 24682input operand is a NaN. These instructions are therefore always used 24683for the respective operations. 24684 24685@item -mnan=2008 24686@itemx -mnan=legacy 24687@opindex mnan=2008 24688@opindex mnan=legacy 24689These options control the encoding of the special not-a-number (NaN) 24690IEEE 754 floating-point data. 24691 24692The @option{-mnan=legacy} option selects the legacy encoding. In this 24693case quiet NaNs (qNaNs) are denoted by the first bit of their trailing 24694significand field being 0, whereas signaling NaNs (sNaNs) are denoted 24695by the first bit of their trailing significand field being 1. 24696 24697The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In 24698this case qNaNs are denoted by the first bit of their trailing 24699significand field being 1, whereas sNaNs are denoted by the first bit of 24700their trailing significand field being 0. 24701 24702The default is @option{-mnan=legacy} unless GCC has been configured with 24703@option{--with-nan=2008}. 24704 24705@item -mllsc 24706@itemx -mno-llsc 24707@opindex mllsc 24708@opindex mno-llsc 24709Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to 24710implement atomic memory built-in functions. When neither option is 24711specified, GCC uses the instructions if the target architecture 24712supports them. 24713 24714@option{-mllsc} is useful if the runtime environment can emulate the 24715instructions and @option{-mno-llsc} can be useful when compiling for 24716nonstandard ISAs. You can make either option the default by 24717configuring GCC with @option{--with-llsc} and @option{--without-llsc} 24718respectively. @option{--with-llsc} is the default for some 24719configurations; see the installation documentation for details. 24720 24721@item -mdsp 24722@itemx -mno-dsp 24723@opindex mdsp 24724@opindex mno-dsp 24725Use (do not use) revision 1 of the MIPS DSP ASE@. 24726@xref{MIPS DSP Built-in Functions}. This option defines the 24727preprocessor macro @code{__mips_dsp}. It also defines 24728@code{__mips_dsp_rev} to 1. 24729 24730@item -mdspr2 24731@itemx -mno-dspr2 24732@opindex mdspr2 24733@opindex mno-dspr2 24734Use (do not use) revision 2 of the MIPS DSP ASE@. 24735@xref{MIPS DSP Built-in Functions}. This option defines the 24736preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}. 24737It also defines @code{__mips_dsp_rev} to 2. 24738 24739@item -msmartmips 24740@itemx -mno-smartmips 24741@opindex msmartmips 24742@opindex mno-smartmips 24743Use (do not use) the MIPS SmartMIPS ASE. 24744 24745@item -mpaired-single 24746@itemx -mno-paired-single 24747@opindex mpaired-single 24748@opindex mno-paired-single 24749Use (do not use) paired-single floating-point instructions. 24750@xref{MIPS Paired-Single Support}. This option requires 24751hardware floating-point support to be enabled. 24752 24753@item -mdmx 24754@itemx -mno-mdmx 24755@opindex mdmx 24756@opindex mno-mdmx 24757Use (do not use) MIPS Digital Media Extension instructions. 24758This option can only be used when generating 64-bit code and requires 24759hardware floating-point support to be enabled. 24760 24761@item -mips3d 24762@itemx -mno-mips3d 24763@opindex mips3d 24764@opindex mno-mips3d 24765Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}. 24766The option @option{-mips3d} implies @option{-mpaired-single}. 24767 24768@item -mmicromips 24769@itemx -mno-micromips 24770@opindex mmicromips 24771@opindex mno-mmicromips 24772Generate (do not generate) microMIPS code. 24773 24774MicroMIPS code generation can also be controlled on a per-function basis 24775by means of @code{micromips} and @code{nomicromips} attributes. 24776@xref{Function Attributes}, for more information. 24777 24778@item -mmt 24779@itemx -mno-mt 24780@opindex mmt 24781@opindex mno-mt 24782Use (do not use) MT Multithreading instructions. 24783 24784@item -mmcu 24785@itemx -mno-mcu 24786@opindex mmcu 24787@opindex mno-mcu 24788Use (do not use) the MIPS MCU ASE instructions. 24789 24790@item -meva 24791@itemx -mno-eva 24792@opindex meva 24793@opindex mno-eva 24794Use (do not use) the MIPS Enhanced Virtual Addressing instructions. 24795 24796@item -mvirt 24797@itemx -mno-virt 24798@opindex mvirt 24799@opindex mno-virt 24800Use (do not use) the MIPS Virtualization (VZ) instructions. 24801 24802@item -mxpa 24803@itemx -mno-xpa 24804@opindex mxpa 24805@opindex mno-xpa 24806Use (do not use) the MIPS eXtended Physical Address (XPA) instructions. 24807 24808@item -mcrc 24809@itemx -mno-crc 24810@opindex mcrc 24811@opindex mno-crc 24812Use (do not use) the MIPS Cyclic Redundancy Check (CRC) instructions. 24813 24814@item -mginv 24815@itemx -mno-ginv 24816@opindex mginv 24817@opindex mno-ginv 24818Use (do not use) the MIPS Global INValidate (GINV) instructions. 24819 24820@item -mloongson-mmi 24821@itemx -mno-loongson-mmi 24822@opindex mloongson-mmi 24823@opindex mno-loongson-mmi 24824Use (do not use) the MIPS Loongson MultiMedia extensions Instructions (MMI). 24825 24826@item -mloongson-ext 24827@itemx -mno-loongson-ext 24828@opindex mloongson-ext 24829@opindex mno-loongson-ext 24830Use (do not use) the MIPS Loongson EXTensions (EXT) instructions. 24831 24832@item -mloongson-ext2 24833@itemx -mno-loongson-ext2 24834@opindex mloongson-ext2 24835@opindex mno-loongson-ext2 24836Use (do not use) the MIPS Loongson EXTensions r2 (EXT2) instructions. 24837 24838@item -mlong64 24839@opindex mlong64 24840Force @code{long} types to be 64 bits wide. See @option{-mlong32} for 24841an explanation of the default and the way that the pointer size is 24842determined. 24843 24844@item -mlong32 24845@opindex mlong32 24846Force @code{long}, @code{int}, and pointer types to be 32 bits wide. 24847 24848The default size of @code{int}s, @code{long}s and pointers depends on 24849the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI 24850uses 64-bit @code{long}s, as does the 64-bit EABI; the others use 2485132-bit @code{long}s. Pointers are the same size as @code{long}s, 24852or the same size as integer registers, whichever is smaller. 24853 24854@item -msym32 24855@itemx -mno-sym32 24856@opindex msym32 24857@opindex mno-sym32 24858Assume (do not assume) that all symbols have 32-bit values, regardless 24859of the selected ABI@. This option is useful in combination with 24860@option{-mabi=64} and @option{-mno-abicalls} because it allows GCC 24861to generate shorter and faster references to symbolic addresses. 24862 24863@item -G @var{num} 24864@opindex G 24865Put definitions of externally-visible data in a small data section 24866if that data is no bigger than @var{num} bytes. GCC can then generate 24867more efficient accesses to the data; see @option{-mgpopt} for details. 24868 24869The default @option{-G} option depends on the configuration. 24870 24871@item -mlocal-sdata 24872@itemx -mno-local-sdata 24873@opindex mlocal-sdata 24874@opindex mno-local-sdata 24875Extend (do not extend) the @option{-G} behavior to local data too, 24876such as to static variables in C@. @option{-mlocal-sdata} is the 24877default for all configurations. 24878 24879If the linker complains that an application is using too much small data, 24880you might want to try rebuilding the less performance-critical parts with 24881@option{-mno-local-sdata}. You might also want to build large 24882libraries with @option{-mno-local-sdata}, so that the libraries leave 24883more room for the main program. 24884 24885@item -mextern-sdata 24886@itemx -mno-extern-sdata 24887@opindex mextern-sdata 24888@opindex mno-extern-sdata 24889Assume (do not assume) that externally-defined data is in 24890a small data section if the size of that data is within the @option{-G} limit. 24891@option{-mextern-sdata} is the default for all configurations. 24892 24893If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G 24894@var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var} 24895that is no bigger than @var{num} bytes, you must make sure that @var{Var} 24896is placed in a small data section. If @var{Var} is defined by another 24897module, you must either compile that module with a high-enough 24898@option{-G} setting or attach a @code{section} attribute to @var{Var}'s 24899definition. If @var{Var} is common, you must link the application 24900with a high-enough @option{-G} setting. 24901 24902The easiest way of satisfying these restrictions is to compile 24903and link every module with the same @option{-G} option. However, 24904you may wish to build a library that supports several different 24905small data limits. You can do this by compiling the library with 24906the highest supported @option{-G} setting and additionally using 24907@option{-mno-extern-sdata} to stop the library from making assumptions 24908about externally-defined data. 24909 24910@item -mgpopt 24911@itemx -mno-gpopt 24912@opindex mgpopt 24913@opindex mno-gpopt 24914Use (do not use) GP-relative accesses for symbols that are known to be 24915in a small data section; see @option{-G}, @option{-mlocal-sdata} and 24916@option{-mextern-sdata}. @option{-mgpopt} is the default for all 24917configurations. 24918 24919@option{-mno-gpopt} is useful for cases where the @code{$gp} register 24920might not hold the value of @code{_gp}. For example, if the code is 24921part of a library that might be used in a boot monitor, programs that 24922call boot monitor routines pass an unknown value in @code{$gp}. 24923(In such situations, the boot monitor itself is usually compiled 24924with @option{-G0}.) 24925 24926@option{-mno-gpopt} implies @option{-mno-local-sdata} and 24927@option{-mno-extern-sdata}. 24928 24929@item -membedded-data 24930@itemx -mno-embedded-data 24931@opindex membedded-data 24932@opindex mno-embedded-data 24933Allocate variables to the read-only data section first if possible, then 24934next in the small data section if possible, otherwise in data. This gives 24935slightly slower code than the default, but reduces the amount of RAM required 24936when executing, and thus may be preferred for some embedded systems. 24937 24938@item -muninit-const-in-rodata 24939@itemx -mno-uninit-const-in-rodata 24940@opindex muninit-const-in-rodata 24941@opindex mno-uninit-const-in-rodata 24942Put uninitialized @code{const} variables in the read-only data section. 24943This option is only meaningful in conjunction with @option{-membedded-data}. 24944 24945@item -mcode-readable=@var{setting} 24946@opindex mcode-readable 24947Specify whether GCC may generate code that reads from executable sections. 24948There are three possible settings: 24949 24950@table @gcctabopt 24951@item -mcode-readable=yes 24952Instructions may freely access executable sections. This is the 24953default setting. 24954 24955@item -mcode-readable=pcrel 24956MIPS16 PC-relative load instructions can access executable sections, 24957but other instructions must not do so. This option is useful on 4KSc 24958and 4KSd processors when the code TLBs have the Read Inhibit bit set. 24959It is also useful on processors that can be configured to have a dual 24960instruction/data SRAM interface and that, like the M4K, automatically 24961redirect PC-relative loads to the instruction RAM. 24962 24963@item -mcode-readable=no 24964Instructions must not access executable sections. This option can be 24965useful on targets that are configured to have a dual instruction/data 24966SRAM interface but that (unlike the M4K) do not automatically redirect 24967PC-relative loads to the instruction RAM. 24968@end table 24969 24970@item -msplit-addresses 24971@itemx -mno-split-addresses 24972@opindex msplit-addresses 24973@opindex mno-split-addresses 24974Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler 24975relocation operators. This option has been superseded by 24976@option{-mexplicit-relocs} but is retained for backwards compatibility. 24977 24978@item -mexplicit-relocs 24979@itemx -mno-explicit-relocs 24980@opindex mexplicit-relocs 24981@opindex mno-explicit-relocs 24982Use (do not use) assembler relocation operators when dealing with symbolic 24983addresses. The alternative, selected by @option{-mno-explicit-relocs}, 24984is to use assembler macros instead. 24985 24986@option{-mexplicit-relocs} is the default if GCC was configured 24987to use an assembler that supports relocation operators. 24988 24989@item -mcheck-zero-division 24990@itemx -mno-check-zero-division 24991@opindex mcheck-zero-division 24992@opindex mno-check-zero-division 24993Trap (do not trap) on integer division by zero. 24994 24995The default is @option{-mcheck-zero-division}. 24996 24997@item -mdivide-traps 24998@itemx -mdivide-breaks 24999@opindex mdivide-traps 25000@opindex mdivide-breaks 25001MIPS systems check for division by zero by generating either a 25002conditional trap or a break instruction. Using traps results in 25003smaller code, but is only supported on MIPS II and later. Also, some 25004versions of the Linux kernel have a bug that prevents trap from 25005generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to 25006allow conditional traps on architectures that support them and 25007@option{-mdivide-breaks} to force the use of breaks. 25008 25009The default is usually @option{-mdivide-traps}, but this can be 25010overridden at configure time using @option{--with-divide=breaks}. 25011Divide-by-zero checks can be completely disabled using 25012@option{-mno-check-zero-division}. 25013 25014@item -mload-store-pairs 25015@itemx -mno-load-store-pairs 25016@opindex mload-store-pairs 25017@opindex mno-load-store-pairs 25018Enable (disable) an optimization that pairs consecutive load or store 25019instructions to enable load/store bonding. This option is enabled by 25020default but only takes effect when the selected architecture is known 25021to support bonding. 25022 25023@item -mmemcpy 25024@itemx -mno-memcpy 25025@opindex mmemcpy 25026@opindex mno-memcpy 25027Force (do not force) the use of @code{memcpy} for non-trivial block 25028moves. The default is @option{-mno-memcpy}, which allows GCC to inline 25029most constant-sized copies. 25030 25031@item -mlong-calls 25032@itemx -mno-long-calls 25033@opindex mlong-calls 25034@opindex mno-long-calls 25035Disable (do not disable) use of the @code{jal} instruction. Calling 25036functions using @code{jal} is more efficient but requires the caller 25037and callee to be in the same 256 megabyte segment. 25038 25039This option has no effect on abicalls code. The default is 25040@option{-mno-long-calls}. 25041 25042@item -mmad 25043@itemx -mno-mad 25044@opindex mmad 25045@opindex mno-mad 25046Enable (disable) use of the @code{mad}, @code{madu} and @code{mul} 25047instructions, as provided by the R4650 ISA@. 25048 25049@item -mimadd 25050@itemx -mno-imadd 25051@opindex mimadd 25052@opindex mno-imadd 25053Enable (disable) use of the @code{madd} and @code{msub} integer 25054instructions. The default is @option{-mimadd} on architectures 25055that support @code{madd} and @code{msub} except for the 74k 25056architecture where it was found to generate slower code. 25057 25058@item -mfused-madd 25059@itemx -mno-fused-madd 25060@opindex mfused-madd 25061@opindex mno-fused-madd 25062Enable (disable) use of the floating-point multiply-accumulate 25063instructions, when they are available. The default is 25064@option{-mfused-madd}. 25065 25066On the R8000 CPU when multiply-accumulate instructions are used, 25067the intermediate product is calculated to infinite precision 25068and is not subject to the FCSR Flush to Zero bit. This may be 25069undesirable in some circumstances. On other processors the result 25070is numerically identical to the equivalent computation using 25071separate multiply, add, subtract and negate instructions. 25072 25073@item -nocpp 25074@opindex nocpp 25075Tell the MIPS assembler to not run its preprocessor over user 25076assembler files (with a @samp{.s} suffix) when assembling them. 25077 25078@item -mfix-24k 25079@itemx -mno-fix-24k 25080@opindex mfix-24k 25081@opindex mno-fix-24k 25082Work around the 24K E48 (lost data on stores during refill) errata. 25083The workarounds are implemented by the assembler rather than by GCC@. 25084 25085@item -mfix-r4000 25086@itemx -mno-fix-r4000 25087@opindex mfix-r4000 25088@opindex mno-fix-r4000 25089Work around certain R4000 CPU errata: 25090@itemize @minus 25091@item 25092A double-word or a variable shift may give an incorrect result if executed 25093immediately after starting an integer division. 25094@item 25095A double-word or a variable shift may give an incorrect result if executed 25096while an integer multiplication is in progress. 25097@item 25098An integer division may give an incorrect result if started in a delay slot 25099of a taken branch or a jump. 25100@end itemize 25101 25102@item -mfix-r4400 25103@itemx -mno-fix-r4400 25104@opindex mfix-r4400 25105@opindex mno-fix-r4400 25106Work around certain R4400 CPU errata: 25107@itemize @minus 25108@item 25109A double-word or a variable shift may give an incorrect result if executed 25110immediately after starting an integer division. 25111@end itemize 25112 25113@item -mfix-r10000 25114@itemx -mno-fix-r10000 25115@opindex mfix-r10000 25116@opindex mno-fix-r10000 25117Work around certain R10000 errata: 25118@itemize @minus 25119@item 25120@code{ll}/@code{sc} sequences may not behave atomically on revisions 25121prior to 3.0. They may deadlock on revisions 2.6 and earlier. 25122@end itemize 25123 25124This option can only be used if the target architecture supports 25125branch-likely instructions. @option{-mfix-r10000} is the default when 25126@option{-march=r10000} is used; @option{-mno-fix-r10000} is the default 25127otherwise. 25128 25129@item -mfix-r5900 25130@itemx -mno-fix-r5900 25131@opindex mfix-r5900 25132Do not attempt to schedule the preceding instruction into the delay slot 25133of a branch instruction placed at the end of a short loop of six 25134instructions or fewer and always schedule a @code{nop} instruction there 25135instead. The short loop bug under certain conditions causes loops to 25136execute only once or twice, due to a hardware bug in the R5900 chip. The 25137workaround is implemented by the assembler rather than by GCC@. 25138 25139@item -mfix-rm7000 25140@itemx -mno-fix-rm7000 25141@opindex mfix-rm7000 25142Work around the RM7000 @code{dmult}/@code{dmultu} errata. The 25143workarounds are implemented by the assembler rather than by GCC@. 25144 25145@item -mfix-vr4120 25146@itemx -mno-fix-vr4120 25147@opindex mfix-vr4120 25148Work around certain VR4120 errata: 25149@itemize @minus 25150@item 25151@code{dmultu} does not always produce the correct result. 25152@item 25153@code{div} and @code{ddiv} do not always produce the correct result if one 25154of the operands is negative. 25155@end itemize 25156The workarounds for the division errata rely on special functions in 25157@file{libgcc.a}. At present, these functions are only provided by 25158the @code{mips64vr*-elf} configurations. 25159 25160Other VR4120 errata require a NOP to be inserted between certain pairs of 25161instructions. These errata are handled by the assembler, not by GCC itself. 25162 25163@item -mfix-vr4130 25164@opindex mfix-vr4130 25165Work around the VR4130 @code{mflo}/@code{mfhi} errata. The 25166workarounds are implemented by the assembler rather than by GCC, 25167although GCC avoids using @code{mflo} and @code{mfhi} if the 25168VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi} 25169instructions are available instead. 25170 25171@item -mfix-sb1 25172@itemx -mno-fix-sb1 25173@opindex mfix-sb1 25174Work around certain SB-1 CPU core errata. 25175(This flag currently works around the SB-1 revision 2 25176``F1'' and ``F2'' floating-point errata.) 25177 25178@item -mr10k-cache-barrier=@var{setting} 25179@opindex mr10k-cache-barrier 25180Specify whether GCC should insert cache barriers to avoid the 25181side effects of speculation on R10K processors. 25182 25183In common with many processors, the R10K tries to predict the outcome 25184of a conditional branch and speculatively executes instructions from 25185the ``taken'' branch. It later aborts these instructions if the 25186predicted outcome is wrong. However, on the R10K, even aborted 25187instructions can have side effects. 25188 25189This problem only affects kernel stores and, depending on the system, 25190kernel loads. As an example, a speculatively-executed store may load 25191the target memory into cache and mark the cache line as dirty, even if 25192the store itself is later aborted. If a DMA operation writes to the 25193same area of memory before the ``dirty'' line is flushed, the cached 25194data overwrites the DMA-ed data. See the R10K processor manual 25195for a full description, including other potential problems. 25196 25197One workaround is to insert cache barrier instructions before every memory 25198access that might be speculatively executed and that might have side 25199effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}} 25200controls GCC's implementation of this workaround. It assumes that 25201aborted accesses to any byte in the following regions does not have 25202side effects: 25203 25204@enumerate 25205@item 25206the memory occupied by the current function's stack frame; 25207 25208@item 25209the memory occupied by an incoming stack argument; 25210 25211@item 25212the memory occupied by an object with a link-time-constant address. 25213@end enumerate 25214 25215It is the kernel's responsibility to ensure that speculative 25216accesses to these regions are indeed safe. 25217 25218If the input program contains a function declaration such as: 25219 25220@smallexample 25221void foo (void); 25222@end smallexample 25223 25224then the implementation of @code{foo} must allow @code{j foo} and 25225@code{jal foo} to be executed speculatively. GCC honors this 25226restriction for functions it compiles itself. It expects non-GCC 25227functions (such as hand-written assembly code) to do the same. 25228 25229The option has three forms: 25230 25231@table @gcctabopt 25232@item -mr10k-cache-barrier=load-store 25233Insert a cache barrier before a load or store that might be 25234speculatively executed and that might have side effects even 25235if aborted. 25236 25237@item -mr10k-cache-barrier=store 25238Insert a cache barrier before a store that might be speculatively 25239executed and that might have side effects even if aborted. 25240 25241@item -mr10k-cache-barrier=none 25242Disable the insertion of cache barriers. This is the default setting. 25243@end table 25244 25245@item -mflush-func=@var{func} 25246@itemx -mno-flush-func 25247@opindex mflush-func 25248Specifies the function to call to flush the I and D caches, or to not 25249call any such function. If called, the function must take the same 25250arguments as the common @code{_flush_func}, that is, the address of the 25251memory range for which the cache is being flushed, the size of the 25252memory range, and the number 3 (to flush both caches). The default 25253depends on the target GCC was configured for, but commonly is either 25254@code{_flush_func} or @code{__cpu_flush}. 25255 25256@item mbranch-cost=@var{num} 25257@opindex mbranch-cost 25258Set the cost of branches to roughly @var{num} ``simple'' instructions. 25259This cost is only a heuristic and is not guaranteed to produce 25260consistent results across releases. A zero cost redundantly selects 25261the default, which is based on the @option{-mtune} setting. 25262 25263@item -mbranch-likely 25264@itemx -mno-branch-likely 25265@opindex mbranch-likely 25266@opindex mno-branch-likely 25267Enable or disable use of Branch Likely instructions, regardless of the 25268default for the selected architecture. By default, Branch Likely 25269instructions may be generated if they are supported by the selected 25270architecture. An exception is for the MIPS32 and MIPS64 architectures 25271and processors that implement those architectures; for those, Branch 25272Likely instructions are not be generated by default because the MIPS32 25273and MIPS64 architectures specifically deprecate their use. 25274 25275@item -mcompact-branches=never 25276@itemx -mcompact-branches=optimal 25277@itemx -mcompact-branches=always 25278@opindex mcompact-branches=never 25279@opindex mcompact-branches=optimal 25280@opindex mcompact-branches=always 25281These options control which form of branches will be generated. The 25282default is @option{-mcompact-branches=optimal}. 25283 25284The @option{-mcompact-branches=never} option ensures that compact branch 25285instructions will never be generated. 25286 25287The @option{-mcompact-branches=always} option ensures that a compact 25288branch instruction will be generated if available. If a compact branch 25289instruction is not available, a delay slot form of the branch will be 25290used instead. 25291 25292This option is supported from MIPS Release 6 onwards. 25293 25294The @option{-mcompact-branches=optimal} option will cause a delay slot 25295branch to be used if one is available in the current ISA and the delay 25296slot is successfully filled. If the delay slot is not filled, a compact 25297branch will be chosen if one is available. 25298 25299@item -mfp-exceptions 25300@itemx -mno-fp-exceptions 25301@opindex mfp-exceptions 25302Specifies whether FP exceptions are enabled. This affects how 25303FP instructions are scheduled for some processors. 25304The default is that FP exceptions are 25305enabled. 25306 25307For instance, on the SB-1, if FP exceptions are disabled, and we are emitting 2530864-bit code, then we can use both FP pipes. Otherwise, we can only use one 25309FP pipe. 25310 25311@item -mvr4130-align 25312@itemx -mno-vr4130-align 25313@opindex mvr4130-align 25314The VR4130 pipeline is two-way superscalar, but can only issue two 25315instructions together if the first one is 8-byte aligned. When this 25316option is enabled, GCC aligns pairs of instructions that it 25317thinks should execute in parallel. 25318 25319This option only has an effect when optimizing for the VR4130. 25320It normally makes code faster, but at the expense of making it bigger. 25321It is enabled by default at optimization level @option{-O3}. 25322 25323@item -msynci 25324@itemx -mno-synci 25325@opindex msynci 25326Enable (disable) generation of @code{synci} instructions on 25327architectures that support it. The @code{synci} instructions (if 25328enabled) are generated when @code{__builtin___clear_cache} is 25329compiled. 25330 25331This option defaults to @option{-mno-synci}, but the default can be 25332overridden by configuring GCC with @option{--with-synci}. 25333 25334When compiling code for single processor systems, it is generally safe 25335to use @code{synci}. However, on many multi-core (SMP) systems, it 25336does not invalidate the instruction caches on all cores and may lead 25337to undefined behavior. 25338 25339@item -mrelax-pic-calls 25340@itemx -mno-relax-pic-calls 25341@opindex mrelax-pic-calls 25342Try to turn PIC calls that are normally dispatched via register 25343@code{$25} into direct calls. This is only possible if the linker can 25344resolve the destination at link time and if the destination is within 25345range for a direct call. 25346 25347@option{-mrelax-pic-calls} is the default if GCC was configured to use 25348an assembler and a linker that support the @code{.reloc} assembly 25349directive and @option{-mexplicit-relocs} is in effect. With 25350@option{-mno-explicit-relocs}, this optimization can be performed by the 25351assembler and the linker alone without help from the compiler. 25352 25353@item -mmcount-ra-address 25354@itemx -mno-mcount-ra-address 25355@opindex mmcount-ra-address 25356@opindex mno-mcount-ra-address 25357Emit (do not emit) code that allows @code{_mcount} to modify the 25358calling function's return address. When enabled, this option extends 25359the usual @code{_mcount} interface with a new @var{ra-address} 25360parameter, which has type @code{intptr_t *} and is passed in register 25361@code{$12}. @code{_mcount} can then modify the return address by 25362doing both of the following: 25363@itemize 25364@item 25365Returning the new address in register @code{$31}. 25366@item 25367Storing the new address in @code{*@var{ra-address}}, 25368if @var{ra-address} is nonnull. 25369@end itemize 25370 25371The default is @option{-mno-mcount-ra-address}. 25372 25373@item -mframe-header-opt 25374@itemx -mno-frame-header-opt 25375@opindex mframe-header-opt 25376Enable (disable) frame header optimization in the o32 ABI. When using the 25377o32 ABI, calling functions will allocate 16 bytes on the stack for the called 25378function to write out register arguments. When enabled, this optimization 25379will suppress the allocation of the frame header if it can be determined that 25380it is unused. 25381 25382This optimization is off by default at all optimization levels. 25383 25384@item -mlxc1-sxc1 25385@itemx -mno-lxc1-sxc1 25386@opindex mlxc1-sxc1 25387When applicable, enable (disable) the generation of @code{lwxc1}, 25388@code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default. 25389 25390@item -mmadd4 25391@itemx -mno-madd4 25392@opindex mmadd4 25393When applicable, enable (disable) the generation of 4-operand @code{madd.s}, 25394@code{madd.d} and related instructions. Enabled by default. 25395 25396@end table 25397 25398@node MMIX Options 25399@subsection MMIX Options 25400@cindex MMIX Options 25401 25402These options are defined for the MMIX: 25403 25404@table @gcctabopt 25405@item -mlibfuncs 25406@itemx -mno-libfuncs 25407@opindex mlibfuncs 25408@opindex mno-libfuncs 25409Specify that intrinsic library functions are being compiled, passing all 25410values in registers, no matter the size. 25411 25412@item -mepsilon 25413@itemx -mno-epsilon 25414@opindex mepsilon 25415@opindex mno-epsilon 25416Generate floating-point comparison instructions that compare with respect 25417to the @code{rE} epsilon register. 25418 25419@item -mabi=mmixware 25420@itemx -mabi=gnu 25421@opindex mabi=mmixware 25422@opindex mabi=gnu 25423Generate code that passes function parameters and return values that (in 25424the called function) are seen as registers @code{$0} and up, as opposed to 25425the GNU ABI which uses global registers @code{$231} and up. 25426 25427@item -mzero-extend 25428@itemx -mno-zero-extend 25429@opindex mzero-extend 25430@opindex mno-zero-extend 25431When reading data from memory in sizes shorter than 64 bits, use (do not 25432use) zero-extending load instructions by default, rather than 25433sign-extending ones. 25434 25435@item -mknuthdiv 25436@itemx -mno-knuthdiv 25437@opindex mknuthdiv 25438@opindex mno-knuthdiv 25439Make the result of a division yielding a remainder have the same sign as 25440the divisor. With the default, @option{-mno-knuthdiv}, the sign of the 25441remainder follows the sign of the dividend. Both methods are 25442arithmetically valid, the latter being almost exclusively used. 25443 25444@item -mtoplevel-symbols 25445@itemx -mno-toplevel-symbols 25446@opindex mtoplevel-symbols 25447@opindex mno-toplevel-symbols 25448Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly 25449code can be used with the @code{PREFIX} assembly directive. 25450 25451@item -melf 25452@opindex melf 25453Generate an executable in the ELF format, rather than the default 25454@samp{mmo} format used by the @command{mmix} simulator. 25455 25456@item -mbranch-predict 25457@itemx -mno-branch-predict 25458@opindex mbranch-predict 25459@opindex mno-branch-predict 25460Use (do not use) the probable-branch instructions, when static branch 25461prediction indicates a probable branch. 25462 25463@item -mbase-addresses 25464@itemx -mno-base-addresses 25465@opindex mbase-addresses 25466@opindex mno-base-addresses 25467Generate (do not generate) code that uses @emph{base addresses}. Using a 25468base address automatically generates a request (handled by the assembler 25469and the linker) for a constant to be set up in a global register. The 25470register is used for one or more base address requests within the range 0 25471to 255 from the value held in the register. The generally leads to short 25472and fast code, but the number of different data items that can be 25473addressed is limited. This means that a program that uses lots of static 25474data may require @option{-mno-base-addresses}. 25475 25476@item -msingle-exit 25477@itemx -mno-single-exit 25478@opindex msingle-exit 25479@opindex mno-single-exit 25480Force (do not force) generated code to have a single exit point in each 25481function. 25482@end table 25483 25484@node MN10300 Options 25485@subsection MN10300 Options 25486@cindex MN10300 options 25487 25488These @option{-m} options are defined for Matsushita MN10300 architectures: 25489 25490@table @gcctabopt 25491@item -mmult-bug 25492@opindex mmult-bug 25493Generate code to avoid bugs in the multiply instructions for the MN10300 25494processors. This is the default. 25495 25496@item -mno-mult-bug 25497@opindex mno-mult-bug 25498Do not generate code to avoid bugs in the multiply instructions for the 25499MN10300 processors. 25500 25501@item -mam33 25502@opindex mam33 25503Generate code using features specific to the AM33 processor. 25504 25505@item -mno-am33 25506@opindex mno-am33 25507Do not generate code using features specific to the AM33 processor. This 25508is the default. 25509 25510@item -mam33-2 25511@opindex mam33-2 25512Generate code using features specific to the AM33/2.0 processor. 25513 25514@item -mam34 25515@opindex mam34 25516Generate code using features specific to the AM34 processor. 25517 25518@item -mtune=@var{cpu-type} 25519@opindex mtune 25520Use the timing characteristics of the indicated CPU type when 25521scheduling instructions. This does not change the targeted processor 25522type. The CPU type must be one of @samp{mn10300}, @samp{am33}, 25523@samp{am33-2} or @samp{am34}. 25524 25525@item -mreturn-pointer-on-d0 25526@opindex mreturn-pointer-on-d0 25527When generating a function that returns a pointer, return the pointer 25528in both @code{a0} and @code{d0}. Otherwise, the pointer is returned 25529only in @code{a0}, and attempts to call such functions without a prototype 25530result in errors. Note that this option is on by default; use 25531@option{-mno-return-pointer-on-d0} to disable it. 25532 25533@item -mno-crt0 25534@opindex mno-crt0 25535Do not link in the C run-time initialization object file. 25536 25537@item -mrelax 25538@opindex mrelax 25539Indicate to the linker that it should perform a relaxation optimization pass 25540to shorten branches, calls and absolute memory addresses. This option only 25541has an effect when used on the command line for the final link step. 25542 25543This option makes symbolic debugging impossible. 25544 25545@item -mliw 25546@opindex mliw 25547Allow the compiler to generate @emph{Long Instruction Word} 25548instructions if the target is the @samp{AM33} or later. This is the 25549default. This option defines the preprocessor macro @code{__LIW__}. 25550 25551@item -mno-liw 25552@opindex mno-liw 25553Do not allow the compiler to generate @emph{Long Instruction Word} 25554instructions. This option defines the preprocessor macro 25555@code{__NO_LIW__}. 25556 25557@item -msetlb 25558@opindex msetlb 25559Allow the compiler to generate the @emph{SETLB} and @emph{Lcc} 25560instructions if the target is the @samp{AM33} or later. This is the 25561default. This option defines the preprocessor macro @code{__SETLB__}. 25562 25563@item -mno-setlb 25564@opindex mno-setlb 25565Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc} 25566instructions. This option defines the preprocessor macro 25567@code{__NO_SETLB__}. 25568 25569@end table 25570 25571@node Moxie Options 25572@subsection Moxie Options 25573@cindex Moxie Options 25574 25575@table @gcctabopt 25576 25577@item -meb 25578@opindex meb 25579Generate big-endian code. This is the default for @samp{moxie-*-*} 25580configurations. 25581 25582@item -mel 25583@opindex mel 25584Generate little-endian code. 25585 25586@item -mmul.x 25587@opindex mmul.x 25588Generate mul.x and umul.x instructions. This is the default for 25589@samp{moxiebox-*-*} configurations. 25590 25591@item -mno-crt0 25592@opindex mno-crt0 25593Do not link in the C run-time initialization object file. 25594 25595@end table 25596 25597@node MSP430 Options 25598@subsection MSP430 Options 25599@cindex MSP430 Options 25600 25601These options are defined for the MSP430: 25602 25603@table @gcctabopt 25604 25605@item -masm-hex 25606@opindex masm-hex 25607Force assembly output to always use hex constants. Normally such 25608constants are signed decimals, but this option is available for 25609testsuite and/or aesthetic purposes. 25610 25611@item -mmcu= 25612@opindex mmcu= 25613Select the MCU to target. This is used to create a C preprocessor 25614symbol based upon the MCU name, converted to upper case and pre- and 25615post-fixed with @samp{__}. This in turn is used by the 25616@file{msp430.h} header file to select an MCU-specific supplementary 25617header file. 25618 25619The option also sets the ISA to use. If the MCU name is one that is 25620known to only support the 430 ISA then that is selected, otherwise the 25621430X ISA is selected. A generic MCU name of @samp{msp430} can also be 25622used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU 25623name selects the 430X ISA. 25624 25625In addition an MCU-specific linker script is added to the linker 25626command line. The script's name is the name of the MCU with 25627@file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc} 25628command line defines the C preprocessor symbol @code{__XXX__} and 25629cause the linker to search for a script called @file{xxx.ld}. 25630 25631The ISA and hardware multiply supported for the different MCUs is hard-coded 25632into GCC. However, an external @samp{devices.csv} file can be used to 25633extend device support beyond those that have been hard-coded. 25634 25635GCC searches for the @samp{devices.csv} file using the following methods in the 25636given precedence order, where the first method takes precendence over the 25637second which takes precedence over the third. 25638 25639@table @asis 25640@item Include path specified with @code{-I} and @code{-L} 25641@samp{devices.csv} will be searched for in each of the directories specified by 25642include paths and linker library search paths. 25643@item Path specified by the environment variable @samp{MSP430_GCC_INCLUDE_DIR} 25644Define the value of the global environment variable 25645@samp{MSP430_GCC_INCLUDE_DIR} 25646to the full path to the directory containing devices.csv, and GCC will search 25647this directory for devices.csv. If devices.csv is found, this directory will 25648also be registered as an include path, and linker library path. Header files 25649and linker scripts in this directory can therefore be used without manually 25650specifying @code{-I} and @code{-L} on the command line. 25651@item The @samp{msp430-elf@{,bare@}/include/devices} directory 25652Finally, GCC will examine @samp{msp430-elf@{,bare@}/include/devices} from the 25653toolchain root directory. This directory does not exist in a default 25654installation, but if the user has created it and copied @samp{devices.csv} 25655there, then the MCU data will be read. As above, this directory will 25656also be registered as an include path, and linker library path. 25657 25658@end table 25659If none of the above search methods find @samp{devices.csv}, then the 25660hard-coded MCU data is used. 25661 25662 25663@item -mwarn-mcu 25664@itemx -mno-warn-mcu 25665@opindex mwarn-mcu 25666@opindex mno-warn-mcu 25667This option enables or disables warnings about conflicts between the 25668MCU name specified by the @option{-mmcu} option and the ISA set by the 25669@option{-mcpu} option and/or the hardware multiply support set by the 25670@option{-mhwmult} option. It also toggles warnings about unrecognized 25671MCU names. This option is on by default. 25672 25673@item -mcpu= 25674@opindex mcpu= 25675Specifies the ISA to use. Accepted values are @samp{msp430}, 25676@samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The 25677@option{-mmcu=} option should be used to select the ISA. 25678 25679@item -msim 25680@opindex msim 25681Link to the simulator runtime libraries and linker script. Overrides 25682any scripts that would be selected by the @option{-mmcu=} option. 25683 25684@item -mlarge 25685@opindex mlarge 25686Use large-model addressing (20-bit pointers, 20-bit @code{size_t}). 25687 25688@item -msmall 25689@opindex msmall 25690Use small-model addressing (16-bit pointers, 16-bit @code{size_t}). 25691 25692@item -mrelax 25693@opindex mrelax 25694This option is passed to the assembler and linker, and allows the 25695linker to perform certain optimizations that cannot be done until 25696the final link. 25697 25698@item mhwmult= 25699@opindex mhwmult= 25700Describes the type of hardware multiply supported by the target. 25701Accepted values are @samp{none} for no hardware multiply, @samp{16bit} 25702for the original 16-bit-only multiply supported by early MCUs. 25703@samp{32bit} for the 16/32-bit multiply supported by later MCUs and 25704@samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs. 25705A value of @samp{auto} can also be given. This tells GCC to deduce 25706the hardware multiply support based upon the MCU name provided by the 25707@option{-mmcu} option. If no @option{-mmcu} option is specified or if 25708the MCU name is not recognized then no hardware multiply support is 25709assumed. @code{auto} is the default setting. 25710 25711Hardware multiplies are normally performed by calling a library 25712routine. This saves space in the generated code. When compiling at 25713@option{-O3} or higher however the hardware multiplier is invoked 25714inline. This makes for bigger, but faster code. 25715 25716The hardware multiply routines disable interrupts whilst running and 25717restore the previous interrupt state when they finish. This makes 25718them safe to use inside interrupt handlers as well as in normal code. 25719 25720@item -minrt 25721@opindex minrt 25722Enable the use of a minimum runtime environment - no static 25723initializers or constructors. This is intended for memory-constrained 25724devices. The compiler includes special symbols in some objects 25725that tell the linker and runtime which code fragments are required. 25726 25727@item -mtiny-printf 25728@opindex mtiny-printf 25729Enable reduced code size @code{printf} and @code{puts} library functions. 25730The @samp{tiny} implementations of these functions are not reentrant, so 25731must be used with caution in multi-threaded applications. 25732 25733Support for streams has been removed and the string to be printed will 25734always be sent to stdout via the @code{write} syscall. The string is not 25735buffered before it is sent to write. 25736 25737This option requires Newlib Nano IO, so GCC must be configured with 25738@samp{--enable-newlib-nano-formatted-io}. 25739 25740@item -mmax-inline-shift= 25741@opindex mmax-inline-shift= 25742This option takes an integer between 0 and 64 inclusive, and sets 25743the maximum number of inline shift instructions which should be emitted to 25744perform a shift operation by a constant amount. When this value needs to be 25745exceeded, an mspabi helper function is used instead. The default value is 4. 25746 25747This only affects cases where a shift by multiple positions cannot be 25748completed with a single instruction (e.g. all shifts >1 on the 430 ISA). 25749 25750Shifts of a 32-bit value are at least twice as costly, so the value passed for 25751this option is divided by 2 and the resulting value used instead. 25752 25753@item -mcode-region= 25754@itemx -mdata-region= 25755@opindex mcode-region 25756@opindex mdata-region 25757These options tell the compiler where to place functions and data that 25758do not have one of the @code{lower}, @code{upper}, @code{either} or 25759@code{section} attributes. Possible values are @code{lower}, 25760@code{upper}, @code{either} or @code{any}. The first three behave 25761like the corresponding attribute. The fourth possible value - 25762@code{any} - is the default. It leaves placement entirely up to the 25763linker script and how it assigns the standard sections 25764(@code{.text}, @code{.data}, etc) to the memory regions. 25765 25766@item -msilicon-errata= 25767@opindex msilicon-errata 25768This option passes on a request to assembler to enable the fixes for 25769the named silicon errata. 25770 25771@item -msilicon-errata-warn= 25772@opindex msilicon-errata-warn 25773This option passes on a request to the assembler to enable warning 25774messages when a silicon errata might need to be applied. 25775 25776@item -mwarn-devices-csv 25777@itemx -mno-warn-devices-csv 25778@opindex mwarn-devices-csv 25779@opindex mno-warn-devices-csv 25780Warn if @samp{devices.csv} is not found or there are problem parsing it 25781(default: on). 25782 25783@end table 25784 25785@node NDS32 Options 25786@subsection NDS32 Options 25787@cindex NDS32 Options 25788 25789These options are defined for NDS32 implementations: 25790 25791@table @gcctabopt 25792 25793@item -mbig-endian 25794@opindex mbig-endian 25795Generate code in big-endian mode. 25796 25797@item -mlittle-endian 25798@opindex mlittle-endian 25799Generate code in little-endian mode. 25800 25801@item -mreduced-regs 25802@opindex mreduced-regs 25803Use reduced-set registers for register allocation. 25804 25805@item -mfull-regs 25806@opindex mfull-regs 25807Use full-set registers for register allocation. 25808 25809@item -mcmov 25810@opindex mcmov 25811Generate conditional move instructions. 25812 25813@item -mno-cmov 25814@opindex mno-cmov 25815Do not generate conditional move instructions. 25816 25817@item -mext-perf 25818@opindex mext-perf 25819Generate performance extension instructions. 25820 25821@item -mno-ext-perf 25822@opindex mno-ext-perf 25823Do not generate performance extension instructions. 25824 25825@item -mext-perf2 25826@opindex mext-perf2 25827Generate performance extension 2 instructions. 25828 25829@item -mno-ext-perf2 25830@opindex mno-ext-perf2 25831Do not generate performance extension 2 instructions. 25832 25833@item -mext-string 25834@opindex mext-string 25835Generate string extension instructions. 25836 25837@item -mno-ext-string 25838@opindex mno-ext-string 25839Do not generate string extension instructions. 25840 25841@item -mv3push 25842@opindex mv3push 25843Generate v3 push25/pop25 instructions. 25844 25845@item -mno-v3push 25846@opindex mno-v3push 25847Do not generate v3 push25/pop25 instructions. 25848 25849@item -m16-bit 25850@opindex m16-bit 25851Generate 16-bit instructions. 25852 25853@item -mno-16-bit 25854@opindex mno-16-bit 25855Do not generate 16-bit instructions. 25856 25857@item -misr-vector-size=@var{num} 25858@opindex misr-vector-size 25859Specify the size of each interrupt vector, which must be 4 or 16. 25860 25861@item -mcache-block-size=@var{num} 25862@opindex mcache-block-size 25863Specify the size of each cache block, 25864which must be a power of 2 between 4 and 512. 25865 25866@item -march=@var{arch} 25867@opindex march 25868Specify the name of the target architecture. 25869 25870@item -mcmodel=@var{code-model} 25871@opindex mcmodel 25872Set the code model to one of 25873@table @asis 25874@item @samp{small} 25875All the data and read-only data segments must be within 512KB addressing space. 25876The text segment must be within 16MB addressing space. 25877@item @samp{medium} 25878The data segment must be within 512KB while the read-only data segment can be 25879within 4GB addressing space. The text segment should be still within 16MB 25880addressing space. 25881@item @samp{large} 25882All the text and data segments can be within 4GB addressing space. 25883@end table 25884 25885@item -mctor-dtor 25886@opindex mctor-dtor 25887Enable constructor/destructor feature. 25888 25889@item -mrelax 25890@opindex mrelax 25891Guide linker to relax instructions. 25892 25893@end table 25894 25895@node Nios II Options 25896@subsection Nios II Options 25897@cindex Nios II options 25898@cindex Altera Nios II options 25899 25900These are the options defined for the Altera Nios II processor. 25901 25902@table @gcctabopt 25903 25904@item -G @var{num} 25905@opindex G 25906@cindex smaller data references 25907Put global and static objects less than or equal to @var{num} bytes 25908into the small data or BSS sections instead of the normal data or BSS 25909sections. The default value of @var{num} is 8. 25910 25911@item -mgpopt=@var{option} 25912@itemx -mgpopt 25913@itemx -mno-gpopt 25914@opindex mgpopt 25915@opindex mno-gpopt 25916Generate (do not generate) GP-relative accesses. The following 25917@var{option} names are recognized: 25918 25919@table @samp 25920 25921@item none 25922Do not generate GP-relative accesses. 25923 25924@item local 25925Generate GP-relative accesses for small data objects that are not 25926external, weak, or uninitialized common symbols. 25927Also use GP-relative addressing for objects that 25928have been explicitly placed in a small data section via a @code{section} 25929attribute. 25930 25931@item global 25932As for @samp{local}, but also generate GP-relative accesses for 25933small data objects that are external, weak, or common. If you use this option, 25934you must ensure that all parts of your program (including libraries) are 25935compiled with the same @option{-G} setting. 25936 25937@item data 25938Generate GP-relative accesses for all data objects in the program. If you 25939use this option, the entire data and BSS segments 25940of your program must fit in 64K of memory and you must use an appropriate 25941linker script to allocate them within the addressable range of the 25942global pointer. 25943 25944@item all 25945Generate GP-relative addresses for function pointers as well as data 25946pointers. If you use this option, the entire text, data, and BSS segments 25947of your program must fit in 64K of memory and you must use an appropriate 25948linker script to allocate them within the addressable range of the 25949global pointer. 25950 25951@end table 25952 25953@option{-mgpopt} is equivalent to @option{-mgpopt=local}, and 25954@option{-mno-gpopt} is equivalent to @option{-mgpopt=none}. 25955 25956The default is @option{-mgpopt} except when @option{-fpic} or 25957@option{-fPIC} is specified to generate position-independent code. 25958Note that the Nios II ABI does not permit GP-relative accesses from 25959shared libraries. 25960 25961You may need to specify @option{-mno-gpopt} explicitly when building 25962programs that include large amounts of small data, including large 25963GOT data sections. In this case, the 16-bit offset for GP-relative 25964addressing may not be large enough to allow access to the entire 25965small data section. 25966 25967@item -mgprel-sec=@var{regexp} 25968@opindex mgprel-sec 25969This option specifies additional section names that can be accessed via 25970GP-relative addressing. It is most useful in conjunction with 25971@code{section} attributes on variable declarations 25972(@pxref{Common Variable Attributes}) and a custom linker script. 25973The @var{regexp} is a POSIX Extended Regular Expression. 25974 25975This option does not affect the behavior of the @option{-G} option, and 25976the specified sections are in addition to the standard @code{.sdata} 25977and @code{.sbss} small-data sections that are recognized by @option{-mgpopt}. 25978 25979@item -mr0rel-sec=@var{regexp} 25980@opindex mr0rel-sec 25981This option specifies names of sections that can be accessed via a 2598216-bit offset from @code{r0}; that is, in the low 32K or high 32K 25983of the 32-bit address space. It is most useful in conjunction with 25984@code{section} attributes on variable declarations 25985(@pxref{Common Variable Attributes}) and a custom linker script. 25986The @var{regexp} is a POSIX Extended Regular Expression. 25987 25988In contrast to the use of GP-relative addressing for small data, 25989zero-based addressing is never generated by default and there are no 25990conventional section names used in standard linker scripts for sections 25991in the low or high areas of memory. 25992 25993@item -mel 25994@itemx -meb 25995@opindex mel 25996@opindex meb 25997Generate little-endian (default) or big-endian (experimental) code, 25998respectively. 25999 26000@item -march=@var{arch} 26001@opindex march 26002This specifies the name of the target Nios II architecture. GCC uses this 26003name to determine what kind of instructions it can emit when generating 26004assembly code. Permissible names are: @samp{r1}, @samp{r2}. 26005 26006The preprocessor macro @code{__nios2_arch__} is available to programs, 26007with value 1 or 2, indicating the targeted ISA level. 26008 26009@item -mbypass-cache 26010@itemx -mno-bypass-cache 26011@opindex mno-bypass-cache 26012@opindex mbypass-cache 26013Force all load and store instructions to always bypass cache by 26014using I/O variants of the instructions. The default is not to 26015bypass the cache. 26016 26017@item -mno-cache-volatile 26018@itemx -mcache-volatile 26019@opindex mcache-volatile 26020@opindex mno-cache-volatile 26021Volatile memory access bypass the cache using the I/O variants of 26022the load and store instructions. The default is not to bypass the cache. 26023 26024@item -mno-fast-sw-div 26025@itemx -mfast-sw-div 26026@opindex mno-fast-sw-div 26027@opindex mfast-sw-div 26028Do not use table-based fast divide for small numbers. The default 26029is to use the fast divide at @option{-O3} and above. 26030 26031@item -mno-hw-mul 26032@itemx -mhw-mul 26033@itemx -mno-hw-mulx 26034@itemx -mhw-mulx 26035@itemx -mno-hw-div 26036@itemx -mhw-div 26037@opindex mno-hw-mul 26038@opindex mhw-mul 26039@opindex mno-hw-mulx 26040@opindex mhw-mulx 26041@opindex mno-hw-div 26042@opindex mhw-div 26043Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of 26044instructions by the compiler. The default is to emit @code{mul} 26045and not emit @code{div} and @code{mulx}. 26046 26047@item -mbmx 26048@itemx -mno-bmx 26049@itemx -mcdx 26050@itemx -mno-cdx 26051Enable or disable generation of Nios II R2 BMX (bit manipulation) and 26052CDX (code density) instructions. Enabling these instructions also 26053requires @option{-march=r2}. Since these instructions are optional 26054extensions to the R2 architecture, the default is not to emit them. 26055 26056@item -mcustom-@var{insn}=@var{N} 26057@itemx -mno-custom-@var{insn} 26058@opindex mcustom-@var{insn} 26059@opindex mno-custom-@var{insn} 26060Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a 26061custom instruction with encoding @var{N} when generating code that uses 26062@var{insn}. For example, @option{-mcustom-fadds=253} generates custom 26063instruction 253 for single-precision floating-point add operations instead 26064of the default behavior of using a library call. 26065 26066The following values of @var{insn} are supported. Except as otherwise 26067noted, floating-point operations are expected to be implemented with 26068normal IEEE 754 semantics and correspond directly to the C operators or the 26069equivalent GCC built-in functions (@pxref{Other Builtins}). 26070 26071Single-precision floating point: 26072@table @asis 26073 26074@item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls} 26075Binary arithmetic operations. 26076 26077@item @samp{fnegs} 26078Unary negation. 26079 26080@item @samp{fabss} 26081Unary absolute value. 26082 26083@item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes} 26084Comparison operations. 26085 26086@item @samp{fmins}, @samp{fmaxs} 26087Floating-point minimum and maximum. These instructions are only 26088generated if @option{-ffinite-math-only} is specified. 26089 26090@item @samp{fsqrts} 26091Unary square root operation. 26092 26093@item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs} 26094Floating-point trigonometric and exponential functions. These instructions 26095are only generated if @option{-funsafe-math-optimizations} is also specified. 26096 26097@end table 26098 26099Double-precision floating point: 26100@table @asis 26101 26102@item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld} 26103Binary arithmetic operations. 26104 26105@item @samp{fnegd} 26106Unary negation. 26107 26108@item @samp{fabsd} 26109Unary absolute value. 26110 26111@item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned} 26112Comparison operations. 26113 26114@item @samp{fmind}, @samp{fmaxd} 26115Double-precision minimum and maximum. These instructions are only 26116generated if @option{-ffinite-math-only} is specified. 26117 26118@item @samp{fsqrtd} 26119Unary square root operation. 26120 26121@item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd} 26122Double-precision trigonometric and exponential functions. These instructions 26123are only generated if @option{-funsafe-math-optimizations} is also specified. 26124 26125@end table 26126 26127Conversions: 26128@table @asis 26129@item @samp{fextsd} 26130Conversion from single precision to double precision. 26131 26132@item @samp{ftruncds} 26133Conversion from double precision to single precision. 26134 26135@item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu} 26136Conversion from floating point to signed or unsigned integer types, with 26137truncation towards zero. 26138 26139@item @samp{round} 26140Conversion from single-precision floating point to signed integer, 26141rounding to the nearest integer and ties away from zero. 26142This corresponds to the @code{__builtin_lroundf} function when 26143@option{-fno-math-errno} is used. 26144 26145@item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud} 26146Conversion from signed or unsigned integer types to floating-point types. 26147 26148@end table 26149 26150In addition, all of the following transfer instructions for internal 26151registers X and Y must be provided to use any of the double-precision 26152floating-point instructions. Custom instructions taking two 26153double-precision source operands expect the first operand in the 2615464-bit register X. The other operand (or only operand of a unary 26155operation) is given to the custom arithmetic instruction with the 26156least significant half in source register @var{src1} and the most 26157significant half in @var{src2}. A custom instruction that returns a 26158double-precision result returns the most significant 32 bits in the 26159destination register and the other half in 32-bit register Y. 26160GCC automatically generates the necessary code sequences to write 26161register X and/or read register Y when double-precision floating-point 26162instructions are used. 26163 26164@table @asis 26165 26166@item @samp{fwrx} 26167Write @var{src1} into the least significant half of X and @var{src2} into 26168the most significant half of X. 26169 26170@item @samp{fwry} 26171Write @var{src1} into Y. 26172 26173@item @samp{frdxhi}, @samp{frdxlo} 26174Read the most or least (respectively) significant half of X and store it in 26175@var{dest}. 26176 26177@item @samp{frdy} 26178Read the value of Y and store it into @var{dest}. 26179@end table 26180 26181Note that you can gain more local control over generation of Nios II custom 26182instructions by using the @code{target("custom-@var{insn}=@var{N}")} 26183and @code{target("no-custom-@var{insn}")} function attributes 26184(@pxref{Function Attributes}) 26185or pragmas (@pxref{Function Specific Option Pragmas}). 26186 26187@item -mcustom-fpu-cfg=@var{name} 26188@opindex mcustom-fpu-cfg 26189 26190This option enables a predefined, named set of custom instruction encodings 26191(see @option{-mcustom-@var{insn}} above). 26192Currently, the following sets are defined: 26193 26194@option{-mcustom-fpu-cfg=60-1} is equivalent to: 26195@gccoptlist{-mcustom-fmuls=252 @gol 26196-mcustom-fadds=253 @gol 26197-mcustom-fsubs=254 @gol 26198-fsingle-precision-constant} 26199 26200@option{-mcustom-fpu-cfg=60-2} is equivalent to: 26201@gccoptlist{-mcustom-fmuls=252 @gol 26202-mcustom-fadds=253 @gol 26203-mcustom-fsubs=254 @gol 26204-mcustom-fdivs=255 @gol 26205-fsingle-precision-constant} 26206 26207@option{-mcustom-fpu-cfg=72-3} is equivalent to: 26208@gccoptlist{-mcustom-floatus=243 @gol 26209-mcustom-fixsi=244 @gol 26210-mcustom-floatis=245 @gol 26211-mcustom-fcmpgts=246 @gol 26212-mcustom-fcmples=249 @gol 26213-mcustom-fcmpeqs=250 @gol 26214-mcustom-fcmpnes=251 @gol 26215-mcustom-fmuls=252 @gol 26216-mcustom-fadds=253 @gol 26217-mcustom-fsubs=254 @gol 26218-mcustom-fdivs=255 @gol 26219-fsingle-precision-constant} 26220 26221@option{-mcustom-fpu-cfg=fph2} is equivalent to: 26222@gccoptlist{-mcustom-fabss=224 @gol 26223-mcustom-fnegs=225 @gol 26224-mcustom-fcmpnes=226 @gol 26225-mcustom-fcmpeqs=227 @gol 26226-mcustom-fcmpges=228 @gol 26227-mcustom-fcmpgts=229 @gol 26228-mcustom-fcmples=230 @gol 26229-mcustom-fcmplts=231 @gol 26230-mcustom-fmaxs=232 @gol 26231-mcustom-fmins=233 @gol 26232-mcustom-round=248 @gol 26233-mcustom-fixsi=249 @gol 26234-mcustom-floatis=250 @gol 26235-mcustom-fsqrts=251 @gol 26236-mcustom-fmuls=252 @gol 26237-mcustom-fadds=253 @gol 26238-mcustom-fsubs=254 @gol 26239-mcustom-fdivs=255 @gol} 26240 26241Custom instruction assignments given by individual 26242@option{-mcustom-@var{insn}=} options override those given by 26243@option{-mcustom-fpu-cfg=}, regardless of the 26244order of the options on the command line. 26245 26246Note that you can gain more local control over selection of a FPU 26247configuration by using the @code{target("custom-fpu-cfg=@var{name}")} 26248function attribute (@pxref{Function Attributes}) 26249or pragma (@pxref{Function Specific Option Pragmas}). 26250 26251The name @var{fph2} is an abbreviation for @emph{Nios II Floating Point 26252Hardware 2 Component}. Please note that the custom instructions enabled by 26253@option{-mcustom-fmins=233} and @option{-mcustom-fmaxs=234} are only generated 26254if @option{-ffinite-math-only} is specified. The custom instruction enabled by 26255@option{-mcustom-round=248} is only generated if @option{-fno-math-errno} is 26256specified. In contrast to the other configurations, 26257@option{-fsingle-precision-constant} is not set. 26258 26259@end table 26260 26261These additional @samp{-m} options are available for the Altera Nios II 26262ELF (bare-metal) target: 26263 26264@table @gcctabopt 26265 26266@item -mhal 26267@opindex mhal 26268Link with HAL BSP. This suppresses linking with the GCC-provided C runtime 26269startup and termination code, and is typically used in conjunction with 26270@option{-msys-crt0=} to specify the location of the alternate startup code 26271provided by the HAL BSP. 26272 26273@item -msmallc 26274@opindex msmallc 26275Link with a limited version of the C library, @option{-lsmallc}, rather than 26276Newlib. 26277 26278@item -msys-crt0=@var{startfile} 26279@opindex msys-crt0 26280@var{startfile} is the file name of the startfile (crt0) to use 26281when linking. This option is only useful in conjunction with @option{-mhal}. 26282 26283@item -msys-lib=@var{systemlib} 26284@opindex msys-lib 26285@var{systemlib} is the library name of the library that provides 26286low-level system calls required by the C library, 26287e.g.@: @code{read} and @code{write}. 26288This option is typically used to link with a library provided by a HAL BSP. 26289 26290@end table 26291 26292@node Nvidia PTX Options 26293@subsection Nvidia PTX Options 26294@cindex Nvidia PTX options 26295@cindex nvptx options 26296 26297These options are defined for Nvidia PTX: 26298 26299@table @gcctabopt 26300 26301@item -m64 26302@opindex m64 26303Ignored, but preserved for backward compatibility. Only 64-bit ABI is 26304supported. 26305 26306@item -misa=@var{ISA-string} 26307@opindex march 26308Generate code for given the specified PTX ISA (e.g.@: @samp{sm_35}). ISA 26309strings must be lower-case. Valid ISA strings include @samp{sm_30} and 26310@samp{sm_35}. The default ISA is sm_35. 26311 26312@item -mmainkernel 26313@opindex mmainkernel 26314Link in code for a __main kernel. This is for stand-alone instead of 26315offloading execution. 26316 26317@item -moptimize 26318@opindex moptimize 26319Apply partitioned execution optimizations. This is the default when any 26320level of optimization is selected. 26321 26322@item -msoft-stack 26323@opindex msoft-stack 26324Generate code that does not use @code{.local} memory 26325directly for stack storage. Instead, a per-warp stack pointer is 26326maintained explicitly. This enables variable-length stack allocation (with 26327variable-length arrays or @code{alloca}), and when global memory is used for 26328underlying storage, makes it possible to access automatic variables from other 26329threads, or with atomic instructions. This code generation variant is used 26330for OpenMP offloading, but the option is exposed on its own for the purpose 26331of testing the compiler; to generate code suitable for linking into programs 26332using OpenMP offloading, use option @option{-mgomp}. 26333 26334@item -muniform-simt 26335@opindex muniform-simt 26336Switch to code generation variant that allows to execute all threads in each 26337warp, while maintaining memory state and side effects as if only one thread 26338in each warp was active outside of OpenMP SIMD regions. All atomic operations 26339and calls to runtime (malloc, free, vprintf) are conditionally executed (iff 26340current lane index equals the master lane index), and the register being 26341assigned is copied via a shuffle instruction from the master lane. Outside of 26342SIMD regions lane 0 is the master; inside, each thread sees itself as the 26343master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or 26344all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD 26345regions). Each thread can bitwise-and the bitmask at position @code{tid.y} 26346with current lane index to compute the master lane index. 26347 26348@item -mgomp 26349@opindex mgomp 26350Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and 26351@option{-muniform-simt} options, and selects corresponding multilib variant. 26352 26353@end table 26354 26355@node OpenRISC Options 26356@subsection OpenRISC Options 26357@cindex OpenRISC Options 26358 26359These options are defined for OpenRISC: 26360 26361@table @gcctabopt 26362 26363@item -mboard=@var{name} 26364@opindex mboard 26365Configure a board specific runtime. This will be passed to the linker for 26366newlib board library linking. The default is @code{or1ksim}. 26367 26368@item -mnewlib 26369@opindex mnewlib 26370This option is ignored; it is for compatibility purposes only. This used to 26371select linker and preprocessor options for use with newlib. 26372 26373@item -msoft-div 26374@itemx -mhard-div 26375@opindex msoft-div 26376@opindex mhard-div 26377Select software or hardware divide (@code{l.div}, @code{l.divu}) instructions. 26378This default is hardware divide. 26379 26380@item -msoft-mul 26381@itemx -mhard-mul 26382@opindex msoft-mul 26383@opindex mhard-mul 26384Select software or hardware multiply (@code{l.mul}, @code{l.muli}) instructions. 26385This default is hardware multiply. 26386 26387@item -msoft-float 26388@itemx -mhard-float 26389@opindex msoft-float 26390@opindex mhard-float 26391Select software or hardware for floating point operations. 26392The default is software. 26393 26394@item -mdouble-float 26395@opindex mdouble-float 26396When @option{-mhard-float} is selected, enables generation of double-precision 26397floating point instructions. By default functions from @file{libgcc} are used 26398to perform double-precision floating point operations. 26399 26400@item -munordered-float 26401@opindex munordered-float 26402When @option{-mhard-float} is selected, enables generation of unordered 26403floating point compare and set flag (@code{lf.sfun*}) instructions. By default 26404functions from @file{libgcc} are used to perform unordered floating point 26405compare and set flag operations. 26406 26407@item -mcmov 26408@opindex mcmov 26409Enable generation of conditional move (@code{l.cmov}) instructions. By 26410default the equivalent will be generated using set and branch. 26411 26412@item -mror 26413@opindex mror 26414Enable generation of rotate right (@code{l.ror}) instructions. By default 26415functions from @file{libgcc} are used to perform rotate right operations. 26416 26417@item -mrori 26418@opindex mrori 26419Enable generation of rotate right with immediate (@code{l.rori}) instructions. 26420By default functions from @file{libgcc} are used to perform rotate right with 26421immediate operations. 26422 26423@item -msext 26424@opindex msext 26425Enable generation of sign extension (@code{l.ext*}) instructions. By default 26426memory loads are used to perform sign extension. 26427 26428@item -msfimm 26429@opindex msfimm 26430Enable generation of compare and set flag with immediate (@code{l.sf*i}) 26431instructions. By default extra instructions will be generated to store the 26432immediate to a register first. 26433 26434@item -mshftimm 26435@opindex mshftimm 26436Enable generation of shift with immediate (@code{l.srai}, @code{l.srli}, 26437@code{l.slli}) instructions. By default extra instructions will be generated 26438to store the immediate to a register first. 26439 26440 26441@end table 26442 26443@node PDP-11 Options 26444@subsection PDP-11 Options 26445@cindex PDP-11 Options 26446 26447These options are defined for the PDP-11: 26448 26449@table @gcctabopt 26450@item -mfpu 26451@opindex mfpu 26452Use hardware FPP floating point. This is the default. (FIS floating 26453point on the PDP-11/40 is not supported.) Implies -m45. 26454 26455@item -msoft-float 26456@opindex msoft-float 26457Do not use hardware floating point. 26458 26459@item -mac0 26460@opindex mac0 26461Return floating-point results in ac0 (fr0 in Unix assembler syntax). 26462 26463@item -mno-ac0 26464@opindex mno-ac0 26465Return floating-point results in memory. This is the default. 26466 26467@item -m40 26468@opindex m40 26469Generate code for a PDP-11/40. Implies -msoft-float -mno-split. 26470 26471@item -m45 26472@opindex m45 26473Generate code for a PDP-11/45. This is the default. 26474 26475@item -m10 26476@opindex m10 26477Generate code for a PDP-11/10. Implies -msoft-float -mno-split. 26478 26479@item -mint16 26480@itemx -mno-int32 26481@opindex mint16 26482@opindex mno-int32 26483Use 16-bit @code{int}. This is the default. 26484 26485@item -mint32 26486@itemx -mno-int16 26487@opindex mint32 26488@opindex mno-int16 26489Use 32-bit @code{int}. 26490 26491@item -msplit 26492@opindex msplit 26493Target has split instruction and data space. Implies -m45. 26494 26495@item -munix-asm 26496@opindex munix-asm 26497Use Unix assembler syntax. 26498 26499@item -mdec-asm 26500@opindex mdec-asm 26501Use DEC assembler syntax. 26502 26503@item -mgnu-asm 26504@opindex mgnu-asm 26505Use GNU assembler syntax. This is the default. 26506 26507@item -mlra 26508@opindex mlra 26509Use the new LRA register allocator. By default, the old ``reload'' 26510allocator is used. 26511@end table 26512 26513@node picoChip Options 26514@subsection picoChip Options 26515@cindex picoChip options 26516 26517These @samp{-m} options are defined for picoChip implementations: 26518 26519@table @gcctabopt 26520 26521@item -mae=@var{ae_type} 26522@opindex mcpu 26523Set the instruction set, register set, and instruction scheduling 26524parameters for array element type @var{ae_type}. Supported values 26525for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}. 26526 26527@option{-mae=ANY} selects a completely generic AE type. Code 26528generated with this option runs on any of the other AE types. The 26529code is not as efficient as it would be if compiled for a specific 26530AE type, and some types of operation (e.g., multiplication) do not 26531work properly on all types of AE. 26532 26533@option{-mae=MUL} selects a MUL AE type. This is the most useful AE type 26534for compiled code, and is the default. 26535 26536@option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this 26537option may suffer from poor performance of byte (char) manipulation, 26538since the DSP AE does not provide hardware support for byte load/stores. 26539 26540@item -msymbol-as-address 26541Enable the compiler to directly use a symbol name as an address in a 26542load/store instruction, without first loading it into a 26543register. Typically, the use of this option generates larger 26544programs, which run faster than when the option isn't used. However, the 26545results vary from program to program, so it is left as a user option, 26546rather than being permanently enabled. 26547 26548@item -mno-inefficient-warnings 26549Disables warnings about the generation of inefficient code. These 26550warnings can be generated, for example, when compiling code that 26551performs byte-level memory operations on the MAC AE type. The MAC AE has 26552no hardware support for byte-level memory operations, so all byte 26553load/stores must be synthesized from word load/store operations. This is 26554inefficient and a warning is generated to indicate 26555that you should rewrite the code to avoid byte operations, or to target 26556an AE type that has the necessary hardware support. This option disables 26557these warnings. 26558 26559@end table 26560 26561@node PowerPC Options 26562@subsection PowerPC Options 26563@cindex PowerPC options 26564 26565These are listed under @xref{RS/6000 and PowerPC Options}. 26566 26567@node PRU Options 26568@subsection PRU Options 26569@cindex PRU Options 26570 26571These command-line options are defined for PRU target: 26572 26573@table @gcctabopt 26574@item -minrt 26575@opindex minrt 26576Link with a minimum runtime environment, with no support for static 26577initializers and constructors. Using this option can significantly reduce 26578the size of the final ELF binary. Beware that the compiler could still 26579generate code with static initializers and constructors. It is up to the 26580programmer to ensure that the source program will not use those features. 26581 26582@item -mmcu=@var{mcu} 26583@opindex mmcu 26584Specify the PRU MCU variant to use. Check Newlib for the exact list of 26585supported MCUs. 26586 26587@item -mno-relax 26588@opindex mno-relax 26589Make GCC pass the @option{--no-relax} command-line option to the linker 26590instead of the @option{--relax} option. 26591 26592@item -mloop 26593@opindex mloop 26594Allow (or do not allow) GCC to use the LOOP instruction. 26595 26596@item -mabi=@var{variant} 26597@opindex mabi 26598Specify the ABI variant to output code for. @option{-mabi=ti} selects the 26599unmodified TI ABI while @option{-mabi=gnu} selects a GNU variant that copes 26600more naturally with certain GCC assumptions. These are the differences: 26601 26602@table @samp 26603@item Function Pointer Size 26604TI ABI specifies that function (code) pointers are 16-bit, whereas GNU 26605supports only 32-bit data and code pointers. 26606 26607@item Optional Return Value Pointer 26608Function return values larger than 64 bits are passed by using a hidden 26609pointer as the first argument of the function. TI ABI, though, mandates that 26610the pointer can be NULL in case the caller is not using the returned value. 26611GNU always passes and expects a valid return value pointer. 26612 26613@end table 26614 26615The current @option{-mabi=ti} implementation simply raises a compile error 26616when any of the above code constructs is detected. As a consequence 26617the standard C library cannot be built and it is omitted when linking with 26618@option{-mabi=ti}. 26619 26620Relaxation is a GNU feature and for safety reasons is disabled when using 26621@option{-mabi=ti}. The TI toolchain does not emit relocations for QBBx 26622instructions, so the GNU linker cannot adjust them when shortening adjacent 26623LDI32 pseudo instructions. 26624 26625@end table 26626 26627@node RISC-V Options 26628@subsection RISC-V Options 26629@cindex RISC-V Options 26630 26631These command-line options are defined for RISC-V targets: 26632 26633@table @gcctabopt 26634@item -mbranch-cost=@var{n} 26635@opindex mbranch-cost 26636Set the cost of branches to roughly @var{n} instructions. 26637 26638@item -mplt 26639@itemx -mno-plt 26640@opindex plt 26641When generating PIC code, do or don't allow the use of PLTs. Ignored for 26642non-PIC. The default is @option{-mplt}. 26643 26644@item -mabi=@var{ABI-string} 26645@opindex mabi 26646Specify integer and floating-point calling convention. @var{ABI-string} 26647contains two parts: the size of integer types and the registers used for 26648floating-point types. For example @samp{-march=rv64ifd -mabi=lp64d} means that 26649@samp{long} and pointers are 64-bit (implicitly defining @samp{int} to be 2665032-bit), and that floating-point values up to 64 bits wide are passed in F 26651registers. Contrast this with @samp{-march=rv64ifd -mabi=lp64f}, which still 26652allows the compiler to generate code that uses the F and D extensions but only 26653allows floating-point values up to 32 bits long to be passed in registers; or 26654@samp{-march=rv64ifd -mabi=lp64}, in which no floating-point arguments will be 26655passed in registers. 26656 26657The default for this argument is system dependent, users who want a specific 26658calling convention should specify one explicitly. The valid calling 26659conventions are: @samp{ilp32}, @samp{ilp32f}, @samp{ilp32d}, @samp{lp64}, 26660@samp{lp64f}, and @samp{lp64d}. Some calling conventions are impossible to 26661implement on some ISAs: for example, @samp{-march=rv32if -mabi=ilp32d} is 26662invalid because the ABI requires 64-bit values be passed in F registers, but F 26663registers are only 32 bits wide. There is also the @samp{ilp32e} ABI that can 26664only be used with the @samp{rv32e} architecture. This ABI is not well 26665specified at present, and is subject to change. 26666 26667@item -mfdiv 26668@itemx -mno-fdiv 26669@opindex mfdiv 26670Do or don't use hardware floating-point divide and square root instructions. 26671This requires the F or D extensions for floating-point registers. The default 26672is to use them if the specified architecture has these instructions. 26673 26674@item -mdiv 26675@itemx -mno-div 26676@opindex mdiv 26677Do or don't use hardware instructions for integer division. This requires the 26678M extension. The default is to use them if the specified architecture has 26679these instructions. 26680 26681@item -march=@var{ISA-string} 26682@opindex march 26683Generate code for given RISC-V ISA (e.g.@: @samp{rv64im}). ISA strings must be 26684lower-case. Examples include @samp{rv64i}, @samp{rv32g}, @samp{rv32e}, and 26685@samp{rv32imaf}. 26686 26687When @option{-march=} is not specified, use the setting from @option{-mcpu}. 26688 26689If both @option{-march} and @option{-mcpu=} are not specified, the default for 26690this argument is system dependent, users who want a specific architecture 26691extensions should specify one explicitly. 26692 26693@item -mcpu=@var{processor-string} 26694@opindex mcpu 26695Use architecture of and optimize the output for the given processor, specified 26696by particular CPU name. 26697Permissible values for this option are: @samp{sifive-e20}, @samp{sifive-e21}, 26698@samp{sifive-e24}, @samp{sifive-e31}, @samp{sifive-e34}, @samp{sifive-e76}, 26699@samp{sifive-s21}, @samp{sifive-s51}, @samp{sifive-s54}, @samp{sifive-s76}, 26700@samp{sifive-u54}, and @samp{sifive-u74}. 26701 26702@item -mtune=@var{processor-string} 26703@opindex mtune 26704Optimize the output for the given processor, specified by microarchitecture or 26705particular CPU name. Permissible values for this option are: @samp{rocket}, 26706@samp{sifive-3-series}, @samp{sifive-5-series}, @samp{sifive-7-series}, 26707@samp{size}, and all valid options for @option{-mcpu=}. 26708 26709When @option{-mtune=} is not specified, use the setting from @option{-mcpu}, 26710the default is @samp{rocket} if both are not specified. 26711 26712The @samp{size} choice is not intended for use by end-users. This is used 26713when @option{-Os} is specified. It overrides the instruction cost info 26714provided by @option{-mtune=}, but does not override the pipeline info. This 26715helps reduce code size while still giving good performance. 26716 26717@item -mpreferred-stack-boundary=@var{num} 26718@opindex mpreferred-stack-boundary 26719Attempt to keep the stack boundary aligned to a 2 raised to @var{num} 26720byte boundary. If @option{-mpreferred-stack-boundary} is not specified, 26721the default is 4 (16 bytes or 128-bits). 26722 26723@strong{Warning:} If you use this switch, then you must build all modules with 26724the same value, including any libraries. This includes the system libraries 26725and startup modules. 26726 26727@item -msmall-data-limit=@var{n} 26728@opindex msmall-data-limit 26729Put global and static data smaller than @var{n} bytes into a special section 26730(on some targets). 26731 26732@item -msave-restore 26733@itemx -mno-save-restore 26734@opindex msave-restore 26735Do or don't use smaller but slower prologue and epilogue code that uses 26736library function calls. The default is to use fast inline prologues and 26737epilogues. 26738 26739@item -mshorten-memrefs 26740@itemx -mno-shorten-memrefs 26741@opindex mshorten-memrefs 26742Do or do not attempt to make more use of compressed load/store instructions by 26743replacing a load/store of 'base register + large offset' with a new load/store 26744of 'new base + small offset'. If the new base gets stored in a compressed 26745register, then the new load/store can be compressed. Currently targets 32-bit 26746integer load/stores only. 26747 26748@item -mstrict-align 26749@itemx -mno-strict-align 26750@opindex mstrict-align 26751Do not or do generate unaligned memory accesses. The default is set depending 26752on whether the processor we are optimizing for supports fast unaligned access 26753or not. 26754 26755@item -mcmodel=medlow 26756@opindex mcmodel=medlow 26757Generate code for the medium-low code model. The program and its statically 26758defined symbols must lie within a single 2 GiB address range and must lie 26759between absolute addresses @minus{}2 GiB and +2 GiB. Programs can be 26760statically or dynamically linked. This is the default code model. 26761 26762@item -mcmodel=medany 26763@opindex mcmodel=medany 26764Generate code for the medium-any code model. The program and its statically 26765defined symbols must be within any single 2 GiB address range. Programs can be 26766statically or dynamically linked. 26767 26768@item -mexplicit-relocs 26769@itemx -mno-exlicit-relocs 26770Use or do not use assembler relocation operators when dealing with symbolic 26771addresses. The alternative is to use assembler macros instead, which may 26772limit optimization. 26773 26774@item -mrelax 26775@itemx -mno-relax 26776Take advantage of linker relaxations to reduce the number of instructions 26777required to materialize symbol addresses. The default is to take advantage of 26778linker relaxations. 26779 26780@item -memit-attribute 26781@itemx -mno-emit-attribute 26782Emit (do not emit) RISC-V attribute to record extra information into ELF 26783objects. This feature requires at least binutils 2.32. 26784 26785@item -malign-data=@var{type} 26786@opindex malign-data 26787Control how GCC aligns variables and constants of array, structure, or union 26788types. Supported values for @var{type} are @samp{xlen} which uses x register 26789width as the alignment value, and @samp{natural} which uses natural alignment. 26790@samp{xlen} is the default. 26791 26792@item -mbig-endian 26793@opindex mbig-endian 26794Generate big-endian code. This is the default when GCC is configured for a 26795@samp{riscv64be-*-*} or @samp{riscv32be-*-*} target. 26796 26797@item -mlittle-endian 26798@opindex mlittle-endian 26799Generate little-endian code. This is the default when GCC is configured for a 26800@samp{riscv64-*-*} or @samp{riscv32-*-*} but not a @samp{riscv64be-*-*} or 26801@samp{riscv32be-*-*} target. 26802 26803@item -mstack-protector-guard=@var{guard} 26804@itemx -mstack-protector-guard-reg=@var{reg} 26805@itemx -mstack-protector-guard-offset=@var{offset} 26806@opindex mstack-protector-guard 26807@opindex mstack-protector-guard-reg 26808@opindex mstack-protector-guard-offset 26809Generate stack protection code using canary at @var{guard}. Supported 26810locations are @samp{global} for a global canary or @samp{tls} for per-thread 26811canary in the TLS block. 26812 26813With the latter choice the options 26814@option{-mstack-protector-guard-reg=@var{reg}} and 26815@option{-mstack-protector-guard-offset=@var{offset}} furthermore specify 26816which register to use as base register for reading the canary, 26817and from what offset from that base register. There is no default 26818register or offset as this is entirely for use within the Linux 26819kernel. 26820@end table 26821 26822@node RL78 Options 26823@subsection RL78 Options 26824@cindex RL78 Options 26825 26826@table @gcctabopt 26827 26828@item -msim 26829@opindex msim 26830Links in additional target libraries to support operation within a 26831simulator. 26832 26833@item -mmul=none 26834@itemx -mmul=g10 26835@itemx -mmul=g13 26836@itemx -mmul=g14 26837@itemx -mmul=rl78 26838@opindex mmul 26839Specifies the type of hardware multiplication and division support to 26840be used. The simplest is @code{none}, which uses software for both 26841multiplication and division. This is the default. The @code{g13} 26842value is for the hardware multiply/divide peripheral found on the 26843RL78/G13 (S2 core) targets. The @code{g14} value selects the use of 26844the multiplication and division instructions supported by the RL78/G14 26845(S3 core) parts. The value @code{rl78} is an alias for @code{g14} and 26846the value @code{mg10} is an alias for @code{none}. 26847 26848In addition a C preprocessor macro is defined, based upon the setting 26849of this option. Possible values are: @code{__RL78_MUL_NONE__}, 26850@code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}. 26851 26852@item -mcpu=g10 26853@itemx -mcpu=g13 26854@itemx -mcpu=g14 26855@itemx -mcpu=rl78 26856@opindex mcpu 26857Specifies the RL78 core to target. The default is the G14 core, also 26858known as an S3 core or just RL78. The G13 or S2 core does not have 26859multiply or divide instructions, instead it uses a hardware peripheral 26860for these operations. The G10 or S1 core does not have register 26861banks, so it uses a different calling convention. 26862 26863If this option is set it also selects the type of hardware multiply 26864support to use, unless this is overridden by an explicit 26865@option{-mmul=none} option on the command line. Thus specifying 26866@option{-mcpu=g13} enables the use of the G13 hardware multiply 26867peripheral and specifying @option{-mcpu=g10} disables the use of 26868hardware multiplications altogether. 26869 26870Note, although the RL78/G14 core is the default target, specifying 26871@option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does 26872change the behavior of the toolchain since it also enables G14 26873hardware multiply support. If these options are not specified on the 26874command line then software multiplication routines will be used even 26875though the code targets the RL78 core. This is for backwards 26876compatibility with older toolchains which did not have hardware 26877multiply and divide support. 26878 26879In addition a C preprocessor macro is defined, based upon the setting 26880of this option. Possible values are: @code{__RL78_G10__}, 26881@code{__RL78_G13__} or @code{__RL78_G14__}. 26882 26883@item -mg10 26884@itemx -mg13 26885@itemx -mg14 26886@itemx -mrl78 26887@opindex mg10 26888@opindex mg13 26889@opindex mg14 26890@opindex mrl78 26891These are aliases for the corresponding @option{-mcpu=} option. They 26892are provided for backwards compatibility. 26893 26894@item -mallregs 26895@opindex mallregs 26896Allow the compiler to use all of the available registers. By default 26897registers @code{r24..r31} are reserved for use in interrupt handlers. 26898With this option enabled these registers can be used in ordinary 26899functions as well. 26900 26901@item -m64bit-doubles 26902@itemx -m32bit-doubles 26903@opindex m64bit-doubles 26904@opindex m32bit-doubles 26905Make the @code{double} data type be 64 bits (@option{-m64bit-doubles}) 26906or 32 bits (@option{-m32bit-doubles}) in size. The default is 26907@option{-m32bit-doubles}. 26908 26909@item -msave-mduc-in-interrupts 26910@itemx -mno-save-mduc-in-interrupts 26911@opindex msave-mduc-in-interrupts 26912@opindex mno-save-mduc-in-interrupts 26913Specifies that interrupt handler functions should preserve the 26914MDUC registers. This is only necessary if normal code might use 26915the MDUC registers, for example because it performs multiplication 26916and division operations. The default is to ignore the MDUC registers 26917as this makes the interrupt handlers faster. The target option -mg13 26918needs to be passed for this to work as this feature is only available 26919on the G13 target (S2 core). The MDUC registers will only be saved 26920if the interrupt handler performs a multiplication or division 26921operation or it calls another function. 26922 26923@end table 26924 26925@node RS/6000 and PowerPC Options 26926@subsection IBM RS/6000 and PowerPC Options 26927@cindex RS/6000 and PowerPC Options 26928@cindex IBM RS/6000 and PowerPC Options 26929 26930These @samp{-m} options are defined for the IBM RS/6000 and PowerPC: 26931@table @gcctabopt 26932@item -mpowerpc-gpopt 26933@itemx -mno-powerpc-gpopt 26934@itemx -mpowerpc-gfxopt 26935@itemx -mno-powerpc-gfxopt 26936@need 800 26937@itemx -mpowerpc64 26938@itemx -mno-powerpc64 26939@itemx -mmfcrf 26940@itemx -mno-mfcrf 26941@itemx -mpopcntb 26942@itemx -mno-popcntb 26943@itemx -mpopcntd 26944@itemx -mno-popcntd 26945@itemx -mfprnd 26946@itemx -mno-fprnd 26947@need 800 26948@itemx -mcmpb 26949@itemx -mno-cmpb 26950@itemx -mhard-dfp 26951@itemx -mno-hard-dfp 26952@opindex mpowerpc-gpopt 26953@opindex mno-powerpc-gpopt 26954@opindex mpowerpc-gfxopt 26955@opindex mno-powerpc-gfxopt 26956@opindex mpowerpc64 26957@opindex mno-powerpc64 26958@opindex mmfcrf 26959@opindex mno-mfcrf 26960@opindex mpopcntb 26961@opindex mno-popcntb 26962@opindex mpopcntd 26963@opindex mno-popcntd 26964@opindex mfprnd 26965@opindex mno-fprnd 26966@opindex mcmpb 26967@opindex mno-cmpb 26968@opindex mhard-dfp 26969@opindex mno-hard-dfp 26970You use these options to specify which instructions are available on the 26971processor you are using. The default value of these options is 26972determined when configuring GCC@. Specifying the 26973@option{-mcpu=@var{cpu_type}} overrides the specification of these 26974options. We recommend you use the @option{-mcpu=@var{cpu_type}} option 26975rather than the options listed above. 26976 26977Specifying @option{-mpowerpc-gpopt} allows 26978GCC to use the optional PowerPC architecture instructions in the 26979General Purpose group, including floating-point square root. Specifying 26980@option{-mpowerpc-gfxopt} allows GCC to 26981use the optional PowerPC architecture instructions in the Graphics 26982group, including floating-point select. 26983 26984The @option{-mmfcrf} option allows GCC to generate the move from 26985condition register field instruction implemented on the POWER4 26986processor and other processors that support the PowerPC V2.01 26987architecture. 26988The @option{-mpopcntb} option allows GCC to generate the popcount and 26989double-precision FP reciprocal estimate instruction implemented on the 26990POWER5 processor and other processors that support the PowerPC V2.02 26991architecture. 26992The @option{-mpopcntd} option allows GCC to generate the popcount 26993instruction implemented on the POWER7 processor and other processors 26994that support the PowerPC V2.06 architecture. 26995The @option{-mfprnd} option allows GCC to generate the FP round to 26996integer instructions implemented on the POWER5+ processor and other 26997processors that support the PowerPC V2.03 architecture. 26998The @option{-mcmpb} option allows GCC to generate the compare bytes 26999instruction implemented on the POWER6 processor and other processors 27000that support the PowerPC V2.05 architecture. 27001The @option{-mhard-dfp} option allows GCC to generate the decimal 27002floating-point instructions implemented on some POWER processors. 27003 27004The @option{-mpowerpc64} option allows GCC to generate the additional 2700564-bit instructions that are found in the full PowerPC64 architecture 27006and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to 27007@option{-mno-powerpc64}. 27008 27009@item -mcpu=@var{cpu_type} 27010@opindex mcpu 27011Set architecture type, register usage, and 27012instruction scheduling parameters for machine type @var{cpu_type}. 27013Supported values for @var{cpu_type} are @samp{401}, @samp{403}, 27014@samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp}, 27015@samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603}, 27016@samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740}, 27017@samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823}, 27018@samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2}, 27019@samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500}, 27020@samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5}, 27021@samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+}, 27022@samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, 27023@samp{power9}, @samp{future}, @samp{powerpc}, @samp{powerpc64}, 27024@samp{powerpc64le}, @samp{rs64}, and @samp{native}. 27025 27026@option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and 27027@option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either 27028endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC 27029architecture machine types, with an appropriate, generic processor 27030model assumed for scheduling purposes. 27031 27032Specifying @samp{native} as cpu type detects and selects the 27033architecture option that corresponds to the host processor of the 27034system performing the compilation. 27035@option{-mcpu=native} has no effect if GCC does not recognize the 27036processor. 27037 27038The other options specify a specific processor. Code generated under 27039those options runs best on that processor, and may not run at all on 27040others. 27041 27042The @option{-mcpu} options automatically enable or disable the 27043following options: 27044 27045@gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol 27046-mpopcntb -mpopcntd -mpowerpc64 @gol 27047-mpowerpc-gpopt -mpowerpc-gfxopt @gol 27048-mmulhw -mdlmzb -mmfpgpr -mvsx @gol 27049-mcrypto -mhtm -mpower8-fusion -mpower8-vector @gol 27050-mquad-memory -mquad-memory-atomic -mfloat128 @gol 27051-mfloat128-hardware -mprefixed -mpcrel -mmma @gol 27052-mrop-protect} 27053 27054The particular options set for any particular CPU varies between 27055compiler versions, depending on what setting seems to produce optimal 27056code for that CPU; it doesn't necessarily reflect the actual hardware's 27057capabilities. If you wish to set an individual option to a particular 27058value, you may specify it after the @option{-mcpu} option, like 27059@option{-mcpu=970 -mno-altivec}. 27060 27061On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are 27062not enabled or disabled by the @option{-mcpu} option at present because 27063AIX does not have full support for these options. You may still 27064enable or disable them individually if you're sure it'll work in your 27065environment. 27066 27067@item -mtune=@var{cpu_type} 27068@opindex mtune 27069Set the instruction scheduling parameters for machine type 27070@var{cpu_type}, but do not set the architecture type or register usage, 27071as @option{-mcpu=@var{cpu_type}} does. The same 27072values for @var{cpu_type} are used for @option{-mtune} as for 27073@option{-mcpu}. If both are specified, the code generated uses the 27074architecture and registers set by @option{-mcpu}, but the 27075scheduling parameters set by @option{-mtune}. 27076 27077@item -mcmodel=small 27078@opindex mcmodel=small 27079Generate PowerPC64 code for the small model: The TOC is limited to 2708064k. 27081 27082@item -mcmodel=medium 27083@opindex mcmodel=medium 27084Generate PowerPC64 code for the medium model: The TOC and other static 27085data may be up to a total of 4G in size. This is the default for 64-bit 27086Linux. 27087 27088@item -mcmodel=large 27089@opindex mcmodel=large 27090Generate PowerPC64 code for the large model: The TOC may be up to 4G 27091in size. Other data and code is only limited by the 64-bit address 27092space. 27093 27094@item -maltivec 27095@itemx -mno-altivec 27096@opindex maltivec 27097@opindex mno-altivec 27098Generate code that uses (does not use) AltiVec instructions, and also 27099enable the use of built-in functions that allow more direct access to 27100the AltiVec instruction set. You may also need to set 27101@option{-mabi=altivec} to adjust the current ABI with AltiVec ABI 27102enhancements. 27103 27104When @option{-maltivec} is used, the element order for AltiVec intrinsics 27105such as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert} 27106match array element order corresponding to the endianness of the 27107target. That is, element zero identifies the leftmost element in a 27108vector register when targeting a big-endian platform, and identifies 27109the rightmost element in a vector register when targeting a 27110little-endian platform. 27111 27112@item -mvrsave 27113@itemx -mno-vrsave 27114@opindex mvrsave 27115@opindex mno-vrsave 27116Generate VRSAVE instructions when generating AltiVec code. 27117 27118@item -msecure-plt 27119@opindex msecure-plt 27120Generate code that allows @command{ld} and @command{ld.so} 27121to build executables and shared 27122libraries with non-executable @code{.plt} and @code{.got} sections. 27123This is a PowerPC 2712432-bit SYSV ABI option. 27125 27126@item -mbss-plt 27127@opindex mbss-plt 27128Generate code that uses a BSS @code{.plt} section that @command{ld.so} 27129fills in, and 27130requires @code{.plt} and @code{.got} 27131sections that are both writable and executable. 27132This is a PowerPC 32-bit SYSV ABI option. 27133 27134@item -misel 27135@itemx -mno-isel 27136@opindex misel 27137@opindex mno-isel 27138This switch enables or disables the generation of ISEL instructions. 27139 27140@item -mvsx 27141@itemx -mno-vsx 27142@opindex mvsx 27143@opindex mno-vsx 27144Generate code that uses (does not use) vector/scalar (VSX) 27145instructions, and also enable the use of built-in functions that allow 27146more direct access to the VSX instruction set. 27147 27148@item -mcrypto 27149@itemx -mno-crypto 27150@opindex mcrypto 27151@opindex mno-crypto 27152Enable the use (disable) of the built-in functions that allow direct 27153access to the cryptographic instructions that were added in version 271542.07 of the PowerPC ISA. 27155 27156@item -mhtm 27157@itemx -mno-htm 27158@opindex mhtm 27159@opindex mno-htm 27160Enable (disable) the use of the built-in functions that allow direct 27161access to the Hardware Transactional Memory (HTM) instructions that 27162were added in version 2.07 of the PowerPC ISA. 27163 27164@item -mpower8-fusion 27165@itemx -mno-power8-fusion 27166@opindex mpower8-fusion 27167@opindex mno-power8-fusion 27168Generate code that keeps (does not keeps) some integer operations 27169adjacent so that the instructions can be fused together on power8 and 27170later processors. 27171 27172@item -mpower8-vector 27173@itemx -mno-power8-vector 27174@opindex mpower8-vector 27175@opindex mno-power8-vector 27176Generate code that uses (does not use) the vector and scalar 27177instructions that were added in version 2.07 of the PowerPC ISA. Also 27178enable the use of built-in functions that allow more direct access to 27179the vector instructions. 27180 27181@item -mquad-memory 27182@itemx -mno-quad-memory 27183@opindex mquad-memory 27184@opindex mno-quad-memory 27185Generate code that uses (does not use) the non-atomic quad word memory 27186instructions. The @option{-mquad-memory} option requires use of 2718764-bit mode. 27188 27189@item -mquad-memory-atomic 27190@itemx -mno-quad-memory-atomic 27191@opindex mquad-memory-atomic 27192@opindex mno-quad-memory-atomic 27193Generate code that uses (does not use) the atomic quad word memory 27194instructions. The @option{-mquad-memory-atomic} option requires use of 2719564-bit mode. 27196 27197@item -mfloat128 27198@itemx -mno-float128 27199@opindex mfloat128 27200@opindex mno-float128 27201Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point 27202and use either software emulation for IEEE 128-bit floating point or 27203hardware instructions. 27204 27205The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7}, 27206@option{-mcpu=power8}), or @option{-mcpu=power9} must be enabled to 27207use the IEEE 128-bit floating point support. The IEEE 128-bit 27208floating point support only works on PowerPC Linux systems. 27209 27210The default for @option{-mfloat128} is enabled on PowerPC Linux 27211systems using the VSX instruction set, and disabled on other systems. 27212 27213If you use the ISA 3.0 instruction set (@option{-mpower9-vector} or 27214@option{-mcpu=power9}) on a 64-bit system, the IEEE 128-bit floating 27215point support will also enable the generation of ISA 3.0 IEEE 128-bit 27216floating point instructions. Otherwise, if you do not specify to 27217generate ISA 3.0 instructions or you are targeting a 32-bit big endian 27218system, IEEE 128-bit floating point will be done with software 27219emulation. 27220 27221@item -mfloat128-hardware 27222@itemx -mno-float128-hardware 27223@opindex mfloat128-hardware 27224@opindex mno-float128-hardware 27225Enable/disable using ISA 3.0 hardware instructions to support the 27226@var{__float128} data type. 27227 27228The default for @option{-mfloat128-hardware} is enabled on PowerPC 27229Linux systems using the ISA 3.0 instruction set, and disabled on other 27230systems. 27231 27232@item -m32 27233@itemx -m64 27234@opindex m32 27235@opindex m64 27236Generate code for 32-bit or 64-bit environments of Darwin and SVR4 27237targets (including GNU/Linux). The 32-bit environment sets int, long 27238and pointer to 32 bits and generates code that runs on any PowerPC 27239variant. The 64-bit environment sets int to 32 bits and long and 27240pointer to 64 bits, and generates code for PowerPC64, as for 27241@option{-mpowerpc64}. 27242 27243@item -mfull-toc 27244@itemx -mno-fp-in-toc 27245@itemx -mno-sum-in-toc 27246@itemx -mminimal-toc 27247@opindex mfull-toc 27248@opindex mno-fp-in-toc 27249@opindex mno-sum-in-toc 27250@opindex mminimal-toc 27251Modify generation of the TOC (Table Of Contents), which is created for 27252every executable file. The @option{-mfull-toc} option is selected by 27253default. In that case, GCC allocates at least one TOC entry for 27254each unique non-automatic variable reference in your program. GCC 27255also places floating-point constants in the TOC@. However, only 2725616,384 entries are available in the TOC@. 27257 27258If you receive a linker error message that saying you have overflowed 27259the available TOC space, you can reduce the amount of TOC space used 27260with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options. 27261@option{-mno-fp-in-toc} prevents GCC from putting floating-point 27262constants in the TOC and @option{-mno-sum-in-toc} forces GCC to 27263generate code to calculate the sum of an address and a constant at 27264run time instead of putting that sum into the TOC@. You may specify one 27265or both of these options. Each causes GCC to produce very slightly 27266slower and larger code at the expense of conserving TOC space. 27267 27268If you still run out of space in the TOC even when you specify both of 27269these options, specify @option{-mminimal-toc} instead. This option causes 27270GCC to make only one TOC entry for every file. When you specify this 27271option, GCC produces code that is slower and larger but which 27272uses extremely little TOC space. You may wish to use this option 27273only on files that contain less frequently-executed code. 27274 27275@item -maix64 27276@itemx -maix32 27277@opindex maix64 27278@opindex maix32 27279Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit 27280@code{long} type, and the infrastructure needed to support them. 27281Specifying @option{-maix64} implies @option{-mpowerpc64}, 27282while @option{-maix32} disables the 64-bit ABI and 27283implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}. 27284 27285@item -mxl-compat 27286@itemx -mno-xl-compat 27287@opindex mxl-compat 27288@opindex mno-xl-compat 27289Produce code that conforms more closely to IBM XL compiler semantics 27290when using AIX-compatible ABI@. Pass floating-point arguments to 27291prototyped functions beyond the register save area (RSA) on the stack 27292in addition to argument FPRs. Do not assume that most significant 27293double in 128-bit long double value is properly rounded when comparing 27294values and converting to double. Use XL symbol names for long double 27295support routines. 27296 27297The AIX calling convention was extended but not initially documented to 27298handle an obscure K&R C case of calling a function that takes the 27299address of its arguments with fewer arguments than declared. IBM XL 27300compilers access floating-point arguments that do not fit in the 27301RSA from the stack when a subroutine is compiled without 27302optimization. Because always storing floating-point arguments on the 27303stack is inefficient and rarely needed, this option is not enabled by 27304default and only is necessary when calling subroutines compiled by IBM 27305XL compilers without optimization. 27306 27307@item -mpe 27308@opindex mpe 27309Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an 27310application written to use message passing with special startup code to 27311enable the application to run. The system must have PE installed in the 27312standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file 27313must be overridden with the @option{-specs=} option to specify the 27314appropriate directory location. The Parallel Environment does not 27315support threads, so the @option{-mpe} option and the @option{-pthread} 27316option are incompatible. 27317 27318@item -malign-natural 27319@itemx -malign-power 27320@opindex malign-natural 27321@opindex malign-power 27322On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option 27323@option{-malign-natural} overrides the ABI-defined alignment of larger 27324types, such as floating-point doubles, on their natural size-based boundary. 27325The option @option{-malign-power} instructs GCC to follow the ABI-specified 27326alignment rules. GCC defaults to the standard alignment defined in the ABI@. 27327 27328On 64-bit Darwin, natural alignment is the default, and @option{-malign-power} 27329is not supported. 27330 27331@item -msoft-float 27332@itemx -mhard-float 27333@opindex msoft-float 27334@opindex mhard-float 27335Generate code that does not use (uses) the floating-point register set. 27336Software floating-point emulation is provided if you use the 27337@option{-msoft-float} option, and pass the option to GCC when linking. 27338 27339@item -mmultiple 27340@itemx -mno-multiple 27341@opindex mmultiple 27342@opindex mno-multiple 27343Generate code that uses (does not use) the load multiple word 27344instructions and the store multiple word instructions. These 27345instructions are generated by default on POWER systems, and not 27346generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian 27347PowerPC systems, since those instructions do not work when the 27348processor is in little-endian mode. The exceptions are PPC740 and 27349PPC750 which permit these instructions in little-endian mode. 27350 27351@item -mupdate 27352@itemx -mno-update 27353@opindex mupdate 27354@opindex mno-update 27355Generate code that uses (does not use) the load or store instructions 27356that update the base register to the address of the calculated memory 27357location. These instructions are generated by default. If you use 27358@option{-mno-update}, there is a small window between the time that the 27359stack pointer is updated and the address of the previous frame is 27360stored, which means code that walks the stack frame across interrupts or 27361signals may get corrupted data. 27362 27363@item -mavoid-indexed-addresses 27364@itemx -mno-avoid-indexed-addresses 27365@opindex mavoid-indexed-addresses 27366@opindex mno-avoid-indexed-addresses 27367Generate code that tries to avoid (not avoid) the use of indexed load 27368or store instructions. These instructions can incur a performance 27369penalty on Power6 processors in certain situations, such as when 27370stepping through large arrays that cross a 16M boundary. This option 27371is enabled by default when targeting Power6 and disabled otherwise. 27372 27373@item -mfused-madd 27374@itemx -mno-fused-madd 27375@opindex mfused-madd 27376@opindex mno-fused-madd 27377Generate code that uses (does not use) the floating-point multiply and 27378accumulate instructions. These instructions are generated by default 27379if hardware floating point is used. The machine-dependent 27380@option{-mfused-madd} option is now mapped to the machine-independent 27381@option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is 27382mapped to @option{-ffp-contract=off}. 27383 27384@item -mmulhw 27385@itemx -mno-mulhw 27386@opindex mmulhw 27387@opindex mno-mulhw 27388Generate code that uses (does not use) the half-word multiply and 27389multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors. 27390These instructions are generated by default when targeting those 27391processors. 27392 27393@item -mdlmzb 27394@itemx -mno-dlmzb 27395@opindex mdlmzb 27396@opindex mno-dlmzb 27397Generate code that uses (does not use) the string-search @samp{dlmzb} 27398instruction on the IBM 405, 440, 464 and 476 processors. This instruction is 27399generated by default when targeting those processors. 27400 27401@item -mno-bit-align 27402@itemx -mbit-align 27403@opindex mno-bit-align 27404@opindex mbit-align 27405On System V.4 and embedded PowerPC systems do not (do) force structures 27406and unions that contain bit-fields to be aligned to the base type of the 27407bit-field. 27408 27409For example, by default a structure containing nothing but 8 27410@code{unsigned} bit-fields of length 1 is aligned to a 4-byte 27411boundary and has a size of 4 bytes. By using @option{-mno-bit-align}, 27412the structure is aligned to a 1-byte boundary and is 1 byte in 27413size. 27414 27415@item -mno-strict-align 27416@itemx -mstrict-align 27417@opindex mno-strict-align 27418@opindex mstrict-align 27419On System V.4 and embedded PowerPC systems do not (do) assume that 27420unaligned memory references are handled by the system. 27421 27422@item -mrelocatable 27423@itemx -mno-relocatable 27424@opindex mrelocatable 27425@opindex mno-relocatable 27426Generate code that allows (does not allow) a static executable to be 27427relocated to a different address at run time. A simple embedded 27428PowerPC system loader should relocate the entire contents of 27429@code{.got2} and 4-byte locations listed in the @code{.fixup} section, 27430a table of 32-bit addresses generated by this option. For this to 27431work, all objects linked together must be compiled with 27432@option{-mrelocatable} or @option{-mrelocatable-lib}. 27433@option{-mrelocatable} code aligns the stack to an 8-byte boundary. 27434 27435@item -mrelocatable-lib 27436@itemx -mno-relocatable-lib 27437@opindex mrelocatable-lib 27438@opindex mno-relocatable-lib 27439Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a 27440@code{.fixup} section to allow static executables to be relocated at 27441run time, but @option{-mrelocatable-lib} does not use the smaller stack 27442alignment of @option{-mrelocatable}. Objects compiled with 27443@option{-mrelocatable-lib} may be linked with objects compiled with 27444any combination of the @option{-mrelocatable} options. 27445 27446@item -mno-toc 27447@itemx -mtoc 27448@opindex mno-toc 27449@opindex mtoc 27450On System V.4 and embedded PowerPC systems do not (do) assume that 27451register 2 contains a pointer to a global area pointing to the addresses 27452used in the program. 27453 27454@item -mlittle 27455@itemx -mlittle-endian 27456@opindex mlittle 27457@opindex mlittle-endian 27458On System V.4 and embedded PowerPC systems compile code for the 27459processor in little-endian mode. The @option{-mlittle-endian} option is 27460the same as @option{-mlittle}. 27461 27462@item -mbig 27463@itemx -mbig-endian 27464@opindex mbig 27465@opindex mbig-endian 27466On System V.4 and embedded PowerPC systems compile code for the 27467processor in big-endian mode. The @option{-mbig-endian} option is 27468the same as @option{-mbig}. 27469 27470@item -mdynamic-no-pic 27471@opindex mdynamic-no-pic 27472On Darwin and Mac OS X systems, compile code so that it is not 27473relocatable, but that its external references are relocatable. The 27474resulting code is suitable for applications, but not shared 27475libraries. 27476 27477@item -msingle-pic-base 27478@opindex msingle-pic-base 27479Treat the register used for PIC addressing as read-only, rather than 27480loading it in the prologue for each function. The runtime system is 27481responsible for initializing this register with an appropriate value 27482before execution begins. 27483 27484@item -mprioritize-restricted-insns=@var{priority} 27485@opindex mprioritize-restricted-insns 27486This option controls the priority that is assigned to 27487dispatch-slot restricted instructions during the second scheduling 27488pass. The argument @var{priority} takes the value @samp{0}, @samp{1}, 27489or @samp{2} to assign no, highest, or second-highest (respectively) 27490priority to dispatch-slot restricted 27491instructions. 27492 27493@item -msched-costly-dep=@var{dependence_type} 27494@opindex msched-costly-dep 27495This option controls which dependences are considered costly 27496by the target during instruction scheduling. The argument 27497@var{dependence_type} takes one of the following values: 27498 27499@table @asis 27500@item @samp{no} 27501No dependence is costly. 27502 27503@item @samp{all} 27504All dependences are costly. 27505 27506@item @samp{true_store_to_load} 27507A true dependence from store to load is costly. 27508 27509@item @samp{store_to_load} 27510Any dependence from store to load is costly. 27511 27512@item @var{number} 27513Any dependence for which the latency is greater than or equal to 27514@var{number} is costly. 27515@end table 27516 27517@item -minsert-sched-nops=@var{scheme} 27518@opindex minsert-sched-nops 27519This option controls which NOP insertion scheme is used during 27520the second scheduling pass. The argument @var{scheme} takes one of the 27521following values: 27522 27523@table @asis 27524@item @samp{no} 27525Don't insert NOPs. 27526 27527@item @samp{pad} 27528Pad with NOPs any dispatch group that has vacant issue slots, 27529according to the scheduler's grouping. 27530 27531@item @samp{regroup_exact} 27532Insert NOPs to force costly dependent insns into 27533separate groups. Insert exactly as many NOPs as needed to force an insn 27534to a new group, according to the estimated processor grouping. 27535 27536@item @var{number} 27537Insert NOPs to force costly dependent insns into 27538separate groups. Insert @var{number} NOPs to force an insn to a new group. 27539@end table 27540 27541@item -mcall-sysv 27542@opindex mcall-sysv 27543On System V.4 and embedded PowerPC systems compile code using calling 27544conventions that adhere to the March 1995 draft of the System V 27545Application Binary Interface, PowerPC processor supplement. This is the 27546default unless you configured GCC using @samp{powerpc-*-eabiaix}. 27547 27548@item -mcall-sysv-eabi 27549@itemx -mcall-eabi 27550@opindex mcall-sysv-eabi 27551@opindex mcall-eabi 27552Specify both @option{-mcall-sysv} and @option{-meabi} options. 27553 27554@item -mcall-sysv-noeabi 27555@opindex mcall-sysv-noeabi 27556Specify both @option{-mcall-sysv} and @option{-mno-eabi} options. 27557 27558@item -mcall-aixdesc 27559@opindex m 27560On System V.4 and embedded PowerPC systems compile code for the AIX 27561operating system. 27562 27563@item -mcall-linux 27564@opindex mcall-linux 27565On System V.4 and embedded PowerPC systems compile code for the 27566Linux-based GNU system. 27567 27568@item -mcall-freebsd 27569@opindex mcall-freebsd 27570On System V.4 and embedded PowerPC systems compile code for the 27571FreeBSD operating system. 27572 27573@item -mcall-netbsd 27574@opindex mcall-netbsd 27575On System V.4 and embedded PowerPC systems compile code for the 27576NetBSD operating system. 27577 27578@item -mcall-openbsd 27579@opindex mcall-netbsd 27580On System V.4 and embedded PowerPC systems compile code for the 27581OpenBSD operating system. 27582 27583@item -mtraceback=@var{traceback_type} 27584@opindex mtraceback 27585Select the type of traceback table. Valid values for @var{traceback_type} 27586are @samp{full}, @samp{part}, and @samp{no}. 27587 27588@item -maix-struct-return 27589@opindex maix-struct-return 27590Return all structures in memory (as specified by the AIX ABI)@. 27591 27592@item -msvr4-struct-return 27593@opindex msvr4-struct-return 27594Return structures smaller than 8 bytes in registers (as specified by the 27595SVR4 ABI)@. 27596 27597@item -mabi=@var{abi-type} 27598@opindex mabi 27599Extend the current ABI with a particular extension, or remove such extension. 27600Valid values are: @samp{altivec}, @samp{no-altivec}, 27601@samp{ibmlongdouble}, @samp{ieeelongdouble}, 27602@samp{elfv1}, @samp{elfv2}, 27603and for AIX: @samp{vec-extabi}, @samp{vec-default}@. 27604 27605@item -mabi=ibmlongdouble 27606@opindex mabi=ibmlongdouble 27607Change the current ABI to use IBM extended-precision long double. 27608This is not likely to work if your system defaults to using IEEE 27609extended-precision long double. If you change the long double type 27610from IEEE extended-precision, the compiler will issue a warning unless 27611you use the @option{-Wno-psabi} option. Requires @option{-mlong-double-128} 27612to be enabled. 27613 27614@item -mabi=ieeelongdouble 27615@opindex mabi=ieeelongdouble 27616Change the current ABI to use IEEE extended-precision long double. 27617This is not likely to work if your system defaults to using IBM 27618extended-precision long double. If you change the long double type 27619from IBM extended-precision, the compiler will issue a warning unless 27620you use the @option{-Wno-psabi} option. Requires @option{-mlong-double-128} 27621to be enabled. 27622 27623@item -mabi=elfv1 27624@opindex mabi=elfv1 27625Change the current ABI to use the ELFv1 ABI. 27626This is the default ABI for big-endian PowerPC 64-bit Linux. 27627Overriding the default ABI requires special system support and is 27628likely to fail in spectacular ways. 27629 27630@item -mabi=elfv2 27631@opindex mabi=elfv2 27632Change the current ABI to use the ELFv2 ABI. 27633This is the default ABI for little-endian PowerPC 64-bit Linux. 27634Overriding the default ABI requires special system support and is 27635likely to fail in spectacular ways. 27636 27637@item -mgnu-attribute 27638@itemx -mno-gnu-attribute 27639@opindex mgnu-attribute 27640@opindex mno-gnu-attribute 27641Emit .gnu_attribute assembly directives to set tag/value pairs in a 27642.gnu.attributes section that specify ABI variations in function 27643parameters or return values. 27644 27645@item -mprototype 27646@itemx -mno-prototype 27647@opindex mprototype 27648@opindex mno-prototype 27649On System V.4 and embedded PowerPC systems assume that all calls to 27650variable argument functions are properly prototyped. Otherwise, the 27651compiler must insert an instruction before every non-prototyped call to 27652set or clear bit 6 of the condition code register (@code{CR}) to 27653indicate whether floating-point values are passed in the floating-point 27654registers in case the function takes variable arguments. With 27655@option{-mprototype}, only calls to prototyped variable argument functions 27656set or clear the bit. 27657 27658@item -msim 27659@opindex msim 27660On embedded PowerPC systems, assume that the startup module is called 27661@file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and 27662@file{libc.a}. This is the default for @samp{powerpc-*-eabisim} 27663configurations. 27664 27665@item -mmvme 27666@opindex mmvme 27667On embedded PowerPC systems, assume that the startup module is called 27668@file{crt0.o} and the standard C libraries are @file{libmvme.a} and 27669@file{libc.a}. 27670 27671@item -mads 27672@opindex mads 27673On embedded PowerPC systems, assume that the startup module is called 27674@file{crt0.o} and the standard C libraries are @file{libads.a} and 27675@file{libc.a}. 27676 27677@item -myellowknife 27678@opindex myellowknife 27679On embedded PowerPC systems, assume that the startup module is called 27680@file{crt0.o} and the standard C libraries are @file{libyk.a} and 27681@file{libc.a}. 27682 27683@item -mvxworks 27684@opindex mvxworks 27685On System V.4 and embedded PowerPC systems, specify that you are 27686compiling for a VxWorks system. 27687 27688@item -memb 27689@opindex memb 27690On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags 27691header to indicate that @samp{eabi} extended relocations are used. 27692 27693@item -meabi 27694@itemx -mno-eabi 27695@opindex meabi 27696@opindex mno-eabi 27697On System V.4 and embedded PowerPC systems do (do not) adhere to the 27698Embedded Applications Binary Interface (EABI), which is a set of 27699modifications to the System V.4 specifications. Selecting @option{-meabi} 27700means that the stack is aligned to an 8-byte boundary, a function 27701@code{__eabi} is called from @code{main} to set up the EABI 27702environment, and the @option{-msdata} option can use both @code{r2} and 27703@code{r13} to point to two separate small data areas. Selecting 27704@option{-mno-eabi} means that the stack is aligned to a 16-byte boundary, 27705no EABI initialization function is called from @code{main}, and the 27706@option{-msdata} option only uses @code{r13} to point to a single 27707small data area. The @option{-meabi} option is on by default if you 27708configured GCC using one of the @samp{powerpc*-*-eabi*} options. 27709 27710@item -msdata=eabi 27711@opindex msdata=eabi 27712On System V.4 and embedded PowerPC systems, put small initialized 27713@code{const} global and static data in the @code{.sdata2} section, which 27714is pointed to by register @code{r2}. Put small initialized 27715non-@code{const} global and static data in the @code{.sdata} section, 27716which is pointed to by register @code{r13}. Put small uninitialized 27717global and static data in the @code{.sbss} section, which is adjacent to 27718the @code{.sdata} section. The @option{-msdata=eabi} option is 27719incompatible with the @option{-mrelocatable} option. The 27720@option{-msdata=eabi} option also sets the @option{-memb} option. 27721 27722@item -msdata=sysv 27723@opindex msdata=sysv 27724On System V.4 and embedded PowerPC systems, put small global and static 27725data in the @code{.sdata} section, which is pointed to by register 27726@code{r13}. Put small uninitialized global and static data in the 27727@code{.sbss} section, which is adjacent to the @code{.sdata} section. 27728The @option{-msdata=sysv} option is incompatible with the 27729@option{-mrelocatable} option. 27730 27731@item -msdata=default 27732@itemx -msdata 27733@opindex msdata=default 27734@opindex msdata 27735On System V.4 and embedded PowerPC systems, if @option{-meabi} is used, 27736compile code the same as @option{-msdata=eabi}, otherwise compile code the 27737same as @option{-msdata=sysv}. 27738 27739@item -msdata=data 27740@opindex msdata=data 27741On System V.4 and embedded PowerPC systems, put small global 27742data in the @code{.sdata} section. Put small uninitialized global 27743data in the @code{.sbss} section. Do not use register @code{r13} 27744to address small data however. This is the default behavior unless 27745other @option{-msdata} options are used. 27746 27747@item -msdata=none 27748@itemx -mno-sdata 27749@opindex msdata=none 27750@opindex mno-sdata 27751On embedded PowerPC systems, put all initialized global and static data 27752in the @code{.data} section, and all uninitialized data in the 27753@code{.bss} section. 27754 27755@item -mreadonly-in-sdata 27756@opindex mreadonly-in-sdata 27757@opindex mno-readonly-in-sdata 27758Put read-only objects in the @code{.sdata} section as well. This is the 27759default. 27760 27761@item -mblock-move-inline-limit=@var{num} 27762@opindex mblock-move-inline-limit 27763Inline all block moves (such as calls to @code{memcpy} or structure 27764copies) less than or equal to @var{num} bytes. The minimum value for 27765@var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit 27766targets. The default value is target-specific. 27767 27768@item -mblock-compare-inline-limit=@var{num} 27769@opindex mblock-compare-inline-limit 27770Generate non-looping inline code for all block compares (such as calls 27771to @code{memcmp} or structure compares) less than or equal to @var{num} 27772bytes. If @var{num} is 0, all inline expansion (non-loop and loop) of 27773block compare is disabled. The default value is target-specific. 27774 27775@item -mblock-compare-inline-loop-limit=@var{num} 27776@opindex mblock-compare-inline-loop-limit 27777Generate an inline expansion using loop code for all block compares that 27778are less than or equal to @var{num} bytes, but greater than the limit 27779for non-loop inline block compare expansion. If the block length is not 27780constant, at most @var{num} bytes will be compared before @code{memcmp} 27781is called to compare the remainder of the block. The default value is 27782target-specific. 27783 27784@item -mstring-compare-inline-limit=@var{num} 27785@opindex mstring-compare-inline-limit 27786Compare at most @var{num} string bytes with inline code. 27787If the difference or end of string is not found at the 27788end of the inline compare a call to @code{strcmp} or @code{strncmp} will 27789take care of the rest of the comparison. The default is 64 bytes. 27790 27791@item -G @var{num} 27792@opindex G 27793@cindex smaller data references (PowerPC) 27794@cindex .sdata/.sdata2 references (PowerPC) 27795On embedded PowerPC systems, put global and static items less than or 27796equal to @var{num} bytes into the small data or BSS sections instead of 27797the normal data or BSS section. By default, @var{num} is 8. The 27798@option{-G @var{num}} switch is also passed to the linker. 27799All modules should be compiled with the same @option{-G @var{num}} value. 27800 27801@item -mregnames 27802@itemx -mno-regnames 27803@opindex mregnames 27804@opindex mno-regnames 27805On System V.4 and embedded PowerPC systems do (do not) emit register 27806names in the assembly language output using symbolic forms. 27807 27808@item -mlongcall 27809@itemx -mno-longcall 27810@opindex mlongcall 27811@opindex mno-longcall 27812By default assume that all calls are far away so that a longer and more 27813expensive calling sequence is required. This is required for calls 27814farther than 32 megabytes (33,554,432 bytes) from the current location. 27815A short call is generated if the compiler knows 27816the call cannot be that far away. This setting can be overridden by 27817the @code{shortcall} function attribute, or by @code{#pragma 27818longcall(0)}. 27819 27820Some linkers are capable of detecting out-of-range calls and generating 27821glue code on the fly. On these systems, long calls are unnecessary and 27822generate slower code. As of this writing, the AIX linker can do this, 27823as can the GNU linker for PowerPC/64. It is planned to add this feature 27824to the GNU linker for 32-bit PowerPC systems as well. 27825 27826On PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU linkers, 27827GCC can generate long calls using an inline PLT call sequence (see 27828@option{-mpltseq}). PowerPC with @option{-mbss-plt} and PowerPC64 27829ELFv1 (big-endian) do not support inline PLT calls. 27830 27831On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr 27832callee, L42}, plus a @dfn{branch island} (glue code). The two target 27833addresses represent the callee and the branch island. The 27834Darwin/PPC linker prefers the first address and generates a @code{bl 27835callee} if the PPC @code{bl} instruction reaches the callee directly; 27836otherwise, the linker generates @code{bl L42} to call the branch 27837island. The branch island is appended to the body of the 27838calling function; it computes the full 32-bit address of the callee 27839and jumps to it. 27840 27841On Mach-O (Darwin) systems, this option directs the compiler emit to 27842the glue for every direct call, and the Darwin linker decides whether 27843to use or discard it. 27844 27845In the future, GCC may ignore all longcall specifications 27846when the linker is known to generate glue. 27847 27848@item -mpltseq 27849@itemx -mno-pltseq 27850@opindex mpltseq 27851@opindex mno-pltseq 27852Implement (do not implement) -fno-plt and long calls using an inline 27853PLT call sequence that supports lazy linking and long calls to 27854functions in dlopen'd shared libraries. Inline PLT calls are only 27855supported on PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU 27856linkers, and are enabled by default if the support is detected when 27857configuring GCC, and, in the case of 32-bit PowerPC, if GCC is 27858configured with @option{--enable-secureplt}. @option{-mpltseq} code 27859and @option{-mbss-plt} 32-bit PowerPC relocatable objects may not be 27860linked together. 27861 27862@item -mtls-markers 27863@itemx -mno-tls-markers 27864@opindex mtls-markers 27865@opindex mno-tls-markers 27866Mark (do not mark) calls to @code{__tls_get_addr} with a relocation 27867specifying the function argument. The relocation allows the linker to 27868reliably associate function call with argument setup instructions for 27869TLS optimization, which in turn allows GCC to better schedule the 27870sequence. 27871 27872@item -mrecip 27873@itemx -mno-recip 27874@opindex mrecip 27875This option enables use of the reciprocal estimate and 27876reciprocal square root estimate instructions with additional 27877Newton-Raphson steps to increase precision instead of doing a divide or 27878square root and divide for floating-point arguments. You should use 27879the @option{-ffast-math} option when using @option{-mrecip} (or at 27880least @option{-funsafe-math-optimizations}, 27881@option{-ffinite-math-only}, @option{-freciprocal-math} and 27882@option{-fno-trapping-math}). Note that while the throughput of the 27883sequence is generally higher than the throughput of the non-reciprocal 27884instruction, the precision of the sequence can be decreased by up to 2 27885ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square 27886roots. 27887 27888@item -mrecip=@var{opt} 27889@opindex mrecip=opt 27890This option controls which reciprocal estimate instructions 27891may be used. @var{opt} is a comma-separated list of options, which may 27892be preceded by a @code{!} to invert the option: 27893 27894@table @samp 27895 27896@item all 27897Enable all estimate instructions. 27898 27899@item default 27900Enable the default instructions, equivalent to @option{-mrecip}. 27901 27902@item none 27903Disable all estimate instructions, equivalent to @option{-mno-recip}. 27904 27905@item div 27906Enable the reciprocal approximation instructions for both 27907single and double precision. 27908 27909@item divf 27910Enable the single-precision reciprocal approximation instructions. 27911 27912@item divd 27913Enable the double-precision reciprocal approximation instructions. 27914 27915@item rsqrt 27916Enable the reciprocal square root approximation instructions for both 27917single and double precision. 27918 27919@item rsqrtf 27920Enable the single-precision reciprocal square root approximation instructions. 27921 27922@item rsqrtd 27923Enable the double-precision reciprocal square root approximation instructions. 27924 27925@end table 27926 27927So, for example, @option{-mrecip=all,!rsqrtd} enables 27928all of the reciprocal estimate instructions, except for the 27929@code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions 27930which handle the double-precision reciprocal square root calculations. 27931 27932@item -mrecip-precision 27933@itemx -mno-recip-precision 27934@opindex mrecip-precision 27935Assume (do not assume) that the reciprocal estimate instructions 27936provide higher-precision estimates than is mandated by the PowerPC 27937ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or 27938@option{-mcpu=power8} automatically selects @option{-mrecip-precision}. 27939The double-precision square root estimate instructions are not generated by 27940default on low-precision machines, since they do not provide an 27941estimate that converges after three steps. 27942 27943@item -mveclibabi=@var{type} 27944@opindex mveclibabi 27945Specifies the ABI type to use for vectorizing intrinsics using an 27946external library. The only type supported at present is @samp{mass}, 27947which specifies to use IBM's Mathematical Acceleration Subsystem 27948(MASS) libraries for vectorizing intrinsics using external libraries. 27949GCC currently emits calls to @code{acosd2}, @code{acosf4}, 27950@code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4}, 27951@code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4}, 27952@code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4}, 27953@code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4}, 27954@code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4}, 27955@code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4}, 27956@code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4}, 27957@code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4}, 27958@code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4}, 27959@code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4}, 27960@code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2}, 27961@code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2}, 27962@code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code 27963for power7. Both @option{-ftree-vectorize} and 27964@option{-funsafe-math-optimizations} must also be enabled. The MASS 27965libraries must be specified at link time. 27966 27967@item -mfriz 27968@itemx -mno-friz 27969@opindex mfriz 27970Generate (do not generate) the @code{friz} instruction when the 27971@option{-funsafe-math-optimizations} option is used to optimize 27972rounding of floating-point values to 64-bit integer and back to floating 27973point. The @code{friz} instruction does not return the same value if 27974the floating-point number is too large to fit in an integer. 27975 27976@item -mpointers-to-nested-functions 27977@itemx -mno-pointers-to-nested-functions 27978@opindex mpointers-to-nested-functions 27979Generate (do not generate) code to load up the static chain register 27980(@code{r11}) when calling through a pointer on AIX and 64-bit Linux 27981systems where a function pointer points to a 3-word descriptor giving 27982the function address, TOC value to be loaded in register @code{r2}, and 27983static chain value to be loaded in register @code{r11}. The 27984@option{-mpointers-to-nested-functions} is on by default. You cannot 27985call through pointers to nested functions or pointers 27986to functions compiled in other languages that use the static chain if 27987you use @option{-mno-pointers-to-nested-functions}. 27988 27989@item -msave-toc-indirect 27990@itemx -mno-save-toc-indirect 27991@opindex msave-toc-indirect 27992Generate (do not generate) code to save the TOC value in the reserved 27993stack location in the function prologue if the function calls through 27994a pointer on AIX and 64-bit Linux systems. If the TOC value is not 27995saved in the prologue, it is saved just before the call through the 27996pointer. The @option{-mno-save-toc-indirect} option is the default. 27997 27998@item -mcompat-align-parm 27999@itemx -mno-compat-align-parm 28000@opindex mcompat-align-parm 28001Generate (do not generate) code to pass structure parameters with a 28002maximum alignment of 64 bits, for compatibility with older versions 28003of GCC. 28004 28005Older versions of GCC (prior to 4.9.0) incorrectly did not align a 28006structure parameter on a 128-bit boundary when that structure contained 28007a member requiring 128-bit alignment. This is corrected in more 28008recent versions of GCC. This option may be used to generate code 28009that is compatible with functions compiled with older versions of 28010GCC. 28011 28012The @option{-mno-compat-align-parm} option is the default. 28013 28014@item -mstack-protector-guard=@var{guard} 28015@itemx -mstack-protector-guard-reg=@var{reg} 28016@itemx -mstack-protector-guard-offset=@var{offset} 28017@itemx -mstack-protector-guard-symbol=@var{symbol} 28018@opindex mstack-protector-guard 28019@opindex mstack-protector-guard-reg 28020@opindex mstack-protector-guard-offset 28021@opindex mstack-protector-guard-symbol 28022Generate stack protection code using canary at @var{guard}. Supported 28023locations are @samp{global} for global canary or @samp{tls} for per-thread 28024canary in the TLS block (the default with GNU libc version 2.4 or later). 28025 28026With the latter choice the options 28027@option{-mstack-protector-guard-reg=@var{reg}} and 28028@option{-mstack-protector-guard-offset=@var{offset}} furthermore specify 28029which register to use as base register for reading the canary, and from what 28030offset from that base register. The default for those is as specified in the 28031relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides 28032the offset with a symbol reference to a canary in the TLS block. 28033 28034@item -mpcrel 28035@itemx -mno-pcrel 28036@opindex mpcrel 28037@opindex mno-pcrel 28038Generate (do not generate) pc-relative addressing when the option 28039@option{-mcpu=future} is used. The @option{-mpcrel} option requires 28040that the medium code model (@option{-mcmodel=medium}) and prefixed 28041addressing (@option{-mprefixed}) options are enabled. 28042 28043@item -mprefixed 28044@itemx -mno-prefixed 28045@opindex mprefixed 28046@opindex mno-prefixed 28047Generate (do not generate) addressing modes using prefixed load and 28048store instructions when the option @option{-mcpu=future} is used. 28049 28050@item -mmma 28051@itemx -mno-mma 28052@opindex mmma 28053@opindex mno-mma 28054Generate (do not generate) the MMA instructions when the option 28055@option{-mcpu=future} is used. 28056 28057@item -mrop-protect 28058@itemx -mno-rop-protect 28059@opindex mrop-protect 28060@opindex mno-rop-protect 28061Generate (do not generate) ROP protection instructions when the target 28062processor supports them. Currently this option disables the shrink-wrap 28063optimization (@option{-fshrink-wrap}). 28064 28065@item -mprivileged 28066@itemx -mno-privileged 28067@opindex mprivileged 28068@opindex mno-privileged 28069Generate (do not generate) code that will run in privileged state. 28070 28071@item -mblock-ops-unaligned-vsx 28072@itemx -mno-block-ops-unaligned-vsx 28073@opindex block-ops-unaligned-vsx 28074@opindex no-block-ops-unaligned-vsx 28075Generate (do not generate) unaligned vsx loads and stores for 28076inline expansion of @code{memcpy} and @code{memmove}. 28077@end table 28078 28079@node RX Options 28080@subsection RX Options 28081@cindex RX Options 28082 28083These command-line options are defined for RX targets: 28084 28085@table @gcctabopt 28086@item -m64bit-doubles 28087@itemx -m32bit-doubles 28088@opindex m64bit-doubles 28089@opindex m32bit-doubles 28090Make the @code{double} data type be 64 bits (@option{-m64bit-doubles}) 28091or 32 bits (@option{-m32bit-doubles}) in size. The default is 28092@option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only 28093works on 32-bit values, which is why the default is 28094@option{-m32bit-doubles}. 28095 28096@item -fpu 28097@itemx -nofpu 28098@opindex fpu 28099@opindex nofpu 28100Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX 28101floating-point hardware. The default is enabled for the RX600 28102series and disabled for the RX200 series. 28103 28104Floating-point instructions are only generated for 32-bit floating-point 28105values, however, so the FPU hardware is not used for doubles if the 28106@option{-m64bit-doubles} option is used. 28107 28108@emph{Note} If the @option{-fpu} option is enabled then 28109@option{-funsafe-math-optimizations} is also enabled automatically. 28110This is because the RX FPU instructions are themselves unsafe. 28111 28112@item -mcpu=@var{name} 28113@opindex mcpu 28114Selects the type of RX CPU to be targeted. Currently three types are 28115supported, the generic @samp{RX600} and @samp{RX200} series hardware and 28116the specific @samp{RX610} CPU. The default is @samp{RX600}. 28117 28118The only difference between @samp{RX600} and @samp{RX610} is that the 28119@samp{RX610} does not support the @code{MVTIPL} instruction. 28120 28121The @samp{RX200} series does not have a hardware floating-point unit 28122and so @option{-nofpu} is enabled by default when this type is 28123selected. 28124 28125@item -mbig-endian-data 28126@itemx -mlittle-endian-data 28127@opindex mbig-endian-data 28128@opindex mlittle-endian-data 28129Store data (but not code) in the big-endian format. The default is 28130@option{-mlittle-endian-data}, i.e.@: to store data in the little-endian 28131format. 28132 28133@item -msmall-data-limit=@var{N} 28134@opindex msmall-data-limit 28135Specifies the maximum size in bytes of global and static variables 28136which can be placed into the small data area. Using the small data 28137area can lead to smaller and faster code, but the size of area is 28138limited and it is up to the programmer to ensure that the area does 28139not overflow. Also when the small data area is used one of the RX's 28140registers (usually @code{r13}) is reserved for use pointing to this 28141area, so it is no longer available for use by the compiler. This 28142could result in slower and/or larger code if variables are pushed onto 28143the stack instead of being held in this register. 28144 28145Note, common variables (variables that have not been initialized) and 28146constants are not placed into the small data area as they are assigned 28147to other sections in the output executable. 28148 28149The default value is zero, which disables this feature. Note, this 28150feature is not enabled by default with higher optimization levels 28151(@option{-O2} etc) because of the potentially detrimental effects of 28152reserving a register. It is up to the programmer to experiment and 28153discover whether this feature is of benefit to their program. See the 28154description of the @option{-mpid} option for a description of how the 28155actual register to hold the small data area pointer is chosen. 28156 28157@item -msim 28158@itemx -mno-sim 28159@opindex msim 28160@opindex mno-sim 28161Use the simulator runtime. The default is to use the libgloss 28162board-specific runtime. 28163 28164@item -mas100-syntax 28165@itemx -mno-as100-syntax 28166@opindex mas100-syntax 28167@opindex mno-as100-syntax 28168When generating assembler output use a syntax that is compatible with 28169Renesas's AS100 assembler. This syntax can also be handled by the GAS 28170assembler, but it has some restrictions so it is not generated by default. 28171 28172@item -mmax-constant-size=@var{N} 28173@opindex mmax-constant-size 28174Specifies the maximum size, in bytes, of a constant that can be used as 28175an operand in a RX instruction. Although the RX instruction set does 28176allow constants of up to 4 bytes in length to be used in instructions, 28177a longer value equates to a longer instruction. Thus in some 28178circumstances it can be beneficial to restrict the size of constants 28179that are used in instructions. Constants that are too big are instead 28180placed into a constant pool and referenced via register indirection. 28181 28182The value @var{N} can be between 0 and 4. A value of 0 (the default) 28183or 4 means that constants of any size are allowed. 28184 28185@item -mrelax 28186@opindex mrelax 28187Enable linker relaxation. Linker relaxation is a process whereby the 28188linker attempts to reduce the size of a program by finding shorter 28189versions of various instructions. Disabled by default. 28190 28191@item -mint-register=@var{N} 28192@opindex mint-register 28193Specify the number of registers to reserve for fast interrupt handler 28194functions. The value @var{N} can be between 0 and 4. A value of 1 28195means that register @code{r13} is reserved for the exclusive use 28196of fast interrupt handlers. A value of 2 reserves @code{r13} and 28197@code{r12}. A value of 3 reserves @code{r13}, @code{r12} and 28198@code{r11}, and a value of 4 reserves @code{r13} through @code{r10}. 28199A value of 0, the default, does not reserve any registers. 28200 28201@item -msave-acc-in-interrupts 28202@opindex msave-acc-in-interrupts 28203Specifies that interrupt handler functions should preserve the 28204accumulator register. This is only necessary if normal code might use 28205the accumulator register, for example because it performs 64-bit 28206multiplications. The default is to ignore the accumulator as this 28207makes the interrupt handlers faster. 28208 28209@item -mpid 28210@itemx -mno-pid 28211@opindex mpid 28212@opindex mno-pid 28213Enables the generation of position independent data. When enabled any 28214access to constant data is done via an offset from a base address 28215held in a register. This allows the location of constant data to be 28216determined at run time without requiring the executable to be 28217relocated, which is a benefit to embedded applications with tight 28218memory constraints. Data that can be modified is not affected by this 28219option. 28220 28221Note, using this feature reserves a register, usually @code{r13}, for 28222the constant data base address. This can result in slower and/or 28223larger code, especially in complicated functions. 28224 28225The actual register chosen to hold the constant data base address 28226depends upon whether the @option{-msmall-data-limit} and/or the 28227@option{-mint-register} command-line options are enabled. Starting 28228with register @code{r13} and proceeding downwards, registers are 28229allocated first to satisfy the requirements of @option{-mint-register}, 28230then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it 28231is possible for the small data area register to be @code{r8} if both 28232@option{-mint-register=4} and @option{-mpid} are specified on the 28233command line. 28234 28235By default this feature is not enabled. The default can be restored 28236via the @option{-mno-pid} command-line option. 28237 28238@item -mno-warn-multiple-fast-interrupts 28239@itemx -mwarn-multiple-fast-interrupts 28240@opindex mno-warn-multiple-fast-interrupts 28241@opindex mwarn-multiple-fast-interrupts 28242Prevents GCC from issuing a warning message if it finds more than one 28243fast interrupt handler when it is compiling a file. The default is to 28244issue a warning for each extra fast interrupt handler found, as the RX 28245only supports one such interrupt. 28246 28247@item -mallow-string-insns 28248@itemx -mno-allow-string-insns 28249@opindex mallow-string-insns 28250@opindex mno-allow-string-insns 28251Enables or disables the use of the string manipulation instructions 28252@code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL} 28253@code{SWHILE} and also the @code{RMPA} instruction. These 28254instructions may prefetch data, which is not safe to do if accessing 28255an I/O register. (See section 12.2.7 of the RX62N Group User's Manual 28256for more information). 28257 28258The default is to allow these instructions, but it is not possible for 28259GCC to reliably detect all circumstances where a string instruction 28260might be used to access an I/O register, so their use cannot be 28261disabled automatically. Instead it is reliant upon the programmer to 28262use the @option{-mno-allow-string-insns} option if their program 28263accesses I/O space. 28264 28265When the instructions are enabled GCC defines the C preprocessor 28266symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the 28267symbol @code{__RX_DISALLOW_STRING_INSNS__}. 28268 28269@item -mjsr 28270@itemx -mno-jsr 28271@opindex mjsr 28272@opindex mno-jsr 28273Use only (or not only) @code{JSR} instructions to access functions. 28274This option can be used when code size exceeds the range of @code{BSR} 28275instructions. Note that @option{-mno-jsr} does not mean to not use 28276@code{JSR} but instead means that any type of branch may be used. 28277@end table 28278 28279@emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}} 28280has special significance to the RX port when used with the 28281@code{interrupt} function attribute. This attribute indicates a 28282function intended to process fast interrupts. GCC ensures 28283that it only uses the registers @code{r10}, @code{r11}, @code{r12} 28284and/or @code{r13} and only provided that the normal use of the 28285corresponding registers have been restricted via the 28286@option{-ffixed-@var{reg}} or @option{-mint-register} command-line 28287options. 28288 28289@node S/390 and zSeries Options 28290@subsection S/390 and zSeries Options 28291@cindex S/390 and zSeries Options 28292 28293These are the @samp{-m} options defined for the S/390 and zSeries architecture. 28294 28295@table @gcctabopt 28296@item -mhard-float 28297@itemx -msoft-float 28298@opindex mhard-float 28299@opindex msoft-float 28300Use (do not use) the hardware floating-point instructions and registers 28301for floating-point operations. When @option{-msoft-float} is specified, 28302functions in @file{libgcc.a} are used to perform floating-point 28303operations. When @option{-mhard-float} is specified, the compiler 28304generates IEEE floating-point instructions. This is the default. 28305 28306@item -mhard-dfp 28307@itemx -mno-hard-dfp 28308@opindex mhard-dfp 28309@opindex mno-hard-dfp 28310Use (do not use) the hardware decimal-floating-point instructions for 28311decimal-floating-point operations. When @option{-mno-hard-dfp} is 28312specified, functions in @file{libgcc.a} are used to perform 28313decimal-floating-point operations. When @option{-mhard-dfp} is 28314specified, the compiler generates decimal-floating-point hardware 28315instructions. This is the default for @option{-march=z9-ec} or higher. 28316 28317@item -mlong-double-64 28318@itemx -mlong-double-128 28319@opindex mlong-double-64 28320@opindex mlong-double-128 28321These switches control the size of @code{long double} type. A size 28322of 64 bits makes the @code{long double} type equivalent to the @code{double} 28323type. This is the default. 28324 28325@item -mbackchain 28326@itemx -mno-backchain 28327@opindex mbackchain 28328@opindex mno-backchain 28329Store (do not store) the address of the caller's frame as backchain pointer 28330into the callee's stack frame. 28331A backchain may be needed to allow debugging using tools that do not understand 28332DWARF call frame information. 28333When @option{-mno-packed-stack} is in effect, the backchain pointer is stored 28334at the bottom of the stack frame; when @option{-mpacked-stack} is in effect, 28335the backchain is placed into the topmost word of the 96/160 byte register 28336save area. 28337 28338In general, code compiled with @option{-mbackchain} is call-compatible with 28339code compiled with @option{-mno-backchain}; however, use of the backchain 28340for debugging purposes usually requires that the whole binary is built with 28341@option{-mbackchain}. Note that the combination of @option{-mbackchain}, 28342@option{-mpacked-stack} and @option{-mhard-float} is not supported. In order 28343to build a linux kernel use @option{-msoft-float}. 28344 28345The default is to not maintain the backchain. 28346 28347@item -mpacked-stack 28348@itemx -mno-packed-stack 28349@opindex mpacked-stack 28350@opindex mno-packed-stack 28351Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is 28352specified, the compiler uses the all fields of the 96/160 byte register save 28353area only for their default purpose; unused fields still take up stack space. 28354When @option{-mpacked-stack} is specified, register save slots are densely 28355packed at the top of the register save area; unused space is reused for other 28356purposes, allowing for more efficient use of the available stack space. 28357However, when @option{-mbackchain} is also in effect, the topmost word of 28358the save area is always used to store the backchain, and the return address 28359register is always saved two words below the backchain. 28360 28361As long as the stack frame backchain is not used, code generated with 28362@option{-mpacked-stack} is call-compatible with code generated with 28363@option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for 28364S/390 or zSeries generated code that uses the stack frame backchain at run 28365time, not just for debugging purposes. Such code is not call-compatible 28366with code compiled with @option{-mpacked-stack}. Also, note that the 28367combination of @option{-mbackchain}, 28368@option{-mpacked-stack} and @option{-mhard-float} is not supported. In order 28369to build a linux kernel use @option{-msoft-float}. 28370 28371The default is to not use the packed stack layout. 28372 28373@item -msmall-exec 28374@itemx -mno-small-exec 28375@opindex msmall-exec 28376@opindex mno-small-exec 28377Generate (or do not generate) code using the @code{bras} instruction 28378to do subroutine calls. 28379This only works reliably if the total executable size does not 28380exceed 64k. The default is to use the @code{basr} instruction instead, 28381which does not have this limitation. 28382 28383@item -m64 28384@itemx -m31 28385@opindex m64 28386@opindex m31 28387When @option{-m31} is specified, generate code compliant to the 28388GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate 28389code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in 28390particular to generate 64-bit instructions. For the @samp{s390} 28391targets, the default is @option{-m31}, while the @samp{s390x} 28392targets default to @option{-m64}. 28393 28394@item -mzarch 28395@itemx -mesa 28396@opindex mzarch 28397@opindex mesa 28398When @option{-mzarch} is specified, generate code using the 28399instructions available on z/Architecture. 28400When @option{-mesa} is specified, generate code using the 28401instructions available on ESA/390. Note that @option{-mesa} is 28402not possible with @option{-m64}. 28403When generating code compliant to the GNU/Linux for S/390 ABI, 28404the default is @option{-mesa}. When generating code compliant 28405to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}. 28406 28407@item -mhtm 28408@itemx -mno-htm 28409@opindex mhtm 28410@opindex mno-htm 28411The @option{-mhtm} option enables a set of builtins making use of 28412instructions available with the transactional execution facility 28413introduced with the IBM zEnterprise EC12 machine generation 28414@ref{S/390 System z Built-in Functions}. 28415@option{-mhtm} is enabled by default when using @option{-march=zEC12}. 28416 28417@item -mvx 28418@itemx -mno-vx 28419@opindex mvx 28420@opindex mno-vx 28421When @option{-mvx} is specified, generate code using the instructions 28422available with the vector extension facility introduced with the IBM 28423z13 machine generation. 28424This option changes the ABI for some vector type values with regard to 28425alignment and calling conventions. In case vector type values are 28426being used in an ABI-relevant context a GAS @samp{.gnu_attribute} 28427command will be added to mark the resulting binary with the ABI used. 28428@option{-mvx} is enabled by default when using @option{-march=z13}. 28429 28430@item -mzvector 28431@itemx -mno-zvector 28432@opindex mzvector 28433@opindex mno-zvector 28434The @option{-mzvector} option enables vector language extensions and 28435builtins using instructions available with the vector extension 28436facility introduced with the IBM z13 machine generation. 28437This option adds support for @samp{vector} to be used as a keyword to 28438define vector type variables and arguments. @samp{vector} is only 28439available when GNU extensions are enabled. It will not be expanded 28440when requesting strict standard compliance e.g.@: with @option{-std=c99}. 28441In addition to the GCC low-level builtins @option{-mzvector} enables 28442a set of builtins added for compatibility with AltiVec-style 28443implementations like Power and Cell. In order to make use of these 28444builtins the header file @file{vecintrin.h} needs to be included. 28445@option{-mzvector} is disabled by default. 28446 28447@item -mmvcle 28448@itemx -mno-mvcle 28449@opindex mmvcle 28450@opindex mno-mvcle 28451Generate (or do not generate) code using the @code{mvcle} instruction 28452to perform block moves. When @option{-mno-mvcle} is specified, 28453use a @code{mvc} loop instead. This is the default unless optimizing for 28454size. 28455 28456@item -mdebug 28457@itemx -mno-debug 28458@opindex mdebug 28459@opindex mno-debug 28460Print (or do not print) additional debug information when compiling. 28461The default is to not print debug information. 28462 28463@item -march=@var{cpu-type} 28464@opindex march 28465Generate code that runs on @var{cpu-type}, which is the name of a 28466system representing a certain processor type. Possible values for 28467@var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6}, 28468@samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8}, 28469@samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11}, 28470@samp{z14}/@samp{arch12}, @samp{z15}/@samp{arch13}, and @samp{native}. 28471 28472The default is @option{-march=z900}. 28473 28474Specifying @samp{native} as cpu type can be used to select the best 28475architecture option for the host processor. 28476@option{-march=native} has no effect if GCC does not recognize the 28477processor. 28478 28479@item -mtune=@var{cpu-type} 28480@opindex mtune 28481Tune to @var{cpu-type} everything applicable about the generated code, 28482except for the ABI and the set of available instructions. 28483The list of @var{cpu-type} values is the same as for @option{-march}. 28484The default is the value used for @option{-march}. 28485 28486@item -mtpf-trace 28487@itemx -mno-tpf-trace 28488@opindex mtpf-trace 28489@opindex mno-tpf-trace 28490Generate code that adds (does not add) in TPF OS specific branches to trace 28491routines in the operating system. This option is off by default, even 28492when compiling for the TPF OS@. 28493 28494@item -mtpf-trace-skip 28495@itemx -mno-tpf-trace-skip 28496@opindex mtpf-trace-skip 28497@opindex mno-tpf-trace-skip 28498Generate code that changes (does not change) the default branch 28499targets enabled by @option{-mtpf-trace} to point to specialized trace 28500routines providing the ability of selectively skipping function trace 28501entries for the TPF OS. This option is off by default, even when 28502compiling for the TPF OS and specifying @option{-mtpf-trace}. 28503 28504@item -mfused-madd 28505@itemx -mno-fused-madd 28506@opindex mfused-madd 28507@opindex mno-fused-madd 28508Generate code that uses (does not use) the floating-point multiply and 28509accumulate instructions. These instructions are generated by default if 28510hardware floating point is used. 28511 28512@item -mwarn-framesize=@var{framesize} 28513@opindex mwarn-framesize 28514Emit a warning if the current function exceeds the given frame size. Because 28515this is a compile-time check it doesn't need to be a real problem when the program 28516runs. It is intended to identify functions that most probably cause 28517a stack overflow. It is useful to be used in an environment with limited stack 28518size e.g.@: the linux kernel. 28519 28520@item -mwarn-dynamicstack 28521@opindex mwarn-dynamicstack 28522Emit a warning if the function calls @code{alloca} or uses dynamically-sized 28523arrays. This is generally a bad idea with a limited stack size. 28524 28525@item -mstack-guard=@var{stack-guard} 28526@itemx -mstack-size=@var{stack-size} 28527@opindex mstack-guard 28528@opindex mstack-size 28529If these options are provided the S/390 back end emits additional instructions in 28530the function prologue that trigger a trap if the stack size is @var{stack-guard} 28531bytes above the @var{stack-size} (remember that the stack on S/390 grows downward). 28532If the @var{stack-guard} option is omitted the smallest power of 2 larger than 28533the frame size of the compiled function is chosen. 28534These options are intended to be used to help debugging stack overflow problems. 28535The additionally emitted code causes only little overhead and hence can also be 28536used in production-like systems without greater performance degradation. The given 28537values have to be exact powers of 2 and @var{stack-size} has to be greater than 28538@var{stack-guard} without exceeding 64k. 28539In order to be efficient the extra code makes the assumption that the stack starts 28540at an address aligned to the value given by @var{stack-size}. 28541The @var{stack-guard} option can only be used in conjunction with @var{stack-size}. 28542 28543@item -mhotpatch=@var{pre-halfwords},@var{post-halfwords} 28544@opindex mhotpatch 28545If the hotpatch option is enabled, a ``hot-patching'' function 28546prologue is generated for all functions in the compilation unit. 28547The funtion label is prepended with the given number of two-byte 28548NOP instructions (@var{pre-halfwords}, maximum 1000000). After 28549the label, 2 * @var{post-halfwords} bytes are appended, using the 28550largest NOP like instructions the architecture allows (maximum 285511000000). 28552 28553If both arguments are zero, hotpatching is disabled. 28554 28555This option can be overridden for individual functions with the 28556@code{hotpatch} attribute. 28557@end table 28558 28559@node Score Options 28560@subsection Score Options 28561@cindex Score Options 28562 28563These options are defined for Score implementations: 28564 28565@table @gcctabopt 28566@item -meb 28567@opindex meb 28568Compile code for big-endian mode. This is the default. 28569 28570@item -mel 28571@opindex mel 28572Compile code for little-endian mode. 28573 28574@item -mnhwloop 28575@opindex mnhwloop 28576Disable generation of @code{bcnz} instructions. 28577 28578@item -muls 28579@opindex muls 28580Enable generation of unaligned load and store instructions. 28581 28582@item -mmac 28583@opindex mmac 28584Enable the use of multiply-accumulate instructions. Disabled by default. 28585 28586@item -mscore5 28587@opindex mscore5 28588Specify the SCORE5 as the target architecture. 28589 28590@item -mscore5u 28591@opindex mscore5u 28592Specify the SCORE5U of the target architecture. 28593 28594@item -mscore7 28595@opindex mscore7 28596Specify the SCORE7 as the target architecture. This is the default. 28597 28598@item -mscore7d 28599@opindex mscore7d 28600Specify the SCORE7D as the target architecture. 28601@end table 28602 28603@node SH Options 28604@subsection SH Options 28605 28606These @samp{-m} options are defined for the SH implementations: 28607 28608@table @gcctabopt 28609@item -m1 28610@opindex m1 28611Generate code for the SH1. 28612 28613@item -m2 28614@opindex m2 28615Generate code for the SH2. 28616 28617@item -m2e 28618Generate code for the SH2e. 28619 28620@item -m2a-nofpu 28621@opindex m2a-nofpu 28622Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way 28623that the floating-point unit is not used. 28624 28625@item -m2a-single-only 28626@opindex m2a-single-only 28627Generate code for the SH2a-FPU, in such a way that no double-precision 28628floating-point operations are used. 28629 28630@item -m2a-single 28631@opindex m2a-single 28632Generate code for the SH2a-FPU assuming the floating-point unit is in 28633single-precision mode by default. 28634 28635@item -m2a 28636@opindex m2a 28637Generate code for the SH2a-FPU assuming the floating-point unit is in 28638double-precision mode by default. 28639 28640@item -m3 28641@opindex m3 28642Generate code for the SH3. 28643 28644@item -m3e 28645@opindex m3e 28646Generate code for the SH3e. 28647 28648@item -m4-nofpu 28649@opindex m4-nofpu 28650Generate code for the SH4 without a floating-point unit. 28651 28652@item -m4-single-only 28653@opindex m4-single-only 28654Generate code for the SH4 with a floating-point unit that only 28655supports single-precision arithmetic. 28656 28657@item -m4-single 28658@opindex m4-single 28659Generate code for the SH4 assuming the floating-point unit is in 28660single-precision mode by default. 28661 28662@item -m4 28663@opindex m4 28664Generate code for the SH4. 28665 28666@item -m4-100 28667@opindex m4-100 28668Generate code for SH4-100. 28669 28670@item -m4-100-nofpu 28671@opindex m4-100-nofpu 28672Generate code for SH4-100 in such a way that the 28673floating-point unit is not used. 28674 28675@item -m4-100-single 28676@opindex m4-100-single 28677Generate code for SH4-100 assuming the floating-point unit is in 28678single-precision mode by default. 28679 28680@item -m4-100-single-only 28681@opindex m4-100-single-only 28682Generate code for SH4-100 in such a way that no double-precision 28683floating-point operations are used. 28684 28685@item -m4-200 28686@opindex m4-200 28687Generate code for SH4-200. 28688 28689@item -m4-200-nofpu 28690@opindex m4-200-nofpu 28691Generate code for SH4-200 without in such a way that the 28692floating-point unit is not used. 28693 28694@item -m4-200-single 28695@opindex m4-200-single 28696Generate code for SH4-200 assuming the floating-point unit is in 28697single-precision mode by default. 28698 28699@item -m4-200-single-only 28700@opindex m4-200-single-only 28701Generate code for SH4-200 in such a way that no double-precision 28702floating-point operations are used. 28703 28704@item -m4-300 28705@opindex m4-300 28706Generate code for SH4-300. 28707 28708@item -m4-300-nofpu 28709@opindex m4-300-nofpu 28710Generate code for SH4-300 without in such a way that the 28711floating-point unit is not used. 28712 28713@item -m4-300-single 28714@opindex m4-300-single 28715Generate code for SH4-300 in such a way that no double-precision 28716floating-point operations are used. 28717 28718@item -m4-300-single-only 28719@opindex m4-300-single-only 28720Generate code for SH4-300 in such a way that no double-precision 28721floating-point operations are used. 28722 28723@item -m4-340 28724@opindex m4-340 28725Generate code for SH4-340 (no MMU, no FPU). 28726 28727@item -m4-500 28728@opindex m4-500 28729Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the 28730assembler. 28731 28732@item -m4a-nofpu 28733@opindex m4a-nofpu 28734Generate code for the SH4al-dsp, or for a SH4a in such a way that the 28735floating-point unit is not used. 28736 28737@item -m4a-single-only 28738@opindex m4a-single-only 28739Generate code for the SH4a, in such a way that no double-precision 28740floating-point operations are used. 28741 28742@item -m4a-single 28743@opindex m4a-single 28744Generate code for the SH4a assuming the floating-point unit is in 28745single-precision mode by default. 28746 28747@item -m4a 28748@opindex m4a 28749Generate code for the SH4a. 28750 28751@item -m4al 28752@opindex m4al 28753Same as @option{-m4a-nofpu}, except that it implicitly passes 28754@option{-dsp} to the assembler. GCC doesn't generate any DSP 28755instructions at the moment. 28756 28757@item -mb 28758@opindex mb 28759Compile code for the processor in big-endian mode. 28760 28761@item -ml 28762@opindex ml 28763Compile code for the processor in little-endian mode. 28764 28765@item -mdalign 28766@opindex mdalign 28767Align doubles at 64-bit boundaries. Note that this changes the calling 28768conventions, and thus some functions from the standard C library do 28769not work unless you recompile it first with @option{-mdalign}. 28770 28771@item -mrelax 28772@opindex mrelax 28773Shorten some address references at link time, when possible; uses the 28774linker option @option{-relax}. 28775 28776@item -mbigtable 28777@opindex mbigtable 28778Use 32-bit offsets in @code{switch} tables. The default is to use 2877916-bit offsets. 28780 28781@item -mbitops 28782@opindex mbitops 28783Enable the use of bit manipulation instructions on SH2A. 28784 28785@item -mfmovd 28786@opindex mfmovd 28787Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for 28788alignment constraints. 28789 28790@item -mrenesas 28791@opindex mrenesas 28792Comply with the calling conventions defined by Renesas. 28793 28794@item -mno-renesas 28795@opindex mno-renesas 28796Comply with the calling conventions defined for GCC before the Renesas 28797conventions were available. This option is the default for all 28798targets of the SH toolchain. 28799 28800@item -mnomacsave 28801@opindex mnomacsave 28802Mark the @code{MAC} register as call-clobbered, even if 28803@option{-mrenesas} is given. 28804 28805@item -mieee 28806@itemx -mno-ieee 28807@opindex mieee 28808@opindex mno-ieee 28809Control the IEEE compliance of floating-point comparisons, which affects the 28810handling of cases where the result of a comparison is unordered. By default 28811@option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is 28812enabled @option{-mno-ieee} is implicitly set, which results in faster 28813floating-point greater-equal and less-equal comparisons. The implicit settings 28814can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}. 28815 28816@item -minline-ic_invalidate 28817@opindex minline-ic_invalidate 28818Inline code to invalidate instruction cache entries after setting up 28819nested function trampolines. 28820This option has no effect if @option{-musermode} is in effect and the selected 28821code generation option (e.g.@: @option{-m4}) does not allow the use of the @code{icbi} 28822instruction. 28823If the selected code generation option does not allow the use of the @code{icbi} 28824instruction, and @option{-musermode} is not in effect, the inlined code 28825manipulates the instruction cache address array directly with an associative 28826write. This not only requires privileged mode at run time, but it also 28827fails if the cache line had been mapped via the TLB and has become unmapped. 28828 28829@item -misize 28830@opindex misize 28831Dump instruction size and location in the assembly code. 28832 28833@item -mpadstruct 28834@opindex mpadstruct 28835This option is deprecated. It pads structures to multiple of 4 bytes, 28836which is incompatible with the SH ABI@. 28837 28838@item -matomic-model=@var{model} 28839@opindex matomic-model=@var{model} 28840Sets the model of atomic operations and additional parameters as a comma 28841separated list. For details on the atomic built-in functions see 28842@ref{__atomic Builtins}. The following models and parameters are supported: 28843 28844@table @samp 28845 28846@item none 28847Disable compiler generated atomic sequences and emit library calls for atomic 28848operations. This is the default if the target is not @code{sh*-*-linux*}. 28849 28850@item soft-gusa 28851Generate GNU/Linux compatible gUSA software atomic sequences for the atomic 28852built-in functions. The generated atomic sequences require additional support 28853from the interrupt/exception handling code of the system and are only suitable 28854for SH3* and SH4* single-core systems. This option is enabled by default when 28855the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A, 28856this option also partially utilizes the hardware atomic instructions 28857@code{movli.l} and @code{movco.l} to create more efficient code, unless 28858@samp{strict} is specified. 28859 28860@item soft-tcb 28861Generate software atomic sequences that use a variable in the thread control 28862block. This is a variation of the gUSA sequences which can also be used on 28863SH1* and SH2* targets. The generated atomic sequences require additional 28864support from the interrupt/exception handling code of the system and are only 28865suitable for single-core systems. When using this model, the @samp{gbr-offset=} 28866parameter has to be specified as well. 28867 28868@item soft-imask 28869Generate software atomic sequences that temporarily disable interrupts by 28870setting @code{SR.IMASK = 1111}. This model works only when the program runs 28871in privileged mode and is only suitable for single-core systems. Additional 28872support from the interrupt/exception handling code of the system is not 28873required. This model is enabled by default when the target is 28874@code{sh*-*-linux*} and SH1* or SH2*. 28875 28876@item hard-llcs 28877Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l} 28878instructions only. This is only available on SH4A and is suitable for 28879multi-core systems. Since the hardware instructions support only 32 bit atomic 28880variables access to 8 or 16 bit variables is emulated with 32 bit accesses. 28881Code compiled with this option is also compatible with other software 28882atomic model interrupt/exception handling systems if executed on an SH4A 28883system. Additional support from the interrupt/exception handling code of the 28884system is not required for this model. 28885 28886@item gbr-offset= 28887This parameter specifies the offset in bytes of the variable in the thread 28888control block structure that should be used by the generated atomic sequences 28889when the @samp{soft-tcb} model has been selected. For other models this 28890parameter is ignored. The specified value must be an integer multiple of four 28891and in the range 0-1020. 28892 28893@item strict 28894This parameter prevents mixed usage of multiple atomic models, even if they 28895are compatible, and makes the compiler generate atomic sequences of the 28896specified model only. 28897 28898@end table 28899 28900@item -mtas 28901@opindex mtas 28902Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}. 28903Notice that depending on the particular hardware and software configuration 28904this can degrade overall performance due to the operand cache line flushes 28905that are implied by the @code{tas.b} instruction. On multi-core SH4A 28906processors the @code{tas.b} instruction must be used with caution since it 28907can result in data corruption for certain cache configurations. 28908 28909@item -mprefergot 28910@opindex mprefergot 28911When generating position-independent code, emit function calls using 28912the Global Offset Table instead of the Procedure Linkage Table. 28913 28914@item -musermode 28915@itemx -mno-usermode 28916@opindex musermode 28917@opindex mno-usermode 28918Don't allow (allow) the compiler generating privileged mode code. Specifying 28919@option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the 28920inlined code would not work in user mode. @option{-musermode} is the default 28921when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2* 28922@option{-musermode} has no effect, since there is no user mode. 28923 28924@item -multcost=@var{number} 28925@opindex multcost=@var{number} 28926Set the cost to assume for a multiply insn. 28927 28928@item -mdiv=@var{strategy} 28929@opindex mdiv=@var{strategy} 28930Set the division strategy to be used for integer division operations. 28931@var{strategy} can be one of: 28932 28933@table @samp 28934 28935@item call-div1 28936Calls a library function that uses the single-step division instruction 28937@code{div1} to perform the operation. Division by zero calculates an 28938unspecified result and does not trap. This is the default except for SH4, 28939SH2A and SHcompact. 28940 28941@item call-fp 28942Calls a library function that performs the operation in double precision 28943floating point. Division by zero causes a floating-point exception. This is 28944the default for SHcompact with FPU. Specifying this for targets that do not 28945have a double precision FPU defaults to @code{call-div1}. 28946 28947@item call-table 28948Calls a library function that uses a lookup table for small divisors and 28949the @code{div1} instruction with case distinction for larger divisors. Division 28950by zero calculates an unspecified result and does not trap. This is the default 28951for SH4. Specifying this for targets that do not have dynamic shift 28952instructions defaults to @code{call-div1}. 28953 28954@end table 28955 28956When a division strategy has not been specified the default strategy is 28957selected based on the current target. For SH2A the default strategy is to 28958use the @code{divs} and @code{divu} instructions instead of library function 28959calls. 28960 28961@item -maccumulate-outgoing-args 28962@opindex maccumulate-outgoing-args 28963Reserve space once for outgoing arguments in the function prologue rather 28964than around each call. Generally beneficial for performance and size. Also 28965needed for unwinding to avoid changing the stack frame around conditional code. 28966 28967@item -mdivsi3_libfunc=@var{name} 28968@opindex mdivsi3_libfunc=@var{name} 28969Set the name of the library function used for 32-bit signed division to 28970@var{name}. 28971This only affects the name used in the @samp{call} division strategies, and 28972the compiler still expects the same sets of input/output/clobbered registers as 28973if this option were not present. 28974 28975@item -mfixed-range=@var{register-range} 28976@opindex mfixed-range 28977Generate code treating the given register range as fixed registers. 28978A fixed register is one that the register allocator cannot use. This is 28979useful when compiling kernel code. A register range is specified as 28980two registers separated by a dash. Multiple register ranges can be 28981specified separated by a comma. 28982 28983@item -mbranch-cost=@var{num} 28984@opindex mbranch-cost=@var{num} 28985Assume @var{num} to be the cost for a branch instruction. Higher numbers 28986make the compiler try to generate more branch-free code if possible. 28987If not specified the value is selected depending on the processor type that 28988is being compiled for. 28989 28990@item -mzdcbranch 28991@itemx -mno-zdcbranch 28992@opindex mzdcbranch 28993@opindex mno-zdcbranch 28994Assume (do not assume) that zero displacement conditional branch instructions 28995@code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the 28996compiler prefers zero displacement branch code sequences. This is 28997enabled by default when generating code for SH4 and SH4A. It can be explicitly 28998disabled by specifying @option{-mno-zdcbranch}. 28999 29000@item -mcbranch-force-delay-slot 29001@opindex mcbranch-force-delay-slot 29002Force the usage of delay slots for conditional branches, which stuffs the delay 29003slot with a @code{nop} if a suitable instruction cannot be found. By default 29004this option is disabled. It can be enabled to work around hardware bugs as 29005found in the original SH7055. 29006 29007@item -mfused-madd 29008@itemx -mno-fused-madd 29009@opindex mfused-madd 29010@opindex mno-fused-madd 29011Generate code that uses (does not use) the floating-point multiply and 29012accumulate instructions. These instructions are generated by default 29013if hardware floating point is used. The machine-dependent 29014@option{-mfused-madd} option is now mapped to the machine-independent 29015@option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is 29016mapped to @option{-ffp-contract=off}. 29017 29018@item -mfsca 29019@itemx -mno-fsca 29020@opindex mfsca 29021@opindex mno-fsca 29022Allow or disallow the compiler to emit the @code{fsca} instruction for sine 29023and cosine approximations. The option @option{-mfsca} must be used in 29024combination with @option{-funsafe-math-optimizations}. It is enabled by default 29025when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine 29026approximations even if @option{-funsafe-math-optimizations} is in effect. 29027 29028@item -mfsrra 29029@itemx -mno-fsrra 29030@opindex mfsrra 29031@opindex mno-fsrra 29032Allow or disallow the compiler to emit the @code{fsrra} instruction for 29033reciprocal square root approximations. The option @option{-mfsrra} must be used 29034in combination with @option{-funsafe-math-optimizations} and 29035@option{-ffinite-math-only}. It is enabled by default when generating code for 29036SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations 29037even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are 29038in effect. 29039 29040@item -mpretend-cmove 29041@opindex mpretend-cmove 29042Prefer zero-displacement conditional branches for conditional move instruction 29043patterns. This can result in faster code on the SH4 processor. 29044 29045@item -mfdpic 29046@opindex fdpic 29047Generate code using the FDPIC ABI. 29048 29049@end table 29050 29051@node Solaris 2 Options 29052@subsection Solaris 2 Options 29053@cindex Solaris 2 options 29054 29055These @samp{-m} options are supported on Solaris 2: 29056 29057@table @gcctabopt 29058@item -mclear-hwcap 29059@opindex mclear-hwcap 29060@option{-mclear-hwcap} tells the compiler to remove the hardware 29061capabilities generated by the Solaris assembler. This is only necessary 29062when object files use ISA extensions not supported by the current 29063machine, but check at runtime whether or not to use them. 29064 29065@item -mimpure-text 29066@opindex mimpure-text 29067@option{-mimpure-text}, used in addition to @option{-shared}, tells 29068the compiler to not pass @option{-z text} to the linker when linking a 29069shared object. Using this option, you can link position-dependent 29070code into a shared object. 29071 29072@option{-mimpure-text} suppresses the ``relocations remain against 29073allocatable but non-writable sections'' linker error message. 29074However, the necessary relocations trigger copy-on-write, and the 29075shared object is not actually shared across processes. Instead of 29076using @option{-mimpure-text}, you should compile all source code with 29077@option{-fpic} or @option{-fPIC}. 29078 29079@end table 29080 29081These switches are supported in addition to the above on Solaris 2: 29082 29083@table @gcctabopt 29084@item -pthreads 29085@opindex pthreads 29086This is a synonym for @option{-pthread}. 29087@end table 29088 29089@node SPARC Options 29090@subsection SPARC Options 29091@cindex SPARC options 29092 29093These @samp{-m} options are supported on the SPARC: 29094 29095@table @gcctabopt 29096@item -mno-app-regs 29097@itemx -mapp-regs 29098@opindex mno-app-regs 29099@opindex mapp-regs 29100Specify @option{-mapp-regs} to generate output using the global registers 291012 through 4, which the SPARC SVR4 ABI reserves for applications. Like the 29102global register 1, each global register 2 through 4 is then treated as an 29103allocable register that is clobbered by function calls. This is the default. 29104 29105To be fully SVR4 ABI-compliant at the cost of some performance loss, 29106specify @option{-mno-app-regs}. You should compile libraries and system 29107software with this option. 29108 29109@item -mflat 29110@itemx -mno-flat 29111@opindex mflat 29112@opindex mno-flat 29113With @option{-mflat}, the compiler does not generate save/restore instructions 29114and uses a ``flat'' or single register window model. This model is compatible 29115with the regular register window model. The local registers and the input 29116registers (0--5) are still treated as ``call-saved'' registers and are 29117saved on the stack as needed. 29118 29119With @option{-mno-flat} (the default), the compiler generates save/restore 29120instructions (except for leaf functions). This is the normal operating mode. 29121 29122@item -mfpu 29123@itemx -mhard-float 29124@opindex mfpu 29125@opindex mhard-float 29126Generate output containing floating-point instructions. This is the 29127default. 29128 29129@item -mno-fpu 29130@itemx -msoft-float 29131@opindex mno-fpu 29132@opindex msoft-float 29133Generate output containing library calls for floating point. 29134@strong{Warning:} the requisite libraries are not available for all SPARC 29135targets. Normally the facilities of the machine's usual C compiler are 29136used, but this cannot be done directly in cross-compilation. You must make 29137your own arrangements to provide suitable library functions for 29138cross-compilation. The embedded targets @samp{sparc-*-aout} and 29139@samp{sparclite-*-*} do provide software floating-point support. 29140 29141@option{-msoft-float} changes the calling convention in the output file; 29142therefore, it is only useful if you compile @emph{all} of a program with 29143this option. In particular, you need to compile @file{libgcc.a}, the 29144library that comes with GCC, with @option{-msoft-float} in order for 29145this to work. 29146 29147@item -mhard-quad-float 29148@opindex mhard-quad-float 29149Generate output containing quad-word (long double) floating-point 29150instructions. 29151 29152@item -msoft-quad-float 29153@opindex msoft-quad-float 29154Generate output containing library calls for quad-word (long double) 29155floating-point instructions. The functions called are those specified 29156in the SPARC ABI@. This is the default. 29157 29158As of this writing, there are no SPARC implementations that have hardware 29159support for the quad-word floating-point instructions. They all invoke 29160a trap handler for one of these instructions, and then the trap handler 29161emulates the effect of the instruction. Because of the trap handler overhead, 29162this is much slower than calling the ABI library routines. Thus the 29163@option{-msoft-quad-float} option is the default. 29164 29165@item -mno-unaligned-doubles 29166@itemx -munaligned-doubles 29167@opindex mno-unaligned-doubles 29168@opindex munaligned-doubles 29169Assume that doubles have 8-byte alignment. This is the default. 29170 29171With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte 29172alignment only if they are contained in another type, or if they have an 29173absolute address. Otherwise, it assumes they have 4-byte alignment. 29174Specifying this option avoids some rare compatibility problems with code 29175generated by other compilers. It is not the default because it results 29176in a performance loss, especially for floating-point code. 29177 29178@item -muser-mode 29179@itemx -mno-user-mode 29180@opindex muser-mode 29181@opindex mno-user-mode 29182Do not generate code that can only run in supervisor mode. This is relevant 29183only for the @code{casa} instruction emitted for the LEON3 processor. This 29184is the default. 29185 29186@item -mfaster-structs 29187@itemx -mno-faster-structs 29188@opindex mfaster-structs 29189@opindex mno-faster-structs 29190With @option{-mfaster-structs}, the compiler assumes that structures 29191should have 8-byte alignment. This enables the use of pairs of 29192@code{ldd} and @code{std} instructions for copies in structure 29193assignment, in place of twice as many @code{ld} and @code{st} pairs. 29194However, the use of this changed alignment directly violates the SPARC 29195ABI@. Thus, it's intended only for use on targets where the developer 29196acknowledges that their resulting code is not directly in line with 29197the rules of the ABI@. 29198 29199@item -mstd-struct-return 29200@itemx -mno-std-struct-return 29201@opindex mstd-struct-return 29202@opindex mno-std-struct-return 29203With @option{-mstd-struct-return}, the compiler generates checking code 29204in functions returning structures or unions to detect size mismatches 29205between the two sides of function calls, as per the 32-bit ABI@. 29206 29207The default is @option{-mno-std-struct-return}. This option has no effect 29208in 64-bit mode. 29209 29210@item -mlra 29211@itemx -mno-lra 29212@opindex mlra 29213@opindex mno-lra 29214Enable Local Register Allocation. This is the default for SPARC since GCC 7 29215so @option{-mno-lra} needs to be passed to get old Reload. 29216 29217@item -mcpu=@var{cpu_type} 29218@opindex mcpu 29219Set the instruction set, register set, and instruction scheduling parameters 29220for machine type @var{cpu_type}. Supported values for @var{cpu_type} are 29221@samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc}, 29222@samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{leon5}, @samp{sparclite}, 29223@samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, 29224@samp{v9}, @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, 29225@samp{niagara2}, @samp{niagara3}, @samp{niagara4}, @samp{niagara7} and 29226@samp{m8}. 29227 29228Native Solaris and GNU/Linux toolchains also support the value @samp{native}, 29229which selects the best architecture option for the host processor. 29230@option{-mcpu=native} has no effect if GCC does not recognize 29231the processor. 29232 29233Default instruction scheduling parameters are used for values that select 29234an architecture and not an implementation. These are @samp{v7}, @samp{v8}, 29235@samp{sparclite}, @samp{sparclet}, @samp{v9}. 29236 29237Here is a list of each supported architecture and their supported 29238implementations. 29239 29240@table @asis 29241@item v7 29242cypress, leon3v7 29243 29244@item v8 29245supersparc, hypersparc, leon, leon3, leon5 29246 29247@item sparclite 29248f930, f934, sparclite86x 29249 29250@item sparclet 29251tsc701 29252 29253@item v9 29254ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4, 29255niagara7, m8 29256@end table 29257 29258By default (unless configured otherwise), GCC generates code for the V7 29259variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler 29260additionally optimizes it for the Cypress CY7C602 chip, as used in the 29261SPARCStation/SPARCServer 3xx series. This is also appropriate for the older 29262SPARCStation 1, 2, IPX etc. 29263 29264With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC 29265architecture. The only difference from V7 code is that the compiler emits 29266the integer multiply and integer divide instructions which exist in SPARC-V8 29267but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally 29268optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and 292692000 series. 29270 29271With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of 29272the SPARC architecture. This adds the integer multiply, integer divide step 29273and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7. 29274With @option{-mcpu=f930}, the compiler additionally optimizes it for the 29275Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With 29276@option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu 29277MB86934 chip, which is the more recent SPARClite with FPU@. 29278 29279With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of 29280the SPARC architecture. This adds the integer multiply, multiply/accumulate, 29281integer divide step and scan (@code{ffs}) instructions which exist in SPARClet 29282but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally 29283optimizes it for the TEMIC SPARClet chip. 29284 29285With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC 29286architecture. This adds 64-bit integer and floating-point move instructions, 292873 additional floating-point condition code registers and conditional move 29288instructions. With @option{-mcpu=ultrasparc}, the compiler additionally 29289optimizes it for the Sun UltraSPARC I/II/IIi chips. With 29290@option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the 29291Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With 29292@option{-mcpu=niagara}, the compiler additionally optimizes it for 29293Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler 29294additionally optimizes it for Sun UltraSPARC T2 chips. With 29295@option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun 29296UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler 29297additionally optimizes it for Sun UltraSPARC T4 chips. With 29298@option{-mcpu=niagara7}, the compiler additionally optimizes it for 29299Oracle SPARC M7 chips. With @option{-mcpu=m8}, the compiler 29300additionally optimizes it for Oracle M8 chips. 29301 29302@item -mtune=@var{cpu_type} 29303@opindex mtune 29304Set the instruction scheduling parameters for machine type 29305@var{cpu_type}, but do not set the instruction set or register set that the 29306option @option{-mcpu=@var{cpu_type}} does. 29307 29308The same values for @option{-mcpu=@var{cpu_type}} can be used for 29309@option{-mtune=@var{cpu_type}}, but the only useful values are those 29310that select a particular CPU implementation. Those are 29311@samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon}, 29312@samp{leon3}, @samp{leon3v7}, @samp{leon5}, @samp{f930}, @samp{f934}, 29313@samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, 29314@samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3}, 29315@samp{niagara4}, @samp{niagara7} and @samp{m8}. With native Solaris 29316and GNU/Linux toolchains, @samp{native} can also be used. 29317 29318@item -mv8plus 29319@itemx -mno-v8plus 29320@opindex mv8plus 29321@opindex mno-v8plus 29322With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The 29323difference from the V8 ABI is that the global and out registers are 29324considered 64 bits wide. This is enabled by default on Solaris in 32-bit 29325mode for all SPARC-V9 processors. 29326 29327@item -mvis 29328@itemx -mno-vis 29329@opindex mvis 29330@opindex mno-vis 29331With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC 29332Visual Instruction Set extensions. The default is @option{-mno-vis}. 29333 29334@item -mvis2 29335@itemx -mno-vis2 29336@opindex mvis2 29337@opindex mno-vis2 29338With @option{-mvis2}, GCC generates code that takes advantage of 29339version 2.0 of the UltraSPARC Visual Instruction Set extensions. The 29340default is @option{-mvis2} when targeting a cpu that supports such 29341instructions, such as UltraSPARC-III and later. Setting @option{-mvis2} 29342also sets @option{-mvis}. 29343 29344@item -mvis3 29345@itemx -mno-vis3 29346@opindex mvis3 29347@opindex mno-vis3 29348With @option{-mvis3}, GCC generates code that takes advantage of 29349version 3.0 of the UltraSPARC Visual Instruction Set extensions. The 29350default is @option{-mvis3} when targeting a cpu that supports such 29351instructions, such as niagara-3 and later. Setting @option{-mvis3} 29352also sets @option{-mvis2} and @option{-mvis}. 29353 29354@item -mvis4 29355@itemx -mno-vis4 29356@opindex mvis4 29357@opindex mno-vis4 29358With @option{-mvis4}, GCC generates code that takes advantage of 29359version 4.0 of the UltraSPARC Visual Instruction Set extensions. The 29360default is @option{-mvis4} when targeting a cpu that supports such 29361instructions, such as niagara-7 and later. Setting @option{-mvis4} 29362also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}. 29363 29364@item -mvis4b 29365@itemx -mno-vis4b 29366@opindex mvis4b 29367@opindex mno-vis4b 29368With @option{-mvis4b}, GCC generates code that takes advantage of 29369version 4.0 of the UltraSPARC Visual Instruction Set extensions, plus 29370the additional VIS instructions introduced in the Oracle SPARC 29371Architecture 2017. The default is @option{-mvis4b} when targeting a 29372cpu that supports such instructions, such as m8 and later. Setting 29373@option{-mvis4b} also sets @option{-mvis4}, @option{-mvis3}, 29374@option{-mvis2} and @option{-mvis}. 29375 29376@item -mcbcond 29377@itemx -mno-cbcond 29378@opindex mcbcond 29379@opindex mno-cbcond 29380With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC 29381Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond} 29382when targeting a CPU that supports such instructions, such as Niagara-4 and 29383later. 29384 29385@item -mfmaf 29386@itemx -mno-fmaf 29387@opindex mfmaf 29388@opindex mno-fmaf 29389With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC 29390Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf} 29391when targeting a CPU that supports such instructions, such as Niagara-3 and 29392later. 29393 29394@item -mfsmuld 29395@itemx -mno-fsmuld 29396@opindex mfsmuld 29397@opindex mno-fsmuld 29398With @option{-mfsmuld}, GCC generates code that takes advantage of the 29399Floating-point Multiply Single to Double (FsMULd) instruction. The default is 29400@option{-mfsmuld} when targeting a CPU supporting the architecture versions V8 29401or V9 with FPU except @option{-mcpu=leon}. 29402 29403@item -mpopc 29404@itemx -mno-popc 29405@opindex mpopc 29406@opindex mno-popc 29407With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC 29408Population Count instruction. The default is @option{-mpopc} 29409when targeting a CPU that supports such an instruction, such as Niagara-2 and 29410later. 29411 29412@item -msubxc 29413@itemx -mno-subxc 29414@opindex msubxc 29415@opindex mno-subxc 29416With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC 29417Subtract-Extended-with-Carry instruction. The default is @option{-msubxc} 29418when targeting a CPU that supports such an instruction, such as Niagara-7 and 29419later. 29420 29421@item -mfix-at697f 29422@opindex mfix-at697f 29423Enable the documented workaround for the single erratum of the Atmel AT697F 29424processor (which corresponds to erratum #13 of the AT697E processor). 29425 29426@item -mfix-ut699 29427@opindex mfix-ut699 29428Enable the documented workarounds for the floating-point errata and the data 29429cache nullify errata of the UT699 processor. 29430 29431@item -mfix-ut700 29432@opindex mfix-ut700 29433Enable the documented workaround for the back-to-back store errata of 29434the UT699E/UT700 processor. 29435 29436@item -mfix-gr712rc 29437@opindex mfix-gr712rc 29438Enable the documented workaround for the back-to-back store errata of 29439the GR712RC processor. 29440@end table 29441 29442These @samp{-m} options are supported in addition to the above 29443on SPARC-V9 processors in 64-bit environments: 29444 29445@table @gcctabopt 29446@item -m32 29447@itemx -m64 29448@opindex m32 29449@opindex m64 29450Generate code for a 32-bit or 64-bit environment. 29451The 32-bit environment sets int, long and pointer to 32 bits. 29452The 64-bit environment sets int to 32 bits and long and pointer 29453to 64 bits. 29454 29455@item -mcmodel=@var{which} 29456@opindex mcmodel 29457Set the code model to one of 29458 29459@table @samp 29460@item medlow 29461The Medium/Low code model: 64-bit addresses, programs 29462must be linked in the low 32 bits of memory. Programs can be statically 29463or dynamically linked. 29464 29465@item medmid 29466The Medium/Middle code model: 64-bit addresses, programs 29467must be linked in the low 44 bits of memory, the text and data segments must 29468be less than 2GB in size and the data segment must be located within 2GB of 29469the text segment. 29470 29471@item medany 29472The Medium/Anywhere code model: 64-bit addresses, programs 29473may be linked anywhere in memory, the text and data segments must be less 29474than 2GB in size and the data segment must be located within 2GB of the 29475text segment. 29476 29477@item embmedany 29478The Medium/Anywhere code model for embedded systems: 2947964-bit addresses, the text and data segments must be less than 2GB in 29480size, both starting anywhere in memory (determined at link time). The 29481global register %g4 points to the base of the data segment. Programs 29482are statically linked and PIC is not supported. 29483@end table 29484 29485@item -mmemory-model=@var{mem-model} 29486@opindex mmemory-model 29487Set the memory model in force on the processor to one of 29488 29489@table @samp 29490@item default 29491The default memory model for the processor and operating system. 29492 29493@item rmo 29494Relaxed Memory Order 29495 29496@item pso 29497Partial Store Order 29498 29499@item tso 29500Total Store Order 29501 29502@item sc 29503Sequential Consistency 29504@end table 29505 29506These memory models are formally defined in Appendix D of the SPARC-V9 29507architecture manual, as set in the processor's @code{PSTATE.MM} field. 29508 29509@item -mstack-bias 29510@itemx -mno-stack-bias 29511@opindex mstack-bias 29512@opindex mno-stack-bias 29513With @option{-mstack-bias}, GCC assumes that the stack pointer, and 29514frame pointer if present, are offset by @minus{}2047 which must be added back 29515when making stack frame references. This is the default in 64-bit mode. 29516Otherwise, assume no such offset is present. 29517@end table 29518 29519@node System V Options 29520@subsection Options for System V 29521 29522These additional options are available on System V Release 4 for 29523compatibility with other compilers on those systems: 29524 29525@table @gcctabopt 29526@item -G 29527@opindex G 29528Create a shared object. 29529It is recommended that @option{-symbolic} or @option{-shared} be used instead. 29530 29531@item -Qy 29532@opindex Qy 29533Identify the versions of each tool used by the compiler, in a 29534@code{.ident} assembler directive in the output. 29535 29536@item -Qn 29537@opindex Qn 29538Refrain from adding @code{.ident} directives to the output file (this is 29539the default). 29540 29541@item -YP,@var{dirs} 29542@opindex YP 29543Search the directories @var{dirs}, and no others, for libraries 29544specified with @option{-l}. 29545 29546@item -Ym,@var{dir} 29547@opindex Ym 29548Look in the directory @var{dir} to find the M4 preprocessor. 29549The assembler uses this option. 29550@c This is supposed to go with a -Yd for predefined M4 macro files, but 29551@c the generic assembler that comes with Solaris takes just -Ym. 29552@end table 29553 29554@node TILE-Gx Options 29555@subsection TILE-Gx Options 29556@cindex TILE-Gx options 29557 29558These @samp{-m} options are supported on the TILE-Gx: 29559 29560@table @gcctabopt 29561@item -mcmodel=small 29562@opindex mcmodel=small 29563Generate code for the small model. The distance for direct calls is 29564limited to 500M in either direction. PC-relative addresses are 32 29565bits. Absolute addresses support the full address range. 29566 29567@item -mcmodel=large 29568@opindex mcmodel=large 29569Generate code for the large model. There is no limitation on call 29570distance, pc-relative addresses, or absolute addresses. 29571 29572@item -mcpu=@var{name} 29573@opindex mcpu 29574Selects the type of CPU to be targeted. Currently the only supported 29575type is @samp{tilegx}. 29576 29577@item -m32 29578@itemx -m64 29579@opindex m32 29580@opindex m64 29581Generate code for a 32-bit or 64-bit environment. The 32-bit 29582environment sets int, long, and pointer to 32 bits. The 64-bit 29583environment sets int to 32 bits and long and pointer to 64 bits. 29584 29585@item -mbig-endian 29586@itemx -mlittle-endian 29587@opindex mbig-endian 29588@opindex mlittle-endian 29589Generate code in big/little endian mode, respectively. 29590@end table 29591 29592@node TILEPro Options 29593@subsection TILEPro Options 29594@cindex TILEPro options 29595 29596These @samp{-m} options are supported on the TILEPro: 29597 29598@table @gcctabopt 29599@item -mcpu=@var{name} 29600@opindex mcpu 29601Selects the type of CPU to be targeted. Currently the only supported 29602type is @samp{tilepro}. 29603 29604@item -m32 29605@opindex m32 29606Generate code for a 32-bit environment, which sets int, long, and 29607pointer to 32 bits. This is the only supported behavior so the flag 29608is essentially ignored. 29609@end table 29610 29611@node V850 Options 29612@subsection V850 Options 29613@cindex V850 Options 29614 29615These @samp{-m} options are defined for V850 implementations: 29616 29617@table @gcctabopt 29618@item -mlong-calls 29619@itemx -mno-long-calls 29620@opindex mlong-calls 29621@opindex mno-long-calls 29622Treat all calls as being far away (near). If calls are assumed to be 29623far away, the compiler always loads the function's address into a 29624register, and calls indirect through the pointer. 29625 29626@item -mno-ep 29627@itemx -mep 29628@opindex mno-ep 29629@opindex mep 29630Do not optimize (do optimize) basic blocks that use the same index 29631pointer 4 or more times to copy pointer into the @code{ep} register, and 29632use the shorter @code{sld} and @code{sst} instructions. The @option{-mep} 29633option is on by default if you optimize. 29634 29635@item -mno-prolog-function 29636@itemx -mprolog-function 29637@opindex mno-prolog-function 29638@opindex mprolog-function 29639Do not use (do use) external functions to save and restore registers 29640at the prologue and epilogue of a function. The external functions 29641are slower, but use less code space if more than one function saves 29642the same number of registers. The @option{-mprolog-function} option 29643is on by default if you optimize. 29644 29645@item -mspace 29646@opindex mspace 29647Try to make the code as small as possible. At present, this just turns 29648on the @option{-mep} and @option{-mprolog-function} options. 29649 29650@item -mtda=@var{n} 29651@opindex mtda 29652Put static or global variables whose size is @var{n} bytes or less into 29653the tiny data area that register @code{ep} points to. The tiny data 29654area can hold up to 256 bytes in total (128 bytes for byte references). 29655 29656@item -msda=@var{n} 29657@opindex msda 29658Put static or global variables whose size is @var{n} bytes or less into 29659the small data area that register @code{gp} points to. The small data 29660area can hold up to 64 kilobytes. 29661 29662@item -mzda=@var{n} 29663@opindex mzda 29664Put static or global variables whose size is @var{n} bytes or less into 29665the first 32 kilobytes of memory. 29666 29667@item -mv850 29668@opindex mv850 29669Specify that the target processor is the V850. 29670 29671@item -mv850e3v5 29672@opindex mv850e3v5 29673Specify that the target processor is the V850E3V5. The preprocessor 29674constant @code{__v850e3v5__} is defined if this option is used. 29675 29676@item -mv850e2v4 29677@opindex mv850e2v4 29678Specify that the target processor is the V850E3V5. This is an alias for 29679the @option{-mv850e3v5} option. 29680 29681@item -mv850e2v3 29682@opindex mv850e2v3 29683Specify that the target processor is the V850E2V3. The preprocessor 29684constant @code{__v850e2v3__} is defined if this option is used. 29685 29686@item -mv850e2 29687@opindex mv850e2 29688Specify that the target processor is the V850E2. The preprocessor 29689constant @code{__v850e2__} is defined if this option is used. 29690 29691@item -mv850e1 29692@opindex mv850e1 29693Specify that the target processor is the V850E1. The preprocessor 29694constants @code{__v850e1__} and @code{__v850e__} are defined if 29695this option is used. 29696 29697@item -mv850es 29698@opindex mv850es 29699Specify that the target processor is the V850ES. This is an alias for 29700the @option{-mv850e1} option. 29701 29702@item -mv850e 29703@opindex mv850e 29704Specify that the target processor is the V850E@. The preprocessor 29705constant @code{__v850e__} is defined if this option is used. 29706 29707If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1} 29708nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5} 29709are defined then a default target processor is chosen and the 29710relevant @samp{__v850*__} preprocessor constant is defined. 29711 29712The preprocessor constants @code{__v850} and @code{__v851__} are always 29713defined, regardless of which processor variant is the target. 29714 29715@item -mdisable-callt 29716@itemx -mno-disable-callt 29717@opindex mdisable-callt 29718@opindex mno-disable-callt 29719This option suppresses generation of the @code{CALLT} instruction for the 29720v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850 29721architecture. 29722 29723This option is enabled by default when the RH850 ABI is 29724in use (see @option{-mrh850-abi}), and disabled by default when the 29725GCC ABI is in use. If @code{CALLT} instructions are being generated 29726then the C preprocessor symbol @code{__V850_CALLT__} is defined. 29727 29728@item -mrelax 29729@itemx -mno-relax 29730@opindex mrelax 29731@opindex mno-relax 29732Pass on (or do not pass on) the @option{-mrelax} command-line option 29733to the assembler. 29734 29735@item -mlong-jumps 29736@itemx -mno-long-jumps 29737@opindex mlong-jumps 29738@opindex mno-long-jumps 29739Disable (or re-enable) the generation of PC-relative jump instructions. 29740 29741@item -msoft-float 29742@itemx -mhard-float 29743@opindex msoft-float 29744@opindex mhard-float 29745Disable (or re-enable) the generation of hardware floating point 29746instructions. This option is only significant when the target 29747architecture is @samp{V850E2V3} or higher. If hardware floating point 29748instructions are being generated then the C preprocessor symbol 29749@code{__FPU_OK__} is defined, otherwise the symbol 29750@code{__NO_FPU__} is defined. 29751 29752@item -mloop 29753@opindex mloop 29754Enables the use of the e3v5 LOOP instruction. The use of this 29755instruction is not enabled by default when the e3v5 architecture is 29756selected because its use is still experimental. 29757 29758@item -mrh850-abi 29759@itemx -mghs 29760@opindex mrh850-abi 29761@opindex mghs 29762Enables support for the RH850 version of the V850 ABI. This is the 29763default. With this version of the ABI the following rules apply: 29764 29765@itemize 29766@item 29767Integer sized structures and unions are returned via a memory pointer 29768rather than a register. 29769 29770@item 29771Large structures and unions (more than 8 bytes in size) are passed by 29772value. 29773 29774@item 29775Functions are aligned to 16-bit boundaries. 29776 29777@item 29778The @option{-m8byte-align} command-line option is supported. 29779 29780@item 29781The @option{-mdisable-callt} command-line option is enabled by 29782default. The @option{-mno-disable-callt} command-line option is not 29783supported. 29784@end itemize 29785 29786When this version of the ABI is enabled the C preprocessor symbol 29787@code{__V850_RH850_ABI__} is defined. 29788 29789@item -mgcc-abi 29790@opindex mgcc-abi 29791Enables support for the old GCC version of the V850 ABI. With this 29792version of the ABI the following rules apply: 29793 29794@itemize 29795@item 29796Integer sized structures and unions are returned in register @code{r10}. 29797 29798@item 29799Large structures and unions (more than 8 bytes in size) are passed by 29800reference. 29801 29802@item 29803Functions are aligned to 32-bit boundaries, unless optimizing for 29804size. 29805 29806@item 29807The @option{-m8byte-align} command-line option is not supported. 29808 29809@item 29810The @option{-mdisable-callt} command-line option is supported but not 29811enabled by default. 29812@end itemize 29813 29814When this version of the ABI is enabled the C preprocessor symbol 29815@code{__V850_GCC_ABI__} is defined. 29816 29817@item -m8byte-align 29818@itemx -mno-8byte-align 29819@opindex m8byte-align 29820@opindex mno-8byte-align 29821Enables support for @code{double} and @code{long long} types to be 29822aligned on 8-byte boundaries. The default is to restrict the 29823alignment of all objects to at most 4-bytes. When 29824@option{-m8byte-align} is in effect the C preprocessor symbol 29825@code{__V850_8BYTE_ALIGN__} is defined. 29826 29827@item -mbig-switch 29828@opindex mbig-switch 29829Generate code suitable for big switch tables. Use this option only if 29830the assembler/linker complain about out of range branches within a switch 29831table. 29832 29833@item -mapp-regs 29834@opindex mapp-regs 29835This option causes r2 and r5 to be used in the code generated by 29836the compiler. This setting is the default. 29837 29838@item -mno-app-regs 29839@opindex mno-app-regs 29840This option causes r2 and r5 to be treated as fixed registers. 29841 29842@end table 29843 29844@node VAX Options 29845@subsection VAX Options 29846@cindex VAX options 29847 29848These @samp{-m} options are defined for the VAX: 29849 29850@table @gcctabopt 29851@item -munix 29852@opindex munix 29853Do not output certain jump instructions (@code{aobleq} and so on) 29854that the Unix assembler for the VAX cannot handle across long 29855ranges. 29856 29857@item -mgnu 29858@opindex mgnu 29859Do output those jump instructions, on the assumption that the 29860GNU assembler is being used. 29861 29862@item -mg 29863@opindex mg 29864Output code for G-format floating-point numbers instead of D-format. 29865@end table 29866 29867@node Visium Options 29868@subsection Visium Options 29869@cindex Visium options 29870 29871@table @gcctabopt 29872 29873@item -mdebug 29874@opindex mdebug 29875A program which performs file I/O and is destined to run on an MCM target 29876should be linked with this option. It causes the libraries libc.a and 29877libdebug.a to be linked. The program should be run on the target under 29878the control of the GDB remote debugging stub. 29879 29880@item -msim 29881@opindex msim 29882A program which performs file I/O and is destined to run on the simulator 29883should be linked with option. This causes libraries libc.a and libsim.a to 29884be linked. 29885 29886@item -mfpu 29887@itemx -mhard-float 29888@opindex mfpu 29889@opindex mhard-float 29890Generate code containing floating-point instructions. This is the 29891default. 29892 29893@item -mno-fpu 29894@itemx -msoft-float 29895@opindex mno-fpu 29896@opindex msoft-float 29897Generate code containing library calls for floating-point. 29898 29899@option{-msoft-float} changes the calling convention in the output file; 29900therefore, it is only useful if you compile @emph{all} of a program with 29901this option. In particular, you need to compile @file{libgcc.a}, the 29902library that comes with GCC, with @option{-msoft-float} in order for 29903this to work. 29904 29905@item -mcpu=@var{cpu_type} 29906@opindex mcpu 29907Set the instruction set, register set, and instruction scheduling parameters 29908for machine type @var{cpu_type}. Supported values for @var{cpu_type} are 29909@samp{mcm}, @samp{gr5} and @samp{gr6}. 29910 29911@samp{mcm} is a synonym of @samp{gr5} present for backward compatibility. 29912 29913By default (unless configured otherwise), GCC generates code for the GR5 29914variant of the Visium architecture. 29915 29916With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium 29917architecture. The only difference from GR5 code is that the compiler will 29918generate block move instructions. 29919 29920@item -mtune=@var{cpu_type} 29921@opindex mtune 29922Set the instruction scheduling parameters for machine type @var{cpu_type}, 29923but do not set the instruction set or register set that the option 29924@option{-mcpu=@var{cpu_type}} would. 29925 29926@item -msv-mode 29927@opindex msv-mode 29928Generate code for the supervisor mode, where there are no restrictions on 29929the access to general registers. This is the default. 29930 29931@item -muser-mode 29932@opindex muser-mode 29933Generate code for the user mode, where the access to some general registers 29934is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this 29935mode; on the GR6, only registers r29 to r31 are affected. 29936@end table 29937 29938@node VMS Options 29939@subsection VMS Options 29940 29941These @samp{-m} options are defined for the VMS implementations: 29942 29943@table @gcctabopt 29944@item -mvms-return-codes 29945@opindex mvms-return-codes 29946Return VMS condition codes from @code{main}. The default is to return POSIX-style 29947condition (e.g.@: error) codes. 29948 29949@item -mdebug-main=@var{prefix} 29950@opindex mdebug-main=@var{prefix} 29951Flag the first routine whose name starts with @var{prefix} as the main 29952routine for the debugger. 29953 29954@item -mmalloc64 29955@opindex mmalloc64 29956Default to 64-bit memory allocation routines. 29957 29958@item -mpointer-size=@var{size} 29959@opindex mpointer-size=@var{size} 29960Set the default size of pointers. Possible options for @var{size} are 29961@samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long} 29962for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers. 29963The later option disables @code{pragma pointer_size}. 29964@end table 29965 29966@node VxWorks Options 29967@subsection VxWorks Options 29968@cindex VxWorks Options 29969 29970The options in this section are defined for all VxWorks targets. 29971Options specific to the target hardware are listed with the other 29972options for that target. 29973 29974@table @gcctabopt 29975@item -mrtp 29976@opindex mrtp 29977GCC can generate code for both VxWorks kernels and real time processes 29978(RTPs). This option switches from the former to the latter. It also 29979defines the preprocessor macro @code{__RTP__}. 29980 29981@item -non-static 29982@opindex non-static 29983Link an RTP executable against shared libraries rather than static 29984libraries. The options @option{-static} and @option{-shared} can 29985also be used for RTPs (@pxref{Link Options}); @option{-static} 29986is the default. 29987 29988@item -Bstatic 29989@itemx -Bdynamic 29990@opindex Bstatic 29991@opindex Bdynamic 29992These options are passed down to the linker. They are defined for 29993compatibility with Diab. 29994 29995@item -Xbind-lazy 29996@opindex Xbind-lazy 29997Enable lazy binding of function calls. This option is equivalent to 29998@option{-Wl,-z,now} and is defined for compatibility with Diab. 29999 30000@item -Xbind-now 30001@opindex Xbind-now 30002Disable lazy binding of function calls. This option is the default and 30003is defined for compatibility with Diab. 30004@end table 30005 30006@node x86 Options 30007@subsection x86 Options 30008@cindex x86 Options 30009 30010These @samp{-m} options are defined for the x86 family of computers. 30011 30012@table @gcctabopt 30013 30014@item -march=@var{cpu-type} 30015@opindex march 30016Generate instructions for the machine type @var{cpu-type}. In contrast to 30017@option{-mtune=@var{cpu-type}}, which merely tunes the generated code 30018for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC 30019to generate code that may not run at all on processors other than the one 30020indicated. Specifying @option{-march=@var{cpu-type}} implies 30021@option{-mtune=@var{cpu-type}}, except where noted otherwise. 30022 30023The choices for @var{cpu-type} are: 30024 30025@table @samp 30026@item native 30027This selects the CPU to generate code for at compilation time by determining 30028the processor type of the compiling machine. Using @option{-march=native} 30029enables all instruction subsets supported by the local machine (hence 30030the result might not run on different machines). Using @option{-mtune=native} 30031produces code optimized for the local machine under the constraints 30032of the selected instruction set. 30033 30034@item x86-64 30035A generic CPU with 64-bit extensions. 30036 30037@item x86-64-v2 30038@itemx x86-64-v3 30039@itemx x86-64-v4 30040These choices for @var{cpu-type} select the corresponding 30041micro-architecture level from the x86-64 psABI. On ABIs other than 30042the x86-64 psABI they select the same CPU features as the x86-64 psABI 30043documents for the particular micro-architecture level. 30044 30045Since these @var{cpu-type} values do not have a corresponding 30046@option{-mtune} setting, using @option{-march} with these values enables 30047generic tuning. Specific tuning can be enabled using the 30048@option{-mtune=@var{other-cpu-type}} option with an appropriate 30049@var{other-cpu-type} value. 30050 30051@item i386 30052Original Intel i386 CPU@. 30053 30054@item i486 30055Intel i486 CPU@. (No scheduling is implemented for this chip.) 30056 30057@item i586 30058@itemx pentium 30059Intel Pentium CPU with no MMX support. 30060 30061@item lakemont 30062Intel Lakemont MCU, based on Intel Pentium CPU. 30063 30064@item pentium-mmx 30065Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support. 30066 30067@item pentiumpro 30068Intel Pentium Pro CPU@. 30069 30070@item i686 30071When used with @option{-march}, the Pentium Pro 30072instruction set is used, so the code runs on all i686 family chips. 30073When used with @option{-mtune}, it has the same meaning as @samp{generic}. 30074 30075@item pentium2 30076Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set 30077support. 30078 30079@item pentium3 30080@itemx pentium3m 30081Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction 30082set support. 30083 30084@item pentium-m 30085Intel Pentium M; low-power version of Intel Pentium III CPU 30086with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks. 30087 30088@item pentium4 30089@itemx pentium4m 30090Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support. 30091 30092@item prescott 30093Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction 30094set support. 30095 30096@item nocona 30097Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE, 30098SSE2 and SSE3 instruction set support. 30099 30100@item core2 30101Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3 30102instruction set support. 30103 30104@item nehalem 30105Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, 30106SSE4.1, SSE4.2 and POPCNT instruction set support. 30107 30108@item westmere 30109Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, 30110SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support. 30111 30112@item sandybridge 30113Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, 30114SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support. 30115 30116@item ivybridge 30117Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, 30118SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C 30119instruction set support. 30120 30121@item haswell 30122Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, 30123SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, 30124BMI, BMI2 and F16C instruction set support. 30125 30126@item broadwell 30127Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, 30128SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2, 30129F16C, RDSEED ADCX and PREFETCHW instruction set support. 30130 30131@item skylake 30132Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, 30133SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, 30134BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and XSAVES 30135instruction set support. 30136 30137@item bonnell 30138Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3 30139instruction set support. 30140 30141@item silvermont 30142Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, 30143SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL and RDRND instruction set support. 30144 30145@item goldmont 30146Intel Goldmont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, 30147SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, 30148XSAVEOPT and FSGSBASE instruction set support. 30149 30150@item goldmont-plus 30151Intel Goldmont Plus CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, 30152SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, 30153XSAVES, XSAVEOPT, FSGSBASE, PTWRITE, RDPID, SGX and UMIP instruction set support. 30154 30155@item tremont 30156Intel Tremont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, 30157SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, 30158XSAVEOPT, FSGSBASE, PTWRITE, RDPID, SGX, UMIP, GFNI-SSE, CLWB, MOVDIRI, 30159MOVDIR64B, CLDEMOTE and WAITPKG instruction set support. 30160 30161@item knl 30162Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, 30163SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, 30164BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, PREFETCHWT1, AVX512F, AVX512PF, 30165AVX512ER and AVX512CD instruction set support. 30166 30167@item knm 30168Intel Knights Mill CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, 30169SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, 30170BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, PREFETCHWT1, AVX512F, AVX512PF, 30171AVX512ER, AVX512CD, AVX5124VNNIW, AVX5124FMAPS and AVX512VPOPCNTDQ instruction 30172set support. 30173 30174@item skylake-avx512 30175Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, 30176SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, 30177BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, 30178CLWB, AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support. 30179 30180@item cannonlake 30181Intel Cannonlake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, 30182SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, 30183RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, 30184XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI, 30185AVX512IFMA, SHA and UMIP instruction set support. 30186 30187@item icelake-client 30188Intel Icelake Client CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, 30189SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, 30190RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, 30191XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI, 30192AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ, 30193AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES instruction set support. 30194 30195@item icelake-server 30196Intel Icelake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, 30197SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, 30198RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, 30199XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI, 30200AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ, 30201AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES, PCONFIG and WBNOINVD instruction 30202set support. 30203 30204@item cascadelake 30205Intel Cascadelake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, 30206SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, 30207BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, CLWB, 30208AVX512VL, AVX512BW, AVX512DQ, AVX512CD and AVX512VNNI instruction set support. 30209 30210@item cooperlake 30211Intel cooperlake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, 30212SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, 30213BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, CLWB, 30214AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VNNI and AVX512BF16 instruction 30215set support. 30216 30217@item tigerlake 30218Intel Tigerlake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, 30219SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, 30220BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, 30221AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI, AVX512IFMA, SHA, CLWB, UMIP, 30222RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ, AVX512BITALG, AVX512VNNI, VPCLMULQDQ, 30223VAES, PCONFIG, WBNOINVD, MOVDIRI, MOVDIR64B, AVX512VP2INTERSECT and KEYLOCKER 30224instruction set support. 30225 30226@item sapphirerapids 30227Intel sapphirerapids CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, 30228SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, 30229FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, 30230AVX512F, CLWB, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VNNI, AVX512BF16, 30231MOVDIRI, MOVDIR64B, AVX512VP2INTERSECT, ENQCMD, CLDEMOTE, PTWRITE, WAITPKG, 30232SERIALIZE, TSXLDTRK, UINTR, AMX-BF16, AMX-TILE, AMX-INT8 and AVX-VNNI 30233instruction set support. 30234 30235@item alderlake 30236Intel Alderlake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, 30237SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, 30238XSAVEOPT, FSGSBASE, PTWRITE, RDPID, SGX, UMIP, GFNI-SSE, CLWB, MOVDIRI, 30239MOVDIR64B, CLDEMOTE, WAITPKG, ADCX, AVX, AVX2, BMI, BMI2, F16C, FMA, LZCNT, 30240PCONFIG, PKU, VAES, VPCLMULQDQ, SERIALIZE, HRESET, KL, WIDEKL and AVX-VNNI 30241instruction set support. 30242 30243@item rocketlake 30244Intel Rocketlake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, 30245SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, 30246RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, 30247XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI, 30248AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ, 30249AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES instruction set support. 30250 30251@item k6 30252AMD K6 CPU with MMX instruction set support. 30253 30254@item k6-2 30255@itemx k6-3 30256Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support. 30257 30258@item athlon 30259@itemx athlon-tbird 30260AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions 30261support. 30262 30263@item athlon-4 30264@itemx athlon-xp 30265@itemx athlon-mp 30266Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE 30267instruction set support. 30268 30269@item k8 30270@itemx opteron 30271@itemx athlon64 30272@itemx athlon-fx 30273Processors based on the AMD K8 core with x86-64 instruction set support, 30274including the AMD Opteron, Athlon 64, and Athlon 64 FX processors. 30275(This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit 30276instruction set extensions.) 30277 30278@item k8-sse3 30279@itemx opteron-sse3 30280@itemx athlon64-sse3 30281Improved versions of AMD K8 cores with SSE3 instruction set support. 30282 30283@item amdfam10 30284@itemx barcelona 30285CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This 30286supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit 30287instruction set extensions.) 30288 30289@item bdver1 30290CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This 30291supersets FMA4, AVX, XOP, LWP, AES, PCLMUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, 30292SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.) 30293 30294@item bdver2 30295AMD Family 15h core based CPUs with x86-64 instruction set support. (This 30296supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCLMUL, CX16, MMX, 30297SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set 30298extensions.) 30299 30300@item bdver3 30301AMD Family 15h core based CPUs with x86-64 instruction set support. (This 30302supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES, 30303PCLMUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 3030464-bit instruction set extensions.) 30305 30306@item bdver4 30307AMD Family 15h core based CPUs with x86-64 instruction set support. (This 30308supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP, 30309AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, 30310SSE4.2, ABM and 64-bit instruction set extensions.) 30311 30312@item znver1 30313AMD Family 17h core based CPUs with x86-64 instruction set support. (This 30314supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX, 30315SHA, CLZERO, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, 30316SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit 30317instruction set extensions.) 30318 30319@item znver2 30320AMD Family 17h core based CPUs with x86-64 instruction set support. (This 30321supersets BMI, BMI2, CLWB, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, 30322MWAITX, SHA, CLZERO, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, 30323SSSE3, SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, RDPID, 30324WBNOINVD, and 64-bit instruction set extensions.) 30325 30326@item znver3 30327AMD Family 19h core based CPUs with x86-64 instruction set support. (This 30328supersets BMI, BMI2, CLWB, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, 30329MWAITX, SHA, CLZERO, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, 30330SSSE3, SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, RDPID, 30331WBNOINVD, PKU, VPCLMULQDQ, VAES, and 64-bit instruction set extensions.) 30332 30333@item btver1 30334CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This 30335supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit 30336instruction set extensions.) 30337 30338@item btver2 30339CPUs based on AMD Family 16h cores with x86-64 instruction set support. This 30340includes MOVBE, F16C, BMI, AVX, PCLMUL, AES, SSE4.2, SSE4.1, CX16, ABM, 30341SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions. 30342 30343@item winchip-c6 30344IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction 30345set support. 30346 30347@item winchip2 30348IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@: 30349instruction set support. 30350 30351@item c3 30352VIA C3 CPU with MMX and 3DNow!@: instruction set support. 30353(No scheduling is implemented for this chip.) 30354 30355@item c3-2 30356VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support. 30357(No scheduling is implemented for this chip.) 30358 30359@item c7 30360VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support. 30361(No scheduling is implemented for this chip.) 30362 30363@item samuel-2 30364VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support. 30365(No scheduling is implemented for this chip.) 30366 30367@item nehemiah 30368VIA Eden Nehemiah CPU with MMX and SSE instruction set support. 30369(No scheduling is implemented for this chip.) 30370 30371@item esther 30372VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support. 30373(No scheduling is implemented for this chip.) 30374 30375@item eden-x2 30376VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support. 30377(No scheduling is implemented for this chip.) 30378 30379@item eden-x4 30380VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, 30381AVX and AVX2 instruction set support. 30382(No scheduling is implemented for this chip.) 30383 30384@item nano 30385Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3 30386instruction set support. 30387(No scheduling is implemented for this chip.) 30388 30389@item nano-1000 30390VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3 30391instruction set support. 30392(No scheduling is implemented for this chip.) 30393 30394@item nano-2000 30395VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3 30396instruction set support. 30397(No scheduling is implemented for this chip.) 30398 30399@item nano-3000 30400VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1 30401instruction set support. 30402(No scheduling is implemented for this chip.) 30403 30404@item nano-x2 30405VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1 30406instruction set support. 30407(No scheduling is implemented for this chip.) 30408 30409@item nano-x4 30410VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1 30411instruction set support. 30412(No scheduling is implemented for this chip.) 30413 30414@item geode 30415AMD Geode embedded processor with MMX and 3DNow!@: instruction set support. 30416@end table 30417 30418@item -mtune=@var{cpu-type} 30419@opindex mtune 30420Tune to @var{cpu-type} everything applicable about the generated code, except 30421for the ABI and the set of available instructions. 30422While picking a specific @var{cpu-type} schedules things appropriately 30423for that particular chip, the compiler does not generate any code that 30424cannot run on the default machine type unless you use a 30425@option{-march=@var{cpu-type}} option. 30426For example, if GCC is configured for i686-pc-linux-gnu 30427then @option{-mtune=pentium4} generates code that is tuned for Pentium 4 30428but still runs on i686 machines. 30429 30430The choices for @var{cpu-type} are the same as for @option{-march}. 30431In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}: 30432 30433@table @samp 30434@item generic 30435Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors. 30436If you know the CPU on which your code will run, then you should use 30437the corresponding @option{-mtune} or @option{-march} option instead of 30438@option{-mtune=generic}. But, if you do not know exactly what CPU users 30439of your application will have, then you should use this option. 30440 30441As new processors are deployed in the marketplace, the behavior of this 30442option will change. Therefore, if you upgrade to a newer version of 30443GCC, code generation controlled by this option will change to reflect 30444the processors 30445that are most common at the time that version of GCC is released. 30446 30447There is no @option{-march=generic} option because @option{-march} 30448indicates the instruction set the compiler can use, and there is no 30449generic instruction set applicable to all processors. In contrast, 30450@option{-mtune} indicates the processor (or, in this case, collection of 30451processors) for which the code is optimized. 30452 30453@item intel 30454Produce code optimized for the most current Intel processors, which are 30455Haswell and Silvermont for this version of GCC. If you know the CPU 30456on which your code will run, then you should use the corresponding 30457@option{-mtune} or @option{-march} option instead of @option{-mtune=intel}. 30458But, if you want your application performs better on both Haswell and 30459Silvermont, then you should use this option. 30460 30461As new Intel processors are deployed in the marketplace, the behavior of 30462this option will change. Therefore, if you upgrade to a newer version of 30463GCC, code generation controlled by this option will change to reflect 30464the most current Intel processors at the time that version of GCC is 30465released. 30466 30467There is no @option{-march=intel} option because @option{-march} indicates 30468the instruction set the compiler can use, and there is no common 30469instruction set applicable to all processors. In contrast, 30470@option{-mtune} indicates the processor (or, in this case, collection of 30471processors) for which the code is optimized. 30472@end table 30473 30474@item -mcpu=@var{cpu-type} 30475@opindex mcpu 30476A deprecated synonym for @option{-mtune}. 30477 30478@item -mfpmath=@var{unit} 30479@opindex mfpmath 30480Generate floating-point arithmetic for selected unit @var{unit}. The choices 30481for @var{unit} are: 30482 30483@table @samp 30484@item 387 30485Use the standard 387 floating-point coprocessor present on the majority of chips and 30486emulated otherwise. Code compiled with this option runs almost everywhere. 30487The temporary results are computed in 80-bit precision instead of the precision 30488specified by the type, resulting in slightly different results compared to most 30489of other chips. See @option{-ffloat-store} for more detailed description. 30490 30491This is the default choice for non-Darwin x86-32 targets. 30492 30493@item sse 30494Use scalar floating-point instructions present in the SSE instruction set. 30495This instruction set is supported by Pentium III and newer chips, 30496and in the AMD line 30497by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE 30498instruction set supports only single-precision arithmetic, thus the double and 30499extended-precision arithmetic are still done using 387. A later version, present 30500only in Pentium 4 and AMD x86-64 chips, supports double-precision 30501arithmetic too. 30502 30503For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse} 30504or @option{-msse2} switches to enable SSE extensions and make this option 30505effective. For the x86-64 compiler, these extensions are enabled by default. 30506 30507The resulting code should be considerably faster in the majority of cases and avoid 30508the numerical instability problems of 387 code, but may break some existing 30509code that expects temporaries to be 80 bits. 30510 30511This is the default choice for the x86-64 compiler, Darwin x86-32 targets, 30512and the default choice for x86-32 targets with the SSE2 instruction set 30513when @option{-ffast-math} is enabled. 30514 30515@item sse,387 30516@itemx sse+387 30517@itemx both 30518Attempt to utilize both instruction sets at once. This effectively doubles the 30519amount of available registers, and on chips with separate execution units for 30520387 and SSE the execution resources too. Use this option with care, as it is 30521still experimental, because the GCC register allocator does not model separate 30522functional units well, resulting in unstable performance. 30523@end table 30524 30525@item -masm=@var{dialect} 30526@opindex masm=@var{dialect} 30527Output assembly instructions using selected @var{dialect}. Also affects 30528which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and 30529extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect 30530order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does 30531not support @samp{intel}. 30532 30533@item -mieee-fp 30534@itemx -mno-ieee-fp 30535@opindex mieee-fp 30536@opindex mno-ieee-fp 30537Control whether or not the compiler uses IEEE floating-point 30538comparisons. These correctly handle the case where the result of a 30539comparison is unordered. 30540 30541@item -m80387 30542@itemx -mhard-float 30543@opindex 80387 30544@opindex mhard-float 30545Generate output containing 80387 instructions for floating point. 30546 30547@item -mno-80387 30548@itemx -msoft-float 30549@opindex no-80387 30550@opindex msoft-float 30551Generate output containing library calls for floating point. 30552 30553@strong{Warning:} the requisite libraries are not part of GCC@. 30554Normally the facilities of the machine's usual C compiler are used, but 30555this cannot be done directly in cross-compilation. You must make your 30556own arrangements to provide suitable library functions for 30557cross-compilation. 30558 30559On machines where a function returns floating-point results in the 80387 30560register stack, some floating-point opcodes may be emitted even if 30561@option{-msoft-float} is used. 30562 30563@item -mno-fp-ret-in-387 30564@opindex mno-fp-ret-in-387 30565@opindex mfp-ret-in-387 30566Do not use the FPU registers for return values of functions. 30567 30568The usual calling convention has functions return values of types 30569@code{float} and @code{double} in an FPU register, even if there 30570is no FPU@. The idea is that the operating system should emulate 30571an FPU@. 30572 30573The option @option{-mno-fp-ret-in-387} causes such values to be returned 30574in ordinary CPU registers instead. 30575 30576@item -mno-fancy-math-387 30577@opindex mno-fancy-math-387 30578@opindex mfancy-math-387 30579Some 387 emulators do not support the @code{sin}, @code{cos} and 30580@code{sqrt} instructions for the 387. Specify this option to avoid 30581generating those instructions. 30582This option is overridden when @option{-march} 30583indicates that the target CPU always has an FPU and so the 30584instruction does not need emulation. These 30585instructions are not generated unless you also use the 30586@option{-funsafe-math-optimizations} switch. 30587 30588@item -malign-double 30589@itemx -mno-align-double 30590@opindex malign-double 30591@opindex mno-align-double 30592Control whether GCC aligns @code{double}, @code{long double}, and 30593@code{long long} variables on a two-word boundary or a one-word 30594boundary. Aligning @code{double} variables on a two-word boundary 30595produces code that runs somewhat faster on a Pentium at the 30596expense of more memory. 30597 30598On x86-64, @option{-malign-double} is enabled by default. 30599 30600@strong{Warning:} if you use the @option{-malign-double} switch, 30601structures containing the above types are aligned differently than 30602the published application binary interface specifications for the x86-32 30603and are not binary compatible with structures in code compiled 30604without that switch. 30605 30606@item -m96bit-long-double 30607@itemx -m128bit-long-double 30608@opindex m96bit-long-double 30609@opindex m128bit-long-double 30610These switches control the size of @code{long double} type. The x86-32 30611application binary interface specifies the size to be 96 bits, 30612so @option{-m96bit-long-double} is the default in 32-bit mode. 30613 30614Modern architectures (Pentium and newer) prefer @code{long double} 30615to be aligned to an 8- or 16-byte boundary. In arrays or structures 30616conforming to the ABI, this is not possible. So specifying 30617@option{-m128bit-long-double} aligns @code{long double} 30618to a 16-byte boundary by padding the @code{long double} with an additional 3061932-bit zero. 30620 30621In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as 30622its ABI specifies that @code{long double} is aligned on 16-byte boundary. 30623 30624Notice that neither of these options enable any extra precision over the x87 30625standard of 80 bits for a @code{long double}. 30626 30627@strong{Warning:} if you override the default value for your target ABI, this 30628changes the size of 30629structures and arrays containing @code{long double} variables, 30630as well as modifying the function calling convention for functions taking 30631@code{long double}. Hence they are not binary-compatible 30632with code compiled without that switch. 30633 30634@item -mlong-double-64 30635@itemx -mlong-double-80 30636@itemx -mlong-double-128 30637@opindex mlong-double-64 30638@opindex mlong-double-80 30639@opindex mlong-double-128 30640These switches control the size of @code{long double} type. A size 30641of 64 bits makes the @code{long double} type equivalent to the @code{double} 30642type. This is the default for 32-bit Bionic C library. A size 30643of 128 bits makes the @code{long double} type equivalent to the 30644@code{__float128} type. This is the default for 64-bit Bionic C library. 30645 30646@strong{Warning:} if you override the default value for your target ABI, this 30647changes the size of 30648structures and arrays containing @code{long double} variables, 30649as well as modifying the function calling convention for functions taking 30650@code{long double}. Hence they are not binary-compatible 30651with code compiled without that switch. 30652 30653@item -malign-data=@var{type} 30654@opindex malign-data 30655Control how GCC aligns variables. Supported values for @var{type} are 30656@samp{compat} uses increased alignment value compatible uses GCC 4.8 30657and earlier, @samp{abi} uses alignment value as specified by the 30658psABI, and @samp{cacheline} uses increased alignment value to match 30659the cache line size. @samp{compat} is the default. 30660 30661@item -mlarge-data-threshold=@var{threshold} 30662@opindex mlarge-data-threshold 30663When @option{-mcmodel=medium} is specified, data objects larger than 30664@var{threshold} are placed in the large data section. This value must be the 30665same across all objects linked into the binary, and defaults to 65535. 30666 30667@item -mrtd 30668@opindex mrtd 30669Use a different function-calling convention, in which functions that 30670take a fixed number of arguments return with the @code{ret @var{num}} 30671instruction, which pops their arguments while returning. This saves one 30672instruction in the caller since there is no need to pop the arguments 30673there. 30674 30675You can specify that an individual function is called with this calling 30676sequence with the function attribute @code{stdcall}. You can also 30677override the @option{-mrtd} option by using the function attribute 30678@code{cdecl}. @xref{Function Attributes}. 30679 30680@strong{Warning:} this calling convention is incompatible with the one 30681normally used on Unix, so you cannot use it if you need to call 30682libraries compiled with the Unix compiler. 30683 30684Also, you must provide function prototypes for all functions that 30685take variable numbers of arguments (including @code{printf}); 30686otherwise incorrect code is generated for calls to those 30687functions. 30688 30689In addition, seriously incorrect code results if you call a 30690function with too many arguments. (Normally, extra arguments are 30691harmlessly ignored.) 30692 30693@item -mregparm=@var{num} 30694@opindex mregparm 30695Control how many registers are used to pass integer arguments. By 30696default, no registers are used to pass arguments, and at most 3 30697registers can be used. You can control this behavior for a specific 30698function by using the function attribute @code{regparm}. 30699@xref{Function Attributes}. 30700 30701@strong{Warning:} if you use this switch, and 30702@var{num} is nonzero, then you must build all modules with the same 30703value, including any libraries. This includes the system libraries and 30704startup modules. 30705 30706@item -msseregparm 30707@opindex msseregparm 30708Use SSE register passing conventions for float and double arguments 30709and return values. You can control this behavior for a specific 30710function by using the function attribute @code{sseregparm}. 30711@xref{Function Attributes}. 30712 30713@strong{Warning:} if you use this switch then you must build all 30714modules with the same value, including any libraries. This includes 30715the system libraries and startup modules. 30716 30717@item -mvect8-ret-in-mem 30718@opindex mvect8-ret-in-mem 30719Return 8-byte vectors in memory instead of MMX registers. This is the 30720default on VxWorks to match the ABI of the Sun Studio compilers until 30721version 12. @emph{Only} use this option if you need to remain 30722compatible with existing code produced by those previous compiler 30723versions or older versions of GCC@. 30724 30725@item -mpc32 30726@itemx -mpc64 30727@itemx -mpc80 30728@opindex mpc32 30729@opindex mpc64 30730@opindex mpc80 30731 30732Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32} 30733is specified, the significands of results of floating-point operations are 30734rounded to 24 bits (single precision); @option{-mpc64} rounds the 30735significands of results of floating-point operations to 53 bits (double 30736precision) and @option{-mpc80} rounds the significands of results of 30737floating-point operations to 64 bits (extended double precision), which is 30738the default. When this option is used, floating-point operations in higher 30739precisions are not available to the programmer without setting the FPU 30740control word explicitly. 30741 30742Setting the rounding of floating-point operations to less than the default 3074380 bits can speed some programs by 2% or more. Note that some mathematical 30744libraries assume that extended-precision (80-bit) floating-point operations 30745are enabled by default; routines in such libraries could suffer significant 30746loss of accuracy, typically through so-called ``catastrophic cancellation'', 30747when this option is used to set the precision to less than extended precision. 30748 30749@item -mstackrealign 30750@opindex mstackrealign 30751Realign the stack at entry. On the x86, the @option{-mstackrealign} 30752option generates an alternate prologue and epilogue that realigns the 30753run-time stack if necessary. This supports mixing legacy codes that keep 307544-byte stack alignment with modern codes that keep 16-byte stack alignment for 30755SSE compatibility. See also the attribute @code{force_align_arg_pointer}, 30756applicable to individual functions. 30757 30758@item -mpreferred-stack-boundary=@var{num} 30759@opindex mpreferred-stack-boundary 30760Attempt to keep the stack boundary aligned to a 2 raised to @var{num} 30761byte boundary. If @option{-mpreferred-stack-boundary} is not specified, 30762the default is 4 (16 bytes or 128 bits). 30763 30764@strong{Warning:} When generating code for the x86-64 architecture with 30765SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be 30766used to keep the stack boundary aligned to 8 byte boundary. Since 30767x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and 30768intended to be used in controlled environment where stack space is 30769important limitation. This option leads to wrong code when functions 30770compiled with 16 byte stack alignment (such as functions from a standard 30771library) are called with misaligned stack. In this case, SSE 30772instructions may lead to misaligned memory access traps. In addition, 30773variable arguments are handled incorrectly for 16 byte aligned 30774objects (including x87 long double and __int128), leading to wrong 30775results. You must build all modules with 30776@option{-mpreferred-stack-boundary=3}, including any libraries. This 30777includes the system libraries and startup modules. 30778 30779@item -mincoming-stack-boundary=@var{num} 30780@opindex mincoming-stack-boundary 30781Assume the incoming stack is aligned to a 2 raised to @var{num} byte 30782boundary. If @option{-mincoming-stack-boundary} is not specified, 30783the one specified by @option{-mpreferred-stack-boundary} is used. 30784 30785On Pentium and Pentium Pro, @code{double} and @code{long double} values 30786should be aligned to an 8-byte boundary (see @option{-malign-double}) or 30787suffer significant run time performance penalties. On Pentium III, the 30788Streaming SIMD Extension (SSE) data type @code{__m128} may not work 30789properly if it is not 16-byte aligned. 30790 30791To ensure proper alignment of this values on the stack, the stack boundary 30792must be as aligned as that required by any value stored on the stack. 30793Further, every function must be generated such that it keeps the stack 30794aligned. Thus calling a function compiled with a higher preferred 30795stack boundary from a function compiled with a lower preferred stack 30796boundary most likely misaligns the stack. It is recommended that 30797libraries that use callbacks always use the default setting. 30798 30799This extra alignment does consume extra stack space, and generally 30800increases code size. Code that is sensitive to stack space usage, such 30801as embedded systems and operating system kernels, may want to reduce the 30802preferred alignment to @option{-mpreferred-stack-boundary=2}. 30803 30804@need 200 30805@item -mmmx 30806@opindex mmmx 30807@need 200 30808@itemx -msse 30809@opindex msse 30810@need 200 30811@itemx -msse2 30812@opindex msse2 30813@need 200 30814@itemx -msse3 30815@opindex msse3 30816@need 200 30817@itemx -mssse3 30818@opindex mssse3 30819@need 200 30820@itemx -msse4 30821@opindex msse4 30822@need 200 30823@itemx -msse4a 30824@opindex msse4a 30825@need 200 30826@itemx -msse4.1 30827@opindex msse4.1 30828@need 200 30829@itemx -msse4.2 30830@opindex msse4.2 30831@need 200 30832@itemx -mavx 30833@opindex mavx 30834@need 200 30835@itemx -mavx2 30836@opindex mavx2 30837@need 200 30838@itemx -mavx512f 30839@opindex mavx512f 30840@need 200 30841@itemx -mavx512pf 30842@opindex mavx512pf 30843@need 200 30844@itemx -mavx512er 30845@opindex mavx512er 30846@need 200 30847@itemx -mavx512cd 30848@opindex mavx512cd 30849@need 200 30850@itemx -mavx512vl 30851@opindex mavx512vl 30852@need 200 30853@itemx -mavx512bw 30854@opindex mavx512bw 30855@need 200 30856@itemx -mavx512dq 30857@opindex mavx512dq 30858@need 200 30859@itemx -mavx512ifma 30860@opindex mavx512ifma 30861@need 200 30862@itemx -mavx512vbmi 30863@opindex mavx512vbmi 30864@need 200 30865@itemx -msha 30866@opindex msha 30867@need 200 30868@itemx -maes 30869@opindex maes 30870@need 200 30871@itemx -mpclmul 30872@opindex mpclmul 30873@need 200 30874@itemx -mclflushopt 30875@opindex mclflushopt 30876@need 200 30877@itemx -mclwb 30878@opindex mclwb 30879@need 200 30880@itemx -mfsgsbase 30881@opindex mfsgsbase 30882@need 200 30883@itemx -mptwrite 30884@opindex mptwrite 30885@need 200 30886@itemx -mrdrnd 30887@opindex mrdrnd 30888@need 200 30889@itemx -mf16c 30890@opindex mf16c 30891@need 200 30892@itemx -mfma 30893@opindex mfma 30894@need 200 30895@itemx -mpconfig 30896@opindex mpconfig 30897@need 200 30898@itemx -mwbnoinvd 30899@opindex mwbnoinvd 30900@need 200 30901@itemx -mfma4 30902@opindex mfma4 30903@need 200 30904@itemx -mprfchw 30905@opindex mprfchw 30906@need 200 30907@itemx -mrdpid 30908@opindex mrdpid 30909@need 200 30910@itemx -mprefetchwt1 30911@opindex mprefetchwt1 30912@need 200 30913@itemx -mrdseed 30914@opindex mrdseed 30915@need 200 30916@itemx -msgx 30917@opindex msgx 30918@need 200 30919@itemx -mxop 30920@opindex mxop 30921@need 200 30922@itemx -mlwp 30923@opindex mlwp 30924@need 200 30925@itemx -m3dnow 30926@opindex m3dnow 30927@need 200 30928@itemx -m3dnowa 30929@opindex m3dnowa 30930@need 200 30931@itemx -mpopcnt 30932@opindex mpopcnt 30933@need 200 30934@itemx -mabm 30935@opindex mabm 30936@need 200 30937@itemx -madx 30938@opindex madx 30939@need 200 30940@itemx -mbmi 30941@opindex mbmi 30942@need 200 30943@itemx -mbmi2 30944@opindex mbmi2 30945@need 200 30946@itemx -mlzcnt 30947@opindex mlzcnt 30948@need 200 30949@itemx -mfxsr 30950@opindex mfxsr 30951@need 200 30952@itemx -mxsave 30953@opindex mxsave 30954@need 200 30955@itemx -mxsaveopt 30956@opindex mxsaveopt 30957@need 200 30958@itemx -mxsavec 30959@opindex mxsavec 30960@need 200 30961@itemx -mxsaves 30962@opindex mxsaves 30963@need 200 30964@itemx -mrtm 30965@opindex mrtm 30966@need 200 30967@itemx -mhle 30968@opindex mhle 30969@need 200 30970@itemx -mtbm 30971@opindex mtbm 30972@need 200 30973@itemx -mmwaitx 30974@opindex mmwaitx 30975@need 200 30976@itemx -mclzero 30977@opindex mclzero 30978@need 200 30979@itemx -mpku 30980@opindex mpku 30981@need 200 30982@itemx -mavx512vbmi2 30983@opindex mavx512vbmi2 30984@need 200 30985@itemx -mavx512bf16 30986@opindex mavx512bf16 30987@need 200 30988@itemx -mgfni 30989@opindex mgfni 30990@need 200 30991@itemx -mvaes 30992@opindex mvaes 30993@need 200 30994@itemx -mwaitpkg 30995@opindex mwaitpkg 30996@need 200 30997@itemx -mvpclmulqdq 30998@opindex mvpclmulqdq 30999@need 200 31000@itemx -mavx512bitalg 31001@opindex mavx512bitalg 31002@need 200 31003@itemx -mmovdiri 31004@opindex mmovdiri 31005@need 200 31006@itemx -mmovdir64b 31007@opindex mmovdir64b 31008@need 200 31009@itemx -menqcmd 31010@opindex menqcmd 31011@itemx -muintr 31012@opindex muintr 31013@need 200 31014@itemx -mtsxldtrk 31015@opindex mtsxldtrk 31016@need 200 31017@itemx -mavx512vpopcntdq 31018@opindex mavx512vpopcntdq 31019@need 200 31020@itemx -mavx512vp2intersect 31021@opindex mavx512vp2intersect 31022@need 200 31023@itemx -mavx5124fmaps 31024@opindex mavx5124fmaps 31025@need 200 31026@itemx -mavx512vnni 31027@opindex mavx512vnni 31028@need 200 31029@itemx -mavxvnni 31030@opindex mavxvnni 31031@need 200 31032@itemx -mavx5124vnniw 31033@opindex mavx5124vnniw 31034@need 200 31035@itemx -mcldemote 31036@opindex mcldemote 31037@need 200 31038@itemx -mserialize 31039@opindex mserialize 31040@need 200 31041@itemx -mamx-tile 31042@opindex mamx-tile 31043@need 200 31044@itemx -mamx-int8 31045@opindex mamx-int8 31046@need 200 31047@itemx -mamx-bf16 31048@opindex mamx-bf16 31049@need 200 31050@itemx -mhreset 31051@opindex mhreset 31052@itemx -mkl 31053@opindex mkl 31054@need 200 31055@itemx -mwidekl 31056@opindex mwidekl 31057These switches enable the use of instructions in the MMX, SSE, 31058SSE2, SSE3, SSSE3, SSE4, SSE4A, SSE4.1, SSE4.2, AVX, AVX2, AVX512F, AVX512PF, 31059AVX512ER, AVX512CD, AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA, AVX512VBMI, SHA, 31060AES, PCLMUL, CLFLUSHOPT, CLWB, FSGSBASE, PTWRITE, RDRND, F16C, FMA, PCONFIG, 31061WBNOINVD, FMA4, PREFETCHW, RDPID, PREFETCHWT1, RDSEED, SGX, XOP, LWP, 310623DNow!@:, enhanced 3DNow!@:, POPCNT, ABM, ADX, BMI, BMI2, LZCNT, FXSR, XSAVE, 31063XSAVEOPT, XSAVEC, XSAVES, RTM, HLE, TBM, MWAITX, CLZERO, PKU, AVX512VBMI2, 31064GFNI, VAES, WAITPKG, VPCLMULQDQ, AVX512BITALG, MOVDIRI, MOVDIR64B, AVX512BF16, 31065ENQCMD, AVX512VPOPCNTDQ, AVX5124FMAPS, AVX512VNNI, AVX5124VNNIW, SERIALIZE, 31066UINTR, HRESET, AMXTILE, AMXINT8, AMXBF16, KL, WIDEKL, AVXVNNI or CLDEMOTE 31067extended instruction sets. Each has a corresponding @option{-mno-} option to 31068disable use of these instructions. 31069 31070These extensions are also available as built-in functions: see 31071@ref{x86 Built-in Functions}, for details of the functions enabled and 31072disabled by these switches. 31073 31074To generate SSE/SSE2 instructions automatically from floating-point 31075code (as opposed to 387 instructions), see @option{-mfpmath=sse}. 31076 31077GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it 31078generates new AVX instructions or AVX equivalence for all SSEx instructions 31079when needed. 31080 31081These options enable GCC to use these extended instructions in 31082generated code, even without @option{-mfpmath=sse}. Applications that 31083perform run-time CPU detection must compile separate files for each 31084supported architecture, using the appropriate flags. In particular, 31085the file containing the CPU detection code should be compiled without 31086these options. 31087 31088@item -mdump-tune-features 31089@opindex mdump-tune-features 31090This option instructs GCC to dump the names of the x86 performance 31091tuning features and default settings. The names can be used in 31092@option{-mtune-ctrl=@var{feature-list}}. 31093 31094@item -mtune-ctrl=@var{feature-list} 31095@opindex mtune-ctrl=@var{feature-list} 31096This option is used to do fine grain control of x86 code generation features. 31097@var{feature-list} is a comma separated list of @var{feature} names. See also 31098@option{-mdump-tune-features}. When specified, the @var{feature} is turned 31099on if it is not preceded with @samp{^}, otherwise, it is turned off. 31100@option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC 31101developers. Using it may lead to code paths not covered by testing and can 31102potentially result in compiler ICEs or runtime errors. 31103 31104@item -mno-default 31105@opindex mno-default 31106This option instructs GCC to turn off all tunable features. See also 31107@option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}. 31108 31109@item -mcld 31110@opindex mcld 31111This option instructs GCC to emit a @code{cld} instruction in the prologue 31112of functions that use string instructions. String instructions depend on 31113the DF flag to select between autoincrement or autodecrement mode. While the 31114ABI specifies the DF flag to be cleared on function entry, some operating 31115systems violate this specification by not clearing the DF flag in their 31116exception dispatchers. The exception handler can be invoked with the DF flag 31117set, which leads to wrong direction mode when string instructions are used. 31118This option can be enabled by default on 32-bit x86 targets by configuring 31119GCC with the @option{--enable-cld} configure option. Generation of @code{cld} 31120instructions can be suppressed with the @option{-mno-cld} compiler option 31121in this case. 31122 31123@item -mvzeroupper 31124@opindex mvzeroupper 31125This option instructs GCC to emit a @code{vzeroupper} instruction 31126before a transfer of control flow out of the function to minimize 31127the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper} 31128intrinsics. 31129 31130@item -mprefer-avx128 31131@opindex mprefer-avx128 31132This option instructs GCC to use 128-bit AVX instructions instead of 31133256-bit AVX instructions in the auto-vectorizer. 31134 31135@item -mprefer-vector-width=@var{opt} 31136@opindex mprefer-vector-width 31137This option instructs GCC to use @var{opt}-bit vector width in instructions 31138instead of default on the selected platform. 31139 31140@table @samp 31141@item none 31142No extra limitations applied to GCC other than defined by the selected platform. 31143 31144@item 128 31145Prefer 128-bit vector width for instructions. 31146 31147@item 256 31148Prefer 256-bit vector width for instructions. 31149 31150@item 512 31151Prefer 512-bit vector width for instructions. 31152@end table 31153 31154@item -mcx16 31155@opindex mcx16 31156This option enables GCC to generate @code{CMPXCHG16B} instructions in 64-bit 31157code to implement compare-and-exchange operations on 16-byte aligned 128-bit 31158objects. This is useful for atomic updates of data structures exceeding one 31159machine word in size. The compiler uses this instruction to implement 31160@ref{__sync Builtins}. However, for @ref{__atomic Builtins} operating on 31161128-bit integers, a library call is always used. 31162 31163@item -msahf 31164@opindex msahf 31165This option enables generation of @code{SAHF} instructions in 64-bit code. 31166Early Intel Pentium 4 CPUs with Intel 64 support, 31167prior to the introduction of Pentium 4 G1 step in December 2005, 31168lacked the @code{LAHF} and @code{SAHF} instructions 31169which are supported by AMD64. 31170These are load and store instructions, respectively, for certain status flags. 31171In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod}, 31172@code{drem}, and @code{remainder} built-in functions; 31173see @ref{Other Builtins} for details. 31174 31175@item -mmovbe 31176@opindex mmovbe 31177This option enables use of the @code{movbe} instruction to implement 31178@code{__builtin_bswap32} and @code{__builtin_bswap64}. 31179 31180@item -mshstk 31181@opindex mshstk 31182The @option{-mshstk} option enables shadow stack built-in functions 31183from x86 Control-flow Enforcement Technology (CET). 31184 31185@item -mcrc32 31186@opindex mcrc32 31187This option enables built-in functions @code{__builtin_ia32_crc32qi}, 31188@code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and 31189@code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction. 31190 31191@item -mmwait 31192@opindex mmwait 31193This option enables built-in functions @code{__builtin_ia32_monitor}, 31194and @code{__builtin_ia32_mwait} to generate the @code{monitor} and 31195@code{mwait} machine instructions. 31196 31197@item -mrecip 31198@opindex mrecip 31199This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions 31200(and their vectorized variants @code{RCPPS} and @code{RSQRTPS}) 31201with an additional Newton-Raphson step 31202to increase precision instead of @code{DIVSS} and @code{SQRTSS} 31203(and their vectorized 31204variants) for single-precision floating-point arguments. These instructions 31205are generated only when @option{-funsafe-math-optimizations} is enabled 31206together with @option{-ffinite-math-only} and @option{-fno-trapping-math}. 31207Note that while the throughput of the sequence is higher than the throughput 31208of the non-reciprocal instruction, the precision of the sequence can be 31209decreased by up to 2 ulp (i.e.@: the inverse of 1.0 equals 0.99999994). 31210 31211Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS} 31212(or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option 31213combination), and doesn't need @option{-mrecip}. 31214 31215Also note that GCC emits the above sequence with additional Newton-Raphson step 31216for vectorized single-float division and vectorized @code{sqrtf(@var{x})} 31217already with @option{-ffast-math} (or the above option combination), and 31218doesn't need @option{-mrecip}. 31219 31220@item -mrecip=@var{opt} 31221@opindex mrecip=opt 31222This option controls which reciprocal estimate instructions 31223may be used. @var{opt} is a comma-separated list of options, which may 31224be preceded by a @samp{!} to invert the option: 31225 31226@table @samp 31227@item all 31228Enable all estimate instructions. 31229 31230@item default 31231Enable the default instructions, equivalent to @option{-mrecip}. 31232 31233@item none 31234Disable all estimate instructions, equivalent to @option{-mno-recip}. 31235 31236@item div 31237Enable the approximation for scalar division. 31238 31239@item vec-div 31240Enable the approximation for vectorized division. 31241 31242@item sqrt 31243Enable the approximation for scalar square root. 31244 31245@item vec-sqrt 31246Enable the approximation for vectorized square root. 31247@end table 31248 31249So, for example, @option{-mrecip=all,!sqrt} enables 31250all of the reciprocal approximations, except for square root. 31251 31252@item -mveclibabi=@var{type} 31253@opindex mveclibabi 31254Specifies the ABI type to use for vectorizing intrinsics using an 31255external library. Supported values for @var{type} are @samp{svml} 31256for the Intel short 31257vector math library and @samp{acml} for the AMD math core library. 31258To use this option, both @option{-ftree-vectorize} and 31259@option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML 31260ABI-compatible library must be specified at link time. 31261 31262GCC currently emits calls to @code{vmldExp2}, 31263@code{vmldLn2}, @code{vmldLog102}, @code{vmldPow2}, 31264@code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2}, 31265@code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2}, 31266@code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2}, 31267@code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, 31268@code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4}, 31269@code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4}, 31270@code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4}, 31271@code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding 31272function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin}, 31273@code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2}, 31274@code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf}, 31275@code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f}, 31276@code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type 31277when @option{-mveclibabi=acml} is used. 31278 31279@item -mabi=@var{name} 31280@opindex mabi 31281Generate code for the specified calling convention. Permissible values 31282are @samp{sysv} for the ABI used on GNU/Linux and other systems, and 31283@samp{ms} for the Microsoft ABI. The default is to use the Microsoft 31284ABI when targeting Microsoft Windows and the SysV ABI on all other systems. 31285You can control this behavior for specific functions by 31286using the function attributes @code{ms_abi} and @code{sysv_abi}. 31287@xref{Function Attributes}. 31288 31289@item -mforce-indirect-call 31290@opindex mforce-indirect-call 31291Force all calls to functions to be indirect. This is useful 31292when using Intel Processor Trace where it generates more precise timing 31293information for function calls. 31294 31295@item -mmanual-endbr 31296@opindex mmanual-endbr 31297Insert ENDBR instruction at function entry only via the @code{cf_check} 31298function attribute. This is useful when used with the option 31299@option{-fcf-protection=branch} to control ENDBR insertion at the 31300function entry. 31301 31302@item -mcall-ms2sysv-xlogues 31303@opindex mcall-ms2sysv-xlogues 31304@opindex mno-call-ms2sysv-xlogues 31305Due to differences in 64-bit ABIs, any Microsoft ABI function that calls a 31306System V ABI function must consider RSI, RDI and XMM6-15 as clobbered. By 31307default, the code for saving and restoring these registers is emitted inline, 31308resulting in fairly lengthy prologues and epilogues. Using 31309@option{-mcall-ms2sysv-xlogues} emits prologues and epilogues that 31310use stubs in the static portion of libgcc to perform these saves and restores, 31311thus reducing function size at the cost of a few extra instructions. 31312 31313@item -mtls-dialect=@var{type} 31314@opindex mtls-dialect 31315Generate code to access thread-local storage using the @samp{gnu} or 31316@samp{gnu2} conventions. @samp{gnu} is the conservative default; 31317@samp{gnu2} is more efficient, but it may add compile- and run-time 31318requirements that cannot be satisfied on all systems. 31319 31320@item -mpush-args 31321@itemx -mno-push-args 31322@opindex mpush-args 31323@opindex mno-push-args 31324Use PUSH operations to store outgoing parameters. This method is shorter 31325and usually equally fast as method using SUB/MOV operations and is enabled 31326by default. In some cases disabling it may improve performance because of 31327improved scheduling and reduced dependencies. 31328 31329@item -maccumulate-outgoing-args 31330@opindex maccumulate-outgoing-args 31331If enabled, the maximum amount of space required for outgoing arguments is 31332computed in the function prologue. This is faster on most modern CPUs 31333because of reduced dependencies, improved scheduling and reduced stack usage 31334when the preferred stack boundary is not equal to 2. The drawback is a notable 31335increase in code size. This switch implies @option{-mno-push-args}. 31336 31337@item -mthreads 31338@opindex mthreads 31339Support thread-safe exception handling on MinGW. Programs that rely 31340on thread-safe exception handling must compile and link all code with the 31341@option{-mthreads} option. When compiling, @option{-mthreads} defines 31342@option{-D_MT}; when linking, it links in a special thread helper library 31343@option{-lmingwthrd} which cleans up per-thread exception-handling data. 31344 31345@item -mms-bitfields 31346@itemx -mno-ms-bitfields 31347@opindex mms-bitfields 31348@opindex mno-ms-bitfields 31349 31350Enable/disable bit-field layout compatible with the native Microsoft 31351Windows compiler. 31352 31353If @code{packed} is used on a structure, or if bit-fields are used, 31354it may be that the Microsoft ABI lays out the structure differently 31355than the way GCC normally does. Particularly when moving packed 31356data between functions compiled with GCC and the native Microsoft compiler 31357(either via function call or as data in a file), it may be necessary to access 31358either format. 31359 31360This option is enabled by default for Microsoft Windows 31361targets. This behavior can also be controlled locally by use of variable 31362or type attributes. For more information, see @ref{x86 Variable Attributes} 31363and @ref{x86 Type Attributes}. 31364 31365The Microsoft structure layout algorithm is fairly simple with the exception 31366of the bit-field packing. 31367The padding and alignment of members of structures and whether a bit-field 31368can straddle a storage-unit boundary are determine by these rules: 31369 31370@enumerate 31371@item Structure members are stored sequentially in the order in which they are 31372declared: the first member has the lowest memory address and the last member 31373the highest. 31374 31375@item Every data object has an alignment requirement. The alignment requirement 31376for all data except structures, unions, and arrays is either the size of the 31377object or the current packing size (specified with either the 31378@code{aligned} attribute or the @code{pack} pragma), 31379whichever is less. For structures, unions, and arrays, 31380the alignment requirement is the largest alignment requirement of its members. 31381Every object is allocated an offset so that: 31382 31383@smallexample 31384offset % alignment_requirement == 0 31385@end smallexample 31386 31387@item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation 31388unit if the integral types are the same size and if the next bit-field fits 31389into the current allocation unit without crossing the boundary imposed by the 31390common alignment requirements of the bit-fields. 31391@end enumerate 31392 31393MSVC interprets zero-length bit-fields in the following ways: 31394 31395@enumerate 31396@item If a zero-length bit-field is inserted between two bit-fields that 31397are normally coalesced, the bit-fields are not coalesced. 31398 31399For example: 31400 31401@smallexample 31402struct 31403 @{ 31404 unsigned long bf_1 : 12; 31405 unsigned long : 0; 31406 unsigned long bf_2 : 12; 31407 @} t1; 31408@end smallexample 31409 31410@noindent 31411The size of @code{t1} is 8 bytes with the zero-length bit-field. If the 31412zero-length bit-field were removed, @code{t1}'s size would be 4 bytes. 31413 31414@item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the 31415alignment of the zero-length bit-field is greater than the member that follows it, 31416@code{bar}, @code{bar} is aligned as the type of the zero-length bit-field. 31417 31418For example: 31419 31420@smallexample 31421struct 31422 @{ 31423 char foo : 4; 31424 short : 0; 31425 char bar; 31426 @} t2; 31427 31428struct 31429 @{ 31430 char foo : 4; 31431 short : 0; 31432 double bar; 31433 @} t3; 31434@end smallexample 31435 31436@noindent 31437For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1. 31438Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length 31439bit-field does not affect the alignment of @code{bar} or, as a result, the size 31440of the structure. 31441 31442Taking this into account, it is important to note the following: 31443 31444@enumerate 31445@item If a zero-length bit-field follows a normal bit-field, the type of the 31446zero-length bit-field may affect the alignment of the structure as whole. For 31447example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a 31448normal bit-field, and is of type short. 31449 31450@item Even if a zero-length bit-field is not followed by a normal bit-field, it may 31451still affect the alignment of the structure: 31452 31453@smallexample 31454struct 31455 @{ 31456 char foo : 6; 31457 long : 0; 31458 @} t4; 31459@end smallexample 31460 31461@noindent 31462Here, @code{t4} takes up 4 bytes. 31463@end enumerate 31464 31465@item Zero-length bit-fields following non-bit-field members are ignored: 31466 31467@smallexample 31468struct 31469 @{ 31470 char foo; 31471 long : 0; 31472 char bar; 31473 @} t5; 31474@end smallexample 31475 31476@noindent 31477Here, @code{t5} takes up 2 bytes. 31478@end enumerate 31479 31480 31481@item -mno-align-stringops 31482@opindex mno-align-stringops 31483@opindex malign-stringops 31484Do not align the destination of inlined string operations. This switch reduces 31485code size and improves performance in case the destination is already aligned, 31486but GCC doesn't know about it. 31487 31488@item -minline-all-stringops 31489@opindex minline-all-stringops 31490By default GCC inlines string operations only when the destination is 31491known to be aligned to least a 4-byte boundary. 31492This enables more inlining and increases code 31493size, but may improve performance of code that depends on fast 31494@code{memcpy} and @code{memset} for short lengths. 31495The option enables inline expansion of @code{strlen} for all 31496pointer alignments. 31497 31498@item -minline-stringops-dynamically 31499@opindex minline-stringops-dynamically 31500For string operations of unknown size, use run-time checks with 31501inline code for small blocks and a library call for large blocks. 31502 31503@item -mstringop-strategy=@var{alg} 31504@opindex mstringop-strategy=@var{alg} 31505Override the internal decision heuristic for the particular algorithm to use 31506for inlining string operations. The allowed values for @var{alg} are: 31507 31508@table @samp 31509@item rep_byte 31510@itemx rep_4byte 31511@itemx rep_8byte 31512Expand using i386 @code{rep} prefix of the specified size. 31513 31514@item byte_loop 31515@itemx loop 31516@itemx unrolled_loop 31517Expand into an inline loop. 31518 31519@item libcall 31520Always use a library call. 31521@end table 31522 31523@item -mmemcpy-strategy=@var{strategy} 31524@opindex mmemcpy-strategy=@var{strategy} 31525Override the internal decision heuristic to decide if @code{__builtin_memcpy} 31526should be inlined and what inline algorithm to use when the expected size 31527of the copy operation is known. @var{strategy} 31528is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets. 31529@var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies 31530the max byte size with which inline algorithm @var{alg} is allowed. For the last 31531triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets 31532in the list must be specified in increasing order. The minimal byte size for 31533@var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the 31534preceding range. 31535 31536@item -mmemset-strategy=@var{strategy} 31537@opindex mmemset-strategy=@var{strategy} 31538The option is similar to @option{-mmemcpy-strategy=} except that it is to control 31539@code{__builtin_memset} expansion. 31540 31541@item -momit-leaf-frame-pointer 31542@opindex momit-leaf-frame-pointer 31543Don't keep the frame pointer in a register for leaf functions. This 31544avoids the instructions to save, set up, and restore frame pointers and 31545makes an extra register available in leaf functions. The option 31546@option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions, 31547which might make debugging harder. 31548 31549@item -mtls-direct-seg-refs 31550@itemx -mno-tls-direct-seg-refs 31551@opindex mtls-direct-seg-refs 31552Controls whether TLS variables may be accessed with offsets from the 31553TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit), 31554or whether the thread base pointer must be added. Whether or not this 31555is valid depends on the operating system, and whether it maps the 31556segment to cover the entire TLS area. 31557 31558For systems that use the GNU C Library, the default is on. 31559 31560@item -msse2avx 31561@itemx -mno-sse2avx 31562@opindex msse2avx 31563Specify that the assembler should encode SSE instructions with VEX 31564prefix. The option @option{-mavx} turns this on by default. 31565 31566@item -mfentry 31567@itemx -mno-fentry 31568@opindex mfentry 31569If profiling is active (@option{-pg}), put the profiling 31570counter call before the prologue. 31571Note: On x86 architectures the attribute @code{ms_hook_prologue} 31572isn't possible at the moment for @option{-mfentry} and @option{-pg}. 31573 31574@item -mrecord-mcount 31575@itemx -mno-record-mcount 31576@opindex mrecord-mcount 31577If profiling is active (@option{-pg}), generate a __mcount_loc section 31578that contains pointers to each profiling call. This is useful for 31579automatically patching and out calls. 31580 31581@item -mnop-mcount 31582@itemx -mno-nop-mcount 31583@opindex mnop-mcount 31584If profiling is active (@option{-pg}), generate the calls to 31585the profiling functions as NOPs. This is useful when they 31586should be patched in later dynamically. This is likely only 31587useful together with @option{-mrecord-mcount}. 31588 31589@item -minstrument-return=@var{type} 31590@opindex minstrument-return 31591Instrument function exit in -pg -mfentry instrumented functions with 31592call to specified function. This only instruments true returns ending 31593with ret, but not sibling calls ending with jump. Valid types 31594are @var{none} to not instrument, @var{call} to generate a call to __return__, 31595or @var{nop5} to generate a 5 byte nop. 31596 31597@item -mrecord-return 31598@itemx -mno-record-return 31599@opindex mrecord-return 31600Generate a __return_loc section pointing to all return instrumentation code. 31601 31602@item -mfentry-name=@var{name} 31603@opindex mfentry-name 31604Set name of __fentry__ symbol called at function entry for -pg -mfentry functions. 31605 31606@item -mfentry-section=@var{name} 31607@opindex mfentry-section 31608Set name of section to record -mrecord-mcount calls (default __mcount_loc). 31609 31610@item -mskip-rax-setup 31611@itemx -mno-skip-rax-setup 31612@opindex mskip-rax-setup 31613When generating code for the x86-64 architecture with SSE extensions 31614disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX 31615register when there are no variable arguments passed in vector registers. 31616 31617@strong{Warning:} Since RAX register is used to avoid unnecessarily 31618saving vector registers on stack when passing variable arguments, the 31619impacts of this option are callees may waste some stack space, 31620misbehave or jump to a random location. GCC 4.4 or newer don't have 31621those issues, regardless the RAX register value. 31622 31623@item -m8bit-idiv 31624@itemx -mno-8bit-idiv 31625@opindex m8bit-idiv 31626On some processors, like Intel Atom, 8-bit unsigned integer divide is 31627much faster than 32-bit/64-bit integer divide. This option generates a 31628run-time check. If both dividend and divisor are within range of 0 31629to 255, 8-bit unsigned integer divide is used instead of 3163032-bit/64-bit integer divide. 31631 31632@item -mavx256-split-unaligned-load 31633@itemx -mavx256-split-unaligned-store 31634@opindex mavx256-split-unaligned-load 31635@opindex mavx256-split-unaligned-store 31636Split 32-byte AVX unaligned load and store. 31637 31638@item -mstack-protector-guard=@var{guard} 31639@itemx -mstack-protector-guard-reg=@var{reg} 31640@itemx -mstack-protector-guard-offset=@var{offset} 31641@opindex mstack-protector-guard 31642@opindex mstack-protector-guard-reg 31643@opindex mstack-protector-guard-offset 31644Generate stack protection code using canary at @var{guard}. Supported 31645locations are @samp{global} for global canary or @samp{tls} for per-thread 31646canary in the TLS block (the default). This option has effect only when 31647@option{-fstack-protector} or @option{-fstack-protector-all} is specified. 31648 31649With the latter choice the options 31650@option{-mstack-protector-guard-reg=@var{reg}} and 31651@option{-mstack-protector-guard-offset=@var{offset}} furthermore specify 31652which segment register (@code{%fs} or @code{%gs}) to use as base register 31653for reading the canary, and from what offset from that base register. 31654The default for those is as specified in the relevant ABI. 31655 31656@item -mgeneral-regs-only 31657@opindex mgeneral-regs-only 31658Generate code that uses only the general-purpose registers. This 31659prevents the compiler from using floating-point, vector, mask and bound 31660registers. 31661 31662@item -mindirect-branch=@var{choice} 31663@opindex mindirect-branch 31664Convert indirect call and jump with @var{choice}. The default is 31665@samp{keep}, which keeps indirect call and jump unmodified. 31666@samp{thunk} converts indirect call and jump to call and return thunk. 31667@samp{thunk-inline} converts indirect call and jump to inlined call 31668and return thunk. @samp{thunk-extern} converts indirect call and jump 31669to external call and return thunk provided in a separate object file. 31670You can control this behavior for a specific function by using the 31671function attribute @code{indirect_branch}. @xref{Function Attributes}. 31672 31673Note that @option{-mcmodel=large} is incompatible with 31674@option{-mindirect-branch=thunk} and 31675@option{-mindirect-branch=thunk-extern} since the thunk function may 31676not be reachable in the large code model. 31677 31678Note that @option{-mindirect-branch=thunk-extern} is compatible with 31679@option{-fcf-protection=branch} since the external thunk can be made 31680to enable control-flow check. 31681 31682@item -mfunction-return=@var{choice} 31683@opindex mfunction-return 31684Convert function return with @var{choice}. The default is @samp{keep}, 31685which keeps function return unmodified. @samp{thunk} converts function 31686return to call and return thunk. @samp{thunk-inline} converts function 31687return to inlined call and return thunk. @samp{thunk-extern} converts 31688function return to external call and return thunk provided in a separate 31689object file. You can control this behavior for a specific function by 31690using the function attribute @code{function_return}. 31691@xref{Function Attributes}. 31692 31693Note that @option{-mindirect-return=thunk-extern} is compatible with 31694@option{-fcf-protection=branch} since the external thunk can be made 31695to enable control-flow check. 31696 31697Note that @option{-mcmodel=large} is incompatible with 31698@option{-mfunction-return=thunk} and 31699@option{-mfunction-return=thunk-extern} since the thunk function may 31700not be reachable in the large code model. 31701 31702 31703@item -mindirect-branch-register 31704@opindex mindirect-branch-register 31705Force indirect call and jump via register. 31706 31707@end table 31708 31709These @samp{-m} switches are supported in addition to the above 31710on x86-64 processors in 64-bit environments. 31711 31712@table @gcctabopt 31713@item -m32 31714@itemx -m64 31715@itemx -mx32 31716@itemx -m16 31717@itemx -miamcu 31718@opindex m32 31719@opindex m64 31720@opindex mx32 31721@opindex m16 31722@opindex miamcu 31723Generate code for a 16-bit, 32-bit or 64-bit environment. 31724The @option{-m32} option sets @code{int}, @code{long}, and pointer types 31725to 32 bits, and 31726generates code that runs on any i386 system. 31727 31728The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer 31729types to 64 bits, and generates code for the x86-64 architecture. 31730For Darwin only the @option{-m64} option also turns off the @option{-fno-pic} 31731and @option{-mdynamic-no-pic} options. 31732 31733The @option{-mx32} option sets @code{int}, @code{long}, and pointer types 31734to 32 bits, and 31735generates code for the x86-64 architecture. 31736 31737The @option{-m16} option is the same as @option{-m32}, except for that 31738it outputs the @code{.code16gcc} assembly directive at the beginning of 31739the assembly output so that the binary can run in 16-bit mode. 31740 31741The @option{-miamcu} option generates code which conforms to Intel MCU 31742psABI. It requires the @option{-m32} option to be turned on. 31743 31744@item -mno-red-zone 31745@opindex mno-red-zone 31746@opindex mred-zone 31747Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated 31748by the x86-64 ABI; it is a 128-byte area beyond the location of the 31749stack pointer that is not modified by signal or interrupt handlers 31750and therefore can be used for temporary data without adjusting the stack 31751pointer. The flag @option{-mno-red-zone} disables this red zone. 31752 31753@item -mcmodel=small 31754@opindex mcmodel=small 31755Generate code for the small code model: the program and its symbols must 31756be linked in the lower 2 GB of the address space. Pointers are 64 bits. 31757Programs can be statically or dynamically linked. This is the default 31758code model. 31759 31760@item -mcmodel=kernel 31761@opindex mcmodel=kernel 31762Generate code for the kernel code model. The kernel runs in the 31763negative 2 GB of the address space. 31764This model has to be used for Linux kernel code. 31765 31766@item -mcmodel=medium 31767@opindex mcmodel=medium 31768Generate code for the medium model: the program is linked in the lower 2 31769GB of the address space. Small symbols are also placed there. Symbols 31770with sizes larger than @option{-mlarge-data-threshold} are put into 31771large data or BSS sections and can be located above 2GB. Programs can 31772be statically or dynamically linked. 31773 31774@item -mcmodel=large 31775@opindex mcmodel=large 31776Generate code for the large model. This model makes no assumptions 31777about addresses and sizes of sections. 31778 31779@item -maddress-mode=long 31780@opindex maddress-mode=long 31781Generate code for long address mode. This is only supported for 64-bit 31782and x32 environments. It is the default address mode for 64-bit 31783environments. 31784 31785@item -maddress-mode=short 31786@opindex maddress-mode=short 31787Generate code for short address mode. This is only supported for 32-bit 31788and x32 environments. It is the default address mode for 32-bit and 31789x32 environments. 31790 31791@item -mneeded 31792@itemx -mno-needed 31793@opindex mneeded 31794Emit GNU_PROPERTY_X86_ISA_1_NEEDED GNU property for Linux target to 31795indicate the micro-architecture ISA level required to execute the binary. 31796@end table 31797 31798@node x86 Windows Options 31799@subsection x86 Windows Options 31800@cindex x86 Windows Options 31801@cindex Windows Options for x86 31802 31803These additional options are available for Microsoft Windows targets: 31804 31805@table @gcctabopt 31806@item -mconsole 31807@opindex mconsole 31808This option 31809specifies that a console application is to be generated, by 31810instructing the linker to set the PE header subsystem type 31811required for console applications. 31812This option is available for Cygwin and MinGW targets and is 31813enabled by default on those targets. 31814 31815@item -mdll 31816@opindex mdll 31817This option is available for Cygwin and MinGW targets. It 31818specifies that a DLL---a dynamic link library---is to be 31819generated, enabling the selection of the required runtime 31820startup object and entry point. 31821 31822@item -mnop-fun-dllimport 31823@opindex mnop-fun-dllimport 31824This option is available for Cygwin and MinGW targets. It 31825specifies that the @code{dllimport} attribute should be ignored. 31826 31827@item -mthread 31828@opindex mthread 31829This option is available for MinGW targets. It specifies 31830that MinGW-specific thread support is to be used. 31831 31832@item -municode 31833@opindex municode 31834This option is available for MinGW-w64 targets. It causes 31835the @code{UNICODE} preprocessor macro to be predefined, and 31836chooses Unicode-capable runtime startup code. 31837 31838@item -mwin32 31839@opindex mwin32 31840This option is available for Cygwin and MinGW targets. It 31841specifies that the typical Microsoft Windows predefined macros are to 31842be set in the pre-processor, but does not influence the choice 31843of runtime library/startup code. 31844 31845@item -mwindows 31846@opindex mwindows 31847This option is available for Cygwin and MinGW targets. It 31848specifies that a GUI application is to be generated by 31849instructing the linker to set the PE header subsystem type 31850appropriately. 31851 31852@item -fno-set-stack-executable 31853@opindex fno-set-stack-executable 31854@opindex fset-stack-executable 31855This option is available for MinGW targets. It specifies that 31856the executable flag for the stack used by nested functions isn't 31857set. This is necessary for binaries running in kernel mode of 31858Microsoft Windows, as there the User32 API, which is used to set executable 31859privileges, isn't available. 31860 31861@item -fwritable-relocated-rdata 31862@opindex fno-writable-relocated-rdata 31863@opindex fwritable-relocated-rdata 31864This option is available for MinGW and Cygwin targets. It specifies 31865that relocated-data in read-only section is put into the @code{.data} 31866section. This is a necessary for older runtimes not supporting 31867modification of @code{.rdata} sections for pseudo-relocation. 31868 31869@item -mpe-aligned-commons 31870@opindex mpe-aligned-commons 31871This option is available for Cygwin and MinGW targets. It 31872specifies that the GNU extension to the PE file format that 31873permits the correct alignment of COMMON variables should be 31874used when generating code. It is enabled by default if 31875GCC detects that the target assembler found during configuration 31876supports the feature. 31877@end table 31878 31879See also under @ref{x86 Options} for standard options. 31880 31881@node Xstormy16 Options 31882@subsection Xstormy16 Options 31883@cindex Xstormy16 Options 31884 31885These options are defined for Xstormy16: 31886 31887@table @gcctabopt 31888@item -msim 31889@opindex msim 31890Choose startup files and linker script suitable for the simulator. 31891@end table 31892 31893@node Xtensa Options 31894@subsection Xtensa Options 31895@cindex Xtensa Options 31896 31897These options are supported for Xtensa targets: 31898 31899@table @gcctabopt 31900@item -mconst16 31901@itemx -mno-const16 31902@opindex mconst16 31903@opindex mno-const16 31904Enable or disable use of @code{CONST16} instructions for loading 31905constant values. The @code{CONST16} instruction is currently not a 31906standard option from Tensilica. When enabled, @code{CONST16} 31907instructions are always used in place of the standard @code{L32R} 31908instructions. The use of @code{CONST16} is enabled by default only if 31909the @code{L32R} instruction is not available. 31910 31911@item -mfused-madd 31912@itemx -mno-fused-madd 31913@opindex mfused-madd 31914@opindex mno-fused-madd 31915Enable or disable use of fused multiply/add and multiply/subtract 31916instructions in the floating-point option. This has no effect if the 31917floating-point option is not also enabled. Disabling fused multiply/add 31918and multiply/subtract instructions forces the compiler to use separate 31919instructions for the multiply and add/subtract operations. This may be 31920desirable in some cases where strict IEEE 754-compliant results are 31921required: the fused multiply add/subtract instructions do not round the 31922intermediate result, thereby producing results with @emph{more} bits of 31923precision than specified by the IEEE standard. Disabling fused multiply 31924add/subtract instructions also ensures that the program output is not 31925sensitive to the compiler's ability to combine multiply and add/subtract 31926operations. 31927 31928@item -mserialize-volatile 31929@itemx -mno-serialize-volatile 31930@opindex mserialize-volatile 31931@opindex mno-serialize-volatile 31932When this option is enabled, GCC inserts @code{MEMW} instructions before 31933@code{volatile} memory references to guarantee sequential consistency. 31934The default is @option{-mserialize-volatile}. Use 31935@option{-mno-serialize-volatile} to omit the @code{MEMW} instructions. 31936 31937@item -mforce-no-pic 31938@opindex mforce-no-pic 31939For targets, like GNU/Linux, where all user-mode Xtensa code must be 31940position-independent code (PIC), this option disables PIC for compiling 31941kernel code. 31942 31943@item -mtext-section-literals 31944@itemx -mno-text-section-literals 31945@opindex mtext-section-literals 31946@opindex mno-text-section-literals 31947These options control the treatment of literal pools. The default is 31948@option{-mno-text-section-literals}, which places literals in a separate 31949section in the output file. This allows the literal pool to be placed 31950in a data RAM/ROM, and it also allows the linker to combine literal 31951pools from separate object files to remove redundant literals and 31952improve code size. With @option{-mtext-section-literals}, the literals 31953are interspersed in the text section in order to keep them as close as 31954possible to their references. This may be necessary for large assembly 31955files. Literals for each function are placed right before that function. 31956 31957@item -mauto-litpools 31958@itemx -mno-auto-litpools 31959@opindex mauto-litpools 31960@opindex mno-auto-litpools 31961These options control the treatment of literal pools. The default is 31962@option{-mno-auto-litpools}, which places literals in a separate 31963section in the output file unless @option{-mtext-section-literals} is 31964used. With @option{-mauto-litpools} the literals are interspersed in 31965the text section by the assembler. Compiler does not produce explicit 31966@code{.literal} directives and loads literals into registers with 31967@code{MOVI} instructions instead of @code{L32R} to let the assembler 31968do relaxation and place literals as necessary. This option allows 31969assembler to create several literal pools per function and assemble 31970very big functions, which may not be possible with 31971@option{-mtext-section-literals}. 31972 31973@item -mtarget-align 31974@itemx -mno-target-align 31975@opindex mtarget-align 31976@opindex mno-target-align 31977When this option is enabled, GCC instructs the assembler to 31978automatically align instructions to reduce branch penalties at the 31979expense of some code density. The assembler attempts to widen density 31980instructions to align branch targets and the instructions following call 31981instructions. If there are not enough preceding safe density 31982instructions to align a target, no widening is performed. The 31983default is @option{-mtarget-align}. These options do not affect the 31984treatment of auto-aligned instructions like @code{LOOP}, which the 31985assembler always aligns, either by widening density instructions or 31986by inserting NOP instructions. 31987 31988@item -mlongcalls 31989@itemx -mno-longcalls 31990@opindex mlongcalls 31991@opindex mno-longcalls 31992When this option is enabled, GCC instructs the assembler to translate 31993direct calls to indirect calls unless it can determine that the target 31994of a direct call is in the range allowed by the call instruction. This 31995translation typically occurs for calls to functions in other source 31996files. Specifically, the assembler translates a direct @code{CALL} 31997instruction into an @code{L32R} followed by a @code{CALLX} instruction. 31998The default is @option{-mno-longcalls}. This option should be used in 31999programs where the call target can potentially be out of range. This 32000option is implemented in the assembler, not the compiler, so the 32001assembly code generated by GCC still shows direct call 32002instructions---look at the disassembled object code to see the actual 32003instructions. Note that the assembler uses an indirect call for 32004every cross-file call, not just those that really are out of range. 32005 32006@item -mabi=@var{name} 32007@opindex mabi 32008Generate code for the specified ABI@. Permissible values are: @samp{call0}, 32009@samp{windowed}. Default ABI is chosen by the Xtensa core configuration. 32010 32011@item -mabi=call0 32012@opindex mabi=call0 32013When this option is enabled function parameters are passed in registers 32014@code{a2} through @code{a7}, registers @code{a12} through @code{a15} are 32015caller-saved, and register @code{a15} may be used as a frame pointer. 32016When this version of the ABI is enabled the C preprocessor symbol 32017@code{__XTENSA_CALL0_ABI__} is defined. 32018 32019@item -mabi=windowed 32020@opindex mabi=windowed 32021When this option is enabled function parameters are passed in registers 32022@code{a10} through @code{a15}, and called function rotates register window 32023by 8 registers on entry so that its arguments are found in registers 32024@code{a2} through @code{a7}. Register @code{a7} may be used as a frame 32025pointer. Register window is rotated 8 registers back upon return. 32026When this version of the ABI is enabled the C preprocessor symbol 32027@code{__XTENSA_WINDOWED_ABI__} is defined. 32028@end table 32029 32030@node zSeries Options 32031@subsection zSeries Options 32032@cindex zSeries options 32033 32034These are listed under @xref{S/390 and zSeries Options}. 32035 32036 32037@c man end 32038 32039@node Spec Files 32040@section Specifying Subprocesses and the Switches to Pass to Them 32041@cindex Spec Files 32042 32043@command{gcc} is a driver program. It performs its job by invoking a 32044sequence of other programs to do the work of compiling, assembling and 32045linking. GCC interprets its command-line parameters and uses these to 32046deduce which programs it should invoke, and which command-line options 32047it ought to place on their command lines. This behavior is controlled 32048by @dfn{spec strings}. In most cases there is one spec string for each 32049program that GCC can invoke, but a few programs have multiple spec 32050strings to control their behavior. The spec strings built into GCC can 32051be overridden by using the @option{-specs=} command-line switch to specify 32052a spec file. 32053 32054@dfn{Spec files} are plain-text files that are used to construct spec 32055strings. They consist of a sequence of directives separated by blank 32056lines. The type of directive is determined by the first non-whitespace 32057character on the line, which can be one of the following: 32058 32059@table @code 32060@item %@var{command} 32061Issues a @var{command} to the spec file processor. The commands that can 32062appear here are: 32063 32064@table @code 32065@item %include <@var{file}> 32066@cindex @code{%include} 32067Search for @var{file} and insert its text at the current point in the 32068specs file. 32069 32070@item %include_noerr <@var{file}> 32071@cindex @code{%include_noerr} 32072Just like @samp{%include}, but do not generate an error message if the include 32073file cannot be found. 32074 32075@item %rename @var{old_name} @var{new_name} 32076@cindex @code{%rename} 32077Rename the spec string @var{old_name} to @var{new_name}. 32078 32079@end table 32080 32081@item *[@var{spec_name}]: 32082This tells the compiler to create, override or delete the named spec 32083string. All lines after this directive up to the next directive or 32084blank line are considered to be the text for the spec string. If this 32085results in an empty string then the spec is deleted. (Or, if the 32086spec did not exist, then nothing happens.) Otherwise, if the spec 32087does not currently exist a new spec is created. If the spec does 32088exist then its contents are overridden by the text of this 32089directive, unless the first character of that text is the @samp{+} 32090character, in which case the text is appended to the spec. 32091 32092@item [@var{suffix}]: 32093Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive 32094and up to the next directive or blank line are considered to make up the 32095spec string for the indicated suffix. When the compiler encounters an 32096input file with the named suffix, it processes the spec string in 32097order to work out how to compile that file. For example: 32098 32099@smallexample 32100.ZZ: 32101z-compile -input %i 32102@end smallexample 32103 32104This says that any input file whose name ends in @samp{.ZZ} should be 32105passed to the program @samp{z-compile}, which should be invoked with the 32106command-line switch @option{-input} and with the result of performing the 32107@samp{%i} substitution. (See below.) 32108 32109As an alternative to providing a spec string, the text following a 32110suffix directive can be one of the following: 32111 32112@table @code 32113@item @@@var{language} 32114This says that the suffix is an alias for a known @var{language}. This is 32115similar to using the @option{-x} command-line switch to GCC to specify a 32116language explicitly. For example: 32117 32118@smallexample 32119.ZZ: 32120@@c++ 32121@end smallexample 32122 32123Says that .ZZ files are, in fact, C++ source files. 32124 32125@item #@var{name} 32126This causes an error messages saying: 32127 32128@smallexample 32129@var{name} compiler not installed on this system. 32130@end smallexample 32131@end table 32132 32133GCC already has an extensive list of suffixes built into it. 32134This directive adds an entry to the end of the list of suffixes, but 32135since the list is searched from the end backwards, it is effectively 32136possible to override earlier entries using this technique. 32137 32138@end table 32139 32140GCC has the following spec strings built into it. Spec files can 32141override these strings or create their own. Note that individual 32142targets can also add their own spec strings to this list. 32143 32144@smallexample 32145asm Options to pass to the assembler 32146asm_final Options to pass to the assembler post-processor 32147cpp Options to pass to the C preprocessor 32148cc1 Options to pass to the C compiler 32149cc1plus Options to pass to the C++ compiler 32150endfile Object files to include at the end of the link 32151link Options to pass to the linker 32152lib Libraries to include on the command line to the linker 32153libgcc Decides which GCC support library to pass to the linker 32154linker Sets the name of the linker 32155predefines Defines to be passed to the C preprocessor 32156signed_char Defines to pass to CPP to say whether @code{char} is signed 32157 by default 32158startfile Object files to include at the start of the link 32159@end smallexample 32160 32161Here is a small example of a spec file: 32162 32163@smallexample 32164%rename lib old_lib 32165 32166*lib: 32167--start-group -lgcc -lc -leval1 --end-group %(old_lib) 32168@end smallexample 32169 32170This example renames the spec called @samp{lib} to @samp{old_lib} and 32171then overrides the previous definition of @samp{lib} with a new one. 32172The new definition adds in some extra command-line options before 32173including the text of the old definition. 32174 32175@dfn{Spec strings} are a list of command-line options to be passed to their 32176corresponding program. In addition, the spec strings can contain 32177@samp{%}-prefixed sequences to substitute variable text or to 32178conditionally insert text into the command line. Using these constructs 32179it is possible to generate quite complex command lines. 32180 32181Here is a table of all defined @samp{%}-sequences for spec 32182strings. Note that spaces are not generated automatically around the 32183results of expanding these sequences. Therefore you can concatenate them 32184together or combine them with constant text in a single argument. 32185 32186@table @code 32187@item %% 32188Substitute one @samp{%} into the program name or argument. 32189 32190@item %" 32191Substitute an empty argument. 32192 32193@item %i 32194Substitute the name of the input file being processed. 32195 32196@item %b 32197Substitute the basename for outputs related with the input file being 32198processed. This is often the substring up to (and not including) the 32199last period and not including the directory but, unless %w is active, it 32200expands to the basename for auxiliary outputs, which may be influenced 32201by an explicit output name, and by various other options that control 32202how auxiliary outputs are named. 32203 32204@item %B 32205This is the same as @samp{%b}, but include the file suffix (text after 32206the last period). Without %w, it expands to the basename for dump 32207outputs. 32208 32209@item %d 32210Marks the argument containing or following the @samp{%d} as a 32211temporary file name, so that that file is deleted if GCC exits 32212successfully. Unlike @samp{%g}, this contributes no text to the 32213argument. 32214 32215@item %g@var{suffix} 32216Substitute a file name that has suffix @var{suffix} and is chosen 32217once per compilation, and mark the argument in the same way as 32218@samp{%d}. To reduce exposure to denial-of-service attacks, the file 32219name is now chosen in a way that is hard to predict even when previously 32220chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s} 32221might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches 32222the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is 32223treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g} 32224was simply substituted with a file name chosen once per compilation, 32225without regard to any appended suffix (which was therefore treated 32226just like ordinary text), making such attacks more likely to succeed. 32227 32228@item %u@var{suffix} 32229Like @samp{%g}, but generates a new temporary file name 32230each time it appears instead of once per compilation. 32231 32232@item %U@var{suffix} 32233Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a 32234new one if there is no such last file name. In the absence of any 32235@samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share 32236the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s} 32237involves the generation of two distinct file names, one 32238for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was 32239simply substituted with a file name chosen for the previous @samp{%u}, 32240without regard to any appended suffix. 32241 32242@item %j@var{suffix} 32243Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is 32244writable, and if @option{-save-temps} is not used; 32245otherwise, substitute the name 32246of a temporary file, just like @samp{%u}. This temporary file is not 32247meant for communication between processes, but rather as a junk 32248disposal mechanism. 32249 32250@item %|@var{suffix} 32251@itemx %m@var{suffix} 32252Like @samp{%g}, except if @option{-pipe} is in effect. In that case 32253@samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at 32254all. These are the two most common ways to instruct a program that it 32255should read from standard input or write to standard output. If you 32256need something more elaborate you can use an @samp{%@{pipe:@code{X}@}} 32257construct: see for example @file{gcc/fortran/lang-specs.h}. 32258 32259@item %.@var{SUFFIX} 32260Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args 32261when it is subsequently output with @samp{%*}. @var{SUFFIX} is 32262terminated by the next space or %. 32263 32264@item %w 32265Marks the argument containing or following the @samp{%w} as the 32266designated output file of this compilation. This puts the argument 32267into the sequence of arguments that @samp{%o} substitutes. 32268 32269@item %V 32270Indicates that this compilation produces no output file. 32271 32272@item %o 32273Substitutes the names of all the output files, with spaces 32274automatically placed around them. You should write spaces 32275around the @samp{%o} as well or the results are undefined. 32276@samp{%o} is for use in the specs for running the linker. 32277Input files whose names have no recognized suffix are not compiled 32278at all, but they are included among the output files, so they are 32279linked. 32280 32281@item %O 32282Substitutes the suffix for object files. Note that this is 32283handled specially when it immediately follows @samp{%g, %u, or %U}, 32284because of the need for those to form complete file names. The 32285handling is such that @samp{%O} is treated exactly as if it had already 32286been substituted, except that @samp{%g, %u, and %U} do not currently 32287support additional @var{suffix} characters following @samp{%O} as they do 32288following, for example, @samp{.o}. 32289 32290@item %I 32291Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}), 32292@option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), 32293@option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options) 32294and @option{-imultilib} as necessary. 32295 32296@item %s 32297Current argument is the name of a library or startup file of some sort. 32298Search for that file in a standard list of directories and substitute 32299the full name found. The current working directory is included in the 32300list of directories scanned. 32301 32302@item %T 32303Current argument is the name of a linker script. Search for that file 32304in the current list of directories to scan for libraries. If the file 32305is located insert a @option{--script} option into the command line 32306followed by the full path name found. If the file is not found then 32307generate an error message. Note: the current working directory is not 32308searched. 32309 32310@item %e@var{str} 32311Print @var{str} as an error message. @var{str} is terminated by a newline. 32312Use this when inconsistent options are detected. 32313 32314@item %n@var{str} 32315Print @var{str} as a notice. @var{str} is terminated by a newline. 32316 32317@item %(@var{name}) 32318Substitute the contents of spec string @var{name} at this point. 32319 32320@item %x@{@var{option}@} 32321Accumulate an option for @samp{%X}. 32322 32323@item %X 32324Output the accumulated linker options specified by @option{-Wl} or a @samp{%x} 32325spec string. 32326 32327@item %Y 32328Output the accumulated assembler options specified by @option{-Wa}. 32329 32330@item %Z 32331Output the accumulated preprocessor options specified by @option{-Wp}. 32332 32333@item %M 32334Output @code{multilib_os_dir}. 32335 32336@item %R 32337Output the concatenation of @code{target_system_root} and @code{target_sysroot_suffix}. 32338 32339@item %a 32340Process the @code{asm} spec. This is used to compute the 32341switches to be passed to the assembler. 32342 32343@item %A 32344Process the @code{asm_final} spec. This is a spec string for 32345passing switches to an assembler post-processor, if such a program is 32346needed. 32347 32348@item %l 32349Process the @code{link} spec. This is the spec for computing the 32350command line passed to the linker. Typically it makes use of the 32351@samp{%L %G %S %D and %E} sequences. 32352 32353@item %D 32354Dump out a @option{-L} option for each directory that GCC believes might 32355contain startup files. If the target supports multilibs then the 32356current multilib directory is prepended to each of these paths. 32357 32358@item %L 32359Process the @code{lib} spec. This is a spec string for deciding which 32360libraries are included on the command line to the linker. 32361 32362@item %G 32363Process the @code{libgcc} spec. This is a spec string for deciding 32364which GCC support library is included on the command line to the linker. 32365 32366@item %S 32367Process the @code{startfile} spec. This is a spec for deciding which 32368object files are the first ones passed to the linker. Typically 32369this might be a file named @file{crt0.o}. 32370 32371@item %E 32372Process the @code{endfile} spec. This is a spec string that specifies 32373the last object files that are passed to the linker. 32374 32375@item %C 32376Process the @code{cpp} spec. This is used to construct the arguments 32377to be passed to the C preprocessor. 32378 32379@item %1 32380Process the @code{cc1} spec. This is used to construct the options to be 32381passed to the actual C compiler (@command{cc1}). 32382 32383@item %2 32384Process the @code{cc1plus} spec. This is used to construct the options to be 32385passed to the actual C++ compiler (@command{cc1plus}). 32386 32387@item %* 32388Substitute the variable part of a matched option. See below. 32389Note that each comma in the substituted string is replaced by 32390a single space. 32391 32392@item %<S 32393Remove all occurrences of @code{-S} from the command line. Note---this 32394command is position dependent. @samp{%} commands in the spec string 32395before this one see @code{-S}, @samp{%} commands in the spec string 32396after this one do not. 32397 32398@item %<S* 32399Similar to @samp{%<S}, but match all switches beginning with @code{-S}. 32400 32401@item %>S 32402Similar to @samp{%<S}, but keep @code{-S} in the GCC command line. 32403 32404@item %:@var{function}(@var{args}) 32405Call the named function @var{function}, passing it @var{args}. 32406@var{args} is first processed as a nested spec string, then split 32407into an argument vector in the usual fashion. The function returns 32408a string which is processed as if it had appeared literally as part 32409of the current spec. 32410 32411The following built-in spec functions are provided: 32412 32413@table @code 32414@item @code{getenv} 32415The @code{getenv} spec function takes two arguments: an environment 32416variable name and a string. If the environment variable is not 32417defined, a fatal error is issued. Otherwise, the return value is the 32418value of the environment variable concatenated with the string. For 32419example, if @env{TOPDIR} is defined as @file{/path/to/top}, then: 32420 32421@smallexample 32422%:getenv(TOPDIR /include) 32423@end smallexample 32424 32425expands to @file{/path/to/top/include}. 32426 32427@item @code{if-exists} 32428The @code{if-exists} spec function takes one argument, an absolute 32429pathname to a file. If the file exists, @code{if-exists} returns the 32430pathname. Here is a small example of its usage: 32431 32432@smallexample 32433*startfile: 32434crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s 32435@end smallexample 32436 32437@item @code{if-exists-else} 32438The @code{if-exists-else} spec function is similar to the @code{if-exists} 32439spec function, except that it takes two arguments. The first argument is 32440an absolute pathname to a file. If the file exists, @code{if-exists-else} 32441returns the pathname. If it does not exist, it returns the second argument. 32442This way, @code{if-exists-else} can be used to select one file or another, 32443based on the existence of the first. Here is a small example of its usage: 32444 32445@smallexample 32446*startfile: 32447crt0%O%s %:if-exists(crti%O%s) \ 32448%:if-exists-else(crtbeginT%O%s crtbegin%O%s) 32449@end smallexample 32450 32451@item @code{if-exists-then-else} 32452The @code{if-exists-then-else} spec function takes at least two arguments 32453and an optional third one. The first argument is an absolute pathname to a 32454file. If the file exists, the function returns the second argument. 32455If the file does not exist, the function returns the third argument if there 32456is one, or NULL otherwise. This can be used to expand one text, or optionally 32457another, based on the existence of a file. Here is a small example of its 32458usage: 32459 32460@smallexample 32461-l%:if-exists-then-else(%:getenv(VSB_DIR rtnet.h) rtnet net) 32462@end smallexample 32463 32464@item @code{sanitize} 32465The @code{sanitize} spec function takes no arguments. It returns non-NULL if 32466any address, thread or undefined behavior sanitizers are active. 32467 32468@smallexample 32469%@{%:sanitize(address):-funwind-tables@} 32470@end smallexample 32471 32472@item @code{replace-outfile} 32473The @code{replace-outfile} spec function takes two arguments. It looks for the 32474first argument in the outfiles array and replaces it with the second argument. Here 32475is a small example of its usage: 32476 32477@smallexample 32478%@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@} 32479@end smallexample 32480 32481@item @code{remove-outfile} 32482The @code{remove-outfile} spec function takes one argument. It looks for the 32483first argument in the outfiles array and removes it. Here is a small example 32484its usage: 32485 32486@smallexample 32487%:remove-outfile(-lm) 32488@end smallexample 32489 32490@item @code{version-compare} 32491The @code{version-compare} spec function takes four or five arguments of the following 32492form: 32493 32494@smallexample 32495<comparison-op> <arg1> [<arg2>] <switch> <result> 32496@end smallexample 32497 32498It returns @code{result} if the comparison evaluates to true, and NULL if it doesn't. 32499The supported @code{comparison-op} values are: 32500 32501@table @code 32502@item >= 32503True if @code{switch} is a later (or same) version than @code{arg1} 32504 32505@item !> 32506Opposite of @code{>=} 32507 32508@item < 32509True if @code{switch} is an earlier version than @code{arg1} 32510 32511@item !< 32512Opposite of @code{<} 32513 32514@item >< 32515True if @code{switch} is @code{arg1} or later, and earlier than @code{arg2} 32516 32517@item <> 32518True if @code{switch} is earlier than @code{arg1}, or is @code{arg2} or later 32519@end table 32520 32521If the @code{switch} is not present at all, the condition is false unless the first character 32522of the @code{comparison-op} is @code{!}. 32523 32524@smallexample 32525%:version-compare(>= 10.3 mmacosx-version-min= -lmx) 32526@end smallexample 32527 32528The above example would add @option{-lmx} if @option{-mmacosx-version-min=10.3.9} was 32529passed. 32530 32531@item @code{include} 32532The @code{include} spec function behaves much like @code{%include}, with the advantage 32533that it can be nested inside a spec and thus be conditionalized. It takes one argument, 32534the filename, and looks for it in the startfile path. It always returns NULL. 32535 32536@smallexample 32537%@{static-libasan|static:%:include(libsanitizer.spec)%(link_libasan)@} 32538@end smallexample 32539 32540@item @code{pass-through-libs} 32541The @code{pass-through-libs} spec function takes any number of arguments. It 32542finds any @option{-l} options and any non-options ending in @file{.a} (which it 32543assumes are the names of linker input library archive files) and returns a 32544result containing all the found arguments each prepended by 32545@option{-plugin-opt=-pass-through=} and joined by spaces. This list is 32546intended to be passed to the LTO linker plugin. 32547 32548@smallexample 32549%:pass-through-libs(%G %L %G) 32550@end smallexample 32551 32552@item @code{print-asm-header} 32553The @code{print-asm-header} function takes no arguments and simply 32554prints a banner like: 32555 32556@smallexample 32557Assembler options 32558================= 32559 32560Use "-Wa,OPTION" to pass "OPTION" to the assembler. 32561@end smallexample 32562 32563It is used to separate compiler options from assembler options 32564in the @option{--target-help} output. 32565 32566@item @code{gt} 32567The @code{gt} spec function takes two or more arguments. It returns @code{""} (the 32568empty string) if the second-to-last argument is greater than the last argument, and NULL 32569otherwise. The following example inserts the @code{link_gomp} spec if the last 32570@option{-ftree-parallelize-loops=} option given on the command line is greater than 1: 32571 32572@smallexample 32573%@{%:gt(%@{ftree-parallelize-loops=*:%*@} 1):%:include(libgomp.spec)%(link_gomp)@} 32574@end smallexample 32575 32576@item @code{debug-level-gt} 32577The @code{debug-level-gt} spec function takes one argument and returns @code{""} (the 32578empty string) if @code{debug_info_level} is greater than the specified number, and NULL 32579otherwise. 32580 32581@smallexample 32582%@{%:debug-level-gt(0):%@{gdwarf*:--gdwarf2@}@} 32583@end smallexample 32584@end table 32585 32586@item %@{S@} 32587Substitutes the @code{-S} switch, if that switch is given to GCC@. 32588If that switch is not specified, this substitutes nothing. Note that 32589the leading dash is omitted when specifying this option, and it is 32590automatically inserted if the substitution is performed. Thus the spec 32591string @samp{%@{foo@}} matches the command-line option @option{-foo} 32592and outputs the command-line option @option{-foo}. 32593 32594@item %W@{S@} 32595Like %@{@code{S}@} but mark last argument supplied within as a file to be 32596deleted on failure. 32597 32598@item %@@@{S@} 32599Like %@{@code{S}@} but puts the result into a @code{FILE} and substitutes 32600@code{@@FILE} if an @code{@@file} argument has been supplied. 32601 32602@item %@{S*@} 32603Substitutes all the switches specified to GCC whose names start 32604with @code{-S}, but which also take an argument. This is used for 32605switches like @option{-o}, @option{-D}, @option{-I}, etc. 32606GCC considers @option{-o foo} as being 32607one switch whose name starts with @samp{o}. %@{o*@} substitutes this 32608text, including the space. Thus two arguments are generated. 32609 32610@item %@{S*&T*@} 32611Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options 32612(the order of @code{S} and @code{T} in the spec is not significant). 32613There can be any number of ampersand-separated variables; for each the 32614wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}. 32615 32616@item %@{S:X@} 32617Substitutes @code{X}, if the @option{-S} switch is given to GCC@. 32618 32619@item %@{!S:X@} 32620Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@. 32621 32622@item %@{S*:X@} 32623Substitutes @code{X} if one or more switches whose names start with 32624@code{-S} are specified to GCC@. Normally @code{X} is substituted only 32625once, no matter how many such switches appeared. However, if @code{%*} 32626appears somewhere in @code{X}, then @code{X} is substituted once 32627for each matching switch, with the @code{%*} replaced by the part of 32628that switch matching the @code{*}. 32629 32630If @code{%*} appears as the last part of a spec sequence then a space 32631is added after the end of the last substitution. If there is more 32632text in the sequence, however, then a space is not generated. This 32633allows the @code{%*} substitution to be used as part of a larger 32634string. For example, a spec string like this: 32635 32636@smallexample 32637%@{mcu=*:--script=%*/memory.ld@} 32638@end smallexample 32639 32640@noindent 32641when matching an option like @option{-mcu=newchip} produces: 32642 32643@smallexample 32644--script=newchip/memory.ld 32645@end smallexample 32646 32647@item %@{.S:X@} 32648Substitutes @code{X}, if processing a file with suffix @code{S}. 32649 32650@item %@{!.S:X@} 32651Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}. 32652 32653@item %@{,S:X@} 32654Substitutes @code{X}, if processing a file for language @code{S}. 32655 32656@item %@{!,S:X@} 32657Substitutes @code{X}, if not processing a file for language @code{S}. 32658 32659@item %@{S|P:X@} 32660Substitutes @code{X} if either @code{-S} or @code{-P} is given to 32661GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and 32662@code{*} sequences as well, although they have a stronger binding than 32663the @samp{|}. If @code{%*} appears in @code{X}, all of the 32664alternatives must be starred, and only the first matching alternative 32665is substituted. 32666 32667For example, a spec string like this: 32668 32669@smallexample 32670%@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@} 32671@end smallexample 32672 32673@noindent 32674outputs the following command-line options from the following input 32675command-line options: 32676 32677@smallexample 32678fred.c -foo -baz 32679jim.d -bar -boggle 32680-d fred.c -foo -baz -boggle 32681-d jim.d -bar -baz -boggle 32682@end smallexample 32683 32684@item %@{%:@var{function}(@var{args}):X@} 32685 32686Call function named @var{function} with args @var{args}. If the 32687function returns non-NULL, then @code{X} is substituted, if it returns 32688NULL, it isn't substituted. 32689 32690@item %@{S:X; T:Y; :D@} 32691 32692If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is 32693given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can 32694be as many clauses as you need. This may be combined with @code{.}, 32695@code{,}, @code{!}, @code{|}, and @code{*} as needed. 32696 32697 32698@end table 32699 32700The switch matching text @code{S} in a @samp{%@{S@}}, @samp{%@{S:X@}} 32701or similar construct can use a backslash to ignore the special meaning 32702of the character following it, thus allowing literal matching of a 32703character that is otherwise specially treated. For example, 32704@samp{%@{std=iso9899\:1999:X@}} substitutes @code{X} if the 32705@option{-std=iso9899:1999} option is given. 32706 32707The conditional text @code{X} in a @samp{%@{S:X@}} or similar 32708construct may contain other nested @samp{%} constructs or spaces, or 32709even newlines. They are processed as usual, as described above. 32710Trailing white space in @code{X} is ignored. White space may also 32711appear anywhere on the left side of the colon in these constructs, 32712except between @code{.} or @code{*} and the corresponding word. 32713 32714The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are 32715handled specifically in these constructs. If another value of 32716@option{-O} or the negated form of a @option{-f}, @option{-m}, or 32717@option{-W} switch is found later in the command line, the earlier 32718switch value is ignored, except with @{@code{S}*@} where @code{S} is 32719just one letter, which passes all matching options. 32720 32721The character @samp{|} at the beginning of the predicate text is used to 32722indicate that a command should be piped to the following command, but 32723only if @option{-pipe} is specified. 32724 32725It is built into GCC which switches take arguments and which do not. 32726(You might think it would be useful to generalize this to allow each 32727compiler's spec to say which switches take arguments. But this cannot 32728be done in a consistent fashion. GCC cannot even decide which input 32729files have been specified without knowing which switches take arguments, 32730and it must know which input files to compile in order to tell which 32731compilers to run). 32732 32733GCC also knows implicitly that arguments starting in @option{-l} are to be 32734treated as compiler output files, and passed to the linker in their 32735proper position among the other output files. 32736 32737@node Environment Variables 32738@section Environment Variables Affecting GCC 32739@cindex environment variables 32740 32741@c man begin ENVIRONMENT 32742This section describes several environment variables that affect how GCC 32743operates. Some of them work by specifying directories or prefixes to use 32744when searching for various kinds of files. Some are used to specify other 32745aspects of the compilation environment. 32746 32747Note that you can also specify places to search using options such as 32748@option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These 32749take precedence over places specified using environment variables, which 32750in turn take precedence over those specified by the configuration of GCC@. 32751@xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint, 32752GNU Compiler Collection (GCC) Internals}. 32753 32754@table @env 32755@item LANG 32756@itemx LC_CTYPE 32757@c @itemx LC_COLLATE 32758@itemx LC_MESSAGES 32759@c @itemx LC_MONETARY 32760@c @itemx LC_NUMERIC 32761@c @itemx LC_TIME 32762@itemx LC_ALL 32763@findex LANG 32764@findex LC_CTYPE 32765@c @findex LC_COLLATE 32766@findex LC_MESSAGES 32767@c @findex LC_MONETARY 32768@c @findex LC_NUMERIC 32769@c @findex LC_TIME 32770@findex LC_ALL 32771@cindex locale 32772These environment variables control the way that GCC uses 32773localization information which allows GCC to work with different 32774national conventions. GCC inspects the locale categories 32775@env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do 32776so. These locale categories can be set to any value supported by your 32777installation. A typical value is @samp{en_GB.UTF-8} for English in the United 32778Kingdom encoded in UTF-8. 32779 32780The @env{LC_CTYPE} environment variable specifies character 32781classification. GCC uses it to determine the character boundaries in 32782a string; this is needed for some multibyte encodings that contain quote 32783and escape characters that are otherwise interpreted as a string 32784end or escape. 32785 32786The @env{LC_MESSAGES} environment variable specifies the language to 32787use in diagnostic messages. 32788 32789If the @env{LC_ALL} environment variable is set, it overrides the value 32790of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE} 32791and @env{LC_MESSAGES} default to the value of the @env{LANG} 32792environment variable. If none of these variables are set, GCC 32793defaults to traditional C English behavior. 32794 32795@item TMPDIR 32796@findex TMPDIR 32797If @env{TMPDIR} is set, it specifies the directory to use for temporary 32798files. GCC uses temporary files to hold the output of one stage of 32799compilation which is to be used as input to the next stage: for example, 32800the output of the preprocessor, which is the input to the compiler 32801proper. 32802 32803@item GCC_COMPARE_DEBUG 32804@findex GCC_COMPARE_DEBUG 32805Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing 32806@option{-fcompare-debug} to the compiler driver. See the documentation 32807of this option for more details. 32808 32809@item GCC_EXEC_PREFIX 32810@findex GCC_EXEC_PREFIX 32811If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the 32812names of the subprograms executed by the compiler. No slash is added 32813when this prefix is combined with the name of a subprogram, but you can 32814specify a prefix that ends with a slash if you wish. 32815 32816If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out 32817an appropriate prefix to use based on the pathname it is invoked with. 32818 32819If GCC cannot find the subprogram using the specified prefix, it 32820tries looking in the usual places for the subprogram. 32821 32822The default value of @env{GCC_EXEC_PREFIX} is 32823@file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to 32824the installed compiler. In many cases @var{prefix} is the value 32825of @code{prefix} when you ran the @file{configure} script. 32826 32827Other prefixes specified with @option{-B} take precedence over this prefix. 32828 32829This prefix is also used for finding files such as @file{crt0.o} that are 32830used for linking. 32831 32832In addition, the prefix is used in an unusual way in finding the 32833directories to search for header files. For each of the standard 32834directories whose name normally begins with @samp{/usr/local/lib/gcc} 32835(more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries 32836replacing that beginning with the specified prefix to produce an 32837alternate directory name. Thus, with @option{-Bfoo/}, GCC searches 32838@file{foo/bar} just before it searches the standard directory 32839@file{/usr/local/lib/bar}. 32840If a standard directory begins with the configured 32841@var{prefix} then the value of @var{prefix} is replaced by 32842@env{GCC_EXEC_PREFIX} when looking for header files. 32843 32844@item COMPILER_PATH 32845@findex COMPILER_PATH 32846The value of @env{COMPILER_PATH} is a colon-separated list of 32847directories, much like @env{PATH}. GCC tries the directories thus 32848specified when searching for subprograms, if it cannot find the 32849subprograms using @env{GCC_EXEC_PREFIX}. 32850 32851@item LIBRARY_PATH 32852@findex LIBRARY_PATH 32853The value of @env{LIBRARY_PATH} is a colon-separated list of 32854directories, much like @env{PATH}. When configured as a native compiler, 32855GCC tries the directories thus specified when searching for special 32856linker files, if it cannot find them using @env{GCC_EXEC_PREFIX}. Linking 32857using GCC also uses these directories when searching for ordinary 32858libraries for the @option{-l} option (but directories specified with 32859@option{-L} come first). 32860 32861@item LANG 32862@findex LANG 32863@cindex locale definition 32864This variable is used to pass locale information to the compiler. One way in 32865which this information is used is to determine the character set to be used 32866when character literals, string literals and comments are parsed in C and C++. 32867When the compiler is configured to allow multibyte characters, 32868the following values for @env{LANG} are recognized: 32869 32870@table @samp 32871@item C-JIS 32872Recognize JIS characters. 32873@item C-SJIS 32874Recognize SJIS characters. 32875@item C-EUCJP 32876Recognize EUCJP characters. 32877@end table 32878 32879If @env{LANG} is not defined, or if it has some other value, then the 32880compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to 32881recognize and translate multibyte characters. 32882 32883@item GCC_EXTRA_DIAGNOSTIC_OUTPUT 32884@findex GCC_EXTRA_DIAGNOSTIC_OUTPUT 32885If @env{GCC_EXTRA_DIAGNOSTIC_OUTPUT} is set to one of the following values, 32886then additional text will be emitted to stderr when fix-it hints are 32887emitted. @option{-fdiagnostics-parseable-fixits} and 32888@option{-fno-diagnostics-parseable-fixits} take precedence over this 32889environment variable. 32890 32891@table @samp 32892@item fixits-v1 32893Emit parseable fix-it hints, equivalent to 32894@option{-fdiagnostics-parseable-fixits}. In particular, columns are 32895expressed as a count of bytes, starting at byte 1 for the initial column. 32896 32897@item fixits-v2 32898As @code{fixits-v1}, but columns are expressed as display columns, 32899as per @option{-fdiagnostics-column-unit=display}. 32900@end table 32901 32902@end table 32903 32904@noindent 32905Some additional environment variables affect the behavior of the 32906preprocessor. 32907 32908@include cppenv.texi 32909 32910@c man end 32911 32912@node Precompiled Headers 32913@section Using Precompiled Headers 32914@cindex precompiled headers 32915@cindex speed of compilation 32916 32917Often large projects have many header files that are included in every 32918source file. The time the compiler takes to process these header files 32919over and over again can account for nearly all of the time required to 32920build the project. To make builds faster, GCC allows you to 32921@dfn{precompile} a header file. 32922 32923To create a precompiled header file, simply compile it as you would any 32924other file, if necessary using the @option{-x} option to make the driver 32925treat it as a C or C++ header file. You may want to use a 32926tool like @command{make} to keep the precompiled header up-to-date when 32927the headers it contains change. 32928 32929A precompiled header file is searched for when @code{#include} is 32930seen in the compilation. As it searches for the included file 32931(@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the 32932compiler looks for a precompiled header in each directory just before it 32933looks for the include file in that directory. The name searched for is 32934the name specified in the @code{#include} with @samp{.gch} appended. If 32935the precompiled header file cannot be used, it is ignored. 32936 32937For instance, if you have @code{#include "all.h"}, and you have 32938@file{all.h.gch} in the same directory as @file{all.h}, then the 32939precompiled header file is used if possible, and the original 32940header is used otherwise. 32941 32942Alternatively, you might decide to put the precompiled header file in a 32943directory and use @option{-I} to ensure that directory is searched 32944before (or instead of) the directory containing the original header. 32945Then, if you want to check that the precompiled header file is always 32946used, you can put a file of the same name as the original header in this 32947directory containing an @code{#error} command. 32948 32949This also works with @option{-include}. So yet another way to use 32950precompiled headers, good for projects not designed with precompiled 32951header files in mind, is to simply take most of the header files used by 32952a project, include them from another header file, precompile that header 32953file, and @option{-include} the precompiled header. If the header files 32954have guards against multiple inclusion, they are skipped because 32955they've already been included (in the precompiled header). 32956 32957If you need to precompile the same header file for different 32958languages, targets, or compiler options, you can instead make a 32959@emph{directory} named like @file{all.h.gch}, and put each precompiled 32960header in the directory, perhaps using @option{-o}. It doesn't matter 32961what you call the files in the directory; every precompiled header in 32962the directory is considered. The first precompiled header 32963encountered in the directory that is valid for this compilation is 32964used; they're searched in no particular order. 32965 32966There are many other possibilities, limited only by your imagination, 32967good sense, and the constraints of your build system. 32968 32969A precompiled header file can be used only when these conditions apply: 32970 32971@itemize 32972@item 32973Only one precompiled header can be used in a particular compilation. 32974 32975@item 32976A precompiled header cannot be used once the first C token is seen. You 32977can have preprocessor directives before a precompiled header; you cannot 32978include a precompiled header from inside another header. 32979 32980@item 32981The precompiled header file must be produced for the same language as 32982the current compilation. You cannot use a C precompiled header for a C++ 32983compilation. 32984 32985@item 32986The precompiled header file must have been produced by the same compiler 32987binary as the current compilation is using. 32988 32989@item 32990Any macros defined before the precompiled header is included must 32991either be defined in the same way as when the precompiled header was 32992generated, or must not affect the precompiled header, which usually 32993means that they don't appear in the precompiled header at all. 32994 32995The @option{-D} option is one way to define a macro before a 32996precompiled header is included; using a @code{#define} can also do it. 32997There are also some options that define macros implicitly, like 32998@option{-O} and @option{-Wdeprecated}; the same rule applies to macros 32999defined this way. 33000 33001@item If debugging information is output when using the precompiled 33002header, using @option{-g} or similar, the same kind of debugging information 33003must have been output when building the precompiled header. However, 33004a precompiled header built using @option{-g} can be used in a compilation 33005when no debugging information is being output. 33006 33007@item The same @option{-m} options must generally be used when building 33008and using the precompiled header. @xref{Submodel Options}, 33009for any cases where this rule is relaxed. 33010 33011@item Each of the following options must be the same when building and using 33012the precompiled header: 33013 33014@gccoptlist{-fexceptions} 33015 33016@item 33017Some other command-line options starting with @option{-f}, 33018@option{-p}, or @option{-O} must be defined in the same way as when 33019the precompiled header was generated. At present, it's not clear 33020which options are safe to change and which are not; the safest choice 33021is to use exactly the same options when generating and using the 33022precompiled header. The following are known to be safe: 33023 33024@gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol 33025-fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol 33026-fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol 33027-pedantic-errors} 33028 33029@item Address space layout randomization (ASLR) can lead to not binary identical 33030PCH files. If you rely on stable PCH file contents disable ASLR when generating 33031PCH files. 33032 33033@end itemize 33034 33035For all of these except the last, the compiler automatically 33036ignores the precompiled header if the conditions aren't met. If you 33037find an option combination that doesn't work and doesn't cause the 33038precompiled header to be ignored, please consider filing a bug report, 33039see @ref{Bugs}. 33040 33041If you do use differing options when generating and using the 33042precompiled header, the actual behavior is a mixture of the 33043behavior for the options. For instance, if you use @option{-g} to 33044generate the precompiled header but not when using it, you may or may 33045not get debugging information for routines in the precompiled header. 33046 33047@node C++ Modules 33048@section C++ Modules 33049@cindex speed of compilation 33050 33051Modules are a C++20 language feature. As the name suggests, they 33052provides a modular compilation system, intending to provide both 33053faster builds and better library isolation. The ``Merging Modules'' 33054paper @uref{https://wg21.link/p1103}, provides the easiest to read set 33055of changes to the standard, although it does not capture later 33056changes. That specification is now part of C++20, 33057@uref{git@@github.com:cplusplus/draft.git}, it is considered complete 33058(there may be defect reports to come). 33059 33060@emph{G++'s modules support is not complete.} Other than bugs, the 33061known missing pieces are: 33062 33063@table @emph 33064 33065@item Private Module Fragment 33066The Private Module Fragment is recognized, but an error is emitted. 33067 33068@item Partition definition visibility rules 33069Entities may be defined in implementation partitions, and those 33070definitions are not available outside of the module. This is not 33071implemented, and the definitions are available to extra-module use. 33072 33073@item Textual merging of reachable GM entities 33074Entities may be multiply defined across different header-units. 33075These must be de-duplicated, and this is implemented across imports, 33076or when an import redefines a textually-defined entity. However the 33077reverse is not implemented---textually redefining an entity that has 33078been defined in an imported header-unit. A redefinition error is 33079emitted. 33080 33081@item Translation-Unit local referencing rules 33082Papers p1815 (@uref{https://wg21.link/p1815}) and p2003 33083(@uref{https://wg21.link/p2003}) add limitations on which entities an 33084exported region may reference (for instance, the entities an exported 33085template definition may reference). These are not fully implemented. 33086 33087@item Language-linkage module attachment 33088Declarations with explicit language linkage (@code{extern "C"} or 33089@code{extern "C++"}) are attached to the global module, even when in 33090the purview of a named module. This is not implemented. Such 33091declarations will be attached to the module, if any, in which they are 33092declared. 33093 33094@item Standard Library Header Units 33095The Standard Library is not provided as importable header units. If 33096you want to import such units, you must explicitly build them first. 33097If you do not do this with care, you may have multiple declarations, 33098which the module machinery must merge---compiler resource usage can be 33099affected by how you partition header files into header units. 33100 33101@end table 33102 33103Modular compilation is @emph{not} enabled with just the 33104@option{-std=c++20} option. You must explicitly enable it with the 33105@option{-fmodules-ts} option. It is independent of the language 33106version selected, although in pre-C++20 versions, it is of course an 33107extension. 33108 33109No new source file suffixes are required or supported. If you wish to 33110use a non-standard suffix (@xref{Overall Options}), you also need 33111to provide a @option{-x c++} option too.@footnote{Some users like to 33112distinguish module interface files with a new suffix, such as naming 33113the source @code{module.cppm}, which involves 33114teaching all tools about the new suffix. A different scheme, such as 33115naming @code{module-m.cpp} would be less invasive.} 33116 33117Compiling a module interface unit produces an additional output (to 33118the assembly or object file), called a Compiled Module Interface 33119(CMI). This encodes the exported declarations of the module. 33120Importing a module reads in the CMI. The import graph is a Directed 33121Acyclic Graph (DAG). You must build imports before the importer. 33122 33123Header files may themselves be compiled to header units, which are a 33124transitional ability aiming at faster compilation. The 33125@option{-fmodule-header} option is used to enable this, and implies 33126the @option{-fmodules-ts} option. These CMIs are named by the fully 33127resolved underlying header file, and thus may be a complete pathname 33128containing subdirectories. If the header file is found at an absolute 33129pathname, the CMI location is still relative to a CMI root directory. 33130 33131As header files often have no suffix, you commonly have to specify a 33132@option{-x} option to tell the compiler the source is a header file. 33133You may use @option{-x c++-header}, @option{-x c++-user-header} or 33134@option{-x c++-system-header}. When used in conjunction with 33135@option{-fmodules-ts}, these all imply an appropriate 33136@option{-fmodule-header} option. The latter two variants use the 33137user or system include path to search for the file specified. This 33138allows you to, for instance, compile standard library header files as 33139header units, without needing to know exactly where they are 33140installed. Specifying the language as one of these variants also 33141inhibits output of the object file, as header files have no associated 33142object file. 33143 33144The @option{-fmodule-only} option disables generation of the 33145associated object file for compiling a module interface. Only the CMI 33146is generated. This option is implied when using the 33147@option{-fmodule-header} option. 33148 33149The @option{-flang-info-include-translate} and 33150@option{-flang-info-include-translate-not} options notes whether 33151include translation occurs or not. With no argument, the first will 33152note all include translation. The second will note all 33153non-translations of include files not known to intentionally be 33154textual. With an argument, queries about include translation of a 33155header files with that particular trailing pathname are noted. You 33156may repeat this form to cover several different header files. This 33157option may be helpful in determining whether include translation is 33158happening---if it is working correctly, it behaves as if it isn't 33159there at all. 33160 33161The @option{-flang-info-module-cmi} option can be used to determine 33162where the compiler is reading a CMI from. Without the option, the 33163compiler is silent when such a read is successful. This option has an 33164optional argument, which will restrict the notification to just the 33165set of named modules or header units specified. 33166 33167The @option{-Winvalid-imported-macros} option causes all imported macros 33168to be resolved at the end of compilation. Without this, imported 33169macros are only resolved when expanded or (re)defined. This option 33170detects conflicting import definitions for all macros. 33171 33172@xref{C++ Module Mapper} for details of the @option{-fmodule-mapper} 33173family of options. 33174 33175@menu 33176* C++ Module Mapper:: Module Mapper 33177* C++ Module Preprocessing:: Module Preprocessing 33178* C++ Compiled Module Interface:: Compiled Module Interface 33179@end menu 33180 33181@node C++ Module Mapper 33182@subsection Module Mapper 33183@cindex C++ Module Mapper 33184 33185A module mapper provides a server or file that the compiler queries to 33186determine the mapping between module names and CMI files. It is also 33187used to build CMIs on demand. @emph{Mapper functionality is in its 33188infancy and is intended for experimentation with build system 33189interactions.} 33190 33191You can specify a mapper with the @option{-fmodule-mapper=@var{val}} 33192option or @env{CXX_MODULE_MAPPER} environment variable. The value may 33193have one of the following forms: 33194 33195@table @gcctabopt 33196 33197@item @r{[}@var{hostname}@r{]}:@var{port}@r{[}?@var{ident}@r{]} 33198An optional hostname and a numeric port number to connect to. If the 33199hostname is omitted, the loopback address is used. If the hostname 33200corresponds to multiple IPV6 addresses, these are tried in turn, until 33201one is successful. If your host lacks IPv6, this form is 33202non-functional. If you must use IPv4 use 33203@option{-fmodule-mapper='|ncat @var{ipv4host} @var{port}'}. 33204 33205@item =@var{socket}@r{[}?@var{ident}@r{]} 33206A local domain socket. If your host lacks local domain sockets, this 33207form is non-functional. 33208 33209@item |@var{program}@r{[}?@var{ident}@r{]} @r{[}@var{args...}@r{]} 33210A program to spawn, and communicate with on its stdin/stdout streams. 33211Your @var{PATH} environment variable is searched for the program. 33212Arguments are separated by space characters, (it is not possible for 33213one of the arguments delivered to the program to contain a space). An 33214exception is if @var{program} begins with @@. In that case 33215@var{program} (sans @@) is looked for in the compiler's internal 33216binary directory. Thus the sample mapper-server can be specified 33217with @code{@@g++-mapper-server}. 33218 33219@item <>@r{[}?@var{ident}@r{]} 33220@item <>@var{inout}@r{[}?@var{ident}@r{]} 33221@item <@var{in}>@var{out}@r{[}?@var{ident}@r{]} 33222Named pipes or file descriptors to communicate over. The first form, 33223@option{<>}, communicates over stdin and stdout. The other forms 33224allow you to specify a file descriptor or name a pipe. A numeric value 33225is interpreted as a file descriptor, otherwise named pipe is opened. 33226The second form specifies a bidirectional pipe and the last form 33227allows specifying two independent pipes. Using file descriptors 33228directly in this manner is fragile in general, as it can require the 33229cooperation of intermediate processes. In particular using stdin & 33230stdout is fraught with danger as other compiler options might also 33231cause the compiler to read stdin or write stdout, and it can have 33232unfortunate interactions with signal delivery from the terminal. 33233 33234@item @var{file}@r{[}?@var{ident}@r{]} 33235A mapping file consisting of space-separated module-name, filename 33236pairs, one per line. Only the mappings for the direct imports and any 33237module export name need be provided. If other mappings are provided, 33238they override those stored in any imported CMI files. A repository 33239root may be specified in the mapping file by using @samp{$root} as the 33240module name in the first active line. Use of this option will disable 33241any default module->CMI name mapping. 33242 33243@end table 33244 33245As shown, an optional @var{ident} may suffix the first word of the 33246option, indicated by a @samp{?} prefix. The value is used in the 33247initial handshake with the module server, or to specify a prefix on 33248mapping file lines. In the server case, the main source file name is 33249used if no @var{ident} is specified. In the file case, all non-blank 33250lines are significant, unless a value is specified, in which case only 33251lines beginning with @var{ident} are significant. The @var{ident} 33252must be separated by whitespace from the module name. Be aware that 33253@samp{<}, @samp{>}, @samp{?}, and @samp{|} characters are often 33254significant to the shell, and therefore may need quoting. 33255 33256The mapper is connected to or loaded lazily, when the first module 33257mapping is required. The networking protocols are only supported on 33258hosts that provide networking. If no mapper is specified a default is 33259provided. 33260 33261A project-specific mapper is expected to be provided by the build 33262system that invokes the compiler. It is not expected that a 33263general-purpose server is provided for all compilations. As such, the 33264server will know the build configuration, the compiler it invoked, and 33265the environment (such as working directory) in which that is 33266operating. As it may parallelize builds, several compilations may 33267connect to the same socket. 33268 33269The default mapper generates CMI files in a @samp{gcm.cache} 33270directory. CMI files have a @samp{.gcm} suffix. The module unit name 33271is used directly to provide the basename. Header units construct a 33272relative path using the underlying header file name. If the path is 33273already relative, a @samp{,} directory is prepended. Internal 33274@samp{..} components are translated to @samp{,,}. No attempt is made 33275to canonicalize these filenames beyond that done by the preprocessor's 33276include search algorithm, as in general it is ambiguous when symbolic 33277links are present. 33278 33279The mapper protocol was published as ``A Module Mapper'' 33280@uref{https://wg21.link/p1184}. The implementation is provided by 33281@command{libcody}, @uref{https://github.com/urnathan/libcody}, 33282which specifies the canonical protocol definition. A proof of concept 33283server implementation embedded in @command{make} was described in 33284''Make Me A Module'', @uref{https://wg21.link/p1602}. 33285 33286@node C++ Module Preprocessing 33287@subsection Module Preprocessing 33288@cindex C++ Module Preprocessing 33289 33290Modules affect preprocessing because of header units and include 33291translation. Some uses of the preprocessor as a separate step either 33292do not produce a correct output, or require CMIs to be available. 33293 33294Header units import macros. These macros can affect later conditional 33295inclusion, which therefore can cascade to differing import sets. When 33296preprocessing, it is necessary to load the CMI. If a header unit is 33297unavailable, the preprocessor issues a warning and continue (when 33298not just preprocessing, an error is emitted). Detecting such imports 33299requires preprocessor tokenization of the input stream to phase 4 33300(macro expansion). 33301 33302Include translation converts @code{#include}, @code{#include_next} and 33303@code{#import} directives to internal @code{import} declarations. 33304Whether a particular directive is translated is controlled by the 33305module mapper. Header unit names are canonicalized during 33306preprocessing. 33307 33308Dependency information can be emitted for macro import, extending the 33309functionality of @option{-MD} and @option{-MMD} options. Detection of 33310import declarations also requires phase 4 preprocessing, and thus 33311requires full preprocessing (or compilation). 33312 33313The @option{-M}, @option{-MM} and @option{-E -fdirectives-only} options halt 33314preprocessing before phase 4. 33315 33316The @option{-save-temps} option uses @option{-fdirectives-only} for 33317preprocessing, and preserve the macro definitions in the preprocessed 33318output. Usually you also want to use this option when explicitly 33319preprocessing a header-unit, or consuming such preprocessed output: 33320 33321@smallexample 33322g++ -fmodules-ts -E -fdirectives-only my-header.hh -o my-header.ii 33323g++ -x c++-header -fmodules-ts -fpreprocessed -fdirectives-only my-header.ii 33324@end smallexample 33325 33326@node C++ Compiled Module Interface 33327@subsection Compiled Module Interface 33328@cindex C++ Compiled Module Interface 33329 33330CMIs are an additional artifact when compiling named module 33331interfaces, partitions or header units. These are read when 33332importing. CMI contents are implementation-specific, and in GCC's 33333case tied to the compiler version. Consider them a rebuildable cache 33334artifact, not a distributable object. 33335 33336When creating an output CMI, any missing directory components are 33337created in a manner that is safe for concurrent builds creating 33338multiple, different, CMIs within a common subdirectory tree. 33339 33340CMI contents are written to a temporary file, which is then atomically 33341renamed. Observers either see old contents (if there is an 33342existing file), or complete new contents. They do not observe the 33343CMI during its creation. This is unlike object file writing, which 33344may be observed by an external process. 33345 33346CMIs are read in lazily, if the host OS provides @code{mmap} 33347functionality. Generally blocks are read when name lookup or template 33348instantiation occurs. To inhibit this, the @option{-fno-module-lazy} 33349option may be used. 33350 33351The @option{--param lazy-modules=@var{n}} parameter controls the limit 33352on the number of concurrently open module files during lazy loading. 33353Should more modules be imported, an LRU algorithm is used to determine 33354which files to close---until that file is needed again. This limit 33355may be exceeded with deep module dependency hierarchies. With large 33356code bases there may be more imports than the process limit of file 33357descriptors. By default, the limit is a few less than the per-process 33358file descriptor hard limit, if that is determinable.@footnote{Where 33359applicable the soft limit is incremented as needed towards the hard limit.} 33360 33361GCC CMIs use ELF32 as an architecture-neutral encapsulation mechanism. 33362You may use @command{readelf} to inspect them, although section 33363contents are largely undecipherable. There is a section named 33364@code{.gnu.c++.README}, which contains human-readable text. Other 33365than the first line, each line consists of @code{@var{tag}: @code{value}} 33366tuples. 33367 33368@smallexample 33369> @command{readelf -p.gnu.c++.README gcm.cache/foo.gcm} 33370 33371String dump of section '.gnu.c++.README': 33372 [ 0] GNU C++ primary module interface 33373 [ 21] compiler: 11.0.0 20201116 (experimental) [c++-modules revision 20201116-0454] 33374 [ 6f] version: 2020/11/16-04:54 33375 [ 89] module: foo 33376 [ 95] source: c_b.ii 33377 [ a4] dialect: C++20/coroutines 33378 [ be] cwd: /data/users/nathans/modules/obj/x86_64/gcc 33379 [ ee] repository: gcm.cache 33380 [ 104] buildtime: 2020/11/16 15:03:21 UTC 33381 [ 127] localtime: 2020/11/16 07:03:21 PST 33382 [ 14a] export: foo:part1 foo-part1.gcm 33383@end smallexample 33384 33385Amongst other things, this lists the source that was built, C++ 33386dialect used and imports of the module.@footnote{The precise contents 33387of this output may change.} The timestamp is the same value as that 33388provided by the @code{__DATE__} & @code{__TIME__} macros, and may be 33389explicitly specified with the environment variable 33390@code{SOURCE_DATE_EPOCH}. @xref{Environment Variables} for further 33391details. 33392 33393A set of related CMIs may be copied, provided the relative pathnames 33394are preserved. 33395 33396The @code{.gnu.c++.README} contents do not affect CMI integrity, and 33397it may be removed or altered. The section numbering of the sections 33398whose names do not begin with @code{.gnu.c++.}, or are not the string 33399section is significant and must not be altered. 33400